1

NCAR/TN???+STR NCAR TECHNICAL NOTE June 2010

Description of the NCAR Community Atmosphere Model (CAM 5.0)

Richard B. Neale Chih-Chieh Chen Andrew Gettelman Peter H. Lauritzen Sungsu Park David L. Williamson Climate And Global Dynamics Division National Center For Atmospheric Research Boulder, Colorado, USA Andrew J. Conley Rolando Garcia Doug Kinnison Jean-Francois Lamarque Dan Marsh Mike Mills Anne K. Smith Simone Tilmes Francis Vitt Atmopsheric Chemistry Division National Center For Atmospheric Research Boulder, Colorado, USA Philip Cameron-Smith Lawrence Livermore National Lab Livermore, California William D. Collins UC Berkeley/Lawrence Berkeley National Laboratory Berkeley, California Michael J. Iacono Atmospheric and Environmental Research, Inc. Lexington, Massachusetts Philip J. Rasch Pacic Northwest National Laboratory Richland, Washington Mark A. Taylor Sandia National Laboratories Albuquerque, New Mexico

NCAR TECHNICAL NOTES

4 5 6 7 8

The Technical Note series provides an outlet for a variety of NCAR manuscripts that contribute in specialized ways to the body of scientic knowledge but which are not suitable for journal, monograph, or book publication. Reports in this series are issued by the NCAR Scientic Divisions; copies may be obtained on request from the Publications Oce of NCAR. Designation symbols for the series include:

EDD:

Engineering, Design, or Development Reports Equipment descriptions, test results, instrumentation, and operating and maintenance manuals. Instructional Aids Instruction manuals, bibliographies, lm supplements, and other research or instructional aids. Program Progress Reports Field program reports, interim and working reports, survey reports, and plans for experiments.

IA:

PPR:

PROC: Proceedings Documentation of symposia, colloquia, conferences, workshops, and lectures. (Distribution may be limited to attendees.) STR: Scientic and Technical Reports Data compilations, theoretical and numerical investigations, and experimental results.

10 11 12 13

The National Center for Atmospheric Research (NCAR) is operated by the University Corporation for Atmospheric Research (UCAR) and is sponsored by the National Science Foundation. Any opinions, ndings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reect the views of the National Science Foundation.

14

Contents
Acknowledgments 1 Introduction 1.1 Brief History . . . . . . . . . 1.1.1 CCM0 and CCM1 . . 1.1.2 CCM2 . . . . . . . . . 1.1.3 CCM3 . . . . . . . . . 1.1.4 CAM3 . . . . . . . . . 1.1.5 CAM4 . . . . . . . . . 1.1.6 Overview of CAM 5.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi 1 1 1 2 3 4 6 7 11

15

16 17 18 19 20 21 22 23

24

2 Coupling of Dynamical Core and Parameterization Suite

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

3 Dynamics 15 3.1 Finite Volume Dynamical Core . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.1.2 The governing equations for the hydrostatic atmosphere . . . . . . . . . 15 3.1.3 Horizontal discretization of the transport process on the sphere . . . . . 17 3.1.4 A vertically Lagrangian and horizontally Eulerian control-volume discretization of the hyd 3.1.5 Optional diusion operators in CAM5 . . . . . . . . . . . . . . . . . . . 24 3.1.6 A mass, momentum, and total energy conserving mapping algorithm . . 26 3.1.7 A geopotential conserving mapping algorithm . . . . . . . . . . . . . . . 28 3.1.8 Adjustment of pressure to include change in mass of water vapor . . . . . 29 3.1.9 Negative Tracer Fixer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.1.10 Global Energy Fixer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.1.11 Further discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.2 Spectral Element Dynamical Core . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.2.1 Continuum Formulation of the Equations . . . . . . . . . . . . . . . . . . 34 3.2.2 Conserved Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.2.3 Horizontal Discretization: Functional Spaces . . . . . . . . . . . . . . . . 36 3.2.4 Horizontal Discretization: Dierential Operators . . . . . . . . . . . . . . 38 3.2.5 Horizontal Discretization: Discrete Inner-Product . . . . . . . . . . . . . 39 3.2.6 Horizontal Discretization: The Projection Operators . . . . . . . . . . . . 40 3.2.7 Horizontal Discretization: Galerkin Formulation . . . . . . . . . . . . . . 40 3.2.8 Vertical Discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.2.9 Discrete formulation: Dynamics . . . . . . . . . . . . . . . . . . . . . . . 42 i

48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88

3.3

3.4

3.2.10 Consistency . . . . . . . . . . . . . . . . . . . . . 3.2.11 Time Stepping . . . . . . . . . . . . . . . . . . . 3.2.12 Dissipation . . . . . . . . . . . . . . . . . . . . . 3.2.13 Discrete formulation: Tracer Advection . . . . . . 3.2.14 Conservation and Compatibility . . . . . . . . . . Eulerian Dynamical Core . . . . . . . . . . . . . . . . . . 3.3.1 Generalized terrain-following vertical coordinates 3.3.2 Conversion to nal form . . . . . . . . . . . . . . 3.3.3 Continuous equations using ln()/t . . . . . . 3.3.4 Semi-implicit formulation . . . . . . . . . . . . . 3.3.5 Energy conservation . . . . . . . . . . . . . . . . 3.3.6 Horizontal diusion . . . . . . . . . . . . . . . . . 3.3.7 Finite dierence equations . . . . . . . . . . . . . 3.3.8 Time lter . . . . . . . . . . . . . . . . . . . . . . 3.3.9 Spectral transform . . . . . . . . . . . . . . . . . 3.3.10 Spectral algorithm overview . . . . . . . . . . . . 3.3.11 Combination of terms . . . . . . . . . . . . . . . 3.3.12 Transformation to spectral space . . . . . . . . . 3.3.13 Solution of semi-implicit equations . . . . . . . . 3.3.14 Horizontal diusion . . . . . . . . . . . . . . . . . 3.3.15 Initial divergence damping . . . . . . . . . . . . . 3.3.16 Transformation from spectral to physical space . . 3.3.17 Horizontal diusion correction . . . . . . . . . . . 3.3.18 Semi-Lagrangian Tracer Transport . . . . . . . . 3.3.19 Mass xers . . . . . . . . . . . . . . . . . . . . . 3.3.20 Energy Fixer . . . . . . . . . . . . . . . . . . . . 3.3.21 Statistics Calculations . . . . . . . . . . . . . . . 3.3.22 Reduced grid . . . . . . . . . . . . . . . . . . . . Semi-Lagrangian Dynamical Core . . . . . . . . . . . . . 3.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . 3.4.2 Vertical coordinate and hydrostatic equation . . . 3.4.3 Semi-implicit reference state . . . . . . . . . . . . 3.4.4 Perturbation surface pressure prognostic variable 3.4.5 Extrapolated variables . . . . . . . . . . . . . . . 3.4.6 Interpolants . . . . . . . . . . . . . . . . . . . . . 3.4.7 Continuity Equation . . . . . . . . . . . . . . . . 3.4.8 Thermodynamic Equation . . . . . . . . . . . . . 3.4.9 Momentum equations . . . . . . . . . . . . . . . . 3.4.10 Development of semi-implicit system equations . . 3.4.11 Trajectory Calculation . . . . . . . . . . . . . . . 3.4.12 Mass and energy xers and statistics calculations

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

43 44 44 45 46 47 47 49 50 51 53 56 58 60 61 61 64 65 66 67 68 69 70 71 75 77 78 78 78 78 79 79 79 80 80 80 82 84 85 91 91

ii

89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133

4 Model Physics 4.1 Conversion to and from dry and wet mixing ratios for trace constituents in the 4.2 Moist Turbulence Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Bulk Moist Richardson Number . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Identication of Convective, Stably Turbulent, and Stable Layers . . . 4.2.3 Turbulent Length Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.4 Steady-State Turbulent Kinetic Energy . . . . . . . . . . . . . . . . . . 4.2.5 Stability Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.6 CL Extension-Merging Procedure . . . . . . . . . . . . . . . . . . . . . 4.2.7 Entrainment Rates at the CL Top and Base Interfaces . . . . . . . . . 4.2.8 Implicit Diusion with Implicit Diusivity . . . . . . . . . . . . . . . . 4.2.9 Implicit Surface Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Shallow Convection Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Reconstruction of Mean Proles and Cloud Condensate Partitioning . . 4.3.2 Source Air Properties of Convective Updraft . . . . . . . . . . . . . . . 4.3.3 Closures at the Cloud Base . . . . . . . . . . . . . . . . . . . . . . . . 4.3.4 Vertical Evolution of A Single Updraft Plume . . . . . . . . . . . . . . 4.3.5 Penetrative Entrainment . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.6 Convective Fluxes at and below the PBL top interface . . . . . . . . . 4.3.7 Grid-Mean Tendency of Conservative Scalars . . . . . . . . . . . . . . . 4.3.8 Grid-Mean Tendency of Non-Conservative Scalars . . . . . . . . . . . . 4.4 Deep Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Updraft Ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Downdraft Ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3 Closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.4 Numerical Approximations . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.5 Deep Convective Momentum Transports . . . . . . . . . . . . . . . . . 4.4.6 Deep Convective Tracer Transport . . . . . . . . . . . . . . . . . . . . 4.5 Evaporation of convective precipitation . . . . . . . . . . . . . . . . . . . . . . 4.6 Cloud Microphysics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6.1 Overview of the microphysics scheme . . . . . . . . . . . . . . . . . . . 4.6.2 Radiative Treatment of Ice . . . . . . . . . . . . . . . . . . . . . . . . . 4.6.3 Formulations for the microphysical processes . . . . . . . . . . . . . . . 4.7 Cloud Macrophysics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7.1 Cloud Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7.2 Cloud Overlaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7.3 Condensation Processes . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8 Aerosols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8.1 Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8.2 Chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8.3 Secondary Organic Aerosol . . . . . . . . . . . . . . . . . . . . . . . . . 4.8.4 Nucleation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8.5 Condensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8.6 Coagulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8.7 Water Uptake . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

93 model 95 . 97 . 98 . 100 . 101 . 101 . 102 . 103 . 104 . 104 . 106 . 108 . 108 . 110 . 111 . 112 . 113 . 114 . 115 . 116 . 118 . 118 . 121 . 122 . 122 . 125 . 125 . 127 . 128 . 128 . 133 . 133 . 141 . 141 . 144 . 146 . 151 . 152 . 153 . 154 . 154 . 154 . 155 . 155

134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156

4.9

4.10

4.11

4.12 4.13

4.8.8 Subgrid Vertical Transport and Activation/Resuspension 4.8.9 Wet Deposition . . . . . . . . . . . . . . . . . . . . . . . 4.8.10 Dry Deposition . . . . . . . . . . . . . . . . . . . . . . . Condensed Phase Optics . . . . . . . . . . . . . . . . . . . . . . 4.9.1 Tropospheric Aerosol Optics . . . . . . . . . . . . . . . . 4.9.2 Stratospheric Volcanic Aerosol Optics . . . . . . . . . . . 4.9.3 Liquid Cloud Optics . . . . . . . . . . . . . . . . . . . . 4.9.4 Ice Cloud Optics . . . . . . . . . . . . . . . . . . . . . . 4.9.5 Snow Cloud Optics . . . . . . . . . . . . . . . . . . . . . Radiative Transfer . . . . . . . . . . . . . . . . . . . . . . . . . 4.10.1 Combination of Aerosol Radiative Properties . . . . . . . 4.10.2 Combination of Cloud Optics . . . . . . . . . . . . . . . 4.10.3 Radiative Fluxes and Heating Rates . . . . . . . . . . . . 4.10.4 Surface Radiative Properties . . . . . . . . . . . . . . . . 4.10.5 Time Sampling . . . . . . . . . . . . . . . . . . . . . . . 4.10.6 Diurnal Cycle and Earth Orbit . . . . . . . . . . . . . . 4.10.7 Solar Spectral Irradiance . . . . . . . . . . . . . . . . . . Surface Exchange Formulations . . . . . . . . . . . . . . . . . . 4.11.1 Land . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.11.2 Ocean . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.11.3 Sea Ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dry Adiabatic Adjustment . . . . . . . . . . . . . . . . . . . . . Prognostic Greenhouse Gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

157 157 158 160 160 160 162 162 164 165 165 165 166 169 169 169 171 176 176 180 181 182 183 185 186 186 186 186 187 187 187 188 188 190 190 190 191 191 192 192 193 196 203 204

157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177

5 Extensions to CAM 5.1 Chemistry . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 Introduction . . . . . . . . . . . . . . . . . . . . 5.1.2 Emissions . . . . . . . . . . . . . . . . . . . . . 5.1.3 Lower boundary conditions . . . . . . . . . . . . 5.1.4 Lightning . . . . . . . . . . . . . . . . . . . . . 5.1.5 Dry deposition . . . . . . . . . . . . . . . . . . 5.1.6 Wet removal . . . . . . . . . . . . . . . . . . . . 5.1.7 Photolytic Approach (Neutral Species) . . . . . 5.1.8 Numerical Solution Approach . . . . . . . . . . 5.2 Superfast Chemistry . . . . . . . . . . . . . . . . . . . 5.2.1 Chemical mechanism . . . . . . . . . . . . . . . 5.2.2 LINOZ . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 PSC ozone loss . . . . . . . . . . . . . . . . . . 5.2.4 Upper boundary condition . . . . . . . . . . . . 5.3 WACCM4.0 Physical Parameterizations . . . . . . . . . 5.3.1 WACCM4.0 Domain and Resolution . . . . . . 5.3.2 Molecular Diusion and Constituent Separation 5.3.3 Gravity Wave Drag . . . . . . . . . . . . . . . . 5.3.4 Turbulent Mountain Stress . . . . . . . . . . . . 5.3.5 QBO Forcing . . . . . . . . . . . . . . . . . . .

iv

178 179 180 181

5.3.6 5.3.7 5.3.8 5.3.9

Radiation . . . . . . . WACCM4.0 chemistry Electric Field . . . . . Boundary Conditions .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

205 207 222 231

182 183 184

6 Initial and Boundary Data 233 6.1 Initial Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233 6.2 Boundary Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234

Appendices
185

A Physical Constants B Acronyms C Resolution and dycore-dependent parameters References

237 239 241 242

186

187

188

vi

189

List of Figures
3.1 3.2 3.3 3.4 3.5 4.1 4.2 4.3 4.4 4.5 4.6 4.7 5.1 5.2

190 191 192 193 194

A graphical illustration of the dierent levels of sub-cycling in CAM5. . . . . . . 24 Tiling the surface of the sphere with quadrilaterals. An inscribed cube is projected to the surface A 4 4 tensor product grid of GLL nodes used within each element, for a degree d = 3 discretiza Vertical level structure of CAM 5.0 . . . . . . . . . . . . . . . . . . . . . . . . . 53 Pentagonal truncation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 62

195 196 197 198 199 200 201

Schematic structure of moist turbulence scheme . . . . . . . . . . . . . . . . . . 99 Schematic structure of shallow cumulus scheme . . . . . . . . . . . . . . . . . . 109 Predicted species for interstitial and cloud-borne component of each aerosol mode in MAM-7. St Predicted species for interstitial and cloud-borne component of each aerosol mode in MAM-3. St Kurucz spectrum. ssf in W/m2 /nm. Source Data: AER. Range from [20, 20000] nm.173 Lean spectrum. Average over 1 solar cycle, May 1, 1996 to Dec 31, 2006. Source Data: Marsh. s LeanKurucz Relative dierence, .5(Lean+Kurucz) between spectra. RRTMG band boundaries are marked with ve

202 203

Global mean distribution of charged constituents during July solar minimum conditions.211 a) Global distribution of ionization rates at 7.3105 hPa, July 1, UT0100 HRS. Contour interva

vii

viii

204

List of Tables
4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 5.1 5.2 5.1 5.2 5.2 5.2 5.2 5.2 5.3 5.4 5.4 5.5 5.6 5.7 Size distributions of primary emissions. . . . . . . . . . . . . . . . . . . . . . . Assumed SOA (gas) yields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hygroscopicity of aerosol components . . . . . . . . . . . . . . . . . . . . . . . Density (kg/m3 ) of aerosol material. . . . . . . . . . . . . . . . . . . . . . . . . Hygroscopicity of aerosol components. . . . . . . . . . . . . . . . . . . . . . . RRTMG SW spectral band boundaries and the solar irradiance in each band. RRTMG LW spectral band boundaries. . . . . . . . . . . . . . . . . . . . . . . Band-level ratio of Solar Irradiances, based on average of one solar cycle . . . Surface uxes for CAM4 superfast chemistry. . . . . . . . . . . . . . . . . . . . WACCM4.0 Neutral Chemical Species (51 computed species + N2 ) . . . . . . (continued) WACCM4.0 Neutral Chemical Species (51 computed species + N2 ) WACCM4.0 Gas-phase Reactions. . . . . . . . . . . . . . . . . . . . . . . . . . (continued) WACCM4.0 Gas-phase Reactions. . . . . . . . . . . . . . . . . . . (continued) WACCM4.0 Gas-phase Reactions. . . . . . . . . . . . . . . . . . . (continued) WACCM4.0 Gas-phase Reactions. . . . . . . . . . . . . . . . . . . (continued) WACCM4.0 Gas-phase Reactions. . . . . . . . . . . . . . . . . . . WACCM4.0 Heterogeneous Reactions on liquid and solid aerosols. . . . . . . . WACCM4.0 Photolytic Reactions. . . . . . . . . . . . . . . . . . . . . . . . . . (continued) WACCM4.0 Photolytic Reactions. . . . . . . . . . . . . . . . . . . Ion-neutral and recombination reactions and exothermicities. . . . . . . . . . . Ionization reactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EUVAC model parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 156 156 160 160 167 168 172 186 213 214 215 216 217 218 219 219 220 221 222 223 223

205 206 207 208 209 210 211 212

213 214 215 216 217 218 219 220 221 222 223 224 225 226

227

C.1 Resolution-dependent parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 241

ix

228

Acknowledgments
The authors wish to acknowledge members of NCARs Atmospheric Modeling and Predictability Section (AMP), Computer Software and Engineering Group (CSEG), and Computation and Information Systems Laboratory (CISL) for their contributions to the development of CAM 5.0. The new model would not exist without the signicant input from members of the CCSM Atmospheric Model Working Group (AMWG) too numerous to mention. Leo Donner (GFDL), Minghua Zhang (SUNY) and Phil Rasch (PNNL) were the co-chairs of the AMWG during the development of CAM 5.0. We would like to acknowledge the substantial contributions to the CAM 5.0 eort from the National Science Foundation, Department of Energy, the National Oceanic and Atmospheric Administration, and the National Aeronautics and Space Administration.

229 230 231 232 233 234 235 236 237 238

xi

xii

239

Chapter 1 Introduction
This report presents the details of the governing equations, physical parameterizations, and numerical algorithms dening the version of the NCAR Community Atmosphere Model designated CAM 5.0. The material provides an overview of the major model components, and the way in which they interact as the numerical integration proceeds. Details on the coding implementation, along with in-depth information on running the CAM 5.0 code, are given in a separate technical report entitled Users Guide to Community Atmosphere ModelCAM 5.0 [Eaton, 2010]. As before, it is our objective that this model provide NCAR and the university research community with a reliable, well documented atmospheric general circulation model. This version of the CAM 5.0 incorporates a number enhancements to the physics package (e.g. adjustments to the deep convection algorithm including the addition of Convective Momentum Transports (CMT), a transition to the nite volume dynamical core as default and the option to run a computationally highly scaleable dynamical core). The ability to transition between CAM-standalone and fully coupled experiment frameworks is much improved in CAM 5.0. We believe that collectively these improvements provide the research community with a signicantly improved atmospheric modeling capability.

240

241 242 243 244 245 246 247 248 249 250 251 252 253 254 255

256

1.1
1.1.1

Brief History
CCM0 and CCM1

257

258 259 260 261 262 263 264 265 266 267 268 269

Over the last twenty years, the NCAR Climate and Global Dynamics (CGD) Division has provided a comprehensive, three-dimensional global atmospheric model to university and NCAR scientists for use in the analysis and understanding of global climate. Because of its widespread use, the model was designated a community tool and given the name Community Climate Model (CCM). The original versions of the NCAR Community Climate Model, CCM0A [Washington, 1982] and CCM0B [Williamson, 1983], were based on the Australian spectral model [Bourke et al., 1977; McAvaney et al., 1978] and an adiabatic, inviscid version of the ECMWF spectral model [Baede et al., 1979]. The CCM0B implementation was constructed so that its simulated climate would match the earlier CCM0A model to within natural variability (e.g. incorporated the same set of physical parameterizations and numerical approximations), but also provided a more exible infrastructure for conducting medium and longrange global forecast studies. The major strength of this latter eort was that all aspects of the model were described

270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288

in a series of technical notes, which included a Users Guide [Sato et al., 1983], a subroutine guide which provided a detailed description of the code [Williamson et al., 1983] a detailed description of the algorithms [Williamson, 1983], and a compilation of the simulated circulation statistics [Williamson and Williamson, 1984]. This development activity rmly established NCARs commitment to provide a versatile, modular, and welldocumented atmospheric general circulation model that would be suitable for climate and forecast studies by NCAR and university scientists. A more detailed discussion of the early history and philosophy of the Community Climate Model can be found in Anthes [1986]. The second generation community model, CCM1, was introduced in July of 1987, and included a number of signicant changes to the model formulation which were manifested in changes to the simulated climate. Principal changes to the model included major modications to the parameterization of radiation, a revised vertical nite-dierencing technique for the dynamical core, modications to vertical and horizontal diusion processes, and modications to the formulation of surface energy exchange. A number of new modeling capabilities were also introduced, including a seasonal mode in which the specied surface conditions vary with time, and an optional interactive surface hydrology that followed the formulation presented by Manabe [1969]. A detailed series of technical documentation was also made available for this version [Williamson et al., 1987; Bath et al., 1987; Williamson and Williamson, 1987; Hack et al., 1989] and more completely describe this version of the CCM.

289

1.1.2

CCM2

290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312

The most ambitious set of model improvements occurred with the introduction of the third generation of the Community Climate Model, CCM2, which was released in October of 1992. This version was the product of a major eort to improve the physical representation of a wide range of key climate processes, including clouds and radiation, moist convection, the planetary boundary layer, and transport. The introduction of this model also marked a new philosophy with respect to implementation. The CCM2 code was entirely restructured so as to satisfy three major objectives: much greater ease of use, which included portability across a wide range of computational platforms; conformance to a plug-compatible physics interface standard; and the incorporation of single-job multitasking capabilities. The standard CCM2 model conguration was signicantly dierent from its predecessor in almost every way, starting with resolution where the CCM2 employed a horizontal T42 spectral resolution (approximately 2.8 x 2.8 degree transform grid), with 18 vertical levels and a rigid lid at 2.917 mb. Principal algorithmic approaches shared with CCM1 were the use of a semi-implicit, leap frog time integration scheme; the use of the spectral transform method for treating the dry dynamics; and the use of a bi-harmonic horizontal diusion operator. Major changes to the dynamical formalism included the use of a terrain-following hybrid vertical coordinate, and the incorporation of a shape-preserving semi-Lagrangian transport scheme [Williamson and Olson, 1994a] for advecting water vapor, as well as an arbitrary number of other scalar elds (e.g. cloud water variables, chemical constituents, etc.). Principal changes to the physics included the use of a -Eddington approximation to calculate solar absorption [Briegleb, 1992]; the use of a Voigt line shape to more accurately treat infrared radiative cooling in the stratosphere; the inclusion of a diurnal cycle to properly account for the interactions between the radiative eects of the diurnal cycle and the surface uxes of sensible and latent heat; the incorporation of a nite heat 2

313 314 315 316 317 318 319 320

capacity soil/sea ice model; a more sophisticated cloud fraction parameterization and treatment of cloud optical properties [Kiehl et al., 1994]; the incorporation of a sophisticated non-local treatment of boundary-layer processes [Holtslag and Boville, 1993a]; the use of a simple mass ux representation of moist convection [Hack, 1994a], and the optional incorporation of the Biosphere-Atmosphere Transfer Scheme (BATS) of Dickinson et al. [1987]. As with previous versions of the model, a Users Guide [Bath et al., 1992] and model description [Hack et al., 1993] were provided to completely document the model formalism and implementation. Control simulation data sets were documented in Williamson [1993].

321

1.1.3

CCM3

322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355

The CCM3 was the fourth generation in the series of NCARs Community Climate Model. Many aspects of the model formulation and implementation were identical to the CCM2, although there were a number of important changes that were incorporated into the collection of parameterized physics, along with some modest changes to the dynamical formalism. Modications to the physical representation of specic climate processes in the CCM3 were motivated by the need to address the more serious systematic errors apparent in CCM2 simulations, as well as to make the atmospheric model more suitable for coupling to land, ocean, and sea-ice component models. Thus, an important aspect of the changes to the model atmosphere was that they address well known systematic biases in the top-of-atmosphere and surface (to the extent that they were known) energy budgets. When compared to the CCM2, changes to the model formulation fell into ve major categories: modications to the representation of radiative transfer through both clear and cloudy atmospheric columns, modications to hydrological processes (i.e., in the form of changes to the atmospheric boundary layer, moist convection, and surface energy exchange), the incorporation of a sophisticated land surface model, the incorporation of an optional slab mixed-layer ocean/thermodynamic sea-ice component, and a collection of other changes to the formalism which did not introduce signicant changes to the model climate. Changes to the clear-sky radiation formalism included the incorporation of minor CO2 bands trace gases (CH4 , N 2 O, CF C11, CF C12) in the longwave parameterization, and the incorporation of a background aerosol (0.14 optical depth) in the shortwave parameterization. All-sky changes included improvements to the way in which cloud optical properties (eective radius and liquid water path) were diagnosed, the incorporation of the radiative properties of ice clouds, and a number of minor modications to the diagnosis of convective and layered cloud amount. Collectively these modication substantially reduced systematic biases in the global annually averaged clear-sky and all-sky outgoing longwave radiation and absorbed solar radiation to well within observational uncertainty, while maintaining very good agreement with global observational estimates of cloud forcing. Additionally, the large warm bias in simulated July surface temperature over the Northern Hemisphere, the systematic over-prediction of precipitation over warm land areas, and a large component of the stationary-wave error in CCM2, were also reduced as a result of cloud-radiation improvements. Modications to hydrological processes included revisions to the major contributing parameterizations. The formulation of the atmospheric boundary layer parameterization was revised (in collaboration with Dr. A. A. M. Holtslag of KNMI), resulting in signicantly improved estimates of boundary layer height, and a substantial reduction in the overall magnitude of the hydrological cycle. Parameterized convection was also modied where this process was repre3

356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395

sented using the deep moist convection formalism of Zhang and McFarlane [1995] in conjunction with the scheme developed by Hack [1994a] for CCM2. This change resulted in an additional reduction in the magnitude of the hydrological cycle and a smoother distribution of tropical precipitation. Surface roughness over oceans was also diagnosed as a function of surface wind speed and stability, resulting in more realistic surface ux estimates for low wind speed conditions. The combination of these changes to hydrological components resulted in a 13% reduction in the annually averaged global latent heat ux and the associated precipitation rate. It should be pointed out that the improvements in the radiative and hydrological cycle characteristics of the model climate were achieved without compromising the quality of the simulated equilibrium thermodynamic structures (one of the major strengths of the CCM2) thanks in part to the incorporation of a Sundqvist [1988] style evaporation of stratiform precipitation. The CCM3 incorporated version 1 of the Land Surface Model (LSM) developed by Bonan [1996] which provided for the comprehensive treatment of land surface processes. This was a one-dimensional model of energy, momentum, water, and CO2 exchange between the atmosphere and land, accounting for ecological dierences among vegetation types, hydraulic and thermal dierences among soil types, and allowing for multiple surface types including lakes and wetlands within a grid cell. LSM replaced the prescribed surface wetness, prescribed snow cover, and prescribed surface albedos in CCM2. It also replaced the land surface uxes in CCM2, using instead ux parameterizations that included hydrological and ecological processes (e.g. soil water, phenology, stomatal physiology, interception of water by plants). The fourth class of changes to the CCM2 included the option to run CCM3 with a simple slab ocean-thermodynamic sea ice model. The model employs a spatially and temporally prescribed ocean heat ux and mixed layer depth, which ensures replication of realistic sea surface temperatures and ice distributions for the present climate. The model allowed for the simplest interactive surface for the ocean and sea ice components of the climate system. The nal class of model modications included a change to the form of the hydrostatic matrix which ensures consistency between and the discrete continuity equation, and a more generalized form of the gravity wave drag parameterization. In the latter case, the parameterization was congured to behave in the same way as the CCM2 parameterization of wave drag, but included the capability to exploit more sophisticated descriptions of this process. One of the more signicant implementation dierences with the earlier model was that CCM3 included an optional message-passing conguration, allowing the model to be executed as a parallel task in distributed-memory environments. This was an example of how the Climate and Global Dynamics Division continued to invest in technical improvements to the CCM in the interest of making it easier to acquire and use in evolving computational environments. As was the case for CCM2, the code was internally documented, obviating the need for a separate technical note that describes each subroutine and common block in the model library. Thus, the Users Guide, the land surface technical note, the CCM3 technical note [Kiehl et al., 1996], the actual code and a series of reviewed scientic publications (including a special issue of the Journal of Climate, Volume 11, Number 6) were designed to completely document CCM3.

396

1.1.4

CAM3

397 398

The CAM3 was the fth generation of the NCAR atmospheric GCM. The name of the model series was changed from Community Climate Model to Community Atmosphere Model to reect 4

399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443

the role of CAM3 in the fully coupled climate system. In contrast to previous generations of the atmospheric model, CAM3 had been designed through a collaborative process with users and developers in the Atmospheric Model Working Group (AMWG). The AMWG includes scientists from NCAR, the university community, and government laboratories. For CAM3, the consensus of the AMWG was to retain the spectral Eulerian dynamical core for the rst ocial release although the code includes the option to run with semi-Lagrange dynamics or with nite-volume dynamics (FV). The addition of FV was a major extension to the model provided through a collaboration between NCAR and NASA Goddards Data Assimilation Oce (DAO). The major changes in the physics included treatment of cloud condensed water using a prognostic formulation with a bulk microphysical component following Rasch and Kristjnsson a [1998a] and a macroscale component following Zhang et al. [2003b]. The Zhang and McFarlane [1995] parameterization for deep convection was retained from CCM3. A new treatment of geometrical cloud overlap in the radiation calculations computed the shortwave and longwave uxes and heating rates for random overlap, maximum overlap, or an arbitrary combination of maximum and random overlap. The calculation was completely separated from the radiative parameterizations. The introduction of the generalized overlap assumptions permitted more realistic treatments of cloud-radiative interactions. The methodology was designed and validated against calculations based upon the independent column approximation (ICA). A new parameterization for the longwave absorptivity and emissivity of water vapor preserved the formulation of the radiative transfer equations using the absorptivity/emissivity method. The components of the method related to water vapor were replaced with new terms calculated with the General Line-by-line Atmospheric Transmittance and Radiance Model (GENLN3). The mean absolute errors in the surface and top-of-atmosphere clear-sky longwave uxes for standard atmospheres were reduced to less than 1 W/m2 . The near-infrared absorption by water vapor was also updated to a parameterization based upon the HITRAN2k line database [Rothman et al., 2003] that incorporated the CKD 2.4 prescription for the continuum. The magnitude of errors in ux divergences and heating rates relative to modern LBL calculations were reduced by approximately seven times compared to the previous CCM3 parameterization. The uniform background aerosol was replaced with a present-day climatology of sulfate, sea-salt, carbonaceous, and soil-dust aerosols. The climatology was obtained from a chemical transport model forced with meteorological analysis and constrained by assimilation of satellite aerosol retrievals. These aerosols aect the shortwave energy budget of the atmosphere. CAM3 also included a mechanism for treating the shortwave and longwave eects of volcanic aerosols. Evaporation of convective precipitation following Sundqvist [1988] was implemented and enhancement of atmospheric moisture through this mechanism was oset by drying introduced by changes in the longwave absorptivity and emissivity. A careful formulation of vertical diusion of dry static energy was also implemented. Additional capabilities included a new thermodynamic package for sea ice in order to mimic the major non-dynamical aspects of CSIM; including snow depth, brine pockets, internal shortwave radiative transfer, surface albedo, ice-atmosphere drag, and surface exchange uxes. CAM3 also allowed for an explicit representation of fractional land and sea-ice coverage that gave a much more accurate representation of ux exchanges from coastal boundaries, island regions, and ice edges. This fractional specication provided a mechanism to account for ux dierences due to sub-grid inhomogeneity of surface types. A new, extensible climatological and time-mean sea-surface temperature boundary data was made available from a blended product using the 5

444 445 446 447 448 449

global HadISST OI dataset prior to 1981 and the Smith/Reynolds EOF dataset post-1981. Coupling was upgraded in order to couple the dynamical core with the parameterization suite in a purely time split or process split manner. The distinction is that in the process split approximation the physics and dynamics are both calculated from the same past state, while in the time split approximations the dynamics and physics are calculated sequentially, each based on the state produced by the other.

450

1.1.5

CAM4

451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486

The CAM4 was the sixth generation of the NCAR atmospheric GCM and had again been developed through a collaborative process of users and developers in the Atmosphere Model Working Group (AMWG) with signcant input from the Chemistry Climate Working Group (Chem-Clim WG) and the Whole Atmosphere Model Working Group (WAMWG). The model had science enhancements from CAM3 and represented an intermediate release version as part of a staged and parallel process in atmospheric model development. In the CAM4 changes to the moist physical representations centered on enhancements to the existing Zhang and McFarlane [1995] deep convection parameterization. The calculation of Convective Available Potential Energy (CAPE) assumed an entraining plume to provide the in-cloud temperature and humidity proles used to determine bouyancy and related cloud closure properties (chapter 4.4). The modication is based on the conservation of moist entropy and mixing methods of Raymond and Blyth [1986, 1992]. It replaced the standard undilute non-entraining plume method used in CAM3 and was employed to increase convection sensitivity to tropospheric moisture and reduce the amplitude of the diurnal cycle of precipitation over land. Sub-grid scale Convective Momentum Transports (CMT) were added to the deep convection scheme following Richter and Rasch [2008] and the methodology of Gregory et al. [1997b] (chapter 4.4.5). CMT aects tropospheric climate mainly through changes to the Coriolis torque. These changes resulted in improvement of the Hadley circulation during northern Winter and it reduced many of the model biases. In an annual mean, the tropical easterly bias, subtropical westerly bias, and the excessive southern hemisphere mid-latitude jet were improved. In combination these modications to the deep-convection lead to signicant improvements in the phase, amplitude and spacial anomaly patterns of the modeled El Ni o, as documented n in Neale et al. [2008]. The calculation of cloud fraction in polar climates was also modied for the CAM4.0. Due to the combination of a diagnostic cloud fraction and prognostic cloud water represntation it was possible to model unphysical extensive cloud decks with near zero in-cloud water in the CAM3. This was particularly pervasize in polar climates in Winter. These calculation inconsitencies and large cloud fractions are signicantly reduced with modications to the calculation of stratiform cloud following Vavrus and Waliser [2008]. In the lower troposphere a freeze-drying process is perfomed whereby cloud fractions were systematically reduced for very low water vaopr amounts. The low cloud reduction caused an Arctic-wide drop of 15 W m2 in surface cloud radiative forcing (CRF) during winter and about a 50% decrease in mean annual Arctic CRF. Consequently, wintertime surface temperatures fell by up to 4 K on land and 2 K over the Arctic Ocean, thus signicantly reducing the CAM3 pronounced warm bias. More generally the radiation calculation was performed using inconsistent cloud fraction and condensate quantities in the CAM3. In CAM4 this was remedied with an updated cloud fraction calculation prior to the radiation call at each physics timestep. The coupled climate performance with 6

487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518

the CAM4.0 physics changes was summarized in the horizontal resolution comparison study of Gent et al. [2009]. For the dynamical core component of CAM4 the nite volume (FV) scheme was made the default due to its superior transport properties [Lin and Rood, 1996]. Modications were made that upgraded the code version to a more recent NASA Goddard supported version. Other changes provided new horizontal grid discretizations (e.g., 1.9x2.5 deg and 0.9x1.25 deg) for optimal computational processor decompostion and polar ltering changes for noise reductions and more continuous (in latitude) ltering. In addition to the existing nite volume and spectralbased dynamical core a new option was also made available that represents the rst scheme released with CAM that removes the computational scalability restrictions associated with a pole convergent latitude-longitude grid and the associated polar ltering requirements. Funded in part by the Department of Energy (DOE) Climate Change Prediction Program the scalable and ecient spectral-element-based atmospheric dynamical core uses the High Order Method Modeling Environment (HOMME) on a cubed sphere grid and was developed by members of the Computational Science Section and the Computational Numerics Group of NCARs Computational and Information Systems Laboratory (CISL). The nite element dynamical core (commonly referred to as the HOMME core) is fully integrated into CCSM coupling architecture and is invaluable for high resolution climate integrations on existing and upcoming massively parallel computing platforms. Model exibility was increased signicantly from the CAM3, both within CAM and the CCSM system as a whole. The method for running thermodynamic sea-ice in CAM-only mode was moved to be maintained entirely within the CICE model of the CCSM4. The single-column version of CAM was given the exibility to be built and run using the same infrastructure as the CAM build and run mechanism. The SCAM GUI run method was no longer supported. The increased coupling exibility also allowed the introduction of a more consistant method for performing slab-ocean model (SOM) experiments. SOM experiments were, by default, now performed using forcing data from an existing CCSM coupled run. This had the advantage of having a closed temperature budget for both the ice and the ocean mixed layer from a coupled run. The methodology was therefore congured to reproduce the fully coupled CCSM climate as opposed to a reproduction of a psuedo-observed climate available with the CAM3-specic SOM method. The CAM3-specic SOM method was no longer made available. For more information regarding updated run methods see the CAM4.0 users guide of Eaton [2010].

519

1.1.6

Overview of CAM 5.0

520

The Community Atmosphere Model CAM has been modied substantially with a range of enhancements and improvements in the representation of physical processes since version 4 (CAM4). In particular, the combination of physical parameterization enhancements makes it possible to simulate full aerosol cloud interactions including cloud droplet activation by aerosols, precipitation processes due to particle size dependant behavior and explicit radiative interaction of cloud particles. As such the CAM 5.0represents the rst version of CAM that is able to simulate the cloud-aerosol indirect radiative eects. The entensive list of physical parameterization improvements are described below: A new moist turbulence scheme (Section 4.2) is included that explicitly simulates stratus-

521 522 523 524 525 526 527 528

529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573

radiation-turbulence interactions, making it possible to simulate full aerosol indirect eects within stratus. It is based on a diagnostic Turbulent Kinetic Energy (TKE) forumlation and uses a 1st order K-diusion scheme with entrainment [Bretherton and Park, 2009a] originally developed at the University of Washington.. The scheme operates in any layer of the atmopshere when the moist Ri ( Richardson number ) is larger than its critical value. A new shallow convection scheme (Section 4.3) uses a realistic plume dilution equation and closure that accurately simulates the spatial distribution of shallow convective activity [Park and Bretherton, 2009]. A steady state convective updraft plume and small fractional area are assumed. An explicit computation of the convective updraft vertcial velocity and updraft fraction is performed using an updraft vertical momentum equation, and thus provides a representation of convective momentum transports. The scheme is specically designed to interact with the new moist turbulence scheme in order to prevent double counting seen in previous CAM parameterizations. The deep convection parameterization is retained from CAM4.0 (Section 4.4). Stratiform microphysical processes (Section 4.6) are represented by a prognostic, two-moment formulation for cloud droplet and cloud ice with mass and number concentrations following the original parameterization of Morrison and Gettelman [2008]. The implimentation in CAM 5.0[Gettelman et al., 2008] determines liquid and ice particle sizes from gamma functions and their evolution in time is subject to grid-scale advection, convective detrainment, turbulent diusion and several microphysical processes. Activation of cloud droplets occurs on an aerosol size distribution based on aerosol chemistry, temperature and vertical velocity. A sub-grid scale vertical velocity is provided through a turbulent kinetic energy approximation. A number of mechanisms are calcuated for ice crystal nucleation [Liu et al., 2007] and combined with modications to allow ice supersaturation [Gettelman et al., 2010b]. The revised cloud macrophysics scheme (Section 4.7,Park et al. [2010]) provides a more transparent treatment of cloud processes and imposes full consistency between cloud fraction and cloud condensate. Separate calculations are performed for liquid and ice stratiform cloud fractions which are assumed to be maximally overlapped. Liquid cloud fraction is based on an assumed triangular distribution of total relative humidity. Ice cloud fraction is based on Gettelman et al. [2010a] and allows supersaturation via a modied relative humidity over ice and the inclusion of ice condensate amount. A new 3-mode modal aerosol scheme (MAM3, Section 4.8, Liu and Ghan [2010]) provides internally mixed representations of number concentrations and mass for Aitkin, accumulation and course aerosol modes which are merged characterizations of the more complex 7-mode version of the scheme. Anthropogenic emissions, dened as originating from industrial, domestic and agriculture activity sectors, are provided from the Lamarque et al. [2010a] IPCC AR5 emission data set. Emissions of black carbon and organic carbon represent an update of Bond et al. [2007] and Junker and Liousse [2008]. Emissions of sulfur dioxide are an update of Smith et al. [2001, 2004]. Injection heights, and size distribution of emissions data are not provided with the raw datasets so the protocols of [Dentener et al., 2006a] are followed for CAM 5.0. AEROCOM emission datastes are used for natural aeroso0l sources. All emission datasets required to run MAM for pre-industrial or 20th century scenarios are available for download. A full inventory of observationally based aerosol emission mass and size is provided in standard available datasets. The 7-mode version of the scheme is also available. 8

574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599

Calculations and specications for the condensed phase optics (aerosols, liquid cloud droplets, hydrometeors and ice crystals) are taken from the microphysics and aerosol parmeterization quantities and provided as input to the radiation scheme (Section 4.9). The radiation scheme (Section 4.10) has been updated to the Rapid Radiative Transfer Method for GCMs (RRTMG, Iacono et al. [2008]; Mlawer et al. [1997]). It employs an ecient and accurate modied correlated-k method for calculating radiative uxes and heating rates in the clear sky and for the condensed phase species. For each short-wave band calculation extinction optical depth, single scattering albedo and asymmetry properties are specied. For each long-wave band massspecic absorption is specied. The aerosol optical properties are dened for each mode of the MAM as described by [Ghan and Zaveri, 2007]. Hygroscopicity characteristics are specied for soluable species. For volcanic aerosols a geometric mean radius is used. Optical properties of aerosols are combined prior to the radiative calculation. Liquid-cloud optics are calculated following Wiscombe [1996] and ice-cloud optics are calculated following Mitchell [2002]. Ice-cloud size optics are extended to allow for radiatively active falling snow. Optical properties of clouds (including separate fractions and in-cloud water contents) are combined prior to the radiative calculation. RRTM separates the short-wave spectrum into 14 bands extending from 0.2 m to 12.2 m, and models sources of extinction for H2 O, O3 , CO2 , O2 , CH4 , N2 and Rayleigh scattering. Solar irradiance is now specied for the short-wave bands from the Lean dataset [Wang et al., 2005]. The long-wave spectrum is separated into 16 bands extending from 3.1 m to 1000 m with molecular sources of absorption for the same species, in addition to CFC-11 (containing multiple CFC species) and CFC-12. RRTMG has extensive modications from the original RRTM in order to provide signicant speed-up for long climate integrations. Chief amongt these is the Monte-Carlo Independent Column Approximation (McICA, Pincus and Morcrette [2003]) that represnts sub-grid scale cloud variability. With these modications RRTMG still retains superior oine agreement with line-by-line calculations when compared to the previous CAM radiation package (CAM-RT). The CAM Chemistry Model (CAM-CHEM) Chemistry in CAM is now fully interactive and implemented in CESM (Section 5.1); in particular, emissions of biogenic compounds and deposition of aerosols to snow, ice, ocean and vegetation are handled through the coupler. The released version of CAM-chem in CESM is using the recently-developed superfast chemistry (Section 5.2), in collaboration with P. Cameron-Smith from LLNL and M. Prather from UCI) to perform centennial scale simulations at a minor cost increase over the base CAM4. These simulations use the recently developed 1850-2005 emissions created in support of CMIP5. The Whole Atmosphere Community Climate Model (WACCM) WACCM4 (Section 5.3), incorporates several improvements and enhancements over the previous version (3.1.9). It can be run coupled to the POP2 and CICE CCSM model components. The models chemistry module (Section 5.1) has been updated according to the latest JPL-2006 recommendations; a quasi-biennial oscillation may be imposed (as an option) by relaxing the winds to observations in the Tropics; heating from stratospheric volcanic aerosols is now computed explicitly; the eects of solar proton events are now included; the eect of unresolved orography is

600

601 602 603 604 605 606 607

608

609 610 611 612 613 614

615 616 617 618

parameterized as a surface stress (turbulent mountain stress) leading to an improvement in the frequency of sudden stratospheric warmings; and gravity waves due to convective and frontal sources are parameterized based upon the occurrence of convection and the diagnosis of regions of frontogenesis in the model.

10

619

Chapter 2 Coupling of Dynamical Core and Parameterization Suite

The CAM 5.0 cleanly separates the parameterization suite from the dynamical core, and makes it easier to replace or modify each in isolation. The dynamical core can be coupled to the parameterization suite in a purely time split manner or in a purely process split one, as described below. Consider the general prediction equation for a generic variable , = D () + P () , t (2.1)

620

621

622 623 624 625

626 627 628 629 630 631 632

where denotes a prognostic variable such as temperature or horizontal wind component. The dynamical core component is denoted D and the physical parameterization suite P . A three-time-level notation is employed which is appropriate for the semi-implicit Eulerian spectral transform dynamical core. However, the numerical characteristics of the physical parameterizations are more like those of diusive processes rather than advective ones. They are therefore approximated with forward or backward dierences, rather than centered three-timelevel forms. The Process Split coupling is approximated by n+1 = n1 + 2tD( n+1 , n , n1 ) + 2tP ( , n1 ) , where P ( , n1) is calculated rst from = n1 + 2tP ( , n1 ) . The Time Split coupling is approximated by = n1 + 2tD( , n , n1 ) , n+1 = + 2tP ( n+1 , ) . (2.4) (2.5) (2.3) (2.2)

633 634 635

The distinction is that in the Process Split approximation the calculations of D and P are both based on the same past state, n1 , while in the Time Split approximations D and P are calculated sequentially, each based on the state produced by the other.

11

636 637 638 639

As mentioned above, the Eulerian core employs the three-time-level notation in (2.2)-(2.5). Eqns. (2.2)-(2.5) also apply to two-time-level nite volume, semi-Lagrangian and spectral element (HOMME) cores by dropping centered n term dependencies, and replacing n-1 by n and 2t by t. The parameterization package can be applied to produce an updated eld as indicated in (2.3) and (2.5). Thus (2.5) can be written with an operator notation n+1 = P ( ) , (2.6)

where only the past state is included in the operator dependency for notational convenience. The implicit predicted state dependency is understood. The Process Split equation (2.2) can also be written in operator notation as n+1 = D n1 , P ( n1 ) n1 2t , (2.7)

where the rst argument of D denotes the prognostic variable input to the dynamical core and the second denotes the forcing rate from the parameterization package, e.g. the heating rate in the thermodynamic equation. Again only the past state is included in the operator dependency, with the implicit predicted state dependency left understood. With this notation the Time Split system (2.5) and (2.5) can be written n+1 = P D n1 , 0 . (2.8)

640 641 642 643 644 645 646 647 648 649 650 651

where M denotes (Moist) precipitation processes, R denotes clouds and Radiation, S denotes the Surface model, and T denotes Turbulent mixing. Each of these in turn is subdivided into various components: M includes an optional dry adiabatic adjustment (normally applied only in the stratosphere), moist penetrative convection, shallow convection, and large-scale stable condensation; R rst calculates the cloud parameterization followed by the radiation parameterization; S provides the surface uxes obtained from land, ocean and sea ice models, or calculates them based on specied surface conditions such as sea surface temperatures and sea ice distribution. These surface uxes provide lower ux boundary conditions for the turbulent mixing T which is comprised of the planetary boundary layer parameterization, vertical diusion, and gravity wave drag. Dening operators following (2.6) for each of the parameterization components, the couplings in CAM 5.0 are summarized as: TIME SPLIT n+1 = T S R M D n1 , 0 PROCESS SPLIT n+1 = D n1 , T (S (R (M ( n1 )))) n1 2t 12 (2.11) (2.10)

The total parameterization package in CAM 5.0 consists of a sequence of components, indicated by P = {M, R, S, T } , (2.9)

652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667

The labels Time Split and Process Split refer to the coupling of the dynamical core with the complete parameterization suite. The components within the parameterization suite are coupled via time splitting in both forms. The Process Split form is convenient for spectral transform models. With Time Split approximations extra spectral transforms are required to convert the updated momentum variables provided by the parameterizations to vorticity and divergence for the Eulerian spectral core, or to recalculate the temperature gradient for the semi-Lagrangian spectral core. The Time Split form is convenient for the nite-volume core which adopts a Lagrangian vertical coordinate. Since the scheme is explicit and restricted to small time-steps by its non-advective component, it sub-steps the dynamics multiple times during a longer parameterization time step. With Process Split approximations the forcing terms must be interpolated to an evolving Lagrangian vertical coordinate every sub-step of the dynamical core. Besides the expense involved, it is not completely obvious how to interpolate the parameterized forcing, which can have a vertical grid scale component arising from vertical grid scale clouds, to a dierent vertical grid. [Williamson, 2002] compares simulations with the Eulerian spectral transform dynamical core coupled to the CCM3 parameterization suite via Process Split and Time Split approximations.

13

14

668

Chapter 3 Dynamics
3.1
3.1.1

669

670

Finite Volume Dynamical Core

Overview

671

672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687

This document describes the Finite-Volume (FV) dynamical core that was initially developed and used at the NASA Data Assimilation Oce (DAO) for data assimilation, numerical weather predictions, and climate simulations. The nite-volume discretization is local and entirely in physical space. The horizontal discretization is based on a conservative ux-form semiLagrangian scheme described by Lin and Rood [1996] (hereafter LR96) and Lin and Rood [1997] (hereafter LR97). The vertical discretization can be best described as Lagrangian with a conservative re-mapping, which essentially makes it quasi-Lagrangian. The quasi-Lagrangian aspect of the vertical coordinate is transparent to model users or physical parameterization developers, and it functions exactly like the coordinate (a hybrid p coordinate) used by other dynamical cores within CAM. In the current implementation for use in CAM, the FV dynamics and physics are time split in the sense that all prognostic variables are updated sequentially by the dynamics and then the physics. The time integration within the FV dynamics is fully explicit, with sub-cycling within the 2D Lagrangian dynamics to stabilize the fastest wave (see section 3.1.4). The transport for tracers, however, can take a much larger time step (e.g., 30 minutes as for the physics).

688

3.1.2

The governing equations for the hydrostatic atmosphere

689 690 691

For reference purposes, we present the continuous dierential equations for the hydrostatic 3D atmospheric ow on the sphere for a general vertical coordinate (e.g., Kasahara [1974]). Using standard notations, the hydrostatic balance equation is given as follows: 1 p + g = 0, z (3.1)

692 693 694

where is the density of the air, p the pressure, and g the gravitational constant. Introducing p the pseudo-density = (i.e., the vertical pressure gradient in the general coordinate), from the hydrostatic balance equation the pseudo-density and the true density are related as follows:

15

=
695 696 697

(3.2)

where = gz is the geopotential. Note that reduces to the true density if = gz, and p the surface pressure Ps if = ( = Ps ). The conservation of total air mass using as the prognostic variable can be written as + V = 0, t (3.3)

698 699

where V = (u, v, d ). Similarly, the mass conservation law for tracer species (or water vapor) dt can be written as (q) + V q = 0, (3.4) t where q is the mass mixing ratio (or specic humidity) of the tracers (or water vapor). Choosing the (virtual) potential temperature as the thermodynamic variable, the rst law of thermodynamics is written as () + V = 0. (3.5) t Letting (, ) denote the (longitude, latitude) coordinate, the momentum equations can be written in the vector-invariant form as follows: 1 1 d u u = v ( + D) + p , t Acos dt 1 1 d v v = u ( + D) + p , t A dt (3.6) (3.7)

700 701 702

703 704

where A is the radius of the earth, is the coecient for the optional divergence damping, D is the horizontal divergence D= 1 (u) + (v cos) , Acos =

705

1 2 u + v2 , 2 and , the vertical component of the absolute vorticity, is dened as follows: = 2 sin + 1 v (u cos) , A cos

706 707 708 709 710 711

where is the angular velocity of the earth. Note that the last term in (3.6) and (3.7) vanishes if the vertical coordinate is a conservative quantity (e.g., entropy under adiabatic conditions [Hsu and Arakawa, 1990] or an imaginary conservative tracer), and the 3D divergence operator becomes 2D along constant surfaces. The discretization of the 2D horizontal transport process is described in section 3.1.3. The complete dynamical system using the Lagrangian control-volume vertical discretization is described in section 3.1.4 and section 3.1.5 describes 16

712 713 714 715 716

the explicit diusion operators available in CAM5. A mass, momentum, and total energy conservative mapping algorithm is described in section 3.1.6 and in section 3.1.7 an alternative geopotential conserving vertical remapping method is described. Sections 3.1.8 and 3.1.9 are on the adjusctment of pressure to include the change in mass of water vapor and on the negative tracer xer in CAM, respectively. Last the global energy xer is described (section 3.1.10).

717

3.1.3

Horizontal discretization of the transport process on the sphere

718 719 720 721

Since the vertical transport term would vanish after the introduction of the vertical Lagrangian control-volume discretization (see section 3.1.4), we shall present here only the 2D (horizontal) forms of the FFSL transport algorithm for the transport of density (3.3) and mixing ratio-like quantities (3.4) on the sphere. The governing equation for the pseudo-density (3.3) becomes 1 + (u) + (v cos) = 0. t Acos (3.8)

722

The nite-volume (integral ) representation of the continuous eld is dened as follows: (t) 1 A2 cos (t; , )A2 cos dd. (3.9)

723 724

Given the exact 2D wind eld V (t; , ) = (U, V ) the 2D integral representation of the conservation law for can be obtained by integrating (3.8) in time and in space n+1 = n 1 2 cos A
t+t t

(t; , ) V dl dt. n

(3.10)

725 726 727 728 729

The above 2D transport equation is still exact for the nite-volume under consideration. To carry out the contour integral, certain approximations must be made. LR96 essentially decomposed the ux integral using two orthogonal 1D ux-form transport operators. Introducing the following dierence operator x q = q(x + x x ) q(x ), 2 2

730 731

and assuming (u , v ) is the time-averaged (from time t to time t + t) V on the C-grid (e.g., Fig. 1 in LR96), the 1-D nite-volume ux-form transport operator F in the -direction is 1 F (u , t, ) = Acos
t+t t

U dt =

t [(u , t; )] , Acos

(3.11)

732 733

where , the time-accumulated (from t to t+t) mass ux across the cell wall, is dened as follows, (u , t; ) = 1 t
t+t t

U dt u (u , t, ),

(3.12)

17

and (u , t; )
734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773

1 t

t+t

dt
t

(3.13)

can be interpreted as a time mean (from time t to time t + t) pseudo-density value of all material that passed through the cell edge from the upwind direction. Note that the above time integration is to be carried out along the backward-in-time trajectory of the cell edge position from t = t + t (the arrival point; (e.g., point B in Fig. 3 of LR96) back to time t (the departure point; e.g., point B in Fig. 3 of LR96). The very essence of the 1D nite-volume algorithm is to construct, based on the given initial cell-mean values of , an approximated subgrid distribution of the true eld, to enable an analytic integration of (3.13). Assuming there is no error in obtaining the time-mean wind (u ), the only error produced by the 1D transport scheme would be solely due to the approximation to the continuous distribution of within the subgrid under consideration (this is not the case in 2D; Lauritzen et al. [2010]). From this perspective, it can be said that the 1D nite-volume transport algorithm combines the time-space discretization in the approximation of the time-mean cell-edge values . The physically correct way of approximating the integral (3.13) must be upwind, in the sense that it is integrated along the backward trajectory of the cell edges. For example, a center dierence approximation to (3.13) would be physically incorrect, and consequently numerically unstable unless articial numerical diusion is added. Central to the accuracy and computational eciency of the nite-volume algorithms is the degrees of freedom that describe the subgrid distribution. The rst order upwind scheme, for example, has zero degrees of freedom within the volume as it is assumed that the subgrid distribution is piecewise constant having the same value as the given volume-mean. The second order nite-volume scheme (e.g., Lin et al. [1994]) assumes a piece-wise linear subgrid distribution, which allows one degree of freedom for the specication of the slope of the linear distribution to improve the accuracy of integrating (3.13). The Piecewise Parabolic Method (PPM, Colella and Woodward [1984]) has two degrees of freedom in the construction of the second order polynomial within the volume, and as a result, the accuracy is signicantly enhanced. The PPM appears to strike a good balance between computational eciency and accuracy. Therefore, the PPM is the basic 1D scheme we chose (see, e.g., Machenhauer [1998]). Note that the subgrid PPM distributions are compact, and do not extend beyond the volume under consideration. The accuracy is therefore signicantly better than the order of the chosen polynomials implies. While the PPM scheme possesses all the desirable attributes (mass conserving, monotonicity preserving, and high-order accuracy) in 1D, it is important that a solution be found to avoid the directional splitting in the multi-dimensional problem of modeling the dynamics and transport processes of the Earths atmosphere. The rst step for reducing the splitting error is to apply the two orthogonal 1D ux-form operators in a directionally symmetric way. After symmetry is achieved, the inner operators are then replaced with corresponding advective-form operators (in CAM5 the inner operators are based on constant cell-average values and not the PPM). A stability analysis of the consequences of using dierent inner and outer operators in the LR96 scheme is given in Lauritzen [2007]. A consistent advective-form operator in the direction can be derived from its ux-form counterpart (F ) as follows:
f (u, t, ) = F (u , t, ) + F (u, t, 1) = F (u, t, ) + Cdef ,

(3.14)

18

774 775 776 777

t u = , (3.15) Acos where Cdef is a dimensionless number indicating the degree of the ow deformation in the direction. The above derivation of f is slightly dierent from LR96s approach, which adopted the traditional 1D advective-form semi-Lagrangian scheme. The advantage of using (3.14) is that computation of winds at cell centers (Eq. 2.25 in LR96) are avoided. Analogously, the 1D ux-form transport operator G in the latitudinal () direction is derived as follows:
Cdef

G(v , t, ) =

1 Acos

t+t t

V cos dt =

t [v cos ] , Acos

(3.16)

and likewise the advective-form operator,

g(v , t, ) = G(v , t, ) + Cdef ,

(3.17)

778 779

t [v cos] . (3.18) Acos To complete the construction of the 2D algorithm on the sphere, we introduce the following short hand notations:
Cdef =

where

1 ( ) = ( )n + g [v , t, ( )n ] , 2 1 ( ) = ( )n + f [u , t, ( )n ] . 2 The 2D transport algorithm (cf, Eq. 2.24 in LR96) can then be written as n+1 = n + F u, t, + G v , t, . Using explicitly the mass uxes (, Y ), (3.21) is rewritten as n+1 = n
780 781 782

(3.19) (3.20)

(3.21)

t Acos

1 1 (u , t; ) + cos Y (v , t; )

(3.22)

where Y , the mass ux in the meridional direction, is dened in a similar fashion as (3.12). The ability of the LR96 scheme to approximate the exact geometry of the uxes for deformational ows is discussed in Machenhauer et al. [2009] and Lauritzen et al. [2010]. It can be veried that in the special case of constant density ow ( = constant) the above equation degenerates to the nite-dierence representation of the incompressibility condition of the time mean wind eld (u , v ), i.e., 1 1 u + (v cos) = 0. (3.23)

19

The fulllment of the above incompressibility condition for constant density ows is crucial to the accuracy of the 2D ux-form formulation. For transport of volume mean mixing ratio-like quantities (q) the mass uxes (, Y ) as dened previously should be used as follows q n+1 =
783 784 785 786

1 n+1

n q n + F (, t, q ) + G(Y, t, q ) .

(3.24)

Note that the above form of the tracer transport equation consistently degenerates to (3.21) if q 1 (i.e., the tracer density equals to the background air density), which is another important condition for a ux-form transport algorithm to be able to avoid generation of noise (e.g., creation of articial gradients) and to maintain mass conservation.

787 788

3.1.4

A vertically Lagrangian and horizontally Eulerian controlvolume discretization of the hydrodynamics

789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812

The very idea of using Lagrangian vertical coordinate for formulating governing equations for the atmosphere is not entirely new. Starr [1945]) is likely the rst to have formulated, in the continuous dierential form, the governing equations using a Lagrangian coordinate. Starr did not make use of the discrete Lagrangian control-volume concept for discretization nor did he present a solution to the problem of computing the pressure gradient forces. In the nite-volume discretization to be described here, the Lagrangian surfaces are treated as the bounding material surfaces of the Lagrangian control-volumes within which the nite-volume algorithms developed in LR96, LR97, and L97 will be directly applied. To use a vertical Lagrangian coordinate system to reduce the 3D governing equations to the 2D forms, one must rst address the issue of whether it is an inertial coordinate or not. For hydrostatic ows, it is. This is because both the right-hand-side and the left-hand-side of the vertical momentum equation vanish for purely hydrostatic ows. Realizing that the earths surface, for all practical modeling purposes, can be regarded as a non-penetrable material surface, it becomes straightforward to construct a terrain-following Lagrangian control-volume coordinate system. In fact, any commonly used terrain-following coordinates can be used as the starting reference (i.e., xed, Eulerian coordinate) of the oating Lagrangian coordinate system. To close the coordinate system, the model top (at a prescribed constant pressure) is also assumed to be a Lagrangian surface, which is the same assumption being used by practically all global hydrostatic models. The basic idea is to start the time marching from the chosen terrain-following Eulerian coordinate (e.g., pure or hybrid -p), treating the initial coordinate surfaces as material surfaces, the nite-volumes bounded by two coordinate surfaces, i.e., the Lagrangian control-volumes, are free vertically, to oat, compress, or expand with the ow as dictated by the hydrostatic dynamics. By choosing an imaginary conservative tracer that is a monotonic function of height and constant on the initial reference coordinate surfaces (e.g., the value of in the hybrid p coordinate used in CAM), the 3D governing equations written for the general vertical coordinate in section 1.2 can be reduced to 2D forms. After factoring out the constant , (3.3), the p conservation law for the pseudo-density ( = ), becomes 1 p + (up) + (vp cos) = 0, t Acos 20 (3.25)

813 814 815 816 817 818

where the symbol represents the vertical dierence between the two neighboring Lagrangian surfaces that bound the nite control-volume. From (3.1), the pressure thickness p of that control-volume is proportional to the total mass, i.e., p = gz. Therefore, it can be said that the Lagrangian control-volume vertical discretization has the hydrostatic balance built-in, and p can be regarded as the pseudo-density for the discretized Lagrangian vertical coordinate system. Similarly, (3.4), the mass conservation law for all tracer species, is 1 (qp) + (uqp) + (vqp cos) = 0, t Acos (3.26)

819

the thermodynamic equation, (3.5), becomes 1 (p) + (up) + (vp cos) = 0, t Acos and (3.6) and (3.7), the momentum equations, are reduced to 1 1 u = v ( + D) + p , t Acos 1 1 v = u ( + D) + p . t A (3.28) (3.29) (3.27)

Given the prescribed pressure at the model top P , the position of each Lagrangian surface Pl (horizontal subscripts omitted) is determined in terms of the hydrostatic pressure as follows:
l

Pl = P +
k=1
820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836

Pk ,

(f or l = 1, 2, 3, ..., N),

(3.30)

where the subscript l is the vertical index ranging from 1 at the lower bounding Lagrangian surface of the rst (the highest) layer to N at the Earths surface. There are N+1 Lagrangian surfaces to dene a total number of N Lagrangian layers. The surface pressure, which is the pressure at the lowest Lagrangian surface, is easily computed as PN using (3.30). The surface pressure is needed for the physical parameterizations and to dene the reference Eulerian coordinate for the mapping procedure (to be described in section 3.1.6). With the exception of the pressure-gradient terms and the addition of a thermodynamic equation, the above 2D Lagrangian dynamical system is the same as the shallow water system described in LR97. The conservation law for the depth of uid h in the shallow water system of LR97 is replaced by (3.25) for the pressure thickness p. The ideal gas law, the mass conservation law for air mass, the conservation law for the potential temperature (3.27), together with the modied momentum equations (3.28) and (3.29) close the 2D Lagrangian dynamical system, which are vertically coupled only by the hydrostatic relation (see (3.54), section 3.1.6). The time marching procedure for the 2D Lagrangian dynamics follows closely that of the shallow water dynamics fully described in LR97. For computational eciency, we shall take advantage of the stability of the FFSL transport algorithm by using a much larger time step (t) for the transport of all tracer species (including water vapor). As in the shallow water 21

837 838 839 840 841

system, the Lagrangian dynamics uses a relatively small time step, = t/m, where m is the number of the sub-cycling needed to stabilize the fastest wave in the system. We shall describe here this time-split procedure for the prognostic variables [p, , u, v; q] on the D-grid. Discretization on the C-grid for obtaining the diagnostic variables, the time-averaged winds (u , v ), is analogous to that of the D-grid (see also LR97). Introducing the following short hand notations (cf, (3.19) and (3.20)): ( ) = ( )n+ i ( ) = ( )n+ i
i1 m i1 1 + g[vi , , ( )n+ m ], 2

i1 1 + f [u , , ( )n+ m ], i 2 and applying directly (3.22), the update of pressure thickness p, using the fractional time step = t/m, can be written as i1 m

pn+ m = pn+

i1 m

Acos

1 1 x (u , ; p ) + cos yi (vi , ; p ) i i i i (f or i = 1, ..., m),

(3.31)

where [x , yi ] are the background air mass uxes, which are then used as input to Eq. 24 for i transport of the potential temperature :

n+ m =

1 p
i n+ m

pn+

i1 m

n+

i1 m

+ F (x , ; ) + G(yi , , ) . i i i

(3.32)

The discretized momentum equations for the shallow water system (cf, Eq. 16 and Eq. 17 in LR97) are modied for the pressure gradient terms as follows: un+ m = un+
i i i1 m

+ yi vi , ;

1 ( D ) + P , Acos 1 ( D ) P , A

(3.33) (3.34)

v n+ m = v n+

i1 m

x u , ; + i i

where is the upwind-biased kinetic energy (as dened by Eq. 18 in LR97), and D , the horizontal divergence on the D-grid, is discretized as follows: D =
i1 i1 1 1 1 un+ m + v n+ m cos Acos

The nite-volume mean pressure-gradient terms in (3.33) and (3.34) are computed as follows: P = d , Acos d

(3.35)

P =

d , A d

(3.36)

22

842 843 844 845 846 847

where = p ( = R/Cp ), and the symbols and indicate that the contour integrations are to be carried out, using the nite-volume algorithm described in L97, in the (, ) and (, ) space, respectively. To complete one time step, equations (3.31-3.34), together with their counterparts on the C-grid are cycled m times using the fractional time step , which are followed by the tracer transport using (3.26) with the large-time-step t. Mass uxes (x , y ) and the winds (u , v ) on the C-grid are accumulated for the large-timestep transport of tracer species (including water vapor) q as q n+1 = 1 q n pn + F (X , t, q ) + G(Y , t, q ) , pn+1 (3.37)

where the time-accumulated mass uxes (X , Y ) are computed as

m

X =
i=1 m

x (u , , p ), i i i

(3.38)

Y =
i=1
848 849

yi (vi , , p ). i

(3.39)

The time-averaged winds (U , V ), dened as follows, are to be used as input for the computations of q and q : 1 U = m
m

u , i
i=1 m vi . i=1

(3.40)

1 V = m

850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865

(3.41)

The use of the time accumulated mass uxes and the time-averaged winds for the largetime-step tracer transport in the manner described above ensures the conservation of the tracer mass and maintains the highest degree of consistency possible given the time split integration procedure. A graphical illustration of the dierent levels of sub-cycling in CAM5 is given on Figure 3.1. The algorithm described here can be readily applied to a regional model if appropriate boundary conditions are supplied. There is formally no Courant number related time step restriction associated with the transport processes. There is, however, a stability condition imposed by the gravity-wave processes. For application on the whole sphere, it is computationally advantageous to apply a polar lter to allow a dramatic increase of the size of the small time step . The eect of the polar lter is to stabilize the short-in-wavelength (and high-in-frequency) gravity waves that are being unnecessarily and unidirectionally resolved at very high latitudes in the zonal direction. To minimize the impact to meteorologically signicant larger scale waves, the polar lter is highly scale selective and is applied only to the diagnostic variables on the auxiliary C-grid and the tendency terms in the D-grid momentum equations. No polar lter is applied directly to any of the prognostic variables.

23

Physics Vertical remapping Tracers Dynamics

t physics t remap t t t t remap t

Figure 3.1: A graphical illustration of the dierent levels of sub-cycling in CAM5.

866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886

The design of the polar lter follows closely that of Suarez and Takacs [1995] for the C-grid Arakawa type dynamical core (e.g., Arakawa and Lamb [1981]). For the CAM 5.0 the fastfourier transform component of the polar ltering has replaced the algebraic form at all ltering latitudes. Because our prognostic variables are computed on the D-grid and the fact that the FFSL transport scheme is stable for Courant number greater than one, in realistic test cases the maximum size of the time step is about two to three times larger than a model based on Arakawa and Lambs C-grid dierencing scheme. It is possible to avoid the use of the polar lter if, for example, the Cubed grid is chosen, instead of the current latitude-longitude grid. rewrite of the rest of the model codes including physics parameterizations, the land model, and most of the post processing packages. The size of the small time step for the Lagrangian dynamics is only a function of the horizontal resolution. Applying the polar lter, for the 2-degree horizontal resolution, a small-time-step size of 450 seconds can be used for the Lagrangian dynamics. From the large-time-step transport perspective, the small-time-step integration of the 2D Lagrangian dynamics can be regarded as a very accurate iterative solver, with m iterations, for computing the time mean winds and the mass uxes, analogous in functionality to a semi-implicit algorithms elliptic solver (e.g., Ringler et al. [2000]). Besides accuracy, the merit of an explicit versus semi-implicit algorithm ultimately depends on the computational eciency of each approach. In light of the advantage of the explicit algorithm in parallelization, we do not regard the explicit algorithm for the Lagrangian dynamics as an impedance to computational eciency, particularly on modern parallel computing platforms.

887

3.1.5

Optional diusion operators in CAM5

The CD-grid discretization method used in the CAM nite-volume dynamical core provides explicit control over the rotational modes at the grid scale, due to monotonicity constraint in the PPM-based advection, but there is no explicit control over the divergent modes at the grid scale [see, e.g., Skamarock, 2010]. Therefore divergence damping terms appear on the right-hand side of the momentum equations ((3.28) and (3.29)): 1 (D) Acos (3.42)

24

and

1 (D) , A

(3.43)

888 889 890 891

respectively, where the strength of the divergence damping is controlled by the coecient given by 2 (A2 ) = , (3.44) t where 2 = 1/128 throughout the atmosphere except in the top model levels where it monotonically increases to approximately 4/128 at the top of the atmosphere. The divergence damping described above is referred to as second-order divergence damping as it eectively damps divergence with a 2 operator. In CAM5 optional fourth-order divergence damping has been implemented where the divergence is eectively damped with a 4 -operator which is usually more scale selective than second-order damping operators. For fourth-order divergence damping the terms and 1 4 2 D Acos 1 4 2 D A , (3.45)

(3.46)

are added to the right-hand side of (3.28) and (3.29), respectively. The horizontal Laplacian 2 -operator in spherical coordinates for a scalar is given by 2 = 1 2 1 + 2 2 cos2 2 A A cos cos . (3.47)

The fourth-order divergence damping coecient is given by 4 = 0.01 A2 cos()

892 893 894 895

/t.

(3.48)

Since divergence damping is added explicitly to the equations of motion it is unstable if the time-step is too large or the damping coecients ( or 4 ) are too large. To stabilize the fourthorder divergence damping the winds used to compute the divergence are ltered using the same FFT ltering which is applied to stabilize the gravity waves. To control potentially excessive polar night jets in high-resolution congurations of CAM, Laplacian damping of the wind components has been added as an option in CAM5. That is, the terms del2 2 u (3.49) and del2 2 v (3.50)

896 897 898 899 900

are added to the right-hand side of the momentum equations (3.28) and (3.29), respectively. The damping coecient del2 is zero throughout the atmosphere except in the top layers where it increases monotonically and smoothly from zero to approximately four times a user-specied damping coecient at the top of the atmosphere (the user-specied damping coecient is typically on the order of 2.5 105 m2 sec1 ). 25

901 902

3.1.6

A mass, momentum, and total energy conserving mapping algorithm

903 904 905 906 907 908 909 910 911 912 913 914 915

The Lagrangian surfaces that bound the nite-volume will eventually deform, particularly in the presence of persistent diabatic heating/cooling, in a time scale of a few hours to a day depending on the strength of the heating and cooling, to a degree that it will negatively impact the accuracy of the horizontal-to-Lagrangian-coordinate transport and the computation of the pressure gradient forces. Therefore, a key to the success of the Lagrangian control-volume discretization is an accurate and conservative algorithm for mapping the deformed Lagrangian coordinate back to a xed reference Eulerian coordinate. There are some degrees of freedom in the design of the vertical mapping algorithm. To ensure conservation, our current (and recommended) mapping algorithm is based on the reconstruction of the mass (pressure thickness p), zonal and meridional winds, tracer mixing ratios, and total energy (volume integrated sum of the internal, potential, and kinetic energy), using the monotonic Piecewise Parabolic sub-grid distributions with the hydrostatic pressure (as dened by (3.30)) as the mapping coordinate. We outline the mapping procedure as follows. Step 1: Dene a suitable Eulerian reference coordinate as a target coordinate. The mass in each layer (p) is then distributed vertically according to the chosen Eulerian coordinate. The surface pressure typically plays an anchoring role in dening the terrain following Eulerian vertical coordinate. The hybrid coordinate used in the NCAR CCM3 [Kiehl et al., 1996] is adopted in the current model setup. Step 2: Construct the piece-wise continuous vertical subgrid proles of tracer mixing ratios (q), zonal and meridional winds (u and v ), and total energy () in the Lagrangian control-volume coordinate, or the source coordinate. The total energy is computed as the sum of the nite-volume integrated geopotential , internal energy (Cv Tv ), and the kinetic energy (K) as follows: = 1 p Cv Tv + + 1 2 u + v2 2 dp. (3.51)

916 917 918 919 920 921

Applying integration by parts and the ideal gas law, the above integral can be rewritten as = 1 p Cp Tv + 1 2 u + v2 2 dp + d (p) (3.52)

= Cp Tv +

1 (p) + K, p

where Tv is the layer mean virtual temperature, K is the layer mean kinetic energy, p is the pressure at layer edges, and Cv and Cp are the specic heat of the air at constant volume and at constant pressure, respectively. The total energy

26

in each grid cell is calculated as i,j,k = Cp Tvi,j,k + 1 2

922 923 924 925

1 1 1 p 1 1 p i,j,k 2 i,j,k 2 + pi,j,k i,j,k+ 2 i,j,k+ 2 2 2 u2 1 ,k + u2 1 ,k vi 1 ,j,k + vi+ 1 ,j,k i,j+ 2 i,j 2 2 2 + 2 2 to create subgrid proles is set by the ag te method. (default), the Piece-wise Parabolic Method (PPM, [1984]) over a pressure coordinate is used and for spline over a logarithmic pressure coordinate is used.

The method employed For te method = 0 Colella and Woodward te method = 1 a cublic

Step 3: Layer mean values of q, (u, v ), and in the Eulerian coordinate system are obtained by integrating analytically the sub-grid distributions, in the vertical direction, from model top to the surface, layer by layer. Since the hydrostatic pressure is chosen as the mapping coordinate, tracer mass, momentum, and total energy are locally and globally conserved. In mapping a variable from the source coordinate to the target coordinate, dierent limiter constraints may be used and they are controlled by two ags, iv and kord. For winds on D-grid, iv should be set to -1. For tracers, iv should be set to 0. For all others, iv = 1. kord directly controls which limiter constraint is used. For kord 7, Huynhs 2nd constraint is used. If kord = 7, the original quasi-monotonic constraint is used. If kord > 7, a full monotonic constraint is used. If kord is less than 7, the variable, lmt, is determined by the following: lmt = kord 3, lmt = max(0, lmt), if(iv = 0) lmt = min(2, lmt).
926 927 928

If lmt = 0, a standard PPM constraint is used. If lmt = 1, an improved full monotonicity constraint is used. If lmt = 2, a positive denite constraint is used. If lmt = 3, the algorithm will do nothing. Step 4: Retrieve virtual temperature in the Eulerian (target) coordinate. Start by computing kinetic energy in the Eulerian coordinate system for each layer. Then substitute kinetic energy and the hydrostatic relationship into (3.52). The layer mean temperature Tv k for layer k in the Eulerian coordinate is then retrieved from the reconstructed total energy (done in Step 3) by a fully explicit integration procedure starting from the surface up to the model top as follows: Tv k = k Kk k+ 1
2

929 930 931 932 933 934

Cp 1 pk 1
2 935

ln pk+ 1 ln pk 1
2 2

(3.53)

pk+ 1 pk 1

where = Rd /Cp and Rd is the gas constant for dry air.

27

To convert the potential virtual temperature v to the layer mean temperature the conversion factor is obtained by equating the following two equivalent forms of the hydrostatic relation for and Tv : = Cp v , (3.54) where = p . The conversion formula between layer mean temperature and layer mean potential temperature is obtained as follows: v =
936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953

= Rd Tv ln p,

(3.55)

lnp Tv .

(3.56)

The physical implication of retrieving the layer mean temperature from the total energy as described in Step 3 is that the dissipated kinetic energy, if any, is locally converted into internal energy via the vertically sub-grid mixing (dissipation) processes. Due to the monotonicity preserving nature of the sub-grid reconstruction the column-integrated kinetic energy inevitably decreases (dissipates), which leads to local frictional heating. The frictional heating is a physical process that maintains the conservation of the total energy in a closed system. As viewed by an observer riding on the Lagrangian surfaces, the mapping procedure essentially performs the physical function of the relative-to-the-Eulerian-coordinate vertical transport, by vertically redistributing (air and tracer) mass, momentum, and total energy from the Lagrangian control-volume back to the Eulerian framework. As described in section 3.1.4, the model time integration cycle consists of m small time steps for the 2D Lagrangian dynamics and one large time step for tracer transport. The mapping time step can be much larger than that used for the large-time-step tracer transport. In tests using the Held-Suarez forcing [Held and Suarez, 1994], a three-hour mapping time interval is found to be adequate. In the full model integration, one may choose the same time step used for the physical parameterizations so as to ensure the input state variables to physical parameterizations are in the usual Eulerian vertical coordinate. In CAM5, vertical remapping takes place at each physics time step.

954

3.1.7

A geopotential conserving mapping algorithm

An alternative vertical mapping approach is available in CAM5. Instead of retrieving temperature by remapped total energy in the Eulerian coordinate, the alternative approach maps temperature directly from the Lagrangian coordinate to the Eulerian coordinate. Since geopotential is dened as = Cp v = Rd Tv ln p,
955 956 957 958 959 960 961

mapping v over or Tv over ln p preserves the geopotential at the model lid. This approach prevents the mapping procedure from generating spurious pressure gradient forces at the model lid. Unlike the energy-conserving algorithm which could produce substantial temperature uctuations at the model lid, the geopotential conserving approach guarantees a smooth (potential) temperature prole. However, the geopotential conserving does not conserve total energy in the remapping procedure. This may be resolved by a global energy xer already implemented in the model (see section 3.1.10). 28

962 963

3.1.8

Adjustment of pressure to include change in mass of water vapor

964 965 966 967 968 969 970 971

The physics parameterizations operate on a model state provided by the dynamics, and are allowed to update specic humidity. However, the surface pressure remains xed throughout the physics updates, and since there is an explicit relationship between the surface pressure and the air mass within each layer, the total air mass must remain xed as well throughout the physics updates. If no further correction were made, this would imply that the dry air mass changed if the water vapor mass changed in the physics updates. Therefore the pressure eld is changed to include the change in water vapor mass due to the physics updates. We impose the restrictions that dry air mass and water mass are conserved as follows: The total pressure p is p = d + e. (3.57) with dry pressure d, water vapor pressure e. The specic humidity is q= e e = , p d+e d = (1 q)p. (3.58)

We dene a layer thickness as k p pk+1/2 pk1/2 , so k d = (1 q k ) k p.

972 973

(3.59)

We are concerned about 3 time levels: qn is input to physics, qn is output from physics, qn+1 is the adjusted value for dynamics. Dry mass is the same at n and n + 1 but not at n. To conserve dry mass, we require that k dn = k dn+1 or
k k (1 qn ) k pn = (1 qn+1 ) k pn+1 .

(3.60) (3.61)

Water mass is the same at n and n + 1, but not at n. To conserve water mass, we require that k k qn k pn = qn+1 k pn+1 . (3.62) Substituting (3.62) into (3.61),
k k (1 qn ) k pn = k pn+1 qn k pn k k k pn+1 = (1 qn + qn ) k pn

(3.63) (3.64)

which yields a modied specic humidity for the dynamics:

k k qn+1 = qn k qn k pn . = k k k pn+1 1 qn + qn

(3.65)

974 975

We note that this correction as implemented makes a small change to the water vapor as well. The pressure correction could be formulated to leave the water vapor unchanged.

29

976

3.1.9

Negative Tracer Fixer

977 978 979 980 981 982 983 984 985 986 987 988

In the Finite Volume dynamical core, neither the monotonic transport nor the conservative vertical remapping guarantee that tracers will remain positive denite. Thus the Finite Volume dynamical core includes a negative tracer xer applied before the parameterizations are calculated. For negative mixing ratios produced by horizontal transport, the model will attempt to borrow mass from the east and west neighboring cells. In practice, most negative values are introduced by the vertical remapping which does not guarantee positive deniteness in the rst and last layer of the vertical column. A minimum value qmin is dened for each tracer. If the tracer falls below that minimum value, it is set to that minimum value. If there is enough mass of the tracer in the layer immediately above, tracer mass is removed from that layer to conserve the total mass in the column. If there is not enough mass in the layer immediately above, no compensation is applied, violating conservation. Usually such computational sources are very small. The amount of tracer needed from the layer above to bring qk up to qmin is qf ill = (qmin qk ) pk pk1 (3.66)

where k is the vertical index, increasing downward. After the lling qkF ILLED = qmin qk1F ILLED = qk1 qf ill (3.67) (3.68)

989 990

Currently qmin = 1.0 1012 for water vapor, qmin = 0.0 for CLDLIQ, CLDICE, NUMLIQ and NUMICE, and qmin = 1.0 1036 for the remaining constituents.

991

3.1.10

Global Energy Fixer

992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008

The nite-volume dynamical core as implemented in CAM and described here conserves the dry air and all other tracer mass exactly without a mass xer. The vertical Lagrangian discretization and the associated remapping conserves the total energy exactly. The only remaining issue regarding conservation of the total energy is the horizontal discretization and the use of the diusive transport scheme with monotonicity constraint. To compensate for the loss of total energy due to horizontal discretization, we apply a global xer to add the loss in kinetic energy due to diusion back to the thermodynamic equation so that the total energy is conserved. The loss in total energy (in ux unit) is found to be around 2 (W/m2 ) with the 2 degrees resolution. The energy xer is applied following the negative tracer xer. The xer is applied on the unstaggered physics grid rather than on the staggered dynamics grid. The energies on these two grids are dicult to relate because of the nonlinear terms in the energy denition and the interpolation of the state variables between the grids. The energy is calculated in the parameterization suite before the state is passed to the nite volume core as described in the beginning of Chapter 4. The xer is applied just before the parameterizations are calculated. The xer is a simplication of the xer in the Eulerian dynamical core described in section 3.3.20. 30

Let minus sign superscript ( ) denote the values at the beginning of the dynamics time step, i.e. after the parameterizations are applied, let a plus sign superscript ( )+ denote the + values after xer is applied, and let a hat ( ) denote the provisional value before adjustment. The total energy over the entire computational domain after the xer is 1 1 +2 2 + + Cp T + + + u + v + + (Lv + Li ) qv + Li q A2 cos d d dp, g 2 0 pt 2 (3.69) where Lv is the latent heat of vaporation, Li is the latent heat of fusion, qv is water vapor mixing ratio, and q is cloud water mixing ratio. E + should equal the energy at the beginning of the dynamics time step E =
+ ps 2
2

1 1 2 2 cp T + + u + v + (Lv + Li ) qv + Li q A2 cos d d dp. g 2 2 0 pt (3.70) + denote the energy of the provisional state provided by the dynamical core before the Let E adjustment. E =

ps

1 1 +2 2 + + cp T + + + + u + v + + (Lv + Li ) qv + Li q A2 cos d d dp. g 2 2 0 pt (3.71) + E . The energy Thus, the total energy added into the system by the dynamical core is E xer then changes dry static energy (s = Cp T + ) by a constant amount over each grid cell to conserve total energy in the entire computational domain. The dry static energy added to each grid cell may be expressed as E+ = s = Therefore, s+ = s+ + s, or Cp T + + + = s+ + s.
1009 2 2

ps

ps pt

2 0

E E+

A2 cos d d dp g

(3.72)

(3.73) (3.74)

This will ensure E + = E . By hydrostatic approximation, the geopotential equation is d = Rd Tv d lnp, and for any arbitrary point between pk+ 1 and pk 1 the geopotential may be written as
2 2

(3.75)

k+ 1
2

d = Rd Tv
2

p pk+ 1
2

d lnp ,

(3.76) (3.77)

= k+ 1 + Rd Tv lnpk+ 1 lnp .
2

31

The geopotential at the mid point of a model layer between pk+ 1 and pk 1 , or the layer mean, 2 2 is
pk + 1 2

k =

pk 1 2
1 pk 2 1 pk + 2

dp dp k+ 1 + Rd Tv lnpk+ 1 lnp
2 2

pk + 1 2

pk 1 2

dp

pk + 1 2

1 pk 2

dp
pk + 1 2 pk 1 2
2

lnp dp
2

= k+ 1 + Rd Tv lnpk+ 1
2 2

pk+ 1 pk 1
2 2 2

= k+ 1 + Rd Tv
2

1 pk 1

lnpk+ 1 lnpk 1
2

pk+ 1 pk 1
2

(3.78)

For layer k, the energy xer will solve the following equation based on (3.74),
+ Cp Tk

+ 1 k+ 2

+ Rd Tk

1+

+ qvk

p+ 1 k 2

lnp+ 1 lnp+ 1 k+ k
2 2

p+ 1 k+ 2

p+ 1 k 2

= s+ + s.

(3.79)

Since the energy xer will not alter the water vapor mixing ratio and the pressure eld,
+ qv = qv , + p+ = p+ .

(3.80) (3.81)

Therefore,
+ Tk

(+ + s) + 1 s k+ Cp + Rd 1 + vk q+ 1 p+ 1 k 2

ln+ p

ln+ 1 p k 2 k+ 1 2 + + p 1 1 p k+ 2 k 2

(3.82)

1010 1011 1012 1013

The energy xer starts from the Earths surface and works its way up to the model top in adjusting the temperature eld. At the surface layer, + 1 = s . After the temperature is k+ 2 adjusted in a grid cell, the geopotential at the upper interface of the cell is updated which is needed for the temperature adjustment in the grid cell above.

1014

3.1.11

Further discussion

1015 1016 1017 1018 1019 1020 1021

There are still aspects of the numerical formulation in the nite volume dynamical core that can be further improved. For example, the choice of the horizontal grid, the computational eciency of the split-explicit time marching scheme, the choice of the various monotonicity constraints, and how the conservation of total energy is achieved. The impact of the non-linear diusion associated with the monotonicity constraint is difcult to assess. All discrete schemes must address the problem of subgrid-scale mixing. The nite-volume algorithm contains a non-linear diusion that mixes strongly when monotonicity 32

1022 1023 1024 1025 1026

principles are locally violated. However, the eect of nonlinear diusion due to the imposed monotonicity constraint diminishes quickly as the resolution matches better to the spatial structure of the ow. In other numerical schemes, however, an explicit (and tunable) linear diusion is often added to the equations to provide the subgrid-scale mixing as well as to smooth and/or stabilize the time marching.

1027

3.2

Spectral Element Dynamical Core

1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062

The CAM includes an optional dynamical core from HOMME, NCARs High-Order Method Modeling Environment [Dennis et al., 2005]. The stand-alone HOMME is used for research in several dierent types of dynamical cores. The dynamical core incorporated into CAM4 uses HOMMEs continuous Galerkin spectral nite element method [Taylor et al., 1997; Fournier et al., 2004; Thomas and Loft, 2005; Wang et al., 2007; Taylor and Fournier, 2010], here abbreviated to the spectral element method (SEM). This method is designed for fully unstructured quadrilateral meshes. The current congurations in the CAM are based on the cubed-sphere grid. The main motivation for the inclusion of HOMME is to improve the scalability of the CAM by introducing quasi-uniform grids which require no polar lters [Taylor et al., 2008]. HOMME is also the rst dynamical core in the CAM which locally conserves energy in addition to mass and two-dimensional potential vorticity [Taylor, 2010]. HOMME represents a large change in the horizontal grid as compared to the other dynamical cores in CAM. Almost all other aspects of HOMME are based on a combination of well-tested approaches from the Eulerian and FV dynamical cores. For tracer advection, HOMME is modeled as closely as possible on the FV core. It uses the same conservation form of the transport equation and the same vertically Lagrangian discretization [Lin, 2004]. The HOMME dynamics are modeled as closely as possible on Eulerian core. They share the same vertical coordinate, vertical discretization, hyper-viscosity based horizontal diusion, top-of-model dissipation, and solve the same moist hydrostatic equations. The main dierences are that HOMME advects the surface pressure instead of its logarithm (in order to conserve mass and energy), and HOMME uses the vector-invariant form of the momentum equation instead of the vorticity-divergence formulation. Several dry dynamical cores including HOMME are evaluated in Lauritzen et al. [2010] using a grid-rotated version of the baroclinic instability test case [Jablonowski and Williamson, 2006]. The timestepping in HOMME is a form of dynamics/tracer/physics subcycling, achieved through the use of multi-stage 2nd order accurate Runge-Kutta methods. The tracers and dynamics use the same timestep which is controlled by the maximum anticipated wind speed, but the dynamics uses more stages than the tracers in order to maintain stability in the presence of gravity waves. The forcing is applied using a time-split approach. The optimal forcing strategy in HOMME has not yet been determined, so HOMME supports several options. The rst option is modeled after the FV dynamical core and the forcing is applied as an adjustment at each physics timestep. The second option is to convert all forcings into tendencies which are applied at the end of each dynamics/tracer timestep. If the physics timestep is larger than the tracer timestep, then the tendencies are held xed and only updated at each physics timestep. Finally, a hybrid approach can be used where the tracer tendencies are applied as in the rst option and the dynamics tendencies are applied as in the second option.

33

1063

3.2.1

Continuum Formulation of the Equations

1064 1065 1066 1067

HOMME uses a conventional vector-invariant form of the moist primitive equations. For the vertical discretization it uses the hybrid pressure vertical coordinate system modeled after 3.3.1 The formulation here diers only in that surface pressure is used as a prognostic variable as opposed to its logarithm. In the -coordinate system, the pressure is given by p() = A()p0 + B()ps . The hydrostatic approximation p/z = g is used to replace the mass density by an coordinate pseudo-density p/. The material derivative in -coordinates can be written (e.g. Satoh [2004], Sec.3.3), X X DX = + u X + Dt t

1068 1069 1070

where the () operator (as well as () and () below) is the two-dimensional gradient on constant -surfaces, / is the vertical derivative, = D/Dt is a vertical ow velocity and u is the horizontal velocity component (tangent to constant z-surfaces, not -surfaces). The -coordinate atmospheric primitive equations, neglecting dissipation and forcing terms can then be written as u + ( + f ) ku + t 1 2 u RTv u + + + p = 0 2 p T RTv T + u T + =0 t cp p p p p + =0 u + t p p p q = 0. q + qu + t (3.83) (3.84) (3.85) (3.86)

1071 1072 1073 1074 1075 1076 1077 1078

p p These are prognostic equations for u, the temperature T , density , and q where q is the specic humidity. The prognostic variables are functions of time t, vertical coordinate and two coordinates describing the surface of the sphere. The unit vector normal to the surface of the sphere is denoted by k. This formulation has already incorporated the hydrostatic equation and the ideal gas law, p = RTv . There is a no-ux ( = 0) boundary condition at = 1 and u, f is a Coriolis term and = Dp/Dt is the = top . The vorticity is denoted by = k pressure vertical velocity. The virtual temperature Tv and variable-of-convenience c are dened p as in 3.3.1. The diagnostic equations for the geopotential height eld is 1

= s +

RTv p d p

(3.87)

where s is the prescribed surface geopotential height (given at = 1). To complete the system, we need diagnostic equations for and , which come from integrating (3.85) with respect to

34

. In fact, (3.85) can be replaced by a diagnostic equation for and a prognostic equation p for surface pressure ps ps + t
1 top

p u
top

d = 0 p u d ,

(3.88) (3.89)

p p = t

where (3.88) is (3.89) evaluated at the model bottom ( = 1) after using that p/t = B()ps /t and (1) = 0, B(1) = 1. Using Eq 3.89, we can derive a diagnostic equation for the pressure vertical velocity = Dp/Dt, = p p + u p + = u p t
top

p u

Finally, we rewrite (3.89) as p = B()

1 top

p u

top

p u

d ,

(3.90)

1079

3.2.2

Conserved Quantities

The equations have innitely many conserved quantities, including mass, tracer mass, potential temperature dened by p MX = X ddA with (X = 1, q or (p/p0 ) T ) and the total moist energy E dened by E=
1080 1081 1082 1083 1084 1085 1086 1087

1 2 u + c T p 2

ddA +

ps s dA

(3.91)

where dA is the spherical area measure. To compute these quantities in their traditional units they should be divided by the constant of gravity g. We have omitted this scaling since g has also been scaled out from (3.83)(3.86). We note that in this formulation of the primitive equations, the pressure p is a moist pressure, representing the eects of both dry air and water vapor. The unforced equations conserve both the moist air mass (X = 1 above) and the dry air mass (X = 1 q ). However, in the presence of a forcing term in (3.86) (representing sources and sinks of water vapor as would be present in a full model) a corresponding forcing term must be added to (3.85) to ensure that dry air mass is conserved. The energy (3.91) is specic to the hydrostatic equations. We have omitted terms from the physical total energy which are constant under the evolution of the unforced hydrostatic equations [Staniforth et al., 2003]. It can be converted into a more universal form involving 1 2 u + c T + , with c dened similarly to c , so that c = cv + (cvv cv )q where cv and cvv v v p v 2 are the specic heats of dry air and water vapor dened at constant volume. We note that cp = R + cv and cpv = Rv + cvv so that c T = c T + RTv . Expanding c T with this expression, p v p 35

integrating by parts with respect to and making use of the fact that the model top is at a constant pressure p RTv d = and thus E=
1088 1089 1090

d =

p d (p) ddA +

=1 =top

1 2 u + c T + v 2

ptop (top ) dA.

(3.92)

The model top boundary term in (3.92) vanishes if ptop = 0. Otherwise it must be included to be consistent with the hydrostatic equations. It is present due to the fact that the hydrostatic momentum equation (3.83) neglects the vertical pressure gradient.

1091

3.2.3

Horizontal Discretization: Functional Spaces

1092 1093 1094 1095 1096 1097

In the nite element method, instead of constructing discrete approximations to derivative operators, one constructs a discrete functional space, and then nds the function in this space which solves the equations of interest in a minimum residual sense. As compared to nite volume methods, there is less choice in how one constructs the discrete derivative operators in this setting, since functions in the discrete space are represented in terms of known basis functions whose derivatives are known, often analytically. Let x and x = x1 e1 + x2 e2 be the Cartesian coordinates and position vector of a point in the reference square [1, 1]2 and let r and r be the coordinates and position vector of a point on the surface of the sphere, denoted by . We mesh using the cubed-sphere grid (Fig. 3.2) rst used in Sadourny [1972]. Each cube face is mapped to the surface of the sphere with the equal-angle gnomonic projection [Rani et al., 1996]. The map from the reference element [1, 1]2 to the cc cube face is a translation and scaling. The composition of these two maps denes a C1 map from the spherical elements to the reference element [1, 1]2 . We denote this map and its inverse by r = r(x; m), x = x(r; m). (3.93)

Figure 3.2: Tiling the surface of the sphere with quadrilaterals. An inscribed cube is projected to the surface of the sphere. The faces of the cubed sphere are further subdivided to form a quadrilateral grid of the desired resolution. Coordinate lines from the gnomonic equal-angle projection are shown.

1098

36

We now dene the discrete space used by the SEM. First we denote the space of polynomials up to degree d in [1, 1]2 by Pd = span(x1 )i (x2 )j = span (x),
i,j=0 I d

where I = {0, . . . , d}2 contains all the degrees and (x) = i1 (x1 )i2 (x2 ), i = 0, . . . , d, are the cardinal functions, namely polynomials that interpolate the tensor-product of degree-d GaussLobatto-Legendre (GLL) nodes = i1 e1 + i2 e2 . The GLL nodes used within an element for d = 3 are shown in Fig. 3.3. The cardinal-function expansion coecients of a function g are its GLL nodal values, so we have g(x) = g( ) (x). (3.94)
I

We can now dene the piecewise-polynomial SEM spaces V0 and V1 as V0 = {f L2 () : f (r(; m)) Pd , m} = span{ (x(; m))}I
m=1 M

(3.95)

and

V1 = C0 () V0 .

Functions in V0 are polynomial within each element but may be discontinuous at element boundaries and V1 is the subspace of continuous function in V0 . We take Md = dim V0 = (d + 1)3 M, and L = dim V1 < Md . We then construct a set of L unique points by
M

{r }L =1
1099 1100

=
m=1

r({ }I ; m),

(3.96)

For every point r , there exists at least one element m and at least one GLL node = x(r ; m). In 2D, if r belongs to exactly one m it is an element-interior node. If it belongs to exactly two m s, it is an element-edge interior node. Otherwise it is a vertex node. Figure 3.3: A 4 4 tensor product grid of GLL nodes used within each element, for a degree d = 3 discretization. Nodes on the boundary are shared by neighboring elements.

1101

We also dene similar spaces for 2D vectors. We introduce two families of spaces, with a subscript of either con or cov, denoting if the contravariant or covariant components of the vectors are piecewise polynomial, respectively. and V0 = {u L2 ()2 : u V0 , = 1, 2} con V1 = C0 ()2 V0 , con con

where u1 , u2 are the contravariant components of u dened below. Vectors in V1 are globally con continuous and their contravariant components are polynomials in each element. Similarly, and V0 = {u L2 ()2 : u V0 , = 1, 2} cov V1 = C0 ()2 V0 . cov cov 37

The SEM is a Galerkin method with respect to the V1 subspace and it can be formulated solely in terms of functions in V1 . In CAM-HOMME, the typical conguration is to run with d = 3 which achieves a 4th order accurate horizontal discretization [Taylor and Fournier, 2010]. All variables in the CAM-HOMME initial condition and history les as well as variables passed to the physics routines are represented by their grid point values at the points {r }L . However, =1 for some intermediate quantities and internally in the dynamical core it is useful to consider the larger V0 space, where variables are represented by their grid point values at the Md mapped GLL nodes. This later representation can also be considered as the cardinal-function (3.94) expansion of a function f local to each element, f (r) =
I
1102 1103 1104

f (r( ; m)) (x(r; m))

(3.97)

since the expansion coecients are the function values at the mapped GLL nodes. Functions f in V0 can be multiple-valued at GLL nodes that are redundant (i.e., shared by more than one element), while for f V1 , the values at any redundant points must all be the same.

1105

3.2.4

Horizontal Discretization: Dierential Operators

We use the standard curvilinear coordinate formulas for vector operators following Heinbockel [2001]. Given the 2 2 Jacobian of the the mapping (3.93) from [1, 1]2 to m , we denote its determinant-magnitude by r . (3.98) J= x A vector v may be written in terms of physical or covariant or contravariant components, v[ ] or v or v , 3 3 3 r ] v= v[ = v g = vg , (3.99) r =1 =1 =1
1106 1107

that are related by v = v g and v = v g , where g = x is a contravariant basis vector r and g = x is a covariant basis vector. The dot product and contravariant components of the cross product are Heinbockel [2001, Table 1] 3 3 1 u v (3.100) uv = u v and (uv) = J ,=1 =1 where {0, 1} is the Levi-Civita symbol. The divergence, covariant coordinates of the gradient and contravariant coordinates of the curl are Heinbockel [2001, eqs. 2.1.1, 2.1.4 & 2.1.6] v = 1 J (Jv ), x (f ) = f x and (v) = 1 J
,

v . x

(3.101)

1108 1109 1110

In the SEM, these operators are all computed in terms of the derivatives with respect to x in the reference element, computed exactly (to machine precision) by dierentiating the local element expansion (3.97). For the gradient, the covariant coordinates of f, f V0 are thus computed 38

1111 1112 1113 1114 1115

exactly within each element. Note that f V0 , but may not be in V1 even for f V1 cov cov due to the fact that its components will be multi-valued at element boundaries because f computed in adjacent elements will not necessarily agree along their shared boundary. In the case where J is constant within each element, the SEM curl of v V0 and the divergence of cov u V0 will also be exact, but as with the gradient, multiple-valued at element boundaries. con For non-constant J, these operators may not be computed exactly by the SEM due to the Jacobian factors in the operators and the Jacobian factors that appear when converting between covariant and contravariant coordinates. We follow Thomas and Loft [2000] and evaluate these operators in the form shown in (3.101). The quadratic terms that appear are rst projected into V0 via interpolation at the GLL nodes and then this interpolant is dierentiated exactly using (3.97). For example, to compute the divergence of v V0 , we rst compute the interpolant con I(Jv ) V0 of Jv , where the GLL interpolant of a product f g derives simply from the product of the GLL nodal values of f and g. This operation is just a reinterpretation of the nodal values and is essentially free in the SEM. The derivatives of this interpolant are then computed exactly from (3.97). The sum of partial derivatives are then divided by J at the GLL nodal values and thus the SEM divergence operator h () is given by v h v = I 1 J I(Jv ) x V0 . (3.102)

Similarly, the gradient and curl are approximated by (f ) (h f ) = and (v) (h v) = f x

,

(3.103) 1 v J x (3.104)

1116 1117 1118 1119

with h f V0 and h v V0 . The SEM is well known for being quite ecient in computing cov con these types of operations. The SEM divergence, gradient and curl can all be evaluated at the (d + 1)3 GLL nodes within each element in O(d) operations per node using the tensor-product property of these points [Deville et al., 2002; Karniadakis and Sherwin, 2005].

1120

3.2.5

Horizontal Discretization: Discrete Inner-Product

Instead of using exact integration of the basis functions as in a traditional nite-element method, the SEM uses a GLL quadrature approximation for the integral over , that we denote by . We can write this integral as a sum of area-weighted integrals over the set of elements {m }M m=1 used to decompose the domain,
M

f g dA =
m=1 m

f g dA.

The integral over a single element m is written as an integral over [1, 1]2 by f g dA =
m [1,1]2

f (r(; m))g(r(; m))Jm dx1 dx2 f g 39

where we approximate the integral over [1, 1]2 by GLL quadrature, fg

m

=
I

wi1 wi2 Jm ( )f (r( ; m))g(r( ; m))

(3.105)

The SEM approximation to the global integral is then naturally dened as

M

f g dA
1121 1122

fg
m=1

= fg

(3.106)

When applied to the product of functions f, g V0 , the quadrature approximation f g denes a discrete inner-product in the usual manner.

1123

3.2.6

Horizontal Discretization: The Projection Operators

1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135

Let P : V0 V1 be the unique orthogonal (self-adjoint) projection operator from V0 onto V1 w.r.t. the SEM discrete inner product (3.106). The operation P is essentially the same as the common procedure in the SEM described as assembly [Karniadakis and Sherwin, 2005, p. 7], or direct stiness summation [Deville et al., 2002, eq. 4.5.8]. Thus the SEM assembly procedure is not an ad-hoc way to remove the redundant degrees of freedom in V0 , but is in fact the natural projection operator P . Applying the projection operator in a nite element method requires inverting the nite element mass matrix. A remarkable fact about the SEM is that with the GLL based discrete inner product and the careful choice of global basis functions, the mass matrix is diagonal [Maday and Patera, 1987]. The resulting projection operator then has a very simple form: at element interior points, it leaves the nodal values unchanged, while at element boundary points shared by multiple elements it is a Jacobian-weighted average over all redundant values [Taylor and Fournier, 2010]. To apply the projection P : V0 V1 to vectors u, one cannot project the covariant cov cov components since the corresponding basis vectors g and g do not necessarily agree along element faces. Instead we must dene the projection as acting on the components using a globally continuous basis such as the latitude-longitude unit vectors and , P (u) = P (u ) + P (u ).

1136

3.2.7

Horizontal Discretization: Galerkin Formulation

The SEM solves a Galerkin formulation of the equations of interest. Given the discrete dierential operators described above, the primitive equations can be written as an ODE for a generic prognostic variable U and right-hand-side (RHS) terms U = RHS. t The SEM solves this equation in integral form with respect to the SEM inner product. That is, for a RHS V0 , the SEM nds the unique U V1 such that t U t = RHS 40 V1 .

As the prognostic variable is assumed to belong to V1 , the RHS will in general belong to V0 since it contains derivatives of the prognostic variables, resulting in the loss of continuity at the element boundaries. If one picks a suitable basis for V1 , this discrete integral equation results in a system of L equations for the L expansion coecients of U . The SEM solves these equations t exactly, and the solution can be written in terms of the SEM projection operator as U = P (RHS) . t
1137 1138

The projection operator commutes with any time-stepping scheme, so the equations can be solved in a two step process, illustrated here for simplicity with the forward Euler method Step 1: U = U t + t RHS Step 2: U t+1 = P (U ) U t+1 V1 U V0

For compactness of notation, we will denote this two step procedure in what follows by P 1
1139 1140 1141 1142 1143 1144 1145 1146 1147

U = RHS. t

Note that P maps a Md dimensional space V0 into a L dimensional space V1 , so here P 1 denotes the left inverse of P . This inverse will never be computed, it is only applied as in step 2 above. This two step Galerkin solution process represents a natural separation between computation and communication for the implementation of the SEM on a parallel computer. The computations in step 1 are all local to the data contained in a single element. Assuming an element-based decomposition so that each processor contains at least one element, no inter-processor communication is required in step 1. All inter-processor communication in HOMME is isolated to the projection operator step, in which element boundary data must be exchanged between adjacent elements.

1148

3.2.8

Vertical Discretization

The vertical coordinate system uses a Lorenz staggering of the variables as shown in 3.4. Let K be the total number of layers, with variables u, T, q, , at layer mid points denoted by k = 1, 2, . . . , K. We denote layer interfaces by k + 1 , k = 0, 1, . . . , K, so that 1/2 = top and 2 K+1/2 = 1. The -integrals will be replaced by sums. We will use to denote the discrete / operator. The operator uses centered dierences to compute derivatives with respect to at layer mid point from layer interface values, (X)k = (Xk+1/2 Xk1/2 )/(k+1/2 k1/2 ). We will use the over-bar notation for vertical averaging, q k+1/2 = (qk+1 + qk )/2. We also introduce p the symbol to denote the discrete pseudo-density given by k = (p)k
1149

. 41

We will use to denote the discrete form of the / operator. We use the discretization given in 3.3.5. This operator acts on quantities dened at layer mid-points and returns a result also at layer mid-points, (X)k = 1 ()k+1/2 (Xk+1 Xk ) + ()k1/2 (Xk Xk1 ) 2k k (3.107)

where k = k+1/2 k1/2 . We use the over-bar notation since the formula can be seen as a -weighted average of a layer interface centered dierence approximation to /. This formulation was constructed in Simmons and Burridge [1981] in order to ensure mass and energy conservation. Here we will use an equivalent expression that can be written in terms of , (X)k = 1 X k
k

X ()k .

(3.108)

1150

3.2.9

Discrete formulation: Dynamics

We discretize the equations exactly in the form shown in (3.83), (3.84), (3.88) and (3.90), obtaining 1 2 RTv u = ( + f ) ku + h u + (u) h (p) t 2 p T RTv P 1 = u h (T ) (T ) + t cp p P 1 P
1 ps K

(3.109) (3.110) (3.111)

j=1

h (u)j j
K i

()i+1/2 = B(i+1/2 )
j=1
1151 1152 1153 1154 1155

h (u)j j

j=1

h (u)j j .

(3.112)

We consider () a single quantity given at layer interfaces and dened by (3.112). The no-ux boundary condition is ()1/2 = ()K+1/2 = 0. In (3.112), we used a midpoint quadrature rule to evaluate the indenite integral from (3.90). In practice can be eliminated from the discrete equations by scaling , but here we retain them so as to have a direct correspondence p with the continuum form of the equations written in terms of . Finally we give the approximations for the diagnostic equations. We rst integrate to layer interface i 1 using the same mid-point rule as used to derive (3.112), and then add an additional 2 1 term representing the integral from i 2 to i:
i1

i = (u h p)i = (u h p)i

j=1 K

h (u)j j + h (u)i Cij h (u)j

i 2

(3.113)

(3.114)

j=1

42

where

and similar for , ( s )i = =

j Cij = j /2 0 RTv p
K

i>j i=j i<j

RTv i + 2 p j=i+1

j
j

(3.115)

Hij
j=1

RTv p

(3.116)
j

where

Similar to 3.3.5, we note that

1156

j Hij = j /2 0

i<j i=j i>j (3.117)

i Cij = j Hji

which ensures energy conservation [Taylor, 2010].

1157

3.2.10

Consistency

It is important that the discrete equations be as consistent as possible. In particular, we need a discrete version of (3.85), the non-vertically averaged continuity equation. Equation (3.112) implicitly implies such an equation. To see this, apply to (3.112) and using that p/t = B()ps /t then we can derive, at layer mid-points, P 1 = h (u) () . t (3.118)

1158 1159

A second type of consistency that has been identied as important is that (3.113), the discrete equation for , be consistent with (3.112), the discrete continuity equation [Williamson and Olson, 1994b]. The two discrete equations should imply a reasonable discretization of = Dp/Dt. To show this, we take the average of (3.112) at layers i1/2 and i+1/2 and combine this with (3.113) (at layer mid-points i) and assuming that B(i ) = B(i1/2 )+B(i+1/2 ) we obtain p 1 P 1 ( )i1/2 + ( )i+1/2 . = i (u h p)i t 2 which, since u h p is given at layer mid-points and at layer interfaces, is the SEM discretiza tion of w = p/t + u h p + .

43

1160

3.2.11

Time Stepping

Applying the SEM discretization to (3.109)-(3.112) results in a system of ODEs. These are solved with an N-stage Runge-Kutta method. This method allows for a gravity-wave based CFL number close to N 1, (normalized so that the largest stable timestep of the Robert ltered Leapfrog method has a CFL number of 1.0). The value of N is chosen large enough so that the dynamics will be stable at the same timestep used by the tracer advection scheme. To determine N, we rst note that the tracer advection scheme uses a less ecient (in terms of maximum CFL) strong stability preserving Runge-Kutta method described below. It is stable at an advective CFL number of 1.4. Let u0 be a maximum wind speed and c0 be the maximum gravity wave speed. The gravity wave and advective CFL conditions are t (N 1)x/c0 ,
1161 1162 1163 1164 1165

t 1.4x/u0 .

In the case where t is chosen as the largest stable timestep for advection, then we require N 1 + 1.4c0 /u0 for a stable dynamics timestep. Using a typical values u0 = 120 m/s and c0 = 340m/s gives N = 5. CAM places additional restrictions on the timestep (such as that the physics timestep must be an integer multiple of t) which also inuence the choice of t and N.

1166

3.2.12

Dissipation

A horizontal hyper-viscosity operator, modeled after 3.3.6 is applied to the momentum and temperature equations. It is applied in a time-split manor after each dynamics timestep. The hyper-viscosity step for vectors can be written as u = 2 u. t An integral form of this equation suitable for the SEM is obtained using a mixed nite element formulation (following Giraldo [1999]) which writes the equation as a system of equations involving only rst derivatives. We start by introduced an auxiliary vector f and using the identity u = ( u) (u), u = ( f ) k(f ) t f = ( u) (u)k. Integrating the gradient and curl operators by parts gives u dA = t ( )( f ) + () k(f ) dA ( )( u) + () k(u) dA. (3.121) (3.122) (3.123) The SEM Galerkin solution of this integral equation is most naturally written in terms of an inverse mass matrix instead of the projection operator. It can be written in terms of the SEM 44 (3.119) (3.120)

f dA =

projection operator by rst testing with the product of the element cardinal functions and the contravariant basis vector = g . With this type of test function, the RHS of (3.122) can be dened as a weak Laplacian operator f = D(u) V0 . The covariant components of f given cov by f = f g are then f (r( ; m)) = 1 (h g )(h u) + (h g ) k(h u). wi1 wi2 Jm ( )

Then the SEM solution to (3.121) and (3.122) is given by u(t + t) = u(t) tP D P D(u)
1167 1168 1169 1170 1171 1172 1173

Because of the SEM tensor product decomposition, the expression for D can be evaluated in only O(d) operations per grid point, and in CAM-HOMME typically d = 3. Following 3.3.6, a correction term is added so the hyper-viscosity does not damp rigid rotation. The hyper-viscosity formulation used for scalars such as T is much simpler, since instead of the vector Laplacian identity we use T = T . Otherwise the approach is identical to that used above so we omit the details. The correction for terrain following coordinates given in 3.3.6 is not yet implemented in CAM-HOMME.

1174

3.2.13

Discrete formulation: Tracer Advection

1175 1176 1177 1178 1179 1180

All tracers, including specic humidity, are advected with a discretized version of (3.86). HOMME uses the vertically Lagrangian approach (see 3.1.4) from Lin [2004]. At the beginning of each timestep, the tracers are assumed to be given on the -coordinate layer mid points. The tracers are advanced in time on a moving vertical coordinate system dened so that = 0. At the end of the timestep, the tracers are remapped back to the -coordinate layer mid points using the monotone remap algorithm from Zerroukat et al. [2005]. The horizontal advection step consists of using the SEM to solve (q) = h (u)q t (3.124)

1181 1182 1183 1184 1185 1186

on the surfaces dened by the layer mid points. The quantity (u) is the mean ux computed during the dynamics update. The mean ux used in (3.124), combined with a suitable mean vertical ux used in the remap stage allows HOMME to preserve mass/tracer-mass consistency: The tracer advection of q with q = 1 will be identical to the advection of implied from (3.118). The mass/tracer-mass consistency capability is not in the version of HOMME included in CAM 4.0, but should be in all later versions. The equation is discretized in time using the optimal 3 stage strong stability preserving (SSP) second order Runge-Kutta method from Spiteri and Ruuth [2002]. The RK-SSP method is chosen because it will preserve the monotonicity properties of the horizontal discretization. RK-SSP methods are convex combinations of forward-Euler timesteps, so each stage s of the RK-SSP timestep looks like (q)s+1 = (q)s th (u)q s 45 (3.125)

1187 1188 1189 1190 1191 1192

Simply discretizing this equation with the SEM will result in locally conservative, high-order accurate but oscillatory transport scheme. A limiter is added to reduce or eliminate these oscillations [Taylor et al., 2009]. HOMME supports both monotone and sign-preserving limiters, but the most eective limiter for HOMME has not yet been determined. The default conguration in CAM4 is to use the sign-preserving limiter to prevent negative values of q coupled with a sign-preserving hyper-viscosity operator which dissipates q 2 .

1193

3.2.14

Conservation and Compatibility

The SEM is compatible, meaning it has a discrete version of the divergence theorem, Stokes theorem and curl/gradient annihilator properties Taylor and Fournier [2010]. The divergence theorem is the key property of the horizontal discretization that is needed to show conservation. For an arbitrary scalar h and vector u at layer mid-points, the divergence theorem (or the divergence/gradient adjoint relation) can be written h u dA + uh dA = 0.

The discrete version obeyed by the SEM discretization, using (3.106), is given by hh u + u h h = 0.
1194 1195 1196 1197

(3.126)

The discrete divergence and Stokes theorem apply locally at the element with the addition of an element boundary integral. The local form is used to show local conservation of mass and that the horizontal advection operator locally conserves the two-dimensional potential vorticity [Taylor and Fournier, 2010]. In the vertical, Simmons and Burridge [1981] showed that the and operators needed to satisfy two integral identities to ensure conservation. For any layer interface velocity which satises 1/2 = K+1/2 = 0 and f, g arbitrary functions of layer mid points. The rst identity is the adjoint property (compatibility) for and ,
K K

i i (f ) +
i=1 i=1

i fi () = 0

(3.127)

which follows directly from the denition of the dierence operator given in (3.108). The second identity we write in terms of ,
K K K

i f g () =
i=1
1198 1199 1200

i f (g) +
i=1 i=1

i g (f )

(3.128)

which is a discrete integrated-by-parts analog of (f g) = f g + gf. Construction of methods with both properties on a staggered unequally spaced grid is the reason behind the complex denition for in (3.108). The energy conservation properties of CAM-HOMME were studied in Taylor [2010] using the aqua planet test case [Neale and Hoskins, 2001a,b]. CAM-HOMME uses
K

E=
i=1

i i

1 2 u + c T p 2 46

+ ps s
i

1201 1202 1203 1204 1205 1206 1207 1208 1209 1210

as the discretization of the total moist energy (3.91). The conservation of E is semi-discrete, meaning that the only error in conservation is the time truncation error. In the adiabatic case (with no hyper-viscosity and no limiters), running from a fully spun up initial condition, the error in conservation decreases to machine precision at a second-order rate with decreasing timestep. In the full non-adiabatic case with a realistic timestep, dE/dt 0.013W/m2 . The CAM physics conserve a dry energy Edry from Boville and Bretherton [2003a] which is not conserved by the moist primitive equations. Although E Edry is small, adiabatic processes in the primitive equations result in a net heating dEdry /dt 0.5W/m2 [Taylor, 2010]. If it is desired that the dynamical core conserve Edry instead of E, HOMME uses the energy xer from 3.3.20.

1211

3.3

Eulerian Dynamical Core

1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224

The hybrid vertical coordinate that has been implemented in CAM 5.0 is described in this u section. The hybrid coordinate was developed by Simmons and Str ng [1981] in order to provide a general framework for a vertical coordinate which is terrain following at the Earths surface, but reduces to a pressure coordinate at some point above the surface. The hybrid coordinate is more general in concept than the modied scheme of Sangster [1960], which is used in the GFDL SKYHI model. However, the hybrid coordinate is normally specied in such a way that the two coordinates are identical. The following description uses the same general development as Simmons and Str ng [1981], u who based their development on the generalized vertical coordinate of Kasahara [1974]. A specic form of the coordinate (the hybrid coordinate) is introduced at the latest possible point. u The description here diers from Simmons and Str ng [1981] in allowing for an upper boundary at nite height (nonzero pressure), as in the original development by Kasahara. Such an upper boundary may be required when the equations are solved using vertical nite dierences.

1225

3.3.1

Generalized terrain-following vertical coordinates

1226 1227 1228 1229

Deriving the primitive equations in a generalized terrain-following vertical coordinate requires only that certain basic properties of the coordinate be specied. If the surface pressure is , then we require the generalized coordinate (p, ) to satisfy: 1. (p, ) is a monotonic function of p. 2. (, ) = 1 3. (0, ) = 0 4. (pt , ) = t where pt is the top of the model. The latter requirement provides that the top of the model will be a pressure surface, simplifying the specication of boundary conditions. In the case that pt = 0, the last two requirements are identical and the system reduces to that described in Simmons and Str ng [1981]. The u boundary conditions that are required to close the system are: (, ) = 0, (pt , ) = (pt ) = 0. 47 (3.129) (3.130)

1230

1231

1232

1233 1234 1235 1236 1237

Given the above description of the coordinate, the continuous system of equations can be written following Kasahara [1974] and Simmons and Str ng [1981]. The prognostic equations u are: = k (n/ cos ) + FH , (3.131) t = (n/ cos ) 2 (E + ) + FH , (3.132) t T 1 T R = (UT ) + cos (V T ) + T + Tv 2 t a cos cp p +Q + FTH + FFH , (3.133) 1 q q = (Uq) + cos (V q) + q + S, (3.134) t a cos2 t p = V d. (3.135) t 1 The notation follows standard conventions, and the following terms have been introduced with n = (nU , nV ): U Tv 1 p R + FU , (3.136) nU = +( + f )V p a V Tv cos p nV = ( + f )U R + FV , (3.137) p a U2 + V 2 , (3.138) E = 2 cos2 (U, V ) = (u, v) cos , (3.139) Rv Tv = 1 + 1 q T , (3.140) R cp v 1 q cp . (3.141) c = 1 + p cp The terms FU , FV , Q, and S represent the sources and sinks from the parameterizations for momentum (in terms of U and V ), temperature, and moisture, respectively. The terms FH and FH represent sources due to horizontal diusion of momentum, while FTH and FFH represent sources attributable to horizontal diusion of temperature and a contribution from frictional heating (see sections on horizontal diusion and horizontal diusion correction). In addition to the prognostic equations, three diagnostic equations are required:
p(1)

= s + R p p = t
p()

Tv d ln p,

(3.142) (3.143) (3.144)

p V d, t p V d. = V p t

1238 1239

Note that the bounds on the vertical integrals are specied as values of (e.g. t , 1) or as functions of p (e.g. p (1), which is the pressure at = 1). 48

1240

3.3.2

Conversion to nal form

1241 1242 1243 1244 1245 1246 1247

Equations (3.129)-(3.144) are the complete set which must be solved by a GCM. However, in order to solve them, the function (p, ) must be specied. In advance of actually specifying (p, ), the equations will be cast in a more convenient form. Most of the changes to the equations involve simple applications of the chain rule for derivatives, in order to obtain terms that will be easy to evaluate using the predicted variables in the model. For example, terms involving horizontal derivatives of p must be converted to terms involving only p/ and horizontal derivatives of . The former can be evaluated once the function (p, ) is specied. The vertical advection terms in (3.133), (3.134), (3.136), and (3.137) may be rewritten as: p = , p (3.145)

1248

since p/ is given by (3.143). Similarly, the rst term on the right-hand side of (3.143) can be expanded as p p = , (3.146) t t and (3.135) invoked to specify /t. The integrals which appear in (3.135), (3.143), and (3.144) can be written more conveniently by expanding the kernel as p V =V p + p V . (3.147)

The second term in (3.147) is easily treated in vertical integrals, since it reduces to an integral in pressure. The rst term is expanded to: V p (p) p =V p p + V =V =V

(3.148)

1249 1250

The second term in (3.148) vanishes because / = 0, while the rst term is easily treated once (p, ) is specied. Substituting (3.148) into (3.147), one obtains: p V = p V + p V . (3.149)

1251 1252

Using (3.149) as the kernel of the integral in (3.135), (3.143), and (3.144), one obtains integrals of the form p V d = = p V + p + p V d (3.150)

V d 49

dp.

The original primitive equations (3.131)-(3.135), together with (3.136), (3.137), and (3.142)(3.144) can now be rewritten with the aid of (3.145), (3.146), and (3.150). = k (n/ cos ) + FH , t = (n/ cos ) 2 (E + ) + FH , t T 1 p T R = (UT ) + cos (V T ) + T + Tv 2 t a cos p cp p +Q + FTH + FFH q 1 p q = (Uq) + cos (V q) + q + S, t a cos2 p = t nU nV
p(1) p dp, p(t ) (t ) p U Tv 1 p = +( + f )V R + FU , p a p p V Tv cos 1 p = ( + f )U R + FV , p a p (1)

(3.151) (3.152)

(3.153) (3.154) (3.155) (3.156) (3.157) (3.158)

V d

p(1)

= s + R
p()

Tv d ln p,
(1)

p p = =

(t ) () (t )

V d p

p(1)

+
p(t ) p()

dp dp,

(3.159)

V d
() (t )

p(t )

p V

V d

p()

dp.
p(t )

(3.160)

1253 1254 1255 1256 1257 1258 1259

Once (p, ) is specied, then p/ can be determined and (3.151)-(3.160) can be solved in a GCM. In the actual denition of the hybrid coordinate, it is not necessary to specify (p, ) explicitly, since (3.151)-(3.160) only requires that p and p/ be determined. It is sucient to specify p(, ) and to let be dened implicitly. This will be done in section 3.3.7. In the case that p(, ) = and t = 0, (3.151)-(3.160) can be reduced to the set of equations solved by CCM1.

1260

3.3.3

Continuous equations using ln()/t

1261 1262 1263 1264 1265

In practice, the solutions generated by solving the above equations are excessively noisy. This problem appears to arise from aliasing problems in the hydrostatic equation (3.158). The ln p integral introduces a high order nonlinearity which enters directly into the divergence equation (3.152). Large gravity waves are generated in the vicinity of steep orography, such as in the Pacic Ocean west of the Andes. 50

1266 1267 1268 1269 1270 1271

The noise problem is solved by converting the equations given above, which use as a prognostic variable, to equations using = ln(). This results in the hydrostatic equation becoming only quadratically nonlinear except for moisture contributions to virtual temperature. Since the spectral transform method will be used to solve the equations, gradients will be obtained during the transform from wave to grid space. Outside of the prognostic equation for , all terms involving will then appear as . Equations (3.151)-(3.160) become: = k (n/ cos ) + FH , t = (n/ cos ) 2 (E + ) + FH , t T 1 p T R = (UT ) + cos (V T ) + T + Tv 2 t a cos p cp p +Q + FTH + FFH , q 1 p q = (Uq) + cos (V q) + q + S, 2 t a cos p = t nU nV 1 p(1) p dp, p(t ) (t ) p U Tv p = +( + f )V R + FU , p a p p V Tv cos p = ( + f )U R + FV , p a p
(1)

(3.161) (3.162) (3.163)

(3.164) (3.165) (3.166) (3.167) (3.168)

V d

p(1)

= s + R
p()

Tv d ln p,
(1)

p p = =
1272 1273

(t ) () (t )

V d p

p(1)

+
p(t ) p()

dp dp,

(3.169)

V d
() (t )

p(t )

p V

V d

p()

dp.
p(t )

(3.170)

The above equations reduce to the standard equations used in CCM1 if = and t = 0. (Note that in this case p/ = p/ = .)

1274

3.3.4

Semi-implicit formulation

The model described by (3.161)-(3.170), without the horizontal diusion terms, together with boundary conditions (3.129) and (3.130), is integrated in time using the semi-implicit leapfrog scheme described below. The semi-implicit form of the time dierencing will be applied to (3.162) and (3.164) without the horizontal diusion sources, and to (3.165). In order to derive the semi-implicit form, one must linearize these equations about a reference state. Isolating

51

the terms that will have their linear parts treated implicitly, the prognostic equations (3.161), (3.162), and (3.165) may be rewritten as: = RTv 2 ln p 2 + X1 , t T R p T = + Tv + Y1 , t cp p p 1 = t
p(1)

(3.171) (3.172) (3.173)

dp + Z1 ,
p(t )

where X1 , Y1 , Z1 are the remaining nonlinear terms not explicitly written in (3.171)-(3.173). The terms involving and may be expanded into vertical integrals using (3.168) and (3.170), while the 2 ln p term can be converted to 2 , giving: p 2 = RT R2 t p RT T = t cp p 1 = t pi
1275 1276 1277 1278

p(1)

T d ln p + X2 ,
p() p(1) p(t ) p()

(3.174) T + Y2 , p (3.175) (3.176)

p() p(t ) p(1)

dp

dp

dp
p(t )

dp + Z2 .
p(t )

Once again, only terms that will be linearized have been explicitly represented in (3.174)-(3.176), and the remaining terms are included in X2 , Y2 , and Z2 . Anticipating the linearization, Tv and c have been replaced by T and cp in (3.174) and (3.175). Furthermore, the virtual temperature p corrections are included with the other nonlinear terms. In order to linearize (3.174)-(3.176), one species a reference state for temperature and pressure, then expands the equations about the reference state: T = T r + T , = r + , p = pr (, r ) + p . (3.177) (3.178) (3.179)

1279 1280 1281 1282

In the special case that p(, ) = , (3.174)-(3.176) can be converted into equations involving only = ln instead of p, and (3.178) and (3.179) are not required. This is a major dierence between the hybrid coordinate scheme being developed here and the coordinate scheme in CCM1. Expanding (3.174)-(3.176) about the reference state (3.177)-(3.179) and retaining only the linear terms explicitly, one obtains: r = R2 T r r t p R Tr T = t cp p r 1 = r t
pr () pr (t ) pr (1)

p
r

pr (1)

+
pr ()

T d ln p +
p () pr (1) pr (t ) r

p (1) r

Tr dp + X3 , pr dpr T r + Y3 , pr

(3.180) (3.181) (3.182)

dp

pr ()

dp

pr (t )

dpr + Z3 .
pr (t )

52

Figure 3.4: Vertical level structure of CAM 5.0 The semi-implicit time dierencing scheme treats the linear terms in (3.180)-(3.182) by averaging in time. The last integral in (3.180) is reduced to purely linear form by the relation dp = d
1283

+ x.

(3.183)

In the hybrid coordinate described below, p is a linear function of , so x above is zero. We will assume that centered dierences are to be used for the nonlinear terms, and the linear terms are to be treated implicitly by averaging the previous and next time steps. Finite dierences are used in the vertical, and are described in the following sections. At this stage only some very general properties of the nite dierence representation must be specied. A layering structure is assumed in which eld values are predicted on K layer midpoints denoted by an integer index, k (see Figure 3.4). The interface between k and k+1 is denoted by a half-integer index, k+1/2 . The model top is at 1/2 = t , and the Earths surface is at K+1/2 = 1. It is further assumed that vertical integrals may be written as a matrix (of order K) times a column vector representing the values of a eld at the k grid points in the vertical. The column vectors representing a vertical column of grid points will be denoted by underbars, the matrices will be denoted by bold-faced capital letters, and superscript T will denote the vector transpose. The nite dierence forms of (3.180)-(3.182) may then be written down as: n+1 = n1 + 2tX n n1 + n+1 n 2 n1 + (T )n+1 (T ) (T )n 2tRH r 2 2 n1 + n+1 n , 2tRhr 2 2 n1 + n+1 n1 n r = T + 2tY 2tD n , 2 2tRbr 2 =
n1

(3.184) (3.185) (3.186)

T n+1
1284 1285 1286 1287 1288

n+1

+ 2tZ 2t

n1 + n+1 n 2

1 pr , r

where ()n denotes a time varying value at time step n. The quantities X n , Y n , and Z n are dened so as to complete the right-hand sides of (3.171)-(3.173). The components of pr are given by pr = pr 1 pr 1 . This denition of the vertical dierence operator will be used in k k+ 2 k 2 subsequent equations. The reference matrices H r and D r , and the reference column vectors br and hr , depend on the precise specication of the vertical coordinate and will be dened later.

1289

3.3.5

Energy conservation

We shall impose a requirement on the vertical nite dierences of the model that they conserve the global integral of total energy in the absence of sources and sinks. We need to derive 53

equations for kinetic and internal energy in order to impose this constraint. The momentum equations (more painfully, the vorticity and divergence equations) without the FU , FV , FH and FH contributions, can be combined with the continuity equation t t p + p V + p =0 (3.187)

to give an equation for the rate of change of kinetic energy: p E = p p EV E p RTv p V p V . p

(3.188)

1290 1291 1292 1293 1294 1295 1296 1297

The rst two terms on the right-hand side of (3.188) are transport terms. The horizontal integral of the rst (horizontal) transport term should be zero, and it is relatively straightforward to construct horizontal nite dierence schemes that ensure this. For spectral models, the integral of the horizontal transport term will not vanish in general, but we shall ignore this problem. The vertical integral of the second (vertical) transport term on the right-hand side of (3.188) should vanish. Since this term is obtained from the vertical advection terms for momentum, which will be nite dierenced, we can construct a nite dierence operator that will ensure that the vertical integral vanishes. The vertical advection terms are the product of a vertical velocity (p/) and the vertical derivative of a eld (/p). The vertical velocity is dened in terms of vertical integrals of elds (3.170), which are naturally taken to interfaces. The vertical derivatives are also naturally taken to interfaces, so the product is formed there, and then adjacent interface values of the products are averaged to give a midpoint value. It is the denition of the average that must be correct in order to conserve kinetic energy under vertical advection in (3.188). The derivation will be omitted here, the resulting vertical advection terms are of the form: p p =
k

1 2pk

k+1/2

(k+1 k ) +

k1/2

(k k1 ) , (3.189) (3.190)

pk = pk+1/2 pk1/2 .
1298 1299 1300 1301 1302

The choice of denitions for the vertical velocity at interfaces is not crucial to the energy conservation (although not completely arbitrary), and we shall defer its denition until later. The vertical advection of temperature is not required to use (3.189) in order to conserve mass or energy. Other constraints can be imposed that result in dierent forms for temperature advection, but we will simply use (3.189) in the system described below. The last two terms in (3.188) contain the conversion between kinetic and internal (potential) energy and the form drag. Neglecting the transport terms, under assumption that global p integrals will be taken, noting that p/p = , and substituting for the geopotential using p (3.168), (3.188) can be written as: t p E = RTv p V p p
p(1)

(3.191) RTv d ln p + . . .
p()

p p V s V 54

The second term on the right-hand side of (3.192) is a source (form drag) term that can be neglected as we are only interested in internal conservation properties. The last term on the right-hand side of (3.192) can be rewritten as p V
p(1) p()

RTv d ln p =

p V

p(1) p()

RTv d ln p

p V

p(1)

RTv d ln p . (3.192)
p()

The global integral of the rst term on the right-hand side of (3.192) is obviously zero, so that (3.192) can now be written as: t p E = RTv p V p + p p V
p(1)

RTv d ln p + ...
p()

(3.193)

We now turn to the internal energy equation, obtained by combining the thermodynamic equation (3.164), without the Q, FTH , and FFH terms, and the continuity equation (3.187): t p c T p = p c TV p p p cp T + RTv . p (3.194)

As in (3.188), the rst two terms on the right-hand side are advection terms that can be neglected under global integrals. Using (3.144), (3.194) can be written as: t
1303 1304 1305 1306 1307 1308

p c T p

= RTv

p V

p p 1 RTv p p

p V

d + ...

(3.195)

The rate of change of total energy due to internal processes is obtained by adding (3.193) and (3.195) and must vanish. The rst terms on the right-hand side of (3.193) and (3.195) obviously cancel in the continuous form. When the equations are discretized in the vertical, the terms will still cancel, providing that the same denition is used for (1/p p/)k in the nonlinear terms of the vorticity and divergence equations (3.166) and (3.167), and in the term of (3.164) and (3.170). The second terms on the right-hand side of (3.193) and (3.195) must also cancel in the global mean. This cancellation is enforced locally in the horizontal on the column integrals of (3.193) and (3.195), so that we require:
1 t

p V

p(1)

RTv d ln p d =
p() t

RTv

p 1 p

p V

d d.

(3.196)

The inner integral on the left-hand side of (3.196) is derived from the hydrostatic equation (3.168), which we shall approximate as
K

k = s + R
=k K

Hk Tv , Hk Tv ,
=1

= s + R

(3.197) (3.198)

= s 1 + RHTv , 55

where Hk = 0 for < k. The quantity 1 is dened to be the unit vector. The inner integral on the right-hand side of (3.196) is derived from the vertical velocity equation (3.170), which we shall approximate as p =
k

p p

K k

V k

=1

Ck p + (V )

(3.199)

where Ck = 0 for > k, and Ck is included as an approximation to 1/pk for k and the symbol is similarly dened as in (3.190). Ck will be determined so that is consistent with the discrete continuity equation following Williamson and Olson [1994a]. Using (3.197) and (3.199), the nite dierence analog of (3.196) is
K

k=1

1 k pk + (V k ) k
K

R
k =1

Hk Tv p

k k ,

=
k=1

pk RTv k k

=1

Ck p + (V )

(3.200)

where we have used the relation V (p/)k = [k pk + (V k ) (p/)k ]/k (see 3.150). We can now combine the sums in (3.200) and simplify to give
K K

(3.201)

k=1 =1 K

k pk + (V k )
K

Hk Tv
k

=
k=1 =1

p + (V )

pk Ck Tv k .

(3.202)

Interchanging the indexes on the left-hand side of (3.202) will obviously result in identical expressions if we require that Hk = Ck p . (3.203)
1309 1310 1311 1312

Given the denitions of vertical integrals in (3.198) and (3.199) and of vertical advection in (3.189) and (3.190) the model will conserve energy as long as we require that C and H satisfy (3.203). We are, of course, still neglecting lack of conservation due to the truncation of the horizontal spherical harmonic expansions.

1313

3.3.6

Horizontal diusion

1314 1315 1316 1317 1318

CAM 5.0 contains a horizontal diusion term for T, , and to prevent spectral blocking and to provide reasonable kinetic energy spectra. The horizontal diusion operator in CAM 5.0 is also used to ensure that the CFL condition is not violated in the upper layers of the model. The horizontal diusion is a linear 2 form on surfaces in the top three levels of the model and a linear 4 form with a partial correction to pressure surfaces for temperature elsewhere. 56

1319 1320 1321 1322

The 2 diusion near the model top is used as a simple sponge to absorb vertically propagating planetary wave energy and also to control the strength of the stratospheric winter jets. The 2 diusion coecient has a vertical variation which has been tuned to give reasonable Northern and Southern Hemisphere polar night jets. In the top three model levels, the 2 form of the horizontal diusion is given by FH = K (2) 2 ( + f ) + 2 ( + f ) /a2 , FTH = K (2) 2 T. FH = K (2) 2 + 2(/a2 ) , (3.204) (3.205) (3.206)

1323 1324 1325 1326

Since these terms are linear, they are easily calculated in spectral space. The undierentiated correction term is added to the vorticity and divergence diusion operators to prevent damping of uniform (n = 1) rotations [Orszag, 1974; Bourke et al., 1977]. The 2 form of the horizontal diusion is applied only to pressure surfaces in the standard model conguration. The horizontal diusion operator is better applied to pressure surfaces than to terrainfollowing surfaces (applying the operator on isentropic surfaces would be still better). Although the governing system of equations derived above is designed to reduce to pressure surfaces above some level, problems can still occur from diusion along the lower surfaces. Partial correction to pressure surfaces of harmonic horizontal diusion (/t = K2 ) can be included using the relations: ln p p 2 = 2 p 2 ln p 2 p p p = p

p +

2 2 p. 2p

(3.207)

Retaining only the rst two terms above gives a correction to the surface diusion which involves only a vertical derivative and the Laplacian of log surface pressure, 2 = 2 p p 2 + ... p (3.208)

Similarly, biharmonic diusion can be partially corrected to pressure surfaces as: 4 = 4 p p 4 + ... p (3.209)

The bi-harmonic 4 form of the diusion operator is applied at all other levels (generally throughout the troposphere) as FH = K (4) 4 ( + f ) ( + f ) 2/a2 FH = K (4) 4 (2/a2 )2 , T p 4 . FTH = K (4) 4 T p
1327 1328

(3.210) (3.211) (3.212)

The second term in FTH consists of the leading term in the transformation of the 4 operator to pressure surfaces. It is included to oset partially a spurious diusion of T over mountains. 57

1329 1330 1331 1332

As with the 2 form, the 4 operator can be conveniently calculated in spectral space. The correction term is then completed after transformation of T and 4 back to gridpoint space. As with the 2 form, an undierentiated term is added to the vorticity and divergence diusion operators to prevent damping of uniform rotations.

1333

3.3.7

Finite dierence equations

1334 1335 1336 1337

The governing equations are solved using the spectral method in the horizontal, so that only the vertical and time dierences are presented here. The dynamics includes horizontal diusion of T, ( + f ), and . Only T has the leading term correction to pressure surfaces. Thus, equations that include the terms in this time split sub-step are of the form = Dyn () (1)i K (2i) 2i , t for ( + f ) and , and T p 2i T = Dyn (T ) (1)i K (2i) 2i T t p , (3.214) (3.213)

where i = 1 in the top few model levels and i = 2 elsewhere (generally within the troposphere). These equations are further subdivided into time split components: n+1 = n1 + 2t Dyn n+1 , n , n1 , = for ( + f ) and , and T n+1 = T n1 + 2t Dyn T n+1 , T n , T n1 T =T T n+1
1338 1339 1340 1341

(3.215) , (3.216) (3.217)

n+1

n+1

= ,

2t (1) K

(2i)

2i

n+1

(3.218) (3.219) (3.220)

(T ) , T p 2i , = T + 2t (1)i K (2i) p

n+1

2t (1) K

(2i)

2i

for T , where in the standard model i only takes the value 2 in (3.220). The rst step from ( )n1 to ( )n+1 includes the transformation to spectral coecients. The second step from ( )n+1 to ()n+1 for and , or ( )n+1 to ( ) for T , is done on the spectral coecients, and the nal step from ( ) to ()n+1 for T is done after the inverse transform to the grid point representation. The following nite-dierence description details only the forecast given by (3.215) and (3.218). The nite-dierence form of the forecast equation for water vapor will be presented later in Section 3c. The general structure of the complete nite dierence equations is determined by the semi-implicit time dierencing and the energy conservation properties described above. In order to complete the specication of the nite dierencing, we require a denition of the vertical coordinate. The actual specication of the generalized vertical coordinate takes advantage of the structure of the equations (3.161)-(3.170). The equations can be nite-dierenced in the vertical and, in time, without having to know the value of anywhere. The quantities that must 58

be known are p and p/ at the grid points. Therefore the coordinate is dened implicitly through the relation: p(, ) = A()p0 + B() , (3.221) p = B() . (3.222) A set of levels k may be specied by specifying Ak and Bk , such that k Ak + Bk , and dierence forms of (3.161)-(3.170) may be derived. The nite dierence forms of the Dyn operator (3.161)-(3.170), including semi-implicit time integration are: n+1 = n1 + 2tk (nn / cos ) , 2tRH r 2 (3.223)

which gives

1342 1343

n+1 = n1 + 2t (nn / cos ) 2 E n + s 1 + RH n (Tv )n

(T )n+1

n+1

(nU )k

(T )n1 + (T )n+1 (T )n 2 n1 + n+1 n , 2tR (br + hr ) 2 2 1 1 n n (UT ) + (V T ) n = (T )n1 2t 2 a cos a cos n1 n+1 + 2tD r n 2 1 = n1 2t n ( n )T pn + (V n )T n n B T n1 + n+1 1 n 2t pr , 2 r 1 p 1 = (k + f ) Vk RTv k p k a 1 2pk p
k+1/2

(3.224) (3.225)

(3.226)

(Uk+1 Uk ) + 1 p p

k1/2

(Uk Uk1 ) (3.227)

+ (FU )k , (nV )k = (k + f ) Uk RTv k 1 2pk p p p

k

cos a p (Vk Vk1 ) (3.228)

k+1/2

(Vk+1 Vk ) +

k1/2

+ (FV )k , RTv = Tk k + k (cp )k 1 2pk

Q (Tk+1 Tk ) + 59 p (Tk Tk1 ) , (3.229)

k+1/2

k1/2

Ek = (uk )2 + (vk )2 , r RTv k Tk + Tv k R , = ) (cp k c p 1 + c pv 1 q k cp

K

(3.230) (3.231)

= Bk+1/2
k+1/2 k =1

[ p + V B ] (3.232) (3.233) (3.234) (3.235)

p =
k

=1

[ p + V B ] ,
k k

1 p p
1 , pk 1 , 2pk

V k

=1

Ck [ p + V B ] ,

Ck =

<k = k,

Hk = Ck p , R r r pr r r Dk = pr Tk Ck + T r Tk1 k+1 Bk1/2 cp 2pr k k pr r T r Tk k Bk+1/2 , + 2pr k+1 k k 1, k = 0, > k, R

n

(3.236) (3.237)

where notation such as (UT ) denotes a column vector with components (Uk Tk )n . In order to complete the system, it remains to specify the reference vector hr , together with the term (1/p p/), which results from the pressure gradient terms and also appears in the semi-implicit reference vector br :

1 p p

b hr
1344 1345 1346 1347 1348 1349 1350

k r

1 p r = T , = 0. =

=
k

Bk , pk

(3.238) (3.239) (3.240)

The matrices C n and H n (i.e. with components Ck and Hk ) must be evaluated at each time step and each point in the horizontal. It is more ecient computationally to substitute the denitions of these matrices into (3.224) and (3.233) at the cost of some loss of generality in the code. The nite dierence equations have been written in the form (3.223)-(3.240) because this form is quite general. For example, the equations solved by Simmons and Str ng [1981] u at ECMWF can be obtained by changing only the vectors and hydrostatic matrix dened by (3.237)-(3.240).

1351

3.3.8

Time lter

1352 1353

The time step is completed by applying a recursive time lter originally designed by [Robert, 1966] and later studied by [Asselin, 1972]. 60

 = n +

n1

2 n + n+1

(3.241)

1354

3.3.9

Spectral transform

1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372

The spectral transform method is used in the horizontal exactly as in CCM1. As shown earlier, the vertical and temporal aspects of the model are represented by nitedierence approximations. The horizontal aspects are treated by the spectraltransform method, which is described in this section. Thus, at certain points in the integration, the prognostic variables ( + f ) , , T, and are represented in terms of coecients of a truncated series of spherical harmonic functions, while at other points they are given by gridpoint values on a corresponding Gaussian grid. In general, physical parameterizations and nonlinear operations are carried out in grid point space. Horizontal derivatives and linear operations are performed in spectral space. Externally, the model appears to the user to be a gridpoint model, as far as data required and produced by it. Similarly, since all nonlinear parameterizations are developed and carried out in gridpoint space, the model also appears as a gridpoint model for the incorporation of physical parameterizations, and the user need not be too concerned with the spectral aspects. For users interested in diagnosing the balance of terms in the evolution equations, however, the details are important and care must be taken to understand which terms have been spectrally truncated and which have not. The algebra involved in the spectral transformations has been presented in several publications [Daley et al., 1976; Bourke et al., 1977; Machenhauer, 1979]. In this report, we present only the details relevant to the model code; for more details and general philosophy, the reader is referred to these earlier papers.

1373

3.3.10

Spectral algorithm overview

The horizontal representation of an arbitrary variable consists of a truncated series of spherical harmonic functions,
M N (m) m m n Pn ()eim , m=M n=|m|

(, ) =

(3.242)

where = sin , M is the highest Fourier wavenumber included in the eastwest representation, and N (m) is the highest degree of the associated Legendre polynomials for longitudinal wavenumber m. The properties of the spherical harmonic functions used in the representation can be found in the review by Machenhauer [1979]. The model is coded for a general pentagonal truncation, illustrated in Figure 3.5, dened by three parameters: M, K, and N, where M is dened above, K is the highest degree of the associated Legendre polynomials, and N is the highest degree of the Legendre polynomials for m = 0. The common truncations are subsets of this pentagonal case: Triangular : Rhomboidal : Trapezoidal : M = N = K, K = N + M, N = K > M.

(3.243)

61

Figure 3.5: Pentagonal truncation parameters

1374 1375 1376

The quantity N (m) in (3.242) represents an arbitrary limit on the two-dimensional wavenumber n, and for the pentagonal truncation described above is simply given by N (m) = min (N + |m|, K). The associated Legendre polynomials used in the model are normalized such that
1 1 m [Pn ()]2 d = 1.

(3.244)

With this normalization, the Coriolis parameter f is f=

1377

o P1 , 0.375

(3.245)

which is required for the absolute vorticity. The coecients of the spectral representation (3.242) are given by
1 m n = 1

1 2

2 0 m (, )eim dPn ()d.

(3.246)

The inner integral represents a Fourier transform, m () = 1 2

2 0

(, )eim d,

(3.247)

which is performed by a Fast Fourier Transform (FFT) subroutine. The outer integral is performed via Gaussian quadrature,
J m n = j=1 m m (j )Pn (j )wj ,

(3.248)

where j denotes the Gaussian grid points in the meridional direction, wj the Gaussian weight at point j , and J the number of Gaussian grid points from pole to pole. The Gaussian grid points (j ) are given by the roots of the Legendre polynomial PJ (), and the corresponding weights are given by 2(1 2 ) j wj = . (3.249) [J PJ1 (j )]2 The weights themselves satisfy
J

wj = 2.0 .
j=1

(3.250)

The Gaussian grid used for the northsouth transformation is generally chosen to allow unaliased computations of quadratic terms only. In this case, the number of Gaussian latitudes J must satisfy J (2N + K + M + 1)/2 for M 2(K N) , J (3K + 1)/2 for M 2(K N) . 62 (3.251) (3.252)

For the common truncations, these become J (3K + 1)/2 for triangular and trapezoidal, J (3N + 2M + 1)/2 for rhomboidal. (3.253) (3.254)

In order to allow exact Fourier transform of quadratic terms, the number of points P in the eastwest direction must satisfy P 3M + 1 . (3.255)
1378 1379

The actual values of J and P are often not set equal to the lower limit in order to allow use of more ecient transform programs. Although in the next section of this model description, we continue to indicate the Gaussian quadrature as a sum from pole to pole, the code actually deals with the symmetric and antisymmetric components of variables and accumulates the sums from equator to pole only. The model requires an even number of latitudes to easily use the symmetry conditions. This may be slightly inecient for some spectral resolutions. We dene a new index, which goes from I at the point next to the south pole to +I at the point next to the north pole and not including 0 (there are no points at the equator or pole in the Gaussian grid), i.e., let I = J/2 and i = j J/2 for j J/2 + 1 and i = j J/2 1 for j J/2; then the summation in (3.248) can be rewritten as
I m n

=
i=I, i=0

m m (i )Pn (i )wi .

(3.256)

The symmetric (even) and antisymmetric (odd) components of m are dened by (E )m = i (O )m = i 1 m m + i , 2 i

1380 1381 1382 1383 1384 1385 1386 1387 1388

The model uses the spectral transform method [Machenhauer, 1979] for all nonlinear terms. However, the model can be thought of as starting from gridpoint values at time t (consistent with the spectral representation) and producing a forecast of the gridpoint values at time t+t (again, consistent with the spectral resolution). The forecast procedure involves computation of the nonlinear terms including physical parameterizations at grid points; transformation via Gaussian quadrature of the nonlinear terms from gridpoint space to spectral space; computation of the spectral coecients of the prognostic variables at time t + t (with the implied spectral truncation to the model resolution); and transformation back to gridpoint space. The details of the equations involved in the various transformations are given in the next section. 63

1 m m i . (3.257) 2 i Since wi is symmetric about the equator, (3.256) can be rewritten to give formulas for the coecients of even and odd spherical harmonics: I m (E )m (i )Pn (i)2wi for n m even, i m n = i=1 (3.258) I (O )m (i )P m (i )2wi for n m odd. n i
i=1

1389

3.3.11

Combination of terms

In order to describe the transformation to spectral space, for each equation we rst group together all undierentiated explicit terms, all explicit terms with longitudinal derivatives, and all explicit terms with meridional derivatives appearing in the Dyn operator. Thus, the vorticity equation (3.223) is rewritten ( + f )n+1 = V + 1 (V ) (1 2 ) (V ) , 2 ) a(1 V = ( + f )n1 , V = 2t nn , V n V = 2t nU . The divergence equation (3.224) is n+1 = D + 1 (D ) + (1 2 ) (D ) 2 D 2 ) a(1 r r r n+1 2 t (RH T + R (b + h ) n+1 ). (3.259)

where the explicit forms of the vectors V, V , and V are given as (3.260) (3.261) (3.262)

(3.263)

The mean component of the temperature is not included in the nexttolast term since the Laplacian of it is zero. The thermodynamic equation (3.226) is T n+1 = T 1 (T ) + (1 2 ) (T ) tD r n+1 . 2 ) a(1 n+1 = PS t pr r
T

(3.264)

The surfacepressure tendency (3.226) is n+1 . (3.265)

The grouped explicit terms in (3.263)(3.265) are given as follows. The terms of (3.263) are D = n1 , D = 2t nn , U n D = 2t nV , D = 2t E n + s 1 + RH r T + t RH r (T ) The terms of (3.264) are T = (T ) + 2t n tD r n1 2 n , n T = 2t(UT ) , T = 2t(V T ) . 64
n n1
n

(3.266) (3.267) (3.268)

n1

2(T )

+ R (br + hr ) n1 2n

(3.269)

(3.270) (3.271) (3.272)

The nonlinear term in (3.265) is PS = n1 2t 1n ( n )T pn + (V n )T n n B t pr

T 1 r

n1 2n .

(3.273)

1390

3.3.12

Transformation to spectral space

1391 1392 1393 1394 1395 1396

Formally, Equations (3.259)-(3.265) are transformed to spectral space by performing the operations indicated in (3.274) to each term. We see that the equations basically contain three types of terms, for example, in the vorticity equation the undierentiated term V, the longitudinally dierentiated term V , and the meridionally dierentiated term V . All terms in the original equations were grouped into one of these terms on the Gaussian grid so that they could be transformed at once. Transformation of the undierentiated term is obtained by straightforward application of (3.246)-(3.248),
J

{V}m = n
m

m Vm (j )Pn (j )wj , j=1

(3.274)

where V (j ) is the Fourier coecient of V with wavenumber m at the Gaussian grid line j . The longitudinally dierentiated term is handled by integration by parts, using the cyclic boundary conditions, (V )
m

1 2

2 o

V im e d,
2

(3.275) (3.276) (3.277)

1 = im 2

V eim d,

so that the Fourier transform is performed rst, then the dierentiation is carried out in spectral space. The transformation to spherical harmonic space then follows (3.280): 1 (V ) a(1 2 )
1397

= im
n j=1

Vm (j )

m Pn (j ) wj , a(1 2 ) j

(3.278)

where Vm (j ) is the Fourier coecient of V with wavenumber m at the Gaussian grid line j . The latitudinally dierentiated term is handled by integration by parts using zero boundary conditions at the poles: 1 (1 2 ) (V ) 2) a(1
m n

1 m (1 2 ) (V )m Pn d, 2) 1 a(1 1 1 dP m = (V )m (1 2 ) n d. 2 d 1 a(1 ) = 65

(3.279) (3.280)

Dening the derivative of the associated Legendre polynomial by

m Hn = (1 2 ) m dPn , d

(3.281)

(3.283) can be written 1 (1 2 ) (V ) a(1 2 )

m n J

(V )m
j=1

m Hn (j ) wj . a(1 2 ) j

(3.282)

Similarly, the 2 operator in the divergence equation can be converted to spectral space by sequential integration by parts and then application of the relationship n(n + 1) m Pn ()eim , a2 to each spherical harmonic function individually so that
m 2 Pn ()eim =

(3.283)

1398

m D n

n(n + 1) = a2

J m Dm (j )Pn (j )wj , j=1

(3.284)

where Dm () is the Fourier coecient of the original grid variable D .

1399

3.3.13

Solution of semi-implicit equations

The prognostic equations can be converted to spectral form by summation over the Gaussian grid using (3.274), (3.278), and (3.282). The resulting equation for absolute vorticity is ( + f )m = VSm , n n (3.285)

where ( + f )m denotes a spherical harmonic coecient of ( + f )n+1 , and the form of VSm , as n n a summation over the Gaussian grid, is given as
J

VSm n

=
j=1

m Vm (j )Pn (j ) + imVm (j )

m H m (j ) Pn (j ) wj . + Vm (j ) n a(1 2 ) a(1 2 ) j j

(3.286)

The spectral form of the divergence equation (3.263) becomes m = DSm + t n n n(n + 1) [RH r T nm + R (br + hr ) m ] , n a2 (3.287)

where m , T m , and m are spectral coecients of n+1 , T n+1 , and n+1 . The Laplacian of n n n the total temperature in (3.263) is replaced by the equivalent Laplacian of the perturbation temperature in (3.287). DSm is given by n
J

DSm n

=
j=1

Dm (j ) +

n(n + 1) m m D (j ) Pn (j ) a2 (3.288)

+imDm (j )

m Pn (j ) H m (j ) wj . Dm (j ) n a(1 2 ) a(1 2 ) j j

66

The spectral thermodynamic equation is T m = TSm tD r m , n n n with TSm dened as n

J

(3.289)

TSm n

=
j=1

m (j )Pn (j )

imTm (j )

m m Pn (j ) Hn (j ) m wj , + T (j ) a(1 2 ) a(1 2 ) j j

(3.290)

while the surface pressure equation is m = PSm m pr n n n where PSm is given by n PSm n =
j=1 J m PSm (j )Pn (j )wj . T

t , r

(3.291)

(3.292)

Equation (3.285) for vorticity is explicit and complete at this point. However, the remaining equations (3.287)(3.291) are coupled. They are solved by eliminating all variables except m : n An m = DSm + t n n where An = I + t2
1400 1401 1402 1403 1404 1405 1406 1407

n(n + 1) [RH r (TS)m + R (br + hr ) (PS)m ] , n n a2 pr

T

(3.293)

n(n + 1) RH r D r + R (br + hr ) 2 a

1 r

(3.294)

which is simply a set of K simultaneous equations for the coecients with given wavenumbers (m, n) at each level and is solved by inverting An . In order to prevent the accumulation of round o error in the global mean divergence (which if exactly zero initially, should remain exactly zero) (Ao )1 is set to the null matrix rather than the identity, and the formal application of m (3.293) then always guarantees o = 0. Once n is known, T m and m can be computed n n o from (3.289) and (3.291), respectively, and all prognostic variables are known at time n+1 as spherical harmonic coecients. Note that the mean component T o is not necessarily zero since o the perturbations are taken with respect to a specied T r .

1408

3.3.14

Horizontal diusion

As mentioned earlier, the horizontal diusion in (3.216) and (3.219) is computed implicitly via time splitting after the transformations into spectral space and solution of the semi-implicit equations. In the following, the and equations have a similar form, so we write only the equation: ( )m = n+1 n (T )m = n
m (1)i 2tK (2i) n m T n+1 n (1)i 2tK (2i)

2i ( )m (1)i ( )m 2/a2 n n 2i (T )m . n

(3.295) (3.296)

67

The extra term is present in (3.295), (3.299) and (3.301) to prevent damping of uniform rotations. The solutions are just
(2i) ( )m = Kn () n+1 n

(T )m n

(2i) = Kn (T )

m , n m T n+1 n ,

(3.297) (3.298) n(n + 1) a2 n(n + 1) a2 n(n + 1) a2 n(n + 1) a2

2

(2) Kn () = (2) Kn (T ) =

1 + 2tDn K (2) 1 + 2tDn K (2) 1 + 2tDn K 1 + 2tDn K

(4)

2 a2 ,

(3.299) (3.300)

(4) Kn

() =

4 4 a
1

(3.301)

(4) Kn (2) (4)

(T ) =

(4)

(3.302)

1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422

Kn () and Kn () are both set to 1 for n = 0. The quantity Dn represents the Courant number limiter, normally set to 1. However, Dn is modied to ensure that the CFL criterion is not violated in selected upper levels of the model. If the maximum wind speed in any of these upper levels is suciently large, then Dn = 1000 in that level for all n > nc , where nc = at max |V |. This condition is applied whenever the wind speed is large enough that nc < K, the truncation parameter in (3.243), and temporarily reduces the eective resolution of the model in the aected levels. The number of levels at which this Courant number limiter may be applied is user-selectable, but it is only used in the top level of the 26 level CAM 5.0 control runs. The diusion of T is not complete at this stage. In order to make the partial correction from to p in (3.210) local, it is not included until gridpoint values are available. This requires that 4 also be transformed from spectral to gridpoint space. The values of the coecients K (2) and K (4) for the standard T42 resolution are 2.5 105 m2 sec1 and 1.0 1016 m4 sec1 , respectively.

1423

3.3.15

Initial divergence damping

1424 1425 1426 1427 1428 1429

Occasionally, with poorly balanced initial conditions, the model exhibits numerical instability during the beginning of an integration because of excessive noise in the solution. Therefore, an optional divergence damping is included in the model to be applied over the rst few days. The damping has an initial e-folding time of t and linearly decreases to 0 over a specied number of days, tD , usually set to be 2. The damping is computed implicitly via time splitting after the horizontal diusion. r = max ( )m = n 1 (tD t)/tD , 0 t (3.303) (3.304)

1 ( )m n 1 + 2tr

68

1430

3.3.16

Transformation from spectral to physical space

m n

After the prognostic variables are completed at time n + 1 in spectral space ( + f )

m

, ( )m , n

(T )m , (n+1 )n they are transformed to grid space. For a variable , the transformation is n given by
M N (m)

(, ) =

m=M

The inner sum is done essentially as a vector product over n, and the outer is again performed by an FFT subroutine. The term needed for the remainder of the diusion terms, 4 , is calculated from N (m) M 2 n(n + 1) m m 4 n+1 = (3.306) n+1 n Pn () eim . a2 m=M
n=|m|

n=|m|

m m n Pn () eim .

(3.305)

In addition, the derivatives of are needed on the grid for the terms involving and V , V = U V + (1 2 ) . a(1 2 ) a(1 2 )
M

(3.307)

These required derivatives are given by = im m=M and using (3.281), (1 2 ) = m=M
M

N (m)

n=|m|

m m Pn () eim , n

(3.308)

N (m)

n=|m|

which involve basically the same operations as (3.306). The other variables needed on the grid are U and V . These can be computed directly from the absolute vorticity and divergence coecients using the relations ( + f )m = n n(n + 1) m m n + fn , a2 n(n + 1) m m n , n = a2 o is f1 = / .375, and 1 (1 2 ) , a a 1 (1 2 ) + . V = a a U= 69 (3.310) (3.311)

m m Hn () eim , n

(3.309)

m in which the only nonzero fn

(3.312) (3.313)

Thus, the direct transformation is

M N (m)

U = V =

a
m=M n=|m|

im 1 m m m n Pn () ( + f )m Hn () eim n n(n + 1) n(n + 1) (3.314) 1 im m ( + f )m Pn () + m H m () eim . n n(n + 1) n(n + 1) n n (3.315)

a o H1 , 2 0.375
M N (m)

a
m=M n=|m|

The horizontal diusion tendencies are also transformed back to grid space. The spectral coecients for the horizontal diusion tendencies follow from (3.295) and (3.296): FTH (T )m = (1)i+1 K 2i 2i (T ) n
m m n

,
i

(3.316) , (3.317) (3.318)

FH (( + f ) )n = (1)i+1 K 2i 2i ( + f ) (1)i ( + f ) 2/a2 FH ( )m = (1) K 2i 2i ( ) (1)i 2/a2 n

1431 1432 1433 1434

using i = 1 or 2 as appropriate for the 2 or 4 forms. These coecients are transformed to grid space following (3.242) for the T term and (3.314) and (3.315) for vorticity and divergence. Thus, the vorticity and divergence diusion tendencies are converted to equivalent U and V diusion tendencies.

1435

3.3.17

Horizontal diusion correction

After gridpoint values are calculated, frictional heating rates are determined from the momentum diusion tendencies and are added to the temperature, and the partial correction of the 4 diusion from to p surfaces is applied to T . The frictional heating rate is calculated from the kinetic energy tendency produced by the momentum diusion FFH = un1 FuH (u)/c v n1FvH (v )/c , p p (3.319)

where FuH , and FvH are the momentum equivalent diusion tendencies, determined from FH and FH just as U and V are determined from and , and c = cp 1 + p cp v 1 q n+1 . cp (3.320)

These heating rates are then combined with the correction, T n+1 Tk = Tk + (2tFFH )k + 2t B p
k

K (4) 4 n+1 .

(3.321)

70

The vertical derivatives of T (where the T p T B p T B p B =

notation is dropped for convenience) are dened by (3.322) (3.323) (3.324)

B1+ 1 (T2 T1 ) , 2 2p1 = Bk+ 1 (Tk+1 Tk ) + Bk 1 (Tk Tk1 ) , 2 2 2pk = BK 1 (TK TK1 ) . 2 2pK

The corrections are added to the diusion tendencies calculated earlier (3.316) to give the total temperature tendency for diagnostic purposes: T FTH (T )k = FTH (T )k + (2tFFH )k + 2tBk p
k

K (4) 4 n+1 .

(3.325)

1436

3.3.18

Semi-Lagrangian Tracer Transport

The forecast equation for water vapor specic humidity and constituent mixing ratio in the system is from (3.164) excluding sources and sinks. dq q p q = + V q + =0 dt t p q q dq = + V q + = 0. dt t
1437 1438 1439

(3.326)

or

(3.327)

Equation (3.327) is more economical for the semi-Lagrangian vertical advection, as does not vary in the horizontal, while p does. Written in this form, the advection equations look exactly like the equations. The parameterizations are time-split in the moisture equation. The tendency sources have already been added to the time level (n 1). The semi-Lagrangian advection step is subdivided into horizontal and vertical advection sub-steps, which, in an Eulerian form, would be written and q q = q n1 + 2t (V q)n
n+1

(3.328)

q = q + 2t n

(3.329)

In the semi-Lagrangian form used here, the general form is q = L q n1 , q n+1 = L (q ) .

1440 1441 1442

(3.330) (3.331)

Equation (3.330) represents the horizontal interpolation of q n1 at the departure point calculated assuming = 0. Equation (3.331) represents the vertical interpolation of q at the departure point, assuming V = 0. 71

The horizontal departure points are found by rst iterating for the mid-point of the trajectory, using winds at time n, and a rst guess as the location of the mid-point of the previous time step k+1 M
1443 1444 1445 1446

k+1 = A tun k , k M M M = A tv
n

a cos k , M /a,

(3.332) (3.333)

k , k M M

where subscript A denotes the arrival (Gaussian grid) point and subscript M the midpoint of the trajectory. The velocity components at k , k are determined by Lagrange cubic interM M polation. For economic reasons, the equivalent Hermite cubic interpolant with cubic derivative estimates is used at some places in this code. The equations will be presented later. Once the iteration of (3.332) and (3.333) is complete, the departure point is given by D = A 2tun (M , M ) a cos M , D = A 2tv n (M , M ) /a, (3.334) (3.335)

1447

where the subscript D denotes the departure point. The form given by (3.332)-(3.335) is inaccurate near the poles and thus is only used for arrival points equatorward of 70 latitude. Poleward of 70 we transform to a local geodesic coordinate for the calculation at each arrival point. The local geodesic coordinate is essentially a rotated spherical coordinate system whose equator goes through the arrival point. Details are provided in Williamson and Rasch [1989]. The transformed system is rotated about the axis through A , 0 and A + , 0 , by an angle A so the equator goes through (A , A ). 2 2 The longitude of the transformed system is chosen to be zero at the arrival point. If the local geodesic system is denoted by ( , ), with velocities (u, v ), the two systems are related by sin sin sin cos v cos u cos v sin sin

= = = =

sin cos A cos sin A cos (A ) , sin cos A + cos sin A cos , sin (A ) cos , v [cos cos A + sin sin A cos (A )] u sin A sin (A ) , = u cos (A ) + v sin sin (A ) .

(3.336) (3.337) (3.338) (3.339) (3.340)

1448 1449 1450 1451

The calculation of the departure point in the local geodesic system is identical to (3.332)(3.335) with all variables carrying a prime. The equations can be simplied by noting that (A , A ) = (0, 0) by design and u (A , A ) = u (A , A ) and v (A , A ) = v (A , A ). The interpolations are always done in global spherical coordinates. The interpolants are most easily dened on the interval 0 1. Dene = (xD xi ) (xi+1 xi ) , (3.341) where x is either or and the departure point xD falls within the interval (xi , xi+1 ). Following (23) of [Rasch and Williamson, 1990] with ri = 3 the Hermite cubic interpolant is given by qD = qi+1 [3 2] 2 di+1 hi 2 (1 ) +qi [3 2 (1 )] (1 )2 + di hi (1 )2 72

(3.342)

1452 1453

where qi is the value at the grid point xi , di is the derivative estimate given below, and hi = xi+1 xi . Following (3.2.12) and (3.2.13) of Hildebrand [1956], the Lagrangian cubic polynomial interpolant used for the velocity interpolation, is given by
2

fD =
j=1

j (xD ) fi+j

(3.343)

where j (xD ) =
1454

(xD xi1 ) . . . (xD xi+j1 ) (xD xi+j+1 ) . . . (xD xi+2 ) (xi+j xi1 ) . . . (xi+j xi+j1 ) (xi+j xi+j+1 ) . . . (xi+j xi+2 )

(3.344)

where f can represent either u or v, or their counterparts in the geodesic coordinate system. The derivative approximations used in (3.342) for q are obtained by dierentiating (3.343) with respect to xD , replacing f by q and evaluating the result at xD equal xi and xi+1 . With these derivative estimates, the Hermite cubic interpolant (3.342) is equivalent to the Lagrangian (3.343). If we denote the four point stencil (xi1 , xi , xi+1 , xi+2 ) by (x1 , x2 , x3 , x4 , ) the cubic derivative estimates are d2 = (x2 x3 )(x2 x4 ) q1 (x1 x2 )(x1 x3 )(x1 x4 ) 1 1 1 q2 (x1 x2 ) (x2 x3 ) (x2 x4 ) (x2 x1 )(x2 x4 ) + q3 (x1 x3 )(x2 x3 )(x3 x4 ) (x2 x1 )(x2 x3 ) q4 (x1 x4 )(x2 x4 )(x3 x4 ) (x3 x2 )(x3 x4 ) q1 (x1 x2 )(x1 x3 )(x1 x4 ) (x3 x1 )(x3 x4 ) q2 (x1 x2 )(x2 x3 )(x2 x4 ) 1 1 1 q3 + (x1 x3 ) (x2 x3 ) (x3 x4 ) (x3 x1 )(x3 x2 ) q4 (x1 x4 )(x2 x4 )(x3 x4 ) (3.345) (3.346) (3.347) (3.348)

and

d3 =

(3.349) (3.350) (3.351) (3.352)

1455 1456 1457 1458 1459 1460

The two dimensional (, ) interpolant is obtained as a tensor product application of the one-dimensional interpolants, with interpolations done rst. Assume the departure point falls in the grid box (i , i+1 ) and (i , i+1 ). Four interpolations are performed to nd q values at (D , j1), (D , j ), (D , j+1), and (D , j+2). This is followed by one interpolation in using these four values to obtain the value at (D , D ). Cyclic continuity is used in longitude. In latitude, the grid is extended to include a pole point (row) and one row across the pole. The 73

1461 1462 1463 1464

pole row is set equal to the average of the row next to the pole for q and to wavenumber 1 components for u and v. The row across the pole is lled with the values from the rst row below the pole shifted in longitude for q and minus the value shifted by in longitude for u and v. n1 Once the departure point is known, the constituent value of q = qD is obtained as indicated in (3.330) by Hermite cubic interpolation (3.342), with cubic derivative estimates (3.343) and (3.344) modied to satisfy the Sucient Condition for Monotonicity with C continuity (SCMO) described below. Dene i q by qi+1 qi i q = . (3.353) xi+1 xi First, if i q = 0 then Then, if either 0 or 0 di = di+1 = 0 . di 3 i q (3.354) (3.355)

1465 1466

di+1 3 (3.356) i q is violated, di or di+1 is brought to the appropriate bound of the relationship. These conditions ensure that the Hermite cubic interpolant is monotonic in the interval [xi , xi+1 ]. The horizontal semi-Lagrangian sub-step (3.330) is followed by the vertical step (3.331). The vertical velocity is obtained from that diagnosed in the dynamical calculations (3.222) by ()k+ 1 =
2

1 k+ 2

pk+1 pk k+1 k

(3.357)

with k = Ak + Bk . Note, this is the only place that the model actually requires an explicit specication of . The mid-point of the vertical trajectory is found by iteration
k+1 k M = A t n M .

(3.358)

Note, the arrival point A is a mid-level point where q is carried, while the used for the interpolation to mid-points is at interfaces. We restrict M by 1 M K , (3.359)

which is equivalent to assuming that q is constant from the surface to the rst model level and above the top q level. Once the mid-point is determined, the departure point is calculated from D = A 2t n (M ) , with the restriction
1467 1468 1469

(3.360) (3.361)

The appropriate values of and q are determined by interpolation (3.342), with the derivative estimates given by (3.343) and (3.344) for i = 2 to K 1. At the top and bottom we assume a zero derivative (which is consistent with (3.359) and (3.361)), di = 0 for the interval k = 1, 74

1 D K .

1470 1471 1472 1473 1474

and i+1 = 0 for the interval k = K 1. The estimate at the interior end of the rst and last grid intervals is determined from an uncentered cubic approximation; that is di+1 at the k = 1 interval is equal to di from the k = 2 interval, and di at the k = K 1 interval is equal to di+1 at the k = K 2 interval. The monotonic conditions (3.355) to (3.356) are applied to the q derivative estimates.

1475

3.3.19

Mass xers

1476 1477 1478 1479 1480 1481 1482 1483 1484 1485

This section describes original and modied xers used for the Eulerian and semi-Lagrangian dynamical cores. Let 0 , p0 and q 0 denote the values of air mass, pressure intervals, and water vapor specic humidity at the beginning of the time step (which are the same as the values at the end of the previous time step.) + , p+ and q + are the values after xers are applied at the end of the time step. , p and q are the values after the parameterizations have updated the moisture eld and tracers. Since the physics parameterizations do not change the surface pressure, and p are also the values at the beginning of the time step. The xers which ensure conservation are applied to the dry atmospheric mass, water vapor specic humidity and constituent mixing ratios. For water vapor and atmospheric mass the desired discrete relations, following Williamson and Olson [1994a] are +
2 3

q + p+ = P , q + p+ =
3 3

(3.362) q p , (3.363)

where P is the dry mass of the atmosphere. From the denition of the vertical coordinate, p = p0 A + B, and the integral
2

(3.364) includes a vertical sum

3

denotes the normal Gaussian quadrature while

followed by Gaussian quadrature. The actual xers are chosen to have the form + (, ) = M + (, ) , (3.365)

preserving the horizontal gradient of , which was calculated earlier during the inverse spectral transform, and q + (, , ) = q + + q + |+ q |. q (3.366) In (3.365) and (3.366) the ( ) denotes the provisional value before adjustment. The form (3.366) forces the arbitrary corrections to be small when the mixing ratio is small and when the change made to the mixing ratio by the advection is small. In addition, the factor is included to make the changes approximately proportional to mass per unit volume [Rasch et al., 1995]. Satisfying

75

(3.362) and (3.363) gives =

3 3 + |+ q q

q p q| p

q + p0 A M +M
3

q + + B q | + B (3.367)

0 A

3 + |+ q q

and

Note that water vapor and dry mass are corrected simultaneously. Additional advected constituents are treated as mixing ratios normalized by the mass of dry air. This choice was made so that as the water vapor of a parcel changed, the constituent mixing ratios would not change. Thus the xers which ensure conservation involve the dry mass of the atmosphere rather than the moist mass as in the case of the specic humidity above. Let denote the mixing ratio of constituents. Historically we have used the following relationship for conservation: + (1 q + )p+ =
3 3

M = P +

q p

+ .

(3.368)

(1 q )p .

(3.369)

The term (1 q)p denes the dry air mass in a layer. Following Rasch et al. [1995] the change made by the xer has the same form as (3.366) + (, , ) = + + + |+ | . Substituting (3.370) into (3.369) and using (3.365) through (3.368) gives =
3

(3.370)

(1 q )p

A,B

+ (1 q + )+ + p
A,B

A,B

+ q + |+ q |p q

A,B

+ |+ |(1 q + )p

+ |+ | q + |+ q |p q

(3.371)

where the following shorthand notation is adopted: ( )p =

A,B 3

( )p0 A + M
3

( )ps B .

(3.372)

We note that there is a small error in (3.369). Consider a situation in which moisture is transported by a physical parameterization, but there is no source or sink of moisture. Under this circumstance q = q 0 , but the surface pressure is not allowed to change. Since (1 q )p = (1 q 0 )p0 , there is an implied change of dry mass of dry air in the layer, and even in circumstances where there is no change of dry mixing ratio there would be an implied change in mass of the tracer. The solution to this inconsistency is to dene a dry air mass only once within the model time step, and use it consistently throughout the model. In this revision, we have chosen to x the dry air mass in the model time step where the surface pressure is updated, e.g. at the end of the model time step. Therefore, we now replace (3.369) with + (1 q + )p+ =
3 3

(1 q 0 )p0 .

(3.373)

76

1486 1487 1488 1489

There is a corresponding change in the rst term of the numerator of (3.371) in which q is replace by q 0 . CAM 5.0uses (3.371) for water substances and constituents aecting the temperature eld to prevent changes to the IPCC simulations. In the future, constituent elds may use a corrected version of (3.371).

1490

3.3.20

Energy Fixer
1 1 +2 2 cp T + + s + u + v+ g 2
3

Following notation in section 3.3.19, the total energy integrals are p+ = E p + S (3.374) (3.375)

E=
3

1 2 1 2 cp T + s + u + v g 2

S=
2

[(F SNT F LNT ) (F SNS F LNS SHF LX H2 O Lv P RECT ) ] t (3.376)

S =
2

[(F SNT F LNT ) (F SNS F LNS SHF LX)] t

(3.377)

+
2

[H2 O Lv (P RECL + P RECC) + H2 O Li (P RESL + P RESC)] t (3.378)

where S is the net source of energy from the parameterizations. F SNT is the net downward solar ux at the model top, F LNT is the net upward longwave ux at the model top, F SNS is the net downward solar ux at the surface, F LNS is the net upward longwave ux at the surface, SHF LX is the surface sensible heat ux, and P RECT is the total precipitation during the time step. From equation (3.365) + (, ) = M + (, ) and from (3.364) p = p0 A + B The energy xer is chosen to have the form T + (, , ) = T + + + u (, , ) = u+ + v (, , ) = v + Then = gE cp T + + s +
3 1 2

(3.379) (3.380)

(3.381) (3.382) (3.383)

u+ + v +
3

p0 A M
3

cp T + + s +
3

1 2

u+ + v +

+ B

cp p0 A + M

cp + B (3.384)

77

1491

3.3.21

Statistics Calculations

At each time step, selected global average statistics are computed for diagnostic purposes when the model is integrated with the Eulerian and semi-Lagrangian dynamical cores. Let 3 denote a global and vertical average and 2 a horizontal global average. For an arbitrary variable , these are dened by
K J I

dV =
3

ijk wj
k=1 j=1 i=1 J I

pk

2I,

(3.385)

and dA =
2

ijk wj /2I,
j=1 i=1 J

(3.386)

where recall that

wj = 2.
j=1

(3.387)

The quantities monitored are:

1/2

global rms ( + f )(s1 ) =

3

( n + f )2 dV
1/2

, ,
1/2

(3.388) (3.389) , (3.390) (3.391) (3.392)

global rms (s1 ) =

3

( n )2 dV
r n 2

global rms T (K) =

3

(T + T ) dV n dA,
2

global average mass times g (Pa) = global average mass of moisture (kg m2 ) =
3

n q n /gdV.

1492

3.3.22

Reduced grid

1493 1494 1495 1496 1497

The Eulerian core and semi-Lagrangian tracer transport can be run on reduced grids. The term reduced grid generally refers to a grid based on latitude and longitude circles in which the longitudinal grid increment increases at latitudes approaching the poles so that the longitudinal distance between grid points is reasonably constant. Details are provided in [Williamson and Rosinski, 2000]. This option provides a saving of computer time of up to 25%.

1498

3.4
3.4.1

Semi-Lagrangian Dynamical Core

Introduction

1499

1500 1501

The two-time-level semi-implicit semi-Lagrangian spectral transform dynamical core in CAM 5.0 evolved from the three-time-level CCM2 semi-Lagrangian version detailed in 78

1502 1503 1504 1505 1506 1507 1508 1509 1510 1511

Williamson and Olson [1994a] hereafter referred to as W&O94. As a rst approximation, to convert from a three-time-level scheme to a two-time-level scheme, the time level index n-1 1 becomes n, the time level index n becomes n+ 1 , and 2t becomes t. Terms needed at n+ 2 2 are extrapolated in time using time n and n-1 terms, except the Coriolis term which is implicit as the average of time n and n+1. This leads to a more complex semi-implicit equation to solve. Additional changes have been made in the scheme to incorporate advances in semi-Lagrangian methods developed since W&O94. In the following, reference is made to changes from the scheme developed in W&O94. The reader is referred to that paper for additional details of the derivation of basic aspects of the semi-Lagrangian approximations. Only the details of the two-time-level approximations are provided here.

1512

3.4.2

Vertical coordinate and hydrostatic equation

The semi-Lagrangian dynamical core adopts the same hybrid vertical coordinate () as the Eulerian core dened by p(, ps ) = A()po + B()ps , (3.393)
1513 1514 1515 1516 1517

where p is pressure, ps is surface pressure, and po is a specied constant reference pressure. The coecients A and B specify the actual coordinate used. As mentioned by Simmons and Burridge [1981] and implemented by Simmons and Str ng [1981] and Simmons and Str ng [1983], the u u coecients A and B are dened only at the discrete model levels. This has implications in the continuity equation development which follows. In the system the hydrostatic equation is approximated in a general way by
K

k = s + R
l=k
1518 1519 1520 1521 1522

Hkl (p) Tvl

(3.394)

where k is the vertical grid index running from 1 at the top of the model to K at the rst model level above the surface, k is the geopotential at level k, s is the surface geopotential, Tv is the virtual temperature, and R is the gas constant. The matrix H, referred to as the hydrostatic matrix, represents the discrete approximation to the hydrostatic integral and is left unspecied for now. It depends on pressure, which varies from horizontal point to point.

1523

3.4.3

Semi-implicit reference state

The semi-implicit equations are linearized about a reference state with constant T r and pr . We s choose T r = 350K, pr = 105 Pa (3.395) s
1524

3.4.4

Perturbation surface pressure prognostic variable

To ameliorate the mountain resonance problem, Ritchie and Tanguay [1996] introduce a perturbation ln ps surface pressure prognostic variable ln ps = ln ps ln p s s ln p = s RT r 79 (3.396) (3.397)

1525 1526 1527 1528 1529 1530

The perturbation surface pressure, ln ps , is never actually used as a grid point variable in the CAM 5.0 code. It is only used for the semi-implicit development and solution. The total ln ps is reclaimed in spectral space from the spectral coecients of s immediately after the semiimplicit equations are solved, and transformed back to spectral space along with its derivatives. This is in part because 4 ln ps is needed for the horizontal diusion correction to pressure surfaces. However the semi-Lagrangian CAM 5.0 default is to run with no horizontal diusion.

1531

3.4.5

Extrapolated variables
1

1 Variables needed at time (n + 2 ) are obtained by extrapolation

( )n+ 2 =

3 1 ( )n ( )n1 2 2

(3.398)

1532

3.4.6

Interpolants

1533 1534 1535 1536

Lagrangian polynomial quasi-cubic interpolation is used in the prognostic equations for the dynamical core. Monotonic Hermite quasi-cubic interpolation is used for tracers. Details are provided in the Eulerian Dynamical Core description. The trajectory calculation uses tri-linear interpolation of the wind eld.

1537

3.4.7

Continuity Equation

The discrete semi-Lagrangian, semi-implicit continuity equation is obtained from (16) of W&O94 modied to be spatially uncentered by a fraction , and to predict ln ps Bl where ( and (
1538 1539 1540

ln psl 1 2

n+1
A

(ln psl )n + 1 p ps +
M2

s RT r
n+1 l A

t =
D2

(1 + ) 1 p ps l l
n+ 1 2

+ (1 )

1 p ps

(3.399)
l D2

Bl n+ 1 (V l s )M2 2 RT r
n+1 A

1 (1 + ) 2

1 pr pr l l s

+ (1 )

1 pr pr l l s

n D2

1 pr pr l l s

1 n+ 2

M2

)l = ( )M 2 2 =
n+ 1

)l+ 1 (
2

)l 1 )A
1 n+ 2

(3.400) )D 2 2
n+ 1

1 (1 + ) ( 2

+ (1 ) (

(3.401)

( )l denotes a vertical dierence, l denotes the vertical level, A denotes the arrival point, D2 the departure point from horizontal (two-dimensional) advection, and M2 the midpoint of that trajectory. 80

The surface pressure forecast equation is obtained by summing over all levels and is related to (18) of W&O94 but is spatially uncentered and uses ln ps
K n+1 (ln ps )A

=
l=1

s Bl (ln psl ) + RT r
n K

1 t 2 D2
K

l=1

(1 )

1 p ps

n l D2

t t
1541 1542

l=1 K

1 l pl ps

1 n+ 2

+ t
M2 l=1

Bl n+ 1 (V l s )M2 2 RT r pr l

(3.402)

l=1

1 1 n+ 1 (1 + ) (l )n+1 + (1 ) (l )n (l )M 2 r A D2 2 ps 2

The corresponding p1s equation for the semi-implicit development follows and is related p to (19) of W&O94, again spatially uncentered and using ln ps . 1 p ps
n+1
1 k+ 2

(1 + )

2 n+1 Bk+ 1 (ln ps )A = 2 t

k

Bl (ln psl )n +
l=1 n

s RT r

D2

l=1 k

(1 ) 1 l pl ps

1 p ps
1 n+ 2

(3.403)
l D2 k

2 2
1543

+2
M2 l=1

l=1 k

Bl n+ 1 (V l s )M2 2 RT r pr l

l=1

1 1 n+ 1 (1 + ) (l )n+1 + (1 ) (l )n (l )M 2 A r D2 2 ps 2
1 p ps

This is not the actual equation used to determine used in the code to calculate eliminated.
1 p ps

in the code. The equation actually

1544 1545

involves only the divergence at time (n+1) with (ln ps )n+1

81

(1 + ) 2 t 2 +2

1 p ps
k

n+1

=
k+ 1 2 K

l=1 k

Bk+ 1
2

Bl (ln psl )n +
l=1 K

s RT r

D2 n

l=1 k

Bk+ 1
2

l=1 K

(1 ) 1 l pl ps

1 p ps
1 n+ 2

l D2

l=1 k

Bk+ 1
2

(3.404)
M2

l=1 K

l=1 k

Bk+ 1

l=1 K

Bl n+ 1 (V l s )M2 2 RT r 1 pr s + 1 n+ 1 (1 + ) (l )n+1 + (1 ) (l )n (l )M 2 A D2 2 2 (V s )n+ 2

1

2
1546

l=1

Bk+ 1
2

pr l

l=1 s RT r

The combination (ln psl )n + (3.401).

1 t 2 RT r

D2

is treated as a unit, and follows from

1547

1548

3.4.8

Thermodynamic Equation

The thermodynamic equation is obtained from (25) of W&O94 modied to be spatially uncentered and to use ln ps . In addition Hortals modication [Temperton et al., 2001] is included, in which d T s ps B (3.405) dt p ref RT r
1549 1550 1551 1552

invariant with time and can commute with the dierential operators.

is subtracted from both sides of the temperature equation. This is akin to horizontal diusion which includes the rst order term converting horizontal derivatives from eta to pressure co ordinates, with (ln ps ) replaced by RTsr , and ps B T taken as a global average so it is p
ref

82

n+1 TA

n TD

ps B()

T p T p

ref

s RT r

n+1

ps B()

T p

ref

s RT r

1 RT r RTv c p p
r

ps B()
1 n+ 2

ref

V s + s

ps B()

T p

ref

D 1 n+ 2
M

+ Qn M 1 p ps
t

RT + cp

pr s pr

d2 ln ps B() + dt p ps p
n+ 1 2 M

(3.406)

RT r pr s cp p r

RT r pr 1 s B() V s r cp p RT r
1553 1554 1555

n+ 1 2 M2

Note that Qn represents the heating calculated to advance from time n to time n + 1 and is valid over the interval. The calculation of ps B T follows that of the ECMWF (Research Manual 3, ECMWF p
ref

1556 1557

Forecast Model, Adiabatic Part, ECMWF Research Department, 2nd edition, 1/88, pp 2.252.26) Consider a constant lapse rate atmosphere

83

T = T0

p p0

R/g

(3.407)
R/g

T 1 R p = T0 p p g p0 ps R T = B T ps B p p g (ps )ref R T = Bk ps B (Tk )ref for (Tk )ref > TC p ref (pk )ref g ps B T p
ref

(3.408) (3.409) (3.410) (3.411) (3.412) (3.413) (3.414) (3.415) (3.416) (3.417) (3.418)

= 0 f or (Tk )ref TC (pk )ref (ps )ref 1013.25mb 288K 1000mb 6.5K/km 216.5K
R/g

(pk )ref = Ak p0 + Bk (ps )ref (Tk )ref = T0 (ps )ref T0 p0 TC = = = = =

1558

3.4.9

Momentum equations

1559 1560 1561 1562

The momentum equations follow from (3) of W&O94 modied to be spatially uncentered, to use oe ln ps , and with the Coriolis term implicit following Ct and Staniforth [1988] and Temperton [1997]. The semi-implicit, semi-Lagrangian momentum equation at level k (but with the level subscript k suppressed) is

84

V n+1 V n A D t

1 2

(1 + ) f k V

n+1
A

+ (1 ) f k V

n
D

+ Fn M
1 n+ 2

1 2

B (1 + ) (s + RH k T v ) + RTv ps ln ps p B + (1 ) (s + RH k T v ) + RTv ps ln ps p

A 1 n+ 2

1 2

(1 + ) [RH r T + k

n+1 RT r ln ps ]A
1

(3.419)

n+ (1 + ) [s + RH r T + RT r ln ps ]A 2 k

+ (1 ) [s + RH r T + RT r ln ps ]n k D
n+ (1 ) [s + RH r T + RT r ln ps ]D 2 k
1563 1564 1

The gradient of the geopotential is more complex than in the system because the hydrostatic matrix H depends on the local pressure: (H k T v ) = H k [(1 + v q) T + v T q] + T v H k (3.420)

where v is (Rv /R 1) and Rv is the gas constant for water vapor. The gradient of T is calculated from the spectral representation and that of q from a discrete cubic approximation that is consistent with the interpolation used in the semi-Lagrangian water vapor advection. In general, the elements of H are functions of pressure at adjacent discrete model levels Hkl = fkl (pl+1/2 , pl , pl1/2 ) The gradient is then a function of pressure and the pressure gradient Hkl = gkl (pl+1/2 , pl , pl1/2 , pl+1/2 , pl , pl1/2 ) (3.422) (3.421)

The pressure gradient is available from (3.393) and the surface pressure gradient calculated from the spectral representation pl = Bl ps = Bl ps ln ps (3.423)
1565

3.4.10

Development of semi-implicit system equations

V n+1 V n 1 A D = t 2 1 2

The momentum equation can be written as (1 + ) f k V

n+1
A

+ (1 ) f k V
n+1

n
D

(1 + ) [RH r T + RT r ln ps ]A k 85

+ RHSV ,

(3.424)

1566

1 where RHSV contains known terms at times (n + 2 ) and (n). By combining terms, 3.424 can be written in general as
n+1 n+1 UA A + VA A = UAA + VAA + UDD + VDD , i j i j i j

(3.425)

where and denote the spherical unit vectors in the longitudinal and latitudinal directions, i j respectively, at the points indicated by the subscripts, and U and V denote the appropriate n+1 combinations of terms in 3.424. Note that UA is distinct from the UA . Following Bates et al. [1990], equations for the individual components are obtained by relating the unit vectors at the departure points (D ,D ) to those at the arrival points (A ,A ): i j i j = u + u iD i j A A A A = v + v , j i j
D A A A A 1567 1568

(3.426) (3.427)

in which the vertical components (k) are ignored. The dependence of s and s on the latitudes and longitudes of the arrival and departure points is given in the Appendix of Bates et al. [1990]. W&O94 followed Bates et al. [1990] which ignored rotating the vector to remain parallel to the earths surface during translation. We include that factor by keeping the length of the vector written in terms of A , A the same as the length of the vector written in terms of D , D . i j i j Thus, (10) of W&O94 becomes UA where =
u v UD A + VD A 2 2 UD + VD 2 u v + UD A + VD A 2
n+1 n+1

VA

u v = UA + A UD + A VD u v = VA + A UD + A VD

(3.428)
1 2

(3.429)

After the momentum equation is written in a common set of unit vectors V n+1 + A 1+ 2 t f k V
n+1
A

1+ 2

t [RH r T + RT r ln ps ]A k

n+1

= R V

(3.430)

Drop the ( )n+1 from the notation, dene A = (1 + ) t and transform to vorticity and divergence R 1 v v cos = (R cos ) a a cos u 1+ n+1 sin + u cos + t2 [RH r T + RT r ln ps ]A k a 2 R 1 u + (R cos ) = v a cos + sin + 86 (3.432) (3.431)

(3.433)

Note that u cos = 1 cos 2 2 a a 1 cos v cos = 2 + 2 a a (3.434) (3.435)

Then the vorticity and divergence equations become + sin + 2 2 a cos 2 2 a 1 R v = (R cos ) = L (3.436) a cos u cos 1+ n+1 t2 [RH r T + RT r ln ps ]A 2 + k 2 a 2 1 Ru = + (R cos ) = M (3.437) a cos v +

sin +

2 2 a

Transform to spectral space as described in the description of the Eulerian spectral transform dynamical core. Note, from (4.5b) and (4.6) on page 177 of Machenhauer [1979]
m m m m m Pn = Dn+1 Pn+1 + Dn Pn1 m Dn =

(3.438) (3.439)

n2 m2 4n2 1

1 2

and from (4.5a) on page 177 of Machenhauer [1979] 1 2 m m m m m P = nDn+1 Pn+1 + (n + 1) Dn Pn1 n (3.440)

Then the equations for the spectral coecients at time n + 1 at each vertical level are
m n 1 m n

im n (n + 1) im 1 n (n + 1)

n n+1 m m m Dn+1 + n1 Dn = L m n n+1 n n n+1 m m m m n+1 Dn+1 n1 Dn n+1 n n (n + 1) 1+ m t RH r T m + RT r ln ps n = Mm k n n 2 a2

m + n+1

(3.441) (3.442)

lnps n

= PSm n

T m = TSm n n

1+ 2 1+ 2

t pr r ps tD r m n

m n

(3.443) (3.444)

87

The underbar denotes a vector over vertical levels. Rewrite the vorticity and divergence equations in terms of vectors over vertical levels. m 1 n n n+1 m m Dn+1 m Dn (3.445) n1 n+1 n n (n + 1) 1+ m t RH r T m + RT r ln ps n = DS m n n 2 2 a im n n+1 m m 1 + m Dn+1 + m Dn = V S m (3.446) n+1 n1 n n (n + 1) n+1 n im n (n + 1) m n+1

m n Dene hm by n

m ghm = RH r Tn + RT r ln ps n n

(3.447)

and Am = 1 n B+ n = B n =
m m

im n (n + 1) n m Dn+1 n+1 n+1 m Dn n

(3.448) (3.449) (3.450)

Then the vorticity and divergence equations are

m Am m + B + n m + B n m n+1 n1 = VSn n n m m

(3.451) (3.452)

Am m B + n m B n m n n n+1 n1
1569 1570

1+ 2

n (n + 1) m ghn = DSm n 2 a

Note that these equations are uncoupled in the vertical, i.e. each vertical level involves variables at that level only. The equation for hm however couples all levels. n 1+ = 2 pr pr s
T

ghm n

t RH D + RT

m + RH r TSm + RT r PSm n n n

(3.453)

Dene C r and HSm so that n ghm = n

1571 1572

1+ 2

tCr m + HSm n n

(3.454)

Let gD denote the eigenvalues of C r with corresponding eigenvectors and is the matrix with columns = 1 2 . . . L (3.455)

1573

and gD the diagonal matrix of corresponding eigenvalues

88

C = gD C r = gD
1

gD = g
r

D1 0 0 D2 . . .. . . . . .

0 0 . . .

(3.456)

0 0 DL

(3.457) (3.458) (3.459) (3.460) (3.461)

m

Then transform m n = 1 m n m n = 1 m n m = 1 hm hn n , , ,
m

VSn = 1 VSm n DSn = 1 DSm n HSn = 1 HSm n

m m m m m m

Am n + B+ n n+1 + B n n1 = VSn n
m m m m m Am n B+ n n+1 B n n1 n

(3.462) (3.463) (3.464) (3.465)

g hn +

m 1+ n (n + 1) m t g hn = DSn 2 2 a m 1+ t1 C r 1 m = HSn n 2 1+ 1 m m m hn + tD n = HS 2 g n

1574 1575 1576 1577

Since D is diagonal, all equations are now uncoupled in the vertical. For each vertical mode, i.e. element of ()m , and for each Fourier wavenumber m we have n a system of equations in n to solve. In following we drop the Fourier index m and the modal element index ( ) from the notation. An n + B+ n n+1 + B n n1 = VSn n (n + 1) 1+ t g hn = DSn An n B+ n n+1 B n n1 2 a2 1+ 1 hn + tD n = HSn 2 g (3.466) (3.467) (3.468)

1578 1579 1580

The modal index ( ) was included in the above equation on D only as a reminder, but will also be dropped in the following. Substitute and h into the equation. 1+ 2
2

An +

(t)2

n (n + 1) gD + B+ n A1 B n+1 + B n A1 B+ n1 n n+1 n1 a2 (3.469)

+ B+ n A1 B+ n+1 n+2 + B n A1 B n1 n2 n+1 n1 n (n + 1) 1+ t HSn + B+ n A1 VSn+1 + B n A1 VSn1 = DSn + n+1 n1 2 2 a 89

1581 1582

which is just two tri-diagonal systems of equations, one for the even and one for the odd ns, and m n N At the end of the system, the boundary conditions are n = m, B n = B m = 0 m m n = m + 1, B n1 = B m = B (m+1)1 = 0 the n2 term is not present, and from the underlying truncation m m N +1 = N +2 = 0 For each m and we have the general systems of equations N +1 or n = m, m + 2, ..., N +2 An n+2 + Bn n Cn n2 = Dn , N +1 n = m + 1, m + 3, ..., or N +2 Cm = Cm+1 = 0 N +1 = N +2 = 0 Assume solutions of the form then Em = FM Am Bm Dm = Bm (3.476) (3.477) n = En n+2 + Fn (3.471)
m

(3.470)

(3.472)

(3.473) (3.474) (3.475)

En =

An , Bn Cn En2 Dn + Cn Fn2 , Bn Cn En2

Fn =

N = FN or N 1 = FN 1 , m n = N 2, N 4, ..., or m+1 n = En n+2 + Fn , m+1 or n = N 3, N 5, ..., m 90

N 2 or n = m + 2, m + 4, ..., N 3 N or n = m + 2, m + 4, ..., N 1

(3.478)

(3.479)

(3.480)

(3.481)

1583

Divergence in physical space is obtained from the vertical mode coecients by m m = n n (3.482)

1584

The remaining variables are obtained in physical space by

m n 1

im n (n + 1)

m = Lm n+1 n

T m = TSm n n lnps n
m

= PSm n

n m m Dn+1 n1 n+1 1+ tD r m n 2 1 + t T pr m n 2 pr s

n+1 m Dn n

(3.483) (3.484) (3.485)

1585

3.4.11

Trajectory Calculation
dR =V dt

The trajectory calculation follows Hortal [1999] Let R denote the position vector of the parcel, (3.486)

which can be approximated in general by Rn = Rn+1 tV M D A Hortals method is based on a Taylors series expansion Rn+1 = Rn + t D A or substituting for dR/dt Rn+1 = Rn + tV n + D D A Approximate dV dt giving VM
1586 1 n+ 2

(3.487)
n

dR dt

+
D

t2 2

d2 R dt2
n

+ ...
D

(3.488)

t2 2

dV dt

+ ...
D

(3.489)

n D

n1 Vn VD A t

(3.490)

n+ 1 2

1 2

2V n V n1

+Vn A

(3.491)

for the trajectory equation.

1587

3.4.12

Mass and energy xers and statistics calculations

1588 1589 1590

The semi-Lagrangian dynamical core applies the same mass and energy xers and statistical calculations as the Eulerian dynamical core. These are described in sections 3.3.19, 3.3.20, and 3.3.21. 91

92

1591

Chapter 4 Model Physics

As stated in chapter 2, the total parameterization package in CAM 5.0 consists of a sequence of components, indicated by P = {M, R, S, T } , (4.1)

1592

1593 1594 1595 1596 1597 1598 1599 1600 1601 1602

where M denotes (Moist) precipitation processes, R denotes clouds and Radiation, S denotes the Surface model, and T denotes Turbulent mixing. Each of these in turn is subdivided into various components: M includes an optional dry adiabatic adjustment normally applied only in the stratosphere, moist penetrative convection, shallow convection, and large-scale stable condensation; R rst calculates the cloud parameterization followed by the radiation parameterization; S provides the surface uxes obtained from land, ocean and sea ice models, or calculates them based on specied surface conditions such as sea surface temperatures and sea ice distribution. These surface uxes provide lower ux boundary conditions for the turbulent mixing T which is comprised of the planetary boundary layer parameterization, vertical diusion, and gravity wave drag. The updating described in the preceding paragraph of all variable except temperature is straightforward. Temperature, however, is a little more complicated and follows the general procedure described by Boville and Bretherton [2003a] involving dry static energy. The state variable updated after each time-split parameterization component is the dry static energy si . Let i be the index in a sequence of I time-split processes. The dry static energy at the end of the ith process is si . The dry static energy is updated using the heating rate Q calculated by the ith process: si = si1 + (t) Qi (si1 , Ti1 , i1 , qi1 , ...) (4.2) In processes not formulated in terms of dry static energy but rather in terms of a temperature tendency, the heating rate is given by Qi = (Ti Ti1 ) / (Cp t). The temperature, Ti , and geopotential, i , are calculated from si by inverting the equation for s s = Cp T + gz = Cp T + (4.3) with the hydrostatic equation
K

1603 1604

k = s + R
l=k
1605

Hkl Tv l

(4.4)

substituted for .

93

1606 1607 1608 1609

The temperature tendencies for each process are also accumulated over the processes. For processes formulated in terms of dry static energy the temperature tendencies are calculated from the dry static energy tendency. Let Ti /t denote the total accumulation at the end of the ith process. Then Ti1 si Ti = + /Cp (4.5) t t t si (si si1 ) /Cp = /Cp (4.6) t t which assumes is unchanged. Note that the inversion of s for T and changes T and . This is not included in the Ti /t above for processes formulated to give dry static energy tendencies.. In processes not formulated in terms of dry static energy but rather in terms of a temperature tendency, that tendency is simply accumulated. After the last parameterization is completed, the dry static energy of the last update is saved. This nal column energy is saved and used at the beginning of the next physics calculation following the Finite Volume dynamical update to calculate the global energy xer associated with the dynamical core. The implication is that the energy inconsistency introduced by sending the T described above to the FV rather than the T returned by inverting the dry static energy is included in the xer attributed to the dynamics. The accumulated physics temperature tendency is also available after the last parameterization is completed, TI /t. An updated temperature is calculated from it by adding it to the temperature at the beginning of the physics. TI = T0 + TI t t (4.7)

1610 1611 1612 1613 1614

This temperature is converted to virtual potential temperature and passed to the Finite Volume dynamical core. The temperature tendency itself is passed to the spectral transform Eulerian and semi-Lagrangian dynamical cores. The inconsistency in the use of temperature and dry static energy apparent in the description above should be eliminated in future versions of the model.

94

1615

4.1

1616

Conversion to and from dry and wet mixing ratios for trace constituents in the model

1617 1618

There are trade os in the various options for the representation of trace constituents in any general circulation model: 1. When the air mass in a model layer is dened to include the water vapor, it is frequently convenient to represent the quantity of trace constituent as a moist mixing ratio m , that is, the mass of tracer per mass of moist air in the layer. The advantage of the representation is that one need only multiply the moist mixing ratio by the moist air mass to determine the tracer air mass. It has the disadvantage of implicitly requiring a change in m whenever the water vapor q changes within the layer, even if the mass of the trace constituent does not. 2. One can also utilize a dry mixing ratio d to dene the amount of constituent in a volume of air. This variable does not have the implicit dependence on water vapor, but does require that the mass of water vapor be factored out of the air mass itself in order to calculate the mass of tracer in a cell. NCAR atmospheric models have historically used a combination of dry and moist mixing ratios. Physical parameterizations (including convective transport) have utilized moist mixing ratios. The resolved scale transport performed in the Eulerian (spectral), and semi-Lagrangian dynamics use dry mixing ratios, specically to prevent oscillations associated with variations in water vapor requiring changes in tracer mixing ratios. The nite volume dynamics module utilizes moist mixing ratios, with an attempt to maintain internal consistency between transport of water vapor and other constituents. There is no right way to resolve the requirements associated with the simultaneous treatment of water vapor, air mass in a layer and tracer mixing ratios. But the historical treatment signicantly complicates the interpretation of model simulations, and in the latest version of CAM we have also provided an alternate representation. That is, we allow the user to specify whether any given trace constituent is interpreted as a dry or wet mixing ratio through the specication of an attribute to the constituent in the physics state structure. The details of the specication are described in the users manual, but we do identify the interaction between state quantities here. At the end of the dynamics update to the model state, the surface pressure, specic humidity, and tracer mixing ratios are returned to the model. The physics update then is allowed to update specic humidity and tracer mixing ratios through a sequence of operator splitting updates but the surface pressure is not allowed to evolve. Because there is an explicit relationship between the surface pressure and the air mass within each layer we assume that water mass can change within the layer by physical parameterizations but dry air mass cannot. We have chosen to dene the dry air mass in each layer at the beginning of the physics update as
0 m pd = (1 qi,k )i,k i,k

1619 1620 1621 1622 1623 1624 1625

1626 1627 1628 1629

1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644

for column i, level k. Note that the specic humidity used is the value dened at the beginning of the physics update. We dene the transformation between dry and wet mixing ratios to be d = (pd /pm )m i,k i,k i,k i,k 95

1645 1646 1647 1648 1649 1650

We note that the various physical parameterizations that operate on tracers on the model (convection, turbulent transport, scavenging, chemistry) will require a specication of the air mass within each cell as well as the value of the mixing ratio in the cell. We have modied the model so that it will use the correct value of p depending on the attribute of the tracer, that is, we use couplets of (m , pm ) or (d , pd ) in order to assure that the process conserves mass appropriately. We note further that there are a number of parameterizations (e.g. convection, vertical diusion) that transport species using a continuity equation in a ux form that can be written generically as F () = (4.8) t p where F indicates a ux of . For example, in convective transports F () might correspond to Mu where Mu is an updraft mass ux. In principle one should adjust Mu to reect the fact that it may be moving a mass of dry air or a mass of moist air. We assume these dierences are small, and well below the errors required to produce equation 4.8 in the rst place. The same is true for the diusion coecients involved in turbulent transport. All processes using equations of such a form still satisfy a conservation relationship t k pk = Fkbot Fktop

1651 1652 1653 1654 1655 1656

1657

provided the appropriate p is used in the summation.

96

1658

4.2

Moist Turbulence Scheme

1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679

Moist turbulence scheme in CAM5 is from Bretherton and Park [2009a] that is a replacement of dry turbulence scheme of Holtslag and Boville [1993b] in CAM3 and CAM4. The role of moist turbulence scheme is to vertically transport heat ( dry static energy s Cp T + g z ), moistures ( qv , ql , qi , nl , ni where nl , ni are number concentrations of cloud liquid droplets and ice crystals ), horizontal momentum ( u, v ), and tracers ( mass and number concentrations of aerosol and chemical species ) by symmetric turbulences. In the symmetric turbulence, updraft and downdraft have similar vertical velocities, fractional areas, and degrees of saturation. Compared to the dry PBL ( Planetary Boundary Layer ) scheme in CAM3 and CAM4, moist turbulence scheme in CAM5 has the following unique characteristics: (1) it is a diagnostic TKEbased ( Turbulent Kinetic Energy, e ) 1st order K-diusion scheme with entrainment parameterization but without counter-gradient transport, (2) it simulates cloud radiation turbulence interactions in an explicit way, which makes it possible to simulate full aerosol indirect eects with direct interactions with cloud macro-microphysics and radiation schemes, (3) using a single set of consistent physical formula, it is operating in any layers above as well as within PBL as long as moist Ri ( Richardson number ) is larger than a critical value Ric =0.19. Thanks to explicit simulation of moist turbulences driven by LW ( Longwave ) radiative cooling at the cloud top, CAM5 does not need a separate formula for stability-based stratus fraction - stratus fraction is computed only using mean relative humidity. It performs much better in the cloudtopped PBL than CAM3/4s dry PBL scheme with similar or superior performance in dry stable and convective PBLs. In order to illucidate conceptual background behind the CAM5s moist turbulence scheme, lets imagine a single symmetric turbulence being perturbed by a static vertical distance l from its equilibrium height. This symmetric turbulence is assumed to be imbedded in the environment without vertical discontinuity such as sharp inversion. If l is suciently smaller than the vertical length scale over which vertical gradient of environmental scalar ( /z ) changes and if turbulent vertical velocity ( w ) is approximated to we can easily derive that turbulent ux e, = l however, atmospheric of any conservative scalar ( ) becomes w e . In reality, stability controls turbulent vertical velocity ( i.e., w will be a product of e and an anisotropic factor of TKE, which is a function of atmospheric stability ) and actual vertical perturbation distance of turbulent updraft and downdraft ( i.e., turbulent mixing length will be a product of a static perturbation distance l and a certain atmospheric stability parameter ). In addition, during vertical displacement, turbulent properties may be changed due to diabatic forcings or mixing with environment. All of these anomalous eects associated with atmospheric stability, diabatic forcings, and mixing are incorporated into a single stability function, S. As a result, turbulent ux of conservative scalar by symmetric turbulences embedded in the vertially-smoothtransitioning environment becomes w = l e S = K (4.9) Thus, computation of turbulent uxes by symmetric turbulence is reduced to the computations of static turbulence length scale ( l ), turbulent kinetic energy ( e ), and stability function ( S ). The product of these 3 terms is so called eddy diusivity, K = l e S. Due to diabatic adjustment of turbulent horizontal momentum to the environment during vertical displacement, 97

1680 1681 1682 1683

1684 1685

S for horizontal momentum ( Sm ) is likely to be smaller than the S for heat and moisture ( Sh ). This means that K is a function of scalar, . If turbulence is embedded in the environment with a sharp vertical transition of stability such as inversion layer at the top of convective PBL, Eqn.(4.9) is inappropriate since turbulent motion will be suppressed in the stable portion of the discontinuous interface. In this case, we use the following entrainment parameterization. w = we = ze we = Ke (4.10)

1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697

where we is entrainment rate and ze is the thickness of the entrainment interfacial layer. Above entrainment parameterization is applied at the top and base interfaces of Convective Layer ( CL. See Fig.1 ) after nishing CL extension-merging procedures that will be explained later. In this case, eddy diusivity is simply a product of ze and we , identical for all scalars. CAM5s moist turbulence scheme consists of 9 main processes: (1) Bulk Moist Richardson number, (2) Initial identication of Convective ( CL ), Stably Turbulent ( STL ), and Stable ( SL ) Layers, (3) Turbulence Length Scale ( l ), (4) Steady-State TKE ( e ), (5) Stability Functions ( S ), (6) CL Extension-Merging Procedures, (7) Entrainment Rates at the CL Top and Base Interfaces ( we ), (8) Implicit Diusion with Implicit Eddy Diusivity, and (9) Implicit Surface Stress. Since many symmetric turbulences exist with dierent vertical length and velocity scales at any interface, the quantities we are trying to parameterize ( l, e, S , we ) should be understood as the ensemble of all symmetric turbulences.

1698

4.2.1

Bulk Moist Richardson Number

1699 1700 1701 1702 1703 1704 1705

Richardson number ( Ri ) is used to diagnose the existence of turbulences. It is dened as the ratio of buoyancy production ( Pb w b = (g/v ) w v ) to shear production ( Ps u u/z w v /z ) at the model interface. P represents energy conversion from mean w v b available potential energy ( APE ) to TKE, while Ps is converison from mean kinetic energy to TKE. If Ri is negative, turbulence is absolutely generated but if it is positive, turbulence can be either generated or dissipated depending on the relative magnitude of |Pb | and |Ps |. Special treatment is necessary for saturated turbulences. If turbulence keeps its unsaturated state during vertical diaplacement, v is a conserved quantity and Eqn.(4.9) can be directly used for computing w v . However, if it is saturated, v decreases within downdraft due to evaporative cooling of cloud droplet, while increases within updraft due to condensation heating of water vapor. The resulting w b including the eects of condensation and evaporation can be represented by the linear combinations of w sc and w qt where sc Cp T + g z Lv ql Ls qi is condensate static energy and qt qv + ql + qi is total specic humidity. Both sc and qt are conserved during vertical displacement and phase change. If we know saturated fractional area at the model interface ( e.g., stratus fraction ), we can write
w b = ch w sc + cq w qt

(4.11) (4.12)

where ch , cm are buoyancy coecients for heat ( sc ) and moisture ( qt ) which are complex functions of temperature and pressure ( Schubert et al. [1979], Bretherton and Park [2009a] ), 98

ch = ch,s a + ch,u (1 a),

cq = cq,s a + cq,u (1 a)

Figure 4.1: The indexing and example of turbulent layer structure in the bottom part of a typical column in the CAM5s moist turbulence scheme. Layer indexing used in CAM5 is shown at left. The bulk Richardson number Ri is used to locate a stable interface ( SI ) with Ri > Ric , stably turbulent interfaces ( STI ) with 0 < Ri < Ric comprising an STL, and unstable convective interfaces ( CI ) with Ri < 0 comprising a CL core. The CL is extended up to an entrainment interfaces ( EI ), at which the turbulent eddy diusivity is computed from an explicitly predicted entrainment rate we . In the interior interfaces of the turbulent layers, the turbulent diusivity K is conventionally using a length scale, diagnosed TKE, and stability function computed from local Ri in an STL and from layer-mean RI in a CL. Thickness of STL and CL are denoted by h, and the thickness of entrainment interface is denoted by ze . See text for further details.

99

subscrits s, u denotes saturated and unsaturated turbulences, and a is stratus fraction. Applying Eqn.(4.9) to each term on the R.H.S. of Eqn.(4.11) and assuming Kh Km , Ri can be written solely in terms of grid mean variables as follows. Ri CPb Ps = CKh N 2 Km S 2 CN 2 S 2 N 2 ch Cc z + cq C qt z, s
1706 1707 1708 1709 1710

(4.13) (4.14)

S 2 (C uz)2 + (Cz)2 v

In the code, ch,s , ch,u , cq,s , cq,u and a are given at the layer mid-points. By averaging two adjacent layers mid-point values, these are converted into model interface values. However, if a in the adjacent upper layer is smaller than the average value, we took the smaller value, which is a necessary procedure to identify stably stratied interface at the top of stratocumulus-capped PBL.

1711 1712

4.2.2

Identication of Convective, Stably Turbulent, and Stable Layers

1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740

Using Ri dened at the interfaces, all model interfaces are grouped into Convective Interface ( CI with Ri 0 ), Stably Turbulent Interface ( STI with 0 Ri Ric ), and Stable Interface ( SI with Ri > Ric ). Here, Ric = 0.19 is a critical Richardson number and turbulence can exist only when Ri < Ric ( see Fig.4.1 ). If several CIs are adjacent ( or even when one CI exists ), they form a single Convective Layer , CL. If several STIs are adjacent ( or even when one STI exists ), they form a single Stably Turbulent Layer , STL. The remaining SIs form Stable Layer , SL where no turbulence exists. After nishing CL extension-merging which will be explained later, the external interfaces surrounding CL will be named as Entrainment Interface ( EI with Ri > 0 ) and the remaining CL interfaces as CL internal interfaces. In CL and STL, we neglect TKE storage. In STL, we further neglect TKE transport. In CL, turbulence can exist from the base to the top interfaces of CL, but in the STL, turbulence can exist from the layer mid-point just below the lowest STI to the layer mid-point just above the highest STI. This denes turbulent layer thickness, h in a slightly dierent way between CL and STL ( see Fig.4.1 ). After CL extension-merging, a single stability function S int is assigned to all CL internal interfaces. We use Eqn.(4.9) to compute eddy diusivity at all interfaces except the top entrainment interface of CL where Eqn.(4.10) is used. If CL is elevated from the surface, Eqn.(4.10) is also applied to the CL base entrainment interface. If any interface is a double entraining interface from above and below, nal eddy diusivity is a simple sum of the two eddy diusivities obtained from above CL and below CL. If surface buoyancy ux is positive ( negative ), surface is considered as a CL internal ( external ) interface and contributes ( does not contribute ) to the computation of internal energetics of CL. Several CLs and STLs can exist in a single grid column. The same physical equation set is used for all CLs and STLs regardless of whether they are based at the surface or elevated. Our moist turbulence scheme, thus, is not a PBL scheme - it is operating in all layers above as well as within the PBL. The conventional PBL is simply a surface-based CL or surface-based STL in our scheme. PBL top height is dened as the top external interface of surface-based CL. If

100

1741 1742 1743 1744 1745 1746 1747 1748

STL is based at the surface instead, PBL top height is dened as the layer mid-point height just above the STL top interface. We dont impose any minimum value on the PBL height. We also dene additional Single Radiatively Driven Convective Layer ( SRCL ) if any single layer satises the following 5 conditions: it has (1) grid-mean LWC larger than 102 [g kg 1] but no LWC in the layer just above it, (2) LW radiative cooling, (3) Ri > Ric at the top interface, (4) positive buoyancy production in the upper half-layer, and (5) it is not within the previously identied CLs. Similar to other CLs, entrainment parameterization is applied at the top and base interfaces of SRCL. Several SRCLs can exist in a single column.

1749

4.2.3

Turbulent Length Scale

1750

Following Blackadar [1962] and Grenier and Bretherton [2001], turbulent length scale is computed as (C1l) = (C1k z) + (C1l ) (4.15) l = h =
1751 1752 1753 1754 1755 1756 1757 1758 1759 1760

(4.16)
int

where k = 0.4 is a von Karman constant, l is asymptotic length scale, h is turbulent layer thickness, and Ri int = l2 N 2 int / l2 S 2 int is the mean Ri averaged over CL internal interfaces int ( denotes vertical average over the CL internal interfaces ). We chose = 3. As explained in the previous section and Fig.4.1, h for CL is dened as the depth between two external interfaces, while h for STL is dened as the distance between the two outmost layers mid-points. When CL is based at surface but surface buoyancy ux is negative, h is dened down to the mid-point of the lowest model layer instead of down to the surface. Our formulation approximates l to l except near the ground, where it asymptotes k z to match surface layer similarity theory. As explained before, the actual turbulent mixing length should be understood as the product of l and stability function S.

0.085 0.085 [ 2 exp(min(0, Ri

at STI )) ] at CI

(4.17)

1761

4.2.4

Steady-State Turbulent Kinetic Energy

1762

We assume steady-state TKE, that is, at each model interface, Pb + Ps + Te D = 0 where dissipation ( D ) and TKE transport ( Te ) are parameterized as D = Ce3/2 b1 l Te = ae C e ( e e)l (4.18) (4.19)

where b1 = 5.8 and e denotes TKE averaged over the whole CL. In case of STL, ae = 0 ( no TKE transport ) while in CL, ae = 1. Ideally, Te should be zero but Eqn.(4.19) only satises this condition approximately. Combining with Pb and Ps , steady-state TKE at any model interface becomes (4.20) e = b1 Cl e (Pb + Ps ) + ae ( e e) 101

1763 1764 1765

At the internal interfaces of CL, Pb = Kh N 2 and Ps = Km S 2 with Kh = l e Sh int and Km = l e Sm int where Sh int and Sm int are mean stability functions averaged over internal interfaces of CL computed by using Ri int and Eqn.(4.23) and (4.24). At entrainment interfaces, Pb = Ke N 2 + fR ch,s FR / and Ps = Ke S 2 with Ke = we ze where ze is the thickness of entrainment interface. In computing Pb at the entrainment interfaces, N 2 is redened using the cloud fraction in the half-layer just below ( above ) the CL top ( base ) entrainment interface. This redenition of N 2 is necessary in order to correctly take into account of buoyancy production associated with the evaporative cooling of entrained airs. FR is grid-mean radiative ux divergence across the CL top layer in unit of [W m2 ] given from the radiation scheme and 0 fR 1 is the fraction of radiative ux divergence conned in the thin transition zone near the top entrainment interface of CL among FR . Following Bretherton and Park [2009a], fR is parameterized as fR = [C (4 + )24 + (6 + )] = 156 qltop ptop /g (4.21) (4.22)

1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785

where is the grid-mean cloud optical depth of CL top layer that has grid-mean LWC of qltop top and a thickness of p . As PBL deepens, cloud can be formed in the layer just above the PBL top ( i.e., ambiguous layer ). In this case, most of radiative ux divergence will be conned at the top of the ambiguous layer. In order to take into account of this case, we simply added fR FR / both in the CL top and ambiguous layers. At the surface interface, Pb,sf c = ch (Fh /) + cm Fq / where Fh is sensible heat ux [J s1 m2 ] and Fq is water vapor ux [kg s1 m2 ] at surface given from the surface ux computation scheme to the moist turbulence scheme, and Ps,sf c = u3 /(k z0 ) where z0 is the mid-point height of the lowest model layer and u is frictional velocity at surface dened as 2 2 u2 = x + y / where x , y is surface momentum ux [kg m s1 s1 m2 ]. Assuming no TKE transport and turbulent length scale l = k z0 , we compute TKE at surface half-layer, esf c using Eqn.(4.20). In order to prevent negative es , we impose a minimum positive value on esf c . By integrating Eqn.(4.20) over the whole CL with an approximation of e e at the entrainment interfaces, we can compute e by solving a cubic equation of e 1/2 . Once e is computed, we can compute e at each internal interfaces of CL using Eqn.(4.20) again. Our moist turbulence scheme computes characteristic excesses ( or standard deviations ) of turbulent updraft vertical velocity ( w ), temperature ( T ), and water vapor ( q ) within e ( w = u /8.5 ) and then PBL or near surface. If PBL is CL ( STL ), we estimate w = T = (Fh //Cp )/w and q = (Fq /)/w . These characteristic convective excesses when PBL is CL are used to dene cumulus source air properties within deep convection scheme.

1786

4.2.5

Stability Functions

1787

Following Galperin et al. [1988], stability functions are parameterized as Sh = [C5 1 + 3 Gh ] Sm = [C1 + 2 Gh (1 + 3 Gh ) (1 + 4 Gh )] 102 (4.23) (4.24)

where 1 = 0.5562, 2 = 4.3643, 3 = 34.6764, 4 = 6.1272, 5 = 0.6986 and Gh N 2 l2 /(2 e) is a nondimensional stability ratio restricted by 3.5334 < Gh < 0.0233. In case of STL, Gh is obtained by combining Eqn.(4.20),(4.23),(4.24) as follows. Ri (1 + 3 Gh ) (1 + 4 Gh ) = 2 b1 Gh [Ri 5 (1 + 4 Gh ) (1 + 2 Gh )]
1788 1789 1790 1791 1792 1793

(4.25)

For Ri > Ric = 0.19, there is no physically realizable solution Gh and the interface is assumed to be non-turbulent. For Ri < Ric , this polynormial has two real roots, but only the larger one is realizable. In case of internal CL, we use the same Eqn.(4.25) but with Ri int = l2 N 2 int / l2 S 2 int to compute Gh int , Sh int and Sm int . For Ri int < 44.5, the allowable upper bound Gh int = 0.0233 is exceeded and stability functions assume their max max maximum values Sh = 3.64 and Sm = 2.76.

1794

4.2.6

CL Extension-Merging Procedure

1795

Several CLs can be identied in a single grid column. A contiguous set of interfaces with negative Ri is initially identied as a CL core. Starting from the lowest CL, each CL is extended rst upward as far as possible, then downward as far as possible from the CL core into any adjacent layers of suciently weak stable stratication. Any external interface of CL is incorporated into the CL if the following criterion is satised. (z l2 N 2 )E < Crinc 1 rinc
1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815

CLint

l2 N 2 dz

(4.26)

where superscript E denotes CL external interface being tested for incorporation into CL, z is the thickness of external interface, and the R.H.S. denotes vertical integration over the CL internal interfaces. We chose rinc = 0.04 to be consistent with a dry convective boundary layer in which the entrainment buoyancy ux is -0.2 of the surface buoyancy ux. Strictly speaking, Eqn.(4.26) compares buoyancy production during TKE dissipation time scale by assuming that Sh of the external interface being tested for merging is the same as the Sh int . The rst interface above ( below ) CL that fails this criterion will be the top ( bottom ) entrainment interface for that CL. No extension-merging is performed for SRCL since SRCL does not have internal interfaces. Above criteria ensures that as long as the initial CL internal core ( CLint ) has net positive buoyancy production, the internal CL after incorporating external interface will also have positive buoyancy production. Our incorporation test also guarantees that if any external interface is unstably stratied, it will be incorporated. Thus, if we incorporate any of the interior of a CL, we will incorporate or merge all of it. If CL is based at surface and surface buoyancy ux is positive, the contribution of surface half-layer is also incorporated into the above integration of CLint . In the surface half-layer, we 3/2 use Gh = (k z0 Pb,sf c )/(2 Sh esf c ) and by combining with Eqn.(4.23), Sh can be computed. Finally, (l2 N 2 )sf c = k z0 Pb,sf c /(Sh esf c ). In computing turbulent length scale from Eqns.(4.15)-(4.17) during CL extension-merging procedure, we simply assume = 0.5 (0.085 + 0.170) = 0.1275 for all merging and merged

103

1816 1817 1818

interfaces and turbulent layer thickness h is xed by the initial value before CL extensionmerging. After nishing all the extension-merging procedure, Sh int and Sm int are computed using Ri int = l2 N 2 int / l2 S 2 int and the updated h.

1819

4.2.7

Entrainment Rates at the CL Top and Base Interfaces

1820

At entrainment interfaces, eddy diusivity is computed using Eqn.(4.10). Entrainment rate we is computed as 3 we = A (Cvl g) Cw E svl h s (4.27) where h = zt zb is the thickness of CL, E svl is the jump of mean liquid virtual static energy svl sl (1 + 0.61 qt ) across the entrainment interfaces at the top ( E svl = svl (kt 1) svl (kt ) ) and base ( E svl = svl (kb 1) svl (kb ) ) of the CL. w is convective velocity dened as
zt 1/3

w = 2.5 and A is an entrainment coecient dened as A=

1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833

zb

Pb dz

(4.28)

0.2 1 + a2 0.8 CLv qltop E svl 0.2

at CL top at CL base

(4.29)

where a2 is a tuning parameter being allowed to be changed between 10 and 60, and we chose a2 = 30. As PBL deepens, cloud can be formed in the ambiguous layer. In order to fully take into account of all possible mixtures between PBL air and free air above inversion, E svl in computing A in Eqn.(4.29) is obtained by using the layer-value just above the ambiguous layer, not the value in the ambiguous layer ( i.e., E svl = svl (kt 2) svl (kt ) in Eqn.(4.29) ). In the top similar context, we take ql = max[l (kt ), ql (kt 1)]. q Due to the way how the model is structured, we dont have information on Kh when entrainment rate is computed. Thus, in computing Pb = Kh N 2 at CL internal interfaces for entrainment parameterization, we use Kh of previous iteration or previous time step. Since we are using a predictor-corrector iteration method, Kh is likely to converge as iteration is proceeded. If eddy diusivity at the entrainment interface is smaller than the value obtained by assuming entrainment interface is STI, the nal eddy diusivity is set to be that of STI.

1834

4.2.8

Implicit Diusion with Implicit Diusivity

1835 1836 1837 1838 1839 1840 1841 1842

CAM5 is using process splitting. At each time step, successive parameterizations operate on the updated state resulting from the previous parameterizations. The parameterizations in CAM5 are called in the following order at each time step: deep convection shallow convection ( computes (1) cumulus fraction and condensate, (2) vertical transport of heat, moisture, momentum, and tracers by asymmetric turbulences ) stratiform macrophysics ( stratus fractions and stratiform net condensation-deposition rates ) stratiform microphysics ( (1) activation of cloud liquid droplets and ice crystals, (2) 104

1843 1844 1845 1846 1847 1848 1849

conversions among cloud liquid droplets, ice crystals, rain, and snow, (3) evaporations of precipitation and sedimented cloud condensates ) wet deposition of aerosols radiation surface uxes ( upward uxes of heat, water vapor, momentum, and tracers at surface ) aerosol and chemical conversion processes ( conversions among various aerosol and chemical species ) turbulent diusion ( vertical transport of heat, moisture, momentum, and tracers by symmetric turbulences ) gravity wave drag dry deposition of aerosols dynamics ( large scale advection of grid mean heat, moisture, momentum, and tracers ). Given a diusivity prole K(z) and an input state (z) updated to include all physical and dynamic processes ( including explicit deposition of surface uxes into the lowest model layer except horizontal momentum ), our moist turbulence scheme diuses using the following implicit backward Euler scheme. C (t + t) t = Cz K(z) Cz (t + t) (4.30)

1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860

subject to specied upward surface uxes of horizontal momentums ( x,tot = x ktms us,i, 1 1 2 y,tot = y ktms vs,i in unit of [kg m s s m ] where the second term on the R.H.S. is turbulent mountain stress obtained by using updated surface wind us,i, vs,i after ith iteration 1 2 ), sensible heat ( Fh in unit of [J s m ] ), and water vapor ( Fq in unit of [kg s1 m2 ] ) where superscript denotes the input value given to the moist turbulence scheme. The eddy diusivity prole K(z) may be computed using the input state variable . However, when a long time step t = 1800 [s] is used as in CAM5, this is not a desirable approach since the physical processes proceeding turbulent diusion scheme ( e.g., radiation ) can dramatically destabilize the input prole (z), resulting in unreasonable K(z). To address this problem, we use an iterative predictor-corrector approach to recalculate eddy diusivities based on an better approximation to the post-diusion state. Lets assume that Ki is diusivity prole obtained from i . When i = 0, 0 and K0 K . Using 0 , we compute K0 and obtain the rst diused prole 1 by solving Eqn.(4.30) . Using 1 , we compute K1 and the predictor-corrector averaged applied to the initial state K0:1 K1 + (1 ) K0 . Using this K0:1 , we diuse the input state again and obtain the second diused prole 2 from which K2 and K0:2 K2 + (1 ) K0:1 are computed. By repeating this process, the predictor-corrector averaged K prole after N iteration becomes

K0:N KN + (1 ) K0:N 1 K0:0 K

1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871

(4.31) (4.32)

We chose N = 4 and = 0.5 to compute the nal eddy diusivity K = K0:N from the eddy diusion scheme. During individual iterative diusion processes of by K = K0:n (1 n N), we diused conservative scalars = s , qt , u , v and reconstructed the diused nonl , qv , ql , qi proles by assuming that (1) qi is not diused and (2) the layer conservative scalars T has homogeneous distribution of cloud condensate across the grid at saturation equilibrium state. Since the initial proles are continuously updated within each iteration, we should also update surface uxes ( x , y , Fh , Fq ) and the proles of stratus fraction and radiative heating rate within each iteration. However, this will hugely increase computation time and make the CAM5 structure much more complex. Thus, we simply ignore the variations of surface uxes, stratus fraction and radiative heating rate during iteration. This simplication can inevitably 105

1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899

cause Ki ( i 1 ) to be computed on the i that has inconsistency among various state variables. Finally, because of the ipping of layer structures and corresponding eddy diusivities between the iterations, our predictor-corrector method may not produce fully convergent K regardless of the iteration number. By choosing = 0.5, however, we address this issue to our best. Once the nal K is obtained from Eqn.(4.31), we diuse the input grid mean scalars ( s, qv , ql , qi , ni , u, v ). Vertical transport of horizontal momentum by symmetric turbulence conserves column-mean horizontal momentum. However, it will change column-mean kinetic energy ( KE ) of the mean wind. In reality, this KE change will be converted into TKE and eventually internal heat energy ( or potential energy, PE ). In CAM5, however, we dont store TKE between time steps because of steady-state TKE assumption and yet require conservation of column-mean total energy, PE+KE. In order to conserve total energy in each column, we computed KE dissipation heating rate in each layer following Boville and Bretherton [2003b] after diusing horizontal momentum, and explicitly added KE dissipation heating into s before diusing s. Since air parcel temperature changes during vertical displacement due to adiabatic compression-expansion, moist turbulence scheme should also handle associated condensationevaporation of cloud droplets during vertical transport. The same should be true for convection and large-scale advection schemes. However, this evaporation-condensation associated with vertical diaplacement of air parcels will be treated in the following stratiform cloud macrophysics. Thus, diusing non-conservative scalars with a phase change ( s, qv , ql , qi , nl , ni ) is not a problem if we admit that reasonable proles of cloud condensates can be restored only after stratiform macro-microphysics. When turbulence transports non-saturated airs into the overlying saturated airs, new cloud droplets can be formed without the change of cloud condensate mass ( so called, cloud droplet activation ). In order to handle adiabatic turbulent vertical transport and concurrent diabatic sources of cloud droplet number in a consistent way, turbulent diusions of nl , aerosol mass and numbers are separately treated by the cloud droplet activation routine within the stratiform microphysics.

1900

4.2.9

Implicit Surface Stress

1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914

In CAM5, surface uxes of various scalars ( s, qv , ql , qi , nl , ni and tracers ) are explicitly deposited into the lowest model layer ( this forms the input to Eqn.(4.30) ) and then implicit vertical diusion is performed using Eqn.(4.30). In case of surface momentum uxes ( x , y ), however, such explicit adding can ip the direction of the lowest model layer wind ( u , vs s ). This is not a physically realizable situation since as wind speed decreases by surface drag, surface drag itself decreases too, preventing ipping of wind in nature. This ipping of the wind in the model can be a source of numerical instability especially when the lowest model layer is thin. Thus, x , y should be added into the lowest model layer in an implicit way. This implicit adding, however, will cause discrepancy between the horizontal momentum that the Earth surface lost ( which are explicit surface momentum ux x , y given to the turbulent diusion scheme ) and the momentum that the atmosphere receives ( which are implicit surface momentum ux ). To conserve horizontal momentum of the whole coupled system, they should be identical. In order to address both the numerical stability and momentum conservation 106

1915 1916 1917

issues, we partitioned the residual surface momentum ux ( = explicit surface momentum ux - implicit surface momentum ux ) over a certain time interval, e.g., 2 hr. This process is called implicit surface stress being detailed below. First, in order to compute implicit surface stress, we compute total surface drag coecient ( ktot ) by summing the normal drag coecient ( knor ) obtained from the lowest model layer wind and the turbulent mountain stress drag coecient ( ktms ) triggered by sub-grid distribution of surface topography. This ktms is computed by separate turbulent mountain stress module. knor = max C
(x )2 + (y )2 max(

( )2 + (s )2 , 1), 104 us v

(4.33) (4.34)

ktot = knor + ktms

Second, a certain fraction of residual stress accumulated upto the current time step from the rst time step is added into the lowest model layer. This changes the wind in the lowest model layer. u+ (t) = u (t) + x,res (t) (Cgp) t, s s = [Cttres ] ,
1918 1919 1920 1921 1922

vs (t) = vs (t) + y,res (t) (Cgp) t, + t tres = 7200 [sec]

(4.35) (4.36)

where x,res (t), y,res (t) are residual stress accumulated upto the current time step from the rst time step, and tres is a time interval over which residual stress is deposited into the atmosphere. With K(z), ktot , and given input wind proles of u(t) and v (t) but with the lowest model layer + + winds of us (t) and vs (t) instead of us (t) and vs (t), we can solve Eqn.(4.30) to obtain implicitly time-marched wind proles, u(t + t) and v(t + t). Finally, the net residual stress accumulated upto the end of current time step which will be used at the next time step becomes
x,res (t + t) = x,res (t) (1 ) + x + knor us (t + t) y,res (t + t) = y,res (t) (1 ) + y + knor vs (t + t)

(4.37) (4.38)

1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936

where us (t + t) and vs (t + t) are implicitly marched winds of the lowest model layer at the end of turbulent diusion scheme at the current time step. At the rst time step, it is x,res (t) = y,res (t) = 0. Our formulation assumes that turbulent mountain stress is fully implicitly added into the atmosphere without generating any residual stress. This assumption causes no conservation problem since turbulent mountain stress is used only within the atmospheric model not in the ocean, sea ice, and land models. One complexity arises because K(z) is iteratively computed at each time step. We assume that all of x,res (t), ktms and knor are not changed within the iteration loop : ktms and knor are obtained from the initial wind prole u (t), vs (t) given to the moist turbulence scheme. In s computing eddy diusivity Ki within each iteration loop, however, we used x,tot = x ktms us,i(t), y,tot = y ktms vs,i (t) where us,i(t), vs,i (t) are iteratively updated wind in the lowest th model layer after i iteration at the current time step. Here, we included turbulent mountain stress in computing eddy diusivity since it is a source of shear production and TKE in the lowest model layer, too.

107

1937

4.3

Shallow Convection Scheme

1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975

Shallow convection scheme in CAM5 is from Park and Bretherton [2009] that is a replacement of Hack [1994b] shallow convection scheme in CAM3 and CAM4. Similar to its precedents, CAM5 performs shallow convection scheme just after deep convection scheme. In general, distinctions between deep and shallow convections are made by the dierences in cloud top height, the existence of convective precipitation and convective downdraft. While named as shallow convection, CAM5s shallow convection scheme does not have any limitation on its cloud top height and convective precipitation. However, because the proceeding deep convection scheme consumes most of Convective Available Potential Energy ( CAPE ) and stabilizes the atmosphere, cloud top height simulated by shallow convection scheme is naturally limited in the tropical regions. In contrast to deep convection scheme, shallow convection scheme does not have a separate formulation for convective downdraft, but have an explicit parameterization of penetrative entrainment in the overshooting zone near cumulus top. Future implementation of convective downdraft as well as renements of other aspects ( e.g., updraft mixing rate and cloud microphysics ) can make shallow convection scheme work for deep convective case, too. The role of shallow convection scheme is to vertically transport heat, moisture, momentum, and tracers by asymmetric turbulences. On the other hands, vertical transport by symmetric turbulences are performed by separate moist turbulence scheme. CAM5s shallow convection scheme is carefully designed to optimally operate with CAM5s moist turbulence scheme without missing or double-counting processes. Similar to the other convection schemes, CAM5 shallow convection scheme assumes (1) steady state convective updraft plume, and (2) small updraft fractional area, so that compensating subsidence entirely exists within the same grid box as convective updraft. One of the unique aspects of CAM5 shallow convection scheme is its ability to compute convective updraft vertical velocity and updraft fractional area by using updraft vertical momentum equation. Computation of updraft vertical velocity enables to compute more rened fractional entrainment-detrainment rates, cloud top height, and penetrative entrainment. While not implemented in the current CAM5s shallow convection scheme, updraft vertical velocity will make it possible to compute activated fraction of aerosol masses and numbers at the cumulus base, more elegant cumulus microphysics, and aerosol-cumulus interactions. CAM5s shallow convection scheme consists of 8 main processes: (1) Reconstruction of mean proles and cloud condensate partitioning, (2) Computation of source air properties of a single ensemble-mean updraft plume at the PBL ( Planetary Boundary Layer ) top, (3) Cloud base mass ux and updraft vertical velocity closures using Convective Inhibition ( CIN ) and TKE ( Turbulent Kinetic Energy ), (4) Vertical evolution of a single entraining-detraining buoyancy sorting plume from the PBL top to the cumulus top, (5) Penetrative entrainment in the overshooting zone near cumulus top, (6) Computation of convective uxes within the PBL, (7) Computation of grid-mean tendencies of conservative scalars, and (8) Computation of grid-mean tendencies of non-conservative scalars. The following sections describe each of these processes in detail.

1976 1977

4.3.1

Reconstruction of Mean Proles and Cloud Condensate Partitioning

1978

108

Figure 4.2: Schematic structure of shallow cumulus scheme describing vertical evolution of a bulk cumulus updraft and its interaction with environment and the subcloud layer. Black dots denote environmental mean virtual potential temperature e,v , from which a e,v prole ( solid line ) is reconstructed. The horizontal solid lines are ux interfaces, where the updraft virtual potential temperature v,u ( open circles ) is computed, from which a cumulus updraft v,u prole ( dashed ) is reconstructed. The model layer and interface indices used in CAM5 are denoted on the right axis. The layer index I indicates the ambiguous layer, and pinv is the reconstructed PBL capping inversion within this layer. Environmental conservative variables reconstructed just above and below the ambiguous layer are denoted by e,I+1/2 and e,I1/2 , respectively. See the text for details.

109

1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

The input state variables to shallow convection scheme are environmental mean qv , ql , qi , T , u, v, and mass and number of aerosols, . While stratus condensate should reside only within cloudy portion, shallow convection scheme assumes uniform distribution of stratus condensate across the grid except when evaporation of precipitation is computed. From the given inputs, we compute condensate potential temperature c = (Lv /Cp /) ql (Ls /Cp /) qi and total specic humidity qt = qv + ql + qi . With respect to vertical displacement involving phase change but without precipitation formation and evaporation of precipitation, c is nearly conserved and qt is completely conserved. Instead of assuming zero slope, we assign a certain slope of c and qt within each layer. In c (k + 1) c (k))/(p(k + 1) p(k)) ] and downward [ (c (k) c (k each layer, upward [ ( 1))/(p(k) p(k 1)) ] slopes are computed. If they have dierent signs or either of two has zero slope, internal slope is set to zero. If they have the same sign, we take the one with minimum absolute slope. In the lowest model layer, internal slope is set to the upward slope, and in the top model layer, it is set to the slope of the layer below. This prole reconstruction is performed indepently to each of c , qt , u, v and . The reconstructed proles conserve mean quantity in each layer but have discontinuity across the model interfaces. Similar prole reconstruction is performed in the moist turbulence scheme. From the reconstructed c and qt , we compute virtual potential temperature v = (1 + 0.61 qv ql qi ) at just below and above each model interface by assuming that ice fraction among cloud condensate is a raming function of temperature between 248 K and 268 K, and saturation specic humidity is a weighting average of two values dened over water and ice. The same temperature pertitioning is applied to cloud condensate within convective updraft. In case of detrained convective condensate, we use 238.15 K and 268.15 K as the two end points of temperature in the cloud condensate ramping function. For computation of radiative properties of cumulus updraft, we repartition in-cumulus condensate into liquid and ice following the partitioning of pre-existing of stratus clouds.

2005

4.3.2

Source Air Properties of Convective Updraft

2006

At the PBL top, we dene source air properties of a single updraft plume. In CAM5, PBL top is located at the top most interface of convective boundary layer, which is diagnosed by the separate moist turbulence scheme. Here, we dene qt,src , c,src , usrc, vsrc , src where the hat denotes convective updraft properties and the subscript src denotes the values of convective updraft source air at the PBL top interface. qt,src is dened as the environmental-mean value in the lowest model layer ( In the below equations, (1) denotes the lowest model layer value ). src is dened in the same way as qt,src . We rst dene condensate virtual potential temperature of source air ( vc = c (1 + 0.61 qt ) ) using the prole-reconstructed minimum value within the PBL ( vc,min ), and from qt,src and vc,src , c,src is computed. usrc and vsrc are dened as the prole-reconstructed values just below the PBL top interface. qt,src = qt (1) c,src = C vc,min (1 + 0.61 qt,src ) (4.39) (4.40)

110

usrc = utop vsrc = vtop src = (1)

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

(4.41) (4.42) (4.43)

includes the mass of individual aerosol species and aerosol numbers in each mode ( Aitken, Accumulation, Coarse ). also contains the numbers of cloud liquid droplets and ice crystals. Since CAM5s cumulus microphysics is the rst moment scheme and the size of in-cumulus and detrained shallow convective condensate are independently specied, vertical convective transport of cloud droplet numbers do not inuence climate simulation in the current CAM5. But we retain this functionality to transport cloud droplet number for future development of higher order cumulus microphysics and aerosol-cumulus interactions. The only unknown source air properties at this stage are updraft mass ux ( Msrc ) and updraft vertical velocity ( wsrc ) which are computed in the next section. Msrc and wsrc allows us to compute updraft fractional area, Asrc .

2017

4.3.3

Closures at the Cloud Base

2018

We assume that turbulent updraft vertical velocity w at the PBL top follows a symmetric Gaussian distribution. The width of the distribution w is determined by the mean TKE within the PBL ( eP BL ) given from the moist turbulence scheme, w = k eP BL + em where k = 1 4 2 2 and em = 5 10 [m s ] is a background minimum TKE. P(w), PDF of w at the PBL top is given as 2 P (w) = C1w 2 exp Cw 2 2 w (4.44)
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029

Among these, only strong updrafts enough to reach to their undiluted Level of Free Convection ( LFC ) are assumed to form a single ensemble mean convective updraft being simulated by shallow convection scheme. The eects of remaining weak updrafts that eventually sink back to the PBL by negative buoyancy are implicitly simulated by the separate moist turbulence scheme through entrainment parameterization. We dene CIN as the strength of potential energy barrier of the undiluted ensemble-mean plume from the PBL top to the undiluted LFC ( see Fig.4.2 ). Then, the minimum vertical velocity of the deatable convective updrafts, or critical vertical velocity becomes wc = 2 a CIN where buoyancy coecient a = 1. In order to reduce the on-and-o behavior of convection between the long model time step t = 1800 [s], CIN is computed using thermodynamic proles at the end of convection time step ( so called, implicit CIN ) as described in Park and Bretherton [2009]. Then, mass ux ( Msrc ), updraft fractional area ( Asrc ), and area-weighted updraft vertical velocity ( wsrc ) of a single ensemble-mean convective updraft at the PBL top can be computed as follows by integrating all deatable plumes with w > wc . Msrc =
wc

2 w P (w) dw = Cw 2 exp Ca CIN w

wc

(4.45)

Asrc =

P (w) dw = C12 erf C a CINw 111

(4.46)

wsrc = C
2030 2031 2032 2033 2034 2035

wc

w P (w) dw

wc

P (w) dw = C Msrc Asrc

(4.47)

Here, we impose additional constraints that (1) Msrc should be smaller than 0.9 of the mass src < 0.9 p(I 1)/g/t where p(I 1) is the in the layer just below the PBL top, that is, M pressure thickness of the layer just below the PBL top, and (2) Asrc and ALCL ( updraft core fractional area at the LCL ) are smaller than 0.1, assuming no lateral mixing from the PBL top to the LCL. From the constrained Msrc and Asrc , we compute the constrained wsrc . As of this, we nished the convective closure at the PBL top.

2036

4.3.4

Vertical Evolution of A Single Updraft Plume

2037

Assuming steady-state updraft plume ( or updraft plume with very small fractional area ), vertical variations of updraft mass ux and conservative scalars can be written as C1M C M p = C p = ( e ) + S + C
2038 2039 2040 2041 2042 2043 2044 2045

(4.48) (4.49)

where pressure coordinate p is dened increasing upward, (, ) are fractional entrainment and detrainment rates, respectively, = qt , c , u, v, is scalar being transported, is updraft value, e is environmental mean value ( note that this is dierent from the grid-mean = A + (1 e unless A = 0 ), S is net diabatic source within cumulus updraft, and C is a direct A) conversion term from environmental to updraft without lateral mass exchange. In case of steady state updraft plume, S changes the column mean total energy, while C conserves the column and C for each component are parameterized as follows. Otherwise, mean total energy. S they are set to zero. q Sqt p = max(l + qi qc,crit, 0) q q Sc p = max (CLv ql + Ls qi Cp (l + qi )) (l + qi qc,crit), 0 Cu = P GF c (C ue p) , Cv = P GF c (Ce p) v (4.50) (4.51) (4.52)

2046 2047 2048 2049 2050 2051 2052 2053 2054 2055

where qc,crit = 0.7 [gkg 1] is maximum cloud condensate amount that cumulus updraft can hold, and P GF c = 0.7 measures the degree to which cumulus updraft adjusts to environment by large scale horizontal pressure gradient force during vertical motion. Above Sqt and Sc assume that if in-cumulus cloud condensate is larger than qc,crit , the excessive condensate is simply precipitated out. This simple cumulus microphysics can be rened using updraft vertical velocity and cloud drop size distribution in future. Following Gregory et al. [1997a], Cu and Cv assume that when cumulus updraft rises across the layer with vertical shear of environmental horizontal wind, updraft gains horizontal momentum increment directly from the environment without lateral mass exchange. We neglect radiative eect and evaporation of convective precipitation within convective updraft. One unique aspect of our shallow convection scheme is to compute updraft vertical velocity for computing (1) updraft fractional area, (2) lateral entrainment and detrainment rates, and 112

2056 2057 2058 2059

(3) cumulus top height and penetrative entrainment rates. Steady state vertical momentum equation is C12 C w 2 p = a B b w 2 (4.53) where B is updraft buoyancy ( B = (g/v ) (v v ) ), and non-dimentional coecients a, b include the partition of perturbation vertical PGF into buoyancy and entrainment drag forces. Without perturbation vertical PGF, a = b = 1 but we use a = 1, b = 2 assuming that perturbation vertical PGF is entirely incorporated into entrainment drag force. Instead of directly parameterizing (, ), we assume that a certain amount of updraft airs ( o M p ) is mixed with the same amount of environmental airs during incremental vertical displacement p, producing a spectrum of mixtures with the same mixing probability P () = 1 where 0 1 ( = 0 is cumulus updraft, = 1 is environmental air ). Among these mixtures, we entrain (1) positively buoyant mixtures and (2) negatively buoyany mixtures with vertical velocity strong enough to reach 0.1 of cumulus top height. This process is so called inertial buoyancy sorting ( Kain and Fritsch [1990], Bretherton et al. [2004] ). This allows us to compute a single critical mixing fraction c : mixtures with c are entrained while the others are detrained. Then, we can derive (, ) as follows. = o 2 c = o (1 c )2 (4.54) (4.55)

where fractional mixing rate o is parameterized as an inverse function of geometric height, o = Cc g z

2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070

(4.56)

where non-dimensional coecient c = 8 and z is geometric height above the surface. In order to simulate deep convection, we can use a smaller value, e.g., c = 4. Cumulus top height necessary to compute c is initially set to the previous time steps value and then recomputed using an iteration loop. Now, we can compute vertical evolution of M , , w. Instead of solving discrete numerical equation, we used the explicit analytical solution by solving the rst order dierential equation to obtain the cumulus updraft properties at the top interface of each layer from the value at the base interface. In solving Eqn.(4.53), we assume a linear prole of B in each layer. At the top interface, we computed updraft fractional area A from M and w, and if A > 0.1, detrainment rate is enhanced such that A is always less than 0.1. Note that this enhancement of detrainment only changes M not w at the top interface.

2071

4.3.5

Penetrative Entrainment

2072

When convective updraft rises into the stably stratied layers ( i.e., Overshooting Zone. See Fig.4.2 ) above the Level of Neutral Buoyancy ( LNB ), some air masses within the overshooting zone are entrained into the layers below. This process is so called penetrative entrainment. We assume that the amount of penetratively entrained airs ( Mpen ) is proportional to the mass involved in the lateral mixing in the overshooting zone and the properties of penetratively 113

entrained airs ( pen ) are identical to the mean environmental values from LNB to LNB + ppen : Mpen = rpen M o ppen (4.57) pen = e (4.58)
2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083

where ppen is vertical overshooting distance of cumulus updraft above LNB and 1 rpen 10 is a tunable non-dimensional penetrative entrainment coecient. In CAM5, we chose rpen = 10. The thickness of overshooting zone above LNB, or the cumulus top height is diagnosed as the level where convective updraft vertical velocity w becomes zero. When convective updraft penetrates into several layers above LNB, Eqn.(4.57) and (4.58) are computed for each layers within penetrative overshooting zone, and all the penetratively entrained mass is deposited into a single layer just below LNB. We neglect convective updraft uxes at the interfaces at and above LNB since most of updraft mass uxes crossing over the LNB are likely to sink down below LNB due to negative updraft buoyancy in the overshooting zone. The thickness of overshooting zone above LNB, or the cumulus top height is diagnosed as the level where convective updraft vertical velocity w becomes zero.

2084

4.3.6

Convective Fluxes at and below the PBL top interface

2085

We view the layer just above the PBL top ( ambiguous layer, I. See Fig.4.2 ) as the accumulation of partial grid layer of PBL air and another partial grid layer of above-PBL air. The interface between these two partial layers, the reconstructed PBL top height pinv , is computed using a simple conservation principle for individual scalar component = qt , c , u, v, as follows. pinv = piI1 r |pI |, r = C e,I e,I+1/2e,I1/2 e,I+1/2 (4.59)
2086 2087 2088

where |pI | is the pressure thickness of the ambiguous layer, pi is the pressure at the model interface, e,I1/2 and e,I+1/2 are the prole-reconstructed environmental values just below the PBL top interface and just above the ambiguous layer, respectively ( See Fig.4.2 ). Convective updraft mass ux Msrc is assumed to be deated from the pinv with src , which enables us to compute convective ux at the pinv . To avoid over stabilizing or destabilizing the ambiguous layer and PBL through cumulus ventilation, this ux is uniformly extracted throughout the whole PBL, which results in the following linear prole of convective ux at model interfaces below pinv . ( )(k) = g Msrc (src e,I1/2 ) Cpi0 pik pi0 pinv , f or 0 k I 1 (4.60)

2089

where is pressure vertical velocity, k = 0 is surface, k = I 1 is PBL top interface index. It is possible for compensating subsidence associated with cumulus updraft mass ux to lower the pinv below the bottom of the ambiguous layer, in which case compensating subsidence will also warm and dry the grid layer below. To diagnose whether compensating subsidence would lower pinv below piI1 during t, we compare the normalized cumulus updraft mass ux, rc = (g Msrc t)/|pI | to r. If rc r, pinv will be lowered down into the layer I 1, replacing 114

 PBL-top air with = e,I1/2 with above-PBL air with = e,I+1/2 . This eect is included by adding the below compensating subsidence ux ( )(k = I 1) = g Msrc (e,I+1/2 e,I1/2 ) 1 Crrc ,
2090 2091 2092

f or rc > r

(4.61)

where we assumed that cumulus mass ux is not strong enough to lower down pinv below piI2 , that is, g Msrc t < r |pI | + |pI1 |. In order to ensure this condition, we impose an upper bound on the convective base mass ux of g Msrc t < 0.9 |pI1 |.

2093

4.3.7

Grid-Mean Tendency of Conservative Scalars

2094

In case of steady state updraft plume approximation with a nite updraft fractional area but compensating subsidence entirely within the same grid box as convective updraft, the budget equation of grid mean conservative scalar represented in ux convergence form becomes C t = Ct Ae e = gCp M ( e ) + Mpen (pen e ) +gMS +Ct Ae e S (4.62) is environmental fractional area and on the R.H.S. the rst and second where Ae = 1 A terms are convergence of convective updraft and penetrative entrainment uxes, respectively, and the third and fourth terms are diabatic sources within convective updraft and environment, respectively. We use the above ux convergence form to compute tendencies of conservative scalars in order to ensure conservation of column-integrated energy during vertical redistribu tion of air masses by convective updraft. M in the third term of the R.H.S. is obtained by averaging updraft mass uxes at the top and base interfaces of each layer. In contrast to , e is discontinuous across the model interface due to prole reconstruction. In order to take into account of the eects of compensating subsidence ( upwelling ) in this ux form, e in the rst ( second ) term on the R.H.S is taken as the reconstructed environmental value just above the top interface ( below the base interface ) of each layer. If downdraft is also considered in future, we should add g /p[Md (d e )] + g Md Sd, on the R.H.S. If = u, v, diabatic sources both within convective updraft ( S ) and environment are zero. Note that a direct conversion term from environment to updraft without lateral mass exchange ( C ) should not be included in this tendency equation in order to conserve column-integrated horizontal momentum. If = qt , c , these diabatic sources are precipitation production within convective updraft ( Eqn.(4.50),(4.51) ) and evaporation of precipitation within environment. Following the formulation in CAM3 and CAM4, we assume that whenever convective precipitation ux exists, it is spread all over the grid. The resulting formulation of evaporation of convective precipitation within environment is Ct (Ae qt,e )S = Ae ke (1 Ue ) (
2107 2108 2109 2110

2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106

FR +

FS )

(4.63)

where FR and FS are grid-mean rain and snow uxes respectively in unit of [kg m2 s1] falling into the model layer from the top interface, and Ue is mean relative humidity within environment obtained using a mean saturation specic humidity that is a weighting average over water and ice, ke = 2 106 [(kg m2 s1 )1/2 s1 ] is evaporation eciency. We also consider snow 115

2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126

melting during fall and corresponding changes of c . This is a very simple formula bypassing the detailed vertical overlap structure of cumulus and stratus clouds. More rened treatment considering vertical cloud overlap will be done in future. Vertical transport of horizontal momentum by convective updraft does not change columnmean horizontal momentum. However, it will change column-mean kinetic energy ( KE ) of the mean wind. In reality, this KE change will be eventually converted into internal heat energy ( or potential energy, PE ). In CAM5, we require conservation of column-mean total energy, PE+KE. In order to satisfy this constraint, we add kinetic energy dissipation heating into c following Boville and Bretherton [2003b]. Similar treatment was made in the moist turbulence scheme. In CAM5, input state variables passed into individual physical schemes is not the grid-mean value including cumulus updraft contribution ( = A + (1 A) e ) but the environmental e ). In order to conserve column-integrated grid-mean mean value without cumulus portion ( energy, we print out Ae e /t instead of e /t from our shallow convection scheme. Under the approximation of very small updraft fractional area ( A 0 and Ae 1 ), it is e /t e /t. In Eqn.(4.63), we also approximate Ae 1. Ae

2127

4.3.8

Grid-Mean Tendency of Non-Conservative Scalars

2128 2129 2130 2131 2132

In contrast to the conservative scalars, we use the following explicit detrainmnet and compensating subsidence tendency form to compute the tendency of non-conservative scalars. We rst compute the tendencies of cloud condensates, and then the tendencies of water vapor ( qv ) and dry static energy ( s ) are extracted from them. Ct (Ae ql,e ) = g (M Mpen ) C ql,e p + g M (l ql,e ) + g Mpen (ql,pen ql,e ) (4.64) q Ct (Ae qi,e ) = g (M Mpen ) C qi,e p + g M (i qi,e ) + g Mpen (qi,pen qi,e ) (4.65) q Ct (Ae qv,e ) = Ct (Ae qt,e ) Ct (Ae ql,e ) Ct (Ae qi,e ) Ct (Ae se ) = Ct (Ae sc,e ) + Lv Ct (Ae ql,e ) + Ls Ct (Ae qi,e ) (4.66) (4.67)

2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145

where condensate static energy sc = Cp c +g z and the rst term on the R.H.S in Eqn.(4.64) and (4.65) is tendency associated with compensating subsidence and upwelling of environmental condensate, and the second and third terms are tendencies due to condensate detrainment from convective updraft and penetrative entrainment masses. If M Mpen > 0 ( M Mpen < 0 ), downward ( upward ) diencing between upper ( lower ) and current layers is used in computing compensating subsidence ( upwelling ) tendency. Any convective updraft condensate detrained into the layers above the LNB are assumed to move down into the layer just below LNB by negative buoyancy and be detrained there. That is, the second term on the R.H.S. in Eqn.(4.64) and (4.65) is zero in the overshooting zone. Similarly, all the penetratively entrained condensate are detrained into the layer just below LNB. That is, the third term on the R.H.S. in Eqn.(4.64) and (4.65) is non-zero only in the layer just below LNB. If environmental condensate is displaced vertically by compensating subsidence/upwelling, phase change should occur due to compression heating/expansion cooling. Ideally, this phase 116

2146 2147 2148 2149 2150 2151 2152 2153 2154

change should be treated within convection scheme but our current scheme does not. However, this phase change of displaced condensate will be treated by separate stratiform macromicrophysics schemes later. The tendencies of cloud droplet number concentration ( nl,e , ni,e ) by compensationg sub sidence/upwelling are treated in a similar way as the tendencies of cloud condensate mass. However, because CAM5s cumulus microphysics is the 1st moment scheme, we dont have any information on the droplet number concentration within cumulus updraft ( nl , ni ). We assume that the eective droplet radius of detrained shallow ( deep ) convective condensate is 8 ( 10 ) and 25 ( 50 ) [m] for liquid and ice respectively.

117

2155

4.4

Deep Convection

2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170

The process of deep convection is treated with a parameterization scheme developed by Zhang and McFarlane [1995] and modied with the addition of convective momentum transports by Richter and Rasch [2008] and a modied dilute plume calculation following Raymond and Blyth [1986, 1992]. The scheme is based on a plume ensemble approach where it is assumed that an ensemble of convective scale updrafts (and the associated saturated downdrafts) may exist whenever the atmosphere is conditionally unstable in the lower troposphere. The updraft ensemble is comprised of plumes suciently buoyant so as to penetrate the unstable layer, where all plumes have the same upward mass ux at the bottom of the convective layer. Moist convection occurs only when there is convective available potential energy (CAPE) for which parcel ascent from the sub-cloud layer acts to destroy the CAPE at an exponential rate using a specied adjustment time scale. For the convenience of the reader we will review some aspects of the formulation, but refer the interested reader to Zhang and McFarlane [1995] for additional detail, including behavioral characteristics of the parameterization scheme. Evaporation of convective precipitation is computed following the procedure described in section 4.5. The large-scale budget equations distinguish between a cloud and sub-cloud layer where temperature and moisture response to convection in the cloud layer is written in terms of bulk convective uxes as cp T t q t = = 1 (Mu Su + Md Sd Mc S) + L(C E) z 1 (Mu qu + Md qd Mc q) + E C , z (4.68) (4.69)

cu

cu

for z zb , where zb is the height of the cloud base. For zs < z < zb , where zs is the surface height, the sub-cloud layer response is written as cp T t q t = = 1 (Mb [S(zb ) Su (zb )] + Md [S(zb ) Sd (zb )]) zb zs 1 (Mb [q(zb ) qu (zb )] + Md [q(zb ) qd (zb )]) , zb zs (4.70) (4.71)

2171 2172 2173 2174

where the net vertical mass ux in the convective region, Mc , is comprised of upward, Mu , and downward, Md , components, C and E are the large-scale condensation and evaporation rates, S, Su , Sd , q, qu , qd , are the corresponding values of the dry static energy and specic humidity, and Mb is the cloud base mass ux.

2175

4.4.1

Updraft Ensemble

The updraft ensemble is represented as a collection of entraining plumes, each with a characteristic fractional entrainment rate . The moist static energy in each plume hc is given by hc = (h hc ), z 118 zb < z < zD . (4.72)

Mass carried upward by the plumes is detrained into the environment in a thin layer at the top of the plume, zD , where the detrained air is assumed to have the same thermal properties as in the environment (Sc = S). Plumes with smaller penetrate to larger zD . The entrainment rate D for the plume which detrains at height z is then determined by solving (4.72), with lower boundary condition hc (zb ) = hb : hc = D (h hb ) D (hc hb ) (z zb ) (4.73) (4.74) (4.75)

(hc hb ) D (hc hb ) = D (h hb ) (z zb ) (hc hb )eD (zzb ) = D (h hb )eD (zzb ) (z zb )

z

(hc hb )eD (zzb ) =

zb

D (h hb )eD (z zb ) dz
z zb

(4.76) (4.77)

(hc hb ) = D

(h hb )eD (z z) dz .

Since the plume is saturated, the detraining air must have hc = h , so that
z

(hb h ) = D
2176

zb

(hb h)eD (z z) dz .

(4.78)

Then, D is determined by solving (4.78) iteratively at each z. The top of the shallowest of the convective plumes, z0 is assumed to be no lower than the mid-tropospheric minimum in saturated moist static energy, h , ensuring that the cloud top detrainment is conned to the conditionally stable portion of the atmospheric column. All condensation is assumed to occur within the updraft plumes, so that C = Cu . Each plume is assumed to have the same value for the cloud base mass ux Mb , which is specied below. The vertical distribution of the cloud updraft mass ux is given by
D

Mu = Mb
0

eD (zzb ) 1 1 (zzb ) e d = Mb , 0 0 (z zb )

(4.79)

where 0 is the maximum detrainment rate, which occurs for the plume detraining at height z0 , and D is the entrainment rate for the updraft that detrains at height z. Detrainment is conned to regions where D decreases with height, so that the total detrainment Du = 0 for z < z0 . Above z0 , Mb D Du = . (4.80) 0 z The total entrainment rate is then just given by the change in mass ux and the total detrainment, Mu Eu = Du . (4.81) z

119

The updraft budget equations for dry static energy, water vapor mixing ratio, moist static energy, and cloud liquid water, , are: (Mu Su ) z (Mu qu ) z (Mu hu ) z (Mu ) z = (Eu Du ) S + LCu = Eu q Du q + Cu = Eu h Du h = Du d + Cu Ru , (4.82) (4.83) (4.84) (4.85)

where (4.84) is formed from (4.82) and (4.83) and detraining air has been assumed to be saturated (q = q and h = h ). It is also assumed that the liquid content of the detrained air is the same as the ensemble mean cloud water (d = ). The conversion from cloud water to rain water is given by Ru = c0 Mu , (4.86)
2177

following Lord et al. [1982], with c0 = 2 103 m1 . Since Mu , Eu and Du are given by (4.79-4.81), and h and h are environmental proles, (4.84) can be solved for hu , given a lower boundary condition. The lower boundary condition is obtained by adding a 0.5 K temperature perturbation to the dry (and moist) static energy at cloud base, or hu = h + cp 0.5 at z = zb . Below the lifting condensation level (LCL), Su and qu are given by (4.82) and (4.83). Above the LCL, qu is reduced by condensation and Su is increased by the latent heat of vaporization. In order to obtain to obtain a saturated updraft at the temperature implied by Su , we dene T as the temperature perturbation in the updraft, then: hu = Su + Lqu Su = S + cp T dq qu = q + T . dT Substituting (4.88) and (4.89) into (4.87), hu = S + Lq + cp 1 + = h + cp (1 + ) T L dq cp dT 1 hu h T = . cp 1 + The required updraft quantities are then Su = S + qu hu h 1+ hu h . = q + L 1+ 120 (4.94) (4.95) L dq cp dT T (4.90) (4.91) (4.92) (4.93) (4.87) (4.88) (4.89)

2178 2179

With Su given by (4.94), (4.82) can be solved for Cu , then (4.85) and (4.86) can be solved for and Ru . The expressions above require both the saturation specic humidity to be q = e , p e e < p , (4.96)

where e is the saturation vapor pressure, and its dependence on temperature (in order to maintain saturation as the temperature varies) to be dq de e d(p e ) = dT p e dT (p e )2 dT 1 de = 1+ pe pe dT q de 1+ . = pe e dT (4.97) (4.98) (4.99)

The deep convection scheme does not use the same approximation for the saturation vapor pressure e as is used in the rest of the model. Instead, e = c1 exp c2 (T Tf ) (T Tf + c3 ) , (4.100)

where c1 = 6.112, c2 = 17.67, c3 = 243.5 K and Tf = 273.16 K is the freezing point. For this approximation, d de = e dT dT c2 (T Tf ) (T Tf + c3 ) c2 c2 (T Tf ) = e (T Tf + c3 ) (T Tf + c3 )2 c2 c3 = e (T Tf + c3 )2 q c2 c3 = q 1 + . e (T Tf + c3 )2 q cp dq = = q 1 + dT L
2180 2181

(4.101) (4.102) (4.103) (4.104)

dq dT

We note that the expression for in the code gives L . RT 2 (4.105)

The expressions for dq /dT in (4.104) and (4.105) are not identical. Also, T Tf + c3 = T and c2 c3 = L/R.

2182

4.4.2

Downdraft Ensemble

Downdrafts are assumed to exist whenever there is precipitation production in the updraft ensemble where the downdrafts start at or below the bottom of the updraft detrainment layer. 121

Detrainment from the downdrafts is conned to the sub-cloud layer, where all downdrafts have the same mass ux at the top of the downdraft region. Accordingly, the ensemble downdraft mass ux takes a similar form to (4.79) but includes a proportionality factor to ensure that the downdraft strength is physically consistent with precipitation availability. This coecient takes the form P = , (4.106) P + Ed
2183 2184 2185 2186

where P is the total precipitation in the convective layer and Ed is the rain water evaporation required to maintain the downdraft in a saturated state. This formalism ensures that the downdraft mass ux vanishes in the absence of precipitation, and that evaporation cannot exceed some fraction, , of the precipitation, where = 0.2.

2187

4.4.3

Closure

The parameterization is closed, i.e., the cloud base mass uxes are determined, as a function of the rate at which the cumulus consume convective available potential energy (CAPE). Since the large-scale temperature and moisture changes in both the cloud and sub-cloud layer are linearly proportional to the cloud base updraft mass ux (e.g. see eq. 4.68 4.71), the CAPE change due to convective activity can be written as A t = Mb F , (4.107)

cu

where F is the CAPE consumption rate per unit cloud base mass ux. The closure condition is that the CAPE is consumed at an exponential rate by cumulus convection with characteristic adjustment time scale = 7200 s: A Mb = . (4.108) F
2188

4.4.4

Numerical Approximations

The quantities Mu,d , , Su,d, qu,d , hu,d are dened on layer interfaces, while Du , Cu , Ru are dened on layer midpoints. S, q, h, are required on both midpoints and interfaces and the interface values k are determined from the midpoint values k as k = log k1 k k1 k . k1 k (4.109)

All of the dierencing within the deep convection is in height coordinates. The dierences are naturally taken as k k+ = k , (4.110) z z z k+ where k and k+ represent values on the upper and lower interfaces, respectively for layer k. The convention elsewhere in this note (and elsewhere in the code) is k = k+ k . Therefore, we avoid using the compact k notation, except for height, and dene dk z z k z k+ = k z , 122 (4.111)

2189 2190 2191 2192

so that dk z corresponds to the variable dz(k) in the deep convection code. Although dierences are in height coordinates, the equations are cast in ux form and the tendencies are computed in units kg m3 s1 . The expected units are recovered at the end by multiplying by gz/p. The environmental proles at midpoints are Sk hk hk q k ek k cp T k + gz k S k + Lq k S k + Lq k ek /(pk ek ) c2 (T k Tf ) = c1 exp (T k Tf + c3 ) L2 q k = q k 1 + . cp RT k 2 = = = = (4.112) (4.113) (4.114) (4.115) (4.116) (4.117)

The environmental proles at interfaces of S, q, q , and are determined using (4.109) if | k1 k | is large enough. However, there are inconsistencies in what happens if | k1 k | is not large enough. For S and q the condition is k = ( k1 + k )/2, For q and the condition is k = k , | k1 k | 106 . (4.119) | k1 k | 106 . max( k1 k ) (4.118)

Interface values of h are not needed and interface values of h are given by hk = S k + Lq k . (4.120)

The unitless updraft mass ux (scaled by the inverse of the cloud base mass ux) is given by dierencing (4.79) as
k Mu =

1 0 (z k zb )

eD (z

k z

b)

(4.121)

M with the boundary condition that Mu + = 1. The entrainment and detrainment are calculated using

mk = u
k Eu k Du

1 (z k

0 zb ) k k+ mu Mu = dk z k k mu Mu . = dk z

eD

k+1

(z k zb )

(4.122) (4.123) (4.124)

2193

k Note that Mu and mk dier only by the value of D . u

123

The updraft moist static energy is determined by dierencing (4.84)

k k+ Mu hk Mu hk+ u u k k = Eu hk Du hk (4.125) kz d 1 k+ k k hk = k Mu hk+ + dk z Eu hk Du hk , (4.126) u u Mu with hM = hM + cp /2, where M is the layer of maximum h. u Once hu is determined, the lifting condensation level is found by dierencing (4.82) and (4.83) similarly to (4.84): 1 k+ k+ k k k Mu Su + dk z Eu S k Du S k (4.127) Su = k Mu 1 k+ k+ k k k Mu qu + dk z Eu q k Du q k . (4.128) qu = k Mu k k k The detrainment of Su is given by Du S k not by Du Su , since detrainment occurs at the environk mental value of S. The detrainment of qu is given by Du q k , even though the updraft is not yet saturated. The LCL will usually occur below z0 , the level at which detrainment begins, but this is not guaranteed. M M The lower boundary conditions, Su = S M + cp /2 and qu = q M , are determined from the rst midpoint values in the plume, rather than from the interface values of S and q. The solution of (4.127) and (4.128) continues upward until the updraft is saturated according to the condition k k qu > q (Tu ), 1 k k Tu = Su gz k . cp

2194

2195 2196 2197 2198

(4.129) (4.130)

The condensation (in units of m1 ) is determined by a centered dierencing of (4.82):

k k k+ k+ Mu Su Mu Su k k k = (Eu Du )S k + LCu kz d k Cu k k k+ k+ 1 Mu Su Mu Su k k (Eu Du )S k = kz L d

(4.131) . (4.132)

The rain production (in units of m1 ) and condensed liquid are then determined by dierencing (4.85) as k k+ Mu k Mu k+ k k k = Du k+ + Cu Ru , (4.133) kz d and (4.86) as k k Ru = c0 Mu k . (4.134) Then
k Mu k k k+ k k k Mu k = Mu k+ dk z Du k+ Cu + c0 Mu k k

(4.135) (4.136) . (4.137)

1 + c0 d z

+d z 1 k+ k k Mu k+ dk z Du k+ Cu = k (1 + c dk z) Mu 0

k+ Mu k+

k Du k+

k Cu

124

2199

4.4.5

Deep Convective Momentum Transports

Sub-grid scale Convective Momentum Transports (CMT) have ben added to the existing deep convection parameterization following Richter and Rasch [2008] and the methodology of Gregory et al. [1997b]. The sub-grid scale transport of momentum can be cast in the same manner as (4.69). Expressing the grid mean horizontal velocity vector, V , tendency due to deep convection transport following Kershaw and Gregory [1997] gives V t = 1 (Mu V u + Md V d Mc V ) , z (4.138)

cu

and neglecting the contribution from the environment the updraft and downdraft budget equation can similarly be written as (Mu V u ) = Eu V Du V u + P u G z (Md V d ) = Ed V + P d , G z (4.139) (4.140)

where P u and P d the updraft and downdraft pressure gradient sink terms parameterized from G G Gregory et al. [1997b] as P u = Cu Mu G Pd G
2200 2201 2202

V z V . = Cd Md z

(4.141) (4.142)

Cu and Cd are tunable parameters. In the CAM 5.0 implementation we use Cu = Cd = 0.4. The value of Cu and Cd control the strength of convective momentum transport. As these coeicients increase so do the pressure gradient terms, and convective momentum transport decreases.

2203

4.4.6

Deep Convective Tracer Transport

2204 2205 2206 2207 2208

The CAM 5.0 provides the ability to transport constituents via convection. The method used for constituent transport by deep convection is a modication of the formulation described in Zhang and McFarlane [1995]. We assume the updrafts and downdrafts are described by a steady state mass continuity equation for a bulk updraft or downdraft (Mx qx ) = Ex qe Dx qx . p (4.143)

2209 2210 2211 2212 2213 2214 2215

The subscript x is used to denote the updraft (u) or downdraft (d) quantity. Mx here is the mass ux in units of Pa/s dened at the layer interfaces, qx is the mixing ratio of the updraft or downdraft. qe is the mixing ratio of the quantity in the environment (that part of the grid volume not occupied by the up and downdrafts). Ex and Dx are the entrainment and detrainment rates (units of s1 ) for the up- and down-drafts. Updrafts are allowed to entrain or detrain in any layer. Downdrafts are assumed to entrain only, and all of the mass is assumed to be deposited into the surface layer. 125

2216 2217

Equation 4.143 is rst solved for up and downdraft mixing ratios qu and qd , assuming the environmental mixing ratio qe is the same as the gridbox averaged mixing ratio q . Given the up- and down-draft mixing ratios, the mass continuity equation used to solve for the gridbox averaged mixing ratio q is q = (Mu (qu q ) + Md (qd q )) . t p (4.144)

2218 2219

These equations are solved for in subroutine CONVTRAN. There are a few numerical details employed in CONVTRAN that are worth mentioning here as well. mixing quantities needed at interfaces are calculated using the geometric mean of the layer mean values. simple rst order upstream biased nite dierences are used to solve 4.143 and 4.144. uxes calculated at the interfaces are constrained so that the resulting mixing ratios are positive denite. This means that this parameterization is not suitable for moving mixing ratios of quantities meant to represent perturbations of a trace constituent about a mean value (in which case the quantity can meaningfully take on positive and negative mixing ratios). The algorithm can be modied in a straightforward fashion to remove this constraint, and provide meaningful transport of perturbation quantities if necessary. the reader is warned however that there are other places in the model code where similar modications are required because the model assumes that all mixing ratios should be positive denite quantities.

2220 2221

2222

2223 2224 2225 2226 2227 2228 2229 2230 2231

126

2232

4.5

Evaporation of convective precipitation

The CAM 5.0 employs a Sundqvist [1988] style evaporation of the convective precipitation as it makes its way to the surface. This scheme relates the rate at which raindrops evaporate to the local large-scale subsaturation, and the rate at which convective rainwater is made available to the subsaturated model layer 1/2 Erk = KE (1 RHk ) (Rrk ) . (4.145)

where RHk is the relative humidity at level k, Rrk denotes the total rainwater ux at level k (which can be dierent from the locally diagnosed rainwater ux from the convective parameterization, as will be shown below), the coecient KE takes the value 0.2 105 (kg m2 s1 )1/2 s1 , and the variable Erk has units of s1 . The evaporation rate Erk is used to determine a local change in qk and Tk , associated with an evaporative reduction of Rrk . Conceptually, the evaporation process is invoked after a vertical prole of Rrk has been evaluated. An evaporation rate is then computed for the uppermost level of the model for which Rrk = 0 using (4.145), where in this case Rrk Rrk . This rate is used to evaluate an evaporative reduction in Rrk which is then accumulated with the previously diagnosed rainwater ux in the layer below, Rrk+1 = Rrk pk g Erk + Rrk+1 . (4.146)

A local increase in the specic humidity qk and a local reduction of Tk are also calculated in accordance with the net evaporation qk = qk + Erk 2t , and Tk = Tk L cp (4.147)

Erk 2t .

(4.148)

The procedure, (4.145)-(4.148), is then successively repeated for each model level in a downward direction where the nal convective precipitation rate is that portion of the condensed rainwater in the column to survive the evaporation process Ps =
2233 2234 2235

RrK

pK g

ErK

/H2 0 .

(4.149)

In global annually averaged terms, this evaporation procedure produces a very small reduction in the convective precipitation rate where the evaporated condensate acts to moisten the middle and lower troposphere.

127

2236

4.6

Cloud Microphysics

2237 2238 2239 2240 2241

The base parameterization of stratiform cloud microphysics is described by Morrison and Gettelman [2008]. Details of the CAM implementation are described by Gettelman et al. [2008]. Modications to handle ice nucleation and ice supersaturation are described by Gettelman et al. [2010b]. The scheme seeks the following: A more exible, self-consistent, physically-based treatment of cloud physics. A reasonable level of simplicity and computational eciency. Treatment of both number concentration and mixing ratio of cloud particles to address indirect aerosol eects and cloud-aerosol interaction. Representation of precipitation number concentration, mass, and phase to better treat wet deposition and scavenging of aerosol and chemical species. The achievement of equivalent or better results relative to the CAM3 microphysics parameterization when compared to observations. The novel aspects of the scheme are an explicit representation of sub-grid cloud water distribution for calculation of the various microphysical process rates, and the diagnostic two-moment treatment of rain and snow.

2242

2243

2244 2245

2246 2247

2248 2249

2250 2251 2252

2253

4.6.1

Overview of the microphysics scheme

2254 2255 2256 2257 2258 2259 2260

The two-moment scheme is based loosely on the approach of Morrison et al. [2005]. This scheme predicts the number concentrations (Nc, Ni) and mixing ratios (qc, qi) of cloud droplets (subscript c) and cloud ice (subscript i). Hereafter, unless stated otherwise, the cloud variables Nc, Ni, qc, and qi represent grid-averaged values; prime variables represent mean in-cloud quantities (e.g., such that Nc = Fcld Nc, where Fcld is cloud fraction); and double prime variables represent local in-cloud quantities. The treatment of sub-grid cloud variability is detailed in section 2.1. The cloud droplet and ice size distributions are represented by gamma functions: (D) = N0 D expD (4.150)

2261 2262 2263 2264 2265

where Dis diameter, N0 is the intercept parameter, is the slope parameter, and = 1/ 2 1 is the spectra shape parameter; is the relative radius dispersion of the size distribution. The parameter for droplets is specied following Martin et al. [1994]. Their observations of maritime versus continental warm stratocumulus have been approximated by the following Nc relationship:
= 0.0005714Nc + 0.2714

(4.151)

2266 2267 2268

where Nc has units of cm3 . The upper limit for is 0.577, corresponding with aNc of 535 cm3 . Note that this expression is uncertain, especially when applied to cloud types other than those observed by Martin et al. [1994]. In the current version of the scheme, = 0 for cloud ice.

128

The spectral parameters N0 and are derived from the predicted N and q and specied : = N ( + 4) 6q ( + 1) N0 =
2269 2270 2271 2272 2273 2274

(1/3)

(4.152) (4.153)

N +1 ( + 1)

where is the Euler gamma function. Note that 4.152 and 4.153 assume spherical cloud particles with bulk density = 1000 kg m3 for droplets and = 500 kg m3 for cloud ice following Reisner et al. [1998]. The eective size for cloud ice needed by the radiative transfer scheme is obtained directly by dividing the third and second moments of the size distribution given by 4.150 and accounting for dierenceds in cloud ice density and that of pure ice. After rearranging terms, this yields de i = 3 i (4.154)

2275 2276 2277

where i = 917 kg m-2 is the bulk density of pure ice. Note that optical properties for cloud droplets are calculated using a lookup table from the N0 and parameters. The droplet eective radius, which is used for output purposes only, is given by re c = ( + 4) 2( + 3) (4.155)

2278 2279

The time evolution of q and N is determined by grid-scale advection, convective detrainment, turbulent diusion, and several microphysical processes: N 1 + [uN] = t N t N t N t N t N t N t N t het (4.156) +

+
nuc

+
evap

+
auto

+
acer

+
accs

+
hom

2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290

q q + t hom t cond evap auto acer accs het (4.157) where t is time, u is the 3D wind vector, is the air density, and D is the turbulent diffusion operator. The symbolic terms on the right hand side of 4.156 and 4.157 represent the grid-average microphysical source/sink terms for N and q. Note that the source/sink terms for q and N are considered separately for cloud water and ice (giving a total of four rate equations), but are generalized here using 4.156 and 4.157 for conciseness. These terms include activation of cloud condensation nuclei or deposition/condensation-freezing nucleation on ice nuclei to form droplets or cloud ice (subscript nuc; N only); ice multiplication via rime-splintering on snow (subscript mult); condensation/deposition (subscript cond; q only), evaporation/sublimation (subscript evap), autoconversion of cloud droplets and ice to form rain and snow (subscript auto), accretion of cloud droplets and ice by rain (subscript accr), accretion of cloud droplets and ice by snow (subscript accs), heterogeneous freezing of droplets to form ice (subscript het), q t + + + + + +

q 1 + [uq] = t

q t

q t

q t

q t

q t

+
mlt

129

2291 2292 2293 2294

homogeneous freezing of cloud droplets (subscript hom), melting (subscript mlt), ice multiplication (subsrcipt mult), sedimentation (subscript sed), and convective detrainment (subscript det). The formulations for these processes are detailed in section 3. Numerical aspects in solving 4.156 and 4.157 are detailed in section 4. Sub-grid cloud variability Sub-grid variability is considered for cloud water but neglected for cloud ice and precipitation at present; furthermore, we neglect sub-grid variability of droplet number concentration for simplicity. We assume that the PDF of in-cloud cloud water, P (qc ), follows a gamma distribution function based on observations of optical depth in marine boundary layer clouds [Barker, 1996; Barker et al., 1996; McFarlane and Klein, 1999]:
P (qc ) 1 qc expqc = ()

2295

2296 2297 2298 2299 2300

(4.158)

2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312

2 where = 1/ 2 ; 2 is the relative variance (i.e., variance divided by qc ); and = /qc (qc is the mean in-cloud cloud water mixing ratio). Note that this PDF is applied to all cloud types treated by the stratiform cloud scheme; the appropriateness of such a PDF for stratiform cloud types other than marine boundary layer clouds (e.g., deep frontal clouds) is uncertain given a lack of observations. Satellite retrievals described by Barker et al. [1996] suggest that > 1 in overcast conditions and 1 (corresponding to an exponential distribution) in broken stratocumulus. The model assumes a constant = 1 for simplicity. A major advantage of using gamma functions to represent sub-grid variability of cloud water is that the grid-average microphysical process rates can be derived in a straightforward manner y as follows. For any generic local microphysical process rate Mp = xqc , replacing qc with P (qc ) from 4.158 and integrating over the PDF yields a mean in-cloud process rate Mp = x

( + y) y q () y c

(4.159)

2313

Thus, each cloud water microphysical process rate in 4.156 and 4.157 is multiplied by a factor E= ( + y) () y (4.160)

2314

Diagnostic treatment of precipitation As described by Ghan and Easter [1992], diagnostic treatment of precipitation allows for a longer time step, since prognostic precipitation is constrained by the Courant criterion for sedimentation. Furthermore, the neglect of horizontal advection of precipitation in the diagnostic approach is reasonable given the large grid spacing ( 100 km) and long time step (15-40 min) of GCMs. A unique aspect of this scheme is the diagnostic treatment of both precipitation mixing ratio qp and number concentration Np . Considering only the vertical dimension, the grid-scale time rates of change of qp and Np are:

2315 2316 2317 2318 2319 2320 2321

130

qp 1 (Vq qp ) = + Sq t z 1 (VN Np ) Np = + SN t z
2322 2323

(4.161) (4.162)

where z is height, Vq and VN are the mass- and number-weighted terminal fallspeeds, respectively, and Sq and SN are the grid-mean source/sink terms for qp and Np , respectively: qp t qp t qp t qp t qp t qp t qp t qp t mult (4.163) +

Sq =

+
auto

+
accw

+
acci

+
het

+
hom

+
mlt

+
evap

2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334

Np t coll auto het hom mlt evap self (4.164) The symbolic terms on the right-hand sides of 4.163 and 4.164 are autoconversion (subscript auto), accretion of cloud water (subscript accw), accretion of cloud ice (subscript acci), heterogeneous freezing (subscript het), homogeneous freezing (subscript hom), melting (subscript mlt), ice multiplication via rime splintering (subsrcipt mult; qp only), evaporation (subscript evap), and self-collection (subscript self; collection of rain drops by other rain drops, or snow crystals by other snow crystals; Np only), and collection of rain by snow (subscript coll). Formulations for these processes are described in section 3. In the diagnostic treatment , (qp /t) =0 and (Np /t) =0 . This allows 4.161 and 4.162 to be expressed as a function of z only. The qp and Np are therefore determined by discretizing and numerically integrating 4.161 and 4.162 downward from the top of the model atmosphere following Ghan and Easter [1992]: SN = + + + + + + 1 a,k Vq,k qp,k = a,k+1 Vq,k+1qp,k+1 + [a,k Sq,k Zk + a,k+1 Sq,k+1 Zk+1] 2 (4.165)

Np t

Np t

Np t

Np t

Np t

Np t

2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345

1 a,k VN,k Np,k = a,k+1 VN,k+1Np,k+1 + [a,k SN,k Zk + a,k+1 SN,k+1Zk+1 ] (4.166) 2 where k is the vertical level (increasing with height, i.e., k + 1 is the next vertical level above k). Since Vq,k , Sq,k , VN,k , and SN,k depend on qp,k and Np,k , 4.165 and 4.166 must be solved by iteration or some other method. The approach of Ghan and Easter [1992] uses values of qp,k and Np,k from the previous time step as provisional estimates in order to calculate Vq,k , VN,k , Sp,k , and SN,k . Final values of qp,k and Np,k are calculated from these values of Vq,k , VN,k , Sq,k and SN,k using 4.165 and 4.166. Here we employ another method that obtains provisional values of qp,k and Np,k from 4.165 and 4.166 assuming Vq,k Vq,k+1 and VN,k VN,k+1. It is also assumed that all source/sink terms in Sq,k and SN,q can be approximated by the values at k + 1, except for the autoconversion, which can be obtained directly at the k level since it does not depend on qp,k or Np,k . If there is no precipitation ux from the level above, then the provisional qp.k and Np,k are calculated using autoconversion at the k level in Sq,k and SN,k ; Vq,k and VN,k are

131

2346 2347 2348 2349 2350 2351 2352 2353 2354

estimated assuming newly-formed rain and snow particles have fallspeeds of 0.45 m/s for rain and 0.36 m/s for snow. Rain and snow are considered separately, and both may occur simultaneously in supercooled conditions (hereafter subscript p for precipitation is replaced by subscripts r for rain and s for snow). The rain/snow particle size distributions are given by 4.150, with the shape parameter = 0, resulting in Marshall-Palmer (exponential) size distributions. The size distribution parameters and N0 are similarly given by 4.152 and 4.153 with = 0. The bulk particle density (parameter in 4.152) is = 1000 kg m3 for rain and = 100 kg m3 for snow following Reisner et al. [1998]. Cloud and precipitation particle terminal fallspeeds The mass- and number-weighted terminal fallspeeds for all cloud and precipitation species are obtained by integration over the particle size distributions with appropriate weighting by number concentration or mixing ratio:
0 a a0 0.54

2355

2356 2357 2358

aD b (D)dD =

VN =

a a0

0.54

a(1 + b + ) (4.167)

(D)dD

b ( + 1)
a a0 0.54

Vq =
2359 2360 2361 2362 2363 2364 2365

6 0

0.54 a aD b+3 (D)dD a0 3 D (D)dD 6 0

a(4 + b + ) (4.168)

b ( + 4)

where a0 is the reference air density at 850 mb and 0 C, a and b are empirical coecients in the diameter-fallspeed relationship V = aD b , where V is terminal fallspeed for an individual particle with diameter D. The air density correction factor is from Heymseld and Banseemer (2007). VN and Vq are limited to maximum values of 9.1 m/s for rain and 1.2 m/s for snow. The a and b coecients for each hydrometeor species are given in Table 2. Note that for cloud water fallspeeds, sub-grid variability of q is considered by appropriately multiplying the VN and Vq by the factor E given by 4.160. Ice Cloud Fraction Several modications have been made to the determination of diagnostic fractional cloudiness in the simulations. The ice and liquid cloud fractions are now calculated separately. Ice and liquid cloud can exist in the same grid box. Total cloud fraction, used for radiative transfer, is determined assuming maximum overlap between the two. The diagnostic ice cloud fraction closure is constructed using a total water formulation of the Slingo [1987a] scheme. There is an indirect dependence of prognostic cloud ice on the ice cloud fraction since the in-cloud ice content is used for all microphysical processes involving ice. The new formulation of ice cloud fraction (CFi ) is calculated using relative humidity (RH) based on total ice water mixing ratio, including the ice mass mixing ratio (qi ) and the vapor mixing ratio (qv ). The RH based on total ice water (RHti ) is then RHti = (qv + qi )/qsat where qsat is the saturation vapor mixing ratio over ice. Because this is for ice clouds only, we do not include ql (liquid mixing ratio). We have tested that the inclusion of ql does not substantially impact the scheme (since there is little liquid present in this regime). 132

2366

2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379

2380

2 Ice cloud fraction is then given by CFi = min(1, RHd ) where

RHd = max 0,
2381 2382 2383 2384 2385 2386 2387 2388

RHimax and RHimin are prescribed maximum and minimum threshold humidities with respect to ice, set at RHimax =1.1 and RHimin =0.8. These are adjustable parameters that reect assumptions about the variance of humidity in a grid box. The scheme is not very sensitive to RHimin . RHimax aects the total ice supersaturation and ice cloud fraction. With RHimax = 1 and qi = 0 the scheme reduces to the Slingo [1987a] scheme. RHti is preferred over RH in RHd because when qi increases due to vapor deposition, it reduces qv , and without any precipitation or sedimentation the decrease in RH would change diagnostic cloud fraction, whereas RHti is constant.

RHti RHimin RHimax RHimin

(4.169)

2389

4.6.2

Radiative Treatment of Ice

2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406

The simulations use a self consistent treatment of ice in the radiation code. The radiation code uses as input the prognostic eective diameter of ice from the cloud microphysics (give eq. # from above). Ice cloud optical properties are calculated based on the modied anomalous diraction approximation (MADA), described in Mitchell [2000, 2002] and Mitchell et al. [2006a]. The mass-weighted extinction (volume extinction coecient/ice water content) and the single scattering albedo, 0 , are evaluated using a look-up table. For solar wavelengths, the asymmetry parameter g is determined as a function of wavelength and ice particle size and shape as described in Mitchell et al. [1996a] and Nousiainen and McFarquhar [2004] for quasi-spherical ice crystals. For terrestrial wavelengths, g was determined following Yang et al. [2005]. An ice particle shape recipe was assumed when calculating these optical properties. The recipe is described in Mitchell et al. [2006b] based on mid-latitude cirrus cloud data from Lawson et al. [2006] and consists of 50% quasi-spherical and 30% irregular ice particles, and 20% bullet rosettes for the cloud ice (i.e. small crystal) component of the ice particle size distribution (PSD). Snow is also included in the radiation code, using the diagnosed mass and eective diameter of falling snow crystals (MG2008). For the snow component, the ice particle shape recipe was based on the crystal shape observations reported in Lawson et al. [2006] at -45C: 7% hexagonal columns, 50% bullet rosettes and 43% irregular ice particles.

2407

4.6.3

Formulations for the microphysical processes

2408

Activation of cloud droplets Activation of cloud droplets, occurs on a multi-modal lognormal aerosol size distribution based on the scheme of Abdul-Razzak and Ghan [2000a]. Activation of cloud droplets occurs if Nc decreases below the number of active cloud condensation nuclei diagnosed as a function of aerosol chemical and physical parameters, temperature, and vertical velocity (see Abdul-Razzak and Ghan [2000a]), and if liquid condensate is present. We use the existing Nc as a proxy for the number of aerosols previously activated as droplets since the actual number of activated aerosols is not tracked as a prognostic variable from time step to time step (for

2409 2410 2411 2412 2413 2414 2415

133

2416 2417 2418 2419 2420

coupling with prescribed aerosol scheme). This approach is similar to that of Lohmann et al. [1999]. Since local rather than grid-scale vertical velocity is needed for calculating droplet activation, a sub-grid vertical velocity wsub is derived from the square root of the Turbulent Kinetic Energy (TKE) following Morrison and Pinto [2005]: 2 T KE (4.170) 3 where TKE is dened using a steady state energy balance (eqn [17] and [28] in Bretherton and Park [2009b]) In regions with weak turbulent diusion, a minimum sub-grid vertical velocity of 10 cm/s is assumed. Some models use the value of w at cloud base to determine droplet activation in the cloud layer (e.g., Lohmann et al. [1999]); however, because of coarse vertical and horizontal resolution and diculty in dening the cloud base height in GCMs, we apply the wsub calculated for a given layer to the droplet activation for that layer. Note that the droplet number may locally exceed the number activated for a given level due to advection of Nc. Some models implicitly assume that the timescale for droplet activation over a cloud layer is equal to the model time step (e.g., Lohmann et al. [1999]), which could enhance sensitivity to the time step. This timescale can be thought of as the timescale for recirculation of air parcels to regions of droplet activation (i.e., cloud base), similar to the timescale for large eddy turnover; here, we assume an activation timescale of 20 min. wsub = Primary ice nucleation Ice crystal nucleation is based on Liu et al. [2007], which includes homogeneous freezing of sulfate competing with heterogeneous immersion freezing on mineral dust in ice clouds (with temperatures below -37 C) [Liu and Penner, 2005]. Because mineral dust at cirrus levels is very likely coated [Wiacek and Peter, 2009], deposition nucleation is not explicitly included in this work for pure ice clouds. Immersion freezing is treated for cirrus (pure ice), but not for mixed phase clouds. The relative eciency of immersion versus deposition nucleation in mixed phase clouds is an unsettled problem, and the omission of immersion freezing in mixed phase clouds may not be appropriate (but is implicitly included in the deposition/condensation nucleation: see below). Deposition nucleation may act at temperatures lower than immersion nucleation (i.e. T<-25C) [Field et al., 2006], and immersion nucleation has been inferred to dominate in mixed phase clouds [Ansmann et al., 2008, 2009; Hoose and Kristjansson, 2010]. We have not treated immersion freezing on soot because while Liu and Penner [2005] assumed it was an ecient mechanism for ice nucleation, more recent studies [Krcher et al., 2007] indicate it is a still highly uncertain. In the mixed phase cloud regime (-37<T<0C), deposition/condensation nucleation is considered based on Meyers et al. [1992], with a constant nucleation rate for T<-20 C. The Meyers et al. [1992] parameterization is assumed to treat deposition/condensation on dust in the mixed phase. Since it is based on observations taken at water saturation, it should include all important ice nucleation mechanisms (such as the immersion and deposition nucleation discussed above) except contact nucleation, though we cannot distinguish all the specic processes. Meyers et al. [1992] has been shown to produce too many ice nuclei during the Mixed Phase 134

2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433

2434

2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455

2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478

Arctic Clouds Experiment (MPACE) by Prenni et al. [2007]. Contact nucleation by mineral dust is included based on Young [1974] and related to the coarse mode dust number. It acts in the mixed phase where liquid droplets are present and and includes Brownian diusion as well as phoretic forces. Hallet-Mossop secondary ice production due to accretion of drops by snow is included following Cotton et al. [1986]. In the Liu and Penner [2005] scheme, the number of ice crystals nucleated is a function of temperature, humidity, sulfate, dust and updraft velocity, derived from tting the results from cloud parcel model experiments. A threshold RHw for homogeneous nucleation was tted as a function of temperature and updraft velocity (see Liu et al. [2007], equation 6). For driving the parameterization, the sub-grid velocity for ice (wsub ) is derived following ewuation 4.170. A minimum of 0.2 m s1 is set for ice nucleation. It is also implicitly assumed that there is some variation in humidity over the grid box. For purposes of ice nucleation, nucleation rates for a grid box are estimated based on the most humid portion of the grid-box. This is assumed to be the grid box average humidity plus a xed value (20% RH). This implies that the local threshold supersaturation for ice nucleation will be reached at a grid box mean value 20% lower than the RH process threshold value. This represents another gross assumption about the RH variability in a model grid box and is an adjustable parameter in the scheme. In the baseline case, sulfate for homogeneous freezing is taken as the portion of the Aitken mode particles with radii greater than 0.1 microns, and was chosen to better reproduce observations (this too can be adjusted to alter the balance of homogeneous freezing). The size represents the large tail of the Aitken mode. In the upper troposphere there is little sulfate in the accumulation mode (it falls out), and almost all sulfate is in the Aitken mode. Deposition/sublimation of ice Several cases are treated below that involve ice deposition in ice-only clouds or mixed-phase clouds in which all liquid water is depleted within the time step. Case [1] Ice only clouds in which qv > qvi where qv is the grid mean water vapor mixing ratio and qvi is the local vapor mixing ratio at ice saturation (qsat ). Case [2] is the same as case [1] (qv > qvi ) but there is existing liquid water depleted by the Bergeron-Findeisen process (ber). Case [3], liquid water is depleted by the Bergeron-Findeisen process and the local liquid is less than local ice saturation (qv qvi ). In Case [4] qv < qvi so sublimation of ice occurs. Case [1]: If the ice cloud fraction is larger than the liquid cloud fraction (including grid cells with ice but no liquid water), or if all new and existing liquid water in mixed-phase clouds is depleted via the Bergeron-Findeisen process within the time step, then vapor depositional ice growth occurs at the expense of water vapor. In the case of a grid cell where ice cloud fraction exceeds liquid cloud fraction, vapor deposition in the pure ice cloud portion of the cell is calculated similarly to eq. [21] in MG08: qi t =
dep

2479

2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492

(qv qvi ) , qv > qvi p

(4.171)

2493 2494 2495

where p = 1 +

Ls dqvi cp dT

is the psychrometric correction to account for the release of latent

heat, Ls is the latent heat of sublimation, cp is the specic heat at constant pressure, dqvi is the dT change of ice saturation vapor pressure with temperature, and is the supersaturation relaxation 135

2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509

timescale associated with ice deposition given by eq. [22] in MG08 (a function of ice crystal surface area and the diusivity of water vapor in air). The assumption for pure ice clouds is that the in-cloud vapor mixing ratio for deposition is equal to the grid-mean value. The same assumption is used in Liu et al. [2007], and while it is uncertain, it is the most straightforward. Thus we do not consider sub-grid variability of water vapor for calculating vapor deposition in pure ice-clouds. The form of the deposition rate in equation 4.171 diers from that used by Rotstayn et al. [2000] and Liu et al. [2007] because they considered the increase in ice mixing ratio qi due to vapor deposition during the time step, and formulated an implicit solution based on this consideration (see eq. [6] in Rotstayn et al. [2000]). However, these studies did not consider sinks for the ice due to processes such as sedimentation and conversion to precipitation when formulating their implicit solution; these sink terms may partially (or completely) balance the source for the ice due to vapor deposition. Thus, we use a simple explicit forward-in-time solution that does not consider changes of qi within the microphysics time step. Case [2]: When all new and existing liquid water is depleted via the Bergeron-Findeisen process (ber) within the time step, the vapor deposition rate is given by a weighted average of the values for growth in mixed phase conditions prior to the depletion of liquid water (rst term on the right hand side) and in pure ice clouds after depletion (second term on the right hand side): 1 (qv qvi ) qc qi qc qi = , qv > qvi (4.172) + 1 t dep t t t ber p where qc is the sum of existing and new liquid condensate mixing ratio, t is the model time step, qi ber is the ice deposition rate in the presence of liquid water (i.e., assuming vapor t mixing ratio is equal to the value at liquid saturation) as described above, and qv is an average of the grid-mean vapor mixing ratio and the value at liquid saturation. Case [3]: If qv qvi then it is assumed that no additional ice deposition occurs after depletion of the liquid water. The deposition rate in this instance is given by: qi t =
dep

2510 2511 2512 2513 2514 2515

qc , qv qvi t

(4.173)

2516 2517

Case [4]: Sublimation of pure ice cloud occurs when the grid-mean water vapor mixing ratio is less than value at ice saturation. In this case the sublimation rate of ice is given by: qi t =
sub

(qv qvi ) , qv < qvi p

(4.174)

2518 2519 2520 2521 2522 2523 2524 2525 2526

Again, the use of grid-mean vapor mixing ratio in equation 4.174 follows the assumption of Liu et al. [2007] that the in-cloud qv is equal to the grid box mean in pure ice clouds. Gridmean deposition and sublimation rates are given by the in-cloud values for pure ice or mixedphase clouds described above, multiplied by the appropriate ice or mixed-phase cloud fraction. Finally, ice deposition and sublimation are limited to prevent the grid-mean mixing ratio from falling below the value for ice saturation in the case of deposition and above this value in the case of sublimation. Cloud water condensation and evaporation are given by the bulk closure scheme within the cloud macrophysics scheme, and therefore not described here. 136

2527

Conversion of cloud water to rain Autoconversion of cloud droplets and accretion of cloud droplets by rain is given by a version of the Khairoutdinov and Kogan [2000] scheme that is modied here to account for sub-grid variability of cloud water within the cloudy part of the grid cell as described previously in section 2.1. Note that the Khairoutdinov and Kogan scheme was originally developed for boundary layer stratocumulus, but is applied here to all stratiform cloud types. The grid-mean autoconversion and accretion rates are found by replacing the qc in Eqs. (29) and (33) of Khairoutdinov and Kogan [2000] with P (qc ) given by equation 4.158 here, integrating the resulting expressions over the cloud water PDF, and multiplying by the cloud fraction. This yields qc t = Fcld ( + 2.47) 2.47 1.79 1350qc Nc () 2.47 ( + 1.15) 67(qc qr )1.15 () 1.15 (4.175) (4.176)

2528 2529 2530 2531 2532 2533 2534 2535 2536

auto

qc t
2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551

accr

= Fcld

The changes in qr due to autoconversion and accretion are given by (qr /t)auto = (qc /t)auto and (qr /t)accr = (qc /t)accr . The changes in Nc and Nr due to autoconversion and accretion (Nc /t)auto , (Nr /t)auto , (Nc /t)accr , are derived from Eqs. (32) and (35) in Khairoutdinov and Kogan [2000]. Since accretion is nearly linear with respect to qc , subgrid variability of cloud water is much less important for accretion than it is for autoconversion. Note that in the presence of a precipitation ux into the layer from above, new drizzle drops formed by cloud droplet autoconversion would be accreted rapidly by existing precipitation particles (rain or snow) given collection eciencies near unity for collision of drizzle with rain or snow (e.g., Pruppacher and Klett [1997]). This may be especially important in models with low vertical resolution, since they cannot resolve the rapid growth of precipitation that occurs over distances much less than the vertical grid spacing. Thus, if the rain or snow mixing ratio in the next level above is greater than 10-6 g kg-1, we assume that autoconversion produces an increase in rain mixing ratio but not number concentration (since the newly-formed drops are assumed to be rapidly accreted by the existing precipitation). Otherwise, autoconversion results in a source of both rain mixing ratio and number concentration. Conversion of cloud ice to snow The autoconversion of cloud ice to form snow is calculated by integration of the cloud ice mass- and number-weighted size distributions greater than some specied threshold size, and transferring the resulting mixing ratio and number into the snow category over some specied timescale, similar to Ferrier [1994]. The grid-scale changes in qi and Ni due to autoconversion are qi t = F
3 2 i N0i Dcs 3Dcs 6Dcs 6D + 2 + 3 + 4 expi Dcs 6auto i i i i

2552

2553 2554 2555 2556 2557

(4.177) (4.178)

auto

Ni t

auto

= F

N0i expi Dcs i auto

137

2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568

where Dcs = 200 m is the threshold size separating cloud ice from snow, i is the bulk density of cloud ice, and auto = 3 min is the assumed autoconversion timescale. Note that this formulation assumes the shape parameter = 0 for the cloud ice size distribution; dierent formulation must be used for other values of . The changes in qs and Ns due to autoconversion are given by (qs /t)auto = (qi /t)auto and (Ns /t)auto = (Ni /t)auto . Accretion of qi and Ni by snow (qi /t)accs , (Ni /t)accs , (qs /t)acci , and (qs /t)acci = (qi /t)accs , are given by the continuous collection equation following Lin et al. [1983], which assumes that the fallspeed of snow cloud ice fallspeed. The collection eciency for collisions between cloud ice and snow is 0.1 following Reisner et al. [1998]. Newly- formed snow particles formed by cloud ice autoconversion are not assumed to be rapidly accreted by existing snowakes, given aggregation eciencies typically much less than unity (e.g., Field et al. [2007]). Other collection processes The accretion of qc and Nc by snow (qc /t)accs , (Nc /t)accs , and (qs /t)accw = (qc /t)accs are given by the continuous collection equation. The collection eciency for droplet-snow collisions is a function of the Stokes number following Thompson et al. [2004] and thus depends on droplet size. Self-collection of snow, (Ns /t)self follows Reisner et al. [1998] using an assumed collection eciency of 0.1. Self-collection of rain(Nr /t)self follows Beheng [1994]. Collisions between rain and cloud ice, cloud droplets and cloud ice, and self-collection of cloud ice are neglected for simplicity. Collection of qr and Nr by snow in subfreezing conditions, (qr /t)coll = (qs /t)coll and (Nr /t)coll , is given by Ikawa and Saito [1990] assuming collection eciency of unity. Freezing of cloud droplets and rain and ice multiplication Heterogeneous freezing of cloud droplets and rain to form cloud ice and snow, respectively, occurs by immersion freezing following Bigg [1953], which has been utilized in previous microphysics schemes (e.g., Reisner et al. [1998], see Eq. A.22, A.55, A.56; Morrison et al. [2005]; Thompson et al. [2008]). Here the freezing rates are integrated over the mass- and numberweighted cloud droplet and rain size distributions and the impact of sub-grid cloud water variability is included as described previously. Homogeneous freezing of cloud droplets to form cloud ice occurs instantaneously at -40C. All rain is assumed to freeze instantaneously at -5 C. Contact freezing of cloud droplets by mineral dust is included based on Young [1974] and related to the coarse mode dust number. It acts in the mixed phase where liquid droplets are present and includes Brownian diusion as well as phoretic forces. Hallet-Mossop ice multiplication (secondary ice production) due to accretion of drops by snow is included following Cotton et al. [1986]. This represents a sink term for snow mixing ratio and source term for cloud ice mixing ratio and number concentration. Melting of cloud ice and snow For simplicity, detailed formulations for heat transfer during melting of ice and snow are not included. Melting of cloud ice occurs instantaneously at 0 C. Melting of snow occurs instantaneously at +2 C. We have tested the sensitivity of both single- column and global results

2569

2570 2571 2572 2573 2574 2575 2576 2577 2578

2579

2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592

2593

2594 2595 2596

138

2597 2598

to changing the specied snow melting temperature from +2 to 0 C and found no signicant changes. Evaporation/sublimation of precipitation Evaporation of rain and sublimation of snow, (qs /t)evap and (qr /t)evap , are given by diffusional mass balance in subsaturated conditions Lin et al. [1983], including ventilation eects. Evaporation of precipitation occurs within the region of the grid cell containing precipitation but outside of the cloudy region. The fraction of the grid cell with evaporation of precipitation is therefore , where Fpre is the precipitation fraction. Fpre is calculated assuming maximum cloud overlap between vertical levels, and neglecting tilting of precipitation shafts due to wind shear (Fpre = Fcld at cloud top). The out-of-cloud water vapor mixing ratio is given by qclr = qv Fcld qs (T ) , Fcld < 1 1 Fcld (4.179)

2599

2600 2601 2602 2603 2604 2605 2606

2607 2608 2609 2610 2611

where qs (T ) is the in-cloud water vapor mixing ratio after bulk condensation/evaporation of cloud water and ice as described previously. As in the older CAM3 microphysics parameterization, condensation/deposition onto rain/snow is neglected. Following Morrison et al. [2005], the evaporation/sublimation of Nr and Ns , (Nr /t)evap and (Ns /t)evap , is proportional to the reduction of qr and qs during evaporation/sublimation. Sedimentation of cloud water and ice The time rates of change of q and N for cloud water and cloud ice due to sedimentation, (qc /t)sed , (qi /t)sed , (Nc /t)sed , and (Ni /t)sed , are calculated with a rst-order forwardin-time-backward-in-space scheme. Numerical stability for cloud water and ice sedimentation is ensured by sub-stepping the time step, although these numerical stability issues are insignicant for cloud water and ice because of the low terminal fallspeeds ( 1 m/s). We assume that the sedimentation of cloud water and ice results in evaporation/sublimation when the cloud fraction at the level above is larger than the cloud fraction at the given level (i.e., a sedimentation ux from cloudy into clear regions), with the evaporation/condensate rate proportional to the dierence in cloud fraction between the levels. Convective detrainment of cloud water and ice The ratio of ice to total cloud condensate detrained from the convective parameterizations, Fdet, is a linear function of temperature between -40 C and -10 C; Fdet = 1 at T < -40 C, and Fdet = 0 at T > -10 C. Detrainment of number concentration is calculated by assuming a mean volume radius of 8 and 32 micron for droplets and cloud ice, respectively. Numerical considerations To ensure conservation of both q and N for each species, the magnitudes of the various sink terms are reduced if the provisional q and N are negative after stepping forward in time. This approach ensures critical water and energy balances in the model, and is similar to the approach employed in other bulk microphysics schemes (e.g., Reisner et al. [1998]. Inconsistencies are possible

2612

2613 2614 2615 2616 2617 2618 2619 2620 2621

2622

2623 2624 2625 2626

2627

2628 2629 2630 2631

139

2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650

because of the separate treatments for N and q, potentially leading to unrealistic mean cloud and precipitation particle sizes. For consistency, N is adjusted if necessary so that mean (numberweighted) particle diameter ( ) remains within a specied range of values for each species. Limiting to a maximum mean diameter can be thought of as an implicit parameterization of particle breakup. For the diagnostic precipitation, the source terms for q and N at a given vertical level are adjusted if necessary to ensure that the vertical integrals of the source terms (from that level to the model top) are positive. In other words, we ensure that at any given level, there isnt more precipitation removed (both in terms of mixing ratio and number concentration) than is available falling from above (this is also the case in the absence of any sources/sinks at that level). This check and possible adjustment of the precipitation and cloud water also ensures conservation of the total water and energy. Our simple adjustment procedure to ensure conservation could potentially result in sensitivity to time step, although as described in section 3, time truncation errors are minimized with appropriate sub-stepping. Melting rates of cloud ice and snow are limited so that the temperature of the layer does not decrease below the melting point (i.e., in this instance an amount of cloud ice or snow is melted so that the temperature after melting is equal to the melting point). A similar approach is applied to ensure that homogeneous freezing does increase the temperature above homogeneous freezing threshold.

140

2651

4.7

Cloud Macrophysics

2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678

Cloud macrophysics is a suite of physical processes that computes (1) cloud fractions in each layer, (2) horizontal and vertical overlapping structures of clouds, (3) net conversion rates of water vapor into cloud condensates. Cloud macrophysics is a process unique for GCM that handles partial cloud fraction. In case of cloud resolving model, for example, cloud fraction in each layer is either 0 or 1, and so there is no need to use special treatment for cloud overlap and partial condensation. Along with convection scheme, correct setting of cloud macrophysics is essential for developing a seamless GCM across the various sizes of horizontal GCM grid. Cloud macrophysics sets a stage for cloud droplet activation and nucleation, cloud microphysics ( i.e., processes controlling conversion from sustained to falling hydrometeors ), wet scavenging of aerosols, radiative transfer, and moist turbulent processes. Cloud macrophysics in CAM3/CAM4 ( cloud macrophysics in CAM3 is nearly identical to the cloud macrophysics in CAM4 ) was constructed to be compatible with and to some degrees to compensate for the incomplete CAM3/CAM4 physics package. For example, (1) without a need to do explicit droplet nucleation and activation processes due to the prescribed cloud droplet radius, CAM3/CAM4 simply assume zero supersaturation within ice stratus, (2) without the information of realistic in-cumulus condensate from shallow and deep convection schemes, CAM3/CAM4 assumes that in-cumulus condensate is identical to in-stratus condensate, and (3) without cloud-radiationturbulence interaction in the dry PBL scheme, CAM3/CAM4 uses additional stability-based stratus fraction as well as RH-based stratus fraction to simulate marine stratocumulus over the subtropical, mid-latitude and Arctic oceans. With the new CAM5 physics addressing these limitations in the CAM3/CAM4 physics, cloud macrophysics should also be revised for consistency among various model physics. Here, we document the revised cloud macrophysics in CAM5. Additional details on CAM5s cloud macrophysics are discussed in Park et al. [2010]. In the following sections, we will document how CAM5 computes (1) cloud fractions - deep cumulus fraction, shallow cumulus fraction, and stratus ( liquid and ice separately ) fractions, (2) horizontal and vertical overlapping structures of clouds, and (3) net condensation rates of water vapor into cloud liquid and ice.

2679

4.7.1

Cloud Fractions

2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690

Cloud fraction is a volume containing hydrometeors sustained in the atmosphere. In CAM5, two types of clouds exist: stratus and cumulus. In nature, these two clouds can be identied by their shapes and turbulent properties. Stratus is horizontally extended with symmetric turbulence properties: fractional area, strength of vertical velocity, vertical extent, and degree of saturation within updraft are similar to those within downdraft. On the other hand, cumulus is vertically stretched with asymmetric turbulence properties: updraft is narrow, strong, and usually saturared while compensating subsidence is broad, weak, and unsaturated. In CAM5, moist turbulence scheme is designed to simulate symmetric turbulences while convection schemes are for simulating asymmetric turbulences. While there is an attempt to treat these two distinct turbulences in a unied way, we stick to the more convectional approach.

141

2691

Deep Cumulus Fraction Similar to CAM3/CAM4, CAM5 computes deep cumulus fraction adp,cu using the following empirical formula. adp,cu = k1,dp loge (1 + k2Mdp,cu ), adp,cu = max(0, min(adp,cu , 0.6)) (4.180)

2692

2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706

where k1,dp is an adjustable parameter given in Appendix C, k2 = 675 and Mdp,cu is convective updraft mass ux [kg m2 s1 ] from deep convection scheme. When identied to be active, Mdp,cu is non-zero from the lowest model layer to the cumulus top. With no further attempt to separate dry and moist deep convection, Eqn.(4.180) can generate empty ( without in-cumulus condensate ) deep convective cloud fraction in the layers below the Lifting Condensation Level ( LCL ). In contrast to stratus fraction that will be discussed later, we compute a single deep cumulus fraction not the separate liquid and ice deep cumulus fractions. We impose a constraint that adp,cu is always smaller than 0.6. Originally, this empirical formula was obtained by including not only cumulus but also stratus generated by detrained cumulus condensate, which by construction results in overestimated cumulus fraction. Thus, we are using a freedom to change the two coecients 0.04 and 675 to simulate convective updraft fractional area only. Currently these coecients are also used as tuning parameters to obtain reasonable regional/global radiation budget and grid-mean LWC/IWC. Shallow Cumulus Fraction In contrast to CAM3/CAM4, CAM5s new shallow convection scheme ( Park and Bretherton, 2009 ) computes vertical velocity as well as mass ux within cumulus updraft. Thus, shallow cumulus fraction ash,cu in CAM5 is directly computed using the denition of convective updraft mass ux: ash,cu = 2 CMsh,cu wu,cu , ash,cu = max(0, min(ash,cu , 0.2)) (4.181)

2707

2708

2709 2710 2711 2712 2713 2714 2715 2716 2717 2718

where Msh,cu is shallow convective mass ux within cumulus updraft [kg m2 s1], is density [kg m3 ] and wu,cu is vertical velocity within cumulus updraft [m s1 ]. Note that a factor 2 is multiplied by considering the dierence between core ( e.g., positively buoyant saturated portions ) updraft fractional area and saturated updraft fractional area estimated from the LES. The details on how to compute Msh,cu and wu,cu are described in Park and Bretherton [2009]. This ash,cu is computed from the LCL of cumulus updraft ( or PBL top if LCL is within the PBL ) to the cumulus top where updraft vertical velocity is zero. So, ash,cu always contains positive cumulus condensate, that is, there is no empty shallow cumulus clouds. Similar to deep cumulus fraction, we compute a single shallow cumulus fraction not the separate liquid and ice shallow cumulus fractions. We impose a constraint that ash,cu is always smaller than 0.2. Liquid Stratus Fraction

2719

2720

142

2721 2722 2723 2724

In CAM3/CAM4, stratus fraction was parameterized as a sum of RH based and Stability based cloud fractions. The latter was necessary because the dry PBL scheme in CAM3/CAM4 cannot moisten upper portion of stratocumulus-topped PBL due to its inability to simulate cloud-radiation-turbulence interactions. The RH-based stratus fraction in CAM3/CAM4 is a quadratic function of grid-mean RH ( Slingo [1987b], Rasch and Kristjansson [1998b] ).
2

ast = CU Uc 1 Uc

(4.182)

where U is grid-mean RH dened using saturation specic humidity over a mixture of cloud water and ice where mixing fraction is a function of temperature, and Uc is a critical RH. Stratus is formed only when U is larger than Uc . Note that CAM3/CAM4 diagnoses a single stratus fraction not the separate liquid and ice stratus fractions in contrast to CAM5. While simple to use, above Eqn.(4.182) has two shortcomings. First, at the limit of ast 1 , we expect that RH in the clear portion ( Ur ) approaches to 1 in nature. However, Eqn.(4.182) does not satisfy this condition unless Uc 1 as shown below: limast 1 Ur = limast 1 C(1 Uc ) ast + Uc ast 1 ast = 0.5 (1 + Uc ) (4.183)

2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735

Second, Eqn.(4.182) is not derived from the explicit subgrid scale distributions of total specic humidity, making it hard to impose internal consistency between stratus fraction and in-stratus condensate. Following Smith [1990], liquid stratus fraction in CAM5 is derived from the assumed triangular distribution of total relative humidity, v = qt,l /qs,w where qt,l is total liquid specic humidity (=qv + ql ) and qs,w is saturation specic humidity over water. Then liquid stratus fraction al,st becomes a function of grid-mean RH over water, Ul ( Park et al. [2010] ). 1 if Ul Ul , 2/3 1 C32 C U U U U l if C16 (5 + Ucl ) Ul Ul , l l cl al,st = 4 cos C13 acos C32 2 CU U U U 2 if Ucl Ul C16 (5 + Ucl ), l cl l cl 0 if Ul Ucl , (4.184) where Ul is RH within liquid stratus ( =1 ) and Ucl is critical RH that liquid stratus is formed when Ul is larger than Ucl . We can easily check limal,st 1 RHr = 1. For a given Ul Ucl , CAM5 ( Eqn.(4.184) ) produces less stratus fraction than CAM3/CAM4 ( Eqn.(4.182) ). In addition, the sensitivity of liquid stratus fraction to the changes of grid-mean RH diers between the two models. Note that Ucl = 1 v where v is the half-width of the triangular distribution. Ideally, subgrid-scale variability v should be internally computed by considering all sources of subgridscale motions from individual physical processes - moist turbulence, detrainment of convective condensate, meso-scale organizations, gravity waves induced by convection or surface inhomogeneity, and etc. In CAM5, however, Ucl is externally specied as a function of height and surface properties and being used as a tuning parameter. We chose Ucl = 0.89 in the layers 143

2736 2737 2738 2739 2740 2741 2742 2743 2744

below 700 hPa ( Low-Level Stratus ) but Ucl = 0.79 over lands when a water-equivalent snow depth is less than 106 [m], Ucl = 0.80 in the layers above 400 hPa ( High-Level Stratus ), and a linearly-interpolated Ucl between 700 hPa and 400 hPa ( Mid-Level Stratus ). In principle, LWC within the liquid stratus can be diagnosed from the assumed triangular PDF ( Smith [1990], Park et al. [2010] ). However, CAM5 uses a separate prognostic condensation scheme for liquid stratus condensation similar to CAM3/CAM4. This ( diagnostic cloud fraction but separate prognostic condensation for liquid stratus ) can cause inconsistency between stratus fraction and in-stratus cloud condensate. We perform additional pseudo condensation-evaporation process to remove this inconsistency as will be discussed later. Ice Stratus Fraction In CAM3/CAM4, a single stratus fraction ast was diagnosed using a mean saturation specic humidity qs = qs,w + (1 ) qs,i where qs,w and qs,i are saturation specic humidities over water and ice, respectively, and 0 1 is a function of temperature. In CAM5, however, we separately diagnose ice stratus fraction ai,st using a modied Slingos formula as below ( Gettelman et al. [2010a] ).
2

2745

2746

ai,st

= CUi Uci Ui Uci Ui = Cqv + qi qs,i

(4.185)

2747 2748 2749 2750 2751 2752 2753 2754

where Ui is grid-mean total RH including ice condensate dened over ice, and Ui is RH within ice stratus. In contrast to liquid condensation that always occurs whenever qv > qs,w , ice nucleation and ice growth processes are not spontaneous and very slow. Thus, the linkage between ice saturation excess s = qv qs,i and the amount of ice condensate is weak. Eqn.(4.185) is an attempt to address these properties of ice processes: supersaturation within ice stratus is taken into account by using Ui > 1, and by including ice condensate in the denition of Ui , ice condensate as well as ice saturation excess contributes to ice stratus fraction. In CAM5, we chose Ui = 1.1 and Uci = 0.80 regardless of heights and the properties of the Earth surface.

2755

4.7.2

Cloud Overlaps

2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766

We have computed 4 independent cloud fractions ( 0 al,st , ai,st 1, 0 ash,cu 0.2, 0 adp,cu 0.6 ) in each layer. The performance of individual physical processes is sensitive how these clouds are distributed in the horizontal plane and vertical column. For example, if al,st and ai,st are maximally-overlapped ( non-overlapped ) in the horizontal, Bergeron-Findeisen conversion process from cloud liquid droplet to ice crystal will be active ( inactive ). If cumulus prefentially grows into the pre-existing stratus ( clear portions ), cumulus will detrain convective condensate into the pre-existing stratus ( clear portions ) without ( with ) evaporation. We can also easily expect that the vertical proles of grid-mean radiative ux, evaporation of precipitation, activation and wet deposition of aerosols are sensitive to the vertical overlapping structures of various clouds. Given the 2-moment stratiform microphysics in CAM5, correct simulations

144

2767 2768 2769 2770

of activation and wet deposition of aerosols become even more important. So, parameterization of cloud overlapping structures is as important as the parameterization of individual cloud fractions. Ideally, all physics schemes should use a single consistent cloud overlapping structure. In this section, we describe the horizontal and vertical overlapping structures of clouds in CAM5. Horizontal Overlap In CAM5, we assume that (1) shallow and deep cumulus fractions are non-overlapped with each other, (2) liquid and ice stratus fractions are maximally overlapped, i.e., ast = max(al,st , ai,st ), and (3) stratus only lls the non-cumulus areas, i.e., a higher occupancy priority is given to the cumulus over stratus in each layer. Stratiform microphysics in CAM5 assumes that stratus LWC/IWC is uniformly distributed over the single stratus fraction ast even though further elaboration is possible. The third assumption above comes from distinct turbulent properties in each clouds: cumulus updraft is strong and grows vertically, and so, if there are any pre-existing stratus on its path, cumulus updraft will push out the pre-existing stratus and occupy the original portion. The assumed horizontal overlapping structure between cumulus and stratus determines the physical stratus fractions. If a is each of 4 cloud fractions computed in the previous section, the physical cloud fraction A of each cloud fraction a becomes Ash,cu = ash,cu 0.2 Acu = Ash,cu + Adp,cu 0.8 Ai,st = (1 Acu ) ai,st 1 Anet = Ast + Acu 1 Al,st = (1 Acu ) al,st 1 Adp,cu = adp,cu 0.6 (4.186)

2771

2772

Ast = max(Al,st , Ai,st) 1

2773 2774 2775 2776 2777

where Ul and Ui in Eqs.(4.184) and (4.185) are now changed to the mean RH averaged over the non-cumulus areas in each layer. In CAM5, state variables saved into the standard physical state arrays are the mean values averaged over the non-cumulus areas, that is, environmental mean not the grid mean. These physical cloud fractions A are passed into various physics schemes following the cloud macrophysics. Vertical Overlap In CAM5, the following physical processes make use of vertical overlap assumption of clouds: (a) deep and shallow convection schemes to compute evaporation of convective precipitations, (b) stratiform microphysics to compute production and evaporation of stratiform precipitation, (c) activation and wet scavenging of aerosols by convective and stratiform precipitations, and (d) radiation scheme. While computations of cloud fractions and horizontal cloud overlaps are substantially revised, CAM5s vertical cloud overlap is similar to CAM3/CAM4, which is summarized below. 145

2778

2779 2780 2781 2782 2783 2784 2785 2786

2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826

(a) CAM5s deep and shallow convection schemes assume that convective precipitation area is always 1 if convective precipitation ux is positive. In reality, however, if vertical shear of horizontal winds is neglected, most of the convective precipitation is likely to fall into the saturated cumulus updraft not into clear portions. Thus, CAM5s vertical cumulus overlap may overestimate evaporation of convective precipitation, leading to excessive water vapor in the atmosphere. (b) CAM5s stratiform microphysics assumes that stratus fraction ast is maximally overlapped in vertical regardless of vertical separation distance, and stratiform precipitation area is the same as maximum stratus fraction in the layers above the current layer as long as precipitation ux is positive. In reality, however, precipitation falling into clear portion can be completely evaporated, so that precipitation area can be smaller than the maximum stratus fraction in the layers above. (c) CAM5s cloud droplet activation routine assumes maximum overlap of stratus fraction between any adjacent layers. CAM5 computes wet scavenging of aerosols by two processes. The rst is the scavenging of activated aerosols within cloud droplets by the production of precipitation. The second is the scavenging of the remaining non-activated aerosols by the precipitation ux. These two processes are separately applied for each convective and stratiform precipitations. For the purpose of wet scavenging of aerosols, CAM5 assumes that (1) convective ( stratiform ) precipitation area at any height is a sum of cumulus ( stratus ) fractions in the layers above weighted by the ratio of net production rate of convective ( stratiform ) precipitation in each layer to the vertically integrated net production rate of convective ( stratiform ) precipitation from the top layer to the layer just above the current layer, and (2) in computing wet scavenging of non-activated aerosols, precipitation ux area at the top interface of each layer is randomly overlapped with the cloud fraction. The second assumption allows CAM5 to bypass the computation of complex overlapping areas between precipitation ux and cloud fractions. (d) CAM5s radiation scheme computes one single cloud fraction and in-cloud LWC/IWC in each layer by combining deep and shallow cumulus and stratus cloud properties through a simple cloud area weighting. Then, it assumes a maximum vertical overlap in each of the 3 regimes representing lower ( p > 700 hPa ), middle ( 400 hPa < p < 700 hPa ), and upper ( p < 400 hPa ) atmospheres, and a random vertical overlap between these 3 regimes. This generates a set of sub-columns in which cloud fraction is either 1 or 0 in each layer. By averaging each sub-columns radiative heating rate, it computes grid-mean radiative heating rate. In principle, all the above 4 processes should use the identical vertical cloud overlapping structure. Due to the contrasting natures of turbulences, cumulus and stratus are likely to have dierent vertical cloud overlap. If vertical shear of horizontal winds is neglected, cumulus fractions are likely to be maximally overlapped over the entire depth of convective updrafts. On the other hand, vertical distance over which stratus is maximally overlapped is likely to be much smaller than the cumulus. Simultaneous treatment of dierent vertical overlapping structures of cumulus and stratus and implementation of the single unied vertical cloud overlap into the CAM is one of the future development plans.

2827

4.7.3

Condensation Processes

2828 2829

This section describes how much water vapor is converted into cloud condensates. This 146

2830 2831

process diers from the conversion of cloud droplet into precipitation, which is treated by the cloud microphysics. Stratus Liquid Similar to CAM3/4 ( Zhang et al. [2003a] ), CAM5 uses prognostic condensation scheme for liquid stratus condensate. The fundamental assumption used for computing grid-mean net condensation rate of water vapor into liquid stratus droplet ( Q ) is that (1) RH over the water within the liquid stratus is always 1, and (2) no liquid stratus droplet exists in the clear portion outside of the liquid stratus fraction. These two conditions will be called saturation equilibrium of liquid stratus. Whenever any GCM grid is perturbed by external forcings, the system always tries to restore the saturation equilibrium state. This allows us to compute the grid-mean net condensation rate of water vapor into liquid stratus condensate for a given set of external forcings. The details of liquid stratus condensation is described in Park et al. [2010]. Lets assume that one GCM grid layer is in saturation equilibrium state at a certain moment. During the model time step t, the layer is perturbed by external forcings ( e.g., stratiform microphysics, radiation, moist turbulence, large-scale advection, and convections ). In order to restore saturation equilibrium, Q should be initiated within the layer. The changes of grid-mean liquid stratus condensate ql,st = Al,st ql,st during t is the sum of grid-mean net condensate rate Q and the grid-mean external forcings of liquid condensates Fl : Q = ql,st Fl = Al,st ql,st + c ql,st Al,st Fl (4.187)

2832

2833 2834 2835 2836 2837 2838 2839 2840 2841 2842

2843 2844 2845 2846

where 0 c 1 is the ratio of in-cloud condensate of newly formed or dissipated stratus to the in-cloud condensate of pre-existing stratus. The denotes time-tendency of . If liquid stratus has homogeneous condensate, it will be c = 1, but it is likely that c < 1 in nature since stratus has non-homogeneous condensate in general. In CAM5, we use c = 0.1. From the two assumptions for saturation equilibrium of liquid stratus, we can derive the following simultaneous linear equations ( Park et al. [2010] ). a11 ql,st + a12 Al,st = b1 a21 ql,st + a22 Al,st = b2 where individual coecients aij and bi are a11 = Al,st a21 = + (CLv Cp ) Al,st b1 = qt,all Tl,all G al,st acu b2 = qt,all Tl,all 147 a22 = (CLv Cp ) c ql,st a12 = G + c ql,st (4.189) (4.188)

with = [C1qs,w ] = C qv qs,w (Cqs,w T ) 2 = (Cqs,w T ) = + CLv Cp (4.190)

G = C11 acu Cal,st Ue and

qt,all = qv,adv + ql,adv + qv,mic + ql,mic Tl,all = Tadv + Tmic CLv Cp (ql,adv + ql,mic ) qt,all = qv,adv + ql,adv + ql,mic ql,mic = [C ql,mic max(Al,st , Ai,st)]
2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873

(4.191) (4.192) (4.193) (4.194)

where denotes local time-tendency, subscript all denotes all the processes except cloud macrophysics, which is the sum of cloud microphysics ( subscript mic ) and the other processes denoted by subscript adv. In Eqn.(4.194), we used max(Al,st , Ai,st) instead of Al,st since stratiform microphysics is formulated based on a single stratus fraction, max(Al,st , Ai,st ). Above formulation was derived assuming that temperature is uniform within the grid, and stratiform microphysics does not change water vapor within the liquid stratus, and all forcings except stratiform microphysics are uniformly applied into the grid. Using Al,st from Eqn.(4.184) and (4.186), we can compute Q if Fl is given. Although the computation of Q explicitly makes use of Al,st , the updated ql,st (t + t) is not necessarily consistent with the updated Al,st (t + t). For example, it can be ql,st (t + t) = 0 but Al,st (t + t) > 0 ( so called empty cloud ) or ql,st (t + t) > 0 but Al,st (t + t) = 0 ( so called innitely dense cloud ). This inconsistency between stratus fraction and in-stratus condensate comes from the combined use of prognostic stratiform condensation and diagnostic stratus fraction schemes with a nite model intergation time step t. In order to prevent these unreasonable situations, we additionally condensate water vapor or evaporate stratus liquid droplets until the in-stratus LWC, ql,st (t + t) falls within the externally specied ranges, 1 0.02 ql,st (t + t) [g kg ] 3. Note that this pseudo condensation evaporation process does not change the grid-mean liquid stratus condensate and is not performed if ql,st (t + t) = 0 and Al,st (t + t) = 0 at the beginning. CAM5 is using two moment stratiform microphysics and so prognoses not only the mass but also the number concentration of cloud droplets. When net condensation occurs ( Q > 0 ), cloud macrophysics does not change droplet number concentration, but when net evaporation occurs ( Q < 0 ), droplet number concentration is reduced in proportion to the decrease of the mass of stratus liquid droplets. Regardless of the sign of Q, however, droplet activation process within stratus is additionally performed at the beginning of cloud microphysics at each time step. Thus, cloud droplet number is consistently generated when Q > 0 in the initially clear layer.

148

2874

Stratus Ice In contrast to liquid stratus, we cannot assume a certain equilibrium state for ice stratus because ice process is much slower than the liquid process. Thus, deposition-sublimation rate between water vapor and ice crystals are computed using an explicit process algorithm. In CAM5, deposition of water vapor into ice crystals ( this is a separate process from the Bergeron-Findeisen conversion of cloud liquid droplets into ice crystals ) only occurs when ice stratus exists at temperature lower than 273.15K. In case of pure ice stratus, in-cloud water vapor is set to the grid-mean water vapor. If in-cloud water vapor is larger than the saturation water vapor over ice, water vapor is deposited into ice crystals. In case of mixed-phase stratus, incloud water vapor is set to the equal weighting average of grid-mean water vapor and saturation water vapor over water. In this case, however, direct deposition of water vapor into ice crystals occurs only after pre-existing cloud liquid droplets are completly depleted into ice crystals by Bergeron-Findeisen process. That is, if Bergeron-Findeisen process is not strong enough to deplete pre-existing cloud liquid droplets, no direct deposition occurs from the water vapor into ice crystals. Sublimation of ice crystals into water vapor occurs regardless of temperature as long as water vapor within the ice stratus is smaller than the saturation water vapor over the pre-existing ice crystals. A constraint is imposed such that direct deposition of water vapor into ice crystals does not reduce grid-mean RH over ice below 1. Additional constraint is imposed such that sublimation should not exceed the available ice crystals and not increase grid-mean RH over ice above 1. See Gettelman et al. [2010a] and the chapter for cloud microphysics for additional details. Condensation within Shallow Cumulus Updraft Condensation within shallow cumulus updraft is described in Park and Bretherton [2009]. Shallow convective updraft rises from the PBL top but condensation occurs from the LCL. If LCL is lower than PBL top, condensation is assumed to occur from the PBL top. During ascent, convective updraft experiences adiabatic cooling, mixing with environmental airs, and precipitation fallout. Vertical evolutions of two conservative scalars qt = qv + ql + qi , c = (Lv /Cp /) ql (Ls /Cp /) qi within convective updraft are explicitly computed using the parameterized entrainment mixing and precipitation processes. From the computed qt , c and saturation specic humidity qs dened as a weighting average of the values over water qs,w and ice qs,i ( the weighting factor is a function of temperature ), we compute condensate amount within convective updraft. Since shallow convective cloud fraction ash,cu is non-zero from the LCL ( or PBL top if LCL is below the PBL ) to the cumulus top, shallow cumulus does not have any empty clouds. Within shallow convection scheme, condensate is partitioned into liquid and ice as a ramping function of temperature between 248K and 268K. However, a separate re-partitioning is performed for convective detrainment ( as a ramping function of temperature between 238.15K and 268.15K ) and for radiative treatment of in-cumulus condensate ( in this case, the repartitioning function is the same as that of stratiform condensate ). When shallow convective condensate is

2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895

2896

2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914

149

2915 2916

detrained into the environment, we assume a xed droplet radius of 8 and 25 [ m] for liquid and ice condensates, respectively. Condensation within Deep Cumulus Updraft Condensates within deep convective updraft is computed in a similar way as shallow convective updraft. When deep convective condensate is detrained into the environment, we assume a xed droplet radius of 10 and 50 [ m] for liquid and ice condensates, respectively. See the chapter for deep convection for additional details.

2917

2918 2919 2920 2921 2922

150

2923

4.8

Aerosols

2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962

Two dierent modal representations of the aerosol were implemented in CAM5. A 7-mode version of the modal aerosol model (MAM-7) serves as a benchmark for the further simplication. It includes Aitken, accumulation, primary carbon, ne dust and sea salt and coarse dust and sea salt modes (4.3). Within a single mode, for example the accumulation mode, the mass mixing ratios of internally-mixed sulfate, ammonium, secondary organic aerosol (SOA), primary organic matter (POM) aged from the primary carbon mode, black carbon (BC) aged from the primary carbon mode, sea salt, and the number mixing ratio of accumulation mode particles are predicted. Primary carbon (OM and BC) particles are emitted to the primary carbon mode and aged to the accumulation mode due to condensation of H2 SO4 , NH3 and SOA (gas) and coagulation with Aitken and accumulation mode (see section below). Aerosol particles exist in dierent attachment states. We mostly think of aerosol particles that are suspended in air (either clear or cloudy air), and these are referred to as interstitial aerosol particles. Aerosol particles can also be attached to (or contained within) dierent hydrometeors, such as cloud droplets. In CAM5, the interstitial aerosol particles and the aerosol particles in stratiform cloud droplets1 (referred to as cloud-borne aerosol particles) are both explicitly predicted, as in Easter et al. [2004]. The interstitial aerosol particle species are stored in the q array of the state variable and are transported in 3 dimensions. The cloud-borne aerosol particle species are stored in the qqcw array of the physics buer and are not transported (except for vertical turbulent mixing), which saves computer time but has little impact on their predicted values Ghan and Easter [2006]. Aerosol water mixing ratio associated with interstitial aerosol for each mode is diagnosed following Kohler theory (see water uptake below), assuming equilibrium with the ambient relative humidity. It also is not transported in 3 dimensions, and is held in the qaerwat array of the physics buer. The size distributions of each mode are assumed to be log-normal, with the mode dry or wet radius varying as number and total dry or wet volume change, and standard deviation prescribed as given in 4.3. The total number of transported aerosol species is 31 for MAM-7. The transported gas species are SO2 , H2 O2 , DMS, H2 SO4 , NH3 , and SOA (gas). For long-term (multiple century) climate simulations a 3-mode version of MAM (MAM-3) is also developed which has only Aitken, accumulation and coarse modes (4.4). For MAM-3 the following assumptions are made: (1) primary carbon is internally mixed with secondary aerosol by merging the primary carbon mode with the accumulation mode; (2) the coarse dust and sea salt modes are merged into a single coarse mode based on the assumption that the dust and sea salt are geographically separated. This assumption will impact dust loading over the central Atlantic transported from Sahara desert because the assumed internal mixing between dust and sea salt there will increase dust hygroscopicity and thus wet removal; (3) the ne dust and sea salt modes are similarly merged with the accumulation mode; and (4) sulfate is partially neutralized by ammonium in the form of NH4 HSO4 , so ammonium is eectively prescribed and NH3 is not simulated. We note that in MAM-3 we predict the mass mixing ratio of sulfate
Note that the explicitly-predicted cloud-borne aerosol particles are for stratiform clouds only, and thus are stratiform-cloud-borne aerosol particles. The convective-cloud-borne aerosol particles in deep and shallow convective clouds are not treated explicitly, and are prescribed as a fraction of the interstitial aerosol particles when calculating wet removal.
1

151

2963 2964

aerosol in the form of NH4 HSO4 while in MAM-7 it is in the form of SO4 . The total number of transported aerosol tracers in MAM-3 is 15.

2965

4.8.1

Emissions

2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999

Anthropogenic (dened here as originating from industrial, domestic and agriculture activity sectors) emissions are from the Lamarque et al. [2010a] IPCC AR5 emission data set. Emissions of black carbon (BC) and organic carbon (OC) represent an update of Bond et al. [2007] and Junker and Liousse [2008]. Emissions of sulfur dioxide are an update of Smith et al. [2001, 2004]. The IPCC AR5 emission data set includes emissions for anthropogenic aerosols and precursor gases: SO2 , primary OM (POM), and BC. However, it does not provide injection heights and size distributions of primary emitted particles and precursor gases for which we have followed the AEROCOM protocols [Dentener et al., 2006a]. We assumed that 2.5% by molar of sulfur emissions are emitted directly as primary sulfate aerosols and the rest as SO2 [Dentener et al., 2006a]. Sulfur from agriculture, domestic, transportation, waste, and shipping sectors is emitted at the surface while sulfur from energy and industry sectors is emitted at 100-300 m above the surface, and sulfur from forest re and grass re is emitted at higher elevations (0-6 km). Sulfate particles from agriculture, waste, and shipping (surface sources), and from energy, industry, forest re and grass re (elevated sources) are put in the accumulation mode, and those from domestic and transportation are put in the Aitken mode. POM and BC from forest re and grass re are emitted at 0-6 km, while those from other sources (domestic, energy, industry, transportation, waste, and shipping) are emitted at surface. Injection height proles for re emissions are derived from the corresponding AEROCOM proles, which vary spatially and temporally. Mass emission uxes for sulfate, POM and BC are converted to number emission uxes for Aitken and accumulation mode at surface or at higher elevations based on AEROCOM prescribed lognormal size distributions as summarized in Table 4.1. The IPCC AR5 data set also does not provide emissions of natural aerosols and precursor gases: volcanic sulfur, DMS, NH3 , and biogenic volatile organic compounds (VOCs). Thus AEROCOM emission uxes, injection heights and size distributions for volcanic SO2 and sulfate and for DMS ux at surface are used. The emission ux for NH3 is prescribed from the MOZART-4 data set [Emmons, 2010]. Emission uxes for isoprene, monoterpenes, toluene, big alkenes, and big alkanes, which are used to derive SOA (gas) emissions (see below), are prescribed from the MOZART-2 data set [Horowitz, 2003]. These emissions represent late 1990s conditions. For years prior to 2000, we use anthropogenic non-methane volatile organic compound (NMVOC) emissions from IPCC AR5 data set and scale the MOZART toluene, bigene, and big alkane emissions by the ratio of year-of-interest NMVOC emissions to year 2000 NMVOC emissions. The emission of sea salt aerosols from the ocean follows the parameterization by Martensson et al. [2003] for aerosols with geometric diameter < 2.8 m. The total particle ux F0 is described by dF0 = W = (Ak Tw + Bk )W dlogDp (4.195)

3000 3001

where Dp is the particle diameter, Tw is the water temperature and Ak and Bk are coecients dependent on the size interval. W is the white cap area:

152

3.41 W = 3.84 104 U10

3002 3003

(4.196)

where U10 is the wind speed at 10 m. For aerosols with a geometric diameter > 2.8 m, sea salt emissions follow the parameterization by Monahan et al. [1986] dF0 B 2 3.41 = 1.373U10 r 3 (1 + 0.0057r 1.05) 101.19e dlogr (4.197)

3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018

where r is the radius of the aerosol at a relative humidity of 80% and B=(0.380-logr)/0.650. All sea salt emissions uxes are calculated for a size interval of dlogDp =0.1 and then summed up for each modal size bin. The cut-o size range for sea salt emissions in MAM-7 is 0.02-0.08 (Aitken), 0.08-0.3 (accumulation), 0.3-1.0 (ne sea salt), and 1.0-10 m (coarse sea salt); for MAM-3 the range is 0.02-0.08 (Aitken), 0.08-1.0 (accumulation), and 1.0-10 m (coarse). Dry, unvegetated soils, in regions of strong winds generate soil particles small enough to be entrained into the atmosphere, and these are referred to here at desert dust particles. The generation of desert dust particles is calculated based on the Dust Entrainment and Deposition Model, and the implementation in the Community Climate System Model has been described and compared to observations [Mahowald et al., 2006a,b; Yoshioka et al., 2007]. The only change to the CAM5 source scheme from the previous studies is the increase in the threshold for leaf area index for the generation of dust from 0.1 to 0.3 m2 /m2 , to be more consistent with observations of dust generation in more productive regions [Okin, 2008]. The cut-o size range for dust emissions is 0.1-2.0 m (ne dust) and 2.0-10 m (coarse dust) for MAM-7; and 0.1-1.0 m (accumulation), and 1.0-10 m (coarse) for MAM-3.

3019

4.8.2

Chemistry

3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036

Simple gas-phase chemistry is included for sulfate aerosol. This includes (1) DMS oxidation with OH and NO3 to form SO2 ; (2) SO2 oxidation with OH to form H2 SO4 (gas); (3) H2 O2 production (HO2 +HO2 ); and (4) H2 O2 loss (H2 O2 photolysis and H2 O2 +OH). The rate coecients for these reactions are provided from the MOZART model [Emmons, 2010]. Oxidant concentrations (O3 , OH, HO2 , and NO3 ) are temporally interpolated from monthly averages taken from MOZART simulations [Lamarque et al., 2010a]. SO2 oxidation in bulk cloud water by H2 O2 and O3 is based on the MOZART treatment [Tie et al., 2001]. The pH value in the bulk cloud water is calculated from the electroneutrality equation between the bulk cloud-borne SO4 and NH4 ion concentrations (summation over modes), and ion concentrations from the dissolution and dissociation of trace gases based on the Henrys law equilibrium. Irreversible uptake of H2 SO4 (gas) to cloud droplets is also calculated [Seinfeld and Pandis, 1998]. The sulfate produced by SO2 aqueous oxidation and H2 SO4 (gas) uptake is partitioned to the cloud-borne sulfate mixing ratio in each mode in proportion to the cloud-borne aerosol number of the mode (i.e., the cloud droplet number associated with each aerosol mode), by assuming droplets associated with each mode have the same size. For MAM-7, changes to aqueous NH4 ion from dissolution of NH3 (g) are similarly partitioned among modes. SO2 and H2 O2 mixing ratios are at the same time reduced due to aqueous phase consumption.

153

3037

4.8.3

Secondary Organic Aerosol

3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048

The simplest treatment of secondary organic aerosol (SOA), which is used in many global models, is to assume xed mass yields for anthropogenic and biogenic precursor VOCs, then directly emit this mass as primary aerosol particles. MAM adds one additional step of complexity by simulating a single lumped gas-phase SOA (gas) species. Fixed mass yields for ve VOC categories of the MOZART-4 gas-phase chemical mechanism are assumed, as shown in Table 4.2. These yields have been increased by an additional 50% for the purpose of reducing aerosol indirect forcing by increasing natural aerosols. The total yielded mass is emitted as the SOA (gas) species. MAM then calculates condensation/evaporation of the SOA (gas) to/from several aerosol modes. The condensation/evaporation is treated dynamically, as described later. The equilibrium partial pressure of SOA (gas), over each aerosol mode m is expressed in terms of Raoults Law as:
Pm = (

ASOA m )P 0 ASOA + 0.1AP OA m m

(4.198)

3049 3050 3051

where ASOA is SOA mass concentration in mode m, AP OA is the primary organic aerosol m m (POA) mass concentration in mode m (10% of which is assumed to be oxygenated), and P 0 is the mean saturation vapor pressure of SOA whose temperature dependence is expressed as: Hvap 1 1 ( )] (4.199) R T 298 where P 0 (298 K) is assumed at 1 1010 atm and the mean enthalpy of vaporization Hvap is assumed at 156 kJ mol1 . Treatment of the gaseous SOA and explicit condensation/evaporation provides (1) a realistic method for calculating the distribution of SOA among dierent modes and (2) a minimal treatment of the temperature dependence of the gas/aerosol partitioning. P 0 (T ) = P 0 (298K) exp[

3052 3053 3054 3055 3056

3057

4.8.4

Nucleation

3058 3059 3060 3061 3062 3063 3064 3065 3066

New particle formation is calculated using parameterizations of binary H2 SO4 -H2 O homogeneous nucleation, ternary H2 SO4 -NH3 -H2 O homogeneous nucleation, and boundary layer nucleation. A binary parameterization [Vehkamaki et al., 2002] is used in MAM-3, which does not predict NH3 , while a ternary parameterization [Merikanto et al., 2007] is used in MAM-7. The boundary layer parameterization, which is used in both versions, uses the empirical 1st order nucleation rate in H2 SO4 from Sihto et al. [2006], with a rst order rate coecient of 1.0 106 s1 as in Wang et al. [2009]. The new particles are added to the Aitken mode, and we use the parameterization of Kerminen and Kulmala [2002] to account for loss of the new particles by coagulation as they grow from critical cluster size to Aitken mode size.

3067

4.8.5

Condensation

3068 3069

Condensation of H2 SO4 vapor, NH3 (MAM-7 only), and the SOA (gas) to various modes is treated dynamically, using standard mass transfer expressions [Seinfeld and Pandis, 1998] that

154

3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090

are integrated over the size distribution of each mode [Binkowski and Shankar, 1995]. An accommodation coecient of 0.65 is used for H2 SO4 [Poschl et al., 1998], and currently, for the other species too. H2 SO4 and NH3 condensation are treated as irreversible. NH3 uptake stops when the NH4 /SO4 molar ratio of a mode reaches 2. SOA (gas) condensation is reversible, with the equilibrium vapor pressure over particles given by Eq. (4.296). In MAM-7, condensation onto the primary carbon mode produces aging of the particles in this mode. Various treatments of the aging process have been used in other models [Cooke and Wilson, 1996; Wilson et al., 2001; Liu et al., 2005; Riemer et al., 2003]. In CAM5 a criterion of 3 mono-layers of sulfate is used to convert a fresh POM/BC particle to the aged accumulation mode. Using this criterion, the mass of sulfate required to age all the particles in the primary carbon mode, MSO4,ageall , is computed. If MSO4,cond condenses on the mode during a time step, we assume that a fraction fage = MSO4,cond / MSO4,ageall has been aged. This fraction of the POM, BC, and number in the mode is transferred to the accumulation mode, along with the condensed soluble species. SOA is included in the aging process. The SOA that condenses in a time step is scaled by its lower hygroscopicity to give a condensed SO4 equivalent. The two continuous growth processes (condensation and aqueous chemistry) can result in Aitken mode particles growing to a size that is nominally within the accumulation mode size range. Most modal aerosol treatments thus transfer part of the Aitken mode number and mass (those particles on the upper tail of the distribution) to the accumulation mode after calculating continuous growth [Easter et al., 2004].

3091

4.8.6

Coagulation

3092 3093 3094 3095 3096 3097 3098 3099 3100 3101

Coagulation of the Aitken, accumulation, and primary carbon modes is treated. Coagulation within each of these modes reduces number but leaves mass unchanged. For coagulation of Aitken with accumulation mode and of primary-carbon with accumulation mode, mass is transferred from Aitken or primary-carbon mode to the accumulation mode. For coagulation of Aitken with primary-carbon mode in MAM-7, Aitken mass is rst transferred to the primarycarbon mode. This ages some of the primary-carbon particles. An aging fraction is calculated as with condensation, then the Aitken mass and the aged fraction of the primary-carbon mass and number are transferred to the accumulation mode. Coagulation rates are calculated using the fast/approximate algorithms of the Community Multiscale Air Quality (CMAQ) model, version 4.6.

3102

4.8.7

Water Uptake

3103 3104 3105 3106 3107 3108

Water uptake is based on the equilibrium Kohler theory [Ghan and Zaveri, 2007] using the relative humidity and the volume mean hygroscopicity for each mode to diagnose the wet volume mean radius of the mode from the dry volume mean radius. The hygroscopity of each component is listed in Table 4.3. The hygroscopicities here are equivalent to the parameters of Petters and Kreidenweis [2007]. Note that the measured solubility of dust varies widely, from 0.03 to 0.26 [Koehler et al., 2009a].

155

Table 4.1: Size distributions of primary emissions. Geometric standard deviation, sg 1.8 See note 1.8 See note 1.8 1.8 1.8
2

Emission Source BC/OM Forest re/grass re Domestic/energy/industry/ transportation/shipping/waste SO4 Forest re/grass re/waste Energy/industry/shipping Domestic/transportation Continuous volcano, 50% in Aitken mode Continuous volcano, 50% in accum. mode
1

Number mode diameter, Dgn (m) 0.080 See note 0.080 See note 0.030 0.030 0.080
2

Demit (m)1 0.134 0.134 0.134 0.261 0.0504 0.0504 0.134

Demit is volume-mean diameter = Dgn exp(1.5ln(sg )2 ) used in number emissions as Enumber = 3 Emass /(/6Demit ) 2 This value is intermediate between the Dentener et al. [2006a] Demit = 0.0504m and Liu et al. [2005] Demit = 0.206m. 3 Adapted from Stier et al. [2005] where 50% of mass goes to accumulation mode with Demit = 0.207 m, and 50% goes to coarse mode with Demit = 3.08m. We put all mass in accumulation mode, and Demit = 0.261m gives same number emissions as Stier et al. [2005]. [Dentener et al. [2006a] put all in coarse mode with Demit = 2.06 m]

Table 4.2: Assumed SOA (gas) yields Species Big Alkanes Big Alkenes Toluene Isoprene Monoterpenes Mass yield 5% 5% 15% 4% 25% Reference Lim and Ziemann [2005] assumed Odum et al. [1997] Kroll et al. [2006] Ng et al. [2007]

Table 4.3: Hygroscopicity of aerosol components Seasalt 1.16 sulfate 0.507 nitrate 0.507 ammonium 0.507 SOA 0.14 POM 0.10 BC 1010 dust 0.068

156

3109

4.8.8

Subgrid Vertical Transport and Activation/Resuspension

3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141

The vertical transport of interstitial aerosols and trace gases by deep convective clouds, using updraft and downdraft mass uxes from the Zhang-McFarlane parameterization, is described in Collins et al. [2004a]. Currently this vertical transport is calculated separately from wet removal, but a more integrated treatment is planned. Cloud-borne aerosols, which are associated with large-scale stratiform cloud, are assumed to not interact with the convective clouds. Vertical transport by shallow convective clouds is treated similarly, using mass uxes from the shallow convection parameterization. Turbulent transport of the aerosol is given a special treatment with respect to other tracers. To strengthen the coupling between turbulent transport and aerosol activation in stratiform clouds, the implicit time integration scheme used for turbulent transport of heat, energy, and momentum is replaced by an explicit scheme for droplets and aerosol. A sub-timestep is calculated for each column based on the minimum turbulent transport time in the column. Turbulent transport is integrated over the sub-time steps using a forward time integration scheme. Aerosol activation converts particles from the interstitial attachment state to the cloudborne state. In stratiform cloud, activation is treated consistently with droplet nucleation, so that the total number of particles activated and transferred to the cloud-borne state equals to the number of droplets nucleated. Activation is parameterized in terms of updraft velocity and the properties of all of the aerosol modes [Abdul-Razzak and Ghan, 2000b], with both mass and number transferred to the cloud-borne state. The updraft velocity is approximated by the square root of the turbulence kinetic energy, with a minimum value of 0.2 m s1 . Activation is assumed to occur as updrafts carry air into the base of the cloud [Ghan et al., 1997] and as cloud fraction increases [Ovtchinnikov and Ghan, 2005]. In addition, activation is assumed to occur as air is continuously cycled through clouds, assuming a cloud regeneration time scale of one hour. Consider a model time step of 20 minutes, so that 1/3 of the cloud is regenerated in a time step. We essentially dissipate then reform 1/3 of cloud each time step. During dissipation, grid-cell mean cloud droplet number is reduced by 1/3, and 1/3 of the cloud-borne aerosols are resuspended and converted to the interstitial state. During regeneration, interstitial aerosols are activated in the new cloud, and cloud droplet number is increased accordingly. The regeneration has small impact on shallow boundary layer clouds, but it noticeably increases droplet number in deeper free-tropospheric clouds where vertical turbulence mixing is slow. Particles are resuspended as aerosol when droplets evaporate. This process is assumed to occur as droplets are transferred below or above cloud and as clouds dissipate.

3142

4.8.9

Wet Deposition

3143 3144 3145 3146 3147 3148 3149 3150

Aerosol wet removal is calculated using the CAM3.5 wet removal routine [Rasch et al., 2000; Barth et al., 2000] with modications for the consistency with cloud macro- and microphysics. The routine treats in-cloud scavenging (the removal of cloud-borne aerosol particles) and belowcloud scavenging (the removal of interstitial aerosol particles by precipitation particles through impaction and Brownian diusion). For in-cloud scavenging, the stratiform and convective cloud fraction, cloud water, and precipitation production proles are used to calculate rst-order loss rate proles for cloud-water. These cloud-water rst-order loss rates are multiplied by solubility factors to obtain aerosol

157

3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177

rst-order loss rates, which are applied to the aerosol proles. The solubility factors can be interpreted as (the fraction of aerosols that are in cloud drops) (an additional tuning factor). In CAM3.5, where the cloud-borne aerosol is not explicitly calculated, a value of 0.3 is used for solubility factors for all aerosol types and sizes. Dierent values are used for the MAM. The stratiform in-cloud scavenging only aects the stratiform-cloud-borne aerosol particles, and these have solubility factors of 1.0. It does not aect the interstitial aerosol particles, and these have solubility factors of 0.0. For convective in-cloud scavenging of MAM aerosols, both a solubility factor and a withinconvective-cloud activation fraction are passed to the wet removal routine. For the stratiformcloud-borne aerosol particles, there is no wet removal by convective clouds, and these factors are zero. For interstitial (with respect to stratiform cloud) aerosol, the solubility factor is 0.5, and the activation fractions are 0.0 for the primary carbon mode, 0.4 for the ne and coarse dust modes, and 0.8 for other modes. The lower values reect lower hygroscopity. These factors are applied to both number and mass species within each mode, with one exception. In MAM-3, dierent activation fractions are applied to the dust and sea salt of the coarse mode (0.4 and 0.8 respectively), and a weighted average is applied to the coarse mode sulfate and number. For below-cloud scavenging, the rst-order removal rate is equal to [ (solubility factor) (scavenging coecient) (precipitation rate) ]. Again, the solubility factor can be viewed as a tuning factor. In CAM3.5, a solubility factor of 0.3 and a scavenging coecient of 0.1 mm1 are used for all aerosols. In MAM, the scavenging coecient for interstitial aerosol is explicitly calculated as in Easter et al. [2004] and thus varies strongly with particle size, with lowest values for the accumulation mode; and the solubility factor is 0.1. For stratiform-cloud-borne aerosol, there is no below-cloud scavenging, and the solubility factor is 0.0. Aerosol that is scavenged at one altitude can be resuspended at a lower altitude if precipitation evaporates. In CAM5, as in CAM3.5, this process is treated for aerosol removed by stratiform in-cloud scavenging. A fraction of the in-cloud scavenged aerosol is resuspended, and the resuspended fraction is equal to the fraction of precipitation that evaporates below cloud.

3178

4.8.10

Dry Deposition

3179 3180 3181 3182 3183 3184 3185

Aerosol dry deposition velocities are calculated using the [Zhang et al., 2001] parameterization with the CAM5 land-use and surface layer information. Gravitational settling velocities are calculated at layers above the surface [Seinfeld and Pandis, 1998]. Both velocities depend on particle wet size and are dierent for mass and number and between modes. The velocities for cloud-borne aerosols are calculated based on droplet sizes. Aerosol mixing ratio changes and uxes from dry deposition and sedimentation throughout a vertical column are then calculated using the CAM5 dust deposition/sedimentation routine.

158

Figure 4.3: Predicted species for interstitial and cloud-borne component of each aerosol mode in MAM-7. Standard deviation for each mode is 1.6 (Aitken), 1.8 (accumulation), 1.6 (primary carbon), 1.8 (ne and coarse soil dust), and 2.0 (ne and coarse sea salt)

Figure 4.4: Predicted species for interstitial and cloud-borne component of each aerosol mode in MAM-3. Standard deviation for each mode is 1.6 (Aitken), 1.8 (accumulation) and 1.8 (coarse mode)

159

Table 4.4: Density (kg/m3 ) of aerosol material. Sea salt 1900 Sulfate 1770 Nitrate Ammonium 1770 1770 SOA 1000 POA 1000 BC Dust 1700 2600

Table 4.5: Hygroscopicity of aerosol components. Sea salt 1.16

3186

Sulfate 0.507

Nitrate 0.507

Ammonium 0.507

SOA 0.14

POA BC Dust 1.e-10 1.e-10 0.068

4.9

Condensed Phase Optics

3187 3188 3189 3190 3191 3192 3193 3194

Condensed phase (aerosols, liquid cloud droplets, hydrometeors, and ice crystal) optics are provided as a mass-specic quantities in m2 /kg. These optics are specied for each band of the shortwave and longwave radiation code. For the shortwave, unscaled extinction, single-scattering albedo, and asymmetry parameter are specied. For the longwave, the mass-specic absorption is specied. Vertical optical depths are computed by multiplying by the mass-specic quantities by the vertical mass path of the corresponding material. For clouds, the in-cloud values of the mixing ratios are used to compute the in-cloud values of cloud optical depths. The radiation does not use grid-cell average optical depths of clouds.

3195

4.9.1

Tropospheric Aerosol Optics

3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212

While the radiation code supports a range of possible aerosol packages, the modal aerosol package is the default conguration, and we will discuss the optics treatment used in that package. Aerosol optical properties for each mode are parameterized in terms of wet refractive index and wet surface mode radius of the mode, as described by [Ghan and Zaveri, 2007], except that volume mixing rather than the Maxwell-Garnett mixing rule is used to calculate the wet refractive index for mixtures of insoluble and soluble particles (We found little dierence between the volume mixing treatment and the Maxwell-Garnett mixing rule.) Refractive indices for water and for most aerosol components are taken from OPAC [Koepke and Schult, 1998], but for black carbon the value (1.95,0.79i) from [Bond and Bergstrom, 2006] is used for solar wavelengths. Densities for each component are listed in Table 4.4. The wet volume mean radius for each mode is calculated from the dry volume mean radius using equilibrium Kohler theory [Ghan and Zaveri, 2007], the relative humidity and the volume mean hygroscopicity. The hygroscopicity of each component is listed in Table 4.5. Note that the measured solubility of dust varies widely, from 0.03 to 0.26 [Koehler et al., 2009b]. The wet surface mode radius is calculated from the wet volume mean radius assuming a wet lognormal size distribution with the same geometric standard deviation as the dry size distribution. The geometric standard deviation is assumed to be constant for each mode.

3213

4.9.2

Stratospheric Volcanic Aerosol Optics

3214 3215

CAM 5.0 species the volcanic aerosol as a mass mixing ratio qV of wet volcanic aerosol to dry air as a function of height, latitude, longitude and time. CAM 5.0 also species a geometric 160

3216 3217

mean radius rg of the volcanic aerosol. The volcanic optics are stored as a lookup table as a function of geometric mean radius. The size distribution is dened by a log-normal size distribution with a geometric mean radius rg and geometric standard deviation g . For the standard version of the optics, g = 1.8 = ln(rg ) [min, max ] min = ln(0.01 106 exp(5/2 (ln g )2 )) max = ln(2.00 106 exp(5/2 (ln g )2 )) (4.200) (4.201) (4.202) (4.203) (4.204)

In other words, re spans the range [0.01,2.0] m. The density of the sulfuric acid / water mixture at 75% / 25% at 215K is = 1.75 103 kg/m3
3218 3219 3220

(4.205)

The index of refraction is that specied by Biermann [Biermann et al., 2000] and is available from the HITRAN [Rothman et al., 2009] database. The index at 75%/25% weight percent (sulfuric acid to water) and at 215K is used. The incomplete gamma weight,
r

L(r) =
0

r n(r )dr /
0

r 2 n(r )dr

(4.206)

can be used to dene the mass-specic aerosol extinction, scattering, and asymmetric scattering,
3 bext = qext (r)dL(r) 4 re 0 3 bsca = qsca (r)dL(r) 4 re 0 3 qgqsc (r)dL(r) basm = 4 re 0 3 = (qext (r) qsca (r))dL(r) 4 re 0

(4.207) (4.208) (4.209) (4.210)

babs
3221 3222 3223 3224

where qext (r), qsca (r), qgqsc (r) are eciencies obtained from the MIEV0 program of Wiscombe [Wiscombe, 1996]. These mass-specic properties are averaged over each frequency band of RRTMG and parameterized in a lookup table with = ln(rg ) as the dependent variable. The vertical optical depths are derived as the product of vertical mass path with mass-specic aerosol properties at runtime. Pdry ext = qV bext () (4.211) g where qV is the mixing ratio of volcanic aerosol. The corresponding scattering optical depth, asymmetric scattering optical depth, and absorption optical depth are derived similarly.

3225 3226

161

3227

4.9.3

Liquid Cloud Optics

For liquid clouds CAM 5.0 species the fraction of each grid cell occupied by liquid cloud droplets Cliq , the ratio of mass of condensed water to wet air in the cloud qliq, and the numbersize distribution in terms of the 2 parameters, and of the gamma distribution, n(D) =
3228

dN +1 = D eD dD ( + 1)

(4.212)

where D is the diameter of the droplets. Both the parameters, and have limited ranges: 2. < < 15. +1 +1 < < 50 106 m 2 106 m (4.213) (4.214)

3229 3230 3231

The liquid cloud optics are specied in terms of a lookup table in and 1/. These optics are computed as size-distribution and spectral-band averages of the quantities (e.g., Qext ) computed by the MIEV0 program [Wiscombe, 1996]. The size-integrated mass-specic extinction coecient, kext , (units m2 /kg) is given by: kext () =
4 0

D 2 Qext (D; , m) n(D) dD D 3 n(D) dD 6 w 0

(4.215)

3232 3233

The corresponding quantities are used to compute mass-specic absorption in the longwave as well as single-scattering albedo and asymmetry parameter. The in-cloud optical depth is then given by: liq () = kext () qliq P g (4.216)

3234 3235 3236 3237 3238

where qliq is the ratio of droplet mass to dry air mass. For RRTMG, the wavenumber average values of liq , liq liq, liq liqgliq on each SW band, and the wavenumber average value of the absorption optical depth, liq (1 liq), on each longwave band. In-cloud water path variability is not treated by the optics.

3239

4.9.4

Ice Cloud Optics

3240 3241 3242 3243 3244 3245 3246 3247 3248

CAM 5.0 species an in-cloud ice water path, an ice cloud fraction, and an eective diameter for ice particles in the cloud. The optics for ice clouds are constructed as a lookup table as a function of eective diameter for each of the shortwave and longwave bands in the radiation code. Ice cloud optical properties have been derived using two approaches: (1) calculations of single ice crystal scattering properties based on electrodynamic theory, followed by their application to assumed ice particle size distributions (PSD) and the representation of PSD optical properties through the eective diameter (De ) of the PSD, and (2) parameterization of scattering/absorption processes in terms of ice particle shape and size, and integrating these expressions 162

3249 3250 3251 3252 3253 3254 3255 3256 3257 3258

over the PSD to produce analytical expressions of PSD optical properties in terms of ice crystal and PSD parameters. In the latter case, the PSD extinction and absorption coecients can be expressed as explicit functions of the ice particle projected area- and mass-dimension power laws and the PSD parameters of the gamma form. The modied anomalous diraction approximation (MADA) uses this second approach to calculate ice cloud optical properties. The development of MADA was motivated by a desire to explicitly represent ice optical properties in terms of the ice PSD and ice crystal shape parameters, given that the ice PSD optical properties cannot be uniquely dened by De [Mitchell, 2002]. MADA was developed from van de Hulsts anomalous diraction theory or ADT [van de Hulst, 1957] through a series of physical insights, which are: 1. The eective photon path through a particle by which its scattering properties can be predicted is given by the ratio of particle projected area/particle volume [Bryant and Latimer, 1969; Mitchell and Arnott, 1994], where volume is dened as particle mass/bulk density of ice (0.917 g/cm3 ). 2. The processes of internal reection and refraction can be viewed as extending the photon path and can be parameterized using a MADA framework [Mitchell et al., 1996b]. 3. The maximum contribution of wave resonance or photon tunneling to absorption and extinction can be estimated as a linear function of the real part of the refractive index for ice, nr . Photon tunneling can then be parameterized in terms of nr , size parameter x and the other MADA parameters described above [Mitchell, 2000]. 4. Edge eects as surface wave phenomena pertain only to extinction and can be represented in terms of the size parameter x as described by [Wu, 1956] and modied by [Mitchell, 2000]. Based on a laboratory ice cloud study [Mitchell et al., 2001], edge eects for nonspherical ice crystals do not appear signicant. The rst insight greatly simplied van de Hulsts ADT, resulting in analytic and integrable expressions for the PSD extinction and absorption coecients as shown in [Mitchell and Arnott, 1994]. This simplied ADT may be more accurate than the original ADT [Mitchell et al., 2006a]. This simplied ADT provided an analytical framework on which the other three insights or processes were expressed. These processes were represented analytically for a single ice particle, and then integrated over the PSD to produce extinction and absorption coecients that account for these processes. These coecients were formulated in terms of ice particle shape (i.e. the ice particle area- and mass-dimension power laws) and the three gamma PSD parameters. The basic MADA equations formulated for ice clouds are given in the appendix of [Mitchell, 2002]. Details regarding their derivation and their physical basis are described in [Mitchell, 2000] and [Mitchell et al., 1996b]. The asymmetry parameter g is not treated by MADA, but was parameterized for solar wavelengths as a function of wavelength and ice particle shape and size, based on raytracing calculations by Andreas Macke, as described in [Mitchell et al., 1996b]. The g parameterization for quasi-spherical ice particles is based on the phase function calculations of [Nousiainen and McFarquhar, 2004]. These parameterizations relate g for a PSD to the ice particle size that divides the PSD into equal projected areas (since scattering depends on projected

3259 3260 3261 3262

3263 3264

3265 3266 3267 3268

3269 3270 3271 3272

3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289

163

3290 3291

area). For terrestrial radiation, CAM 5.0 g values for ice are based on the g parameterization described in [Yang et al., 2005]. Tests of MADA While this treatment of ice optical properties began and evolved through van de Hulsts original insights formulated in ADT, optical properties predicted by MADA closely agree with those predicted by other ice optics schemes based on electrodynamic theory. As described in [Mitchell et al., 2001, 2006a], MADA has been tested in a laboratory ice cloud experiment where the MADA extinction error was 3% on average relative to the FTIR measured extinction eciency over the 2-14 m wavelength range. These same laboratory PSD were used to calculate the absorption eciencies using MADA and T-matrix, which diered by 6% on average over the wavelength range 2-18 m (size parameter range 2-22). In corresponding T-matrix calculations of the single-scattering albedo, the mean MADA error was 2.5%. In another test, MADA absorption errors relative to the Finite Dierence Time Domain (FDTD) method (i.e. [Yang et al., 2005] over the wavelength range 3-100 m were no greater than 15% for six ice particle shapes. Finally, the absorption coecients predicted by MADA and the [Fu et al., 1998] and the [Yang et al., 2005] ice optics schemes generally agreed within 5%. Application to CAM 5.0 The MADA-based ice optics scheme described above is not used explicitly in CAM 5.0, but was used to generate a look-up table of optical properties as a function of eective diameter, De . The PSD optical properties consist of the mass-normalized extinction coecient (volume extinction coecient / ice water content), the single-scattering albedo and the asymmetry parameter for bands covering all solar and terrestrial wavelengths. The radiation bands coincide with those used in RRTMG. The ice refractive index values used are from [Warren and Brandt, 2008]. Since MADA is formulated to accept any ice particle shape recipe, a shape recipe corresponding to that observed for mid-latitude cirrus clouds at 45 C (see [Lawson et al., 2006]) was assumed for ice particles larger than 60 m: 7% hexagonal columns, 50% bullet rosettes and 43% irregular ice particles. At smaller sizes, the shape recipe consists of 50% quasi-spherical, 30% irregular and 20% bullet rosette ice crystals, based on in-situ measurements in tropical cirrus [P. Lawson, 2005, personal communication]. The eective diameter is dened in a way that is universal for both ice and water clouds, which is essentially the photon path characterizing the PSD [Mitchell, 2002]: De = 3 IWC 2 i A (4.217)

3292

3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305

3306

3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318

3319 3320

where IWC is the ice water content (g/cm3 ), i is the bulk ice density (0.917 g/cm3 ) and A is the total projected area of the PSD (cm2 /cm3 ).

3321

4.9.5

Snow Cloud Optics

3322 3323

CAM 5.0 species snow as a cloud fraction of snow, an eective diameter of snow, and an in-cloud mass mixing ratio of snow. The snow optics are identical to the optics for ice clouds.

164

3324

4.10

Radiative Transfer

3325 3326 3327 3328 3329

Radiative transfer calculations in the longwave and shortwave are provided by the radiation code RRTMG [Iacono et al., 2008; Mlawer et al., 1997]. This is an accelerated and modied version of the correlated k-distribution model, RRTM. The condensed phase radiative parameterizations are external to the radiation package, however the gas optics and radiative transfer solver are provided within RRTMG.

3330

4.10.1

Combination of Aerosol Radiative Properties

The number Na of aerosol species is arbitrary; however in the standard conguration there are 3 modes. The radiative properties are combined before being passed to the radiative transfer solver. If the extinction optical depth of species i in band b is ib and the single-scattering albedo is ib and the asymmetry parameter is gib then the aerosol optics are combined as follows:
Na

b =
i=1 Na

ib ib ib /b
i=1 Na

(4.218) (4.219) (4.220)

b = gb =
i=1
3331 3332

ib ib gib /(b b )

where b is the total aerosol extinction optical depth in band b, b is the total single-scattering albedo in band b, and gb is the asymmetry parameter in band b.

3333

4.10.2

Combination of Cloud Optics

CAM 5.0 are species three dierent types of clouds: ice clouds, liquid clouds, and snow clouds. Each of these clouds has a separate cloud fraction Cliq, Cice , Csnow , as well as an in-cloud radiative characterization in terms of optical depths i , single-scattering albedo i and asymmetry parameter gi . The optics are smeared together into a total cloud fraction C as follows: C = max{Cliq, Cice , Csnow } c =
ttype

(4.221) (4.222) (4.223) (4.224)

t Ct /C

c =
ttype

tb tb Ct /(c C) tb tb gtb Ct /(c c C)

ttype

gc =
3334

where C, c , c , gc are the combined cloud radiative parameters.

165

3335

4.10.3

Radiative Fluxes and Heating Rates

3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354

Radiative uxes and heating rates in CAM 5.0 are calculated using RRTMG[Iacono et al., 2008]. This model utilizes the correlated k-distribution technique to calculate irradiance and heating rate eciently in broad spectral intervals, while realizing the objective of retaining a high level of accuracy relative to measurements and high-resolution line-by-line models. Sub-grid cloud characterization in RRTMG is treated in both the longwave and shortwave spectral regions with McICA, the Monte-Carlo Independent Column Approximation [Pincus and Morcrette, 2003], using the maximum-random cloud overlap assumption. The thermodynamic state, gas concentrations, cloud fraction, condensed phase optics, and aerosol properties are specied elsewhere. The CAM 5.0 surface model provides both the surface albedo, area-averaged for each atmospheric column, and the upward longwave surface ux, which incorporates the surface emissivity, for input to the radiation. The bulk aerosol package of CAM4 continues to be supported by this radiation code as an option, however a description of this optional conguration is not provided in this document. To provide uxes at the top of the atmosphere, RRTMG uses with an additional layer above the CAM 5.0 model top in both the longwave and shortwave. This extra layer is specied by replicating the composition of the highest CAM 5.0 layer into a layer that extends from the top of the model to 104 hPa. RRTMG does not treat non-LTE (local thermodynamic equilibrium) eects in the upper atmosphere. It provides accurate uxes and heating rates up to about 0.1 hPa, above which non-LTE eects become more signicant. Shortwave Radiative Transfer RRTMG divides the solar spectrum into 14 shortwave bands that extend over the spectral range from 0.2 m to 12.2 m (820 to 50000 cm1 ). Modeled sources of extinction (absorption and scattering) are H2O, O3, CO2, O2, CH4, N2, clouds, aerosols, and Rayleigh scattering. The model uses a two-stream -Eddington approximation assuming homogeneously mixed layers, while accounting for both absorption and scattering in the calculation of reectance and transmittance. The model distinguishes the direct solar beam from scattered (diuse) radiation. The scattering phase function is parameterized using the Henyey-Greenstein approximation to represent the forward scattering fraction as a function of the asymmetry parameter. This deltascaling is applied to the total irradiance as well as to the direct and diuse components. The latter are consistent with the direct and diuse components of the surface albedo, which are applied to the calculation of surface reectance. The shortwave version of RRTMG used in CAM5 is derived from RRTM SW [Clough et al., 2005]. It utilizes a reduced complement of 112 quadrature points (g-points) to calculate radiative transfer across the 14 spectral bands, which is half of the 224 g-points used in RRTM SW, to enhance computational performance with little impact on accuracy. The number of g-points needed within each band varies depending on the strength and complexity of the absorption in each spectral interval. Total uxes are accurate to within 1-2 W/m2 relative to the standard RRTM SW (using DISORT with 16 streams) in clear sky and in the presence of aerosols and within 6 W/m2 in overcast sky. RRTM SW with DISORT is itself accurate to within 2 W/m2 of the data-validated multiple scattering model, CHARTS [Moncet and Clough, 1997]. Input absorption coecient data for the k-distributions used by RRTMG are obtained directly from the line-by-line radiation model LBLRTM [Clough et al., 2005]. 166

3355

3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377

Table 4.6: RRTMG SW spectral band boundaries and the solar irradiance in each band. Band Index 1 2 3 4 5 6 7 8 9 10 11 12 13 14
3378 3379 3380 3381 3382 3383 3384 3385 3386 3387

Band Band Band Band Solar Min Max Min Max Irradiance (m) (m) (cm1 ) (cm1 ) (W/m2 ) 3.077 3.846 2600 3250 12.11 2.500 3.077 3250 4000 20.36 2.150 2.500 4000 4650 23.73 1.942 2.150 4650 5150 22.43 1.626 1.942 5150 6150 55.63 1.299 1.626 6150 7700 102.93 1.242 1.299 7700 8050 24.29 0.778 1.242 8050 12850 345.74 0.625 0.778 12850 16000 218.19 0.442 0.625 16000 22650 347.20 0.345 0.442 22650 29000 129.49 0.263 0.345 29000 38000 50.15 0.200 0.263 38000 50000 3.08 3.846 12.195 820 2600 12.89

RRTMG shortwave utilizes McICA, the Monte-Carlo Independent Column Approximation, to represent sub-grid scale cloud variability such as cloud fraction and cloud overlap. An external sub-column generator is used to dene the stochastic cloud arrays used by the McICA technique. The Kurucz solar source function is used in the shortwave model, which assumes a total solar irradiance (TSI) at the top of the atmosphere of 1368.22 W/m2 . However, this value is scaled in each spectral band through the specication of a time-varying solar spectral irradiance as discussed below. The TSI assumed in each RRTMG shortwave band is listed in the table below, along with the spectral band boundaries in m and wavenumbers. Shortwave radiation is only calculated by RRTMG when the cosine of the zenith angle is larger than zero, that is, when the sun is above the horizon. Longwave Radiative Transfer The infrared spectrum in RRTMG is divided into 16 longwave bands that extend over the spectral range from 3.1 m to 1000.0 m (10 to 3250 cm1 ). The band boundaries are listed in the table below. The model calculates molecular, cloud and aerosol absorption and emission. Scattering eects are not presently included. Molecular sources of absorption are H2O, CO2, O3, N2O, CH4, O2, N2 and the halocarbons CFC-11 and CFC-12. CFC-11 is specied by CAM5 as a weighed sum of multiple CFCs (other than CFC-12). The water vapor continuum is treated with the CKD v2.4 continuum model. For completeness, band 16 includes a small adjustment to add the infrared contribution from the spectral interval below 3.1 m. The longwave version of RRTMG [Iacono et al., 2008, 2003, 2000] used in CAM5 has been modied from RRTM LW [Mlawer et al., 1997] to enhance its computational eciency with minimal eect on the accuracy. This includes a reduction in the total number of g-points from 256 to 140. The number of g-points used within each band varies depending on the strength 167

3388

3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400

Table 4.7: RRTMG LW spectral band boundaries. Band Index 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419

Band Band Band Band Min Max Min Max 1 (m) (m) (cm ) (cm1 ) 28.57 1000.0 10 350 20.00 28.57 350 500 15.87 20.00 500 630 14.29 15.87 630 700 12.20 14.29 700 820 10.20 12.20 820 980 9.26 10.20 980 1080 8.47 9.26 1080 1180 7.19 8.47 1180 1390 6.76 7.19 1390 1480 5.56 6.76 1480 1800 4.81 5.56 1800 2080 4.44 4.81 2080 2250 4.20 4.44 2250 2380 3.85 4.20 2380 2600 3.08 3.85 2600 3250

and complexity of the absorption in each band. Fluxes are accurate to within 1.0 W/m2 at all levels, and cooling rate generally agrees within 0.1 K/day in the troposphere and 0.3 K/day the stratosphere relative to the line-by-line radiative transfer model, LBLRTM [Clough et al., 2005; Clough and Iacono, 1995]. Input absorption coecient data for the k-distributions used by RRTMG are obtained directly from LBLRTM. This model also utilizes McICA, the Monte-Carlo Independent Column Approximation [Pincus and Morcrette, 2003], to represent sub-grid scale cloud variability such as cloud fraction and cloud overlap. An external sub-column generator is used to dene the stochastic cloud arrays needed by the McICA technique. Within the longwave radiation model, the surface emissivity is assumed to be 1.0. However, the radiative surface temperature used in the longwave calculation is derived with the StefanBoltzmann relation from the upward longwave surface ux that is input from the land model. Therefore, this value may include some representation of surface emissivity less than 1.0 if this condition exists in the land model. RRTMG longwave also provides the capability of varying the surface emissivity within each spectral band, though this feature is not presently utilized. Longwave radiative transfer is performed over a single (diusivity) angle (secant =1.66) for one upward and one downward calculation. RRTMG includes an accuracy adjustment in proles with very high water vapor that slightly varies the diusivity angle in some bands as a function of total column water vapor.

168

3420

4.10.4

Surface Radiative Properties

3421 3422 3423 3424 3425 3426 3427 3428

For the shortwave, the surface albedoes are specied at every grid point at every time step. The albedoes are partitioned for the spectral ranges [2.0, 0.7]m and [0.7,12.0]m. In addition they are partitioned between the direct and diuse beam. In the longwave, the surface is assumed to have an emissivity of 1.0 within the radiation model. However, the radiative surface temperature used in the longwave calculation is derived with the Stefan-Boltzmann relation from the upward longwave surface ux that is input from the surface models. Therefore, this value may include some representation of surface emissivity less than 1.0, if this condition exists in surface models (e.g. the land model).

3429

4.10.5

Time Sampling

3430 3431

Both the shortwave and longwave radiation is computed at hourly intervals by default. The heating rates and uxes are assumed to be constant between time steps.

3432

4.10.6

Diurnal Cycle and Earth Orbit

In CAM 5.0, the diurnal cycle and earth orbit is computed using the method of [Berger, 1978]. Using this formulation, the insolation can be determined for any time within 106 years of 1950 AD. The insolation at the top of the model atmosphere is given by SI = S0 2 cos ,
3433 3434 3435 3436

(4.225)

where S0 is the solar constant, is the solar zenith angle, and 2 is the distance factor (square of the ratio of mean to actual distance that depends on the time of year). A time series of the solar spectral irradiance at 1 a.u. for 1870-2100 based upon [Wang et al., 2005] is included with the standard model and is in section 4.10.7. We represent the annual and diurnal cycle of solar insolation with a repeatable solar year of exactly 365 days and with a mean solar day of exactly 24 hours, respectively. The repeatable solar year does not allow for leap years. The expressions dening the annual and diurnal variation of solar insolation are: cos = sin sin cos cos cos(H) = arcsin(sin sin ) 1 e2 = 1 + e cos( ) = + (4.226) (4.227) (4.228) (4.229)

169

where H e = = = = = = = = = latitude in radians solar declination in radians hour angle of sun during the day obliquity true longitude of the earth relative to vernal equinox eccentricity factor longitude of the perihelion + 180 longitude of perihelion based on the xed equinox general precession

(4.230)

The hour angle H in the expression for cos depends on the calendar day d as well as model longitude: H = 2 d + , (4.231) 360
3437 3438 3439 3440 3441

where = model longitude in degrees starting from Greenwich running eastward. Note that the calendar day d varies continuously throughout the repeatable year and is updated every model time step. The values of d at 0 GMT for January 1 and December 31 are 0 and 364, respectively. This would mean, for example, that a model calendar day d having no fraction (such as 182.00) would refer to local midnight at Greenwich, and to local noon at the date line (180 longitude). The obliquity may be approximated by an empirical series expansion of solutions for the Earths orbit
47

= +
j=1
3442 3443

Aj cos (fj t + j )

(4.232)

where Aj , fj , and j are determined by numerical tting. The term = 23.320556, and t is the time (in years) relative to 1950 AD. Since the series expansion for the eccentricity e is slowly convergent, it is computed using e= (e cos )2 + (e sin )2 (4.233)

The terms on the right-hand side may also be written as empirical series expansions: e cos sin
19

=
j=1

Mj

cos sin

(gj t + j )

(4.234)

where Mj , gj , and j are estimated from numerical tting. Once these series have been computed, the longitude of perihelion is calculated using = arctan e sin e cos (4.235)

170

The general precession is given by another empirical series expansion

78

= t+ +
j=1
3444 3445

Fj sin fj t + j

(4.236)

where = 50.439273, = 3.392506, and Fj , fj , and j are estimated from the numerical solution for the Earths orbit. The calculation of requires rst determining two mean longitudes for the orbit. The mean longitude m0 at the time of the vernal equinox is :

m0 = 2

e e3 + 2 8 2 e 4 e3 + 8

(1 + ) sin( ) 1 + 2 1 + 3 sin(2 ) sin(3 ) (4.237)

where =

2 (d dve ) (4.238) 365 where dve = 80.5 is the calendar day for the vernal equinox at noon on March 21. The true longitude is then given by: m = m0 + = m + + 2e e3 4 sin(m )

1 e2 . The mean longitude is

3446 3447 3448 3449 3450 3451 3452

5 e2 sin [2(m )] (4.239) 4 13 e3 + sin [3(m )] 12 The orbital state used to calculate the insolation is held xed over the length of the model integration. This state may be specied in one of two ways. The rst method is to specify a year for computing t. The value of the year is held constant for the entire length of the integration. The year must fall within the range of 1950 106. The second method is to specify the eccentricity factor e, longitude of perihelion 180, and obliquity . This set of values is sucient to specify the complete orbital state. Settings for AMIP II style integrations under 1995 AD conditions are = 23.4441, e = 0.016715, and 180 = 102.7.

3453

4.10.7

Solar Spectral Irradiance

3454 3455 3456 3457 3458 3459

The reference spectrum assumed by RRTMG is the Kurucz spectrum. CAM 5.0 species the solar spectral irradiance in a le, based on the work of Lean [Wang et al., 2005]. The Kurucz spectrum can be seen in gure 4.5. The Lean data seen in gure 4.6 is time-varying and the graphed values are an average over one solar cycle. These two spectra postulate dierent values of the total solar irradiance. A graph of the relative dierence between them can be seen in gure 4.7. 171

Solar Irradiance Kurucz Lean Total 1368.60 1366.96 In RRTMG bands 1368.14 1366.39 > 12195 nm 0.46 0.46 [120, 200] nm 0 0.11 EUV 0 0.0047

RRTMG Band Index 14 1 2 3 4 5 6 7 8 9 10 11 12 13

high , low , Kurucz Lean Lean Relative Lean(t) Max % Lean(t) Max nm nm W/m2 W/m2 - Kurucz % Variation Flux 12195 3846 12.79 12.78 -0.01 -0.08 0.16 0.020 3846 3077 12.11 11.99 -0.12 -1.00 0.02 0.003 3077 2500 20.36 20.22 -0.14 -0.69 0.03 0.007 2500 2151 23.73 23.49 -0.24 -1.02 0.02 0.005 2151 1942 22.43 22.17 -0.26 -1.17 0.01 0.003 1942 1626 55.63 55.61 -0.02 -0.04 0.02 0.011 1626 1299 102.9 102.9 0.0 0. 0.02 0.019 1299 1242 24.29 24.79 0.50 2.06 0.04 0.011 1242 778 345.7 348.9 3.2 0.93 0.06 0.226 778 625 218.1 218.2 0.1 0.05 0.11 0.238 625 441 347.2 344.9 -2.3 -0.67 0.13 0.463 441 345 129.5 130.0 0.5 0.39 0.26 0.340 345 263 50.15 47.41 -2.74 -5.78 0.45 0.226 263 200 3.120 3.129 0.009 0.29 4.51 0.141

Table 4.8: Band-level ratio of Solar Irradiances, based on average of one solar cycle

172

Figure 4.5: Kurucz spectrum. ssf in W/m2 /nm. Source Data: AER. Range from [20, 20000] nm.
3460 3461 3462 3463 b The heating in each band b is scaled by the ratio, Lean(t)b , where Kuruczb is assumed by Kurucz RRTMG as specied in table 4.8 , and Lean(t)b is the solar irradiance specied by the timedependent solar spectral irradiance le. Lean(t)14 includes the Lean irradiance longward of 12195 nm to capture irradiance in the very far infrared.

173

Figure 4.6: Lean spectrum. Average over 1 solar cycle, May 1, 1996 to Dec 31, 2006. Source Data: Marsh. ssf in W/m2 /nm. Range from [120, 99975] nm.

174

Figure 4.7: Relative dierence, marked with vertical lines.

LeanKurucz .5(Lean+Kurucz)

between spectra. RRTMG band boundaries are

175

3464

4.11

Surface Exchange Formulations

3465 3466 3467 3468 3469 3470 3471 3472 3473

The surface exchange of heat, moisture and momentum between the atmosphere and land, ocean or ice surfaces are treated with a bulk exchange formulation. We present a description of each surface exchange separately. Although the functional forms of the exchange relations are identical, we present the descriptions of these components as developed and represented in the various subroutines in CAM 5.0. The dierences in the exchange expressions are predominantly in the denition of roughness lengths and exchange coecients. The description of surface exchange over ocean follows from Bryan et al. [1996], and the surface exchange over sea ice is discussed in the sea-ice model documentation. Over lakes, exchanges are computed by a lake model embedded in the land surface model described in the following section.

3474

4.11.1

Land

3475 3476 3477 3478 3479 3480 3481

In CAM 5.0, the NCAR Land Surface Model (LSM) [Bonan, 1996] has been replaced by the Community Land Model CLM2 [Bonan et al., 2002]. This new model includes components treating hydrological and biogeochemical processes, dynamic vegetation, and biogeophysics. Because of the increased complexity of this new model and since a complete description is available online, users of CAM 5.0 interested in CLM should consult this documentation at http://www.cgd.ucar.edu/tss/clm/. A discussion is provided here only of the component of CLM which controls surface exchange processes. Land surface uxes of momentum, sensible heat, and latent heat are calculated from MoninObukhov similarity theory applied to the surface (i.e. constant ux) layer. The zonal x and meridional y momentum uxes (kg m1 s2 ), sensible heat H (W m2 ) and water vapor E (kg m2 s1 ) uxes between the surface and the lowest model level z1 are: us u1 ram vs v1 y = 1 (v w ) = 1 u2 (v1 /Va ) = 1 ram s 1 H = 1 cp (w ) = 1 cp u = 1 cp rah qs q1 E = 1 (w q ) = 1 u q = 1 raw x = 1 (uw ) = 1 u2 (u1 /Va ) = 1 ram = Va /u2 rah = (1 s )/u raw = (q1 qs )/u q (4.240) (4.241) (4.242) (4.243)

(4.244) (4.245) (4.246)

3482 3483 3484 3485 3486

where 1 , u1 , v1 , 1 and q1 are the density (kg m3 ), zonal wind (m s1 ), meridional wind (m s1 ), air potential temperature (K), and specic humidity (kg kg1 ) at the lowest model level. By denition, the surface winds us and vs equal zero. The symbol 1 represents temperature, and q1 is specic humidity at surface. The terms ram , rah , and raw are the aerodynamic resistances (s m1 ) for momentum, sensible heat, and water vapor between the lowest model level at height

176

3487 3488 3489 3490 3491 3492

z1 and the surface at height z0m + d [z0h + d]. Here z0m [z0h ] is the roughness length (m) for momentum [scalar] uxes, and d is the displacement height (m). For the vegetated fraction of the grid, s = Taf and qs = qaf , where Taf and qaf are the air temperature and specic humidity within canopy space. For the non-vegetated fraction, s = Tg and qs = qg , where Tg and qg are the air temperature and specic humidity at ground surface. These terms are described by Dai et al. [2001]. Roughness lengths and zero-plane displacement The aerodynamic roughness z0m is used for wind, while the thermal roughness z0h is used for heat and water vapor. In general, z0m is dierent from z0h , because the transfer of momentum is aected by pressure uctuations in the turbulent waves behind the roughness elements, while for heat and water vapor transfer no such dynamical mechanism exists. Rather, heat and water vapor must ultimately be transferred by molecular diusion across the interfacial sublayer. Over bare soil and snow cover, the simple relation from Zilitinkevich [1970] can be used [Zeng and Dickinson, 1998]: ln u z0m 0.45 z0m =a z0h a = 0.13 = 1.5 105m2 s1 Rn H Lv E = 0 (4.247) (4.248) (4.249)

3493

Over canopy, the application of energy balance (4.250)

3494 3495 3496 3497 3498 3499 3500 3501

(where Rn is the net radiation absorbed by the canopy) is equivalent to the use of dierent z0m versus z0h over bare soil, and hence thermal roughness is not needed over canopy [Zeng et al., 1998]. The roughness z0m is proportional to canopy height, and is also aected by fractional vegetation cover, leaf area index, and leaf shapes. The roughness is derived from the simple relationship z0m = 0.07 hc , where hc is the canopy height. Similarly, the zero-plane displacement height d is proportional to canopy height, and is also aected by fractional vegetation cover, leaf area index, and leaf shapes. The simple relationship d/hc = 2/3 is used to obtain the height. Monin-Obukhov similarity theory (1) Turbulence scaling parameters A length scale (the Monin-Obukhov length) L is dened by L= v u2 kgv (4.251)

3502

3503

where k is the von K`rman constant, and g is the gravitational acceleration. L > 0 indicates a stable conditions, L < 0 indicates unstable conditions, and L = applies to neutral conditions. The virtual potential temperature v is dened by v = 1 (1 + 0.61q1 ) = Ta 177 ps pl
R/cp

(1 + 0.61q1 )

(4.252)

where T1 and q1 are the air temperature and specic humidity at height z1 respectively, 1 is the atmospheric potential temperature, pl is the atmospheric pressure, and ps is the surface pressure. The surface friction velocity u is dened by u2 = [u w + v w ]1/2 The temperature scale and v and a humidity scale q are dened by = w /u v = q = w q /u
w v /u 2 2

(4.253)

(4.254) (4.255) (4.256)

= + 0.61q
3504

(w + 0.61w q )/u

(where the mean temperature serves as a reference temperature in this linearized form of v ). The stability parameter is dened as z1 d , L 2. The scalar wind speed is dened as = , if
1/3

(4.257)

with the restriction that 100

2 2 Va2 = u2 + v1 + Uc 1

(4.258) 0 (stable) (4.259) , if < 0 (unstable) .

0.1 ms1 Uc =
3505 3506 3507 3508

w = zi gv v u

Here w is the convective velocity scale, zi is the convective boundary layer height, and = 1. The value of zi is taken as 1000 m (2) Flux-gradient relations [Zeng et al., 1998] The ux-gradient relations are given by: k(z1 d) = h () z k(z1 d) q = q () q z h = q m () = h () = (1 16)1/4 for < 0 1 + 5 for 0 < < 1 (1 16)1/2 for < 0 1 + 5 for 0 < < 1 (4.260) (4.261) (4.262) (4.263) (4.264)

Under very unstable conditions, the ux-gradient relations are taken from Kader and Yaglom [1990]: m = 0.7k 2/3 ()1/3 h = 0.9k
4/3

(4.265) (4.266)

()

1/3

178

3509 3510 3511

To ensure the functions m () and h () are continuous, the simplest approach (i.e., without considering any transition regions) is to match the above equations at m = 1.574 for m () and h = 0.465 for h () . Under very stable conditions (i.e., > 1 ), the relations are taken from Holtslag et al. [1990]: m = h = 5 + (4.267)

3512

(3) Integral forms of the ux-gradient relations Integration of the wind prole yields: Va = fM () = fM () = fM () = fM () = u fM () k m L m (m ) + 1.14[()1/3 (m )1/3 ] ln z0m z0m z1 d m () + m , ln z0m L z1 d ln + 5 , z0m L + 5 + [5 ln() + 1] , ln z0m (4.268) , < m = 1.574 m < < 0 0< <1 >1 (4.268a) (4.268b) (4.268c) (4.268d)

Integration of the potential temperature prole yields: 1 s = fT () = fT () = fT () = fT () = fT () k h L h (h ) + 0.8[(h )1/3 ()1/3 ] ln z0h z0h z1 d h () + h , ln z0h L z1 d ln + 5 , z0h L ln + 5 + [5 ln() + 1] , z0h (4.269) , < h = 0.465 h < < 0 0< <1 >1 (4.269a) (4.269b) (4.269c) (4.269d)

The expressions for the specic humidity proles are the same as those for potential temperature except that (1 s ), and z0h are replaced by (q1 qs ), q and z0q respectively. The stability functions for < 0 are m = 2 ln 1+ 2 + ln 1 + 2 2 1 + 2 2 2 tan1 + 2 (4.270) (4.271)

h = q = 2 ln where

= (1 16)1/4 179

(4.272)

3513 3514 3515 3516 3517 3518 3519 3520 3521

Note that the CLM code contains extra terms involving z0m /, z0h /, and z0q / for completeness. These terms are very small most of the time and hence are omitted in Eqs. 4.268 and 4.269. In addition to the momentum, sensible heat, and latent heat uxes, land surface albedos and upward longwave radiation are needed for the atmospheric radiation calculations. Surface albedos depend on the solar zenith angle, the amount of leaf and stem material present, their optical properties, and the optical properties of snow and soil. The upward longwave radiation is the dierence between the incident and absorbed uxes. These and other aspects of the land surface uxes have been described by Dai et al. [2001].

3522

4.11.2

Ocean

The bulk formulas used to determine the turbulent uxes of momentum (stress), water (evaporation, or latent heat), and sensible heat into the atmosphere over ocean surfaces are ( , E, H) = A | v| (CD v, CE q, Cp CH ),
3523 3524 3525 3526 3527

(4.273)

where A is atmospheric surface density and Cp is the specic heat. Since CAM 5.0 does not allow for motion of the ocean surface, the velocity dierence between surface and atmosphere is v = v A , the velocity of the lowest model level. The potential temperature dierence is = A Ts , where Ts is the surface temperature. The specic humidity dierence is q = qA qs (Ts ), where qs (Ts ) is the saturation specic humidity at the sea-surface temperature. In (4.273), the transfer coecients between the ocean surface and the atmosphere are computed at a height ZA and are functions of the stability, : C(D,E,H) = ln
2

ZA Z0m

ln

ZA Z0(m,e,h)

(m,s,s)

(4.274)

where = 0.4 is von Krmns constant and Z0(m,e,h) is the roughness length for momentum, a a evaporation, or heat, respectively. The integrated ux proles, m for momentum and s for scalars, under stable conditions ( > 0) are m () = s () = 5. For unstable conditions ( < 0), the ux proles are m () =2 ln[0.5(1 + X)] + ln[0.5(1 + X 2 )] 2 tan1 X + 0.5, s () =2 ln[0.5(1 + X 2 )], X =(1 16)1/4 . The stability parameter used in (4.275)(4.278) is = g ZA u2 Q + 1 v ( + qA ) , (4.279) (4.275)

(4.276) (4.277) (4.278)

180

where the virtual potential temperature is v = A (1 + qA ); qA and A are the lowest level atmospheric humidity and potential temperature, respectively; and = 0.606. The turbulent velocity scales in (4.279) are u =CD | v|, | v| (Q , ) =C(E,H) ( q, ). u
1/2

(4.280)

Over oceans, Z0e = 9.5 105 m under all conditions and Z0h = 2.2 109 m for > 0, Z0h = 4.9 105 m for 0, which are given in Large and Pond [1982]. The momentum roughness length depends on the wind speed evaluated at 10 m as Zom = 10 exp U10 = UA 1 +
3528 3529 3530 3531 3532 3533

c4 + c5 + c6 U10 U10 ZA m 10

,
1

N C10 ln

(4.281)

where c4 = 0.0027 m s1 , c5 = 0.000142, c6 = 0.0000764 m1 s, and the required drag coecient 1 N at 10-m height and neutral stability is C10 = c4 U10 + c5 + c6 U10 as given by Large et al. [1994]. The transfer coecients in (4.273) and (4.274) depend on the stability following (4.275) (4.278), which itself depends on the surface uxes (4.279) and (4.280). The transfer coecients also depend on the momentum roughness, which itself varies with the surface uxes over oceans (4.281). The above system of equations is solved by iteration.

3534

4.11.3

Sea Ice

3535 3536

The uxes between the atmosphere and sea ice are described in detail in the sea-ice model documentation.

181

3537

4.12

Dry Adiabatic Adjustment

If a layer is unstable with respect to the dry adiabatic lapse rate, dry adiabatic adjustment is performed. The layer is stable if T T < . (4.282) p p In nitedierence form, this becomes where Tk+1 Tk < C1k+1(Tk+1 + Tk ) + , C1k+1 = (pk+1 pk ) . 2pk+1/2 (4.283) (4.284)

If there are any unstable layers in the top three model layers, the temperature is adjusted so that (4.283) is satised everywhere in the column. The variable represents a convergence criterion. The adjustment is done so that sensible heat is conserved, cp (Tk pk + Tk+1 pk+1 ) = cp (Tk pk + Tk+1 pk+1 ), and so that the layer has neutral stability: Tk+1 Tk = C1k+1 (Tk+1 + Tk ) . (4.286) (4.285)

As mentioned above, the hats denote the variables after adjustment. Thus, the adjusted temperatures are given by Tk+1 = and where pk pk+1 Tk + Tk+1 , pk+1 + pk C2k+1 pk+1 + pk C2k+1 (4.287) (4.288) (4.289)

Tk = C2k+1Tk+1 , C2k+1 = 1 C1k+1 . 1 + C1k+1

Whenever the two layers undergo dry adjustment, the moisture is assumed to be completely mixed by the process as well. Thus, the specic humidity is changed in the two layers in a conserving manner to be the average value of the original values, qk+1 = qk = (qk+1 pk+1 + qk pk )/(pk+1 + pk ).
3538 3539 3540 3541 3542 3543 3544 3545 3546 3547

(4.290)

The layers are adjusted iteratively. Initially, = 0.01 in the stability check (4.283). The column is passed through from k = 1 to a user-speciable lower level (set to 3 in the standard model conguration) up to 15 times; each time unstable layers are adjusted until the entire column is stable. If convergence is not reached by the 15th pass, the convergence criterion is doubled, a message is printed, and the entire process is repeated. If exceeds 0.1 and the column is still not stable, the model stops. As indicated above, the dry convective adjustment is only applied to the top three levels of the standard model. The vertical diusion provides the stabilizing vertical mixing at other levels. Thus, in practice, momentum is mixed as well as moisture and potential temperature in the unstable case. 182

3548

4.13

Prognostic Greenhouse Gases

3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583

The principal greenhouse gases whose longwave radiative eects are included in CAM 5.0 are H2 O, CO2 , O3 , CH4 , N2 O, CFC11, and CFC12. The prediction of water vapor is described elsewhere in this chapter, and CO2 is assumed to be well mixed. Monthly O3 elds are specied as input, as described in chapter 6. The radiative eects of the other four greenhouse gases (CH4 , N2 O, CFC11, and CFC12) may be included in CAM 5.0 through specied concentration distributions [Kiehl et al., 1998] or prognostic concentrations [Boville et al., 2001]. The specied distributions are globally uniform in the troposphere. Above a latitudinally and seasonally specied tropopause height, the distributions are zonally symmetric and decrease upward, with a separate latitude-dependent scale height for each gas. Prognostic distributions are computed following Boville et al. [2001]. Transport equations for the four gases are included, and losses have been parameterized by specied zonally symmetric loss frequencies: q/t = (y, z, t)q. Monthly averaged loss frequencies, , are obtained from the two-dimensional model of Garcia and Solomon [1994]. We have chosen to specify globally uniform surface concentrations of the four gases, rather than their surface uxes. The surface sources are imperfectly known, particularly for CH4 and N2 O in preindustrial times. Even given constant sources and reasonable initial conditions, obtaining equilibrium values for the loading of these gases in the atmosphere can take many years. CAM 5.0 was designed for tropospheric simulation with relatively coarse vertical resolution in the upper troposphere and lower stratosphere. It is likely that the rate of transport into the stratosphere will be misrepresented, leading to erroneous loading and radiative forcing if surface uxes are specied. Specifying surface concentrations has the advantage that we do not need to worry much about the atmospheric lifetime. However, we cannot examine observed features such as the interhemispheric gradient of the trace gases. For climate change experiments, the specied surface concentrations are varied but the stratospheric loss frequencies are not. Oxidation of CH4 is an important source of water vapor in the stratosphere, contributing about half of the ambient mixing ratio over much of the stratosphere. Although CH4 is not generally oxidized directly into water vapor, this is not a bad approximation, as shown by Le Texier et al. [1988]. In CAM 5.0, it is assumed that the water vapor (volume mixing ratio) source is twice the CH4 sink. This approach was also taken by Mote et al. [1993] for middle atmosphere studies with an earlier version of the CCM. This part of the water budget is of some importance in climate change studies, because the atmospheric CH4 concentrations have increased rapidly with time and this increase is projected to continue into the next century (e.g., Alcamo et al. [1995]) The representation of stratospheric water vapor in CAM 5.0 is necessarily crude, since there are few levels above the tropopause. However, the model is capable of capturing the main features of the CH4 and water distributions.

183

184

3584

Chapter 5 Extensions to CAM

3585

185

3586

5.1
5.1.1

Chemistry
Introduction

3587

3588 3589 3590 3591 3592

In this Section, we provide a description of the neutral constituent chemical processes adopted in CAM-chem and WACCM4.0. This section will contain a description of constituent: 1) surface boundary conditions; 2) numerical algorithms used to solve the corresponding set of ordinary dierential equations. 3) gas-phase and heterogeneous reactions; and 4) wet and dry deposition removal processes;

3593

5.1.2

Emissions

3594 3595 3596 3597 3598 3599 3600

Surface emissions are used in as a ux boundary condition for the diusion equation of all applicable tracers in the planetary boundary-layer scheme. The surface ux les used in the released version are discussed in Lamarque et al. [2010b] and conservatively remapped from their original resolution (monthly data available every decade at 0.5x0.5) to (monthly data every year at 1.9x2.5). In addition, natural emissions from MOZART-4 were added to all relevant species. Finally, isoprene emissions are calculated interactively (within CLM (point 2) above), leading to approximately (emissions depend on light and temperature) 500 Tg/year. Anthro. x x Natural Interactive x x x x

CH2 O CO DMS ISOP NO SO2

x x

Table 5.1: Surface uxes for CAM4 superfast chemistry.

3601 3602 3603 3604 3605

Additional emissions (volcanoes and aircraft) are included as three-dimension arrays, conservatively-remapped to the CAM-chem grid. The volcanic emission are from Dentener et al. [2006b] and the aircraft (NO2 ) emissions are from Lamarque et al. [2010b]. In the case of volcanic emissions (SO2 and SO2 ), an assumed 2% of the total sulfur mass is directly released as SO2 .

3606

5.1.3

Lower boundary conditions

3607 3608 3609 3610 3611 3612

The concentration of specic long-lived chemical tracers (methane and longer lifetimes, in addition to hydrogen and methyl bromide) are xed in the lowest model layer using the reconstructed concentrations (CMIP5 recommended data) available from http://www.iiasa.ac.at/web-apps/tnt/RcpDb/dsd?Action=htmlpage&page=download. As this dataset only provides annual average values, a specied seasonal cycle (based on present-day observations) is added to methane and carbon dioxide.

186

3613

5.1.4

Lightning

3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625

The lightning parameterization diers slightly from that used in MOZART-2 [Horowitz et al., 2003]. The lightning strength still depends on cloud top height, with a stronger dependence over land than ocean [Price et al., 1997a]. The denition of ocean grid boxes has been rened to include only boxes surrounded by ocean, so that the land parameterization is extended one grid box beyond the continents Price and Rind [1992]. Flash frequency is determined by area, not grid box. The vertical distribution of NO emissions has been modied from that given by Pickering et al. [1998], to have a reduced proportion of the emissions emitted near the surface, similar to that used by DeCaria et al. [2006]. In addition, the strength of intra-cloud (IC) lightning strikes is assumed to be equal to cloud-to-ground strikes, as recommended by Ridley et al. [2005]. Because the lightning NO source is very resolution-dependent, it can be scaled under nonstandard resolutions to a produce 5-7 Tg(N)/year globally.

3626

5.1.5

Dry deposition

3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648

The list of species aected by dry deposition is subject to user selection. Dry deposition velocities are computed interactively (i.e. are inuenced by variations in temperature, solar insolation and precipitation), based on the resistance-based parameterization of Wesely [1989], Walmsley and Wesely [1996], and Wesely and Hicks [2000]. The calculation of surface resistances over land uses the vegetation distribution as dened in CLM. In addition, as the parameters in the Wesely [1989] parameterization are season-dependent (to take into account the specic role of changes in vegetation cover), a season index is computed from the monthlyaveraged leaf-area index input le to CLM. In the case of CLM-CN or DGVM (where the LAI is prognostic), this le is still read and seasonality is still dened accordingly. Deposition over the ocean is computed separately in CAM (but using the same formulation), with the overall deposition velocity computed as the weighted (by the land/ocean fraction) mean between the land and ocean values. The deposition velocity calculation has been extended from the aforementioned references to take into account special cases for CO, H2 and PAN. In the case of CO and H2 ,surface uptake is caused by the oxidation by soil bacteria or enzymes [Yonemura et al., 2000]. This has been parameterized using the approach of Sanderson et al. [2003], which denes the deposition velocity by a linear or quadratic function in soil moisture content (or its logarithm), depending on the land cover type. In the case of PAN, new laboratory experiments have indicated a strong uptake of PAN by leaves [Teklemariam and Sparks, 2004]. Using the results from that study, we have included a leaf uptake of PAN that is vegetation-dependent, based on Sparks et al. (2003). Results from this parameterization agreed with observations during a eld experiment [Turnipseed et al., 2006].

3649

5.1.6

Wet removal

3650 3651 3652

Wet deposition for gas-phase species is represented as a rst-order loss process within the chemistry operator, with loss rates computed based on the large-scale and convective precip- itation rates in CAM. Soluble species (H2 O2 , HNO3 , CH2 O, SO2 ) undergo wet removal by in-

187

3653 3654 3655 3656

cloud scavenging, using the parameterization of Giorgi and Chameides [1985] based on their temperature-dependent eective Henrys law constants. In addition, highly soluble species (HNO3 , H2 O2 ) are also removed by below-cloud washout, using the formulation described in detail by Brasseur et al. [1998].

3657

5.1.7

Photolytic Approach (Neutral Species)

3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687

The calculation of the photolysis coecients is divided into two regions: (1) 120 nm to 200 nm (33 wavelength intervals); (2) 200 nm to 750 nm (67 wavelength intervals). The total photolytic rate constant (J) for each absorbing species is derived during model execution by integrating the product of the wavelength dependent exo-atmospheric ux (Fexo); the atmospheric transmission function (or normalized actinic ux) (NA ), which is unity at the top of atmosphere in most wavelength regions; the molecular absorption cross-section (); and the quantum yield (). The exoatmospheric ux over these wavelength intervals can be specied from observations and varied over the 11-year solar sunspot cycle (see section 5.3.6). The wavelength-dependent transmission function is derived as a function of the model abundance of ozone and molecular oxygen. For wavelengths greater than 200 nm a normalized ux lookup table (LUT) approach is used, based on the 4-stream version of the Stratosphere, Troposphere, Ultraviolet (STUV) radiative transfer model (S. Madronich, personal communication). The transmission function is interpolated from the LUT as a function of altitude, column ozone, surface albedo, and zenith angle. The temperature and pressure dependences of the molecular cross sections and quantum yields for each photolytic process are also represented by a LUT in this wavelength region. At wavelengths less than 200 nm, the wavelength-dependent cross section and quantum yields for each species are specied and the transmission function is calculated explicitly for each wavelength interval. There are two exceptions to this approach. In the case of J(NO) and J(O2 ), detailed photolysis parameterizations are included inline. In the Schumann-Runge Band region (SRBs), the parameterization of NO photolysis in the -bands is based on Minschwaner and Siskind [1993]. This parameterization includes the eect of self-absorption and subsequent attenuation of atmospheric transmission by the model-derived NO concentration. For J(O2 ), the SRB and Lyman-alpha parameterizations are based on Koppers and Murtagh [1996] and Chabrillat and Kockarts [1997], respectively. The photolytic reactions included in WACCM4.0 are listed in Table 4. In most all cases the photolytic rate constants are taken from JPL02-25 [Sander, S. P., et al., 2003]. Exceptions to this condition are described in the comment section for any given reaction. In addition, tropospheric photolysis rates can be computed interactively following Tie et al. [1992]. Users interested in using this capability have to contact the Chemistry-CLimate Working Group Liaison as this is an unsupported option.

3688

5.1.8

Numerical Solution Approach

Chemical and photochemical processes are expressed by a system of time-dependent ordinary dierential equations at each point in the spatial grid, of the following form: dy = P (y, t) L(y, t) y dt (5.1)

188

y(t) = {yi (t)} i = 1, 2, . . . , N where y is the vector of all solution variables (chemical species), N is the number of variables in the system, and yi represents the ith variable. P and L represent the production and loss rates, which are, in general, non-linear functions of the yi . This system of equations is solved via two algorithms: an explicit forward Euler method:
n+1 n yi = yi + t fi (tn , y n )

(5.2)

in the case of species with long lifetimes and weak forcing terms (e.g., N2 O), and a more robust implicit backward Euler method:
n+1 n yi = yi + t fi (tn+1 , y n+1)

(5.3)

for species that comprise asti system with short lifetimes and strong forcings (e.g., OH). Here n represents the time step index. Each method is rst order accurate in time and conservative. The overall chemistry time step, t = tn+1 tn , is xed at 30 minutes. Preprocessing software requires the user to assign each solution variable, yi, to one of the solution schemes. The discrete analogue for methods (5.2) and (5.3) above results in two systems of algebraic equations at each grid point. The solution to these algebraic systems for equation (5.2) is straightforward (i.e., explicit). The algebraic system from the implicit method (5.3) is quadratically non-linear. This system can be written as: G(y n+1 ) = y n+1 y n t f (tn+1 , y n+1 ) = 0 (5.4)

Here G is an N-valued, non-linear vector function, where N equals the number of species solved via the implicit method. The solution to equation (5.4) is solved with a Newton- Raphson iteration approach as shown below:
n+1 n+1 n+1 ym+1 = ym J G(ym ); m = 0, 1, . . . , M

(5.5)

Where m is the iteration index and has a maximum value of ten. The elements of the Jacobian matrix J are given by: Gi Jij = yj
3689 3690 3691 3692 3693 3694 3695 3696 3697

The iteration and solution of equation (5.5) is carried out with a sparse matrix solution algorithm. This process is terminated when the given solution variable changes in a relative measure by less than a prescribed fractional amount. This relative error criterion is set on a species by species basis, and is typically 0.001; however, for some species (e.g., O3 ), where a tighter error criterion is desired, it is set to 0.0001. If the iteration maximum is reached (for any species) before the error criterion is met, the time step is cut in half and the solution to equation (5.5) is iterated again. The time step can be reduced ve times before the solution is accepted. This approach is based on the work of Sandu et al. [1996] and Sandu et al. [1997]; see also Brasseur et al. [1999].

189

3698

5.2
5.2.1

Superfast Chemistry
Chemical mechanism

3699

3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712

The super-fast mechanism was developed for coupled climate model usage, and is based on an updated version of the full non-methane hydrocarbon eects (NMHC) chemical mechanism for the troposphere and stratosphere used in the Lawrence Livermore National Laboratory oline 3D global chemistry-transport model (IMPACT) [Rotman et al., 2004]. The super-fast mechanism includes 15 photochemically active trace species (O3 , OH, HO2 , H2 O2 , NO, NO2 , HNO3 , CO, CH2 O, CH3 O2 , CH3 OOH, DMS, SO2 , SO4 , and C5 H8 ) that allow us to calculate the major terms by which global change operates in tropospheric ozone and sulfate photochemistry. The families selected are Ox, HOx, NOy, the CH4 oxidation suite plus isoprene (to capture the main NMHC eects), and a group of sulfur species to simulate natural and anthropogenic sources leading to sulfate aerosol. Sulfate aerosols is handled following Tie et al. [2005]. In this scheme, CH4 concentrations are read in from a le and uses CAM3.5 simulations Lamarque et al. [2010b]. The super-fast mechanism was validated by comparing the super-fast and full mechanisms in side-by-side simulations.

3713

5.2.2

LINOZ

3714 3715 3716 3717 3718 3719 3720 3721 3722 3723

Linoz is linearized ozone chemistry for stratospheric modeling [McLinden et al., 2000]. It calculates the net production of ozone (i.e., production minus loss) as a function of only three independent variables: local ozone concentration, temperature, and overhead column ozone). A zonal mean climatology for these three variables as well as the other key chemical variables such a total odd-nitrogen methane abundance is developed from satellite and other in situ observations. A relatively complete photochemical box model Prather [1992] is used to integrate the radicals to a steady state balance and then compute the net production of ozone. Small perturbations about the chemical climatology are used to calculate the coecients of the rstorder Taylor series expansion of the net production in terms of local ozone mixing ratio (f), temperature (T), and overhead column ozone (c). df df (P L) (P L) (f f o ) + (T T o ) f T o o

= (P L)o + +

(5.6)

(P L) (c co ) c o

3724 3725 3726 3727 3728 3729 3730 3731

The photochemical tendency for the climatology is denoted by (P L)o , and the climatology values for the independent variables are denoted by fo , co , and To , respectively. Including these four climatology values and the three partial derivatives, Linoz is dened by seven tables. Each table is specied by 216 atmospheric proles: 12 months by 18 latitudes (85o S to 85oN). For each prole, quantities are evaluated at every 2 km in pressure altitude from z = 10 to 58 km (z = 16 km log1 0 (1000/p)). These tables (calculated for each decade, 1850-2000 to take into account changes in CH4 and N2O) are automatically remapped onto the CAM-chem grid with the mean vertical properties for each CAM-chem level calculated as the mass-weighted average

190

3732 3733

of the interpolated Linoz proles. Equation (1) is implemented for the chemical tendency of ozone in CAM-chem.

3734

5.2.3

PSC ozone loss

3735 3736 3737 3738 3739 3740 3741 3742

In the superfast chemistry, we incorporate the PSCs parameterization scheme of Cariolle et al. [1990] when the temperature falls below 195 K and the sun is above the horizon at stratospheric altitudes. The O3 loss scales as the squared stratospheric chlorine loading (normalized by the 1980 level threshold). In this formulation PSC activation invokes a rapid e-fold of O3 based on a photochemical model, but only when t he temperature stays below the PSC threshold. The stratospheric chlorine loading (1850-2005) is input in the model using equivalent eective stratospheric chlorine (EESC) [Newman et al., 2007] table based on observed mixing ratios at the surface.

3743

5.2.4

Upper boundary condition

3744

The model top is considered a rigid lid (no ux across that boundary) for all chemical species.

191

3745

5.3

WACCM4.0 Physical Parameterizations

3746 3747 3748 3749 3750

In WACCM4.0, we extend the physical parameterizations used in CAM4 by adding constituent separation velocities to the molecular (vertical) diusion and modifying the gravity spectrum parameterization. Both of these parameterizations are present, but not used, in CAM4. In addition, we replace the CAM4 parameterizations for both solar and longwave radiation above 65 km, and add neutral and ion chemistry models.

3751

5.3.1

WACCM4.0 Domain and Resolution

WACCM4.0 has 66 vertical levels from the ground to 5.1106 hPa, as in the previous WACCM versions. As in CAM4, the vertical coordinate is purely isobaric above 100 hPa, but is terrain following below that level. At any model grid point, the local pressure p is determined by p(i, j, k) = A(k) p0 + B(k) ps (i, j)
3752 3753 3754 3755 3756

(5.7)

where A and B are functions of model level, k, only; p0 = 103 hPa is a reference surface pressure; and ps is the predicted surface pressure, which is a function of model longitude and latitude (indexed by i and j). The nite volume dynamical core uses locally material surfaces for its internal vertical coordinate and remaps (conservatively interpolates) to the hybrid surfaces after each time step. Within the physical and chemical parameterizations, a local pressure coordinate is used, as described by (5.7). However, in the remainder of this note we refer to the vertical coordinate in terms of log-pressure altitude p0 Z = H log . (5.8) p The value adopted for the scale height, H = 7 km, is representative of the real atmosphere up to 100 km, above that altitude temperature increases very rapidly and the typical scale height becomes correspondingly larger. It is important to distinguish Z from the geopotential height z, which is obtained from integration of the hydrostatic equation. In terms of log-pressure altitude, the model top level is found at Z = 140 km (z 150 km). It should be noted that the solution in the top 15-20 km of the model is undoubtedly aected by the presence of the top boundary. However, it should not be thought of as a sponge layer, since molecular diusion is a real process and is the primary damping on upward propagating waves near the model top. Indeed, this was a major consideration in moving the model top well above the turbopause. Considerable eort has been expended in formulating the upper boundary conditions to obtain realistic solutions near the model top and all of the important physical and chemical processes for that region have been included. The standard vertical resolution is variable; it is 3.5 km above about 65 km, 1.75 km around the stratopause (50 km), 1.1-1.4 km in the lower stratosphere (below 30 km), and 1.1 km in the troposphere (except near the ground where much higher vertical resolution is used in the planetary boundary layer). Two standard horizontal resolutions are supported in WACCM4.0: the 4 5 (latitude longitude) low resolution version has 72 longitude and 46 latitude points; the 1.9 2.5 medium resolution version has 96 longitude and 144 latitude points. A 0.9 1.25 high resolution 192

3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775

3776 3777 3778 3779 3780 3781 3782

version of WACCM4.0 has had limited testing, and is not yet supported, due to computational cost constraints. The 4 5 version has been used extensively for MLT studies, where it gives very similar results to the 1.9 2.5 version. However, caution should be exercised in using 4 5 results below the stratopause, since the meridional resolution may not be sucient to represent adequately the dynamics of either the polar vortex or synoptic and planetary waves. At all resolutions, the time step is 1800 s for the physical parameterizations. Within the nite volume dynamical core, this time step is subdivided as necessary for computational stability.

3783

5.3.2

Molecular Diusion and Constituent Separation

3784 3785 3786 3787 3788 3789 3790

The vertical diusion parameterization in CAM4 provides the interface to the turbulence parameterization, computes the molecular diusivities (if necessary) and nally computes the tendencies of the input variables. The diusion equations are actually solved implicitly, so the tendencies are computed from the dierence between the nal and initial proles. In WACCM4.0, we extend this parameterization to include the terms required for the gravitational separation of constituents of diering molecular weights. The formulation for molecular diusion follows Banks and Kockarts [1973] A general vertical diusion parameterization can be written in terms of the divergence of diusive uxes: 1 (u, v, q) = (Fu , Fv , Fq ) t z 1 s = FH + D t z (5.9) (5.10)

where s = cp T + gz is the dry static energy, z is the geopotential height above the local surface (does not include the surface elevation) and D is the heating rate due to the dissipation of resolved kinetic energy in the diusion process. The diusive uxes are dened as: Fu,v = Km FH Fq (u, v), z s t + KH H , = KH z q t = Kq + Kq q + sep ux. z (5.11) (5.12) (5.13)

t The viscosity Km and diusivities Kq,H are the sums of: turbulent components Km,q,H , which m dominate below the mesopause; and molecular components Km,q,H , which dominate above 120 km. The non-local transport terms q,H are given by the ABL parameterization and and the kinetic energy dissipation is u v 1 D Fu . (5.14) + Fv z z
3791 3792 3793

t The treatment of the turbulent diusivities Km,q,H , the energy dissipation D and the nonlocal transport terms H,q is described in the CAM 5.0 Technical Description and will be omitted here.

193

3794

Molecular viscosity and diusivity The empirical formula for the molecular kinematic viscosity is
m Km = 3.55 107 T 2/3 /,

(5.15)

and the molecular diusivity for heat is

m m K H = P r Km ,

(5.16)

where Pr is the Prandtl number and we assume Pr = 1 in WACCM4.0. The constituent diusivities are m Kq = T 1/2 Mw /, (5.17)
3795

where Mw is the molecular weight. Diusive separation velocities As the mean free path increases, constituents of dierent molecular weights begin to separate in the vertical. In WACCM4.0, this separation is represented by a separation velocity for each constituent with respect mean air. Since WACCM4.0 extends only into the lower thermosphere, we avoid the full complexity of the separation problem and represent mean air by the usual dry air mixture used in the lower atmosphere (Mw = 28.966) Banks and Kockarts [1973]. Discretization of the vertical diusion equations In CAM4, as in previous version of the CCM, (5.95.12) are cast in pressure coordinates, using dp = gdz, (5.18)

3796

3797 3798 3799 3800 3801

3802

and discretized in a time-split form using an Euler backward time step. Before describing the numerical solution of the diusion equations, we dene a compact notation for the discrete equations. For an arbitrary variable , let a subscript denote a discrete time level, with current step n and next step n+1 . The model has L layers in the vertical, with indexes running from top to bottom. Let k denote a layer midpoint quantity and let k denote the value at the interface above (below) k. The relevant quantities, used below, are then: k+ k k k+ k n+ n t k,l = = = = = = = = = = ( k + k+1 )/2, ( k1 + k )/2, k+ k , k+1 k , k k1 , (n + n+1 )/2, n+1 n , tn+1 tn , 1, k = l, 0, k = l. 194 k (1, 2, 3, ..., L 1) k (2, 3, 4..., L)

Like the continuous equations, the discrete equations are required to conserve momentum, total energy and constituents. Neglecting the nonlocal transport terms, the discrete forms of (5.95.10) are: k Fu,v,q n (u, v, q)k = g k t p k k n s FH = g k + Dk . t p For interior interfaces, 1 k L 1,
k+ Fu,v = k+ Fq,H =
3803 3804 3805 3806 3807

(5.19) (5.20)

g2 Km g2 Kq,H

k+ n

k+ (u, v)n+1 k+ p k+ (u, v)n+1 k+ . n k+ p

(5.21) (5.22)

L+ Surface uxes Fu,v,q,H are provided explicitly at time n by separate surface models for land, 1 ocean, and sea ice while the top boundary uxes are usually Fu,v,q,H = 0. The turbulent t diusion coecients Km,q,H and non-local transport terms q,H are calculated for time n by the turbulence model (identical to CAM4). The molecular diusion coecients, given by (5.155.17) are also evaluated at time n.

3808

Solution of the vertical diusion equations

t Neglecting the discretization of Km,q,H , D and q,H , a series of time-split operators is dened by (5.195.22). Once the diusivities (Km,q,H ) and the non-local transport terms (q,H ) have been determined, the solution of (5.195.22), proceeds in several steps.

3809 3810 3811

3812

1. update the bottom level values of u, v, q and s using the surface uxes; 2. invert (5.19) and (5.21) for u, vn+1 ; 3. compute D and use to update the s prole; 4. invert (5.19,5.20) and (5.22) for sn+1 and qn+1 Note that since all parameterizations in CAM4 return tendencies rather modied proles, the actual quantities returned by the vertical diusion are n (u, v, s, q)/t. Equations (5.195.22) constitute a set of four tridiagonal systems of the form
k+1 k k1 k Ak n+1 + B k n+1 C k n+1 = n ,

3813

3814

3815

3816 3817

(5.23)

where n indicates u, v, q, or s after updating from time n values with the nonlocal and boundary uxes. The super-diagonal (Ak ), diagonal (B k ) and sub-diagonal (C k ) elements of (5.23) are: Ak = Bk Ck 1 t g 2 2 K k p k+ p = 1 + Ak + C k , 1 t = k k g 2 2 K p p 195
k+ n

(5.24) (5.25)

k n

(5.26)

The solution of (5.23) has the form

k k1 n+1 = E k n+1 + F k ,

(5.27) (5.28)

or,
k+1 k n+1 = E k+1 n+1 + F k+1 .

Substituting (5.28) into (5.23),

k n+1 =

Ck k + Ak F k+1 k1 n+1 + n . B k Ak E k+1 B k Ak E k+1 Ck = , B k Ak E k+1 k n + Ak F k+1 , = B k Ak E k+1

(5.29)

Comparing (5.27) and (5.29), we nd E

k

L > k > 1, L > k > 1.

(5.30) (5.31)

Fk

The terms E k and F k can be determined upward from k = L, using the boundary conditions E L+1 = F L+1 = AL = 0.
k Finally, (5.29) can be solved downward for n+1 , using the boundary condition

(5.32)

C 1 = 0 E 1 = 0.
3818 3819 3820 3821 3822 3823

(5.33)

CCM1-3 used the same solution method, but with the order of the solution reversed, which k1 k+1 merely requires writing (5.28) for n+1 instead of n+1 . The order used here is particularly convenient because the turbulent diusivities for heat and all constituents are the same but their molecular diusivities are not. Since the terms in (5.30-5.31) are determined from the bottom upward, it is only necessary to recalculate Ak , C k , E k and 1/(B k Ak E k+1 ) for each constituent within the region where molecular diusion is important.

3824

5.3.3

Gravity Wave Drag

3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836

Vertically propagating gravity waves can be excited in the atmosphere where stably stratied air ows over an irregular lower boundary and by internal heating and shear. These waves are capable of transporting signicant quantities of horizontal momentum between their source regions and regions where they are absorbed or dissipated. Previous GCM results have shown that the large-scale momentum sinks resulting from breaking gravity waves play an important role in determining the structure of the large-scale ow. CAM4 incorporates a parameterization for a spectrum of vertically propagating internal gravity waves based on the work of Lindzen [1981], Holton [1982], Garcia and Solomon [1985] and McFarlane [1987]. The parameterization solves separately for a general spectrum of monochromatic waves and for a single stationary wave generated by ow over orography, following McFarlane [1987]. The spectrum is omitted in the standard tropospheric version of CAM4, as in previous versions of the CCM. Here we describe the modied version of the gravity wave spectrum parameterization used in WACCM4.0. 196

3837

Adiabatic inviscid formulation Following Lindzen [1981], the continuous equations for the gravity wave parameterization are obtained from the two-dimensional hydrostatic momentum, continuity and thermodynamic equations in a vertical plane: u= +u , (5.34) t x x u + x +u t x W = 0, Z + N 2w = 0 . Z (5.35) (5.36)

Where N is the local Brunt-Visll frequency, and W is the vertical velocity in log pressure a a a height (Z) coordinates. Eqs. (5.34)(5.36) are linearized about a large scale background wind u, with perturbations u, w , and combined to obtain: +u t x
2

2 w 2 w + N2 2 = 0 . Z 2 x

(5.37)

Solutions to (5.37) are assumed to be of the form: w = w eik(xct) eZ/2H , (5.38)

where H is the scale height, k is the horizontal wavenumber and c is the phase speed of the wave. Substituting (5.38) into (5.37), one obtains: k 2 (u c)2 Neglecting equation:
1 2H

1 + Z 2H

w k2N 2 w = 0 .

(5.39)

compared to

in (5.39), one obtains the nal form of the two dimensional wave d2 w + 2 w = 0 , 2 dZ = N . (u c)
Z

(5.40) (5.41)

with the coecient dened as:

The WKB solution of (5.40) is:

w = A1/2 exp i
0

dz

(5.42)

and the full solution, from (5.38), is: w (Z, t) = A1/2 exp i
0 Z

dz

eik(xct) eZ/2H .

(5.43)

197

The constant A is determined from the wave amplitude at the source (z = 0), The Reynolds stress associated with (5.43) is: 2 (Z) = (0) = u w = |A|2 0 sgn() , k
3838 3839 3840 3841 3842

(5.44)

and is conserved, while the momentum ux u w = (m/k) w w grows exponentially with altitude as exp(Z/H), per (5.43). We note that the vertical ux of wave energy is cgz E = (U c) (Andrews et al. [1987]), where cgz is the vertical group velocity, so that deposition of wave momentum into the mean ow will be accompanied by a transfer of energy to the background state. Saturation condition The wave amplitude in (5.43) grows as eZ/2H until the wave becomes unstable to convective overturning, Kelvin-Helmholtz instability, or other nonlinear processes. At that point, the wave amplitude is assumed to be limited to the amplitude that would trigger the instability and the wave is saturated. The saturation condition used in CAM4 is from McFarlane [1987], based on a maximum Froude number (Fc ), or streamline slope. k |u c|3 |u w | = Fc2 , (5.45) 2 N where is the saturation stress and Fc2 = 0.5. In WACCM4.0, Fc2 = 1 and is omitted hereafter. Following Lindzen [1981], within a saturated region the momentum tendency can be determined analytically from the divergence of : e u = u w , t Z k (u c)3 1 , e 2 N Z k (u c)3 e , (5.46) 2 NH where e is an eciency factor. For a background wave spectrum, e represents the temporal and spatial intermittency in the wave sources. The analytic solution (5.46) is not used in WACCM4.0; it is shown here to illustrate how the acceleration due to breaking gravity waves depends on the intrinsic phase speed. In the model, the stress prole is computed at interfaces and dierenced to get the specic force at layer midpoints. Diusive damping In addition to breaking as a result of instability, vertically propagating waves can also be damped by molecular diusion (both thermal and momentum) or by radiative cooling. Because the intrinsic periods of mesoscale gravity waves are short compared to IR relaxation time scales throughout the atmosphere, we ignore radiative damping. We take into account the molecular m viscosity, Km , such that the stress prole is given by: (Z) = (Zt ) exp 198 2 H
Z 0

3843

3844 3845 3846 3847 3848

3849

i dz

(5.47)

where Zt denotes the top of the region, below Z, not aected by thermal dissipation or molecular diusion. The imaginary part of the local vertical wavenumber, i is then: i =
3850 3851 3852

m N 3 Km . 2k(u c)4

(5.48)

In WACCM4.0, (5.475.48) are only used within the domain where molecular diusion is important (above 75 km). At lower altitudes, molecular diusion is negligible, i 0, and is conserved outside of saturation regions. Transport due to dissipating waves When the wave is dissipated, either through saturation or diusive damping, there is a transfer of wave momentum and energy to the background state. In addition, a phase shift is introduced between the waves vertical velocity eld and its temperature and constituent perturbations so that uxes of heat and constituents are nonzero within the dissipation region. The nature of the phase shift and the resulting transport depends on the dissipation mechanism; in WACCM4.0, we assume that the dissipation can be represented by a linear damping on the potential temperature and constituent perturbations. For potential temperature, , this leads to: +u t x + w = , z (5.49)

3853

where is the dissipation rate implied by wave breaking, which depends on the waves group velocity, cgz (see Garcia [2001]): = cgz (u c)2 =k . 2H 2HN (5.50)

Substitution of (5.50) into (5.49) then yields the eddy heat ux: w =
3854 3855 3856 3857 3858 3859 3860 3861

ww . 2 (u c)2 + 2 z k

(5.51)

Similar expressions can be derived for the ux of chemical constituents, with mixing ratio substituted in place of potential temperature in (5.51). We note that these wave uxes are always downgradient and that, for convenience of solution, they may be represented as vertical diusion, with coecient Kzz equal to the term in brackets in (5.51), but they do not represent turbulent diusive uxes but rather eddy uxes. Any additional turbulent uxes due to wave breaking are ignored. To take into account the eect of localization of turbulence (e.g., Fritts and Dunkerton [1985]; McIntyre [1989]), (5.51) is multiplied times an inverse Prandtl number, P r1 ; in WACCM4.0 we use P r1 = 0.25. Heating due to wave dissipation The vertical ux of wave energy density, E , is related to the stress according to: cgz E = (u c) , 199 (5.52)

3862

where cgz is the vertical group velocity [Andrews et al., 1987]. Therefore, the stress divergence /Z that accompanies wave breaking implies a loss of wave energy. The rate of dissipation of wave energy density is: E 1 (u c) = (u c) . t cgz t Z For a saturated wave, the stress divergence is given by (5.46), so that: k (U c)4 E = (u c) = e . (5.54) t Z 2NH This energy loss by the wave represents a heat source for the background state, as does the change in the background kinetic energy density implied by wave drag on the background ow: u2 k u (u c)3 K =u = e , t 2 t Z 2NH which follows directly from (5.46). The background heating rate, in K sec1 , is then: Qgw = 1 K E . + cp t t k c (c u)3 , 2NH (5.55) (5.53)

(5.56)

Using (5.54) (5.55), this heating rate may be expressed as: Qgw =
3863 3864 3865

1 1 c = cp Z cp

(5.57)

where cp is the specic heat at constant pressure. In WACCM4.0, Qgw is calculated for each component of the gravity wave spectrum using the rst equality in (5.57), i.e., the product of the phase velocity times the stress divergence. Orographic source function For orographically generated waves, the source is taken from McFarlane [1987]: k 2 h 0 N0 u0 , (5.58) 2 0 where h0 is the streamline displacement at the source level, and 0 , N0 , and u0 are also dened at the source level. For orographic waves, the subgrid-scale standard deviation of the orography is used to estimate the average mountain height, determining the typical streamline displacement. An upper bound is used on the displacement (equivalent to dening a separation streamline) which corresponds to requiring that the wave not be supersaturated at the source level: u0 h0 = min(2, ). (5.59) N0 The source level quantities 0 , N0 , and u0 are dened by vertical averages over the source region, taken to be 2, the depth to which the average mountain penetrates into the domain: g = |uw |0 =
2

3866

0 =
0
3867 3868

dz,

{, N, u, v} .

(5.60)

The source level wind vector (u0 , v0 ) determines the orientation of the coordinate system in (5.34)(5.36) and the magnitude of the source wind u0 . 200

3869

Non-orographic source functions The source spectrum for non-orographic gravity waves is no longer assumed to be a specied function of location and season, as was the case with the earlier version of the model described by Garcia et al. [2007]. Instead, gravity waves are launched according to trigger functions that depend on the atmospheric state computed in WACCM4 at any given time and location, as discussed by Richter et al. [2010]. Two trigger functions are used: convective heat release (which is a calculated model eld) and a frontogenesis function, Hoskins [1982], which diagnoses regions of strong wind eld deformation and temperature gradient using the horizontal wind components and potential temperature eld calculated by the model. In the case of convective excitation, the method of Beres et al. [2005] is used to determine a phase speed spectrum based upon the properties of the convective heating eld. A spectrum is launched whenever the deep convection parameterization in WACCM4 is active, and the vertical prole of the convective heating, together with the mean wind eld in the heating region, are used to determine the phase speed spectrum of the momentum ux. Convectively generated waves are launched at the top of the convective region (which varies according to the depth of the convective heating calculated in the model). Waves excited by frontal systems are launched whenever the frontogenesis trigger function exceeds a critical value (see Richter et al. [2010]). The waves are launched from a constant source level, which is specied to be 600 mb. The momentum ux phase speed spectrum is given by a Gaussian function in phase speed: s (c) = b exp c Vs cw
2

3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884

(5.61)

centered on the source wind, Vs = |Vs |, with width cw = 30 m/s. A range of phase speeds with specied width and resolution is used: c Vs + [dc , 2dc , ... cmax ] ,
3885 3886 3887 3888 3889

(5.62)

with dc = 2.5 m s1 and cmax = 80 m s1 , giving 64 phase speeds. Note that c = Vs is retained in the code for simplicity, but has a critical level at the source and, therefore, s (c = Vs ) = 0. Note also that b is a tunable parameter; in practice this is set such that the height of the polar mesopause, which is very sensitive to gravity wave driving, is consistent with observations. In WACCM4, b = 1.5 x 103 Pa. Above the source region, the saturation condition is enforced separately for each phase speed, ci , in the momentum ux spectrum: k |u ci |3 (ci ) i = Fc2 . 2 N (5.63)

3890

Numerical approximations The gravity wave drag parameterization is applied immediately after the nonlinear vertical diusion. The interface Brunt-Visll frequency is a a a N k+
2

g2 T k+

1 k+ T k+ k+ cp p 201

(5.64)

Where the interface density is: k+ =

3891 3892 3893 3894 3895 3896 3897

RT k+ . pk+

(5.65)

The midpoint Brunt-Visll frequencies are N k = (N k+ + N k )/2. a a a The level for the orographic source is an interface determined from an estimate of the vertical penetration of the subgrid mountains within the grid box. The subgrid scale standard deviation of the orography, h , gives the variation of the mountains about the mean elevation, which denes the Earths surface in the model. Therefore the source level is dened as the interface, ks 1/2, for which z ks + < 2h < z ks , where the interface heights are dened from the midpoint heights by z k+ = (z k z k+1 ). The source level wind vector, density and Brunt-Visll frequency are determined by vertical a a a integration over the region from the surface to interface ks + 1/2:
K

0 =
k=ks

k k p ,

{, N, u, v} .

(5.66)

The source level background wind is u0 =

2 (u2 + v0 ), the unit vector for the source wind is 0

(x0 , y0 ) = (u0 , v0 )/u0 , and the projection of the midpoint winds onto the source wind is uk = uk x0 + v k y0 .
3898 3899 3900 3901 3902 3903 3904 3905

(5.67)

(5.68)

Assuming that u0 > 2 m s1 and 2 h > 10 m, then the orographic source term, g is given by (5.58) and (5.59), with Fc2 =1 and k = 2/105 m1 . Although the code contains a provision for a linear stress prole within a low level deposition region, this part of the code is not used in the standard model. The stress proles are determined by scanning up from the bottom of the model to the top. The stress at the source level is determined by (5.58). The saturation stress, at each interface is determined by (5.63), and = 0 if a critical level is passed. A critical level is contained within a layer if (uk+ c )/(uk c ) < 0. Within the molecular diusion domain, the imaginary part of the vertical wavenumber is given by (5.48). The interface stress is then determined from the stress on the interface below by: R k = min ( )k , k+ exp 2i T k k ln p . (5.69) g Below the molecular diusion domain, the exponential term in (5.69) is omitted. Once the complete stress prole has been obtained, the forcing of the background wind is determined by dierentiating the prole during a downward scan: uk k =g k < t p uk t
max

3906

uk t

max

(5.70) (5.71)

= min

|c uk | , 500 m s1 day1 . 2t 202

3907 3908 3909 3910 3911 3912 3913

The rst bound on the forcing comes from requiring that the forcing not be large enough to push the wind more than half way towards a critical level within a time step and takes the place of an implicit solution. This bound is present for numerical stability, it comes into play when the time step is too large for the forcing. It is not feasible to change the time step, or to write an implicit solver, so an a priori bound is used instead. The second bound is used to constrain the forcing to lie within a physically plausible range (although the value used is extremely large) and is rarely invoked. When any of the bounds in (5.70) are invoked, conservation of stress is violated. In this case, stress conservation is ensured by decreasing the stress on the lower interface to match the actual stress divergence in the layer: k+ = k + uk k p . t g (5.72)

3914 3915

This has the eect of pushing some of the stress divergence into the layer below, a reasonable choice since the waves are propagating up from below. Finally, the vector momentum forcing by the gravity waves is determined by projecting the background wind forcing with the unit vectors of the source wind: Vk = (x0 , y0) E t uk . t (5.73)

3916 3917 3918 3919 3920

In addition, the frictional heating implied by the momentum tendencies, c1p Vk Vk /t, is added to the thermodynamic equation. This is the correct heating for orographic (c = 0) waves, but not for waves with c = 0, since it does not account for the wave energy ux. This ux is accounted for in some middle and upper atmosphere versions of CAM4, but also requires accounting for the energy ux at the source.

3921

5.3.4

Turbulent Mountain Stress

3922 3923 3924 3925 3926 3927 3928 3929 3930 3931

An important dierence between WACCM4 and earlier versions is the addition of surface stress due to unresolved orography. A numerical model can compute explicitly only surface stresses due to resolved orography. At the standard 1.9 x 2.5 (longitude x latitude) resolution used by WACCM4 only the gross outlines of major mountain ranges are resolved. To address this problem, unresolved orography is parameterized as turbulent surface drag, using the concept of eective roughness length developed by Fiedler and Panofsky [1972]. Fiedler and Panofsky dened the roughness length for heterogeneous terrain as the roughness length that homogenous terrain would have to give the correct surface stress over a given area. The concept of eective roughness has been used in several Numerical Weather Prediction models (e.g., Wilson [2002]; Webster et al. [2003]). In WACCM4 the eective roughness stress is expressed as: = Cd |V|V , where is the density and Cd is a turbulent drag coecient, Cd = f (Ri ) k 2 ln2 203
z+z0 z0

(5.74)

(5.75)

3932 3933 3934 3935 3936 3937 3938

k is von Krmns constant; z is the height above the surface; z0 is an eective roughness length, a a dened in terms of the standard deviation of unresolved orography; and f (Ri ) is a function of the Richardson number (see Richter et al. [2010] for details). The stress calculated by (5.74) is used the models nonlocal PBL scheme to evaluate the PBL height and nonlocal transport, per Eqs. (3.10)(3.12) of Holstlag and Boville [1993]. This calculation is carried out only over land, and only in grid cells where the height of topography above sea level, z, is nonzero.

3939

5.3.5

QBO Forcing

3940 3941 3942 3943 3944

WACCM4 has several options for forcing a quasi-biennial oscillation (QBO) by applying a momentum forcing in the tropical stratosphere. The parameterization relaxes the simulated winds to a specied wind eld that is either xed or varies with time. The parameterization can also be turned o completely. The namelist variables and input les can be selected to choose one of the following options: Idealized QBO East winds, used for perpetual xed-phase of the QBO, as described by Matthes et al. [2010]. Idealized QBO West winds, as above but for the west phase. Repeating idealized 28-month QBO, also described by Matthes et al. [2010]. QBO for the years 1953-2004 based on the climatology of Giorgetta [see: http://www.pa.op.dlr.de/CCMVal/Forcings/qbo data ccmval/u prole 195301-200412.html, 2004]. QBO with a 51-year repetition, based on the 1953-2004 climatology of Giorgetta, which can be used for any calendar year, past or future. The relaxation of the zonal wind is based on Balachandran and Rind [1995] and is described in Matthes et al. [2010]. The input winds are specied at the equator and the parameterization extends latitudinally from 22 N to 22 S, as a Gaussian function with a half width of 10 centered at the equator. Full vertical relaxation extends from 86 to 4 hPa with a time constant of 10 days. One model level below and above this altitude range, the relaxation is half as strong and is zero for all other levels. This procedure constrains the equatorial winds to more realistic values while allowing resolved and parameterized waves to continue to propagate. The xed or idealized QBO winds (rst 3 options) can be applied for any calendar period. The observed input (Giorgetta climatology) can be used only for the model years 1953-2004. The winds in the nal option were determined from the Giorgetta climatology for 1954-2004 via ltered spectral decomposition of that climatology. This gives a set of Fourier coecients that can be expanded for any day and year. The expanded wind elds match the climatology during the years 1954-2004.

3945 3946

3947

3948

3949 3950 3951

3952 3953

3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966

204

3967

5.3.6

Radiation

3968 3969 3970 3971 3972

The radiation parameterizations in CAM4 are quite accurate up to 65 km, but deteriorate rapidly above that altitude. Because 65 km is near a local minimum in both shortwave heating and longwave cooling, it is a particularly convenient height to merge the heating rates from parameterizations for the lower and upper atmosphere. Therefore, we retain the CAM4 parameterizations below 65 km and use new parameterizations above. The merged shortwave and longwave radiative heatings are determined from Q = w1 QCAM 3 + w2 QM LT , (5.76)

where w1 (z < zb ) = 1, w2 (z > zt ) = 1 and z = log(105 /p) is the pressure scale height. The CAM4 radiation parameterizations are used below zb and the MLT parameterizations are used above zt . For zb < z < zt , w2 = 1 w1 and

w1 = 1 tanh
3973 3974 3975 3976 3977 3978

z zb zw

(5.77)

where zw is the transition width. The merging was developed and tested separately for shortwave and longwave radiation and the constants are slightly dierent. For longwave radiation, the constants are zb = 8.57, zt = 10 and zw = 0.71. For shortwave radiation, the constants are zb = 9, zt = 10 and zw = 0.75. These constants give smooth heating proles. Note that a typical atmospheric scale height of H = 7 km places the transition zones between 60 and 70 km. Longwave radiation WACCM4.0 retains the longwave (LW) formulation used in CAM4 [Kiehl and Briegleb, 1991]. However, in the MLT longwave radiation uses the parameterization of Fomichev et al. [1998] for CO2 and O3 cooling and the parameterization of Kockarts [1980] for NO cooling at 5.3 m. As noted above, the LW heating/cooling rates produced by these parameterizations are merged smoothly at 65 km with those produced by the standard CAM4 LW code, as recently revised by Collins et al. [2002]. In the interactive chemistry case all of the gases (O, O2 , O3 , N2 , NO, and CO2 ) that are required by these parameterizations, are predicted within WACCM4.0. Shortwave radiation WACCM4.0 uses a combination of solar parameterizations to specify spectral irradiances over two spectral intervals. The rst spectral interval covers soft x-ray and extreme ultraviolet irradiances (wavelengths between 0.05 nm to Lyman- (121.6 nm)) and is calculated using the parameterization of Solomon and Qiang [2005]. The parameterizations take as input the 10.7 cm solar radio ux (f 10.7) and its 81-day average (f 10.7a). Daily values of f 10.7 are obtained from NOAAs Space Environment Center (www.sec.noaa.gov). The irradiance of the jth spectral interval is: Fj = Fj0 1 + Rj (f 10.7 + f 10.7a) Fmin 2 205 (5.78)

3979

3980 3981 3982 3983 3984 3985 3986

3987

3988 3989 3990 3991 3992 3993

3994 3995 3996 3997 3998

where Fmin = 80. Fj0 and Rj are taken from Table A1 of Solomon and Qiang [2005]. Fluxes for the second interval between Lyman- (121.6 nm) and 100 m. are specied using an empirical model of the wavelength-dependent sunspot and facular inuences [Lean, 2000; Wang et al., 2005]. Spectral resolution is 1 nm between 121.6 nm and 750nm, 5 nm between 750nm and 5m, 10 nm between 5m and 10m, and 50 nm between 10m and 100 m. In the troposphere, stratosphere and lower mesosphere (z < 65km) WACCM4.0 retains the CAM4 shortwave heating (200 nm to 4.55 m) which is calculated from the net shortwave spectral ux into each layer Collins et al. [2004b]. The solar spectrum for the CAM4 heating calculation is divided into 19 intervals [Collins, 1998]. The heating in these intervals must be adjusted to match the irradiances calculated for the upper part of the model, and those used in the photolysis calculations. This is achieved by applying a scaling (Sj ) to the solar heating in the jth CAM4 spectral interval using the spectrum from Lean [2000] and Wang et al. [2005]: Sj = Fj Fjref , (5.79)

3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012

where Fj is the spectral irradiance (W/m2 /nm) integrated over the jth band, and Fjref is the same integral taken over a reference spectrum calculated from annual mean uxes over a 3-solarcycle period from XX to YY. In the MLT region, shortwave heating is the sum of the heating due to absorption of photons and subsequent exothermic chemical reactions that are initiated by photolysis. The majority of energy deposited by an absorbed photon goes into breaking molecular bonds, rather than into translational energy of the absorbing molecule (heat). Chemical heating results when constituents react to form products of lower total chemical potential energy. This heating can take place months after the original photon absorption and thousands of kilometers away. Heating rates range from 1 K/day near 75 km to 100-300 K/day near the top of the model domain. It is clear that quenching of O(1 D) is a large source of heating throughout the MLT. Above 100 km ion reactions and reactions involving atomic nitrogen are signicant sources of heat, while below that level OX (= O + O3 ) and HOX (= H + OH + HO2 ) reactions are the dominant producers of chemical heating. Heating within the MLT from the absorption of radiation that is directly thermalized is calculated over the wavelength range of 0.05 nm to 350 nm. For wavelengths less than Lyman, it is assumed that 5% of the energy of each absorbed photon is directly thermalized: QEU V = (cp )1
k

nk
j

Jk (j )

hc , j

(5.80)

4013 4014 4015 4016 4017 4018 4019 4020

where = 0.05. Here is mass density, cp is the specic heat of dry air, n is the number density of the absorbing species, and J is the photolysis/photoionization rate. The total heating is the sum of k photolysis reactions and j wavelengths intervals. At these wavelengths absorption of a photon typically leads to photoionization, with the resulting photoelectron having sucient energy to ionize further molecules. Calculation of Jij and ionization rates from photoelectrons is calculated based on the parameterization of Solomon and Qiang [2005]. In a similar manner, the heating rate within the aurora (QAU R ) is calculated as the product of the total ionization rate, 35 eV per ion pair, and the same heating eciency of 5%.

206

Between Lyman- and 350 nm the energy required to break molecular bonds is explicitly accounted for. The heating rate is thus dened as: QU V = (cp )1
k
4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041

nk
j

Jk (j ){

hc BDEk }, j

(5.81)

where BDE is the bond dissociation energy. In addition to these sources of heat, WACCM4.0 calculates heating by absorption in the nearinfrared by CO2 (between 1.05 to 4.3 m), which has its largest contribution near 70km and can exceed 1 K/day [Fomichev et al., 2004]. Heating from this process is calculated using the parameterization of Ogibalov and Fomichev [2003]. Finally, the heating produced by collisions of electrons and neutrals (Joule heating) is also calculated using the predicted ion and electron concentrations. This is described in section 5.3.8. Local heating rates from joule heating can be very large in the auroral regions, reaching over 103 K/day in the upper levels of the model. Airglow, radiation produced when excited atoms or molecules spontaneously emit, is accounted for in WACCM4.0 for emissions of O2 (1 ), O2 (1 ), and vibrationally excited OH. Airglow from the excited molecular oxygen species are handled explicitly; radiative lifetimes for O2 (1 ) and O2 (1 ) are 2.58104 s1 and 0.085 s1 respectively. However, modeling of the many possible vibrational transitions of OH is impractical in a model as large as WACCM4.0. Energy losses from the emission of vibrationally excited OH are therefore accounted for by applying an eciency factor to the exothermicity of the reaction that produces vibrationally excited OH; the reaction of hydrogen and ozone. In other words, the reaction H + O3 produces ground state OH only, but the chemical heating from the reaction has been reduced to take into consideration that some of the chemical potential energy has been lost in airglow. This approach is the same one used by Mlynczak and Solomon [1993] and we use their recommended eciency factor of 60%. Any energy lost through airglow is assumed to be lost to space, and so represents an energy pathway that does not generate heat. Volcanic Heating The sulfate aerosol heating is a function of a prescribed aerosol distribution varying in space and time that has a size distribution similar to that seen after a volcanic eruption [Tilmes et al., 2009]. The H2 SO4 mass distribution is calculated from the prescribed sulfate surface area density (SAD) assuming a lognormal size distribution, number of particles per cm-3, and distribution width (see section 3.6.2). The H2SO4 mass distribution is then passed to the radiative transfer code (CAMRT), which in turn calculates heating and cooling rates.

4042

4043 4044 4045 4046 4047 4048

4049

5.3.7

WACCM4.0 chemistry

4050

Chemical Mechanism (Neutral Species) WACCM4.0 includes a detailed neutral chemistry model for the middle atmosphere based on the Model for Ozone and Related Chemical Tracers, Version 3 [Kinnison et al., 2006]. The mechanism represents chemical and physical processes in the troposphere through the lower thermosphere. The species included within this mechanism are contained within the OX , NOX , HOX , ClOX , and BrOX chemical families, along with CH4 and its degradation products. This 207

4051 4052 4053 4054 4055

4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067

mechanism contains 52 neutral species, one invariant (N2 ), 127 neutral gas-phase reactions, 48 neutral photolytic reactions, and 17 heterogeneous reactions on three aerosol types (see below). Lists of the chemical species are given in Table 1. The rst column lists the symbolic name (as used in the mechanism); the second column lists the species atomic composition; the third column designates which numerical solution approach is used (i.e., explicit or implicit); the fourth column lists any deposition processes (wet or dry) for that species; and the fth column indicates whether the surface (or upper) boundary condition is xed vmr or ux, or if a species has an in-situ ux (from lightning or aircraft emissions). The gas-phase reactions included in the WACCM4.0 middle atmosphere chemical mechanism are listed in Table 2. In most all cases the chemical rate constants are taken from JPL06-2 [Sander, S. P., et al., 2006]. Exceptions to this condition are described in the comment section for any given reaction. Heterogeneous reactions on four dierent aerosols types are also represented in the WACCM4.0 chemical mechanism (see Table 3): 1) liquid binary sulfate (LBS); 2) Supercooled ternary solution (STS); 3) Nitric acid trihydrate (NAT); and 4) water-ice. There are 17 reactions, six reactions on liquid sulfate aerosols (LBS or STS), ve reactions on solid NAT aerosols, and six reactions on solid water-ice aerosols. The rate constants for these 17 heterogeneous reactions can be divided up into two types: 1) rst order; and 2) pseudo second order. For rst order hydrolysis reactions (Table 3, reactions 1-3, 7-8, 11, and 12-14), the heterogeneous rate constant is derived in the following manner: 1 k = V SAD 4 (5.82)

4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080

Where V = mean velocity; SAD = surface area density of LBS, STS, NAT, or water-ice, and = reaction probability for each reaction. The units for this rate constant are s1 . Here the H2 O abundance is in excess and assumed not change relative to the other reactant trace constituents. The mean velocity is dependent on the molecular weight of the non-H2 O reactant (i.e., N2 O5 , ClONO2 , or BrONO2 ). The SAD for each aerosol type is described in section 7. The reaction probability is dependent on both composition and temperature for sulfate aerosol (see JPL06-2). The reaction probability is a xed quantity for NAT and water-ice aerosols and is listed in Table 3. Multiplying the rate constant times the concentration gives a loss rate in units of molecules cm3 sec1 for the reactants and is used in the implicit solution approach. The non-hydrolysis reaction (Table 3, reactions 4-6, 9-10, and 15-17) are second order reactions. Here, the rst order rate constant (equation 6) is divided by the HCl concentration, giving it the typical bimolecular rate constant unit value of cm3 molecule1 sec1 . This approach assumes that all the HCl is in the aerosol particle. Stratospheric Aerosols Heterogeneous processes on liquid sulfate aerosols and solid polar stratospheric clouds (Type 1a, 1b, and 2) are included following the approach of Considine et al. [2000]. This approach assumes that the condensed phase mass follows a lognormal size distribution taken from Considine et al. [2000], ln(r/r0 )2 N0 N(r) = exp (5.83) 2 2 r 2 208

4081

4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124

where N is the aerosol number density (particles cm3 ); r and r0 are the particle radius and median radius respectively; and is the standard deviation of the lognormal distribution. N0 and r0 are supplied for each aerosol type. The aerosol surface area density (SAD) is the second moment of this distribution. At model temperatures (Tmodel ) greater than 200 K, liquid binary sulfate (LBS) is the only aerosol present. The surface area density (SAD) for LBS is derived from observations from SAGE, SAGE-II and SAMS [Thomason et al., 1997] as updated by Considine [World Meteorological Organization, 2003]. As the model atmosphere cools, the LBS aerosol swells, taking up both HNO3 and H2 O to give STS aerosol. The Aerosol Physical Chemistry Model (ACPM) is used to derive STS composition Tabazadeh et al. [1994]. The STS aerosol median radius and surface area density is derived following the approach of Considine et al. [2000]. The width of the STS size distribution ( = 1.6) and number density (10 particles cm3 ) are prescribed according to measurements from Dye et al. [1992]. The STS aerosol median radius can swell from approximately 0.1 m to approximately 0.5 m. There is no aerosol settling assumed for this type of aerosol. The median radius is used in derivation of sulfate aerosol reaction probability coecients. Both the LBS and STS surface area densities are used for the calculation of the rate constants as listed in Table 3; reactions (1)-(6). Solid nitric acid containing aerosol formation is allowed when the model temperature reaches a prescribed super saturation ratio of HNO3 over NAT [Hansen and Mauersberger, 1988]. This ratio is set to 10 in WACCM4.0 [Peter et al., 1991]. There are three methods available to handle the HNO3 uptake on solid aerosol. The rst method directly follows Considine et al. [2000, 2004]. Here, after the supersaturation ratio assumption is met, the available condensed phase HNO3 is assumed to reside in the solid NAT aerosol. The derivation of the NAT median radius and surface area density follows the same approach as the STS aerosol, by assuming: a lognormal size distribution, a width of a distribution ( = 1.6; Dye et al. [1992]), and a number density (0.01 particles cm3 ; Tabazadeh et al. [2000]). The NAT radius settles with a value of r0 ranging between 2 and 5 m; this value depends on the model temperature and subsequent amount of condensed phase HNO3 formed. This NAT median radius r0 is also used to derive the terminal velocity for settling of NAT (section 8) and the eventual irreversible denitrication. The NAT surface area density is used to calculate the rate constants for heterogeneous reactions 7-11 (Table 3). Since the available HNO3 is included inside the NAT aerosol, there is no STS aerosol present. However, there are still heterogeneous reactions occurring on the surface of LBS aerosols. If the calculated atmospheric temperature, T , becomes less than or equal to the saturation temperature (Tsat ) for water vapor over ice (e.g., Marti and Mauersberger [1993]), water-ice aerosols can form. In WACCM4.0 the condensed phase H2 O is derived in the prognotic water routines of CAM and passed into the chemistry module. Using this condensed phase H2 O, the median radius and the surface area density for water-ice are again derived following the approach of Considine et al. [2000]. The water-ice median radius and surface area density assumes a lognormal size distribution, a width of a distribution = 1.6 [Dye et al., 1992], and a number density of 0.001 particles cm3 [Dye et al., 1992]. The value of r0 is typically 10m. The waterice surface area density is used for the calculation of the rate constants for reactions 12-17 (Table 3).

209

4125

Sedimentation of Stratospheric Aerosols The sedimentation of HNO3 in stratospheric aerosols follows the approach described in Considine et al. [2000]. The following equation is used to derive the ux (F ) of HNO3 , as NAT aerosol, across model levels in units of molecules cm2 sec1 . Fi = Vi Ci exp(8 ln2 i ), (5.84)

4126 4127 4128 4129 4130 4131 4132 4133 4134 4135

where i = 1 for NAT; Vi is the terminal velocity of the aerosol particles (cm s1 ); C is the condensed-phase concentration of HNO3 (molecules cm3 ); is the width of the lognormal size distribution for NAT (see discussion above). The terminal velocity is dependent on the given aerosol: 1) mass density; 2) median radius; 3) shape; 4) dynamic viscosity; and 5) Cunningham correction factor for spherical particles (see Fuch [1964] and Kasten [1968] for the theory behind the derivation of terminal velocity). For each aerosol type the terminal velocity could be calculated, however, in WACCM4.0 this quantity is only derived for NAT. Settling of HNO3 contain in STS is not derived based on the assumption that the median radius is too small to cause any signicant denitrication and settling of condensed phase H2 O is handled in the CAM4 prognostic water routines. Ion Chemistry WACCM4.0 includes a six constituent ion chemistry model (O+ , O+ , N+ , N+ , NO+ , and elec2 2 trons) that represents the the E-region ionosphere. The global mean ion and electron distributions simulated by WACCM4.0 for solar minimum conditions are shown in Figure 5.1, which clearly shows that the dominant ions in this region are NO+ and O+ . Ion-neutral and recombi2 nation reactions included in WACCM4.0 are listed in Table 5.3.7. The reaction rate constants for these reactions are taken from R.G.Roble [1995]. Ionization sources include not only the aforementioned absorption of extreme ultraviolet and soft x-ray photons, and photoelectron impact, but also energetic particles precipitation in the auroral regions. The latter is calculated by a parameterization based on code from the NCAR TIME-GCM model [Roble and Ridley, 1987] that rapidly calculates ion-pair production rates, including production in the polar cusp and polar cap. The parameterization takes as input hemispheric power (HP), the estimated power in gigawatts deposited in the polar regions by energetic particles. Currently WACCM4.0 uses a parameterization of HP (in GW) based on an empirical relationships between HP and the Kp planetary geomagnetic index. For Kp 7, WACCM4.0 uses the relationship obtained by Zhang and Paxton [2008] from TIMED/GUVI observations: HP = 16.82 Kp exp (0.32) 4.86 (5.85)

4136

4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149

For Kp > 7, WACCM4.0 linearly interpolates HP, assuming HP equals to 300 when Kp is 9, based on NOAA satellite measurements: HP = 153.13 +
4150 4151

Kp 7 (300 153.13) 97

(5.86)

Kp is also available from NOAAs Space Environment Center and covers the period from 1933 to the present, making it ideal for long-term retrospective simulations. 210

Figure 5.1: Global mean distribution of charged constituents during July solar minimum conditions.

Figure 5.2: a) Global distribution of ionization rates at 7.3105 hPa, July 1, UT0100 HRS. Contour interval is 2103 cm3 s1 . b) Simultaneous global mean ionization rates (cm3 s1 ) versus pressure. 211

4152 4153 4154 4155 4156 4157

Total ionization rates at 110km during July for solar maximum conditions are shown in Figure 5.2a. The broad region of ionization centered in the tropics is a result of EUV ionization, and has a peak value of almost 103 at 22 N. Ionization rates from particle precipitation can exceed this rate by 40% but are limited to the high-latitudes, as can been seen by the two bands that are approximately aligned around the magnetic poles. The global mean ionization rate (Figure 5.2b) An important aspect of including ionization processes (both in the aurora and by energetic photons and photoelectrons), is that it leads to a more accurate representation of thermospheric nitric oxide. Not only does nitric oxide play an important role in the energy balance of the lower thermosphere through emission at 5.3 m, it might also be transported to the upper stratosphere, where it can aect ozone concentrations. Nitric oxide is produced through quenching of N(2 D): N(2 D) + O2 NO + O(1 D) + 1.84eV (5.87)

4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170

N(2 D) is produced either via recombination of NO+ (see Table 5.3.7) or directly by ionization of molecular nitrogen. The branching ratio between N(2 D) and ground-state atomic nitrogen for the photoionization process is critical in determining the eectiveness of NO production. If ground-state atomic nitrogen is produced then it can react with NO to produce molecular nitrogen and eectively remove to members of the NOx family. In WACCM4.0 60% of the atomic nitrogen produced is in the excited state, which implies absorption of EUV results in a net source of NO. Also shown are maxima at high latitudes due to auroral ionization. WACCM4.0 reproduces many of the features of the Nitric Oxide Empirical Model (NOEM) distribution [Marsh et al., 2004], which is based on data from the Student Nitric Oxide Explorer satellite [Barth et al., 2003] In particular, larger NO in the winter hemisphere (a result of less photolytic loss), and a more localized NO maximum in the Northern Hemisphere (related to the lesser oset of geographic and magnetic poles, and so less spread when viewed as a geographic zonal mean).

212

Table 5.2: WACCM4.0 Neutral Chemical Species (51 computed species + N2 ) no. Symbolic Name Chemical Formula Numerics Deposition Boundary Condition 1 O O(3 P) Implicit ubvmr 1 2 O1D O( D) Implicit 3 O3 O3 Implicit dry 4 O2 O2 Implicit ubvmr 5 O2 1S O2 (1 ) Implicit 6 O2 1D O2 (1 ) Implicit 7 H H Implicit ubvmr 8 OH OH Implicit 9 HO2 HO2 Implicit 10 H2 H2 Implicit vmr, ubvmr 11 H2O2 H2 O2 Implicit dry, wet 12 N N Implicit ubvmr 13 N2D N(2 D) Implicit from TIME-GCM 14 N2 N2 Invariant 15 NO NO Implicit ux, ubvmr, lux, airux 16 NO2 NO2 Implicit dry 17 NO3 NO3 Implicit 18 N2O5 N2 O5 Implicit 19 HNO3 HNO3 Implicit dry, wet 20 HO2NO2 HO2 NO2 Implicit dry, wet 21 CL Cl Implicit 22 CLO ClO Implicit 23 CL2 Cl2 Implicit 24 OCLO OClO Implicit 25 CL2O2 Cl2 O2 Implicit 26 HCL HCl Implicit wet 27 HOCL HOCl Implicit wet 28 ClONO2 ClONO2 Implicit wet 29 BR Br Implicit 30 BRO BrO Implicit 31 HOBR HOBr Implicit wet 32 HBR HBr Implicit wet 33 BrONO 2 BrONO2 Implicit wet 34 BRCL BrCl Implicit

213

Table 5.1: (continued) WACCM4.0 Neutral Chemical Species (51 computed species + N2 ) no. Symbolic Name Chemical Formula Numerics Deposition Boundary Condition 35 CH4 CH4 Implicit vmr, airux 36 CH3O2 CH3 O2 Implicit 37 CH3OOH CH3 OOH Implicit dry, wet 38 CH2O CH2 O Implicit dry, wet ux 39 CO CO Explicit dry ux, ubvmr, airux 40 CH3CL CH3 Cl Explicit vmr 41 CH3BR CH3 Br Explicit vmr 42 CFC11 CFCl3 Explicit vmr 43 CFC12 CF2 Cl2 Explicit vmr 44 CFC113 CCl2 FCClF2 Explicit vmr 45 HCFC22 CHClF2 Explicit vmr 46 CCL4 CCl4 Explicit vmr 47 CH3CCL3 CH3 CCl3 Explicit vmr 48 CF2CLBR CBr2 F2 (Halon-1211) Explicit vmr 49 CF3BR CBrF3 (Halon-1301) Explicit vmr 50 H2O H2 O Explicit ux 51 N2O N2 O Explicit vmr 52 CO2 CO2 Explicit vmr, ubvmr Deposition: wet = wet deposition included dry = surface dry deposition included If there is no designation in the deposition column, this species is not operated on by wet or dry deposition algorthims. Boundary Condition: ux = ux lower boundary conditions vmr = xed volume mixing ratio (vmr) lower boundary condition ubvmr = xed vmr upper boundary condition lux = lightning emission included for this species airux= aircraft emissions included for this species If there is no designation in the Boundary Conditions column, this species has a zero ux boundary condition for the top and bottom of the model domain.

214

no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

Table 5.2: WACCM4.0 Gas-phase Reactions. Reactions Comments Oxygen Reactions O + O2 + M O 3 + M JPL-06 O + O3 2 O 2 JPL-06 O + O + M O2 + M Smith and Robertson (2008) O2 (1 ) + O O2 (1 ) + O JPL-06 1 O2 1S + O2 O2 ( ) + O2 JPL-06 1 1 O2 ( ) + N2 O2 ( ) + N2 JPL-06 O2 (1 ) + O3 O2 (1 ) + O3 JPL-06 1 1 O2 ( ) + CO2 O2 ( ) + CO2 JPL-06 1 O2 ( ) O2 JPL-06 O2 (1 ) + O O2 + O JPL-06 1 O2 ( ) + O2 2 O2 JPL-06 O2 (1 ) + N2 O2 + N2 JPL-06 1 O2 ( ) O2 JPL-06 1 O( D) + N2 O + N2 JPL-06 O(1 D)+ O2 O + O2 (1 ) JPL-06 O(1 D)+ O2 O + O2 JPL-06 O(1 D)+ H2 O 2 OH JPL-06 O(1 D) + N2 O 2 NO JPL-06 1 O( D) + N2 O N2 + O2 JPL-06 O(1 D) + O3 2 O2 JPL-06 1 O( D) + CFC11 3 Cl JPL-06; Bloomeld [1994] for quenching of O(1 D) O(1 D) + CFC12 2 Cl JPL-06; Bloomeld [1994] 1 O( D) + CFC113 3 Cl JPL-06; Bloomeld [1994] O(1 D) + HCFC22 Cl JPL-06; Bloomeld [1994] 1 O( D) + CCl4 4 Cl JPL-06 1 O( D) + CH3 Br Br JPL-06 O(1 D) + CF2 ClBr Cl + Br JPL-06 1 O( D) + CF3 Br Br JPL-06 O(1 D) + CH4 CH3 O2 + OH JPL-06 1 O( D) + CH4 CH2 O + H + HO2 JPL-06 1 O( D) + CH4 CH2 O + H2 JPL-06 O(1 D) + H2 H + OH JPL-06 O(1 D) + HCl Cl + OH JPL-06 O(1 D) + HBr Br + OH JPL-06

215

no. 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56

Table 5.2: (continued) WACCM4.0 Gas-phase Reactions. Reactions Comments Nitrogen Radicals N(2 D) + O2 NO + O(1 D) JPL-06 2 N( D) + O N + O JPL-06 N + O2 NO + O JPL-06 N + NO N2 + O JPL-06 N + NO2 N2 O + O JPL-06 NO + O + M NO2 + M JPL-06 NO + HO2 NO2 + OH JPL-06 NO + O3 NO2 + O2 JPL-06 NO2 + O NO + O2 JPL-06 NO2 + O + M NO3 + M JPL-06 NO2 + O3 NO3 + O2 JPL-06 NO2 + NO3 + M N2 O5 + M JPL-06 N2 O5 + M NO2 + NO3 + M JPL-06 NO2 + OH + M HNO3 + M JPL-06 HNO3 + OH NO3 + H2 O JPL-06 NO2 + HO2 + M HO2 NO2 + M JPL-06 NO3 + NO 2 NO2 JPL-06 NO3 + O NO2 + O2 JPL-06 NO3 + OH NO2 + HO2 JPL-06 NO3 + HO2 NO2 + OH + O2 JPL-06 HO2 NO2 + OH NO2 + H2 O + O2 JPL-06 HO2 NO2 + M HO2 + NO2 + M JPL-06

216

no. 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85

Table 5.2: (continued) WACCM4.0 Gas-phase Reactions. Reactions Comments Hydrogen Radicals H + O2 + M HO2 + M JPL-06 H + O3 + M OH + O2 JPL-06 H + HO2 2 OH JPL-06 H + HO2 H2 + O2 JPL-06 H + HO2 H2 O + O JPL-06 OH + O H + O2 JPL-06 OH + O3 HO2 + O2 JPL-06 OH + HO2 H2 O + O2 JPL-06 OH + OH H2 O + O JPL-06 OH + OH + M H2 O2 + M JPL-06 OH + H2 H2 O + H JPL-06 OH + H2 O2 H2 O + HO2 JPL-06 HO2 + O OH + O2 JPL-06 HO2 + O3 OH + 2O2 JPL-06 HO2 + HO2 H2 O2 + O2 JPL-06 H2 O2 + O OH + HO2 JPL-06 Chlorine Radicals Cl + O3 ClO + O2 JPL-06 Cl + H2 HCl + H JPL-06 Cl + H2 O2 HCl + HO2 JPL-06 Cl + HO2 HCl + O2 JPL-06 Cl + HO2 ClO + OH JPL-06 Cl + CH2 O HCl + HO2 + CO JPL-06 Cl + CH4 CH3 O2 + HCl JPL-06 ClO + O Cl + O2 JPL-06 ClO + OH Cl + HO2 JPL-06 ClO + OH HCl + O2 JPL-06 ClO + HO2 HOCl + O2 JPL-06 ClO + NO NO2 + Cl JPL-06 ClO + NO2 + M ClONO2 + M JPL-06

217

no. 86 87 88 89 90 91 92 93 94 95 96 97 98 no. 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114

Table 5.2: (continued) WACCM4.0 Gas-phase Reactions. Reactions Comments Chlorine Radicals Continued ClO + ClO 2 Cl + O2 JPL-06 ClO + ClO Cl2 + O2 JPL-06 ClO + ClO Cl + OClO JPL-06 ClO + ClO + M Cl2 O2 + M JPL-06 Cl2 O2 + M 2 ClO + M JPL-06 HCl + OH H2 O + Cl JPL-06 HCl + O Cl + OH JPL-06 HOCl + O ClO + OH JPL-06 HOCl + Cl HCl + ClO JPL-06 HOCl + OH ClO + H2 O JPL-06 ClONO2 + O ClO + NO3 JPL-06 ClONO2 + OH HOCl + NO3 JPL-06 ClONO2 + Cl Cl2 + NO3 JPL-06 Reactions Comments Bromine Radicals Br + O3 BrO + O2 JPL-06 Br + HO2 HBr + O2 JPL-06 Br + CH2 O HBr + HO2 + CO JPL-06 BrO + O Br + O2 JPL-06 BrO + OH Br + HO2 JPL-06 BrO + HO2 HOBr + O2 JPL-06 BrO + NO Br + NO2 JPL-06 BrO + NO2 + M BrONO2 + M JPL-06 BrO + ClO Br + OClO JPL-06 BrO + ClO Br + Cl + O2 JPL-06 BrO + ClO BrCl + O2 JPL-06 BrO + BrO 2 Br + O2 JPL-06 HBr + OH Br + H2 O JPL-06 HBr + O Br + OH JPL-06 HOBr + O BrO + OH JPL-06 BrONO2 + O BrO + NO3 JPL-06

218

no. 115 116 117 118 119 120 121 122 123 124 125 126 127

Table 5.2: (continued) WACCM4.0 Gas-phase Reactions. Reactions Comments Halogen Radicals CH3 Cl + Cl HO2 + CO + 2HCl JPL-06 CH3 Cl + OH Cl + H2 O + HO2 JPL-06 CH3 CCl3 + OH 3 Cl + H2 O JPL-06 HCFC22 + OH Cl + H2 O + HO2 JPL-06 CH3 Br + OH Br + H2 O + HO2 JPL-06 CH4 and Derivatives CH4 + OH CH3 O2 + H2 O JPL-06 CH3 O2 + NO CH2 O + NO2 + HO2 JPL-06 CH3 O2 + HO2 CH3 OOH + O2 JPL-06 CH3 OOH + OH 0.7 CH3 O2 + 0.3 OH + 0.3 CH2 O + H2 O JPL-06 CH2 O + NO3 CO + HO2 + HNO3 JPL-06 CH2 O + OH CO + H2 O + H JPL-06 CH2 O + O OH + HO2 + CO JPL-06 CO + OH H + CO2 JPL-06

Table 5.3: WACCM4.0 Heterogeneous Reactions on liquid and solid aerosols. no. Reaction Comments Sulfate Aerosol 1 N2 O5 + H2 O 2 HNO3 JPL-06; f (sulfuric acid wt %) 2 ClONO2 + H2 O HOCl + HNO3 JPL-06; f (T, P, HCl, H2 O, r) 3 BrONO2 + H2 O HOBr + HNO3 JPL-06; f (T, P, H2 O, r) 4 ClONO2 + HCl Cl2 + HNO3 JPL-06; f (T, P, HCl, H2 O, r) 5 HOCl + HCl Cl2 + H2 O JPL-06; f (T, P, HCl, HCl, H2 O, r) 6 HOBr + HCl BrCl + H2 O JPL-06; f (T, P, HCl, HOBr, H2 O, r) NAT Aerosol 7 N2 O5 + H2 O 2 HNO3 JPL-06; = 4 104 8 ClONO2 + H2 O HOCl + HNO3 JPL-06; = 4 103 9 ClONO2 + HCl Cl2 + HNO3 JPL-06; = 0.2 10 HCl + HCl Cl2 + H2 O JPL-06; = 0.1 11 BrONO2 + H2 O HOBr + HNO3 JPL-06; = 0.3 Water-Ice Aerosol 12 N2 O5 + H2 O 2 HNO3 JPL-06; = 0.02 13 ClONO2 + H2 O HOCl + HNO3 JPL-06; = 0.3 14 BrONO2 + H2 O HOBr + HNO3 JPL-06; = 0.3 15 ClONO2 + HCl Cl2 + HNO3 JPL-06; = 0.3 16 HOCl + HCl Cl2 + H2 O JPL-06; = 0.2 17 HOBr + HCl BrCl + H2 O JPL-06; = 0.3

219

Table 5.4: WACCM4.0 Photolytic Reactions. no. 1 Reactants O2 + h Products O + O(1 D) Comments Ly-: Chabrillat and Kockarts (1997, 1998) (Ly-): Lacoursiere et al. (1999) SRB: Koppers and Murtaugh (1996) For wavelength regions not Ly- or SRB, (120-205nm): Brasseur and Solomon (1986); (205-240 nm): Yoshino et al. (1988) see above (120-136.5nm): Tanaka et al. (1953); (136.5-175nm): Ackerman (1971); (175-847nm): WMO (1985); except for (185-350nm): Molina and Molina (1986) (<280nm): Marsh (1999) (>280nm): JPL-06. see above JPL-06 Minschwaner et al. (1993) JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 (Ly-): Slanger et al. (1982); (105-145nm): Stief et al. (1975); (>145): JPL-06 (120-182nm): Yoshino et al. (1996); (183-194nm): Cantrell et al. (1997)

2 3

O2 + h O3 + h

2O O(1 D) + O2

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

O3 + h N2 O + h NO + h NO + h NO2 + h N2 O5 + h N2 O5 + h HNO3 + h NO3 + h NO3 + h HO2 NO2 + h HO2 NO2 + h CH3 OOH + h CH2 O + h CH2 O + h H2 O + h

O + O2 O(1 D) + N2 N+O NO+ + e NO + O NO2 + NO3 NO + O + NO3 OH + NO2 NO2 + O NO + O2 OH + NO3 NO2 + HO2 CH2 O + H + OH CO + 2 H CO + H2 H + OH

220

Table 5.4: (continued) WACCM4.0 Photolytic Reactions. no. 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 Reactants H2 O + h H2 O + h H2 O2 + h Cl2 + h ClO + h OClO + h Cl2 O2 + h HOCl + h HCl + h ClONO2 + h ClONO2 + h BrCl + h BrO + h HOBr + h BrONO2 + h BrONO2 + h CH3 Cl + h CCl4 + h CH3 CCl3 + h CFC11 + h CFC12 + h CFC113 + h HCFC22 + h CH3 Br + h CF3 Br + h CF2 ClBr + h CO2 + h CH4 + h CH4 + h Products H2 + O(1 D) H+2O 2 OH 2 Cl Cl + O O + ClO Cl + ClOO Cl + OH Cl + H Cl + NO3 ClO + NO2 Br + Cl Br + O Br + OH Br + NO3 BrO + NO2 Cl + CH3 O2 4 Cl 3 Cl 3 Cl 2 Cl 3 Cl Cl Br + CH3 O2 Br Br + Cl CO + O H + CH3 O2 H2 + 0.18 CH2 O + 0.18 O + 0.44 CO2 + 0.44 H2 + 0.38 CO + 0.05 H2 O Comments (see above) (see above) JPL-06 JPL-06 JPL-06 JPL-06 Burkholder et al. (1990); Stimpe et al. (2004) JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 JPL-06 (120-167): Nakata, et al. (1965); (167-199): Human (1971) : JPL-06; based on Brownsword et al. (1997) see above

221

Table 5.5: Ion-neutral and recombination reactions and exothermicities. Reaction O + + O2 O + + O 2 O+ + N2 NO+ + N N+ + O NO+ + N(2 D) 2 O+ + N NO+ + O 2 O+ + NO NO+ + O2 2 N+ + O2 O+ + N 2 N+ + O2 NO+ + O N+ + O O+ + N N+ + O2 O+ + N2 2 2 O+ + N2 NO+ + NO 2 N+ + O O+ + N2 2 NO+ + e 0.2N + 0.8N(2 D) + O O+ + e 1.15O + 0.85O(1 D) 2 N+ + e 1.1N + 0.9N(2 D) 2 H (kJ mol1 ) 150.11 105.04 67.53 406.16 271.38 239.84 646.28 95.55 339.59 82.389 508.95 354.83

4171

5.3.8

Electric Field

4172 4173 4174 4175

The global electric eld is based on a composite of two empirical models for the dierent latitude regions: at high latitude the Weimer95 model [Weimer, 1995], and at low- and midlatitude the Scherliess model [Scherliess et al., 2002]. In the following the dierent models are described since the model is not published to date. Low- and midlatitude electric potential model The low- and mid latitude electric eld model was developed by L dger Scherliess u [Scherliess et al., 2002]. Its based on Incoherent Scatter Radar data (ISR) from Jicamarca, Arecibo, Saint Santin, Millstone Hill, and the MU radar in Shigaraki. The electric eld is calculated for a given year, season, UT, Sa , local time, and with longitudinal/latitudinal variation. The empirical model is constructed from a model for low solar ux (Sa = 90) and a high solar ux model (Sa = 180). The global electric potential is expressed according to Richmond et al. [1980] by
2 2 n 12 m Aklmn Pn (sin)fm ( k=0 l=2 m=n n=1

4176

(d, T, t, ) =

2t ) 24

(5.88)

2(d + 9) 2T fl ( )fk ( ) 24 365.24

222

Table 5.6: Ionization reactions. O + h O+ + e O + e O+ + e + e N + hv N+ + e O2 + h O+ + e 2 O2 + e O+ + e + e 2 O2 + h O + O+ + e O2 + e O + O+ + e + e N2 + h N+ + e 2 N2 + e N+ + e + e 2 N2 + h N + N+ + e N2 + e N + N+ + e + e N2 + h N(2 D) + N+ + e N2 + e N(2 D) + N+ + e + e

wavelength interval nm 0.05 - 0.4 0.4 - 0.8 0.8 - 1.8 1.8 - 3.2 3.2 - 7.0 7.0 - 15.5 15.5 - 22.4 22.4 - 29.0 29.0 - 32.0 32.0 - 54.0 54.0 - 65.0 65.0 - 79.8 65.0 - 79.8 79.8 - 91.3 79.8 - 91.3 79.8 - 91.3 91.3 - 97.5 91.3 - 97.5 91.3 - 97.5 97.5 - 98.7 98.7 - 102.7 102.7 - 105.0 105.0 - 121.0

Table 5.7: EUVAC model parameters. Fi0 Ri ph cm2 s1 5.010e+01 6.240e-01 1.000e+04 3.710e-01 2.000e+06 2.000e-01 2.850e+07 6.247e-02 5.326e+08 1.343e-02 1.270e+09 9.182e-03 5.612e+09 1.433e-02 4.342e+09 2.575e-02 8.380e+09 7.059e-03 2.861e+09 1.458e-02 4.830e+09 5.857e-03 1.459e+09 5.719e-03 1.142e+09 3.680e-03 2.364e+09 5.310e-03 3.655e+09 5.261e-03 8.448e+08 5.437e-03 3.818e+08 4.915e-03 1.028e+09 4.955e-03 7.156e+08 4.422e-03 4.482e+09 3.950e-03 4.419e+09 5.021e-03 4.235e+09 4.825e-03 2.273e+10 3.383e-03

223

with fm () = 2 sin(m) fm () = 1 fm () = 2 cos(m) m>0 m=0 m<0 (5.89) (5.90) (5.91)

the day of the year is denoted by d, universal time by T , magnetic local time by t, and geomagm netic latitude . The values of d, T , and t are expressed as angles between 0 and 2. Pn are fully normalized Legendre polynomials. Due to the assumption that the geomagnetic eld lines are highly conducting, the n + m odd coecients are set to zero to get a symmetrical electric potential about the magnetic equator. The coecients Aklmn are found by a leastsquare t for low and high solar ux. The solar cycle dependence is introduced by inter- and extrapolation of the sets of coecients Alow for Sa = 90 and Ahigh for Sa = 180. klmn klmn Aklmn = Alow + SaM [Ahigh Alow ] klmn klmn klmn with SaM = a90 a180
4177 4178 4179

(5.92)

arctan[(Sa 65)2 /902] a90 a180 a90 = arctan[(90 65)2/902 ] = arctan[(180 65)2 /902 ]

(5.93) (5.94) (5.95)

We are using the daily F10.7 number for Sa . SaM levels o at high and low solar ux numbers, and therefore the model does not predict unrealistic high or low electric potential values. The geomagnetic eld is described by modied apex coordinates [Richmond, 1995] which already take into account the distortion of the magnetic eld. Modied apex coordinates have a reference height associated with them, which in our case is set to 130 km. The electric eld E and the electromagnetic drift velocity vE can be expressed by quantities mapped to the reference height, e.g. by Ed1 , Ed2 and ve1 , ve2 . These quantities are not actual electric eld or electromagnetic drift velocity components, but rather the representation of the electric eld or electromagnetic drift velocities by being constant along the geomagnetic eld line. The elds in an arbitrary direction I can be expressed by I E = I d1 Ed1 + I d2 Ed2 I vE = I e1 ve1 + I e2 ve2 (5.96) (5.97)

4180 4181 4182 4183 4184 4185 4186 4187

The basis vector d1 and e1 are in moreorless magnetic eastward direction and d2 and e2 in downward/ equatorward direction. The base vectors vary with height, di is decreasing and ei increasing with altitude. Therefore when the base vectors are applied to the mapped eld at the reference height, e.g. Ed1 , Ed2 and ve1 , ve2 , they already take into account the height and directional variation of the corresponding quantity. Note that the modied apex coordinates are using the International Geomagnetic Reference Field (IGRF), and in the WACCM4 code the IGRF is only dened between the years 1900 and 2000. The description of the IGRF can be updated every 5 years to be extended in time. 224

4188

Highlatitude electric potential model The highlatitude electric potential model from Weimer [Weimer, 1995] is used. The model is based on spherical harmonic coecients that were derived by least square tting of measurements from the Dynamics Explorer 2 (DE2) satellite. The variation of the spherical harmonic coecients with the interplanetary magnetic eld (IMF) clock angle, IMF strength, solar wind velocity and season can be reproduced by a combination of Fourier series and multiple linear regression formula. The nal model varies with magnetic latitude, magnetic local time, season, IMF strength and direction, and solar wind velocity. For our purpose we have set the solar wind speed to a constant value of 400 km/s and only consider the eects of IMF Bz (By = 0). Since the IMF conditions are not known all the time, we developed an empirical relation between Bz and the Kp index and the solar ux number Sa . Both, the Kp index and the daily solar ux number F10.7 , are known in the WACCM4 model.
2 Bz (Kp , F10.7 ) = 0.085Kp 0.08104Kp + 0.4337+

0.00794F10.7 0.00219Kp F10.7

(5.98)

4189 4190 4191 4192

Note that the Weimer model uses an average year of 365.24 days/year and an average month of 30.6001 days/month. The boundary of the Weimer model is at 46o magnetic latitude. The model was developed for an averaged northern and southern hemisphere. The By value and the season are reversed to get the values for the other hemisphere. Combing low/ midlatitude with the high latitude electric potential After the low/midlatitude electric potential mid and the high latitude potential hgh are calculated, both patterns are combined to be smooth at the boundary. The boundary between high and mid latitude bnd is dened to lie where the electric eld magnitude E from hgh equals 15 mV/m. After nding the longitudinal variation of the high latitude boundary bnd , its shifted halfway towards 54o magnetic latitude. The width of the transition zone 2trs from high to mid latitude varies with magnetic local time. First, the high and mid latitude electric potential are adjusted by a constant factor such that the average for the high and mid latitude electric potential along the boundary bnd are the same. The combined electric potential is dened by || < bnd trs mid = hgh (5.99) || > bnd + trs Fint (mid , hgh ) bnd trs || bnd + trs with Fint (mid , hgh ) = 1 1 [ {mid (, bnd trs ) + 2mid (, )} 3 2trs {bnd || + trs } + (hgh (, bnd + trs )+ 2hgh (, )) {bnd + || + trs }] (5.100)

4193

225

4194

Calculation of electric eld The electric eld can be derived from the electric potential by E = (5.101)

The more-or-less magnetic eastward electric eld component Ed1 and the in general downward/ equatorward Ed2 component are calculated. These components are constant along the magnetic eld line. They are calculated at a reference height hr = 130 km with R = Rearth + hr . The electric eld does not vary much with altitude, and therefore we assume in the code that the electric eld is constant in height. Ed1 = Ed2
4195

1 Rcos 1 = R sin I

(5.102) (5.103)

with sin I = 2 sin [4 3 cos2 ]0.5 . Calculation of electrodynamic drift velocity The electric eld is calculated on a 2o 2o degree geomagnetic grid with the magnetic longitude represented by the magnetic local time (MLT) from 0 MLT to 24 MLT. Therefore, the magnetic local time of the geographic longitudes of the WACCM4 grid has to be determined rst to map from the geomagnetic to the geographic WACCM4 grid. The magnetic local time is calculated by using the location of the geomagnetic dipole North pole, the location of the subsolar point, and the apex longitude of the geographic WACCM4 grid point. A bilinear interpolation is used for the mapping. Note that every processor calculates the global electric eld, which is computationally inexpensive. Otherwise, to calculate the electric eld some communication between the dierent processors would be necessary to get the spatial derivatives. The mapped electric eld is rotated into the geographic direction by E = d1 Ed1 + d2 Ed2 (5.104)

4196

with the components of E being the geographic eastward, westward and upward electric eld. At high altitudes the ionneutral collision frequency in is small in relation to the angular gyrofrequency of the ions i (in i ), and the electronneutral collision frequency en is much smaller than the angular gyrofrequency of the electrons e (en e ), due to the decrease in neutral density with increasing altitude. Therefore, the ion drift vi perpendicular to the geomagnetic eld can be simplied by the electrodynamic drift velocity vE vi vE =
4197

E Bo B2 o

(5.105)

with Bo the geomagnetic main eld from IGRF.

226

4198

Ion drag calculation The following is written according to the source code. Two subroutines iondrag calc exist in the code, one uses the calculated ion drag coecients if WACCM MOZART is used, and the other one uses look-up tables for the ion drag coecients 1 and 2 . It is assumed that the electron Te and ion Ti temperature is equal to the neutral temperature Tn . Ti = Te = Tn The dip angle I of the geomagnetic eld is calculated by I = arctan
2 Bnorth

(5.106)

Bz 2 + Beast

(5.107)

with a minimum dip angle |I| 0.17. The declination is D = arctan Beast Bnorth (5.108)

The magnetic eld component Bz , Beast , Bnorth are determined from the International Geomagnetic Reference Field (IGRF). The collision frequencies in units of s1 are determined by, e.g. Schunk and Nagy [2000] 1 + = 2.59 1011 NO2 O2 O2 1 O+ O2 = 6.64 1010 NO2 1 N O+ O2 = 4.27 1010 NO2 1 O+ O = 3.67 1011 NO 1 N O+ O NO 1 + NO O2 O 1 + NN2 O2 N2 1 N O+ N2 NN2 1 O+ N2 NN2 = 2.44 1010 = 2.31 1010 = 4.13 1010 = 4.34 1010 = 6.82 1010 Ti + Te 1 0.064log10 2 Ti + Te 2
2

Ti + Te 1 0.73log10 2

Ti + Te 2

(5.109) (5.110) (5.111) fcor (5.112) (5.113) (5.114) (5.115) (5.116) (5.117)

with Nn the number density for the neutral n in units of 1/cm3 , and the temperature in Kelvins. + The collisions frequencies for O2 O2 and O+ O are resonant, all other are nonresonant. The 227

arbitrary correction factor fcor multiplies the O+ O collision frequency and is set to fcor = 1.5 which has been found to improve agreement between calculated and observed winds and electron densities in the upper thermosphere in other models. The mean mass mmid [g/mole] at the midpoints of the height level is calculated in the Mozart module. The number densities [1/cm3 ] are Nmmid mmrO2 (5.118) NO2 = mO2 Nmmid mmrO NO = (5.119) mO Nmmid mmrN2 (5.120) NN2 = mN2 + Nmmid mmrO2 + NO2 = (5.121) + mO2 NO+ = Nmmid mmrO+ mO+ Nmmid mmre Ne = me (5.122) (5.123)

with mmr the mass mixing ratio, and N the total number density in units of 1/cm3 . The pressure [dyne/cm2 ] and the mean mass at the midpoint mmid in units of g/mole are p = 10 pmid pm Nmmid = kB Tn (5.124) (5.125)

with the factor 10 to convert from [Pa] to [dyne/cm2 ], and kB the Boltzmann constant. The collision frequencies are
+ + + + O2 = O2 O2 + O2 O + O2 N2

(5.126) (5.127) (5.128) Te )+ (5.129)

O+ = O+ O2 + O+ O + O+ N2 N O+ = N O+ O2 + N O+ O + N O+ N2 en = 2.33 1011 NN2 Te (1 1.21 104 Te )+ 8.9 1011 NO 1.82 1010 NO2 Te (1 + 5.7 104Te ) Te (1 + 3.6 102

The ratios r between collision frequency and gyro frequency are + O2 + rO 2 = + O2 O+ rO + = O+ N O+ rN O + = N O+ en re = e 228

(5.130) (5.131) (5.132) (5.133)

with the gyro frequency for ions i = eB/mi and for electrons e = eB/me . The Pedersen conductivity [ S/m] is + rO 2 e rO + + P = [NO+ + NO2 + 2 2 B 1 + rO + 1 + rO + 2 (5.134) rN O + re NN O+ ] + Ne 2 2 1 + rN O + 1 + re The Hall conductivity [S/m] is e 1 1 + H = [NO+ NO2 2 2 B 1 + rO + 1 + rO +
2

1 1 ] + Ne NN O+ 2 2 1 + rN O + 1 + re The ion drag coecients are 1 = P B 2 H B 2 2 =

(5.135)

(5.136) (5.137)

m with = N NA , and NA the Avagadro number. The ion drag tensor in magnetic direction mag is

mag =

mag mag xx xy mag mag yy yx

1 2 sinI 2 sinI 1 sin2 I

(5.138)

with the xdirection in magnetic east, and ydirection magnetic north in the both hemispheres. The ion drag tensor can be rotated in geographic direction by using the rotation matrix R R= cos D sin D sin D cos D xx xy yx yy (5.139)

Applying the rotation to the ion drag tensor Rmag R1 leads to = mag yx = (5.140) (5.141)

mag cos2 D + mag sin2 D mag + (mag mag ) sin D cos D xx yy xy yy xx mag mag + (yy xx ) sin D cos D mag cos2 D + mag sin2 D yy xx JB = 1 (vE un ) + 2 b (vE un )

The ion drag acceleration ai due to the Amp`re force is e ai = (5.142)

with un the neutral wind velocity perpendicular to the geomagnetic eld and b the unit vector of the geomagnetic eld. The tendencies on the neutral wind are calculated by vEn = vEn t 229 (5.143)

For stability an implicit scheme is used with vEn (t + t) vEn (t) = vEn (t + t) t which leads to ( 1 1 I + )vEn (t + t) = vEn (t) t t (5.145) (5.144)

with I the unit matrix. Solving for vEn (t + t) gives vEn (t + t) = The tendencies are determined by vEn vEn (t + t) vEn (t) 1 1 1 = = [ ( I + )1 1]vEn (t) t t t t t The tensor
1 I t

1 1 ( I + )1 vEn (t) t t

(5.146)

(5.147)

+ is 11 12 21 22 =
1 t

+ xx yx

1 t

xy + yy

(5.148)

1 1 Det = t t 11 22 12 21 The tendencies applied to the neutral winds with vEn = (uE un , vE vn ) gives dt u i = 1 Det ( (vE vn ) (uE un )) + uE un 22 t t 12 1 Det ( (uE un ) (vE vn )) + vE vn dt vi = 11 t t 21

(5.149)

(5.150) (5.151)

4199 4200

The electromagnetic energy transfer to the ionosphere is J E = J E + un J B (5.152)

The rst term on the right hand side denotes the Joule heating, which is the electromagnetic energy transfer rate in the frame of reference of the neutral wind. The second term represents the generation of kinetic energy due to the Amp`re force. Since the electric eld is small along e the magnetic eld line, we consider only the perpendicular component to the magnetic eld of the Joule heating J E. The electric eld in the frame of the neutral wind u can be written as E = E + u B (5.153)

230

The Joule heating can be expressed by J E = P E2 with E2 = B 2 ( EB u )2 B2 (5.155) (5.154)

and EB the electromagnetic drift velocity vE with the components uE and vE . The Joule B2 heating QJ is QJ = (uE un )2 xx + (uE un )(vE vn )(xy yx )+ (vE vn )2 yy
4201

(5.156)

Note, that the vertical velocity components are not taken into account here.

4202

5.3.9

Boundary Conditions

4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231

The upper boundary conditions for momentum and for most constituents are the usual zero ux conditions used in CAM4. However, in the energy budget of the thermosphere, much of the SW radiation at wavelengths <120 nm is absorbed above 145 km (the upper boundary of the model), where LW radiation is very inecient. This energy is transported downward by molecular diusion to below 120 km, where it can be dissipated more eciently by LW emission. Imposing a zero ux upper boundary condition on heat omits a major term in the heat budget and causes the lower thermosphere to be much too cold. Instead, we use the Mass Spectrometer-Incoherent Scatter (MSIS) model [Hedin, 1987, 1991] to specify the temperature at the top boundary as a function of season and phase of the solar cycle. The version of the MSIS model used in WACCM4.0is NRLMSISE-00 [see http://uapwww.nrl.navy.mil/models web/msis/msis home.htm]. For chemical constituents, surface mixing ratios of CH4 , N2 O, CO2 , H2 , CFC-11, CFC-12, CFC-113, HCFC-22, H-1211, H-1301, CCl4 , CH3 CCH3 , CH3 Cl, and CH3 Br are specied from observations. The model accounts for surface emissions of NOX and CO based on the emission inventories described in Horowitz et al. [2003]. The NOX source from lightning is distributed according to the location of convective clouds based on Price et al. [1997a] and Price et al. [1997b], with a vertical prole following Pickering et al. [1998]. Aircraft emissions of NOX and CO are included in the model and based on Friedl [1997]. At the upper boundary, a zero-ux upper boundary condition is used for most species whose mixing ratio is negligible in the lower thermosphere, while mixing ratios of other species are specied from a variety of sources. The MSIS model is used to specify the mixing ratios of O, O2 , H, and N; as in the case of temperature, the MSIS model returns values of these constituents as functions of season and phase of the solar cycle. CO and CO2 are specied at the upper boundary using output from the TIME-GCM [Roble and Ridley, 1994]. NO is specied using data from the Student Nitric Oxide Explorer (SNOE) satellite [Barth et al., 2003], which has been parameterized as a function of latitude, season, and phase of the solar cycle in the Nitric Oxide Empirical Model (NOEM) of Marsh et al. [2004]. Finally, a global-mean value (typical of the sunlit lower thermosphere) is specied for species such as H2 O, whose abundance near the top of the model is very small under sunlit conditions, but which can be rapidly transported upward 231

4232 4233 4234

by diusive separation in polar night (since they are lighter than the background atmosphere). In these cases, a zero-ux boundary condition leads to unrealistically large mixing ratios at the model top in polar night.

232

4235

Chapter 6 Initial and Boundary Data

6.1 Initial Data

4236

4237

In this section, we describe how the time integration is started from data consistent with the spectral truncation. The land surface model requires its own initial data, as described by Bonan [1996]. The basic initial data for the model consist of values of u, v, T, q, , and s on the Gaussian grid at time t = 0. From these, U, V, T , and are computed on the grid using (3.139), and (3.177). The Fourier coecients of these variables U m , V m , T m , m , and m are determined s m via an FFT subroutine (3.277), and the spherical harmonic coecients Tn , m , and (s )m are n n determined by Gaussian quadrature (3.278). The relative vorticity and divergence spherical harmonic coecients are determined directly from the Fourier coecients U m and V m using the relations, = V 1 U 1 , 2 ) a(1 a U 1 V 1 + . = 2 ) a(1 a (6.1) (6.2)

4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253

The relative vorticity and divergence coecients are obtained by Gaussian quadrature directly, using (3.282) for the derivative terms and (3.285) for the derivatives. Once the spectral coecients of the prognostic variables are available, the gridpoint values of , , T , , and s may be calculated from (3.308), the gradient from (3.311) and (3.312), and U and V from (3.317) and (3.318). The absolute vorticity is determined from the relative vorticity by adding the appropriate associated Legendre function for f (3.245). This process gives gridpoint elds for all variables, including the surface geopotential, that are consistent with the spectral truncation even if the original gridpoint data were not. These gridpoint values are then convectively adjusted (including the mass and negative moisture corrections). The rst time step of the model is forward semiimplicit rather than centered semiimplicit, so only variables at t = 0 are needed. The model performs this forward step by setting the variables at time t = t equal to those at t = 0 and by temporarily dividing 2t by 2 for this time step only. This is done so that formally the code and the centered prognostic equations of chapter 3 also describe this rst forward step and no additional code is needed for this special step. The model loops through as indicated sequentially in chapter 3. The time step 2t is set to its original value before beginning the second time step. 233

4254

6.2

Boundary Data

4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284

In addition to the initial gridpoint values described in the previous section, the model also requires lower boundary conditions. The required data are surface temperature (Ts ) at each ocean point, the surface geopotential at each point, and a ag at each point to indicate whether the point is land, ocean, or sea ice. The land surface model requires its own boundary data, as described by Bonan [1996]. A surface temperature and three subsurface temperatures must also be provided at non-ocean points. For the uncoupled conguration of the model, a seasonally varying seasurface temperature, and seaice concentration dataset is used to prescribe the time evolution of these surface quantities. This dataset prescribes analyzed monthly mid-point mean values of SST and ice concentration for the period 1950 through 2001. The dataset is a blended product, using the global HadISST OI dataset prior to 1981 and the Smith/Reynolds EOF dataset post-1981 (see Hurrell, 2002). In addition to the analyzed time series, a composite of the annual cycle for the period 1981-2001 is also available in the form of a mean climatological dataset. The sea surface temperature and sea ice concentrations are updated every time step by the model at each grid point using linear interpolation in time. The mid-month values have been evaluated in such a way that this linear time interpolation reproduces the mid-month values. Earlier versions of the global atmospheric model (the CCM series) included a simple landocean-sea ice mask to dene the underlying surface of the model. It is well known that uxes of fresh water, heat, and momentum between the atmosphere and underlying surface are strongly aected by surface type. The CAM 5.0 provides a much more accurate representation of ux exchanges from coastal boundaries, island regions, and ice edges by including a fractional specication for land, ice, and ocean. That is, the area occupied by these surface types is described as a fractional portion of the atmospheric grid box. This fractional specication provides a mechanism to account for ux dierences due to sub-grid inhomogeneity of surface types. In CAM 5.0 each atmospheric grid box is partitioned into three surface types: land, sea ice, and ocean. Land fraction is assigned at model initialization and is considered xed throughout the model run. Ice concentration data is provided by the external time varying dataset described above, with new values determined by linear interpolation at the beginning of every time-step. Any remaining fraction of a grid box not already partitioned into land or ice is regarded as ocean. Surface uxes are then calculated separately for each surface type, weighted by the appropriate fractional area, and then summed to provide a mean value for a grid box: F T = ai F i + ao F o + al F l , (6.3)

4285 4286 4287 4288 4289 4290 4291 4292 4293

where F denotes the surface ux of the arbitrary scalar quantity , a denotes fractional area, and the subscripts T, i, o, and l respectively denote the total, ice, ocean, and land components of the uxes. For each time-step the aggregated grid box uxes are passed to the atmosphere and all ux arrays which have been used for the accumulations are reset to zero in preparation for the next time-step. The fractional land values for CAM 5.0 were calculated from Navy 10-Min Global Elevation Data. An area preserving binning algorithm was used to interpolate from the high-resolution Navy dataset to standard model resolutions. The radiation parameterization requires monthly mean ozone volume mixing ratios to be specied as a function of the latitude grid, 23 vertical pressure levels, and time. The ozone path 234

4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305

lengths are evaluated from the mixingratio data. The path lengths are interpolated to the model layer interfaces for use in the radiation calculation. As with the seasurface temperatures, the seasonal version assigns the monthly averages to the midmonth date and updates them every 12 hours via linear interpolation. The actual mixing ratios used in the standard version were derived by Chervin [1986] from analysis of D tsch [1986]. u The sub-grid scale standard deviation of surface orography is specied in the following manner. The variance is rst evaluated from the global Navy 10 topographic height data over an intermediate grid (e.g. 2 2 grid for T42 and lower resolutions, 1.67 1.67 for T63, and 1.0 1.0 for T106 resolution) and is assumed to be isotropic. Once computed on the appropriate grid, the standard deviations are binned to the CAM 5.0 grid (i.e., all values whose latitude and longitude centers fall within each grid box are averaged together). Finally, the standard deviation is smoothed twice with a 121 spatial lter. Values over ocean are set to zero.

235

236

4306

Appendix A Physical Constants

Following the American Meteorological Society convention, the model uses the International System of Units (SI) (see August 1974 Bulletin of the American Meteorological Society, Vol. 55, No. 8, pp. 926-930). a g ts B k N R mair R mv Rv cp zvir Lv Li H2 O cpv Tmelt pstd air = = = = = = = = = = = = = = = = = = = = = = = 6.37122 106 m 9.80616 m s2 3.14159265358979323846 86164.0 s 2 /ts [s1 ] 5.67 108 W m2 K4 1.38065 1023 JK1 6.02214 1026 k N [JK1 ] 28.966 kg R /mair [J kg1 K1 ] 18.016 kg R /mv [J kg1 K1 ] 1.00464 103 J kg1 K1 2/5 Rv /R 1 2.501 106 J kg1 3.337 105 J kg1 1.0 103 kg m3 1.81 103 J kg1 K1 273.16 K 1.01325 105 Pa pstd /(R Tmelt ) [kgm3 ] Radius of earth Acceleration due to gravity Pi Earth s sidereal day Earth s angular velocity Stefan Boltzmann constant Boltzmann constant Avogadro s number Universal gas constant Molecular weight of dry air Gas constant for dry air Molecular weight of water vapor Gas constant for water vapor Specic heat of dry air at constant pressure Von Karman constant Ratio of gas constants for water vapor and dry air Latent heat of vaporization Latent heat of fusion Density of liquid water Specic heat of water vapor at constant pressure Melting point of ice Standard pressure Density of dry air at standard pressure/temperature

4307

4308

The model code denes these constants to the stated accuracy. We do not mean to imply that these constants are known to this accuracy nor that the low-order digits are signicant to the physical approximations employed.

4309 4310

237

238

4311

Appendix B Acronyms
ABL AMIP AMWG BATS CAM CAPE CCM CCN CCSM CFC CFL CGD CGS CKD CLM CMS CSIM CWP DAO DAS DISORT ECMWF EOF FASCODE FFSL FFT FV/fv GCM GENLN GEOS GFDL GSFC GMT Atmospheric Boundary Layer Atmospheric Model Intercomparison Project Atmospheric Model Working Group Biosphere-Atmosphere Transfer Scheme Community Atmosphere Model Convectively Available Potential Energy Community Climate Model Cloud Condensation Nucleus Community Climate System Model Chloro-Fluoro Carbon Courant-Friedrichs-Levy Condition NCAR Climate and Global Dynamics Division Centimeters/grams/seconds Clough-Kneizys-Davies Community Land Model (NCAR) Climate Modeling Section Community Sea-Ice Model Condensed Water Path (NASA Goddard) Data Assimilation Oce Data Assimilation System DIScrete-Ordinate method Radiative Transfer European Centre for Medium Range Forecasts Empirical Orthogonal Function FASt atmosphere Signature Code Flux-Form Semi-Lagrangian Transport Fast Fourier Transform Finite Volume General Circulation Model General Line-by-line Atmospheric Transmittance and Radiance Model Goddard Earth Observing System Geophysical Fluid Dynamics Laboratory Goddard Space Flight Center Greenwich Mean Time 239

4312

4313

4314

HadISST HITRAN ICA IPCC KNMI LBL LCL LSM MATCH M/R NASA NCAR NCEP NOAA NWP OI OPAC PBL PCMDI PPM RHS RMS SCMO SI SOM SST TOA TOM UCAR WKB

Hadley Centre for Climate Prediction and Research SST High-resolution Transmission Molecular Absorption Database Independent Column Approximation International Panel on Climate Change Royal Netherlands Meteorological Institute Line by line Lifting condensation level Land Surface Model Model for Atmospheric Transport and Chemistry Maximum/Random overlap National Space Administration National Center for Atmospheric Research National Center for Environmental Prediction National Oceanographic and Atmospheric Administration Numerical Weather Prediction Optimal Interpolation Optical Properties of Aerosols and Clouds Planetary Boundary Layer Program for Climate Model Diagnosis and Intercomparison Piece-wise Parabolic Method Right Hand Side Root-mean Square Sucient Condition for Monotonicity International System of Units Slab Ocean Model Sea-surface temperature Top Of Atmosphere Top Of Model University Corporation for Atmospheric Research Wentzel-Kramer-Brillouin approximation

240

4315

Appendix C Resolution and dycore-dependent parameters

The following adjustable parameters dier between various nite volume resolutions in the CAM 5.0. Refer to the model code for parameters relevant to alternative dynamical cores. Table C.1: FV 1 deg FV 2 deg 2.e-4 2.e-4 18.e-6 9.5e-6 5.e-6 5.e-6 .92 .91 .77 .80 0.10 0.10 750.e2 750.e2 1.0e-4 1.0e-4 3.5E-3 3.5E-3 1.0E-6 1.0E-6 1.0 0.5 Resolution-dependent parameters Description threshold for autoconversion of warm ice threshold for autoconversion of cold ice stratiform precipitation evaporation eciency parameter minimum RH threshold for low stable clouds minimum RH threshold for high stable clouds parameter for deep convection cloud fraction top of area dened to be mid-level cloud shallow convection precip production eciency parameter deep convection precipitation production eciency parameter convective precipitation evaporation eciency parameter Stokes ice sedimentation fall speed (m/s)

4316

4317

4318 4319

Parameter qic,warm qic,cold ke,strat low RHmin high RHmin k1,deep pmid c0,shallow c0,deep ke,conv vi

241

242

4320

Bibliography
Abdul-Razzak, H., and S. J. Ghan, A parameterization of aerosol activation 2. multiple aerosol types, J. Geophys. Res., 105 (D5), 68376844, 2000a. Abdul-Razzak, H., and S. J. Ghan, A parameterization of aerosol activation 2. multiple aerosol types, Journal of Geophysical Research-Atmospheres, 105 (D5), 68376844, 2000b, 294KT Times Cited:82 Cited References Count:27. Alcamo, J., A. Bouwman, J. Edmonds, A. Grubler, T. Morita, and A. Sugandhy, An evaluation of the ipcc is92 emission scenarios, in Climate Change 1994: Radiative Forcing of Climate Change and an Evaluation of the IPCC IS92 Emission Scenarios, edited by J. Houghton, 247304, Cambridge University Press, 1995. Andrews, D. G., J. R. Holton, and C. B. Leovy, Middle Atmosphere Dynamics. Academic Press, 1987. Ansmann, A. et al., Inuence of Saharan dust on cloud glaciation in southern Morocco during the Saharan Mineral Dust Experiment, J. Geophys. Res., 113 (D04210), 2008. Ansmann, A., M. Tesche, P. Seifert, D. Althausen, R. Engelmann, J. Furntke, U. Wandinger, I. Mattis, and D. M ller, Evolution of the ice phase in tropical altocumulus: SAMUM lidar u observations over Cape Verde, J. Geophys. Res., 114 (D17208), 2009. Anthes, R. A., Summary of workshop on the NCAR Community Climate/forecast Models 1426 July 1985, Boulder, Colorado, Bull. Am. Meteorol. Soc., 67, 94198, 1986. Arakawa, A., and V. R. Lamb, A potential enstrophy and energy conserving scheme for the shallow-water equations, Mon. Wea. Rev., 109, 1836, 1981. Asselin, R., Frequency lter for time integrations, Mon. Wea. Rev., 100, 487490, 1972. Baede, A. P. M., M. Jarraud, and U. Cubasch, Adiabatic formulation and organization of ECMWFs model, Technical Report 15, ECMWF, Reading, U.K., 1979. Balachandran, N., and D. Rind, Modeling the eects of uv variability and the qbo on the troposphere-stratosphere system. part i: The middle atmosphere, J. Climate, 8, 20582079, 1995. Banks, P. M., and G. Kockarts, Aeronomy, Part B. Academic Press, San Diego, Calif., 1973, 355 pp.

4321 4322

4323 4324 4325

4326 4327 4328 4329

4330 4331

4332 4333

4334 4335 4336

4337 4338

4339 4340

4341

4342 4343

4344 4345 4346

4347 4348

243

4349 4350 4351

Barker, H., A parameterization for computing grid-averaged solar uxes for inhomogenous marine boundary layer clouds. part I: Methodology and homgenous biases, J. Atmos. Sci., 53, 22892303, 1996. Barker, H., B. A. Weilicki, and L. Parker, A parameterization for computing grid-averaged solar uxes for inhomogenous marine boundary layer clouds. part II: Validation using satellite data, J. Atmos. Sci., 53, 23042316, 1996. Barth, C. A., K. D. Manko, S. M. Bailey, and S. Solomon, Global observations of nitric oxide in the thermosphere, J. Geophys. Res., 108, doi:10.1029/2002JA009458, 2003. Barth, M. C., P. J. Rasch, J. T. Kiehl, C. M. Benkovitz, and S. E. Schwartz, Sulfur chemistry in the National Center for Atmospheric Research Community Climate Model: Description, evaluation, features and sensitivity to aqueous chemistry, J. Geophys. Res., 105, 13871415, 2000. Bates, J. R., F. H. M. Semazzi, R. W. Higgins, and S. R. Barros, Integration of the shallow water equations on the sphere using a vector semi-Lagrangian scheme with a multigrid solver, Mon. Wea. Rev., 118, 16151627, 1990. Bath, L., J. Rosinski, and J. Olson, Users Guide to NCAR CCM2, Technical Report NCAR/TN379+IA, National Center for Atmospheric Research, Boulder, CO, 156 pp., 1992. Bath, L. M., M. A. Dias, D. L. Williamson, G. S. Williamson, and R. J. Wolski, Users Guide to NCAR CCM1, Technical Report NCAR/TN-286+IA, National Center for Atmospheric Research, Boulder, CO, 173 pp., 1987. Beheng, K. D., A parameterization of warm cloud microphysical conversion processes, Atmos. Res., 33, 193206, 1994. Beres, J. H., R. R.Garcia, B. A. Boville, and F. Sassi, Implementation of a gravity wave source spectrum parameterization dependent on the properties of convection in the whole atmosphere community climate model (waccm), J. Geophys. Res., 110, doi:10.1029/2004JD005504, 2005. Berger, A. L., Long-term variations of daily insolation and quaternary climatic changes, J. Atmos. Sci., 35, 23622367, 1978. Biermann, U. M., B. P. Luo, and T. Peter, Absorption spectra and optical constants of binary and ternary solutions of h2so4, hno3, and h2o in the mid infrared at atmospheric temperatures, The Journal of Physical Chemistry A, 104 (4), 783793, 01, 2000. Bigg, E. K., The supercooling of water, Proc. Phys. Soc. B, 66, 688694, 1953. Binkowski, F. S., and U. Shankar, The regional particulate matter model .1. model description and preliminary results, Journal of Geophysical Research-Atmospheres, 100, 2619126209, 1995. Blackadar, A. K., The vertical distribution of wind and turbulent exchange in a neutral atmosphere, J. Geophys. Res., 67, 30953102, 1962. 244

4352 4353 4354

4355 4356

4357 4358 4359 4360

4361 4362 4363

4364 4365

4366 4367 4368

4369 4370

4371 4372 4373

4374 4375

4376 4377 4378

4379

4380 4381 4382

4383 4384

4385 4386 4387

Bloomeld, P., Inferred lifetimes, in NASA Report on Concentrations, Lifetimes and Trends of CFCs, Halons and Related Species, edited by J. Kaye, S. Penkett,, and F. Osmond, NASA, Washington, D.C., 1994. Bonan, G. B., A land surface model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: Technical description and users guide, Technical Report NCAR/TN417+STR, National Center for Atmospheric Research, Boulder, CO, 150 pp., 1996. Bonan, G. B., K. W. Oleson, M. Vertenstein, S. Levis, X. Zeng, Y. Dai, R. E. Dickinson, , and Z.-L. Yang, The land surface climatology of the Community Land Model coupled to the NCAR Community Climate Model, J. Climate, 15, 31233149, 2002. Bond, T. C., and R. W. Bergstrom, Light absorption by carbonaceous particles: An investigative review, Aerosol. Sci. Technol., 40, 2767, 2006. Bond, T. C., E. Bhardwaj, R. Dong, R. Jogani, S. K. Jung, C. Roden, D. G. Streets, and N. M. Trautmann, Historical emissions of black and organic carbon aerosol from energyrelated combustion, 1850-2000, Global Biogeochemical Cycles, 21 (2), , 2007, 174KO Times Cited:24 Cited References Count:98. Bourke, W., B. McAvaney, K. Puri, and R. Thurling, Global modeling of atmospheric ow by spectral methods, in Methods in Computational Physics, Vol. 17, 267324, Academic Press, New York, 1977. Boville, B. A., and C. S. Bretherton, Heating and dissipation in the NCAR community atmosphere model, J. Climate, 16, 38773887, 2003a. Boville, B. A., and C. S. Bretherton, Heating and kinetic energy dissipation in the ncar community atmosphere model, J. Climate, 16, 38773887, 2003b. Boville, B. A., J. T. Kiehl, P. J. Rasch, and F. O. Bryan, Improvements to the ncar csm-1 for transient climate simulations, J. Climate, 14, 164179, 2001. Brasseur, G. P., D. A. Hauglustaine, S. Walters, P. J. Rasch, J.-F. Muller, C. Granier, and X.-X. Tie, Mozart: A global chemical transport model for ozone and related chemical tracers, part 1: Model description, J. Geophys. Res., 103, 28,26528,289, 1998. Brasseur, G. P., J. J. Orlando, and G. S. Tyndall, Atmospheric Chemistry and Global Change. Oxford University Press, 1999. Bretherton, C. S., J. R. McCaa, and H. Grenier, A new parameterization for shallow cumulus convection and its application to marine subtropical cloud-topped boundary layer. part i: Description and 1d results, Mon. Wea. Rev., 132, 864882, 2004. Bretherton, C. S., and S. Park, A new moist turbulence parameterization in the community atmosphere model, J. Climate, 22, 34223448, 2009a. Bretherton, C. S., and S. Park, A new moist turbulence parameterization in the Community Atmosphere Model, J. Clim., 22, 34223448, 2009b. 245

4388 4389 4390

4391 4392 4393

4394 4395

4396 4397 4398 4399

4400 4401 4402

4403 4404

4405 4406

4407 4408

4409 4410 4411

4412 4413

4414 4415 4416

4417 4418

4419 4420

4421 4422

Briegleb, B. P., Delta-Eddington approximation for solar radiation in the NCAR Community Climate Model, J. Geophys. Res., 97, 76037612, 1992. Bryan, F. O., B. G. Kauman, W. G. Large, and P. R. Gent, The NCAR CSM Flux Coupler, Technical Report NCAR/TN-424+STR, National Center for Atmospheric Research, Boulder, Colorado, 58 pp., 1996. Bryant, F. D., and P. Latimer, Optical eciencies of large particles of arbitrary shape and orientation, J. Colloid Interface Sci., 30, 291304, 1969. Cariolle, D., A. Lasserre-Bigorry, J.-F. Royer, and J.-F. Geleyn, A general circulation model simulation of the springtime antarctic ozone decrease and its impact on mid-latitudes., J. Geophys. Res., 95, 18831898, 1990. Chabrillat, S., and G. Kockarts, Simple parameterization of the absorption of the solar lymanalpha line, Geophys. Res. Lett., 24, 26592662, 1997. Chervin, R. M., Interannual variability and seasonal climate predictability, J. Atmos. Sci., 43, 233251, 1986. Clough, S. A., M. W. Shephard, E. H. Mlawer, J. S. Delamere, M. J. Iacono, K. Cady-Pereira, S. Boukabara, and P. D. Brown, Atmospheric radiative transfer modeling: a summary of the aer codes, J. Quant. Spectrosc. Radiat. Transfer, 91, 233244, 2005. Clough, S., and M. Iacono, Line-by-line calculation of atmospheric uxes and cooling rates 2. application to carbon dioxide, ozone, methane, nitrous oxide and the halocarbons., J. Geophys. Res., 100, 1651916535, 1995. Colella, P., and P. R. Woodward, The piecewise parabolic method (ppm) for gas-dynamical simulations, J. Comp. Phys., 54, 174201, 1984. Collins, W. D., A global signature of enhanced shortwave absorption by clouds, J. Geophys. Res., 103(D24), 31,66931,680, 1998. Collins, W. D., P. J. Rasch, B. A. Boville, J. J. Hack, J. R. MaCaa, D. L. Williamson, J. T. Kiehl, B. P. Briegleb, C. Bitz, S. J. Lin, M. Zhang, and Y. Dai, Description of the ncar community atmosphere model (cam 3.0), Technical report, National Center for Atmospheric Research, 2004a. Collins, W. D., P. J. Rasch, B. A. Boville, J. J. Hack, J. R. McCaa, D. L. Williamson, J. T. Kiehl, B. P. Briegleb, C. Bitz, S.-J. Lin, M. Zhang, and Y. Dai, Description of the NCAR Community Atmosphere Model (CAM3). Nat. Cent. for Atmos. Res., Boulder, Colo., 2004b. Collins, W., J. Hackney, and D. Edwards, A new parameterization for infrared emission and absorption of water vapor in the national center for atmospheric research community atmospheric model, J. Geophys. Res., 107, doi: 10.1029/2001JD001365, 2002.

4423 4424 4425

4426 4427

4428 4429 4430

4431 4432

4433 4434

4435 4436 4437

4438 4439 4440

4441 4442

4443 4444

4445 4446 4447 4448

4449 4450 4451

4452 4453 4454

246

4455 4456 4457 4458

Considine, D. B., A. R. Douglass, D. E. Kinnison, P. S. Connell, and D. A. Rotman, A polar stratospheric cloud parameterization for the three dimensional model of the global modeling initiative and its response to stratospheric aircraft emissions, J. Geophys. Res., 105, 3955 3975, 2000. Considine, D. B., P. S. Connell, D. Bergmann, D. A. Rotman, and S. Strahan, Sensitivity of Global Modeling Initiative CTM predictions of Antarctic ozone recovery to GCM and DAS generated meteorological elds, J. Geophys. Res., 109, doi:10.1029/20003JD004487, 2004. Cooke, W. F., and J. J. N. Wilson, A global black carbon aerosol model, J. Geophys. Res., 101, 1939519409, 1996. Ct, J., and A. Staniforth, A two-time-level semi-Lagrangian semi-implicit scheme for spectral oe models, Mon. Wea. Rev., 116, 20032012, 1988. Cotton, W. R., G. J. Tripoli, R. M. Rauber, and E. A. Mulvihill, Numerical simulation of the eects of varying ice crystal nucleation rates and aggregation processes on orographic snowfall, J. Appl. Meteor., 25, 16581680, 1986. Dai, Y., X. Zeng, R. E. Dickinson, and coauthors, The Common Land Model: Documentation and Users Guide, http://climate.eas.gatech.edu/dai/clmdoc.pdf, 2001. Daley, R., C. Girard, J. Henderson, and I. Simmonds, Short-term forecasting with a multi-level spectral primitive equation model. Part Imodel formulation, Atmosphere, 14, 98116, 1976. DeCaria, A. J., K. E. Pickering, G. L. Stenchikov, and L. E. Ott, Lightning-generated nox and its impact on tropospheric ozone production: A three-dimensional modeling study of a stratosphere-troposphere experiment: Radiation, aerosols and ozone (sterao-a) thunderstorm, J. Geophys. Res., 110, D14303, 2006. Dennis, J., A. Fournier, W. F. Spotz, A. St.-Cyr, M. A. Taylor, S. J. Thomas, and H. Tufo, High resolution mesh convergence properties and parallel eciency of a spectral element atmospheric dynamical core, Int. J. High Perf. Comput. Appl., 19, 225235, 2005. Dentener, F., S. Kinne, T. Bond, O. Boucher, J. Cofala, S. Generoso, P. Ginoux, S. Gong, J. J. Hoelzemann, A. Ito, L. Marelli, J. E. Penner, J. P. Putaud, C. Textor, M. Schulz, G. R. van derWerf, and J. Wilson, Emissions of primary aerosol and precursor gases in the years 2000 and 1750 prescribed data-sets for aerocom, Atmospheric Chemistry and Physics, 6, 43214344, 2006a, 087YG Times Cited:81 Cited References Count:52. Dentener, F., D. Stevenson, K. Ellingsen et al., Global atmospheric environment for the next generation, Environmental Science and Technology, 40, 35863594, 2006b. Deville, M. O., P. F. Fischer, and E. H. Mund, High Order Methods for Incompressible Fluid Flow. Cambridge University Press, 1 edition, 8, 2002. Dickinson, R. E., A. Henderson-Sellers, P. J. Kennedy, and M. F. Wilson, Biosphere-atmosphere transfer scheme (BATS) for the NCAR Community Climate Model, Technical Report NCAR/TN-275+STR, National Center for Atmospheric Research, Boulder, Colorado, 69 pp., 1987. 247

4459 4460 4461

4462 4463

4464 4465

4466 4467 4468

4469 4470

4471 4472

4473 4474 4475 4476

4477 4478 4479

4480 4481 4482 4483 4484

4485 4486

4487 4488

4489 4490 4491 4492

4493 4494

D tsch, H. U., Vertical ozone distribution on a global scale, Pure Appl. Geophys., 116, 511529, u 1986. Dye, J. E., D.Baumgardner, B. W. Gandrun, S. R. Kawa, K. K. Kelly, M. Loewenstein, G. V. Ferry, K. R. Chan, and B. L. Gary, Particle size distributions in arctic polar stratospheric clouds, growth and freezing of sulfuring acid droplets, and implications for cloud formation, J. Geophys. Res., 97, 80158034, 1992. Easter, R. C., S. J. Ghan, Y. Zhang, R. D. Saylor, E. G. Chapman, N. S. Laulainen, H. AbdulRazzak, L. R. Leung, X. D. Bian, and R. A. Zaveri, Mirage: Model description and evaluation of aerosols and trace gases, Journal of Geophysical Research-Atmospheres, 109 (D20), , 2004, 867ZR Times Cited:37 Cited References Count:280. Eaton, B. E., Users Guide to the Community Atmosphere Model AM4.0, Technical report, National Center for Atmospheric Research, Boulder, Colorado, http://www.ccsm.ucar.edu/models/ccsm4.0/cam/docs/users_guide/book1.html, 2010. Emmons, e. a., L. K., Description and evaluation of the model for ozone and related chemical tracers, version 4 (mozart-4), Geosci. Model Dev, 3, 4367, 2010. Ferrier, B., A double-moment multiple-phase four-class bulk ice scheme. part I: Description, J. Atmos. Sci., 51, 249280, 1994. Fiedler, F., and H. A. Panofsky, The geostrophic drag coecient and the eective roughness length, Quart. J. Roy. Meteor. Soc., 98, 213220, 1972. Field, P. R., O. Mhler, P. Connolly, M. Krmer, R. Cotton, A. J. Heymseld, H. Saatho, o and M. Schnaiter, Some ice nucleation characteristics of Asian and Saharan desert dust, Atmospheric Chemistry and Physics, 6 (10), 29913006, 2006. Field, P. R., A. J. Heymseld, and A. Bansemer, Shattering and particle inter-arrival times measured by optical array probes in ice clouds, J. Atmos. Oceanic Technol., 23, 13571371, 2007. Fomichev, V. I., J. P. Blanchet, and D. S. Turner, Matrix parameterization of the 15 m co2 band cooling in the middle and upper atmosphere for variable co2 concentration, J. Geophys. Res., 103, 1150511528, 1998. Fomichev, V. I., V. P. Ogibalov, and S. R. Beagley, Solar heating by the near-IR co2 bands in the mesosphere, Geophys. Res. Lett., 31, L21102, 2004. Fournier, A., M. A. Taylor, and J. Tribbia, The spectral element atmosphere model (SEAM): High-resolution parallel computation and localized resolution of regional dynamics, Mon. Wea. Rev., 132, 726748, 2004. Friedl, R., editor, Atmospheric eects of subsonic aircraft: Interim assessment report of the advanced subsonic technology program. 1997.

4495 4496 4497 4498

4499 4500 4501 4502

4503 4504 4505

4506 4507

4508 4509

4510 4511

4512 4513 4514

4515 4516 4517

4518 4519 4520

4521 4522

4523 4524 4525

4526 4527

248

4528 4529

Fritts, D. C., and T. J. Dunkerton, Fluxes of heat and constituents due to convectively unstable gravity waves, jas, 42, 549556, 1985. Fu, Q., P. Yang, and W. B. Sun, An accurate parameterization of the infrared radiative properties of cirrus clouds for climate models, J. Climate, 11, 22232237, 1998. Fuch, N. A., The Mechanics of Aerosols. Pergamon, Oxford, 1964. Galperin, B., L. H. Kanta, S. Hassid, and A. Rosati, A quasi-equilibrium turbulent energy model for geophysical ows, J. Atmos. Sci., 45, 5562, 1988. Garcia, R. R., Parameterization of planetary wave breaking in the middle atmosphere, jas, 2001. Garcia, R. R., D. R. Marsh, D. E. Kinnison, B. A. Boville, and F. Sassi, Simulation of secular trends in the middle atmosphere, 1950-2003, J. Geophys. Res., 112, doi:10.1029/2006JD007485, 2007. Garcia, R. R., and S. Solomon, The eect of breaking gravity waves on the dynamical and chemical composition of the mesosphere and lower thermosphere, J. Geophys. Res., 90, 3850 3868, 1985. Garcia, R. R., and S. Solomon, A new numerical model of the middle atmosphere. part ii: Ozone and related species, J. Geophys. Res., 99, 1293712951, 1994. Gent, P. R., S. Yeager, R. B. Neale, S. Levis, and D. Bailey, Improvements in a half degree atmopshere/land version of the CCSM., Climate Dynamics, 79, 2558, 2009. Gettelman, A., H. Morrison, and S. J. Ghan, A new two-moment bulk stratiform cloud microphysics scheme in the NCAR Community Atmosphere Model (CAM3), Part II: Single-column and global results, J. Clim., 21 (15), 36603679, 2008. Gettelman, A., A. X. Liu, S. J. Ghan, H. Morrison, S. Park, A. J. Conley, S. A. Klein, J. Boyle, D. L. Mitchell, and J. L. F. Li, Global simulations of ice nucleation and ice supersaturation with an improved cloud scheme in the community atmosphere model, J. Geophys. Res., 108, ????????, 2010a. Gettelman, A. et al., Multi-model assessment of the upper troposphere and lower stratosphere: Tropics and trends, in press J. Geophys. Res., 2010b. Ghan, S. J., and R. C. Easter, Computationally ecient approximations to stratiform cloud microphysics parameterization, Mon. Weather Rev., 120 (1572-1582), 1992. Ghan, S. J., and R. C. Easter, Impact of cloud-borne aerosol representation on aerosol direct and indirect eects, Atmospheric Chemistry and Physics, 6, 41634174, 2006, 086FO Times Cited:5 Cited References Count:47. Ghan, S. J., L. R. Leung, R. C. EasterEaster, and K. AbdulRazzak, Prediction of cloud droplet number in a general circulation model, J. Geophys Res-Atmos, 102, 2177721794, 1997.

4530 4531

4532

4533 4534

4535

4536 4537 4538

4539 4540 4541

4542 4543

4544 4545

4546 4547 4548

4549 4550 4551 4552

4553 4554

4555 4556

4557 4558 4559

4560 4561

249

4562 4563

Ghan, S. J., and R. A. Zaveri, Parameterization of optical properties for hydrated internallymixed aerosol, J. Geophys. Res., 112, DOI 10.1029/2006JD007927, 2007. Giorgi, F., and W. Chameides, The rainout parameterization in a photochemical model, J. Geophys. Res., 90, 78727880, 1985. Giraldo, F. X., Trajectory calculations for spherical geodesic grids in cartesian space, Mon. Wea. Rev., 127, 16511662, 1999. Gregory, D., R. Kershaw, and P. M. Inness, Parameterization of momentum transport by convection. ii: Tests in single column and general circulation models, Q. J. R. Meteorol. Soc., 123, 11531184, 1997a. Gregory, D., R. Kershaw, and P. M. Inness, Parametrization of momentum transport by convection. II: Tests in single-column and general circulation models, Q. J. R. Meteorol. Soc., 123, 11531183, 1997b. Grenier, H., and C. S. Bretherton, A moist pbl parameterization for large-scale models and its application to subtropical cloud-topped marine boundary layer, Mon. Wea. Rev., 129, 357377, 2001. Hack, J. J., Parameterization of moist convection in the National Center for Atmospheric Research Community Climate Model (CCM2), J. Geophys. Res., 99, 55515568, 1994a. Hack, J. J., Parameterization of moist convection in the national center for atmospheric research community climate model (ccm2), J. Geophys. Res., 99, 55515568, 1994b. Hack, J. J., L. M. Bath, G. W. Williamson, and B. A. Boville, Modications and enhancements to the NCAR Community Climate Model (CCM1), Technical Report NCAR/TN-336+STR, National Center for Atmospheric Research, Boulder, Colorado, 97 pp., 1989. Hack, J. J., B. A. Boville, B. P. Briegleb, J. T. Kiehl, P. J. Rasch, and D. L. Williamson, Description of the NCAR Community Climate Model (CCM2), Technical Report NCAR/TN382+STR, National Center for Atmospheric Research, 120 pp., 1993. Hedin, A. E., Extension of the MSIS thermosphere model into the middle and lower atmosphere, J. Geophys. Res., 96, 1159, 1991. Hedin, A., MSIS-86 thermospheric model, J. Geophys. Res., 92, 4649, 1987. Heinbockel, J. H., Introduction to Tensor Calculus and Continuum Mechanics. Traord Publishing, Victoria, B.C., 12, 2001. Held, I. M., and M. J. Suarez, A proposal for the intercomparison of the dynamical cores of atmospheric general circulation models, Bull. Am. Meteorol. Soc., 75, 18251830, 1994. Hildebrand, F. B., Introduction to Numerical Analysis. McGraw-Hill, New York, New York, 1956, 511 pp.

4564 4565

4566 4567

4568 4569 4570

4571 4572 4573

4574 4575 4576

4577 4578

4579 4580

4581 4582 4583

4584 4585 4586

4587 4588

4589

4590 4591

4592 4593

4594 4595

250

4596 4597

Holstlag, A. M., and B. A. Boville, Local versus nonlocal boundary layer diusion in a global climate model, J. Climate, 6, 18251842, 1993. Holton, J. R., The role of gravity wave induced drag and diusion in the momentum budget of the mesosphere, J. Atmos. Sci., 39, 791799, 1982. Holtslag, A. A. M., and B. A. Boville, Local versus nonlocal boundary-layer diusion in a global climate model, J. Climate, 6, 18251842, 1993a. Holtslag, A. A. M., and B. A. Boville, Local versus nonlocal boundary layer diusion in a global model, J. Climate, 6, 18251842, 1993b. Holtslag, A. A. M., E. I. F. deBruijn, and H.-L. Pan, A high resolution air mass transformation model for short-range weather forecasting, Mon. Wea. Rev., 118, 15611575, 1990. Hoose, C., and J. E. Kristjansson, A classical-theory-based parameterization of heterogeneous ice nucleation by mineral dust, soot and biological particles in a global climate model, in press, J. Atmos. Sci., 2010. Horowitz, e. a., L., A global simulation of tropospheric ozone and related tracers: Description and evaluation of mozart, version 2, Journal of Geophysical Research-Atmospheres, 108, 4784,, 2003. Horowitz, L., S. Walters, D. Mauzerall, L. Emmons, P. Rasch, C. Granier, X. Tie, J.-F. Lamarque, M. Schultz, G. Tyndall, J. Orlando, and G. Brasseur, A global simulation of tropospheric ozone and related tracers: Description and evaluation of mozart, version 2, J. Geophys. Res., 108, 4784, 2003. Hortal, M. 1999. Aspects of the numerics of the ECMWF model. in Proceedings of ECMWF Seminar: Recent developments in numerical methods for atmospheric modelling, 711 September 1998, 127143, . Hoskins, B. J., The mathematical theory of frontogenesis, Ann. Rev. Fluid Mech., 14, 131151, 1982. Hsu, Y.-J. G., and A. Arakawa, Numerical modeling of the atmosphere with an isentropic vertical coordinate, Mon. Wea. Rev., 118, 19331959, 1990. Iacono, M., E. Mlawer, S. Clough, and J.-J. Morcrette, Impact of an improved longwave radiation model, rrtm, on the energy budget and thermodynamic properties of the ncar community climate model, ccm3, J. Geophys. Res., 105, 1487314890, 2000. Iacono, M., J. Delamere, E. Mlawer, and S. Clough, Evaluation of upper tropospheric water vapor in the ncar community climate model (ccm3) using modeled and observed hirs radiances, J. Geophys. Res., 108(D2), 4037, 2003. Iacono, M., J. Delamere, E. Mlawer, M. Shephard, S. Clough, and W. Collins, Radiative forcing by long-lived greenhouse gases: Calculations with the aer radiative transfer models, J. Geophys. Res., 2008. 251

4598 4599

4600 4601

4602 4603

4604 4605

4606 4607 4608

4609 4610 4611

4612 4613 4614 4615

4616 4617 4618

4619 4620

4621 4622

4623 4624 4625

4626 4627 4628

4629 4630 4631

4632 4633

Ikawa, M., and K. Saito, description of the nonhydrostatic model developed at the forecast research department of the mri, Technical Report 28, Meteorological Institute MRI, 1990. Jablonowski, C., and D. L. Williamson, A baroclinic instability test case for atmospheric model dynamical cores, Q. J. R. Meteorol. Soc., 132, 29432975, 2006. Junker, C., and C. Liousse, A global emission inventory of carbonaceous aerosol from historic records of fossil fuel and biofuel consumption for the period 1860-1997, Atmospheric Chemistry and Physics, 8 (5), 11951207, 2008, 273CS Times Cited:8 Cited References Count:52. Kader, B. A., and A. M. Yaglom, Mean elds and uctuation moments in unstably stratied turbulent boundary layers, J. Fluid Mech., 212, 637662, 1990. Kain, J. S., and J. M. Fritsch, A one-dimensional entraining/detraining plume model and its application in convective parameterization, J. Atmos. Sci., 47, 27842802, 1990. Krcher, B., O. Mhler, P. J. DeMott, S. Pechtl, and F. Yu, Insights into the role of soot aerosols a o in cirrus cloud formation, Atmos. Chem. Phys., 7, 42034227, 2007. Karniadakis, G. E., and S. J. Sherwin, Spectral/hp Element Methods for Computational Fluid Dynamics (Numerical Mathematics and Scientic Computation). Oxford University Press, USA, 2 edition, 8, 2005. Kasahara, A., Various vertical coordinate systems used for numerical weather prediction, Mon. Wea. Rev., 102, 509522, 1974. Kasten, F., Falling speed of aerosol particles, J. Appl. Meteorol., 7, 944947, 1968. Kerminen, V. M., and M. Kulmala, Analytical formulae connecting the real and the apparent nucleation rate and the nuclei number concentration for atmospheric nucleation events, Journal of Aerosol Science, 33 (4), 609622, 2002, 535VC Times Cited:67 Cited References Count:26. Kershaw, R., and D. Gregory, Parametrization of momentum transport by convection. I: Theory and cloud modelling results, Q. J. R. Meteorol. Soc., 123, 11331151, 1997. Khairoutdinov, M. F., and Y. Kogan, A new cloud physics parameterization in a large-eddy simulation model of marine stratocumulus, Mon. Weather Rev., 128, 229243, 2000. Kiehl, J. T., and B. P. Briegleb, A new parameterization of the absorbance due to the 15 micron band system of carbon dioxide, J. Geophys. Res., , 90139019, 1991. Kiehl, J. T., J. J. Hack, and B. P. Briegleb, The simulated Earth radiation budget of the National Center for Atmospheric Research Community Climate Model CCM2 and comparisons with the Earth Radiation Budget Experiment (ERBE), J. Geophys. Res., 99, 2081520827, 1994. Kiehl, J. T., J. Hack, G. Bonan, B. Boville, B. Briegleb, D. Williamson, and P. Rasch, Description of the NCAR Community Climate Model (CCM3), Technical Report NCAR/TN420+STR, National Center for Atmospheric Research, Boulder, Colorado, 152 pp., 1996.

4634 4635

4636 4637 4638

4639 4640

4641 4642

4643 4644

4645 4646 4647

4648 4649

4650

4651 4652 4653 4654

4655 4656

4657 4658

4659 4660

4661 4662 4663

4664 4665 4666

252

4667 4668 4669

Kiehl, J. T., J. J. Hack, G. B. Bonan, B. B. Boville, D. L. Williamson, and P. J. Rasch, The National Center for Atmospheric Research Community Climate Model: CCM3, J. Climate, 11, 11311149, 1998. Kinnison, D. E., G. P. Brasseur, S. Walters, R. R. Garcia, F. Sassi, B. A. Boville, D. Marsh, L. Harvey, C. Randall, W. Randel, J. F. Lamarque, L. K. Emmons, P. Hess, J. Orlando, J. Tyndall, and L. Pan, Sensitivity of chemical tracers to meteorological parameters in the mozart-3 chemical transport model, J. Geophys. Res., in press, 2006. Kockarts, G., Nitric oxide cooling in the terrestrial thermosphere, Geophys. Res. Lett., 7, 137 140, 1980. Koehler, K. A., S. M. Kreidenweis, P. J. DeMott, M. D. Petters, A. J. Prenni, and C. M. Carrico, Hygroscopicity and cloud droplet activation of mineral dust aerosol, Geophysical Research Letters, 36, , 2009a, 435FB Times Cited:4 Cited References Count:25. Koehler, K. A., S. M. Kreidenweis, P. J. DeMott, M. D. Petters, A. J. Prenni, and C. M. Carrico, Hygroscopicity and cloud droplet activation of mineral dust aerosol., Geophys. Res. Lett., 36, 2009b. Koepke, M. H. P., and I. Schult, Optical properties of aerosols and clouds: The software package opac, Bull. Am. Meteorol. Soc., 79, 831844, 1998. Koppers, G. A. A., and D. P. Murtagh, Model studies of the inuence of O2 photodissociation parameterizations in the schumann-runge bands on ozone related photolysis in the upper atmosphere, Ann. Geophys., 14, 6879, 1996. Kroll, J. H., N. L. Ng, S. M. Murphy, R. C. Flagan, and J. H. Seinfeld, Secondary organic aerosol formation from isoprene photooxidation, Environmental Science and Technology, 40 (6), 18691877, 2006, 024RN Times Cited:121 Cited References Count:62. Lamarque, J. F., T. C. Bond, V. Eyring, C. Granier, A. Heil, Z. Klimont, D. Lee, C. Liousse, A. Mieville, B. Owen, M. G. Schultz, D. Shindell, S. J. Smith, E. Stehfest, J. vanAardenne, O. R. Cooper, M. Kainuma, N. Mahowald, J. R. McConnell, V. Naik, K. Riahi, and D. P. vanVuuren, Historical (18502000) gridded anthropogenic and biomass burning emissions of reactive gases and aerosols: methodology and application, Atmospheric Chemistry and Physics Discussion, 10, 49635019, 2010a. Lamarque, J., T. C. Bond, V. Eyring, C. Granier, A. Heil, Z. Klimont, D. Lee, C. Liousse, A. Mieville, B. Owen, M. G. Schultz, D. Shindell, S. J. Smith, E. Stehfest, J. van Aardenne, O. R. Cooper, M. Kainuma, N. Mahowald, J. R. McConnell, V. Naik, K. Riahi, and D. P. van Vuuren, Historical (1850-2000) gridded anthropogenic and biomass burning emissions of reactive gases and aerosols: methodology and application, Atmospheric Chemistry & Physics Discussions, 10, 49635019, Feb., 2010b. Large, W. G., J. C. McWilliams, and S. C. Doney, Oceanic vertical mixing: A review and a model with a nonlocal boundary layer parameterization, Rev. Geophys., 32, 363403, 1994.

4670 4671 4672 4673

4674 4675

4676 4677 4678

4679 4680 4681

4682 4683

4684 4685 4686

4687 4688 4689

4690 4691 4692 4693 4694 4695

4696 4697 4698 4699 4700 4701

4702 4703

253

4704 4705

Large, W. G., and S. Pond, Sensible and latent heat ux measurements over the ocean, J. Phys. Oceanogr., 12, 464482, 1982. Lauritzen, P., C. Jablonowski, M. Taylor, and R. Nair, Rotated versions of the jablonowski steady-state and baroclinic wave test cases: A dynamical core intercomparison, Journal of Advances in Modeling Earth Systems, In Press, 2010. Lauritzen, P. H., A stability analysis of nite-volume advection schemes permitting long time steps, Mon. Wea. Rev., 135, 26582673, 2007. Lauritzen, P. H., P. A. Ullrich, and R. D. Nair, Atmospheric transport schemes: Desirable properties and a semi-lagrangian view on nite-volume discretizations, in: P.H. Lauritzen, R.D. nair, C. Jablonowski, M. Taylor (Eds.), Numerical techniques for global atmospheric models, Lecture Notes in Computational Science and Engineering, Springer, 2010, to appear. Lawson, R. P., B. Barker, B. Pilson, and Q. Mo, In situ observations of the microphysical properties of wave, cirrus and anvil clouds. Part 2: Cirrus cloud, J. Atmos. Sci., 63, 3186 3203, 2006. Le Texier, H., S. Solomon, and R. R. Garcia, Role of molecular hydrogen and methane oxidation in the water vapor budget of the stratosphere, Q. J. R. Meteorol. Soc., 114, 281295, 1988. Lean, J., Evolution of the suns spectral irradiance since the maunder minimum, Geophys. Res. Lett., 27 (16), 24252428, 2000. Lim, Y. B., and P. J. Ziemann, Products and mechanism of secondary organic aerosol formation from reactions of n-alkanes with oh radicals in the presence of nox, Environmental Science and Technology, 39 (23), 92299236, 2005, 989ZK Times Cited:53 Cited References Count:48. Lin, S.-J., W. C. Chao, Y. C. Sud, and G. K. Walker, A class of the van leer-type transport schemes and its applications to the moisture transport in a general circulation model, Mon. Wea. Rev., 122, 15751593, 1994. Lin, S.-J., and R. B. Rood, Multidimensional ux form semi-lagrangian transport schemes, Mon. Wea. Rev., 124, 20462070, 1996. Lin, S.-J., and R. B. Rood, An explicit ux-form semi-lagrangian shallow water model on the sphere, Q. J. R. Meteorol. Soc., 123, 25312533, 1997. Lin, S.-J., A vertically lagrangian nite-volume dynamical core for global models, Mon. Wea. Rev., 132, 22932397, 2004. Lin, Y.-L., R. R. Farley, and H. D. Orville, Bulk parameterization of the snow eld in a cloud model, J. Clim. Appl. Meteorol., 22, 10651092, 1983. Lindzen, R. S., Turbulence and stress due to gravity wave and tidal breakdown, J. Geophys. Res., 86, 97019714, 1981. Liu, X., J. E. Penner, S. J. Ghan, and M. Wang, Inclusion of ice microphysics in the NCAR Community Atmosphere Model version 3 (CAM3), J. Clim., 20, 45264547, 2007. 254

4706 4707 4708

4709 4710

4711 4712 4713 4714

4715 4716 4717

4718 4719

4720 4721

4722 4723 4724

4725 4726 4727

4728 4729

4730 4731

4732 4733

4734 4735

4736 4737

4738 4739

4740 4741

Liu, X., and S. Ghan, A modal aerosol model implementation in the Community Atmosphere Model, version 5 (CAM5), J. Atmos. Sci., 2010. Liu, X., and J. E. Penner, Ice nucleation parameterization for global models, Meteor. Z., 14 (499-514), 2005. Liu, X. H., J. E. Penner, and M. Herzog, Global modeling of aerosol dynamics: Model description, evaluation, and interactions between sulfate and nonsulfate aerosols, Journal of Geophysical Research-Atmospheres, 110 (D18), , 2005, 968RL Times Cited:39 Cited References Count:207. Lohmann, U., J. Feichter, C. C. Chuang, and J. Penner, Prediction of the number of cloud droplets in the echam gcm, J. Geophys. Res., 104 (D8), 91699198, 1999. Lord, S. J., W. C. Chao, and A. Arakawa, Interaction of a cumulus cloud ensemble with the large-scale environment. part iv: The discrete model, J. Atmos. Sci., 39, 104113, 1982. Machenhauer, B., The spectral method, in Numerical Methods Used in Atmospheric Models, World Meteorological Organization, Geneva, Switzerland, 1979. Machenhauer, B. 1998. MPI workshop on conservative transport schemes. in Report No. 265, . Max-planck-Institute for Meteorology. Machenhauer, B., E. Kaas, and P. H. Lauritzen, Finite volume methods in meteorology, in: R. Temam, J. Tribbia, P. Ciarlet (Eds.), Computational methods for the atmosphere and the oceans, Handbook of Numerical Analysis, 14, 2009, Elsevier, 2009, pp.3-120. Maday, Y., and A. T. Patera, Spectral element methods for the incompressible Navier Stokes equations, in State of the Art Surveys on Computational Mechanics, edited by A. K. Noor, and J. T. Oden, 71143, ASME, New York, 1987. Mahowald, N. M., D. R. Muhs, S. Levis, P. J. Rasch, M. Yoshioka, C. S. Zender, and C. Luo, Change in atmospheric mineral aerosols in response to climate: Last glacial period, preindustrial, modern, and doubled carbon dioxide climates, Journal of Geophysical ResearchAtmospheres, 111 (D10), , 2006a, 050EP Times Cited:68 Cited References Count:123. Mahowald, N. M., M. Yoshioka, W. D. Collins, A. J. Conley, D. W. Fillmore, and D. B. Coleman, Climate response and radiative forcing from mineral aerosols during the last glacial maximum, pre-industrial, current and doubled-carbon dioxide climates, Geophysical Research Letters, 33 (20), , 2006b, 100WD Times Cited:16 Cited References Count:21. Manabe, S., Climate and the ocean circulation: 1. The atmospheric circulation and the hydrology of the earths surface, Mon. Wea. Rev., 97, 739774, 1969. Marsh, D., S. Solomon, and A. Reynolds, Empirical model of nitric oxide in the lower thermosphere, J. Geophys. Res., 109, doi:10.1029/2003JA010199, 2004.

4742 4743

4744 4745 4746 4747

4748 4749

4750 4751

4752 4753

4754 4755

4756 4757 4758

4759 4760 4761

4762 4763 4764 4765

4766 4767 4768 4769

4770 4771

4772 4773

255

4774 4775 4776 4777

Martensson, E. M., E. D. Nilsson, G. deLeeuw, L. H. Cohen, and H. C. Hansson, Laboratory simulations and parameterization of the primary marine aerosol production, Journal of Geophysical Research-Atmospheres, 108 (D9), , 2003, 682QU Times Cited:51 Cited References Count:32. Marti, J., and K. Mauersberger, A survey and new measurements of ice vapor pressure at temperatures between 170 and 250 K, Geophys. Res. Lett., 20, 363366, 1993. Martin, G. M., D. W. Johnson, and A. Spice, The measurement and parameterization of eective radius of droplets in warm stratocumulus clouds, J. Atmos. Sci., 51, 18231842, 1994. Matthes, K., D. R. Marsh, R. R. Garcia, D. E. Kinnison, F. Sassi, and S. Walters, The role of the qbo in modulating the inuence of the 11-year solar cycle on the atmosphere using constant forcings, J. Geophys. Res., ??, doi:10.1029/2009JD013020, 2010. McAvaney, B. J., W. Bourke, and K. Puri, A global spectral model for simulation of the general circulation, J. Atmos. Sci., 35, 15571583, 1978. McFarlane, N. A., The eect of orographically excited gravity wave drag on the general circulation of the lower stratosphere and troposphere, J. Atmos. Sci., 44, 17751800, 1987. McFarlane, R. P. S. A., and S. A. Klein, Albedo bias and the horizontal variability of clouds in subtropical marine boundary layers: Observations from ships and satellite, J. Geophys. Res., 104, 61386191, 1999. McIntyre, M. E., On dynamics and transport near the polar mesopause in summer, J. Geophys. Res., 94, 14,61714,628, 1989. McLinden, C., S. Olsen, B. Hannegan, O. Wild, and M. Prather, Stratospheric ozone in 3-d models: A simple chemistry and the cross-tropopause ux, J. Geophys. Res., 105, 14,653 14,665, 2000. Merikanto, J., I. Napari, H. Vehkamaki, T. Anttila, and M. Kulmala, New parameterization of sulfuric acid-ammonia-water ternary nucleation rates at tropospheric conditions, Journal of Geophysical Research-Atmospheres, 112 (D15), , 2007, 202AA Times Cited:16 Cited References Count:39. Meyers, M. P., P. J. DeMott, and W. R. Cotton, New primary ice-nucleation parameterizations in an explicit cloud model, J. Applied Met., 31, 708721, 1992. Minschwaner, K., and D. E. Siskind, A new calculation of nitric oxide photolysis in the stratosphere, mesosphere, and lower thermosphere, J. Geophys. Res., 98, 2040120412, 1993. Mitchell, D. L., Parameterization of the Mie extinction and absorption coecients for water clouds, J. Atmos. Sci., 57, 13111326, 2000. Mitchell, D. L., Eective diameter in radiation transfer: General denition, applications and limitations, J. Atmos. Sci., 59, 23302346, 2002.

4778 4779

4780 4781

4782 4783 4784

4785 4786

4787 4788

4789 4790 4791

4792 4793

4794 4795 4796

4797 4798 4799 4800

4801 4802

4803 4804

4805 4806

4807 4808

256

4809 4810 4811

Mitchell, D. L., and W. P. Arnott, A model predicting the evolution of ice particle size spectra and radiative properties of cirrus clouds. part ii: Dependence of absorption and extinction on ice crystal morphology, J. Atmos. Sci., , 817832, 1994. Mitchell, D. L., A. Macke, and Y. Liu, Modeling cirrus clouds. Part II: Treatment of radiative properties, J. Atmos. Sci., 53, 26972988, 1996a. Mitchell, D. L., A. Macke, and Y. Liu, Modeling cirrus clouds. part ii: Treatment of radiative properties, J. Atmos. Sci., 53, 29672988, 1996b. Mitchell, D. L., W. P. Arnott, C. Schmitt, A. J.Baran, S. Havemann, and Q. Fu, Contributions of photon tunneling to extinction in laboratory grown hexagonal columns, J. Quant. Spectrosc. Radiat. Transfer, 70, 761776, 2001. Mitchell, D. L., A. J. Baran, W. P. Arnott, and C. Schmitt, Testing and comparing the modied anomalous diraction approximation, J. Atmos. Sci., 59, 23302346, 2006a. Mitchell, D. L., R. P. dEntremont, and R. P. Lawson, Passive thermal retrievals of ice and liquid water path, eective size and optical depth and their dependence on particle and size distribution shape, 12th AMS Conference on Atmospheric Radiation, July, 2006b. Mlawer, E., S. Taubman, P. Brown, M. Iacono, and S. Clough, Radiative transfer for inhomogeneous atmospheres: Rrtm, a validated correlated-k model for the longwave, J. Geophys. Res., 102, 1666316682, 1997. Mlynczak, M. G., and S. Solomon, A detailed evaluation of the heating eciency in the middle atmosphere, J. Geophys. Res., 98, 10,51710,541, 1993. Monahan, E. C., D. E. Spiel, and K. L. Davidson, A model of marine aerosol generation via whitecaps and wave disruption, in Oceanic Whitecaps, edited by E. C. Monahan, G. M. Niocaill,, and D. Reidel, Norwell, Mass., 1986. Moncet, J.-L., and S. Clough, Accelerated monochromatic radiative transfer for scattering atmospheres: Application of a new model to spectral radiance observations, J. Geophys. Res., 102 (21), 853866, 1997. Morrison, H., J. A. Curry, and V. I. Khvorostyanov, A new double-moment microphysics parameterization for application in cloud and climate models. part i: Description, J. Atmos. Sci., 62, 16651677, 2005. Morrison, H., and A. Gettelman, A new two-moment bulk stratiform cloud microphysics scheme in the NCAR Community Atmosphere Model (CAM3), Part I: Description and numerical tests, J. Clim., 21 (15), 36423659, 2008. Morrison, H., and J. O. Pinto, Mesoscale modeling of springtime arctic mixed-phase stratiform clouds using a new two-moment bulk microphysics scheme, J. Atmos. Sci., 62, 36833704, 2005.

4812 4813

4814 4815

4816 4817 4818

4819 4820

4821 4822 4823

4824 4825 4826

4827 4828

4829 4830 4831

4832 4833 4834

4835 4836 4837

4838 4839 4840

4841 4842 4843

257

4844 4845

Mote, P. W., J. R. Holton, J. M. Russell, and B. A. Boville, A comparison of observed (haloe) and modeled (ccm2) methane and stratospheric water vapor, grl, 20, 14191422, 1993. Neale, R. B., J. H. Richter, and M. Jochum, The impact of convection on ENSO: From a delayed oscillator to a series of events, J. Climate, 21, 59045924, 2008. Neale, R. B., and B. J. Hoskins, A standard test case for AGCMs including their physical parametrizations: I: The proposal, Atmos. Sci. Lett., 1, 101107, 2001a. Neale, R. B., and B. J. Hoskins, A standard test case for AGCMs including their physical parametrizations: II: Results for the Met Oce Model, Atmos. Sci. Lett., 1, 108114, 2001b. Newman, P. A., J. S. Daniel, D. W. Waugh, and E. R. Nash, A new formulation of equivalent eective stratospheric chlorine (EESC), Atmos. Chem. Phys., 7 (17), 45374552, 2007. Ng, N. L., P. S. Chhabra, A. W. H. Chan, J. D. Surratt, J. H. Kroll, A. J. Kwan, D. C. McCabe, P. O. Wennberg, A. Sorooshian, S. M. Murphy, N. F. Dalleska, R. C. Flagan, and J. H. Seinfeld, Eect of nox level on secondary organic aerosol (soa) formation from the photooxidation of terpenes, Atmospheric Chemistry and Physics, 7 (19), 51595174, 2007, 235MS Times Cited:42 Cited References Count:63. Nousiainen, T., and G. M. McFarquhar, Light scattering by quasi-spherical ice crystals, J. Atmos. Sci., 61, 22292248, 2004. Odum, J. R., T. P. W. Jungkamp, R. J. Grin, H. J. L. Forstner, R. C. Flagan, and J. H. Seinfeld, Aromatics, reformulated gasoline, and atmospheric organic aerosol formation, Environmental Science and Technology, 31 (7), 18901897, 1997, Xh483 Times Cited:168 Cited References Count:29. Ogibalov, V. P., and V. I. Fomichev, Parameterization of solar heating by the near IR co2 bands in the mesosphere, Adv. Space Res., 32, 759764, 2003. Okin, G. S., A new model of wind erosion in the presence of vegetation, Journal of Geophysical Research-Earth Surface, 113 (F2), , 2008, 276QS Times Cited:5 Cited References Count:41. Orszag, S. A., Fourier series on spheres, Mon. Wea. Rev., 102, 5675, 1974. Ovtchinnikov, M., and S. J. Ghan, Parallel simulations of aerosol inuence on clouds using cloud-resolving and single-column models, Journal of Geophysical Research-Atmospheres, 110 (D15), , 2005, 916EG Times Cited:9 Cited References Count:45. Park, S., and C. S. Bretherton, The university of washington shallow convection and moist turbulence schemes and their impact on climate simulations with the community atmosphere model, J. Climate, 22, 34493469, 2009. Park, S., C. S. Bretherton, and P. J. Rasch, The revised cloud macrophysics in the community atmosphere model, J. Climate, 000, 00000000, 2010. Peter, T., C. Bruhl, and P. J. Crutzen, Increase in the PSC formation probability caused by high-ying aircraft, J. Geophys. Res., 18, 14651468, 1991. 258

4846 4847

4848 4849

4850 4851

4852 4853

4854 4855 4856 4857 4858

4859 4860

4861 4862 4863 4864

4865 4866

4867 4868

4869

4870 4871 4872

4873 4874 4875

4876 4877

4878 4879

4880 4881

Petters, M., and S. Kreidenweis, A single parameter representation of hygroscopic growth and ccn activity, Atmos. Chem. Phys., 7, 19611971, 2007. Pickering, K., Y. Wang, W.-K. Tao, C. Price, and J. Mueller, Vertical distributions of lightning NOx for use in regional and global chemical transport models, J. Geophys. Res., 103, 31,203 31,216, 1998. Pincus, R. H. W. B., and J.-J. Morcrette, A fast, exible, approximation technique for computing radiative transfer in inhomogeneous cloud elds, J. Geophys. Res., 108(D13), 4376, 2003. Poschl, U., M. Canagaratna, J. T. Jayne, L. T. Molina, D. R. Worsnop, C. E. Kolb, and M. J. Molina, Mass accommodation coecient of h2so4 vapor on aqueous sulfuric acid surfaces and gaseous diusion coecient of h2so4 in n-2/h2o, Journal of Physical Chemistry A, 102 (49), 1008210089, 1998, 146ZX Times Cited:40 Cited References Count:41. Prather, M., Catastrophic loss of stratospheric ozone in dense volcanic clouds, J. Geophys. Res., 97, 18710,191, 1992. Prenni, A. J., J. Y. Harrington, M. Tjernstrom, P. J. DeMott, A. Avramov, C. N. Long, S. M. Kreidenweis, P. Q. Olsson, and J. Verlinde, Can ice-nucleating aerosols eect Arctic seasonal climate?, Bull. Am. Meteorol. Soc., 88, 541550, 2007. Price, C., J. Penner, and M. Prather, NOx from lightning. Part 1: Global distribution based on lightning physics, J. Geophys. Res., 102, 59295941, 1997a. Price, C., J. Penner, and M. Prather, NOx from lightning. Part 2: Constraints from the global atmospheric electric circuit, J. Geophys. Res., 102, 59435952, 1997b. Price, C., and D. Rind, A simple lightning parameterization for calculating global lightning distributions, J. Geophys. Res., 97, 99199933, 1992. Pruppacher, H. R., and J. D. Klett, Microphysics of Clouds and Precipitation. Kluwer Academic, 1997. Rani, M., R. Purser, and F. Mesinger, A global shallow-water model using an expanded cc spherical cube: Gnomonic versus conformal coordinates, Q. J. R. Meteorol. Soc., 122, 959 982, 1996. Rasch, P. J., B. A. Boville, and G. P. Brasseur, A three-dimensional general circulation model with coupled chemistry for the middle atmosphere, J. Geophys. Res., 100, 90419071, 1995. Rasch, P. J., M. C. Barth, J. T. Kiehl, S. E. Schwartz, and C. M. Benkovitz, A description of the global sulfur cycle and its controlling processes in the National Center for Atmospheric Research Community Climate Model, Version 3, J. Geophys. Res., 105, 13671385, 2000. Rasch, P. J., and J. E. Kristjnsson, A comparison of the CCM3 model climate using diagnosed a and predicted condensate parameterizations, J. Climate, 11, 15871614, 1998a. Rasch, P. J., and J. E. Kristjansson, A comparison of the ccm3 model climate using diagnosed and predicted condensate parameterizations, J. Climate, 11, 15871614, 1998b. 259

4882 4883 4884

4885 4886

4887 4888 4889 4890

4891 4892

4893 4894 4895

4896 4897

4898 4899

4900 4901

4902 4903

4904 4905 4906

4907 4908

4909 4910 4911

4912 4913

4914 4915

4916 4917

Rasch, P. J., and D. L. Williamson, On shape-preserving interpolation and semi-Lagrangian transport, SIAM J. Sci. Stat. Comput., 11, 656687, 1990. Raymond, D. J., and A. M. Blyth, A stochastic mixing model for non-precipitating cumulus clouds, J. Atmos. Sci., 43, 27082718, 1986. Raymond, D. J., and A. M. Blyth, Extension of the stochastic mixing model to cumulonimbus clouds, J. Atmos. Sci., 49, 19681983, 1992. Reisner, J., R. Rasmussen, and R. T. Bruintjes, Explicit forecasting of supercooled liquid water in winter storms using the mm5 forecast model, Q. J. R. Meteorol. Soc., 124, 10711107, 1998. R.G.Roble, Energetics of the mesosphere and thermosphere, in The upper mesosphere and lower thermosphere : a review of experiment and theory, edited by R. Johnson, and T. Killeen, Geophys. Monogr., 87, 121, 1995. Richmond, A., Ionospheric electrodynamics using magnetic apex coordinates, J. Geomag. Geoelectric., 47, 191, 1995. Richmond, A., M. Blanc, B. Emery, R. Wand, B. Fejer, R. Woodman, S. Ganguly, P. Amayenc, R. Behnke, C. Calderon, and J. Evans, An empirical model of quiet-day ionospheric electric elds at middle and low latitudes, J. Geophy. Res, 85, 4658, 1980. Richter, J. H., F. Sassi, and R. R. Garcia, Towards a physically-based wave source parameterization in a general circulation model, J. Atmos. Sci., 67, 136156, 2010. Richter, J. H., and P. J. Rasch, Eects of convective momentum transport on the atmospheric circulation in the community atmosphere model, version 3, J. Climate, 21, 14871499, 2008. Ridley, B., K. Pickering, and J. Dye, Comments on the parameterization of lightning-produced no in global chemistry-transport models, Atmos. Environ., 39, 61846187, 2005. Riemer, N., H. Vogel, B. Vogel, and F. Fiedler, Modeling aerosols on the mesoscale-gamma: Treatment of soot aerosol and its radiative eects, Journal of Geophysical ResearchAtmospheres, 108 (D19), , 2003, 732EM Times Cited:17 Cited References Count:90. Ringler, T. D., R. P. Heikes, , and D. A. Randall, Modeling the atmospheric general circulation using a spherical geodesic grid: A new class of dynamical cores, Mon. Wea. Rev., 128, 2471 2490, 2000. Ritchie, H., and M. Tanguay, A comparison of spatially averaged Eulerian and semi-Lagrangian treatments of mountains, Mon. Wea. Rev., 124, 167181, 1996. Robert, A. J., The integration of a low order spectral form of the primitive meteorological equations, J. Meteorol. Soc. Japan, 44, 237245, 1966. Roble, R., and E. Ridley, An auroral model for the ncar thermospheric general circulation model (TGCM), Ann. Geophys., 87, 369, 1987.

4918 4919

4920 4921

4922 4923 4924

4925 4926 4927

4928 4929

4930 4931 4932

4933 4934

4935 4936

4937 4938

4939 4940 4941

4942 4943 4944

4945 4946

4947 4948

4949 4950

260

4951 4952 4953

Roble, R., and E. Ridley, A thermosphere-ionosphere-mesosphere-electrodynamics general circulation model (TIME-GCM): Equinox solar cycle minimum simulations (30-500 km), Geophys. Res. Lett., 21, 417420, 1994. Rothman, L. S., A. Barbe, D. C. Benner, L. R. Brown, C. Camy-Peyret, M. R. Carleer, K. Chance, C. Clerbaux, V. Dana, V. M. Devi, A. Fayt, J.-M. Flaud, R. R. Gamache, A. Goldman, D. Jacquemart, K. W. Jucks, W. J. Laerty, J.-Y. Mandin, S. T. Massie, V. Nemtchinov, D. A. Newnham, A. Perrin, C. P. Rinsland, J. Schroeder, K. M. Smith, M. A. H. Smith, K. Tang, R. A. Toth, J. V. Auwera, P. Varanasi, and K. Yoshino, The HITRAN molecular spectroscopic database: Edition of 2000 including updates of 2001, J. Quant. Spectrosc. Radiat. Transfer, 82, 2003. Rothman, L., I. Gordon, A. Barbe, D. Benner, P. Bernath, M. Birk, V. Boudon, L. Brown, A. Campargue, J.-P. Champion, K. Chance, L. Coudert, V. Dana, V. Devi, S. Fally, J.-M. Flaud, R. Gamache, A. Goldman, D. Jacquemart, I. Kleiner, N. Lacome, W. Laerty, J.-Y. Mandin, S. Massie, S. Mikhailenko, C. Miller, N. Moazzen-Ahmadi, O. Naumenko, A. Nikitin, J. Orphal, V. Perevalov, A. Perrin, A. Predoi-Cross, C. Rinsland, M. Rotger, M. Simeckov, M. Smith, K. Sung, S. Tashkun, J. Tennyson, R. Toth, A. Vandaele, and J. V. Auwera, The hitran 2008 molecular spectroscopic database, Journal of Quantitative Spectroscopy and Radiative Transfer, 110 (9-10), 533 572, 2009, HITRAN. Rotstayn, L. D., B. F. Ryan, and J. J. Katzfey, A scheme for calculation of the liquid fraction in mixed-phase stratiform clouds in large-scale models, Mon. Weather Rev., 128 (4), 10701088, 2000. Sadourny, R., Conservative nite-dierence approximations of the primitive equations on quasiuniform spherical grids, Mon. Wea. Rev., 100 (2), 136144, 1972. Sander, S. P., et al., Chemical kinetics and photochemical data for use in atmospheric studies. Evaluation number 14, Technical Report Publication 02-25, Jet Propulsion Laboratory, Pasadena, CA, 2003. Sander, S. P., et al., Chemical kinetics and photochemical data for use in atmospheric studies. Evaluation number 15, Technical Report Publication 06-02, Jet Propulsion Laboratory, Pasadena, CA, 2006. Sanderson, M., W. Collins, R. Derwent, and C. Johnson, Simulation of global hydrogen levels using a lagrangian three-dimensional model, J. Atmos. Chem., 46 (1), 1528, SEP, 2003. Sandu, A., F. A. Potra, V. Damian-Iordache, and G. R. Carmichael, Ecient implementation of fully implicit methods for atmospheric chemistry, J. Comp. Phys., 129, 101110, 1996. Sandu, A., J. Verwer, M. vanLoon, G. Carmichael, F. Potra, D. Dabdub, and J. Seinfeld, Benchmarking sti ODE solvers for atmospheric chemistry problems I: Implicit versus explicit, Atmospheric Environment, 31, 31513166, 1997. Sangster, W. E., A meteorological coordinate system in which the Earths surface is a coordinate surface, Ph.D. thesis, University of Chicago, Department of Geophysical Sciences, 1960. 261

4954 4955 4956 4957 4958 4959 4960

4961 4962 4963 4964 4965 4966 4967 4968

4969 4970 4971

4972 4973

4974 4975 4976

4977 4978 4979

4980 4981

4982 4983

4984 4985 4986

4987 4988

4989 4990 4991

Sato, R. K., L. M. Bath, D. L. Williamson, and G. S. Williamson, Users guide to NCAR CCMOB, Technical Report NCAR/TN-211+IA, National Center for Atmospheric Research, Boulder, Colorado, 133 pp., 1983. Satoh, M., Atmospheric Circulation Dynamics and Circulation Models (Springer Praxis Books / Environmental Sciences). Springer, 1 edition, 6, 2004. Scherliess, L., A. Richmond, and B. Fejer. 2002. Global empirical ionospheric electric eld model. draft, Center for Atmospheric and Space Science, Utah State University, Logan. Schubert, W. H., J. S. Wakeeld, E. J. Steiner, and S. K. Cox, Marine stratocumulus convection. part i: Governing equations and horizontally homogeneous solutions, J. Atmos. Sci., 36, 1286 1307, 1979. Schunk, R., and A. Nagy, Ionospheres: Physics, Plasma Physics, and Chemistry. Cambridge, University Press, 2000. Seinfeld, J. H., and S. N. Pandis, Atmospheric Chemistry and Physics: From Air Pollution to Climate Change. John Wiley, Hoboken, N. J, 1998. Sihto, S. L., M. Kulmala, V. M. Kerminen, M. D. Maso, T. Petaja, I. Riipinen, H. Korhonen, F. Arnold, R. Janson, M. Boy, A. Laaksonen, and K. E. J. Lehtinen, Atmospheric sulphuric acid and aerosol formation: implications from atmospheric measurements for nucleation and early growth mechanisms, Atmospheric Chemistry and Physics, 6, 40794091, 2006. Simmons, A. J., and D. M. Burridge, An energy and angular momentum conserving vertical nite-dierence scheme and hybrid vertical coordinates, Mon. Wea. Rev., 109, 758766, 1981. Simmons, A. J., and R. Str ng, An energy and angular-momentum conserving nite-dierence u scheme, hybrid coordinates and medium-range weather prediction, Technical Report ECMWF Report No. 28, European Centre for MediumRange Weather Forecasts, Reading, U.K., 68 pp., 1981. Simmons, A. J., and R. Str ng, Numerical forecasts of stratospheric warming events using a u model with hybrid vertical coordinate, Q. J. R. Meteorol. Soc., 109, 81111, 1983. Skamarock, W., Kinetic energy spectra and model lters, in: P.H. Lauritzen, R.D. nair, C. Jablonowski, M. Taylor (Eds.), Numerical techniques for global atmospheric models, Lecture Notes in Computational Science and Engineering, Springer, 2010, 2010, to appear. Slingo, J. M., The development and verication of a cloud prediction scheme for the ECMWF model, Q. J. R. Meteorol. Soc., 113, 899927, 1987a. Slingo, J. M., The development and verication of a cloud prediction scheme for the ecmwf model, Q. J. R. Meteorol. Soc., 113, 899927, 1987b. Smith, R. N. B., A scheme for predicting layer clouds and their water content in a general circulation model, Q. J. R. Meteorol. Soc., 116, 435460, 1990.

4992 4993

4994 4995

4996 4997 4998

4999 5000

5001 5002

5003 5004 5005 5006

5007 5008

5009 5010 5011 5012

5013 5014

5015 5016 5017

5018 5019

5020 5021

5022 5023

262

5024 5025

Smith, S. J., H. Pitcher, and T. M. L. Wigley, Global and regional anthropogenic sulfur dioxide emissions, Glob. Biogeochem. Cycles, 29, 99119, 2001. Smith, S. J., R. Andres, E. Conception, and J. Lurz, Historical sulfur dioxide emissions 18502000: Methods and results, Technical report, Pacic Northwest National Laboratory, Joint Global Change Research Institute, 2004. Solomon, S., and L. Qiang, Solar extreme-ultraviolet irradiance for general circulation models, J. Geophys. Res., 110, A10306, doi:10.1029/2005JA011160, 2005. Spiteri, R. J., and S. J. Ruuth, A new class of optimal high-order strong-stability-preserving time discretization methods, SIAM J. Numer. Anal., 40 (2), 469491, 2002. Staniforth, A., N. Wood, and C. Girard, Energy and energy-like invariants for deep nonhydrostatic atmospheres, Q. J. R. Meteorol. Soc., 129, 34953499, 2003. Starr, V. P., A quasi-lagrangian system of hydrodynamical equations, J. Meteor., 2, 227237, 1945. Stier, P., J. Feichter, S. Kinne, S. Kloster, E. Vignati, J. Wilson, L. Ganzeveld, I. Tegen, M. Werner, Y. Balkanski, M. Schulz, O. Boucher, A. Minikin, and A. Petzold, The aerosolclimate model echam5-ham, Atmospheric Chemistry and Physics, 5, 11251156, 2005, 912EM Times Cited:156 Cited References Count:125. Suarez, M. J., and L. L. Takacs, Documentation of the aries/geos dynamical core: Version 2, Technical Report Technical Memorandum 104 606 Vol. 5, NASA, 1995. Sundqvist, H., Parameterization of condensation and associated clouds in models for weather prediction and general circulation simulation, in Physically-based Modeling and Simulation of Climate and Climate Change, Vol. 1, edited by M. E. Schlesinger, 433461, Kluwer Academic, 1988. Tabazadeh, A., R. P. Turco, and M. Z. Jacobson, A model for studying the composition and chemical eects of stratospheric aerosols, J. Geophys. Res., 99, 12,89712,914, 1994. Tabazadeh, A., M. L. Santee, M. Y. Danilin, H. C. Pumphrey, P. A. Newman, P. J. Hamill, and J. L. Mergenthaler, Quantifying denitrication and its eect on ozone recovery, Science, 288, 14071411, 2000. Taylor, M. A., J. Edwards, and A. St.Cyr, Petascale atmospheric models for the community climate system model: New developments and evaluation of scalable dynamical cores, J. Phys. Conf. Ser., 125 (012023), 2008. Taylor, M. 2010. Conservation of mass and energy for the moist atmospheric primitive equations on unstructured grids. in Numerical Techniques for Global Atmospheric Models, Springer Lecture Notes in Computational Science and Engineering, . Springer. Taylor, M., J. Tribbia, and M. Iskandarani, The spectral element method for the shallow water equations on the sphere, J. Comp. Phys., 130, 92108, 1997. 263

5026 5027 5028

5029 5030

5031 5032

5033 5034

5035 5036

5037 5038 5039 5040

5041 5042

5043 5044 5045 5046

5047 5048

5049 5050 5051

5052 5053 5054

5055 5056 5057

5058 5059

5060 5061 5062

Taylor, M. A., A. St.Cyr, and A. Fournier. 2009. A non-oscillatory advection operator for the compatible spectral element method. in Computational Science ICCS 2009 Part II, Lecture Notes in Computer Science 5545, 273282, , Berlin / Heidelberg. Springer. Taylor, M. A., and A. Fournier, A compatible and conservative spectral element method on unstructured grids, J. Comp. Phys., In Press, , 2010. Teklemariam, T., and J. Sparks, Gaseous uxes of peroxyacetyl nitrate (PAN) into plant leaves, PLANT CELL AND ENVIRONMENT, 27 (9), 11491158, SEP, 2004. Temperton, C., Treatment of the Coriolis terms in semi-Lagrangian spectral models, in Atmospheric and Ocean Modelling. The Andr J. Robert Memorial Volume, edited by C. Lin, e R. Laprise,, and H. Ritchie, 293302, Canadian Meteorological and Oceanographic Society, Ottawa, Canada, 1997. Temperton, C., M. Hortal, and A. Simmons, A two-time-level semi-Lagrangian global spectral model, Q. J. R. Meteorol. Soc., 127, 111127, 2001. Thomas, S., and R. Loft, Parallel semi-implicit spectral element methods for atmospheric general circulation models, J. Sci. Comput., 15, 499518, 2000. Thomas, S., and R. Loft, The NCAR spectral element climate dynamical core: Semi-implicit Eulerian formulation, J. Sci. Comput., 25, 307322, 2005. Thomason, L. W., L. R. Poole, and T. Deshler, A global climatology of stratospheric aerosol surface area density measurements: 1984-1994, J. Geophys. Res., 102, 89678976, 1997. Thompson, G. M., R. M. Rasmussen, and K. Manning, Explicit forecasts of winter precipitation using and improved bulk microphysics scheme. part I: Description and sensitivity analysis, Mon. Weather Rev., 132, 519542, 2004. Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. part ii: Implementation of a new snow parameterization, Monthly Weather Review, 136 (12), 50955115, 2008. Tie, X., S. Madronich, S. Walters, D. P. Edwards, P. Ginoux, N. Mahowald, R. Zhang, C. Lou, and G. Brasseur, Assessment of the global impact of aerosols on tropospheric oxidants, J. Geophys. Res., 110, D03204, 1992. Tie, X., G. Brasseur, L. Emmons, L. Horowitz, and D. Kinnison, Eects of aerosols on tropospheric oxidants: A global model study, Journal of Geophysical Research-Atmospheres, 106 (D19), 2293122964, 2001, 481UJ Times Cited:64 Cited References Count:92. Tie, X. X., S. Madronich, S. Walters, D. P. Edwards, P. Ginoux, N. Mahowald, R. Y. Zhang, C. Lou, and G. Brasseur, Assessment of the global impact of aerosols on tropopshic ozidants, J. Geophys. Res., 110 (D03204), 2005. Tilmes, S., R. Garcia, D. Kinnisen, A. Gettelman, and P. J. Rasch, Impact of geoengineered aerosols on the troposphere and stratosphere, J. Geophys. Res., 114, 2009. 264

5063 5064

5065 5066

5067 5068 5069 5070

5071 5072

5073 5074

5075 5076

5077 5078

5079 5080 5081

5082 5083 5084

5085 5086 5087

5088 5089 5090

5091 5092 5093

5094 5095

5096 5097 5098

Turnipseed, A., G. Huey, E. Nemitz, R. Stickel, J. Higgs, D. Tanner D. andSlusher, J. Sparks, F. Flocke, and A. Guenther, Eddy covariance uxes of peroxyacyl nitrates (pans) and noy to a coniferous forest, J. Geophys. Res., 111, D09304, 2006. van de Hulst, H. C., Light Scattering by Small Particles. Dover, 1957. Vavrus, S., and D. Waliser, An improved parameterization for simulating arctic cloud amount in the CCSM3 climate model, J. Climate, 21, 56735687, 2008. Vehkamaki, H., M. Kulmala, I. Napari, K. E. J. Lehtinen, T. T., N. Noppel, and A. Laaksonen, an improved parameterization for sulfuric acid-water nucleation rates for tropospheric and stratospheric conditions, Journal of Geophysical Research-Atmospheres, 107, 4622, 2002. Walmsley, J., and M. Wesely, Modication of coded parametrizations of surface resistances to gaseous dry deposition, Atmos. Environ., 30 (7), 11811188, APR, 1996. Wang, H., J. J. Tribbia, F. Baer, A. Fournier, and M. A. Taylor, A spectral element version of CAM2, Mon. Wea. Rev., 135, 38253840, 2007. Wang, M. H., J. E. Penner, and X. H. Liu, Coupled impact aerosol and NCAR CAM3 model: Evaluation of predicted aerosol number and size distribution, Journal of Geophysical ResearchAtmospheres, 114, D06302, 2009. Wang, Y.-M., J. Lean, and N. Sheeley, Modeling the suns magnetic eld and irradiance since 1713, The Astrophysical Journal, 625 (1), 2005. Warren, S., and R. E. Brandt, Optical constants of ice from the ultraviolet to the microwave: A revised compilation, J. Geophys. Res., 113, 2008. Washington, W. M., Documentation for the Community Climate Model (CCM), Version , Technical Report NTIS No. PB82 194192, National Center for Atmospheric Research, Boulder, Colorado, 1982. Webster, S., A. R. Brown, D. R. Cameron, and P. Jones, Improvements to the representation of orography in the met oce unied model, Quart. J. Roy. Meteor. Soc., 129, 19892010, 2003. Weimer, D., Models of high-latitude electric potentials derived with the least error t of spherical harmonic coecients, J. Geophy. Res, 100, 19,595, 1995. Wesely, M., Parameterization of surface resistances to gaseous dry deposition in regional-scale numerical models, Atmos. Environ., 23, 12931304, 1989. Wesely, M., and B. Hicks, A review of the current status of knowledge on dry deposition, Atmos. Environ., 34 (12-14), 22612282, 2000. Wiacek, A., and T. Peter, On the availability of uncoated mineral dust ice nuclei in cold cloud regions, Geophys. Res. Lett., 36 (L17801), 2009.

5099

5100 5101

5102 5103 5104

5105 5106

5107 5108

5109 5110 5111

5112 5113

5114 5115

5116 5117 5118

5119 5120

5121 5122

5123 5124

5125 5126

5127 5128

265

5129 5130 5131

Williamson, D. L., Description of NCAR Community Climate Model (CCM0B), Technical Report NCAR/TN-210+STR, National Center for Atmospheric Research, Boulder, Colorado, NTIS No. PB83 23106888, 88 pp., 1983. Williamson, D. L., Time-split versus process-split coupling of parameterizations and dynamical core, Mon. Wea. Rev., 130, 20242041, 2002. Williamson, D. L., L. M. Bath, R. K. Sato, T. A. Mayer, and M. L. Kuhn, Documentation of NCAR CCM0B program modules, Technical Report NCAR/TN-212+IA, National Center for Atmospheric Research, Boulder, Colorado, NTIS No. PB83 263996, 198 pp., 1983. Williamson, D. L., J. T. Kiehl, V. Ramanathan, R. E. Dickinson, and J. J. Hack, Description of NCAR Community Climate Model (CCM1), Technical Report NCAR/TN-285+STR, National Center for Atmospheric Research, Boulder, Colorado, 112 pp., 1987. Williamson, D. L., and J. G. Olson, Climate simulations with a semi-Lagrangian version of the NCAR Community Climate Model, Mon. Wea. Rev., 122, 15941610, 1994a. Williamson, D. L., and J. G. Olson, Climate simulations with a semi-lagrangian version of the NCAR community climate model, Mon. Wea. Rev., 122, 15941610, 1994b. Williamson, D. L., and P. J. Rasch, Two-dimensional semi-Lagrangian transport with shapepreserving interpolation, Mon. Wea. Rev., 117, 102129, 1989. Williamson, D. L., and J. M. Rosinski, Accuracy of reduced grid calculations, Q. J. R. Meteorol. Soc., 126, 16191640, 2000. Williamson, D. L., and G. S. Williamson, Circulation statistics from January and July simulations with the NCAR Community Climate Model (CCM0B), Technical Report NCAR/TN224+STR,, National Center for Atmospheric Research, Boulder, Colorado, NTIS No. PB85 165637/AS, 112 pp., 1984. Williamson, G. S., CCM2 datasets and circulation statistics, Technical Report NCAR/TN391+STR, National Center for Atmospheric Research, Boulder, Colorado, 85 pp., 1993. Williamson, G. S., and D. L. Williamson, Circulation statistics from seasonal and perpetual January and July simulations with the NCAR Community Climate Model (CCM1): R15, Technical Report NCAR/TN-302+STR, National Center for Atmospheric Research, Boulder, Colorado, 199 pp., 1987. Wilson, J., C. Cuvelier, and F. Raes, A modeling study of global mixed aerosol elds, Journal of Geophysical Research-Atmospheres, 106 (D24), 34,08134,108, 2001. Wilson, J. D., Representing drag on unresolved terrain as a distributed momentum sink, J. Atmos. Sci., 59, 16291637, 2002. Wiscombe, W. J., Mie scattering calculations: Advances in technique and fast, vector-speed computer codes., Technical Report Tech. Note. NCAR/TN-140+STR, NCAR, 1996.

5132 5133

5134 5135 5136

5137 5138 5139

5140 5141

5142 5143

5144 5145

5146 5147

5148 5149 5150 5151

5152 5153

5154 5155 5156 5157

5158 5159

5160 5161

5162 5163

266

5164 5165

World Meteorological Organization, Scientic Assessment of Ozone Depletion: 2002 Global Ozone Research and Monitoring Project, Report no. 47, 498 pp., Geneva, 2003. Wu, T. T., High frequency scattering, Phys. Rev., 104, 12011212, 1956. Yang, P., H. Wei, H. L. Huang, B. A. Baum, Y. X. Hu, G. W. Kattawar, M. I. Mishchenko, and Q. Fu, Scattering and absorption property database for nonspherical ice particles in the nearthrough far-infrared spectral region, Appl. Opt., 44, 55125523, 2005. Yonemura, S., S. Kawashima, and H. Tsuruta, Carbon monoxide, hydrogen, and methane uptake by soils in a temperate arable eld and a forest, J. Geophys. Res., 105 (D11), 14,34714,362, JUN 16, 2000. Yoshioka, M., N. M. Mahowald, A. J. Conley, W. D. Collins, D. W. Fillmore, C. S. Zender, and D. B. Coleman, Impact of desert dust radiative forcing on sahel precipitation: Relative importance of dust compared to sea surface temperature variations, vegetation changes, and greenhouse gas warming, Journal of Climate, 20 (8), 14451467, 2007, 157SY Times Cited:33 Cited References Count:80. Young, K. C., The role of contact nucleation in ice phase initiation of clouds, J. Atmos. Sci., 31, 768776, 1974. Zeng, X., and R. E. Dickinson, Eect of surface sublayer on surface skin temperature and uxes, J. Climate, 11, 537550, 1998. Zeng, X., M. Zhao, and R. E. Dickinson, Intercomparison of bulk aerodynamic algorithms for the computation of sea surface uxes using TOGA COARE and TAO data, J. Climate, 11, 26282644, 1998. Zerroukat, M., N. Wood, and A. Staniforth, A monotonic and positive-denite lter for a semi-lagrangian inherently conserving and ecient (slice) scheme, Q. J. R. Meteorol. Soc., 131 (611), 29232936, 2005. Zhang, G. J., and N. A. McFarlane, Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model, AtmosphereOcean, 33, 407446, 1995. Zhang, L. M., S. L. Gong, J. Padro, and L. Barrie, A size-segregated particle dry deposition scheme for an atmospheric aerosol module, Atmospheric Environment, 35 (3), 549560, 2001, 395TK Times Cited:136 Cited References Count:49. Zhang, M., W. Lyn, C. S. Bretherton, J. J. Hack, and P. J. Rasch, A modied formulation of fractional stratiform condensation rate in the ncar community atmosphere model (cam2), J. Geophys. Res., 000, 10000000, 2003a. Zhang, M., W. Lin, C. S. Bretherton, J. J. Hack, and P. J. Rasch, A modied formulation of fractional stratiform condensation rate in the NCAR community atmospheric model CAM2, J. Geophys. Res., 108 (D1), 2003b.

5166

5167 5168 5169

5170 5171 5172

5173 5174 5175 5176 5177

5178 5179

5180 5181

5182 5183 5184

5185 5186 5187

5188 5189 5190

5191 5192 5193

5194 5195 5196

5197 5198 5199

267

5200 5201

Zhang, Y., and L. Paxton, An empirical kp-dependent global auroral model based on timed/guvi fuv data, Journal of Atmospheric and Solar-Terrestrial Physics, 70 (8-9), 1231 1242, 2008. Zilitinkevich, S. S., Dynamics of the Atmospheric Boundary Layer. Gidrometeoizdat, Leningrad, 1970, 292 pp.

5202 5203

268