Users Guide Rttov12 v1.3

RTTOV v12 Users Guide Doc ID : NWPSAF-MO-UD-037
Version : 1.3
Date : 05/03/2019
RTTOV v12 Users Guide
James Hocking, Peter Rayer, David Rundle and Roger Saunders

Met Office, Exeter, UK
&
Marco Matricardi and Alan Geer
ECMWF
&
Pascal Brunel and Jérôme Vidot
MétéoFrance
This documentation was developed within the context of the EUMETSAT Satellite
Application Facility on Numerical Weather Prediction (NWP SAF), under the Cooperation
Agreement dated 7 December 2016, between EUMETSAT and the Met Office, UK, by one
or more partners within the NWP SAF. The partners in the NWP SAF are the Met Office,
ECMWF, DWD and Météo France.
Copyright 2019, EUMETSAT, All Rights Reserved.
Change record
Version Date Author / changed by Remarks
0.1 12/08/16 JH First draft for v12 beta.
1.0 05/12/16 JH Updates after beta.
1.1 31/01/17 JH Updates after DRI.
1.2 03/04/18 JH Updates for v12.2.
1.3 05/03/19 JH Updates for v12.3.
1
Version : 1.3
Date : 05/03/2019
TABLE OF CONTENTS
1. INTRODUCTION AND SCOPE ................................................................................................ 5
2. OVERVIEW OF RTTOV V12 .................................................................................................... 5
3. SUMMARY OF RTTOV CONCEPTS .................................................................................... 12
4. CHANGES FROM PREVIOUS VERSIONS .......................................................................... 14

4.1. CHANGES FROM RTTOV V12.2 14
4.2. CHANGES FROM RTTOV V12.1 16
4.3. CHANGES FROM RTTOV V11 18
5. FORTRAN-90 UNIX/LINUX INSTALLATION AND TESTING ........................................ 22
5.1 UNPACKING THE CODE 23
5.2 COMPILING THE CODE 25
5.3 RUNNING THE TEST SUITE 29
6. RTTOV TYPES, SUBROUTINES, LIBARARIES AND EXECUTABLES ........................ 33
6.1. LINKING YOUR CODE AGAINST RTTOV 33
6.2. DOXYGEN DOCUMENTATION 34
7. RUNNING RTTOV V12 FOR YOUR APPLICATIONS ....................................................... 37
7.1. SET RTTOV OPTIONS 38
7.2. INITIALISE COEFFICIENT STRUCTURES 38
7.3. SET UP INPUT PROFILES 40
7.4. SPECIFYING THE CHANNELS TO SIMULATE 46
7.5. SPECIFYING SURFACE EMISSIVITY 47
7.6. SPECIFYING SURFACE REFLECTANCE FOR SOLAR SIMULATIONS 50
7.7. ALLOCATION OF TRAJECTORY STRUCTURES. 52
7.8. OUTPUT ARRAYS FROM RTTOV V12 52
7.9. CALLING THE TANGENT LINEAR (TL), ADJOINT (AD) AND JACOBIAN (K) MODELS 56
7.10. MULTI-THREADED EXECUTION 58
7.11. RTTOV PERFORMANCE 58
7.12. SUMMARY OF STEPS FOR RUNNING RTTOV V12 60
8. DETAILS OF SPECIFIC RTTOV CAPABILITIES .............................................................. 61
8.1. SIMULATION OF CLEAR AIR RADIANCES FOR INFRARED AND MICROWAVE CHANNELS 61
8.2. SIMULATION OF CLEAR AIR RADIANCES FOR VISIBLE AND NEAR-INFRARED CHANNELS 61
8.3. SIMPLE CLOUD 63
8.4. DEFINITION OF SURFACE EMISSIVITY 63
8.5. SIMULATION OF VISIBLE AND IR CLOUD-AFFECTED RADIANCES 65
8.6. SIMULATION OF VISIBLE AND IR AEROSOL-AFFECTED RADIANCES 73
8.7. SIMULATION OF MICROWAVE RADIANCES SCATTERED BY CLOUD AND PRECIPITATION 76
8.8. SIMULATION OF HYPERSPECTRAL IR SOUNDER RADIANCES USING PC-RTTOV 78
8.9. SIMULATION OF HYPERSPECTRAL IR SOUNDER RADIANCES USING HTFRTC 80
8.10. INCLUSION OF NON-LOCAL THERMODYNAMIC EQUILIBRIUM EFFECTS 82
8.11. OPTION TO TREAT SURFACE AS A LAMBERTIAN REFLECTOR 83
8.12. ZEEMAN EFFECT FOR SSMIS AND AMSU-A 84
8.13. SIMULATION OF SSU RADIANCES 85
8.14. SIMULATIONS WITH VARIABLE SO2 85
9. LIMITATIONS OF RTTOV V12 ............................................................................................. 86
2
Version : 1.3
Date : 05/03/2019
10. REPORTING AND KNOWN BUGS FOR RTTOV V12 ....................................................... 86
11. FREQUENTLY ASKED QUESTIONS ................................................................................... 86
12. GLOSSARY ................................................................................................................................ 87
13. REFERENCES ........................................................................................................................... 88
14. ANNEXES ................................................................................................................................... 92

Annex A - Coefficient information and conversion tools ........................................................ 92
1. RTTOV_COEF_INFO.EXE ................................................................................................ 92
2. RTTOV_MFASIS LUT_INFO.EXE..................................................................................... 92
3. RTTOV_CONV_COEF_.EXE ............................................................................................. 92
4. RTTOV_ASCII2BIN_SCATTCOEF.EXE............................................................................ 94
5. RTTOV11_CONV_COEF_11TO12.EXE............................................................................ 94
6. RTTOV11_CONV_COEF_12TO11.EXE............................................................................ 94
Annex B – RTTOV_ERRORHANDLING interface ................................................................. 95
Annex C – Coefficient allocation and deallocation subroutines ............................................ 96
1. RTTOV_READ_COEFS interface ...................................................................................... 96
2. RTTOV_READ_COEFS_HTFRTC interface ..................................................................... 98
3. RTTOV_DEALLOC_COEFS interface ............................................................................... 98
4. RTTOV_READ_SCATTCOEFFS interface ........................................................................ 98
5. RTTOV_DEALLOC_SCATTCOEFFS interface ................................................................. 99
Annex D – RTTOV allocation/deallocation and initialisation subroutines .......................... 100
1. RTTOV_ALLOC_DIRECT interface ................................................................................ 100
2. RTTOV_ALLOC_TL interface .......................................................................................... 102
3. RTTOV_ALLOC_AD interface ......................................................................................... 103
4. RTTOV_ALLOC_K interface ............................................................................................ 104
5. RTTOV_ALLOC_PROF interface .................................................................................... 105
6. RTTOV_INIT_PROF interface ......................................................................................... 105
7. RTTOV_ALLOC_RAD interface ....................................................................................... 105
8. RTTOV_INIT_RAD interface............................................................................................ 106
9. RTTOV_ALLOC_TRANSMISSION interface ................................................................... 106
10. RTTOV_INIT_TRANSMISSION interface ...................................................................... 106
11. RTTOV_ALLOC_PCCOMP interface ............................................................................ 107
12. RTTOV_INIT_PCCOMP interface ................................................................................. 107
13. RTTOV_ALLOC_OPT_PARAM interface ...................................................................... 108
14. RTTOV_INIT_OPT_PARAM interface ........................................................................... 108
15. RTTOV_ALLOC_TRAJ interface.................................................................................... 109
16. RTTOV_ALLOC_SCATT_PROF interface ..................................................................... 110
17. RTTOV_INIT_SCATT_PROF interface.......................................................................... 110
18. RTTOV_ALLOC_EMIS_RET_TERMS interface ............................................................ 110
19. RTTOV_SCATT_SETUPINDEX interface...................................................................... 111
Annex E – Optical parameter calculation subroutines......................................................... 112
1. RTTOV_BPR_INIT interface ............................................................................................ 112
2. RTTOV_BPR_CALC interface.......................................................................................... 112
3. RTTOV_BPR_DEALLOC interface .................................................................................. 112
4. RTTOV_LEGCOEF_CALC interface ............................................................................... 113
Annex F – Emissivity atlas subroutines ................................................................................ 114
3
Version : 1.3
Date : 05/03/2019
1. RTTOV_SETUP_EMIS_ATLAS interface ........................................................................ 114

2. RTTOV_GET_EMIS interface .......................................................................................... 115
3. RTTOV_DEALLOCATE_EMIS_ATLAS interface............................................................ 117
Annex G – BRDF atlas subroutines ...................................................................................... 118
1. RTTOV_SETUP_BRDF_ATLAS interface ....................................................................... 118
2. RTTOV_GET_BRDF interface ......................................................................................... 118
3. RTTOV_DEALLOCATE_BRDF_ATLAS interface .......................................................... 119
Annex H – RTTOV_GET_PC_PREDICTINDEX interface .................................................. 120
Annex I – RTTOV_DIRECT interface................................................................................... 121
Annex J – RTTOV_K interface ............................................................................................. 123
Annex K – RTTOV_TL interface ........................................................................................... 125
Annex L – RTTOV_AD interface .......................................................................................... 127
Annex M – RTTOV_SCATT interface ................................................................................... 129
Annex N – RTTOV Utility routines ....................................................................................... 132
1. RTTOV_USER_OPTIONS_CHECKINPUT interface ...................................................... 132
2. RTTOV_USER_PROFILE_CHECKINPUT interface ...................................................... 132
3. RTTOV_PRINT_OPTS interface ...................................................................................... 133
4. RTTOV_PRINT_OPTS_SCATT interface ......................................................................... 133
5. RTTOV_PRINT_INFO interface ...................................................................................... 133
6. RTTOV_PRINT_PROFILE interface................................................................................ 134
7. RTTOV_PRINT_CLD_PROFILE interface ...................................................................... 134
8. RTTOV_PRINT_RADIANCE_QUALITY interface .......................................................... 134
9. CREATE_AER_CLIM_PROF.EXE .................................................................................. 135
10. RTTOV_AER_CLIM_PROF ........................................................................................... 135
11. RTTOV_ZUTILITY ......................................................................................................... 136
12. RTTOV_OBS_TO_PC.EXE ............................................................................................ 138
13. RTTOV_SCATT_EMIS_RETRIEVAL ............................................................................. 138
14. RTTOV_MAKE_SCAERCOEF.EXE .............................................................................. 139
15. RTTOV_SCALE_REF_GAS_PROF ............................................................................... 139
Annex O – RTTOV v12 derived types ................................................................................... 140
4
Version : 1.3
Date : 05/03/2019
1. Introduction and Scope

This document is structured as follows. Section 2 gives a broad overview of the RTTOV v12 fast radiative transfer
model. Section 3 provides a brief introduction to the key concepts involved in running RTTOV v12 and it is
recommended to all users. Section 4 provides details of all changes since RTTOV v11: this is intended to be a reference
for users upgrading from v11. Section 5 gives the instructions for compiling RTTOV, verifying the build and, more
generally, for running the test suite. Section 6 provides a reference for the RTTOV libraries, executables, subroutines
and derived types. Section 7 provides a detailed step-by-step guide for implementing RTTOV in your own application.
Section 8 gives details of the various simulation capabilities of RTTOV: you should consult the parts of this section
relevant to your application. Section 9 lists the limitations of RTTOV v12. The procedure for reporting bugs or learning
about known bugs is given in section 10. Finally a frequently asked questions (FAQ) section and glossary are provided
in sections 11 and 12. This document relates to version 12 of the RTTOV code and all its sub-versions (12.x). The
document is only updated with new minor releases of RTTOV v12: the document version is given in the header. To
obtain a copy of the RTTOV v12 code go to http://nwpsaf.eu/site/register/ and register on the website. Once you are
registered you can log-in and click on the “Software Downloads” link to subscribe to RTTOV v12: once you agree to
the licence you will be able to download the code immediately.
RTTOV v12 is an evolution of RTTOV v11, adding and upgrading many features as documented here. The RTTOV
v11 code is still available, but cannot be guaranteed to be upgraded for new instruments and capability. The format of
the coefficient files has changed since RTTOV v11 and as such RTTOV v10/v11 coefficient files cannot be read by
RTTOV v12 and vice versa. Exectuables are included in the package to convert RTTOV v10/v11-format rtcoef files to
RTTOV v12 format and vice versa (where possible): these are described in Annex A. Existing RTTOV v11 coefficient
files will remain available on the website. However the latest v12 coefficients provided with the package and available
via the RTTOV v12 web page are the recommended ones for RTTOV v12.
The RTTOV v12 scientific and validation report describes or gives links to the scientific basis of the model and also
describes in more detail any new scientific changes made. It also documents the test results carried out on the new code
before delivery. The most up to date versions of these reports, including this user guide, can be viewed at the NWP SAF
web site: http://nwpsaf.eu/site/software/rttov/ in pdf format on the RTTOV v12 page. There is also a RTTOV v12
performance report which documents the run times of RTTOV v12 on a few platforms and compares these to the
equivalent RTTOV v11 run times.
The NWP SAF web forum can be found here: http://nwpsaf.eu/site/forums/. Information about RTTOV including bugs
and updates is posted here and users are encouraged to post questions and comments about the software.
2. Overview of RTTOV v12

This section gives a brief overview of the RTTOV v12 model. More details can be found in the references given in this
section. RTTOV v12 is a development of the fast radiative transfer model for TOVS, RTTOV, originally developed at
ECMWF in the early 90's (Eyre, 1991) for TOVS. Subsequently the original code has gone through several
developments (e.g. Saunders et al., 1999; Matricardi et al., 2001), more recently within the EUMETSAT NWP Satellite
Application Facility (SAF), of which RTTOV v12 is the latest version (Saunders et al., 2018). The model allows rapid
simulations (~1 ms for 40 channel ATOVS on a desktop PC) of radiances for satellite visible, infrared or microwave
nadir scanning radiometers given an atmospheric profile of temperature, variable gas concentrations, cloud and surface
properties, referred to as the state vector. The only mandatory variable gas for RTTOV v12 is water vapour. Optionally
ozone, carbon dioxide, nitrous oxide, methane, carbon monoxide and sulphur dioxide can be variable with all other
constituents assumed to be constant. The state vector for RTTOV v12 is given in Annex O. RTTOV v12 can accept
input profiles on any defined set of pressure levels. The range of temperatures and water vapour concentrations over
which the optical depth computations are valid depends on the training datasets which were used. This is defined in the
coefficient file and for RTTOV v12 is mainly based on the 91-level 83 diverse profile dataset from ECMWF analyses
for temperature, water vapour and ozone (http://nwpsaf.eu/site/software/atmospheric-profile-data/). For other gases a
range of profile datasets were used based on models and measurements. The limits for temperature, water vapour and
ozone are given in Table 1. These are derived from the strict profile dataset minimum/maximum envelopes by applying
a stretching factor (+/-10% for temperature max/min respectively, and +/-20% for each gas max/min). The RTTOV v12
coefficient files contain the strict min/max envelopes: the limits that are applied within RTTOV are calculated when the
5
Version : 1.3
Date : 05/03/2019
coefficients are read in. More details on the profile datasets used for the different gases can be found in Matricardi
(2008). For RTTOV v12 some trace gas datasets were stretched in order to improve their validity over the whole
satellite era from 1970 and into the 2020s. This is described in the RTTOV v12 Science and Validation Report. The
training profiles used for sulphur dioxide are also described in the Science and Validation Report. The majority of
RTTOV v12 coefficient files are based on the 54 levels shown in Table 1, though coefficients for some hyperspectral
sounders are also available on 101 levels.
The spectral range of the RTTOV v12 model in the visible/infrared is 0.4-50 microns (200 – 25000 cm-1), governed by
the range of the LBLRTM line-by-line datasets on which the coefficients are based. In the microwave the frequency
range is from 10 – 800 GHz which is covered using the Liebe-89 MPM line-by-line model. The full list of currently
supported platforms and sensors is given in Tables 2 and 3. This list is updated as new sensors are launched. For the
visible/IR sensors, the channel order can either be decreasing or increasing with wavelength and in some cases (e.g.
MTSAT imager) it is not monotonic wavelength order, rather being defined according to instrument convention. The
channel order is indicated in Table 3. New or updated coefficient files will be made available from the RTTOV pages
on the NWP SAF web site for the latest RTTOV versions. Coefficients for sensors that are not currently supported may
be requested via the NWP SAF helpdesk.
An important feature of the RTTOV model is that it not only computes the forward (or direct) radiative transfer
calculation but also the gradient of the radiances with respect to the state vector variables at the location in state space
specified by the input state vector values. Given a state vector, x, a radiance vector, y, is computed:
y  H (x) (1)
where H is the radiative transfer model (also referred to as the observation operator). The Jacobian matrix H gives the
change in radiance δy for a change in any element of the state vector δx assuming a linear relationship about a given
atmospheric state x0:
δy  H(x0 )δx (2)
The elements of H contain the partial derivatives yi/xj where the subscript i refers to channel number and j to position
in state vector. The Jacobian gives the top-of-atmosphere radiance change for each channel given unit perturbations at
each respective level of the profile vectors and in each of the surface/cloud parameters. It shows clearly, for a given
profile, which layers in the atmosphere are most sensitive to changes in temperature and variable gas concentrations for
each channel. rttov_k (and its associated subroutines ending in K) compute the H(x0) matrix for each input profile.
It is not always necessary to store and access the full Jacobian matrix H and so the RTTOV package has routines to only
output the tangent linear values δy, the change in top of atmosphere radiances yn for each channel n, for a given change
in atmospheric profile, δx, about an initial atmospheric state x0.
 y1 y y y 
y( x0 )  x , x 2 , x 3 ,.....x nchan  (3)
 x x x x 
The tangent linear routines all have TL as an ending. Conversely the adjoint routines (ending in AD) compute the
change in any scalar quantity up to nel elements of the state vector (e.g. T, q, ozone, surface variables etc) δx for an
assumed atmospheric state, x0, given a change in the radiances, δy.
 x1 x x x 
x( x0 )  y , y 2 , y 3 ,.....y nel  (4)
 y y y y 
These routines are normally used as part of the variational assimilation of radiances. Information about the RTTOV TL,
AD and K models is given in section 7.9 below and some general information on TL/AD and K codes is available at:
http://cimss.ssec.wisc.edu/itwg/groups/rtwg/fastrt.html. If you only want to compute radiances with the forward model
the TL/AD/K routines are not required.
6
Version : 1.3
Date : 05/03/2019
The core of RTTOV simulates clear-sky radiances. For channels with a significant thermally emitted contribution, the
top of the atmosphere upwelling clear-sky radiance, LClr(v,θ), at a frequency v and viewing angle θ can be written as:
B( ,T) (5)

L ( , ) =  s ( , )  s ( , )B( ,T s ) +  s B( ,T)d + (1-  s ( , )) s ( , )  s d
Clr 1 2 1
 2
where s is the surface to space transmittance, s is the surface emissivity and B(v,T) is the Planck function for a
frequency v and temperature T.
The transmittances, , are computed by means of a linear regression in optical depth based on variables from the input
profile vector as described in Matricardi et al. (2001) for RTTOV v7 predictors, Matricardi (2003) for RTTOV v8
predictors and those given in Matricardi (2005) or the RTTOV v9 science plan for RTTOV v9 predictors. The code
supports any of these predictor sets with the selection being made according to the coefficient file supplied to the
program. Table 4 gives more information on the coefficient files available for each instrument type and the underlying
predictor sets. More details on the performance of the different predictor sets are given in the RTTOV v9 Science and
Validation Report. No changes to the optical depth predictors were made for RTTOV v12, but SO2 has been added as a
new optional variable gas as described in the RTTOV v12 Science and Validation Report.
In addition to the clear-air simulations described above (and in section 8.1) there are options to include solar radiation
(section 8.2), for a simple visible/IR cloud scheme based on a single cloud top pressure and cloud fraction (section 8.3),
for visible/infrared cloud- and aerosol-affected radiances (sections 8.5 and 8.6) and for cloud- and precipitation-
affected microwave radiances (section 8.7).
7
Version : 1.3
Date : 05/03/2019
Level Pressure Tmax Tmin Qmax Qmin O3max O3min O3Ref

Number hPa K K ppmv* ppmv* ppmv* ppmv* ppmv*
1 0.01 245.95 143.66 5.24 0.91 1.404 0.014 0.296
2 0.01 252.13 154.19 6.03 1.08 1.410 0.069 0.321
3 0.03 263.71 168.42 7.42 1.35 1.496 0.108 0.381
4 0.06 280.12 180.18 8.10 1.58 1.670 0.171 0.527
5 0.13 299.05 194.48 8.44 1.80 2.064 0.228 0.769
6 0.23 318.64 206.21 8.59 1.99 2.365 0.355 1.074
7 0.41 336.24 205.66 8.58 2.49 2.718 0.553 1.471
8 0.67 342.08 197.17 8.34 3.01 3.565 0.731 1.991
9 1.08 340.84 189.50 8.07 3.30 5.333 0.716 2.787
10 1.67 334.68 179.27 7.89 3.20 7.314 0.643 3.756
11 2.50 322.50 176.27 7.75 2.92 9.191 0.504 4.864
12 3.65 312.51 175.04 7.69 2.83 10.447 0.745 5.953
13 5.19 303.89 173.07 7.58 2.70 12.336 1.586 6.763
14 7.22 295.48 168.38 7.53 2.54 12.936 1.879 7.109
15 9.84 293.33 166.30 7.36 2.46 12.744 1.322 7.060
16 13.17 287.05 163.47 7.20 2.42 11.960 0.719 6.574
17 17.33 283.36 161.49 6.96 2.20 11.105 0.428 5.687
18 22.46 280.93 161.47 6.75 1.71 9.796 0.278 4.705
19 28.69 282.67 162.09 6.46 1.52 8.736 0.164 3.870
20 36.17 279.93 162.49 6.14 1.31 7.374 0.107 3.111
21 45.04 273.15 164.66 5.90 1.36 6.799 0.055 2.478
22 55.44 265.93 166.19 6.21 1.30 5.710 0.048 1.907
23 67.51 264.70 167.42 9.17 1.16 4.786 0.043 1.440
24 81.37 261.95 159.98 17.89 0.36 4.390 0.038 1.020
25 97.15 262.43 163.95 20.30 0.01 3.619 0.016 0.733
26 114.94 259.57 168.59 33.56 0.01 2.977 0.016 0.604
27 134.83 259.26 169.71 102.24 0.01 2.665 0.016 0.489
28 156.88 260.13 169.42 285.00 0.01 2.351 0.013 0.388
29 181.14 262.27 170.63 714.60 0.01 1.973 0.010 0.284
30 207.61 264.45 174.11 1464.00 0.01 1.481 0.013 0.198
31 236.28 270.09 177.12 2475.60 0.01 1.075 0.016 0.145
32 267.10 277.93 181.98 4381.20 0.01 0.774 0.015 0.110
33 300.00 285.18 184.76 6631.20 0.01 0.628 0.015 0.086
34 334.86 293.68 187.69 9450.00 1.29 0.550 0.016 0.073
35 371.55 300.12 190.34 12432.00 1.52 0.447 0.015 0.063
36 409.89 302.63 194.40 15468.00 2.12 0.361 0.015 0.057
37 449.67 304.43 198.46 18564.00 2.36 0.284 0.015 0.054
38 490.65 307.20 201.53 21684.00 2.91 0.247 0.015 0.052
39 532.58 312.17 202.74 24696.00 3.67 0.199 0.015 0.050
40 575.15 315.55 201.61 27480.00 3.81 0.191 0.012 0.050
41 618.07 318.26 189.95 30288.00 6.82 0.171 0.010 0.049
42 661.00 321.71 189.95 32796.00 6.07 0.128 0.009 0.048
43 703.59 327.95 189.95 35328.00 6.73 0.124 0.009 0.047
44 745.48 333.77 189.95 37692.00 8.71 0.117 0.008 0.046
45 786.33 336.46 189.95 39984.00 8.26 0.115 0.008 0.045
46 825.75 338.54 189.95 42192.00 7.87 0.113 0.008 0.043
47 863.40 342.55 189.95 44220.00 7.53 0.111 0.007 0.041
48 898.93 346.23 189.95 46272.00 7.23 0.108 0.006 0.040
49 931.99 349.24 189.95 47736.00 6.97 0.102 0.006 0.038
50 962.26 349.92 189.95 51264.00 6.75 0.099 0.006 0.034
51 989.45 350.09 189.95 49716.00 6.57 0.099 0.006 0.030
52 1013.29 350.09 189.95 47208.00 6.41 0.094 0.006 0.028
53 1033.54 350.09 189.95 47808.00 6.29 0.094 0.006 0.027
54 1050.00 350.09 189.95 47640.00 6.19 0.094 0.006 0.027
*The gas units here are ppmv with respect to dry air because RTTOV is trained on gas profiles in these units and this is
the context in which these limits are applied within RTTOV.
Table 1. Pressure levels adopted for RTTOV v12 54 level coefficients and profile limits within which the
transmittance calculations are valid. The default ozone profile is also given in the right hand column. Note
that some coefficients are based on a standard 101 level pressure profile which extends down to 1100 hPa.
8
Version : 1.3
Date : 05/03/2019
Platform RTTOV ID Sat ID range Platform RTTOV ID Sat ID range

NOAA* 1 1 - 20 Himawari 31 8-9
DMSP 2 8 - 19 MTG 32 1
Meteosat 3 1-7 Saral 33 1
GOES 4 4 - 19 Metop-SG 34 1
GMS 5 1-5 Landsat 35 4,5,7,8
FY2 6 2 - 5,7 Jason 36 2
TRMM 7 1 GPM 37 1
ERS 8 1-2 INSAT-1 38 1-4
EOS 9 1-2 INSAT-2 39 1-5
METOP 10 1-2 INSAT-3 40 4-5
ENVISAT 11 1 Reserved 41 -
MSG 12 1-4 DSCOVR 42 1
FY1 13 3-4 CLARREO 43 1
ADEOS 14 2 TICFIRE 44 1
MTSAT 15 1-2 Reserved 45 -
CORIOLIS 16 1 ISS 46 1
JPSS (SNPP) 17 0 HJ1 47 2
GIFTS 18 1 GEOKOMPSAT2 48 1
Sentinel3 19 1 GCOM-C 49 1
MeghaTropique 20 1 SMOS 50 1
Kalpana 21 1 ORS 51 6
Meteor 22 25 FY4 52 1
FY3 23 1-4 TROPICS 53 0
COMS 24 1 GF5 54 1
METEOR-M 25 1-2 HY2 55 1
GOSAT 26 1 CloudSat 56 1
CALIPSO 27 1 CloudCore 57 1
Reserved 28 - FORUM 58 1
GCOM-W 29 1 TEMPEST-D 59 1
Nimbus 30 3-7
*“NOAA-5” is TIROS-N
Table 2. Platforms supported by RTTOV as of March 2019. Platforms in italics are not yet supported in the
RTTOV v12 distribution but can be requested.
Sensor Chans => Sensor Chans =>

Sensor RTTOV ID RTTOV v12 Chans RTTOV v12 Chans
(IR/MW-only) (VIS/NIR/IR)
HIRS 0 1-19 => 1-19 -
MSU 1 1-4 => 1-4 -
SSU 2 1-3 => 1-3 -
AMSU-A 3 1-15 => 1-15 -
AMSU-B 4 1-5 => 1-5 -
AVHRR 5 3b-5 => 1 to 3 1-6 => 1-6
SSMI 6 1-7 => 1-7 -
VTPR1 7 1-8 => 1-8 -
Spare 8 - -
TMI 9 1-9 => 1-9 -
SSMIS 10 1-24 => 1-24 -
AIRS 11 1-2378 => 1-2378 1-2378 => 1-2378
HSB 12 1-4 => 1-4 -
9
Version : 1.3
Date : 05/03/2019
MODIS 13 (20-25, 27-36) => 1-16 1-36 => 1-36

ATSR 14 1-3 => 1-3 1-7 => 1-7
MHS 15 1-5 => 1-5 -
IASI 16 1-8461 => 1-8461 1-8461 => 1-8461
AMSR-E 17 1-12 => 1-12 -
1 => 1 1-2 => 1-2
GMS imager 18
1-3 => 1-3 1-4 => 1-4
ATMS 19 1-22 => 1-22 -
MVIRI 20 1-2 => 1-2 -
SEVIRI 21 4-11 => 1-8 1-12 => 1-12
GOES imager 22 2-5 => 1-4 1-5 => 1-5
GOES sounder 23 1-18 => 1-18 -
MTSAT imager 24 1-4 => 1-4 1-5 => 1-5
FY2-3/4 VISSR 25 2-5 => 1-4 1-5 => 1-5
FY1 MVISR 26 1-3 => 1-3 -
CrIS 27 1-1305 => 1-1305 1-1305 => 1-1305
CrIS-FSR 28 1-2211 => 1-2211 1-2211 => 1-2211
VIIRS 29 16-22 => 1-7 1-22 => 1-22
WINDSAT 30 1-16 => 1-16 -
GIFTS 31 - -
SSM-T1 32 1-7 => 1-7 -
SSM-T2 33 1-5 => 1-5 -
SAPHIR 34 1-6 => 1-6 -
MADRAS 35 1-9 => 1-9 -
Reserved 36 - -
VHRR 37 2-3 => 1-2 1-3 => 1-3
INSAT imager 38 3-6 => 1-4 1-6 => 1-6
INSAT sounder 39 1-18 => 1-18 1-19 => 1-19
MWTS 40 1-4 => 1-4 -
MWHS 41 1-5 => 1-5 -
IRAS 42 1-20 => 1-20 1-26 => 1-26
MWRI 43 1-10 => 1-10 -
ABI 44 7-16 => 1-10 1-16 => 1-16
COMS MI 45 2-5 => 1-4 1-5 => 1-5
MSUMR 46 1-3 => 1-3 -
TANSO-FTS 47 - -
IIR 48 1-3 => 1-3 -
ESA MWR 49 1-2 => 1-2 -
Reserved 50-53 - -
SCAMS 54 1-5 => 1-5 -
SMMR 55 1-10 => 1-10 -
AHI 56 7-16 => 1-10 1-16 => 1-16
MTG IRS 57 1-1738 => 1-1738 1-1738 => 1-1738
AltiKa 58 1-2 => 1-2 -
IASI-NG 59 1-16921 => 1-16921 1-16921 => 1-16921
Landsat TM 60 6 => 1 -
MTG FCI 61 9-16 => 1-8 1-16 => 1-16
AMSR1 62 1-16 => 1-16 -
AMSR2 63 1-14 => 1-14 -
FY2-2 VISSR 64 1-2 =>1-2 -
SLSTR 65 7-9 => 1-3 1-9 => 1-9
TIRS 66 10-11 => 1-2 -
AMR 67 1-3 => 1-3 -
OLI 68 - 1-9 => 1-9
IRIS 69 1-862 => 1-862 -
ICI 70 1-13 => 1-13 -
GMI 71 1-13 => 1-13 -
MWTS-2 72 1-13 => 1-13 -
MWHS-2 73 1-15 => 1-15 -
10
Version : 1.3
Date : 05/03/2019
ASTER 74 10-14 => 1-5 (1-2,3N,3B,4-14) => 1-15

Reserved 75 - -
MTVZA-GY 76 1-29 => 1-29 -
MetImage 77 11-20 => 1-9 1-20 => 1-20
MWS 78 1-24 => 1-24 -
MWI 79 1-26 => 1-26 -
EPIC 80 - -
MRIR 81 2-5 => 1-4 -
SI 82 1-579 => 1-579 1-579 => 1-579
Reserved 83-86 - -
MERSI-1 87 20 => 1 -
MERSI-2 88 20-25 => 1-6 1-25 => 1-25
ECOSTRESS 89 1-5 => 1-5 -
IRMSS 90 4 => 1 -
OLCI 91 - See coef file headers
THIR 92 1-2 => 1-2 -
AMI 93 7-16 => 1-10 1-16 => 1-16
IKFS2 94 1 - 2701 => 1-2701 1 - 2701 => 1-2701
LI 95 - 1-2 => 1-2
SGLI 96 18-19 => 1-2 -
HIRAS 97 - -
GIIRS 98 - -
7/8-14 => 1-7 1-7/8,9-14 => 1-13
AGRI 99 (coefs include only one 3.9um (coefs include only one 3.9um
channel) channel)
PMR 100 - -
MIRAS 101 1-2 => 1-2 -
COWVR 102 1-12 => 1-12 -
TROPICS 103 1-12 => 1-12 -
VIMS 104 9-12 => 1-4 -
DPR 105 1-4 => 1-4 -
HY2 MWRI 106 1-9 => 1-9 -
CPR 107 1 => 1 -
SAPHIR NG 108 - -
FORUM 109 - -
TEMPEST-D 110 - -
Reserved 111 - -
*channels 19-21 are only simulated accurately with Zeeman coefficient file
The channel order in the coefficient files can be determined using the rttov_coef_info.exe executable (see Annex A), or
by looking at the following web page which gives the spectral response functions and passbands used by RTTOV and
the linked pages show the channel order in the coefficient files:
https://nwpsaf.eu/site/software/rttov/download/coefficients/spectral-response-functions/
Table 3. Instruments supported by RTTOV v12 as of March 2019. Sensors in italics are not yet supported in
the RTTOV v12 distribution but can be requested. “IR/MW only” refers to v7/v8 predictor files;
“VIS/NIR/IR” refers to v9 predictor solar-compatible files.
11
Version : 1.3
Date : 05/03/2019
3. Summary of RTTOV concepts

This section provides a brief overview of the steps required when running RTTOV. Existing users should read this
section for information on coefficient files. New users should read this section to familiarise themselves with the key
concepts related to RTTOV. You should also look through the code contained in src/test/example_fwd.F90 and/or
src/test/example_k.F90 as these provide fully-commented templates for running the RTTOV forward and K models.
The src/test/ directory also contains examples for running other kinds of simulation. A detailed description of the steps
involved in running RTTOV is given in section 7. There is an “RTTOV Quick Start Guide” in the docs/ directory which
is intended to guide new users through a simple clear-sky forward model simulation.
All aspects of RTTOV simulations can be configured at run-time via the rttov_options structure (described fully in
Annex O). This provides a list of logical flags and other settings to control various aspects of the simulation such as
whether solar radiation should be included, whether clouds and/or aerosol scattering should be included for IR
instruments, and whether the RTTOV interpolator should be used.
Calculation of clear-sky optical depths is carried out by a predictor-based regression scheme. The coefficients for the
optical depth regression are instrument-specific and are stored in RTTOV coefficient files whose names begin
“rtcoef_”. There are three flavours of predictors denoted “v7”, “v8” and “v9” which determine the variable trace gases
available for the simulation and whether solar simulations are available. The coefficients for v9 predictors are trained
over a wider range of zenith angles to allow for solar simulations which means the maximum allowable zenith angle is
larger (potentially relevant for geostationary satellites) compared to v7 and v8 predictors. The predictor versions are
summarised in Table 4.
The latest RTTOV coefficients have been trained using an updated profile dataset (see the RTTOV v12 Science and
Validation Report). The reference profiles for trace gases (CO2, CH4 and N2O in particular) are appropriate for
contemporary simulations. When simulating older instruments you may wish to specify a more appropriate CO2 profile.
Coefficients are now available with v8 predictors (variable ozone and CO2) for all IR sensors that enable you to vary
the CO2 profile. In addition the v9 predictor coefficients for visible/IR sensors now have variable ozone and CO2.
Previously they only allowed variable ozone: the new coefficients reproduce radiances from the line-by-line model as
well as or better than the ozone-only coefficients. The only exception to this is the 3.9µm channel for SEVIRI so ozone-
only SEVIRI v9 predictor coefficients are available (with “o3” in the filename).
RTTOV accounts for thermal emission for all channels at wavelengths above 3µm (referred to as “thermal” channels).
Thermal emission is ignored at wavelengths below 3µm. Solar radiation is only included in channels at wavelengths
below 5µm. These are referred to as “solar” channels. Channels below 3µm are solar-only channels, while those in the
range 3-5µm are mixed thermal+solar channels.
Important notes on v7/v8 vs v9 predictor coefficients:

 Where both v7/v8 and v9 predictor files exist for an instrument (see Table 3), the v7/v8 predictor files support
only the IR channels, while the v9 predictor file supports all visible, near-IR and IR instrument channels with
wavelengths above 0.4µm.
 The RTTOV channel numbering in the coefficient files begins at one. The v7 and v8 predictor files have the
same channel numbering, but in many cases the channel numbering for the IR channels is not the same for the
v7/v8 and v9 predictor files. Table 3 provides information on how the instrument channel numbers map to the
RTTOV channel numbers and this can also be found by examining the headers of the coefficient file.
 Simulated IR radiances for the same options and profile will show small differences when different predictor
versions are used. It is entirely reasonable to use the v9 predictor coefficients for simulating IR channels only.
However it is not the case that, for example, v9 predictors are generally superior to v7 or v8 predictors. For IR-
only simulations you may find the v7 or v8 predictors are the optimal choice. There are plots illustrating the fit
of the optical depth regression in terms of BT for each coefficient file on the RTTOV web site:
https://www.nwpsaf.eu/site/software/rttov/download/coefficients/comparison-with-lbl-simulations/
12
Version : 1.3
Date : 05/03/2019
The accuracy of simulations for very broad channels (e.g. SEVIRI channel 4 at 3.9 microns) is poor with significant
biases noted (~1-2K) (see e.g. Brunel and Turner, 2003). To mitigate this, the line-by-line optical depths in the
coefficient generation are weighted with the Planck function across the instrument channel and the coefficients are then
computed for these Planck-weighted optical depths resulting in much reduced biases. Whether coefficients are Planck-
weighted or not for a channel can be determined by examining the PLANCK_WEIGHTED section in the coefficient
file (if it is not present there are no Planck-weighted channels).
The following page provides the definitive reference for RTTOV v12 coefficient files:
http://nwpsaf.eu/site/software/rttov/download/coefficients/coefficient-download/
The coefficients files are read in using the rttov_read_coefs subroutine. If Principal Component (PC), cloud or aerosol
simulations are being performed, the additional associated coefficients for those simulations are read in the same call.
You must then allocate some derived types or structures to hold various input and output quantities. These structures are
all described fully in Annex O. The main ones are:
 chanprof structure – array holding a list of channel and profile indices to simulate
 profile structure – array holding the input atmospheric and surface variables
 emissivity structure – array holding the surface emissivity
 reflectance structure – array holding the surface reflectance
 radiance structure – to hold the output simulated radiances
 transmittance structure – to hold the output simulated transmittances
In a typical application you would then loop over input profiles. For each profile, the input profile data is read into the
profile, emissivity and reflectance structures. Then RTTOV is called and the simulated radiances are written out or
stored. It is possible to pass multiple profiles into RTTOV in a single call if desired. There is no fixed limit on the
number of profiles you can pass into RTTOV in one go: this is limited only by the memory available (some kinds of
simulations take more memory than others). Once all simulations are complete, you should call a number of
deallocation subroutines to release allocated memory.
v7 predictors v8 predictors v9 predictors

Max zenith angle 75 degrees 75 degrees 85* degrees
Solar computations? No No Yes
MW instruments All MW sensors. N/A N/A
Coefs on 54 levels except
for Zeeman coef files.
Variable gases: H2O
Hyperspectral Coefs on 54 and 101 levels Coefs on 101 levels Coefs on 101 levels
sounders Variable gases: H2O, O3 Variable gases: H2O, O3, CO2 Variable gases: H2O, O3, CO2,
CO, NO2, CH4, SO2
IR instruments All IR instruments except All IR instruments Selected instruments (see final
(including those with SSU, PMR IR channels only column in Table 3)
visible/near-IR IR channels only Coefs on 54 levels (SSU on 51 Coefs on 54 levels
channels) Coefs on 54 levels levels) Visible/near-IR/IR channels
Variable gases: H2O, O3 Variable gases: H2O, O3, CO2 Variable gases: H2O, O3, CO2
Table 4. Flavours of RTTOV coefficients.
*
For some instruments the coefficients are trained up to a slightly smaller maximum angle as it improves the overall
accuracy of the optical depth calculations: see the comments in the coefficient file headers.
13
Version : 1.3
Date : 05/03/2019
4. Changes from previous versions
4.1. Changes from RTTOV v12.2

This section provides details of the differences between RTTOV v12.2 and v12.3. Users will be able to switch v12.3 for
v12.2 in their Fortran application without having to make any code changes and the outputs will be the same. Users of
the C++/Python wrapper will have to make minor modifications to their existing code to make it work with v12.3 (see
below). Users running MFASIS or HTFRTC simulations will need to download new coefficient files for use with v12.3.
RTTOV v12.3 provides a number of new options and capabilities: these are summarised here and described more fully
elsewhere in the user guide.
Main new/updated features in v12.3
Clear-sky/general simulation updates:

 New CAMEL climatology atlas for IR emissivities: emissivity data based on 2000-2016 climatology,
improved treatment of snow, and includes emissivity standard deviations computed from multi-year
climatology (see sections 7.5, 8.4 and Annex G).
 New option opts%rt_all%lambertian_fixed_angle to switch between fixed or parameterised effective angle
for downwelling radiance with the Lambertian surface option (see section 8.11).
 A new specularity profile skin variable has been added which specifies the weighting for a linear combination
of specular and Lambertian downwelling radiances with the Lambertian surface option (see section 8.11).
 A new subroutine has been added (rttov_scale_ref_gas_prof) which makes it easy to use scaled copies of the
RTTOV reference profiles for optional trace gases in simulations (see Annex N).
Visible/IR scattering updates: (sections 8.5 and 8.6)

 MFASIS has been modified to allow for variable water vapour in affected sensor channels (currently applies to
SEVIRI 0.8 micron channel only) which greatly improves the fit to the training data. New MFASIS LUTs for
all sensors must be downloaded from the RTTOV coefficients web page for use with v12.3.
 An updated parameterisation of the Baran ice cloud property database has been added for visible/IR cloudy
simulations which is more spectrally consistent and extends the parameterisation to the far-IR.
 A new option opts%rt_ir%grid_box_avg_cloud has been added. If false (the default) then input cloud
concentrations for visible/IR cloud simulations should represent the concentration of cloud in the cloudy
fraction of the layer (the behaviour of previous versions of RTTOV). If the option is true cloud concentrations
for visible/IR cloud simulations should be layer grid-box-average values (consistent with RTTOV-SCATT),
which is commonly the quantity output by GCMs. This does not affect simulations with
opts%rt_ir%user_cld_opt_param true.
 A new executable has been created (rttov_make_scaercoef.exe) which enables the generation of custom
aerosol optical property files for use with RTTOV (see Annex N).
Technical updates:
 The wrapper has been updated to support the new options, the specularity profile variable, and to enable
simulations with user-generated aerosol optical property files. NB The specularity profile variable has been
added to the “skin” variable array for the Profile and Profiles classes and this means that user code which calls
the RTTOV v12.2 wrapper will need minor updates to work with v12.3. See the wrapper user guide.
 The GUI has been updated to support the new code options, the new specularity profile variable, and to enable
choice among the available emissivity atlases.
HTFRTC updates (see section 8.9):

 all RTTOV trace gases may now optionally vary in the simulation; ozone is now optional
 options to calculate overcast radiances and simple cloud (as RTTOV does)
 option to supply emissivities; IR emissivity atlas interface has been updated to work with HTFRTC
 support for all RTTOV gas units options, do_lambertian option with new specularity profile variable, and the
plane_parallel option
 multi-threaded simulations via parallel interfaces enabled
 NB a new static coefficient file must be downloaded from the coefficients web page for use with v12.3.
14
Version : 1.3
Date : 05/03/2019
Additional changes since v12.1
Changes in behaviour
The CrIS FSR instrument ID has changed from 27 to 28.
When apply_reg_limits is true, RTTOV does not print out warnings when input profiles exceed regression limits (this
reverts to the behaviour of v12.1 and earlier).
The rttov_user_profile_checkinput subroutine no longer warns when ice water contents are above zero, but below the
minimum value for the selected ice_scheme (such small values are unlikely to be problematic).
The rttov_print_opts subroutine output is more concise: irrelevant options are not printed (e.g. cloud-related options
are not printed if addclouds is false).
Some internal changes in RTTOV have been made to improve the validity checks on input data: in particular the
checking of input profile data for unphysical values now includes tests on cloud/aerosol variables (these were not
checked previously).
Other new options
The number of PC scores to compute for PC-RTTOV may now be specified in a new option
opts%rt_ir%pc%npcscores. If the value of this option is <1 then RTTOV derives npcscores from the size of the
arrays in pccomp which is the previous behaviour. The advantage of using this option is that pccomp can be sized to
the maximum required, but RTTOV can then be called on fewer profiles without having to re-size pccomp. This is a
common scenario when passing batches of different numbers of profiles to RTTOV.
A new option opts%rt_all%rad_down_lin_tau has been added which switches between linear-in-tau and layer-
average radiance expressions for the downwelling specular radiance calculation. The difference in TOA radiances
between the two options is very small (order 1mK), but the layer-average is slightly faster and the code is simpler and is
consistent with Lambertian downwelling term which always uses the layer-average source term. The removal of the
linear-in-tau code for downwelling radiances is under consideration for v13.
A new option opts%rt_mw%clw_cloud_top has been added which allows the lower pressure limit for MW CLW
absorption calculations to be changed. Previously this was hard-coded at 322hPa above which input CLW values were
ignored, and this remains the default (see section 7.3).
A new option opts%rt_ir%rayleigh_single_scatt has been added which allows the Rayleigh single-scattering
calculation to be enabled/disabled for visible channels (the TOA radiances still include extinction due to Rayleigh
scattering regardless of this option).
A new option opts%rt_ir%cldstr_low_cloud_top has been added which allows the upper pressure limit for the
opts%rt_ir%cldstr_simple option to be changed for visible/IR cloud simulations. This setting applies to all channels
being simulated. Both options should be used with caution: the cldstr_simple option is primarily intended for mid- to
upper-tropospheric channels. (See Annex O for a description of these options in the options structure).
New RTTOV-SCATT option opts_scatt%cc_threshold which specifies the effective cloud fraction threshold below
which scattering is ignored. Previously this was hard-coded at 0.05, and this remains the default. See section 8.7.
Bug fixes
The bugs in RTTOV v12.2 listed on the website have been fixed:
https://www.nwpsaf.eu/site/software/rttov/rttov-v12/code-updates/
15
Version : 1.3
Date : 05/03/2019
4.2. Changes from RTTOV v12.1

This section provides details of the differences between RTTOV v12.1 and v12.2. The vast majority of users will be
able to switch v12.2 for v12.1 in their application without having to make any code changes and the outputs will be the
same aside from certain specific differences which are described below. Users of HTFRTC will have to update their
code for v12.2 as the interface for HTFRTC calls has changed. RTTOV v12.2 provides a number of new options and
capabilities: these are summarised here and described more fully elsewhere in the user guide.
Main new/updated features in v12.2
Clear-sky/general simulation updates:

 New option for solar sea surface BRDF model (section 7.6).
 New options for MW cloud liquid water (CLW) absorption calculations: new permittivity parameterisations
and new option to perform calculations on user levels rather than coefficient levels (section 7.3).
 MW optical depth coefficient files now contain band correction coefficients.
RTTOV-SCATT (section 8.7):

 New option to carry out calculations on radiances instead of brightness temperatures.
 New optional output structure containing information required to perform all-sky emissivity retrievals. A new
subroutine is included which carries out the retrieval calculation.
 New options for CLW permittivity in “Mietable” generation code (same options as for CLW absorption) and
updated Mietable files using new recommended option.
 New options for non-spherical particle optical properties in “Mietable” generation code: inclusion of the ARTS
single scattering database including many options for representing ice, snow, hail and graupel.
 New Mietable file for MetopSG ICI.
 New OpenMP parallel interface to RTTOV-SCATT.
Visible/IR scattering simulations:

 New option to use MFASIS fast visible cloud scattering parameterisation (section 8.5).
 New cloud liquid water optical properties parameterised in terms of particle effective diameter (section 8.5).
 Explicit cloud/aerosol optical properties are optionally active variables in the TL/AD/K models (section 7.9).
 New scaer files containing optical properties for 9 CAMS aerosol species (section 8.6).
Principal Components simulations:

 New PC-RTTOV coefficients enabling all variable trace gases (except SO2) and aerosol simulations using
OPAC aerosol properties (section 8.8).
 The HTFRTC interface has been rewritten to improve performance (section 8.9).
Technical updates:
 Python wrapper is now compatible with both Python 2 and 3. The wrapper has also been updated to enable
MFASIS simulations and to work with the new CAMS aerosol species (see docs/rttov-wrapper.pdf).
 The RTTOV GUI has been updated to enable MFASIS simulations and to work with the new CAMS aerosol
species (see docs/rttov_gui_v12.pdf)
 Build system updated to optionally allow compilation against an external LAPACK library (section 5.2).
Additional changes since v12.1
Changes in behaviour and interfaces
In order to enable the TL/AD/K of the explicit aerosol/cloud optical properties, new optional arguments were added to
the rttov_tl, rttov_ad, rttov_k, rttov_parallel_tl, rttov_parallel_ad, rttov_parallel_k, rttov_alloc_tl,
rttov_alloc_ad, and rttov_alloc_k subroutines. The vast majority of user code will be unaffected, but it is possible that
you will have to modify your code to specify optional arguments to these subroutines by name.
16
Version : 1.3
Date : 05/03/2019
Existing user code for RTTOV v12.1 using the object oriented wrapper (the “Rttov” and similar C++ classes or the
pyrttov package) will work with v12.2. However to support the new capabilities the underlying Fortran interface has
changed so user code which calls the wrapper interface subroutines directly will need to be updated for v12.2.
There is a change in behaviour related to the TL/AD/K of surface emissivity and BRDF when physical
emissivity/BRDF models are used. This affects the TL/AD/K when using FASTEM (any version, all surface types),
TESSEM2 (sea surfaces), IREMIS (sea surfaces), the solar BRDF model (sea surfaces). In v12.1 any non-zero
emissivity or BRDF TL input was used in preference to the TL calculated by each of these models from the input
profiles_tl. RTTOV v12.2 always computes the TL interally in these cases and ignores any input emissivity_tl and
reflectance_tl. In v12.1 RTTOV calculated emissivity and reflectance AD/K outputs for these cases. In v12.2 the
emissivity_ad/k and reflectance_ad/k outputs are always zero for these cases representing the fact that RTTOV
computed the surface emissivity/BRDF from the input profile variables. This does not affect other cases, for example
when calcemis/calcrefl are false or when using ISEM.
New outputs and options
There are new outputs in the rttov_radiance structure. The radiance%plane_parallel logical is set to true if RTTOV
used a strict plane-parallel geometry (true for example for DOM or MFASIS simulations). The radiance%quality(:)
integer array has one element per simulated radiance and is used as a bit-mask to flag possible issues with radiances
such as when optical depth regression limits are exceeded (see section 7.8). There is a new subroutine
rttov_print_radiance_quality which prints a human-readable interpretation of a given quality output (see Annex N).
In previous versions of RTTOV the input surface elevation (profiles(:)%elevation) was associated with the pressure
level equal to or immediately below the surface pressure (profiles(:)%s2m%p). The geopotential height of each
pressure level is computed relative to this fixed point. There is a new option opts%config%fix_hgpl which, if true,
associates the surface elevation with the 2m pressure regardless of whether or not it lies on a pressure level. Strictly
speaking this is more correct and it eliminates certain very small discontinuities in forward model radiances as the 2m
pressure varies. The impact on forward model radiances is small (up to ~0.01K, usually much smaller).
RTTOV-SCATT
It is recommended to use the new opts_scatt%lradiance option to carry out computations in radiance. The RTTOV-
SCATT code has also been updated to apply band corrections.
The lgradp switch to make pressure an active TL/AD/K variable is now an option for RTTOV-SCATT.
There are new subroutines rttov_print_opts_scatt and rttov_print_cld_profile which can be used to print out the
contents of the opts_scatt and RTTOV-SCATT cld_profiles structures (similar to the existing subroutines for the
RTTOV options and profiles structures). These may be useful for debugging (see Annex N).
Principal Components simulations
The new PC-RTTOV coefficient files must be used with the latest associated rtcoef file. In fact for IASI
(rtcoef_metop_2_iasi_pcrttov_compat.H5) this is the same as the the one released with RTTOV v12.1 except that the
PC compatibility version has been updated. RTTOV performs checks at runtime to ensure the PC coefficients are
compatible with the options.
The interface to HTFRTC has been rewritten for RTTOV v12.2. The HTFRTC code is now included within the
RTTOV source so there is no need to compile the HTFRTC model separately. There is a new subroutine
rttov_read_coefs_htfrtc which is used to read HTFRTC data files before calling rttov_direct or rttov_k. Users of
HFRTC should read section 8.9 which gives full details of the new implementation.
Bug fixes
The bugs in RTTOV v12.1 listed on the website have been fixed:
https://www.nwpsaf.eu/site/software/rttov/rttov-v12/code-updates/
17
Version : 1.3
Date : 05/03/2019
4.3. Changes from RTTOV v11

This section provides details of the differences in the user interface between RTTOV v11 and RTTOV v12.1. In
particular, this section should be useful if you wish to replace v11 with v12 in your application as it lists the changes to
the user-interface. Many of the changes are superficial and have been made to improve the consistency in the interface
to RTTOV.
Coefficient files
The format of the optical depth (“rtcoef”) coefficient files has been updated for v12. Executables are available to
convert between v10/v11 format and v12 format files (see Annex A). The visible/IR aerosol and cloud scattering
coefficient files have also changed for v12. The Mietable files for RTTOV-SCATT remain the same as those for v11.
Note that in previous versions of RTTOV the optical depth coefficient files contained the input profile “regression
limits”: by default RTTOV outputs warnings if the input profiles exceed these limits. The limits are calculated from the
strict minimum/maximum envelope of the diverse training profile set by applying a stretching factor (+/-10% for
temperature max/min respectively, and +/-20% for each gas max/min). The RTTOV v12 coefficient files contain the
strict min/max envelopes: the regression limits that are applied within RTTOV are calculated when the coefficients are
read in and the same stretching factors are applied as in v11. As such there is no change in behaviour for RTTOV v12
compared to v11 in this respect, but this is one of the differences between the v11 and v12 coefficient files.
Changes to RTTOV derived type names
Some derived types have been renamed to improve consistency:

 profile_type -> rttov_profile (input profile data for RTTOV)
 radiance_type -> rttov_radiance (output radiance data from RTTOV)
 radiance2_type -> rttov_radiance2 (additional output radiance data from RTTOV)
 transmission_type -> rttov_transmission (output transmittance data from RTTOV)
 profile_cloud_type -> rttov_profile_cloud (input cloud profile data for RTTOV-SCATT)
Changes to the rttov_options structure
The opts%rt_mw%do_lambertian and opts%rt_ir%do_lambertian options have been removed. The

opts%rt_all%do_lambertian flag applies to all sensor types and should be used instead. See below regarding the
change in the behaviour of this option.
The default FASTEM version is now 6. The value in opts%rt_mw%fastem_version must be a valid FASTEM version
number (1-6) or zero for the new TESSEM2 model.
The default IR sea surface emissivity model is the new physically-based model “IREMIS” which depends on wind
speed and skin temperature as well as zenith angle. All visible/IR coefficient files contain coefficients for this model
(SSU and PMR are the only exceptions). The older ISEM model is still available. The model is specified in
opts%rt_ir%ir_sea_emis_model: 1=>ISEM, 2=>IREMIS.
There are various new options related to the new multiple-scattering code (see sections 8.5 and 8.6 and Annex O): the
default treatment for scattering of solar radiation is the new multiple scattering solver.
Changes to the rttov_profile structure
The profiles(:)%gas_units “compatibility-mode” option (zero) has been removed. A value of zero or less now
corresponds to ppmv over dry air. The values for kg/kg over moist air (one) and ppmv over moist air (two) remain the
same as for v11.3. The default is now ppmv over moist air (two).
The profiles(:)%snow_frac variable has been moved to profiles(:)%skin%snow_fraction. As in v11.3 this is only
used by the IR emissivity atlases. The profiles(:)%soil_moisture profile variable has been moved to
profiles(:)%skin%soil_moisture, but is still unused.
18
Version : 1.3
Date : 05/03/2019
There is a new option to select the units of input cloud/aerosol profiles for visible/IR simulations. This is specified in
the logical flag profiles(:)%mmr_cldaer. By default this is true so that the units for clouds and aerosols are kg/kg.
Setting the flag to false reverts to the units used in previous RTTOV versions (g/m3 for clouds, cm-3 for aerosols). See
sections 8.5 and 8.6.
The profiles(:)%ish variable has been renamed profiles(:)%ice_scheme and there are just two valid values now: 1 =>
use SSEC database properties for ice cloud (new in v12); 2 => use 2014 Baran ice cloud parameterisation (same as v11
for IR channels, but now also extended to visible wavelengths). See section 8.5. The old Hexagonal/Aggregate and
Baran 2013 parameterisations have been removed.
Changes to rttov_opt_param structure (visible/IR scattering optical properties)
The index order in the explicit optical property arrays for VIS/IR scattering has been reversed for efficiency. These
structures also have new members related to the new scattering options (see section 8.5).
Changes to subroutine interfaces
The rttov_read_coefs subroutine has a new final optional argument named path which can be used to specify the
directory containing the coefficient files when you use the instrument argument to this subroutine to specify the
coefficient file via the platform and sensor IDs. Previously the coefficient file had to be in the current directory.
Previously some allocation subroutines took nlayers as an argument and some took nlevels. Now all affected RTTOV
allocation subroutines take nlevels as an argument. This changes the interface to: rttov_alloc_rad,
rttov_alloc_transmission and rttov_alloc_opt_param. This doesn't change the size of any of the RTTOV input or
output arrays. See Annex D.
The rttov_alloc_opt_param subroutine takes an additional nmom argument specifying the number of Legendre
coefficients for the phase functions (see Annex D and sections 8.5 and 8.6).
The rttov_alloc_direct, rttov_alloc_tl, rttov_alloc_ad, and rttov_alloc_k subroutines take additional optional
arguments to specify the number of Legendre coefficients for explicit aerosol and cloud optical property structures (see
Annex D).
The radiancedata and radiancedata2 arguments to rttov_direct, rttov_tl, rttov_ad, rttov_k, the parallel interfaces,
rttov_alloc_direct, rttov_alloc_tl, rttov_alloc_ad and rttov_alloc_k have been renamed radiance and radiance2 (see
Annexes D, I, J, K and L).
The order of the arguments to the rttov_bpr_init and rttov_bpr_calc subroutines have changed so that the err
argument is now first to be consistent with other subroutines (see Annex E). The rttov_bpr_calc subroutine now
accepts an optional nthreads argument: if RTTOV is compiled with OpenMP this subroutine can exploit multiple
cores/CPUs.
The rttov_read_scattcoeffs and rttov_scatt_setupindex subroutines now take the top-level coefficient structure
(derived type rttov_coefs) as an argument like all other subroutines instead of the optical depth coefficient structure
(derived type rttov_coef). See Annexes C and D.
The rttov_read_scattcoeffs subroutine now takes opts_scatt (RTTOV-SCATT options structure) as an argument and
respects the setting of the opts_scatt%config%verbose option. This subroutine has a new optional argument file_coef
which is the path to an ASCII or binary Mietable file. This allows Mietables to be read in a similar manner to other
RTTOV coefficient files by specifying the filename explicitly. If this argument is omitted the Mietable reading works as
in v11 (the filename is constructed based on the supplied optical depth coefficients structure). In this latter case there is
now an optional path argument to specify an alternative directory for the Mietable file. See Annex C.
The rttov_alloc_scatt_prof subroutine now returns an error status (similar to other allocation subroutines) to indicate
whether the allocation was successful (see Annex C).
The subroutine interfaces to the land surface emissivity and BRDF atlases have also changed: these are described
below.
19
Version : 1.3
Date : 05/03/2019
List of subroutines with interface changes:

rttov_read_coefs rttov_bpr_calc rttov_read_scattcoeffs
rttov_alloc_rad rttov_direct rttov_scatt_setupindex
rttov_alloc_transmission rttov_tl rttov_alloc_scatt_prof
rttov_alloc_opt_param rttov_ad rttov_setup_emis_atlas
rttov_alloc_direct rttov_k rttov_get_emis
rttov_alloc_tl rttov_parallel_direct rttov_deallocate_emis_atlas
rttov_alloc_ad rttov_parallel_tl rttov_setup_brdf_atlas
rttov_alloc_k rttov_parallel_ad rttov_get_brdf
rttov_bpr_init rttov_parallel_k rttov_deallocate_brdf_atlas
Changes to emissivity and BRDF atlases
The TELSEM atlas has been replaced by the new TELSEM2 atlas which includes emissivities for climatological sea-ice
and may be used for all instruments supported by RTTOV (including MetopSG ICI). The atlas files are still in ASCII
format and TELSEM2 can be used without compiling RTTOV against the HDF5 library (unlike RTTOV v11). The
CNRM MW atlas has also been updated: data are now available for AMSU, MHS, SSMI/S and ATMS and the input
files are now in HDF5 format. For IR sensors the UWIRemis atlas remains available and the new CAMEL IR atlas has
been added.
The emissivity and BRDF atlases now work differently. You must declare an instance of derived type
rttov_emis_atlas_data for emissivity atlases (defined in mod_rttov_emis_atlas) or rttov_brdf_atlas_data for the
BRDF atlas (defined in mod_rttov_brdf_atlas) to hold the atlas data for a particular month/instrument. These data
structures are then passed into the RTTOV atlas interface subroutines. This enables data from different atlases for
different months/instruments to be initialised and used concurrently.
As a result of this update the interfaces to the atlas interface subroutines have changed significantly. The atlas “version”
numbers have been replaced by atlas “ID” numbers: for the IR atlases, 1 => UWIRemis, 2 => CAMEL atlas; for the
MW atlases, 1 => TELSEM2, 2 => CNRM MW atlas. See Annexes F and G and section 7.5 and the example code in
src/test/example_atlas_fwd.F90.
Other changes to behaviour and outputs
When calling the TL model for visible and IR sensors, if the input emissivity_tl(:)%emis_in or
reflectance_tl(:)%refl_in is non-zero then this input perturbation will be used instead of any calculated
emissivity/reflectance perturbation that would result from RTTOV's internal emissivity/reflectance calculations. This
makes the behaviour for VIS/IR sensors consistent with the FASTEM TL for MW sensors which has behaved this way
in previous versions of RTTOV. NB In RTTOV v12.2 this behaviour was changed: see above for details.
For solar radiation over sea surfaces where calcrefl(:) is true RTTOV provides a calculation of the BRDF for the direct
surface-reflected solar beam using a sunglint model. In RTTOV v11 this same BRDF was used when calculating the
downward-scattered surface-reflected radiation contribution. However this sunglint BRDF is not appropriate for this
radiance contribution: it is too large within the sunglint region and is too small (almost zero) outside the sunglint region.
In RTTOV v12, for visible/near-IR channels the BRDF for scattered radiation is taken from a fixed reflectance
spectrum for water (the same one used by the BRDF atlas for sea surfaces) and for channels with both thermal emission
and solar contributions this BRDF is computed as (1-emissivity)/π which is consistent with the treatment of
downwelling atmospheric emission in these channels. In addition, for visible/near-IR channels, the BRDF derived from
the water reflectance spectrum is added to the sunglint BRDF for the direct surface-reflected beam to mitigate a dark
bias observed in simulated reflectances outside the sunglint region. This changes the radiances from those in RTTOV
v11 for all visible/near-IR channels where the sunglint model is used and for mixed thermal+solar channels with single-
scattering for clouds/aerosols where the sunglint model is used. This does not apply to the new DOM solver for which
the surface is treated as strictly Lambertian.
The opts%rt_all%do_lambertian option is not applicable with sea surface emissivity models and as such it is not
applied for channels where the IR or MW sea surface emissivity models are being used.
When the interpolator is used, any profile values obtained by extrapolating the input profile up to the top of the
coefficient profile (in most cases 0.005hPa) are always clipped to the regression limits for the purposes of the optical
20
Version : 1.3
Date : 05/03/2019
depth regression. Previously the default constant-value extrapolation could result in values beyond the range of the
training profiles.
The internal calculation for the local path angle along the surface-satellite path has been modified: previously the
calculated path angle at the surface was not identical to the user-specified zenith angle and, for solar simulations, the
surface-satellite and surface-sun path angles were not identical when the satellite and solar zenith angles were identical.
These inconsistencies have been removed. In most cases the impact is of the order of hundredths of a Kelvin, but in
some cases at the highest zenith angles differences of a few tenths of a Kelvin have been observed.
Some other internal changes have been made which have a small impact (a few hundredths of a Kelvin or less) on
simulated radiances:
 Values of some physical constants have been updated with latest values from NIST
(http://physics.nist.gov/cuu/Constants/)
 In previous versions of RTTOV the 2m water vapour variable (profiles(:)%s2m%q) was never used with v9
predictor coefficient files. The 2m q variable is now used for any coefficient file if the opts%rt_all%use_q2m
option is true.
 The method of calculating geopotential height has been modified to avoid numerical problems if the input
profile has very thick layers near the top of the atmosphere.
Additional capabilities of RTTOV v12 over RTTOV v11
Visible/IR scattering updates (sections 8.5 and 8.6):

 Discrete Ordinates Method (DOM) multiple-scattering solver for clouds and aerosols: there are independent
options available for the treatment of scattering for thermal emission and solar radiation.
 Ice optical properties from the SSEC database replace the old Hexagonal/Aggregate ice shapes.
 Baran2014 ice parameterisation extended to visible/near-IR wavelengths.
 New option to supply cloud and aerosol inputs as mass mixing ratios.
Surface emissivity and BRDF updates (sections 7.5 and 8.4 and Annexes F and G):
 New MW sea surface emissivity model, TESSEM2, intended for use with ICI.
 New IR sea surface emissivity model including wind speed and Tskin dependency.
 Updated TELSEM MW land surface and sea-ice emissivity atlas (TELSEM2).
 Updated CNRM MW land surface emissivity atlas.
 New IR land surface emissivity atlas (CAMEL).
 More flexible interface to land surface emissivity and BRDF atlases allowing data from different atlases and for
different months and instruments to be loaded concurrently.
Other science updates:

 New coefficients available allowing SO2 as an optional trace gas (section 8.14).
 Updated NLTE correction (section 8.10).
 Updated PC-RTTOV coefficients including the NLTE correction (section 8.8).
 Capability to call the HTFRTC PC-based RT model through a PC-RTTOV-like interface (see docs/rttov-
htfrtc_interface.pdf).
Technical updates:
 The Python/C++ wrapper now exploits the more flexible interface to the emissivity/BRDF atlases. The wrapper
also now supports calls to RTTOV-SCATT and visible/IR scattering simulations where the optical properties are
explicitly provided to RTTOV (see docs/rttov-wrapper.pdf).
 The GUI has been updated to support the new capabilities of RTTOV v12 (see docs/rttov_gui_v12.pdf).
 Improved consistency in subroutine interfaces and derived type naming (see above for the differences to v11).
 Optimisation of gas optical depth calculation in particular for v9 predictors with all trace gases.
 Doxygen markup has been added to document user-level subroutines and data types (see Section 6).
21
Version : 1.3
Date : 05/03/2019
5. FORTRAN-90 UNIX/LINUX installation and testing

RTTOV v12 is designed for UNIX/Linux systems. The software is now successfully tested on Intel, IBM, Cray and
Apple Mac systems and for a range of Fortran 90 compilers listed in the report on the RTTOV platforms/compilers
tested.
The following system components are needed before running RTTOV v12:
 UNIX or Linux operating system
 Fortran 90 compiler
 Perl v5.6 or later
 make utilities
 gzip and gunzip
 About 100 Mbytes of free disk space is the minimum required (although more is necessary for hyperspectral IR
sounder coefficient files and atlas data files).
 Memory requirements are strongly dependent on the instrument being simulated and the kind of simulations being
performed: the simplest clear-sky direct model simulations require of the order of 10Mbytes.
 It is recommended to compile RTTOV against the HDF5 library (v1.8.8 or higher) so that all RTTOV features are
available (see section 5.2 below).
 The Python interface and the RTTOV GUI require that f2py is installed. The GUI has additional requirements: see
the GUI user guide in the docs/ directory. The Python wrapper is compatible with Python 2 and 3, but Python 2 is
recommended for the GUI (see Known Issues below).
 HTFRTC requires RTTOV to be compiled against the NetCDF v4 library.
The RTTOV v12 package is available as a compressed tar file named rttov123.tar.gz which should be copied to your
‘top’ RTTOV directory (e.g. ~user/rttov12) and extracted. This will create a number of sub-directories (see below).
RTTOV v12 will not work with older versions of some compilers. The following list gives the versions of several
common compilers known to be compatible:
 gfortran – v4.4.6 and later
 ifort – v16.0.1 and later
 pgf90 – v16.10, v18.7
 NAG – v6.1, v6.2
 Cray Fortran – v8.3.4
 IBM – xlf95 v13.1 (RTTOV v12 beta tested)
Known compiler issues:

 The RTTOV GUI can be compiled and run with Python 3, but issues such as frozen windows have been
observed when running the GUI under Python 3 and so Python 2 is strongly recommended for the GUI. (Note
that the Python wrapper is fully compatible with Python 2 and 3).
 NAG v6.0: cannot read nested derived types in namelists (compiler bug) and so cannot run the test suite. The
test and example programs outside of the test harness (rttov_test.pl) run fine which suggests the core of
RTTOV is compiled successfully and can be used with this compiler. NAG v6.1 does not have this problem.
 RTTOV v12 is known to NOT work with:
o gfortran v4.4.5 and earlier
o pgf90 v11.7
22
Version : 1.3
Date : 05/03/2019
5.1 Unpacking the code

This is achieved using the command:
$ tar xvf rttov123.tar.gz
The following subdirectories are created:

brdf_data/ BRDF atlas data (data must be downloaded from web site)
build/ Scripts used in building RTTOV and files containing flags for various
compilers/architectures
data/ Various ancillary data files
docs/ Documentation
emis_data/ Emissivity atlas data (data must be downloaded from web site)
gui/ RTTOV GUI source code
src/ The RTTOV source code
rtcoef_rttov12/ RTTOV v12 coefficient files (see below)
rttov_test/ test scripts, input profiles for tests, and reference output for tests
wrapper/ example code calling RTTOV from Python and C++
As described in section 3 and Table 4 there are three predictor versions on which RTTOV coefficients are based. A
comprehensive list of RTTOV v12 coefficients is available via the RTTOV v12 web page which provides details of
which features are supported by each file:
http://nwpsaf.eu/site/software/rttov/download/coefficients/coefficient-download/
You should refer to this web page to determine the most appropriate coefficient file for your application.
The rtcoef_rttov12/ directory contains sub-directories for each kind of coefficient file:
rttov7pred54L/ v7 predictor files on 54 levels (most IR/MW optical depth predictor coefficient files)
rttov7pred101L/ v7 predictor files on 101 levels (hi-res IR sounder files)
rttov8pred51L/ v8 predictor files on 51 levels (SSU only, allows variable O 3 and CO2)
rttov8pred54L/ v8 predictor files on 54 levels (allows variable O3 and CO2)
rttov9pred54L/ v9 predictor files on 54 levels (VIS/NIR/IR sensors, solar calculations, variable O3 and CO2)
rttov9pred101L/ v9 predictor files on 101 levels (hi-res IR sounder files allowing more variable gases)
cldaer_ir/ cloud and aerosol scattering coefficient files for coefficient files containing only IR channels
cldaer_visir/ cloud and aerosol scattering coefficient files for v9 predictor coefficient files containing
visible and IR channels
mfasis_lut/ MFASIS look-up table (LUT) files
mietable/ MW scattering coefficient files
pc/ Principal Components (PC-RTTOV) coefficient files
htfrtc/ HTFRTC coefficient files
In addition to the coefficient files listed in Table 4 there are:

 Zeeman coefficient files for SSMI/S (by default expected in the rttov7pred54L/ directory even though they
are based on different numbers of levels)
 PC coefficients for IASI, AIRS and IASI-NG. The PC-RTTOV coefficients must be used with the v9
predictor 101 level optical depth coefficient files with which they were trained. These are in the
rttov9pred101L/ directory and contain “pcrttov_compat” in the filename.
 Aerosol and cloud scattering coefficients for most visible/IR sensors; note that the channels in the scattering
coefficients file must match the channels in the rtcoef and so there are separate scattering files for use with IR-
only coefficient files and visible+IR coefficient files.
 MFASIS LUT files for use with the MFASIS fast visible scattering model.
 Scattering coefficients for most MW instruments.
 Input files for the HTFRTC model.
23
Version : 1.3
Date : 05/03/2019
The RTTOV distribution includes all of the VIS/IR and MW coefficient files supported by RTTOV at the time of the
release. The hi-res IR sounder coefficient files, scattering, PC-RTTOV, and HTFRTC files are not included in the
distribution and nor are the emissivity and BRDF atlas datasets. These are all available from the RTTOV web site.
For the purposes of running the test suite you should ensure coefficient files are placed in the appropriate directories.
An interactive script rttov_coef_download.sh is available in the rtcoef_rttov12/ directory which can be used to
download any or all coefficients into the standard locations in the coefficients directory. Note that you only need to
download the coefficients required for the simulations you wish to carry out. For example, there is no need to download
any hi-res IR sounder coefficients unless you want to run simulations for an instrument of that kind.
Compatibility with v10/v11:

 The format of the rtcoef optical depth coefficient files has changed for v12 and as such v10/v11-format files
cannot be used directly with v12. Executables to convert coefficients between v10/v11-format and v12-format
are described in Annex A.
 The IR scattering coefficient files are not backwards-compatible and cannot be converted for use with v10/v11.
 The MW scattering coefficient “mietable” files remain the same as for v10/v11.
 The format of PC-RTTOV coefficient files has not changed since RTTOV v11. You can use RTTOV v11 PC-
RTTOV coefficients with RTTOV v12. It is important to use the corresponding v9 predictor 101 level optical
depth coefficient files with which the PC coefficients were trained: in RTTOV v12 these have
“pcrttov_compat” in their filename and, at the time of release, the v12 AIRS and IASI-NG files are identical
to the corresponding RTTOV v11 coefficient files. The v12 IASI rtcoef file for PC-RTTOV includes the new
NLTE coefficients, but aside from that is identical to the corresponding v11 file.
All coefficients are supplied in ASCII format except for hypersepctral IR sounders and MFASIS LUT files: due to the
large file sizes the HDF5 format is used. It is possible to convert coefficient files between ASCII, HDF5 and Fortran
unformatted (“binary”) formats, the latter two being more efficient (though note that the binary format is not portable
between systems). It is also possible to extract a subset of channels which reduces the file size and can improve
performance, particularly for hyperspectral sounders. The rttov_conv_coef.exe program performs these tasks and is
described in Annex A.
RTTOV-SCATT Mietable files may be converted to binary format using the rttov_asciii2bin_scattcoef.exe executable
(see Annex A).
HTFRTC files cannot be converted to alternative formats.
24
Version : 1.3
Date : 05/03/2019
5.2 Compiling the code

An interactive shell script is available to compile RTTOV v12 which asks some questions and then runs the necessary
commands to do the compilation:
$ cd src
$ ../build/rttov_compile.sh
There must be a file in the build/arch/ directory containing the compilation flags you wish to use. There are a number
of example files for various common compilers or you can create a new one: more details on this are given below in the
section “Creating an architecture configuration file”. In order to make use of multi-threaded execution via the
rttov_parallel_* routines RTTOV must be compiled with OpenMP. This involves supplying a suitable flag to an
appropriate compiler. There are compiler flag files in build/arch/ for compiling with OpenMP support with gfortran,
pgf90, ifort and NAG.
RTTOV may be compiled immediately without requiring any external libraries. However some features of RTTOV
have external dependencies:
Reading HDF5 coefficient files : requires the HDF5 library.

Emissivity*/BRDF atlases : require the HDF5 library.
RTTOV GUI : requires the HDF5 library.
Python interface and RTTOV GUI : require that f2py is installed.
HTFRTC : requires the NetCDF v4 library.
*The TELSEM2 MW emissivity atlas may be used without any external dependencies as the atlas data files are in ASCII
format, but the other atlases require the HDF5 library.
Compiling with the HDF5 library is recommended as all functionality is then available. Before compiling with HDF5
you must first edit the build/Makefile.local file with the location of the HDF5 library. This involves specifying the
path to the library installation in the “HDF5_PREFIX” variable and uncommenting one “FFLAGS_HDF5” definition
and one “LDFLAGS_HDF5” definition appropriate to your build of the library. Similarly, if you want to compile
against the NetCDF library (to use HTFRTC) or against an external LAPACK library, you must also edit the relevant
sections of build/Makefile.local with the details of these libraries before running the rttov_compile.sh script.
Once the code is compiled you will find bin/ and lib/ directories in your top-level RTTOV directory containing the
RTTOV binaries and libraries. One library is created for each subfolder within src/ and you should link all required
libraries in your application (at the very least librttov12_main and librttov12_coef_io – see section 6 for more
information). Tables 7 and 8 list all libraries and executables produced by the build process.
The file src/test/Makefile_examples is an example stand-alone Makefile for the example_*.F90 demonstration
programs. It has a section at the top which describes the variables that should be edited with paths appropriate for your
system. This is intended as a demonstration of how to link your own code against the RTTOV libraries: the
example_*.F90 executables are compiled by the RTTOV build process so it is not necessary to use this Makefile to
compile the example code.
Notes on compiling with the HDF5 library:

1. The HDF library must be built with the Fortran interface (see the HDF documentation).
2. In build/Makefile.local the FFLAGS_HDF5 variable defines the _RTTOV_HDF macro. It is important to
supply this macro to the compiler so that the sections of code which do HDF5 I/O are included in the
compilation. For most Linux-based Fortran compilers this is achieved by passing –D_RTTOV_HDF as seen in
the FFLAGS_HDF5 variable, but for XLF on AIX it is passed using –WF,–D_RTTOV_HDF.
3. Note that if you do NOT compile with the HDF5 library, you must NOT supply the _RTTOV_HDF macro to
the compiler (so the HDF5 lines in Makefile.local must be commented out).
4. If the code was previously compiled without HDF5 first then you should select a clean compilation if the
recompilation fails.
5. Before running RTTOV ensure the HDF5 library you compiled RTTOV against is in your
$LD_LIBRARY_PATH or equivalent.
25
Version : 1.3
Date : 05/03/2019
Compiling RTTOV manually
It is recommended to use the build/rttov_compile.sh script described in the previous section, but manual compilation
of RTTOV is possible. As noted above, if compiling with HDF5 you must first edit the build/Makefile.local file with
the location of the HDF5 library, and likewise for the NetCDF library.
RTTOV makes use of some LAPACK subroutines: the source code for these is included with RTTOV and by default
you do not have to worry about this. However if you wish instead to compile against an external library containing the
LAPACK subroutines you can specify this library in build/Makefile.local in a similar manner to the HDF5 library.
The general compilation procedure is then as follows:
$ cd src
$ ../build/Makefile.PL RTTOV_HDF=1 RTTOV_F2PY=1 RTTOV_USER_LAPACK=1
$ make ARCH=myarch INSTALLDIR=myinstalldir
The arguments in italics are optional.
The second step (running Makefile.PL to regenerate the RTTOV Makefiles) is not required if you are compiling
RTTOV for the first time "out-of-the-box" and you do not require either HDF5 or Python-related code to be compiled
and you do not want to compile against an external LAPACK library. However, if you are compiling RTTOV with the
HDF5 library or a LAPACK library or you want to compile the RTTOV GUI or the RTTOV Python interface, you must
run Makefile.PL with one or more of the arguments shown above:
RTTOV_HDF=1 - required if compiling RTTOV against the HDF5 library

RTTOV_F2PY=1 - required if compiling the RTTOV GUI or the RTTOV Python interface
RTTOV_USER_LAPACK=1 - required if compiling RTTOV against an external LAPACK library
The arguments to "make" are:
ARCH - this argument is optional: if omitted RTTOV is compiled with gfortran, otherwise "myarch" should correspond
to the name of one of the files in the build/arch/ directory. These files contain build flags for various common
compilers/platforms. You can add new ones: see below for details. Using *-openmp flags will enable multi-threaded
execution via the RTTOV parallel interfaces: this is generally recommended.
INSTALLDIR - by default the build process creates output directories (e.g. lib/ and bin/) in the top-level RTTOV
directory. You can optionally specify "myinstalldir" to be another path relative to the top-level RTTOV directory to
contain the lib/, bin/and other subdirectories (useful if compiling RTTOV using more than one set of compiler flags).
Some examples are given below, all run from within the src/ directory:
Compile RTTOV with gfortran without any external dependencies:

$ make
Compile all RTTOV code excluding the GUI and Python interface with gfortran compiler flags with the HDF5 library:
First edit build/Makefile.local with the location of your HDF5 installation.

$ ../build/Makefile.PL RTTOV_HDF=1
$ make
Compile all RTTOV code with gfortran-openmp compiler flags with the HDF5 library:
First edit build/Makefile.local with the location of your HDF5 installation.

$ ../build/Makefile.PL RTTOV_HDF=1 RTTOV_F2PY=1
$ make ARCH=gfortran-openmp
26
Version : 1.3
Date : 05/03/2019
Compilation: more details
This section is not essential reading, but contains some additional information about compilation which you may find
useful.
General notes
Programs should be compiled with the C-style preprocessor options enabled to make use of the #include statements for
subroutine declarations. Note for most compilers this implies you need .F90 as the file extension which is what is
provided. For users with HP compilers it may be necessary to convert the .F90 file extensions to .f90 for all the
routines.
Specifying an installation directory
As described above, by default the build process creates new directories (bin/, lib/ etc) in the top-level RTTOV
directory. It is possible to specify a subdirectory where the new directories will be placed by supplying the
INSTALLDIR argument to “make”. This feature is useful if compiling RTTOV with different compiler flags or with
different compilers. After compilation and testing the build directories can be moved to an arbitrary location (i.e.
outside the RTTOV directory).
Specifying external dependencies
The file build/Makefile.local is used to specify the locations of external libraries, in particular the HDF5 library. As
noted above, for the HDF code to be compiled requires the _RTTOV_HDF macro to be passed to the compiler. It is
equally important that if the HDF code is not required this macro is not supplied to the compiler.
The file contains templates for linking against the HDF5 library, the HTFRTC model (see the separate user guide for
information about this), and a LAPACK library. An example for DrHook is also included: note that in the case of
DrHook, the RTTOV source code includes yomhook.F90, a dummy routine, which must be removed from the
src/main/ directory if you want to run with DrHook enabled.
Regenerating the Makefiles
If the RTTOV code dependencies change for some reason or, if source files are added or removed from the src/
directory, or if you want to include (or remove) the HDF5 capability the Makefiles must be regenerated. This is easily
achieved as follows:
$ cd src/
$ ../build/Makefile.PL RTTOV_HDF=1 RTTOV_F2PY=1 RTTOV_USER_LAPACK=1
$ make ARCH=myarch INSTALLDIR=mydir clean
where RTTOV_HDF=1, RTTOV_F2PY=1 and RTTOV_USER_LAPACK=1 are as described above. It is important to

remember that if RTTOV_HDF=1 was supplied to Makefile.PL, then Makefile.local must supply the _RTTOV_HDF
macro to the compiler, and likewise, if RTTOV_HDF=1 is not supplied to Makefile.PL, Makefile.local must not
supply the macro.
It is good practice to do a “make clean” after running Makefile.PL to avoid problems when recompiling.
27
Version : 1.3
Date : 05/03/2019
Creating an architecture configuration file
If the required architecture is not included in the build/arch/ directory bundled with RTTOV or if you would like to
customise the installation of RTTOV it is possible to create a new configuration file. This configuration file must be
installed in the build/arch/ directory and define the following macros:
FC : the name of the Fortran 95 compiler.

FC77 : the name of the Fortran 77 compiler; this might be the Fortran 95 compiler, possibly with some special options.
CC: the name of the C compiler.
LDFLAGS_ARCH : specific flags to pass to the linker.
FFLAGS_ARCH : specific flags for the Fortran compiler.
CFLAGS_ARCH: specific flags for the C compiler.
AR : the command to create a library from object files.
NB The Fortran 77 and C compilers are used to compile specific source files. However, the RTTOV v12 software must
be compiled with a Fortran 95 compiler.
This configuration file may also define the following macros:
FFLAG_MOD: this is the flag used by the Fortran 95 compiler to locate module files; it defaults to -I, but it is possible to
override this setting.
CPP: the name of the pre-processor; defaults to cpp.
Specific flags for some RTTOV source files; defining FFLAGS_ARCH_a will force the build system to compile unit
a.F90 with these specific flags.
In order to use the GUI or the Python interface RTTOV must be compiled with f2py support. The following macros
should be defined:
F2PY: this defines the f2py command and specifies the Fortran compiler being used (see the f2py documentation for
relevant compiler names).
F2PYFLAGS_ARCH: compiler flags to pass to the F2PY compilation. This should specify the PIC (position
independent code) flag in the appropriate form for the relevant Fortran compiler.
F2PYLDFLAGS_ARCH: linker flags for the F2PY compilation.
The existing files in build/arch/ provide useful templates.
28
Version : 1.3
Date : 05/03/2019
5.3 Running the test suite

You should first navigate to the rttov_test/ directory which contains the relevant scripts and data for testing RTTOV.
Unless otherwise specified all files and directories referred to in this section may be found in this directory.
RTTOV has a comprehensive and flexible test suite based on the rttov_test.exe executable which is run via the
rttov_test.pl script. This allows most aspects of RTTOV to be configured and tested from the command-line and can
compare the simulated outputs with reference data. Some shell scripts are included which run the test suite for a range
of instruments, profiles and options. These are described in the “Core RTTOV testing” section below. The output from
these tests can be visualised using a graphical interface written in Python (rttov_test_plot.py) which is described in the
RTTOV test suite documentation in the docs/ directory.
In addition there are several stand-alone shell scripts which can be used to run the example_*.exe demonstration
programs and also the additional test and example executables for RTTOV-SCATT and the emissivity and BRDF
atlases. These are also described below.
Verifying the RTTOV build
NB Due to the way the Intel Fortran compiler manages memory, users compiling with ifort on Linux may need to
increase the stack size by executing the following command before all tests will run correctly:
$ ulimit –s unlimited
In addition, if running the PC-RTTOV K model with multiple threads under ifort, you may need to increase the
OpenMP stack size as well to allow tests to run:
$ export OMP_STACKSIZE=1000M
A number of shell scripts are provided which run the RTTOV test suite for various instruments to test particular aspects
of RTTOV. Some of the scripts require coefficient files to be downloaded from the website (for example for hi-res IR
sounders, visible/IR or MW scattering coefficients or PC-RTTOV coefficients). Note that the majority of hi-res IR
sounder tests defined in the test suite expect coefficient files in HDF5 format.
In order to verify your RTTOV installation you can run the following script without downloading any additional
coefficient files:
$ ./test_rttov12.sh ARCH=myarch BIN=bindir
The ARCH parameter should match the one used when you compiled RTTOV. The BIN parameter is optional. It is
only required if the INSTALLDIR parameter was supplied when compiling RTTOV i.e. if the location of bin/ is not in
the top-level RTTOV directory. If specified BIN must give the location of the directory containing binary executables
relative to the top-level RTTOV distribution directory (e.g. if you specified INSTALLDIR=install/gfortran when
building RTTOV then you should use BIN=install/gfortran/bin).
It is also possible to execute the test suite in a distributed computing environment (e.g. supercomputer) via a scheduler:
you can optionally supply, for example, SCHED_CMD=aprun. Each executable run by rttov_test.pl will be called via
aprun in this case. Additional arguments can be supplied using quotes: e.g. SCHED_CMD=”aprun --abc”.
The script above runs the RTTOV direct, TL, AD and K models for a range of instruments and compares the results to
the supplied reference data. The test suite reports whether each individual test was successful or not. There may be
cases where there are differences in the least significant digits between test output and the reference output due to
compiler-dependent rounding errors (especially in the Jacobian output from the K model): these will be reported as
differences, but are not cause for concern. The rttov_test/ directory contains several other shell scripts can be used to
test particular types of RTTOV simulations, but note that some will require you to download the relevant coefficient
files from the website:
29
Version : 1.3
Date : 05/03/2019
test_fwd.sh tests the forward model for a wide range of instruments

test_rttov12.sh tests the full code (direct/TL/AD/K) for a range of instruments
test_rttov12_hires.sh tests the full code for hi-res IR sounders
test_solar.sh tests visible/near-IR solar simulations
test_multi_instrument.sh tests RTTOV running for multiple instruments together
test_pc.sh tests the PC-RTTOV calculations
test_htfrtc.sh tests the HTFRTC calculations
You can run any of these scripts in exactly the same way as described above for test_rttov12.sh. Note that the tests
expect HDF5 format coefficient files for hyperspectral IR sounders. The script test_core.sh calls all of the above scripts
and this is run in the same way as the individual scripts. However it is not necessary to run all the scripts to verify your
RTTOV installation: calling test_rttov12.sh is sufficient.
There are also two scripts which test the coefficient I/O (which take the same arguments as the other scripts):
test_coef_io.sh tests the coefficient input/output code (this test has no reference data)
test_coef_io_hdf.sh tests the HDF5 coefficient input/output code (this test has no reference data)
It is not generally necessary to run these unless you need to verify that the coefficients are being read correctly on your
system. These require hyperspectral IR sounder coefficient files in ASCII format so you would usually have to convert
the HDF5 files to ASCII format – same filename, but with extension “.dat” – before calling these scripts (see
rttov_conv_coef.exe in Annex A).
A full description of the RTTOV v12 test suite may be found in docs/rttov-test.pdf (under the top-level RTTOV
directory). A brief overview is given here, but it is not necessary to read this to use RTTOV.
The tests.0/ directory contains data required to run the tests: for each instrument this defines profile data, the channel
and profile lists, specification of surface emissivity and reflectance, a reference to the RTTOV coefficients, and so on.
Note that a small number of tests defined in tests.0/ are intended for developers only and the relevant coefficient files
are not available (this mostly involves coefficient files that have historically been available, but are no longer
generated). It is certainly not necessary to run every test defined in tests.0/ and the tests which are run by the scripts
listed above use coefficients which are available in the package or on the website (though some files need to be
converted to ASCII format as noted above).
Test outputs for the myarch architecture are located in tests.1.myarch/ – these are created when the tests are run. Test
reference output created on the NWP SAF test platforms is held in directories with names ending in .2.
The rttov_test.exe binary executable created during the building of RTTOV (and located in the bin/ directory of the
build) is used to run one or more tests. It is controlled by the rttov_test.pl perl script. A typical test run involves a
command like:
$ ./rttov_test.pl ARCH=myarch BIN=bindir TEST_LIST=hirs/001,avhrr/001 DIRECT=1
The ARCH and BIN parameters are the same as described above. The TEST_LIST parameter provides a list of tests
defined in tests.0/ to run. In this case, only the direct code is being tested (DIRECT=1). The test suite documentation
provides a complete list of parameters which may be supplied to rttov_test.pl which allow almost all aspects of
RTTOV to be tested. The full list of parameters may be listed by typing:
$ ./rttov_test.pl ARCH=myarch HELP=1
30
Version : 1.3
Date : 05/03/2019
Running examples of code calling RTTOV v12
Several examples of running the RTTOV forward model and an example of calling the K model are provided which are
intended to form a basis for your own applications. They are all in the src/test/ directory:
 example_fwd.F90 – simple example for clear-sky simulations

 example_k.F90 – simple example calling K model for clear-sky simulations.
 example_atlas_fwd.F90 – same as example_fwd.F90, but demonstrates use of emissivity and BRDF atlases
 example_cld_file_fwd.F90 – visible/IR cloud scattering simulations using a cloud coefficient file.
 example_cld_param_fwd.F90 – visible/IR cloud scattering simulations by passing the scattering parameters into
RTTOV explicitly.
 example_cld_mfasis_fwd.F90 – visible MFASIS cloud scattering simulations.
 example_aer_file_fwd.F90 – visible/IR aerosol scattering simulations using a aerosol coefficient file.
 example_aer_param_fwd.F90 – visible/IR aerosol scattering simulations by passing the scattering parameters into
RTTOV explicitly.
 example_rttovscatt_fwd.F90 – demonstrates calling RTTOV-SCATT
 example_pc_fwd.F90 – demonstrates calling PC-RTTOV
 example_htfrtc_fwd.F90 – demonstrates calling HTFRTC
Each of these programs may be run via a shell script in the rttov_test/ directory with the name run_example_*.sh
corresponding to the executable name. Near the top of each script is a small section where inputs may be configured
such as the coefficient file and its location and the name of the input file(s) for profile data. The scripts may be run by
typing (for example):
$ ./run_example_fwd.sh ARCH=myarch BIN=bindir
The ARCH and BIN arguments are described above in the “Verifying the RTTOV build” section. Test reference
outputs are in test_example.2/. Input files for the script are in test_example.1/, and this is also where the test outputs
are written. The outputs consist of files named output_example_*.dat.myarch and diff files named
diff_example_*.myarch showing the differences between the test outputs and the reference outputs. The diff files
should typically have zero size. In some cases they might show differences in the least significant digits which result
from compiler-dependent factors and are not cause for concern.
RTTOV-SCATT testing and example code
The test_rttovscatt.sh shell script may be used to verify the RTTOV-SCATT code. You may need to edit the first few
lines of this script to specify the location of the RTTOV coefficient files (by default assumed to be in
rtcoef_rttov12/rttov7pred54L/ and rtcoef_rttov12/mietable/). The script may then be run by typing:
$ ./test_rttovscatt.sh ARCH=myarch BIN=bindir
Test reference output is in test_rttovscatt.2/. Input files for the script are in the test_rttovscatt.1/ directory, and this is
also where the test output is written. The output consists of files named output.NN.rttov_scatt.myarch and
diff.NN.myarch (where NN is 01, 02, etc), the latter being diff files showing differences compared to the test reference
data. The script will exit cleanly if no internal errors are found. The diff files should typically have zero size if no errors
occurred.
There is also an example program mw_scatt/example_rttovscatt.F90 demonstrating how to perform direct and
Jacobian calculations with RTTOV-SCATT. Once test_rttovscatt.sh has been run, the required links to coefficient files
are set up within test_rttovscatt.1/. You may then call example_rttovscatt.exe (located in bin/) from this directory to
run the example code. There is no reference output for this example program.
31
Version : 1.3
Date : 05/03/2019
Emissivity and BRDF atlas testing
To use the atlases RTTOV must be compiled with HDF5 (see section 5.2). The only exception is the TELSEM2 atlas as
the data files for this atlas are in ASCII format. There is one test script per atlas:
 test_uwiremis_atlas.sh UWIRemis IR emissivity atlas

 test_camel_atlas.sh CAMEL 2007 IR emissivity atlas
 test_camel_clim_atlas.sh CAMEL climatology IR emissivity atlas
 test_telsem2_atlas.sh TELSEM2 MW atlas and interpolator
 test_cnrm_mw_atlas.sh CNRM MW atlas
 test_brdf_atlas.sh BRDF atlas
These test scripts each run a test program for the respective atlas. The test programs initialise the atlas, return
emissivity/BRDF values for a series of profiles/locations and then deallocate the atlas. The emissivities/BRDFs are
written to an output file. You may need to edit the first few lines of each script to specify the location of the RTTOV
coefficient files (by default assumed to be in rtcoef_rttov12/rttov7pred54L/), and the location of the emissivity and
BRDF atlas data files (by default assumed to be in emis_data/ and brdf_data/). Note that the IR emissivity atlas tests
require the all of the IR atlas files to be downloaded (including the covariance files and the angular correction files).
The scripts are all run in the same way, for example:
$ ./test_uwiremis_atlas.sh ARCH=myarch BIN=bindir
For the emissivity atlases test reference output is in test_emis_atlas.2/. Input files for the scripts are in the
test_emis_atlas.1/ directory, and this is also where the test output is written. The output consists of files named
output_*_atlas.NN.myarch according to the name of the atlas where NN is 01, 02, etc. The scripts also write diff files
named diff_*_atlas.NN.myarch showing the difference between the test output and the reference output. The
difference files should have zero size.
Similarly, the BRDF atlas test reference output is in test_brdf_atlas.2/. Input files for the scripts are in the
test_brdf_atlas.1/ directory, and this is also where the test output is written. The output consists of a file named
output_brdf_atlas.1.myarch. The script also writes a diff file named diff_brdf_atlas.1.myarch showing the
difference between the test output and the reference output. The difference files should have zero size.
32
Version : 1.3
Date : 05/03/2019
6. RTTOV types, subroutines, libararies and executables

This section summarises the output from the build process and the derived types and subroutines intended for your own
programs.
The build process produces a library for every directory within src/ that is included in the build target provided to
make. Table 7 lists all libraries created in the lib/ directory by the compilation and the associated user-level subroutines
contained therein. All user-level subroutine interfaces are detailed in the Annexes.
Table 6 provides a list of derived types (structures) which you may use in your application. All types are defined in the
src/main/rttov_types.F90 module. Annex O provides details of all the derived types. The two exceptions to this are
rttov_emis_atlas_data which is stored in src/emis_atlas/mod_rttov_emis_atlas.F90 and rttov_brdf_atlas_data
which is stored in src/brdf_atlas/mod_rttov_brdf_atlas.F90. These are used to hold data for the land surface atlases.
Table 8 gives a list of all executables produced by the build process in bin/ and their purpose. Most test executables are
intended to be called via the Perl and shell scripts found in rttov_test/. See section 5.3 for more information on calling
test programs. Aside from the test programs, the most commonly-used executables are the coefficient conversion tools
(see Annex A).
When calling RTTOV from Python the file rttov_wrapper_f2py.so must be in your current directory or in your
$PYTHONPATH. Similarly, when using the GUI the file rttov_gui_f2py.so must be in the current directory or
$PYTHONPATH. These files are found in the lib/ directory after compilation. See the separate user guides for the
wrapper and the GUI in the docs/ directory.
6.1. Linking your code against RTTOV

When linking against RTTOV you only need to link against the RTTOV libraries which contain subroutines that you
have used (see Table 7). The order in which the libraries are linked is important: if library A depends on library B then A
must appear before B in the linking step. You can link the RTTOV libraries as follows:
-lrttov12_wrapper -lrttov12_mw_scatt -lrttov12_brdf_atlas

-lrttov12_emis_atlas -lrttov12_other -lrttov12_parallel
-lrttov12_coef_io -lrttov12_hdf -lrttov12_main
-lhdf5hl_fortran -lhdf5_hl -lhdf5_fortran -lhdf5
The file src/test/Makefile_examples is an example Makefile for the example_*.F90 executables which may be used as
a template for compiling your own code which calls RTTOV. As noted in section 5.2 the example executables are built
when RTTOV is compiled: this example Makefile is only for demonstration purposes.
33
Version : 1.3
Date : 05/03/2019
6.2. Doxygen documentation

Doxygen markup has been added to the user-level subroutines and data types. If you have Doxygen installed you can
generate the documentation by running the following from the top-level RTTOV directory:
$ doxygen docs/doxygen_config_user
The resulting documentation can be found in docs/doxygen_doc_user/. In particular HTML documentation can be
found in docs/doxygen_doc_user/html/index.html. This may be helpful as a reference when reading sections 7 and 8:
section 7 describes the technical aspects of calling RTTOV including subroutine calls while section 8 describes the
various kinds of simulation RTTOV can carry out.
Type name Purpose

rttov_options RTTOV options structure to configure simulations.
rttov_options_scatt Limited set of options for configuring RTTOV-SCATT.
rttov_coefs Coefficients structure for optical depth, cloud, aerosol and PC coefficients.
rttov_scatt_coef Coefficients structure for RTTOV-SCATT Mie table coefficients.
rttov_chanprof Define channel/profile indexes to simulate.
rttov_profile Input profile.
rttov_profile_cloud Input cloudy profile for RTTOV-SCATT.
rttov_opt_param Explicit aerosol/cloud optical parameter profiles.
rttov_emissivity Input/output surface emissivity values.
rttov_reflectance Input/output surface BRDF values and cloud-top BRDF for simple cloud scheme.
rttov_traj Holds various internal variables for repeated RTTOV calls (optional).
rttov_radiance Calculated radiances.
rttov_radiance2 Secondary direct model radiances (optional).
rttov_transmission Calculated transmittances.
rttov_scatt_emis_retrieval_type Optional RTTOV-SCATT output structure containing data enabling all-sky
emissivity retrievals.
rttov_pccomp Calculated PC scores and reconstructed radiances.
rttov_emis_atlas_data Holds data for an emissivity atlas (from mod_rttov_emis_atlas.F90)
rttov_brdf_atlas_data Holds data for a BRDF atlas (from mod_rttov_brdf_atlas.F90)
Table 6. User-level derived types.
34
Version : 1.3
Date : 05/03/2019
Library name User-level subroutines Description

librttov12_main.a rttov_direct, rttov_tl, rttov_ad, rttov_k Core library, always required.
rttov_alloc_direct, rttov_alloc_tl Interfaces to the direct, TL, AD, K models.
rttov_alloc_ad, rttov_alloc_k Allocation and initialisation subroutines for
rttov_alloc_prof, rttov_init_prof RTTOV structures.
rttov_alloc_rad, rttov_init_rad User “checkinput” subroutines to verify input
rttov_alloc_transmission options and profiles.
rttov_init_transmission Subroutine to set the logical unit for error
rttov_alloc_pccomp, rttov_init_pccomp messages.
rttov_alloc_traj, rttov_alloc_opt_param
rttov_init_opt_param
rttov_user_options_checkinput
rttov_user_profile_checkinput
rttov_errorhandling
librttov12_hdf.a N/A Required if RTTOV was compiled with HDF5
capability.
librttov12_coef_io.a rttov_read_coefs, rttov_read_coefs_htfrtc Coef input/output, always required.
rttov_dealloc_coefs Read coefficient files, deallocate memory and
rttov_get_pc_predictindex get PC-RTTOV predictor channel sets.
librttov12_parallel.a rttov_parallel_direct, rttov_parallel_tl Parallel interfaces for multi-threaded execution
rttov_parallel_ad, rttov_parallel_k using OpenMP.
librttov12_mw_scatt.a rttov_scatt, rttov_scatt_tl, rttov_scatt_ad Interface to RTTOV-SCATT for MW scattering
rttov_parallel_scatt, rttov_parallel_scatt_tl simulations.
rttov_parallel_scatt_ad Allocation and initialisation subroutines for
rttov_read_scattcoeffs RTTOV-SCATT structures.
rttov_dealloc_scattcoeffs Subroutines for reading RTTOV-SCATT
rttov_alloc_scatt_prof coefficient files and deallocating memory.
rttov_init_scatt_prof Subroutine for performing all-sky emissivity
rttov_alloc_emis_ret_terms retrieval calculation.
rttov_scatt_setupindex
rttov_scatt_emis_retrieval
librttov12_emis_atlas.a rttov_setup_emis_atlas, rttov_get_emis Interface to emissivity atlases.
rttov_deallocate_emis_atlas
librttov12_brdf_atlas.a rttov_setup_brdf_atlas, rttov_get_brdf Interface to BRDF atlas.
rttov_deallocate_brdf_atlas
librttov12_other.a rttov_bpr_calc, rttov_bpr_init Ancillary subroutines:
rttov_bpr_dealloc  calculate “bpr” parameter for IR scattering
rttov_legcoef_calc simulations with explicit optical parameters
rttov_print_opts, rttov_print_opts_scatt  calculate Legendre coefficients for a phase
rttov_print_profile function for visible/IR scattering simulations
rttov_print_cld_profile  print out contents of options, profile and coef
rttov_print_info structures for debugging
rttov_print_radiance_quality  print info for radiance quality output
rttov_scale_ref_gas_prof  generate scaled copies of the RTTOV
rttov_aer_clim_prof background trace gas profiles
rttov_zutility  generate climatological aerosol profiles
 obtain values of magnetic field strength for
Zeeman simulations
librttov12_wrapper.a Link against this library when compiling your
See separate wrapper user guide for
own C++ code which uses the RTTOV wrapper.
wrapper API.
rttov_wrapper_f2py.so F2PY library for Python interface to RTTOV.
rttov_gui_f2py.so N/A F2PY library for RTTOV GUI.
The remaining libraries are not intended for linking:
librttov12_test.a, librttov12_mw_scatt_coef.a, librttov12_coef_io_11.a, librttov12_gui.a
Table 7. Libraries produced by the build process and associated user-level subroutines.
35
Version : 1.3
Date : 05/03/2019
Executable name Purpose

Useful executables.
rttov_coef_info.exe Print out information about a given rtcoef_ coefficient file.
rttov_mfasis_lut_info.exe Print out information about a given MFASIS LUT file.
rttov_conv_coef.exe Convert coefficients between formats (ASCII, binary, HDF5) and extract channels
to reduce file sizes.
rttov11_conv_coef_11to12.exe Convert v10/v11 rtcoef coefficient files to v12 format.
rttov11_conv_coef_12to11.exe Convert v12 rtcoef coefficient files to v10/v11 format.
rttov_ascii2bin_scattcoef.exe Convert ASCII RTTOV-SCATT Mie tables to binary format.
rttov_make_scaercoef.exe Generate visible/IR scaercoef aerosol optical property files.
rttov_scatt_make_coef.exe Generate RTTOV-SCATT Mie table files.
create_aer_clim_prof.exe Generate a file containing climatological aerosol profiles from combinations of
the RTTOV pre-defined particle types.
rttov_obs_to_pc.exe Demonstrating conversion of observations to PC-space for PC assimilation
applications.
Executables for test suite and example program files: these should be run via the supplied scripts in rttov_test/.
rttov_test.exe Main test suite executable: should be called via rttov_test.pl.
example_fwd.exe Example program demonstrating simple forward model call.
example_k.exe Example program demonstrating simple K model call.
example_atlas_fwd.exe Example program demonstrating forward model call with use of emissivity and
BRDF atlases.
example_cld_file_fwd.exe Example program demonstrating forward model call with clouds specified using
pre-defined particle types.
example_cld_param_fwd.exe Example program demonstrating forward model call with cloud parameter
profiles specified explicitly.
example_cld_mfasis_fwd.exe Example program demonstrating forward model call for MFASIS cloud
simulations.
example_aer_file_fwd.exe Example program demonstrating forward model call with aerosols specified using
pre-defined particle types.
example_aer_param_fwd.exe Example program demonstrating forward model call with aerosol parameter
profiles specified explicitly.
example_rttovscatt_fwd.exe Example program demonstrating forward model call for RTTOV-SCATT.
example_pc_fwd.exe Example program demonstrating forward model call for PC-RTTOV.
example_htfrtc_fwd.exe Example program demonstrating forward model call for HTFRTC.
example_rttovscatt.exe Example program demonstrating calling RTTOV-SCATT.
rttovscatt_test.exe RTTOV-SCATT test executable, should be run using the test_rttovscatt.sh script.
rttov_uwiremis_atlas_test.exe Test program for UWIRemis IR emissivity atlas.
rttov_camel_atlas_test.exe Test program for CAMEL IR emissivity atlas.
rttov_camel_clim_atlas_test.exe Test program for CAMEL climatology IR emissivity atlas.
rttov_telsem2_atlas_test.exe Test program for TELSEM2 MW emissivity atlas.
rttov_cnrm_mw_atlas_test.exe Test program for CNRM MW emissivity atlas.
rttov_brdf_atlas_test.exe Test program for BRDF atlas.
Additional executables.
rttov_test_get_pc_predictindex.exe Test program for rttov_get_pc_predictindex subroutine, can be used to obtain the
PC-RTTOV predictor channel numbers.
rttov_gui_test_run.exe Test program for GUI functionality (not generally required).
rttov_make_opt_param.exe Used by test suite to create explicit optical property input files from cloud/aerosol
profiles and associated coefficient files.
Table 8. Executables produced by the RTTOV build process.
36
Version : 1.3
Date : 05/03/2019
7. Running RTTOV v12 for your applications

RTTOV v12 includes interfaces to allow much RTTOV functionality to be called from Python, C or C++ code. This
allows you to run RTTOV without writing any Fortran code. There is a separate document in the docs/ directory which
describes the wrapper interface. However, it is still important to be aware of the contents of this section of the user
guide as you will not be able to run RTTOV correctly without understanding it.
There are several example programs which can serve as a basis for your application:
 example_fwd.F90 clear-sky with/without solar radiation
 example_k.F90 clear-sky K model simulation
 example_atlas_fwd.F90 as example_fwd but using the emissivity and BRDF atlases
 example_cld_file_fwd.F90 visible/IR cloud scattering via pre-defined particle types
 example_cld_param_fwd.F90 visible/IR cloud scattering by explicitly providing optical parameters
 example_cld_mfasis_fwd.F90 visible MFASIS cloud scattering
 example_aer_file_fwd.F90 visible/IR aerosol scattering via pre-defined particle types
 example_aer_param_fwd.F90 visible/IR aerosol scattering by explicitly providing optical parameters
 example_rttovscatt_fwd.F90 RTTOV-SCATT
 example_pc_fwd.F90 PC-RTTOV
 example_htfrtc_fwd.F90 HTFRTC
These may be found in the src/test/ directory and section 5.3 describes how to run them.
The module rttov_types defines the derived types used by RTTOV that you will need in your program. These are listed
in Table 6 and Annex O gives full details. There are also two additional derived types contained in
mod_rttov_emis_atlas and mod_rttov_brdf_atlas which are required when using the emissivity and BRDF atlases:
see section 6, sections 7.5 and 7.6, and example_atlas_fwd.F90.
The module parkind1 defines the standard RTTOV integer, real and logical kinds (jpim, jprb and jplm respectively).
The default RTTOV integer kind is 32-bit and the default real kind is double precision (64-bit): it is not recommended
to change the latter to single precision (32-bit). This is particularly true if running the AD or K models as this can
significantly affect the adjoint/Jacobian output.
The rttov_types and parkind1 modules should always be used in your code which calls RTTOV. You will also see that
the example code uses the rttov_const module: this contains a large number of parameters (constants) used by RTTOV,
some of which may be useful in your program. The files rttov_types.F90, parkind1.F90 and rttov_const.F90 can be
found in src/main/.
The rttov_options derived type (in module rttov_types) holds a number of flags controlling various aspects of
RTTOV. It is also important to allocate the various input and output arrays for rttov_direct to the correct dimensions
(see example_fwd). Figure 1 gives a process diagram of what routines to call when running RTTOV v12.
It is recommended that you look at the header section of the coefficient file for the sensor you wish to simulate as there
is useful information there such as the definition of channel numbers and the polarisation (for MW sensors) assumed for
each channel for that instrument etc. You can use the rttov_coef_info.exe executable (see Annex A) to view the
headers of any coefficient file.
The following sections describe the recommended steps to be taken in coding a program which calls RTTOV v12.
These involve preparing structures and arrays with the necessary input data for RTTOV, and creating appropriate
structures and arrays to hold the output of the model. Users of RTTOV v11 will notice that the interfaces to some
routines have changed slightly, but the fundamental sequence of calls remains the same. To provide some context, the
syntax for calling rttov_direct is provided here with the arguments containing input data highlighted in bold. The
subroutine interface is described fully in Annex I.
call rttov_direct(errorstatus, chanprof, opts, profiles, coefs, transmission,

radiance, radiance2, calcemis, emissivity, calcrefl, reflectance, aer_opt_param,
cld_opt_param, traj, pccomp, channels_rec)
37
Version : 1.3
Date : 05/03/2019
7.1. Set RTTOV options

You must first declare a variable of the rttov_options derived type, say opts. This structure contains a number of flags
with which you can configure various aspects of RTTOV. The full structure is described in Annex O. It is important to
set options covering the widest range of required types of simulation before reading the coefficients and allocating
RTTOV structures. For example, if cloudy IR simulations are required then set opts%rt_ir%addclouds to true at the
start before reading the coefficients and allocating the profiles(:) structure. This ensures that the cloudy coefficients are
read in and the cloudy profile array variables are allocated. You can subsequently set the addclouds option to false to
run clear-sky simulations without reallocating structures or re-reading coefficients. If you wish to read MFASIS LUT
files you must set opts%rt_ir%vis_scatt_model=3 (i.e. select MFASIS as the solar scattering model) before reading
the coefficient files in.
You can also initialise the logical unit for error/warning messages. This is performed by rttov_errorhandling, an
optional subroutine which can be called at any time (see Annex B).
7.2. Initialise coefficient structures

The coefs structure (derived type rttov_coefs) contains data read from the optical depth (rtcoef) coefficient file and,
optionally data from the visible/IR cloud and/or aerosol (sccldcoef, scaercoef) coefficient files, MFASIS LUTs
(rttov_mfasis_), and the PC-RTTOV (pccoef) coefficient file for the instrument being simulated.
You should declare an instance of the rttov_coefs type and read the coefficients for the desired instrument by calling
rttov_read_coefs (see Annex C). After reading the coefficients you may wish to call the subroutine
rttov_user_options_checkinput (see Annex N) which checks the consistency of the input options with the coefficient
file and reports any issues: this can be useful for debugging purposes.
If fast performance is required when reading the coefficient files it is better to access Fortran unformatted or HDF5
formatted coefficient files. The program rttov_conv_coef.exe (located in the bin/ directory of the build) can be used to
convert between the three formats. It can also produce ASCII, unformatted or HDF5 coefficient files for a subset of
channels which can be useful in particular for hyperspectral sounders. The command-line arguments for this tool are
described in Annex A.
Another program, rttov11_conv_coef_11to12.exe, is available to convert version 10/11 coefficient files to version 12-
compatible files: this routine is also documented in Annex A.
The coefficient file defines the variable trace gases allowed in the input profile (see Table 4). There are two points to be
aware of in specifying what gaseous absorption needs to be included in the computation. The first is that the flag in the
options structure for the gas of interest must be set to .true. The second is that the coefficient file supplied must contain
the coefficients for the gas of interest. The fewer gases simulated, the faster the code will run. In all cases, water vapour
is a mandatory input. All other trace gases (O3, CO2, CO, N2O, CH4, SO2) are optional. Clear-sky MW simulations may
also use a cloud liquid water input profile which is treated purely as an absorber in a similar manner to the gases.
If the coefficient file contains coefficients for a particular gas (e.g. coefs%coef%nozone > 0) then a profile may be
supplied for that gas. If you wish to provide profiles for the gas the relevant flag in opts (e.g.
opts%rt_ir%ozone_data) should be set to true. If the opts flag is set to false the RTTOV reference profile is used
instead. This is necessary because the coefficients for the fixed gases will not include any gases which have specific
variable gas coefficients and so the calculation will be in error if the variable gas calculation is not included. Note none
of the MW coefficients include ozone as a variable gas. The reference profiles are contained in the
REFERENCE_PROFILE section of the ASCII coefficient files: this section lists a “reference” profile and a
“background” profile for each gas. It is the latter which is used when an optional gas input is omitted and in fact for all
gases except SO2 the reference and background profiles are identical. For HDF5 coefficients the background gas
profiles are in the /COEF/BKG_PRFL_MR dataset.
The predictors also determine whether solar computations are allowed. The v9 predictors were specifically designed to
work well over the wider range of zenith angles required for solar simulations. Therefore only v9 predictor files are
compatible with solar simulations. For all non-hires IR sensor v9 predictor coefficient files, the only optional variable
trace gases are ozone and CO2.
38
Version : 1.3
Date : 05/03/2019
Figure 1. Process diagram of user program calling RTTOV v12 forward model.
39
Version : 1.3
Date : 05/03/2019
7.3. Set up input profiles

The program src/test/example_fwd.F90 is a useful guide on how to set up the input profile(s). Allocate a profiles array
(derived type rttov_profile) with size equal to the number of profiles you wish RTTOV to process in each call. The
rttov_alloc_prof subroutine (see Annex D) should be called to allocate the various arrays within the profile structure.
This can also initialise all the profile variables to zero if requested, which is recommended. Alternatively the
rttov_alloc_direct/tl/ad/k subroutines can be used to allocate all input arrays to the direct, TL, AD and K models (see
Annex D). The members of the array profiles as listed in Table 12 should be populated with data: some are mandatory
as indicated and some are optional depending on various factors such as the flags set in opts and the coefficient file
being used. Table 12 also indicates which profile variables are treated as constants for the tangent linear and which are
active in the TL calculation (and hence which may be non-zero in the Jacobian).
Figure 2. Internal RTTOV coefficient levels and optical depth computations.
Figure 3. Interpolation to and from to user levels.
Pressure levels and interpolation
RTTOV carries out the optical depth calculation on a fixed set of levels (usually 54 or 101 levels) defined by the
coefficient file. You can provide input profiles on these fixed sets of pressure levels (see Table 1 for the 54 levels) in
which case no interpolation is required. However the input profiles may be supplied on arbitrary pressure levels (see for
example Figures 2 and 3). In this case you must set opts%interpolation%addinterp to true. Note that this increases
the run-time of RTTOV, although the interpolation is faster if all input profiles use the same pressure levels. RTTOV
interpolates the input profile data onto the fixed coefficient levels, calculates the optical depths on the coefficient levels
and then interpolates the optical depths back onto the input pressure levels. The radiative transfer integration is carried
out on the input pressure levels. Input profiles can be on different pressure levels, but for each call to RTTOV the
number of levels must be the same for all profiles.
The interpolation method is given in Rochon et al. (2007) and is also described in the RTTOV v9 Science and
Validation Report (see the RTTOV website). An important feature of the Rochon interpolation scheme is that all input
levels in an interpolation contribute to the interpolated profile with appropriate weight which means that Jacobians do
40
Version : 1.3
Date : 05/03/2019
not exhibit excessively large or small sensitivity to particular levels purely as a result of the interpolation. However, it
has been found that the Rochon interpolation scheme can result in significant oscillations in the computed temperature
Jacobians in the case when the input levels are more densely spaced than the coefficient levels. RTTOV provides
additional interpolation options which are intended to mitigate this effect. These options are listed in Table 9 below.
The interpolation option is set in the opts%interpolation%interp_mode variable. The default is the standard Rochon
interpolator that was originally implemented in RTTOV v9. Despite the oscillations observed in the Jacobians it is
worth noting that the default interpolation option (mode 1) has not been observed to have a negative impact in
applications such as 1DVar retrievals or data assimilation systems. This is a result of the vertical correlations in the
background smoothing out the impact of the oscillations in the Jacobians. Therefore if you are happily using the original
interpolator then there is no compelling reason to change. However, modes 4 and 5 result in smooth temperature
Jacobians and may be considered as alternatives when the input profile levels are more densely spaced than the
coefficient levels. In the opposite case (coefficient levels are more dense), then modes 1 or 3 are recommended. A more
detailed description and comparison of the interpolation modes is given in Hocking (2014). You may wish to carry out
experiments to determine the best interpolation option for your application. You can of course interpolate your input
profiles onto the RTTOV coefficient levels before input to RTTOV (and therefore switch off the RTTOV interpolation)
if you find this beneficial for your application.
interp Profile interp Optical depth interp Description

mode (user->coef levels) (coef->user levels)
1 Rochon Rochon on optical Default, original RTTOV interpolation method,
depths Jacobians may show oscillations. Reasonable choice
when user levels are sparse compared to coef levels.
2 Log-linear Log-linear on optical May be beneficial in execution time for direct-model
depths calculations, but not suitable for TL/AD/K.
3 Rochon Log-linear on optical Similar to mode 1, but with somewhat reduced
depths oscillations. Reasonable choice when user levels are
sparse compared to coef levels.
4 Rochon Rochon on weighting No oscillations, but most computationally expensive
function method. Reasonable choice when user levels are dense
compared to coef levels.
5 Rochon Log-linear on No oscillations, but Jacobians may show small
weighting function “artefacts” due to interpolation, only slightly more
expensive than mode 1. Reasonable choice when user
levels are dense compared to coef levels.
Table 9. Interpolation options available in RTTOV.
There are several points which need to be considered if the internal profile interpolation is used:
i. Ideally the input profile should cover the whole atmosphere with an adequate number of levels, at least close to
the number of coefficient levels or more. A coarse layering will reduce the accuracy of the calculations.
Certainly, for accurate simulations, the input profile should span the range of pressures over which the
weighting functions of the channels being simulated are significantly greater than zero.
ii. If there are pressure levels in the input profile below the surface then they may contribute to the simulated
radiance. In particular, the level immediately below the surface will contribute in many situations as the values
of profile variables on the levels immediately above and below the surface may be used to determine the
values for the partial layer above the surface. However, due to the interpolation method, it is possible for levels
even below this to contribute to the simulation.
iii. If the input profile is below the top of the coefficient file the input profile is extrapolated assuming a constant
value of the uppermost user value for all levels above the top. Alternatively there is an option to carry out
extrapolation based on the coefficient file regression limits (see below). In both cases the extrapolated values
are clipped so that extrapolated values never exceed the regression limits.
iv. When calling the tangent linear model with interpolation, profiles_tl%p (i.e. the pressure perturbations in the
perturbation profile – see Annex K) can be non-zero (and for sigma levels should be). In this case
opts%interpolation%lgradp must be set to true. If the interpolation is not used, input levels are assumed to
be on fixed pressure levels, so profiles_tl%p should be zero. This applies similarly to the adjoint and Jacobian
models: the output profiles_ad/_k%p will be zero unless interpolation is used and lgradp is true.
41
Version : 1.3
Date : 05/03/2019
Specification of the surface pressure
The location of the surface is defined by the surface pressure in profile%s2m%p. There is no restriction on the value
of the surface pressure (other than that it must be physically reasonable). RTTOV accounts for emission, absorption and
(where relevant) scattering from the top of the atmosphere down to the surface pressure. Ideally the surface pressure
should be less than or equal to the bottom pressure level in the input profile. If the surface pressure lies between two
input pressure levels, RTTOV calculates the contribution from the partial layer bounded by the surface pressure and the
first pressure level above this. RTTOV also allows the surface pressure to lie below the bottom pressure level: in this
case the calculated optical depth of the bottom layer is extrapolated down to the surface pressure.
The RTTOV profile structure also includes an elevation variable (profile%elevation): this is used in the calculations
which determine the local path angles of the radiation in each atmospheric layer. It is the surface pressure that
determines the surface location within the vertical profile rather than the elevation.
Trace gases
Table 4 in section 3 lists the gases available with each type of coefficient file. If you are supplying profiles for an
optional trace gas you must set the corresponding options flag to true: e.g. if you supply ozone data then you must set
opts%rt_ir%ozone_data to true. If the flag is true you must supply a valid input profile. If a coefficient file with trace
gas coefficients is used and suitable trace gas profile concentrations are not available then you can set the corresponding
data flag to false. In this case RTTOV uses the reference trace gas profile for the calculation. This applies to all gases
except water vapour which is always mandatory.
RTTOV v12.3 has a new subroutine rttov_scale_ref_gas_prof which provides a method of supplying scaled copies of
the RTTOV background profiles for the optional gases. This could be useful, for example, if you do not have a CO2
profile, but wish to specify a lower concentration corresponding to a specific time in the past: this new subroutine
allows you to specify the maximum CO2 ppmv value and the RTTOV background profile will be scaled to have this
maximum value. In order to use this subroutine, you must specify the relevant gas flags (e.g. ozone_data or co2_data)
that you wish to include in the simulation. You should then populate the profiles(:) structure with all of the data (p(:),
t(:), q(:), etc) except for the trace gas(es) for which you intend to use the scaled reference profiles. Just before the call to
RTTOV, you call rttov_scale_ref_gas_prof passing the coefficients structure, the profiles(:) array, and specifying the
relevant arguments for scaling the gas(es) of interest. The subroutine then populates those gas arrays in the profiles(:)
structures. Annex N provides mores information about the subroutine.
It is possible to specify the units of the gas abundances in the input profile structure. This is set in the
profiles(:)%gas_units variable: all profiles passed into RTTOV in a single call must use the same gas units. The units
apply to all trace gas input profiles and the 2m (surface) gas variables. Table 10 lists the available options for gas units.
Note that the units of gas Jacobians are consistent with the setting of gas_units so that if gas_units = 1, for example,
then water vapour Jacobians are in units of [Kelvin / (kg/kg)].
profiles(:)%gas_units Description
2 ppmv over moist air (default)
1 kg/kg over moist air
0 (or less) ppmv over dry air: this is intended primarily for coefficient generation and testing, but
is a valid input option.
Table 10: Options for gas units
Cloud liquid water (CLW) for MW simulations
For MW simulations it is possible to supply a cloud liquid water (CLW) profile in the profiles(:)%clw(:) array which is
treated as a purely absorbing medium. It is important to note that this is not the same as CLW supplied to RTTOV-
SCATT for scattering simulations (see section 8.7). If you wish to include CLW in the “clear-sky” simulations you
must set opts%rt_mw%clw_data to true. By default RTTOV ignores any CLW content above 322hPa. This limit is
specified in the opts%rt_mw%clw_cloud_top option. The Jacobian model generates non-zero Jacobians even for all
levels with zero CLW content, but only those levels with pressures greater than the limit specified in clw_cloud_top.
Historically in RTTOV the CLW optical depth calculation has been carried out on the coefficient levels. However this
is not required and arguably it is better to avoid interpolating cloud profiles. In RTTOV v12.2 a new option was
introduced, opts%rt_mw%clw_calc_on_coef_lev, which gives the option to do the calculation on coefficient levels
42
Version : 1.3
Date : 05/03/2019
(true, the default) or on the input levels (false, recommended). RTTOV v12.2 also introduced a choice of liquid water
permittivity parameterisations: these are selected using the new opts%rt_mw%clw_scheme option:
 1 => Liebe (1989), same as previous versions of RTTOV (default)

 2 => Rosenkranz (2015), supports frequencies up to 1000GHz, can increase run-time by ~50% (recommended)
 3 => Turner, Kneifel, Cadeddu or TKC (2016)
These same options are now available for generating RTTOV-SCATT Mietable files. Rosenkranz (2015) is the
recommended option, and is the one used to generate the latest RTTOV-SCATT Mietable files.
Limits applied to input profile data
All profile variables must lie within the “hard” limits specified in section 1.9 of rttov_const.F90 (see Table 11). This
means, for example, that all input gas profiles (water vapour, ozone, etc) must have values greater than 1E-11 ppmv at
every level. This applies to all levels in the input profile, even those which lie below the specified surface pressure (in
profile%s2m%p), as the optical depth calculations are carried out for every layer. Input profile variables which lie
outside these hard limits will cause RTTOV to abort the simulation and report a non-zero error status.
The clear-sky gas optical depth calculations (excluding CLW for MW sensors) are based on regressions derived from
line-by-line calculations run for a training set of diverse profiles. The minimum/maximum envelopes of the training
profile data are contained in the coefficient files. RTTOV calculates “regression limits” by applying stretches to the
envelope values (+/-10% for temperature, +/-20% for each gas). By default RTTOV compares the input profiles to these
regression limits and will print warnings when the regression limits are exceeded, but the simulations are performed
regardless. A new feature in RTTOV v12.2 was the output radiance%quality array which will indicate if the
regression limits are exceeded for any quantity for a given channel (the flags will not be set if the do_checkinput
option is false – see below). See section 7.8 for more information about the quality output.
RTTOV provides some options for checking input profile data:

- The printed warnings about profiles exceeding the regression limits can be suppressed by setting the
opts%config%verbose option to false (default: true).
- Input profiles can be clipped to the regression limits when the limits are exceeded by setting
opts%config%apply_reg_limits to true (default: false). This also prevents printed warnings being output.
- You can turn off all the internal checking of profiles against both the hard limits described above and the
regression limits by setting opts%config%do_checkinput to false (default: true). This is not recommended
unless you are using the rttov_user_profile_checkinput subroutine described below.
Note that when opts%config%apply_reg_limits is set to true the profile values are only clipped to the regression
limits for the purposes of the gas absorption optical depth calculation. The source function for the radiative transfer
equation is always based on the unadjusted input profiles. Also note that when warnings are printed out for gases, the
values are in ppmv over dry air regardless of the setting of the gas_units variable. You may wish to perform your own
sensitivity study to decide what setting of the apply_reg_limits option is best for your application.
PC-RTTOV uses its own set of regression limits, but they operate in the same way as for standard RTTOV simulations
(see section 8.8).
The rttov_user_profile_checkinput subroutine (Annex N) can be used to test profiles against both the hard limts and
the regression limits before calling RTTOV. This allows out-of-bounds profiles to be rejected without running full
simulations if desired. In this case you can set opts%config%do_checkinput to false to turn off the internal RTTOV
profile checking.
When using the RTTOV interpolator it may be the case that your input profile top is lower than the top of the RTTOV
coefficient levels (usually 0.005hPa). In this case RTTOV must extrapolate the input profile values to the top coefficient
level. By default constant-value extrapolation is used: if any extrapolated values lie outside the regression limits they
are clipped to the limits to avoid generating warning messages. RTTOV provides an alternative option for profile
extrapolation: by setting opts%interpolation%reg_limit_extrap to true, the profile interpolation step identifies the
relative position of each interpolated profile variable at the top of the input profile with respect to the coefficient
regression limits at that level. Each profile variable is then extrapolated upwards to top coefficient level by maintaining
the relative position between the regression limits at every level. This is performed independently for each profile
43
Version : 1.3
Date : 05/03/2019
variable. This option may be useful for obtaining physically reasonable profiles at the top of the atmosphere when the
input data does not reach the top-most coefficient levels. If the interpolated value at the top input level lies beyond the
regression limits, then every extrapolated value will be set to the corresponding limit value, similar to the
apply_reg_limits functionality. The apply_reg_limits and reg_limit_extrap options may be switched on or off
independently.
Variable Minimum Maximum

Temperature (inc. 2m T) 90 K 400 K
Water vapour (inc. 2m q) 1E-11 ppmv 60000 ppmv
O3 1E-11 ppmv 1000 ppmv
CO2 1E-11 ppmv 1000 ppmv
CO 1E-11 ppmv 10 ppmv
N2O 1E-11 ppmv 10 ppmv
CH4 1E-11 ppmv 50 ppmv
SO2 1E-11 ppmv 1000 ppmv
Cloud liquid water (clear-sky MW only) 0 kg/kg 1 kg/kg
2m pressure 400 hPa 1100 hPa
10m wind speed 0 m/s 100 m/s
Cloud top pressure (simple cloud; only 50 hPa 1100 hPa
applies if cfraction > 0)
Magnetic field Be (Zeeman only) 0.2 Gauss 0.7 Gauss
Table 11: Hard limits for input profiles variables as defined in rttov_const.F90.
Input values must not lie outside the interval [minimum, maximum].
Input profile arrays Description Units Mand When used Variable

atory? for TL?
profiles(i) % nlevels Number of pressure levels Y N
profiles(i) % nlayers Number of atmospheric layers (i.e. Y N
nlevels – 1)
profiles(i) % gas_units Units of gas abundances (must be 0,1,2 Y N
the same for all profiles)
profiles(i) % p(:) Pressure levels hPa Y Y if opts %
interpolation
% lgradp
true
profiles(1) % t(:) Temperatures on levels K Y Y
profiles(i) % q(:) Water vapour conc on levels ppmv or Y Y
kg/kg
profiles(i) % o3(:) Ozone conc on levels ppmv or N If ozone_data flag Y
.true.
kg/kg
profiles(i) % co2(:) CO2 conc on levels ppmv or N If co2_data flag Y
.true.
kg/kg
profiles(i) % n2o(:) N2O conc on levels ppmv or N If n2o_data flag Y
.true.
kg/kg
profiles(i) % co(:) CO conc on levels ppmv or N If co_data flag Y
.true.
kg/kg
profiles(i) % ch4(:) CH4 conc on levels ppmv or N If ch4_data flag Y
.true.
kg/kg
profiles(i) % so2(:) SO2 conc on levels ppmv or N If so2_data flag Y
.true.
kg/kg
profiles(i) % clw(:) Microwave cloud liquid water kg/kg N MW clear-sky Y
only, if clw_data
treated as absorbing medium; not
flag .true.
used with RTTOV-SCATT.
profiles(i) % mmr_cldaer Logical flag to set cloud and aerosol T/F N VIS/IR cloud or N
aerosol with coef
units: true => kg/kg (cld+aer);
files
false => g.m-3 (cld), cm-3 (aer)
(must be the same for all profiles)
44
Version : 1.3
Date : 05/03/2019
profiles(i) % aerosols(:,:) Aerosol components/species conc kg/kg or N VIS/IR aerosol Y

scattering with
indices are (aerosol_type, on layers cm-3
aerosol coef file
layers)
profiles(i) % cloud(:,:) Cloud water/ice content on layers kg/kg or N VIS/IR cloud Y
scattering with
indices are (cloud_type, g.m-3
cloud coef file
layers)
unless user inputs
optical params
profiles(i) % cfrac(:,:) Cloud fractional cover on layers 0-1 N VIS/IR cloud Y
scattering
profiles(i) % clw_scheme CLW particle scheme: OPAC (1) or 1-2 N VIS/IR cloud N
scattering with coef
“Deff” scheme (2)
file
profiles(i) % ice_scheme Ice particle scheme: SSEC database 1-2 N VIS/IR cloud N
(1) or Baran scheme (2)
file
profiles(i) % idg Ice water content effective diameter 1-4 N VIS/IR cloud N
scheme
file and
ice_scheme=1
profiles(i) % clwde(:) Cloud liquid water particle effective microns N Required with Y
VIS/IR cloud
diameter
file and
clw_scheme=2
profiles(i) % icede(:) Ice particle effective diameter microns N Optional with Y
VIS/IR cloud
file and
ice_scheme=1
profiles(i) % s2m % p Surface pressure hPa Y Y
profiles(i) % s2m % t 2m temperature K Y Y
profiles(i) % s2m % q 2m water vapour ppmv or N If use_q2m flag is Y
.true.
kg/kg
profiles(i) % s2m % o 2m ozone ppmv or N Currently never N
used.
kg/kg
profiles(i) % s2m % u, 10m wind u, v components m/s N Sea surface Y
emissivity and
profiles(i) % s2m % v
BRDF models
(except ISEM)
profiles(i) % s2m % wfetc Wind fetch (length of water over m N Sea surface BRDF Y
models
which the wind has blown)
profiles(i) % skin % Surface type (land = 0, sea = 1, 0-2 Y Emissivity and N
BRDF models and
surftype seaice = 2)
atlases
profiles(i) % skin % Water type (fresh = 0, ocean = 1) 0-1 N Surface BRDF N
models and BRDF
watertype
atlas
profiles(i) % skin % t Surface skin temperature K Y Y
profiles(i) % skin % salinity Ocean salinity Practical N FASTEM 4-6, Y
TESSEM2
salinity
unit
profiles(i) % skin % Ocean foam fraction 0-1 N FASTEM if Y
supply_foam_fracti
foam_fraction
on flag .true.
profiles(i) % skin % Surface snow cover fraction 0-1 N IR emis atlases N
snow_fraction
profiles(i) % skin % Surface soil moisture m3/m3 N Not used N
soil_moisture
profiles(i) % skin % FASTEM land/sea-ice parameters N FASTEM for Y
land/sea-ice surface
fastem(1:5) (see Table 21)
types
profiles(i) % skin % Specularity (weighting for specular 0-1 N If do_lambertian N
flag is .true.
specularity and Lambertian downwelling
radiances).
45
Version : 1.3
Date : 05/03/2019
profiles(i) % cfraction Cloud fraction for simple cloud 0-1 N Simple cloud Y
VIS/IR only
profiles(i) % ctp Cloud top pressure for simple cloud hPa N Simple cloud Y
VIS/IR only (if
cfraction > 0)
profiles(i) % zenangle Satellite zenith angle deg Y N
profiles(i) % azangle Satellite azimuth angle (0-360; deg N Using FASTEM or N
solar option
measured clockwise, east=+90)
profiles(i) % sunzenangle Solar zenith angle deg N Solar option, NLTE N
option
profiles(i) % sunazangle Solar azimuth angle (0-360; deg N Solar option N
measured clockwise, east=+90)
profiles(i) % latitude Latitude (-90 to +90) deg Y For radiation path N
geometry and
emis/BRDF atlases
profiles(i) % longitude Longitude (0-360) deg N Emis/BRDF atlases N
profiles(i) % elevation Elevation km Y For radiation path N
geometry
profiles(i) % Be Earth magnetic field strength Gauss N Zeeman N
profiles(i) % cosbk Cosine of the angle between the N Zeeman N
Earth magnetic field and wave
propagation direction
profiles(i) % date(1:3) Date of the profile as year (e.g. N Used with solar N
calculations.
2013), month (1-12), and day (1-31)
profiles(i) % time(1:3) Time of profile as hour, minute, N Not used N
second.
Table 12. Profile input parameters for user profile i.
7.4. Specifying the channels to simulate

As described above, the profiles(:) array contains a list of the profiles for which to calculate radiances. RTTOV offers
the flexibility to calculate radiances for a different set of channels for each profile in profiles(:), though often in
practice radiances for the same set of channels will be calculated for every profile. You should allocate a chanprof(:)
array (derived type rttov_chanprof): this defines which channels are simulated for each profile. The size is the total
number of channels to simulate over all profiles in each call to RTTOV. Each element in the array has two members:
chanprof(j)%prof (the profile index), and chanprof(j)%chan (the channel index), where j runs from 1 up to the total
number of radiances to calculate per call to RTTOV. Table 13 illustrates how chanprof(:) should be set up for three
different sensors and for 2 profiles per RTTOV call: all channels for the first profile are specified, followed by all
channels for the second profile, and so on. Note that for Principal Components calculations the channels you must
simulate are determined by opts%rt_ir%pc%ipcbnd and opts%rt_ir%pc%ipcreg as described in section 8.8 and in
this case you must simulate the same channels for every profile.
The RTTOV channel numbering always begins at 1 for any coefficient file regardless of the original instrument channel
numbering. This means that with IR-only and visible/IR coefficient files for the same instrument, RTTOV may use
different indices to refer to the same channel (see for example SEVIRI in Table 3). If a subset of n possibly non-
consecutive channels is read from the coefficient file (by supplying the channels(:) argument to rttov_read_coefs),
then this subset of channels is identified in chanprof(:)%chan by the numbers 1 to n rather than by their original
channel index in the coefficient file. For example, if coefficients for only channels 3 and 5 of a sensor with five
channels are read from the coefficient file, then chanprof(1:2)%chan should typically be (/1, 2/) corresponding to the
channels 3 and 5 respectively. Similarly, if the rttov_conv_coef.exe executable is used to create a coefficient file
containing a subset of n possibly non-consecutive instrument channels, then these will be identified by the indices 1 to n
when reading this new coefficient file using rttov_read_coefs.
If Principal Component computations are being carried out (see Section 8.8) you may request radiances to be
reconstructed from the PC scores by setting opts%rt_ir%pc%addradrec to true. In this case the array
channels_rec(:) must be supplied containing the indices of the channels for which reconstructed radiances are required.
The reconstructed radiance channel numbers are again always counted from 1. If you passed the channels_rec(:)
argument to rttov_read_coefs to specify some subset of n possibly non-consecutive channels for which radiances may
be reconstructed, then this subset of channels is identified in the channels_rec(:) argument to rttov_direct (and
46
Version : 1.3
Date : 05/03/2019
TL/AD/K) by the numbers 1 to n just as for chanprof(:)%chan. If the PC calculations are not being used or
reconstructed radiances are not required the channels_rec(:) argument should be omitted.
Input structure HIRS (2 profiles/call) SSM/I (2 profiles/call) AMSU-B (2 profiles/call)

size(chanprof) 38 14 10
size(profiles) 2 2 2
chanprof(:) % chan 1,2,3 ..,19,1,2,3..,19 1,2,3,4,5,6,7,1,2,3,4,5,6,7 1,2,3,4,5,1,2,3,4,5
chanprof(:) % prof 1,1,1,…,1,2,2,2…,2 1,1,1,1,1,1,1,2,2,2,2,2,2,2 1,1,1,1,1,2,2,2,2,2
Table 13. Examples of chanprof(:) inputs for RTTOV.
7.5. Specifying surface emissivity

This section describes how to specify surface emissivity when calling RTTOV. Section 8.4 provides more information
about the emissivity models and the atlases. RTTOV allows you to provide values for the surface emissivity or
alternatively RTTOV can provide values for surface emissivity. The emissivity naturally depends on the surface type
(land, sea or sea-ice) which is specified in the profiles(:)%skin%surfacetype variable (see Annex O).
RTTOV takes as input a logical array calcemis(:) and an array emissivity(:) of type rttov_emissivity. Both arrays must
be allocated with the same number of elements as chanprof(:). For each radiance j to be calculated, if calcemis(j) is
true RTTOV will supply a surface emissivity internally. If calcemis(j) is false then RTTOV will use the value in
emissivity(j)%emis_in for the surface emissivity: this allows you to supply your own emissivity values to RTTOV.
Microwave emissivities
For MW instruments RTTOV has two internal emissivity models: FASTEM and TESSEM2.
FASTEM provides emissivities for both sea surfaces and for land/sea-ice surfaces. There are 6 versions of the sea
surface model FASTEM available which can be chosen by setting opts%rt_mw%fastem_version to a value between 1
and 6. FASTEM-6 is the default and is the recommended version for channels below 200GHz. For channels above
200GHz TESSEM2 is recommended – see below. Although FASTEM-6 is recommended you may wish to carry out
experiments to see if an earlier version gives better results for your application as some users have found, for example,
FASTEM-3 to yield better results. The older FASTEM versions are included for users who are happily using them and
do not wish to change.
The profile variables used by FASTEM for sea surface emissivities are the 10m wind speed, skin temperature, salinity,
zenith angle and azimuth angle. By default FASTEM calculates a value for the ocean surface foam fraction: you can
optionally supply an explicit value for the foam fraction in the profile structure by setting
opts%rt_mw%supply_foam_fraction to true. For land and sea-ice surfaces FASTEM provides emissivities according
to the parameters specified in profiles(i)%skin%fastem(1:5). Section 8.4 provides more information about the
different versions of the FASTEM sea surface model and the land/sea-ice emissivity parameters.
TESSEM2 is a sea-surface emissivity model (Prigent et al, 2016) which is recommended specifically for use with
instrument channels above 200GHz (in particular the Metop-SG ICI instrument) although it can be used for any
instrument. The profile variables used by TESSEM2 are the 10m wind speed, skin temperature, salinity and zenith
angle (note there is no azimuthal dependence). TESSEM2 is selected for sea-surface emissivities by setting
opts%rt_mw%fastem_version to zero. Note that for land and sea-ice surface types the FASTEM land/sea-ice model
is still used where calcemis(:) is true.
Infrared emissivities
For IR instruments RTTOV has two internal sea surface emissivity models: ISEM and IREMIS.
The ISEM model was in previous RTTOV versions: emissivity is parameterised only in terms of zenith angle. RTTOV
v12 introduces a new more physically-based model, IREMIS (described in the RTTOV v12 Science and Validation
Report), which uses zenith angle, 10m wind speed and skin temperature as inputs. The IR emissivity model is selected
via opts%rt_ir%ir_sea_emis_model: 1 selects ISEM and 2 (the default) selects IREMIS. PC-RTTOV (see section
47
Version : 1.3
Date : 05/03/2019
8.8) currently uses its own emissivity model which uses zenith angle and 10m wind speed as inputs: RTTOV
automatically selects this emissivity model for PC-RTTOV when calcemis(:) is true over sea surfaces.
Over land and sea-ice emissivities are assigned default values of 0.98 (land) and 0.99 (sea-ice) where calcemis(:) is
true.
Tangent linear, adjoint and Jacobian emissivities
For the tangent linear model the emissivity perturbation is calculated within RTTOV when using FASTEM, TESSEM2
or IREMIS. Any input emissivity perturbation in emissivity_tl(:)%emis_in is ignored. ISEM depends only on zenith
angle and as such is inactive in the TL, AD and K models. When using ISEM or when supplying emissivities to
RTTOV you can specify input perturbations in emissivity_tl(:)%emis_in. The emissivity perturbations used in the TL
calculations are output in emissivity_tl(:)%emis_out. Table 14 summarises the options for the input of surface
emissivity including the tangent linear emissivity output.
In the adjoint and Jacobian models the emissivity_ad/k(:)emis_in output data are zero when using FASTEM,
TESSEM2 or IREMIS. If you use ISEM or you supply emissivities to RTTOV then emissivity_ad/k(:)%emis_in
contain the emissivity adjoint/Jacobian on exit.
Land surfaces
RTTOV provides IR and MW land surface emissivity atlases. These are called externally to RTTOV and the
emissivities are passed to RTTOV in emissivity(j)%emis_in with calcemis(:) set to false. Each atlas provides monthly
climatological emissivity values. There are three IR and two MW emissivity atlases each of which has slightly different
inputs, outputs and configuration options.
UWIRemis, CAMEL 2007, and CAMEL climatology IR emissivity atlases

In the IR there are three atlases: the University of Wisconsin UWIRemis atlas (first included in RTTOV v10) described
in Borbas et al. (2010), the CAMEL 2007 atlas (first included in RTTOV v12.1) described in the RTTOV v12 Science
and Validation Report. and the CAMEL climatology atlas (new in v12.3) described in Borbas and Feltz (2019). These
atlases work in very similar ways, but the CAMEL 2007 and CAMEL climatology atlases both have a resolution of
0.05° in lat/lon while the UWIRemis atlas resolution is 0.1°. The atlases can optionally include a correction for zenith
angle effects (see Borbas 2014): this is specified when the atlases are initialised. By default the atlases, once loaded, can
be used with any IR instrument. However if you are only using the atlas with a single instrument you can supply the
corresponding coefficients structure to the atlas initialisation call. This allows some pre-calculations to be done that
result in emissivities being returned much more rapidly from the atlas: it is recommended to use this option where
possible, but the resulting atlas data can only be used with the instrument for which they were initialised. The PC-
RTTOV land/sea coefficients were trained using the UWIRemis atlas so it is recommended (though not strictly
mandatory) to use this atlas to provide land and sea-ice emissivities for PC-RTTOV.
Surface type:
- sea: returns negative values
- sea-ice: a fixed ice emissivity spectrum is used
- land: returns emissivities (where it has data) linearly combined with a snow emissivity spectrum according to
the skin%snow_fraction profile variable (see section 8.4 for more information about the treatment of snow).
Inputs:
- mandatory: latitude, longitude, skin%surftype
- skin%snow_fraction (this modifies emissivities if the value is >0, but see section 8.4 for more information
regarding the CAMEL climatology atlas)
- zenangle, sunzenangle (only required if angular correction was requested; sunzenangle only needs to be <85º
(day) or >85º (night), the specific value does not matter for the atlas)
Outputs:
- emissivities (negative values where atlas has no data)
- optional: emissivity standard deviation (requires a flag at setup, requires more memory)
- optional: quality flag
48
Version : 1.3
Date : 05/03/2019
TELSEM2 MW atlas and interpolator

This is an updated version of the TELSEM atlas and interpolator included in previous versions of RTTOV. TELSEM2
(Wang et al, 2016) is valid for all MW satellite instruments that RTTOV can simulate including Metop-SG ICI. When
loaded the atlas data may be used with any MW instrument. TELSEM2 includes emissivities for climatological sea-ice
and therefore it ignores the specified surftype and simply returns a value if it has data. The native resolution of the
emissivities is 0.25° lat/lon, but optionally a different resolution may be specified to return emissivity values integrated
over a larger area. For frequencies below 19GHz the atlas returns the emissivity value for 19GHz. For frequencies
above 85GHz the atlas returns the emissivity at 85GHz except for certain classes of sea-ice (see the reference for more
information about this atlas).
Surface type:
- sea/sea-ice/land: returns emissivity values if it has data, otherwise negative values.
Inputs:
- mandatory: latitude, longitude, zenangle
Outputs:
- optional: emissivity standard deviation for each channel
- optional: emissivity covariance matrix for all channels specified (one matrix per profile)
CNRM MW atlas
The CNRM MW atlas is described in Karbou et al. (2006) and Karbou et al. (2010). Unlike the other atlases the CNRM
atlas is only compatible with a subset of satellite instruments. These are: AMSU-A, AMSU-B/MHS, SSMI/S and
ATMS. In addition the CNRM atlas datasets are generated from data gathered over a specific year. The year is specified
when setting up the dataset. Currently data are available for 2015 (the default if unspecified when setting up the atlas)
and 2014.
Surface type:
- sea/sea-ice: returns negative values
- land: returns emissivity values if it has data, otherwise negative values.
Inputs:
- mandatory: latitude, longitude, zenangle, skin%surftype
Outputs:
The input variables used by each atlas are summarised in Table 15.
A common interface is provided to both IR and MW atlases. Three subroutines are used to access the atlases:
 rttov_setup_emis_atlas : this allocates the necessary arrays and reads the data for a given atlas for a given
month. In some cases the atlas data are specific to a particular sensor.
 rttov_get_emis : this returns surface emissivity values at a given latitude/longitude for the specified channels
of the given instrument. This is called when populating the emissivity(:)%emis_in array for input to RTTOV.
 rttov_deallocate_emis_atlas : deallocates memory used by the atlas and is called once the atlas is no longer
required.
The interface to the atlases has changed in RTTOV v12. To initialise data for an emissivity atlas you must first declare a
variable of type rttov_emis_atlas_data which is defined in the module mod_rttov_emis_atlas. This variable is passed
into each of the above subroutines and it contains the atlas data for a specific month and (where relevant) for a specific
instrument. This enables you to allocate data for multiple months and/or instruments simultaneously which makes the
atlases more flexible.
The interfaces of these subroutines are described in Annex F. Note that if an atlas does not contain emissivity data for
the given surface type or lat/lon location it will return negative values for the corresponding emissivities. You should
check the returned emissivities and handle negative values accordingly, for example by setting the corresponding
49
Version : 1.3
Date : 05/03/2019
elements of calcemis(:) to true. An example of using the atlas subroutines with a forward model call can be found in the
program src/test/example_atlas_fwd.F90.
Note that if calculations are only being done for visible/near-IR channels (i.e. channels with no significant thermally
emitted component, with wavelengths less than 3µm) the calcemis and emissivity arguments may be omitted in the call
to RTTOV. When simulating both solar and thermally emissive channels it is recommended to set
emissivity(:)%emis_in to zero for any channels with wavelengths less than 3µm.
calcemis Input ε Forward Output ε Tangent Linear Output ε

Land/sea-ice/ISEM: no ε
calculated.
Land=0.98/sea-ice=0.99
Infrared IREMIS: ε computed from u, v,
True 0 Sea = εISEM or εIREMIS
channels sst about εIREMIS.
or εPC-RTOV
PC-RTTOV: ε computed from u,
v about εPC-RTTOV.
False ε atlas/user ε atlas/user No ε calculated
Land/sea-ice: ε about εFASTEM
Land/sea-ice: computed from
Sea: ε computed from
coefs in
Microwave True 0 prof % skin % fastem(1:5) u, v, sst, salinity (and
channels Sea = ε FASTEM or ε TESSEM2 optionally foam_fraction) about
εFASTEM or ε TESSEM2
False ε atlas/user ε atlas/user No ε calculated.
Table 14. Input and output values of ε and ε arrays for infrared and microwave channels for forward and tangent
linear surface emissivity routines.
7.6. Specifying surface reflectance for solar simulations

Solar simulations are described in section 8.2. The input and output of surface reflectances is very similar to that for
emissivities and again depends on the surface type (land, sea or sea-ice) which is specified in the
profiles(:)%skin%surfacetype variable (see Annex O). RTTOV takes bi-directional reflectance function (BRDF)
values as input. These are defined as the ratio of out-going radiance (towards the satellite) to incoming solar irradiance
(treating the sun as a point source).
Analogously to the emissivity inputs, RTTOV takes a logical array argument calcrefl(:) and an array argument
reflectance(:) of type rttov_reflectance, both of which must be allocated with the same number of elements as
chanprof(:). Where calcrefl(:) is true RTTOV will provide surface BRDF values as follows.
Over sea surfaces RTTOV calculates the reflectance due to sun-glint as described in Matricardi (2003). RTTOV v12.2
introduced a new option opts%rt_ir%solar_sea_brdf_model between 1 the JONSWAP wave spectrum (Hasselman et
al, 1973), as implemented in previous RTTOV versions and 2 the Elfouhaily et al 1998 wave spectrum, new in v12.2.
JONSWAP remains the default, but the Elfouhaily wave spectrum is recommended as it significantly reduces bias
compared to observations. The profile variables used are the 10m wind speed, wind fetch and watertype along with the
satellite and solar zenith and azimuth angles. A BRDF derived from the USGS water reflectance spectra (Clark et al,
2007) is added to this for visible/near-IR channels. These spectra are returned by the BRDF atlas (see below) for sea
surface profiles. The resulting BRDF is appropriate for the direct surface-reflected solar beam, but for downward-
scattered radiation (molecular Rayleigh scattering, aerosol, clouds) this sun-glint model is not appropriate. For
visible/near-IR channels the BRDF for downward scattered radiation is taken from the USGS reflectance spectra, and
for channels with both emissive and solar radiation the BRDF is calculated as (1-emissivity)/π which is consistent with
the treatment of downwelling atmospheric emission. This does not apply to the new DOM solver for which the surface
is treated as strictly Lambertian (see section 8.5).
For land/sea-ice surfaces the BRDF is calculated as (1-emissivity)/π for channels with emissive and solar contributions
and for visible/near-IR channels the surface reflectance is set to 0.3/π or 0.8/π for land and sea-ice surface types
respectively (note these latter values are necessarily crude and it is recommended to supply land surface BRDF values
50
Version : 1.3
Date : 05/03/2019
instead, for example from the BRDF atlas – see below). The same BRDF value is used for the direct solar reflectance
and for downward-scattered radiation.
As for emissivity you can provide your own BRDF values in reflectance(:)%refl_in in which case the corresponding
elements of calcrefl(:) should be set to false. In this case the same input BRDF value is used for the direct solar
reflectance and for downward-scattered radiation. Currently it is not possible to input separate BRDF values for the
direct and downward scattered radiation. For sea surface BRDFs calculated by RTTOV, the output reflectance is the
direct BRDF calculated by the sun-glint model. You cannot access the diffuse BRDF for downward scattered radiation,
but the same fixed reflectance spectra are returned by the BRDF atlas for sea surface profiles so you can obtain the
diffuse BRDF values from the atlas if required (see below).
For channels where you supply BRDF values (i.e. calcrefl(:) is false), you can supply input reflectance perturbations in
reflectance_tl(:)%refl_in. If you are using the solar sea BRDF model the BRDF perturbation is computed from the
relevant input profiles_tl(:) variables and any input reflectance perturbations in reflectance_tl(:)%refl_in are ignored.
The reflectance perturbations used in the TL calculations are output in reflectance_tl(:)%refl_out.
In the adjoint and Jacobian models the reflectance_ad/k(:)refl_in output data are zero when using the solar sea BRDF
model. If you supply BRDFs to RTTOV then emissivity_ad/k(:)%emis_in contain the BRDF adjoint/Jacobian on exit.
RTTOV provides a BRDF atlas (Vidot and Borbas, 2013) which is similar in many ways to the IR emissivity atlases. It
provides monthly climatological land surface BRDF values based on the MODIS BRDF kernel product including
climatological snow. The atlas is called outside of RTTOV and the BRDFs are input to RTTOV in
reflectance(:)%refl_in in with the corresponding elements of calcrefl(:) set to false.
Surface type:
- sea: returns Lambertian BRDFs from USGS ocean or fresh water spectrum (no sunglint)
- sea-ice: returns negative values
- land: returns BRDF values if it has data (inc. climatological snow), otherwise negative values
Inputs:
- mandatory: latitude, longitude, skin%surftype, zenangle, sunzenangle, azangle, sunazangle
- skin%watertype (used for sea surface types)
Outputs:
- optional: bi-hemispherical (black-sky) albedo
- optional: quality flag
The interface to the BRDF atlas is similar to the emissivity atlases:

 rttov_setup_brdf_atlas : this allocates the necessary arrays and reads the data for a given atlas for a given
month. The atlas data may be loaded for a specific sensor or they can be initialised for use with any sensor
with visible channels.
 rttov_get_brdf : this returns surface BRDF values at a given latitude/longitude in the required channels. This
should be called when populating the reflectance(:)%refl_in array for input to RTTOV.
 rttov_deallocate_brdf_atlas : deallocates atlas arrays and is called once the atlas is no longer required.
As for the emissivity atlases the interface for the BRDF has also been modified since RTTOV v11. To initialise data for
the BRDF atlas you must first declare a variable of type rttov_brdf_atlas_data which is defined in the module
mod_rttov_brdf_atlas. This variable is passed into each of the above subroutines and to hold the atlas data for a
specific month and, optionally, a specific instrument.
The interfaces of these subroutines are described in Annex G. Note that if an atlas does not contain BRDF data for the
given surface type or lat/lon location it will return negative values for the corresponding BRDFs. You should check the
returned BRDFs and handle negative values accordingly, for example by setting the corresponding elements of
calcrefl(:) to true. An example of using the atlas subroutines with a forward model call can be found in the program
src/test/example_atlas_fwd.F90.
51
Version : 1.3
Date : 05/03/2019
Note that if solar calculations are not required then the calcrefl and reflectance arguments may be omitted in the call to
RTTOV.
TELSEM2 MW atlas latitude, longitude, zenangle

CNRM MW atlas latitude, longitude, zenangle, skin%surftype
UWIRemis and CAMEL latitude, longitude, skin%surftype, skin%snow_fraction
IR emissivity atlases zenangle, sunzenangle – sat/sun zenith angles only required with optional
angular correction; sunzenangle only needs to be <85º (day) or >85º (night)
BRDF atlas latitude, longitude, skin%surftype, skin%watertype
zenangle, sunzenangle, azangle, sunazangle
Table 15. Profile variables required by the emissivity and BRDF atlases.
7.7. Allocation of trajectory structures.

If multiple calls to RTTOV are being made it may be more efficient to allocate some of the internal data structures
before calling RTTOV and then to deallocate these structures once all calls to RTTOV have been made. This can be
achieved using the optional traj argument to rttov_direct. This can be allocated using the rttov_alloc_traj subroutine
or using the rttov_alloc_direct/tl/ad/k subroutines described in Annex D. The trajectory structure consists of a number
of structures used internally by RTTOV. You should make a second call to rttov_alloc_traj to deallocate the traj
structure once it is no longer required. It is important to note that the traj structure is sized according to the number of
channels, profiles and levels, and according to certain input options such as the active trace gases, so if any of these
values change for any calls to RTTOV a new traj structure must be allocated. It is also required that the coefs variable
(of type rttov_coefs) be declared as a TARGET when passing the traj arguments into RTTOV.
Originally this facility was introduced for super-computers, especially for the NEC SX8, because some Fortran
compilers generate complicated assembly code for memory allocation which can take a lot of execution time. Whether
this capability offers performance benefit is dependent on the compiler and architecture. If you are interested in
minimising run-time it is recommended to carry out a test on your system to see if this offers a benefit: it has been
observed to be beneficial on some Intel/Linux-based systems. Use of the traj structure should never be detrimental to
performance because RTTOV calls the rttov_alloc_traj subroutine internally if the traj argument is not present. Note
that the traj structure may not be used in conjunction with the parallel interface to RTTOV.
7.8. Output arrays from RTTOV v12

The syntax for the call to rttov_direct is given again, this time with the output arguments in bold:
call rttov_direct(errorstatus, chanprof, opts, profiles, coefs, transmission,

radiance, radiance2, calcemis, emissivity, calcrefl, reflectance, aer_opt_param,
cld_opt_param, traj, pccomp, channels_rec)
errorstatus is an integer error return code. If the value of errorstatus is non-zero, a fatal error occurred during
processing (more generally success is indicated by a return code errorstatus_success and failure by errorstatus_fatal,
both of which are contained in the rttov_const module).
Instances transmission and radiance of the derived types rttov_transmission and rttov_radiance should be declared.
Optionally an instance radiance2 of type rttov_radiance2 may be declared. The radiance structure contains the
primary radiance, BT and reflectance outputs (the output reflectance quantity is defined in section 8.2) and is
mandatory, while the radiance2 structure holds secondary radiance outputs which are only calculated by the direct
model (not the TL/AD/K) for non-scattering, non-solar, non-PC-RTTOV calculations. The transmission and radiance
structures may be initialised using the rttov_alloc_direct/tl/ad/k subroutines, or the individual allocation subroutines
rttov_alloc_transmission and rttov_alloc_rad (see Annex D). Annex O defines fully these output radiance and
transmittance structures. Tables 17 and 18 list the output arrays and highlight those most commonly used. The table also
indicates array dimensions for rttov_direct and gradient routines (note that nchanprof is the size of the chanprof(:)
array, nlevel is the number of vertical levels in the profile, and nlayer = nlevel - 1). A number of the radiance outputs
are defined on layers such as rad%overcast. These represent radiances from/to to the level bounding the bottom of
each layer. All values relate to the pressure levels defined by the user except for the layer containing the surface where
the value is the radiance calculated from/to the 2m surface pressure rather than the profile level immediately below. (If
the surface pressure lies on a profile level then there is no anomaly).
52
Version : 1.3
Date : 05/03/2019
Note that the rad%overcast values for channels with significant thermally emitted and solar reflected contributions
contain no reflected solar contribution in accordance with the assumption that the overcast layer is perfectly emissive.
The rad%overcast output is not calculated for visible/IR aerosol or PC simulations. Also note that in the case of
visible/IR aerosol simulations, the “clear-sky” outputs (such as rad%clear and rad%bt_clear) contain the aerosol-
affected radiances.
New in RTTOV v12.2 was the radiance%quality output array. This is a bit mask which is used to flag potential issues
with radiances. If there are no issues the quality value corresponding to a given output radiance is zero. However one or
more bits may be set indicating specific warnings. The bit positions are contained in the rttov_const module and are
listed in Table 16 below. In Fortran you can test for a specific bit using the Fortran BTEST function. For example to
check whether the regression limits were exceeded for channel i you can use:
IF (BTEST(radiance%quality(i), qflag_reg_limit)) THEN…
There is a new subroutine rttov_print_radiance_quality (see Annex N) which prints out a human-readable
interpretation of any warning flags set in a given radiance%quality array element which may be useful for debugging.
The transmission structure contains separate transmittances for the thermally emitted and solar calculations. This is
because the solar calculations are derived from different optical depth predictor calculations. For mixed thermal+solar
channels the thermal and solar transmittances for the surface-satellite path will therefore not be exactly equal, and in the
case of channels using Planck-weighted coefficients may be significantly different.
The emissivity(:)%emis_out and reflectance(:)%refl_out arrays contain the surface emissivities and surface BRDFs
used by RTTOV.
Finally, if Principal Components calculations are being performed (section 8.8), you should allocate an instance of the
pccomp structure (derived type rttov_pccomp) using the rttov_alloc_pccomp subroutine (Annex D). This structure
will contain the computed PC scores, and (if opts%rt_ir%pc%addradrec is true) the reconstructed radiances. This
structure is also defined in Annex O. The pccomp argument is not required if opts%rt_ir%pc%addpc is false.
Bit position Description

qflag_reg_limits Gas regression limits exceeded in optical depth computation.
qflag_pc_aer_reg_limits PC-RTTOV aerosol regression limits exceeded.
qflag_mfasis_zenangle MFASIS maximum zenith angle exceeded (by satellite or solar zenith angle).
qflag_mfasis_sumzenangle MFASIS maximum sum of satellite and solar zenith angles exceeded.
qflag_mfasis_hiscaangle MFASIS maximum scattering angle exceeded.
qflag_mfasis_loscaangle MFASIS minimum scattering angle exceeded.
qflag_mfasis_geometry_bounds MFASIS scattering angle outside LUT limits.
qflag_mfasis_opdpedia_bounds MFASIS optical depth or effective diameter value outside LUT limits.
Table 16. Bit positions (contained in the rttov_const module) associated with RTTOV warnings. See section 7.3 for
information on gas regression limits, section 8.8 for information on PC-RTTOV aerosol regression limits, and section
8.5 for information about the MFASIS flags.
53
Version : 1.3
Date : 05/03/2019
rttov_radiance Radiances in units of mW/cm-1/sr/sq.m

Type Array name Contents
real clear(nchanprof) Clear sky top of atmosphere radiance output for
each channel
real total(nchanprof) Clear+cloudy top of atmosphere radiance for given
cloud top pressure and fraction for each channel
(simple cloud scheme) or full cloudy radiance (if
opts%rt_ir%addclouds or opts%rt_ir%addaerosl is
true)
real cloudy(nchanprof) Cloudy top of atmosphere radiance for 100%
fraction for each channel at given cloud top
pressure (for simple cloud scheme) or same as total
(if opts%rt_ir%addclouds or opts%rt_ir%addaerosl
is true).
real overcast(nlayer,nchanprof) Level to space overcast radiance at the level
bounding the bottom of each layer. For IR channels
(wavelengths above 3µm) this assumes an opaque
black cloud. For solar-only channels at wavelengths
less than 3µm this consists of reflected solar
radiation according to assumptions described in
section 8.3. This is not calculated for PC or
visible/IR aerosol simulations or scattering
simulations using the DOM solver..
rttov_radiance Brightness Temperatures deg K

real bt(nchanprof) BT equivalent to total (clear+cloudy) top of
atmosphere radiance output for each channel
(channels with thermally emitted contribution only)
real bt_clear(nchanprof) BT equivalent to clear top of atmosphere radiance
output for each channel (channels with thermally
emitted contribution only)
Table 17. Main RTTOV output arrays. The green rows are those most commonly used. (Continued below)
54
Version : 1.3
Date : 05/03/2019
rttov_radiance Bi-directional reflectance factors (BRF) unitless

real refl(nchanprof) BRF equivalent to total (clear+cloudy) top of
(solar-affected channels only).
real refl_clear(nchanprof) BRF equivalent to clear top of atmosphere radiance
output for each channel (solar-affected channels
only).
rttov_transmission Transmittances 0-1

real tau_total(nchanprof) Transmittance from surface to top of atmosphere
(TOA). Only populated for channels with a
significant thermally emitted contribution.
real tau_levels(nlevel, nchanprof) Transmittance from each user pressure level to
TOA. Only populated for channels with a significant
thermally emitted contribution.
real tausun_total_path2(nchanprof) Transmittance for combined sun-surface-satellite
path. Only populated for solar-affected channels.
real tausun_levels_path2(nlevel, Transmittance from TOA to each user pressure level
nchanprof) to TOA along combined sun-surface-satellite path.
Only populated for solar-affected channels.
real tausun_total_path1(nchanprof) Transmittance from surface to TOA. Only populated
for solar-affected channels.
real tausun_levels_path1(nlevel, Transmittance from each user pressure level to
nchanprof) TOA. Only populated for solar-affected channels.
Output Emissivity 0-1

real emissivity%emis_out(nchanprof) Emissivity vales used in RTTOV calculation (same as
input emissivity values if calcemis is false).
Output surface BRDF
real reflectance%refl_out(nchanprof) Surface BRDF values used in RTTOV calculation (same
as input BRDF values if calcrefl is false).
Table 17 (continued). Main RTTOV output arrays. The green rows are those most commonly used. See Annex O for
more details about the output transmittances.
rttov_radiance2 Radiances in units of mW/cm-1/sr/sq.m

real upclear(nchanprof) Clear sky upwelling radiance at top of atmosphere
including surface emission term, but omitting
downwelling reflected radiance term.
real dnclear(nchanprof) Clear sky downwelling radiance at surface.
real refldnclear(nchanprof) Reflected clear sky downwelling radiance
contribution to top of atmosphere radiance.
real up(nlayer,nchanprof) Summed upwelling atmospheric emission term at
top of atmosphere for layers down to the level
bounding the bottom of each layer.
real down(nlayer,nchanprof) Summed downwelling atmospheric emission term
at bottom of layer for layers down to the level
bounding the bottom of each layer.
real surf(nlayer,nchanprof) Radiance emitted by a black cloud at the level
bounding the bottom of each layer; for the surface
layer this is evaluated for the surface skin
temperature.
Table 18. Secondary RTTOV output radiance arrays.
55
Version : 1.3
Date : 05/03/2019
7.9. Calling the tangent linear (TL), adjoint (AD) and Jacobian (K) models
The tangent linear (TL), adjoint (AD) and Jacobian (K) models are typically used in assimilation and retrieval
applications. They provide information about the gradient of the forward model given a particular input profile. The TL,
AD and K models share some features:
 All three models take the same input atmospheric profile and surface parameters as the direct model
 All three models output the direct model radiances, BTs and reflectances
 As for the direct model all three models may be run using multiple threads via OpenMP (see section 7.10)
 Not all profile variables are “active” in the TL, AD and K models: Table 12 indicates which variables can be
perturbed in the TL and hence which may be non-zero in the Jacobian and AD output.
RTTOV TL model
The RTTOV TL model (see Annex K for the interface) takes as input a profiles_tl(:) profile structure containing a set
of perturbations to the input profile. The TL model calculates the linearisation of the direct model evaluated for the
given input profile and it outputs the change in satellite seen radiances that result from the given profile perturbation for
this linearisation.
The profiles_tl(:) array should be allocated to be the same size as the input profiles(:) array. As an example the
subroutine src/test/rttov_make_profile_inc.F90 is used to generate profile perturbations for the RTTOV test suite. For
RTTOV-SCATT TL (section 8.7, Annex M) the cld_profiles_tl(:) array should be allocated to be the same size as the
input profiles(:) and cld_profiles(:) arrays and it should be initialised with the cloud profile perturbations.
In RTTOV v12.2 the visible/IR aerosol/cloud explicit optical properties (method 2 described in sections 8.5 and 8.6)
were enabled as active variables in the TL model. You can optionally supply perturbations to the optical properties via
the aer/cld_opt_param_tl structure. Note that in this case the aer/cld_opt_param_tl argument to the tangent linear
model is optional: if omitted the optical properties are treated as static variables as in earlier versions of RTTOV.
For channels where you supply emissivity values to RTTOV, you can specify an emissivity perturbation in
emissivity_tl(:)%emis_in. This also applies when using ISEM because this is a static model depending only on zenith
angle. For channels where you are using FASTEM, TESSEM2 or IREMIS, RTTOV calculates the emissivity
perturbation from the relevant input profiles_tl(:) variables and any input emissivity perturbation is ignored. The
emissivity perturbation used in the TL calculation is written to emissivity_tl(:)%emis_out.
Similarly, for solar calculations you can supply the BRDF perturbation in reflectance_tl(:)%refl_in for channels for
which you input BRDF values. When using the sea surface BRDF model the reflectance perturbation is computed
internally and any input perturbation is ignored. The BRDF perturbation used in the TL calculation is written to
reflectance_tl(:)%refl_out.
The output radiance, brightness temperature and reflectance perturbations are contained in the radiance_tl structure.
For PC-RTTOV (see section 8.8) the output perturbations in PC scores and optionally reconstructed radiances are in the
pccomp_tl structure.
RTTOV AD and K models
The file src/test/example_k.F90 provides en example of calling the RTTOV K model for a clear-sky simulation.
The RTTOV AD model (see Annex L for the interface) takes as input the gradient of some scalar function (e.g. a cost
function) with respect to the satellite seen radiance (or BT) in radiance_ad and it outputs the gradient of the same
scalar function with respect to the profile variables in profiles_ad(:). In this case the profiles_ad(:) array must be
allocated to be the same size as the input profiles(:) array.
The RTTOV K model (see Annex J for the interface) takes as input a radiance (or BT) perturbation in radiance_k and
it outputs the gradient of each forward model radiance with respect to each input profile variable evaluated for the given
input profile in profiles_k(:). The gradients are scaled by the perturbations in radiance_k: typically the input
perturbation is set to a value of 1 in radiance or BT units for each channel. For the K model, profiles_k(:) must be
56
Version : 1.3
Date : 05/03/2019
allocated to be the same size as the chanprof(:) array: each element of the array contains the gradient of the forward
model for the corresponding channel.
Aside from the size of the profiles_ad/k(:) arrays the AD and K models are very similar. The input perturbations in
radiance_ad/k may be supplied either in radiance or BT: if the opts%rt_all%switchrad flag is false the input
perturbations must be in radiance_ad/k%total(:) (in units of radiance), otherwise the input perturbations must be
specified in radiance_ad/k%bt(:) (in Kelvin). For solar-only channels (those with wavelength < 3µm) the input
perturbation is always in radiance regardless of the setting of switchrad. Note that there is never any problem in
providing perturbations in both the total(:) and the bt(:) arrays: RTTOV uses the correct perturbation for each channel
according to the setting of switchrad.
In general all AD/K input and output structures (aside from the input perturbations) should be initialised to zero before
calling the AD/K models. It is important to remember to reinitialise these structures before each call when making
multiple calls to the AD/K subroutines. The following structures and arrays should always be set to zero:
 profiles_ad/k structure (use rttov_init_prof subroutine)
 transmission_ad/k structure (use rttov_init_transmission subroutine)
 radiance_ad/k structure (use rttov_init_rad subroutine before specifying the perturbations in total(:) and/or
bt(:); for PC-RTTOV all members of the structure should be set to zero – see below)
 emissivity_ad/k structure (set both emis_in(:) and emis_out(:) member arrays to zero)
 reflectance_ad/k structure (set both refl_in(:) and refl_out(:) member arrays to zero)
 , see Annex D)
In RTTOV v12.2 the visible/IR aerosol/cloud explicit optical properties (method 2 described in sections 8.5 and 8.6)
were enabled as active variables in the adjoint and Jacobian models. When using these arguments you should ensure the
aer/cld_opt_param_ad/k structures are zeroed using rttov_init_opt_param subroutine with the zero_only argument.
On output these structures contain the adjoint/Jacobian of the optical properties. Note that in this case the
aer/cld_opt_param_ad/k arguments to the adjoint and Jacobian models are optional: if omitted the optical properties
are treated as static variables as in earlier versions of RTTOV.
See Annex D for all rttov_init_* subroutine interfaces.
For channels where you supply an input emissivity or for which you use ISEM, the emissivity_ad/k(:)%emis_in array
contains the surface emissivity AD/Jacobians on exit. For channels where you use FASTEM, TESSEM2 or IREMIS
this array is zero because the emissivity is a function of input profile variables.
Similarly, for channels where you supply an input BRDF, the reflectance_ad/k(:)%refl_in array contains the surface
BRDF AD/Jacobians on exit. For channels where you use the solar sea BRDF model this array is zero.
For RTTOV-SCATT (section 8.7, Annex M) the rttov_scatt_ad subroutine doubles as the adjoint and Jacobian model.
If the size of the profiles_ad(:) array is the same size as the input profiles(:) array then it calculates the adjoint.
Otherwise profiles_ad(:) must be the same size as the chanprof(:) array in which case rttov_scatt_ad calculates the
Jacobian. In either case the cld_profiles_ad(:) array must be the same size as profiles_ad(:) and on exit it contains the
adjoint or Jacobian of the cloud profile variables. The cld_profiles_ad(:) array should always be initialised to zero
before calling rttov_scatt_ad.
When calling rttov_scatt_ad the behaviour is as if the switchrad flag is always true and the input perturbations are
always in terms of brightness temperature. This is always true, even when internally computations are being done as
radiances. Hence for RTTOV-SCATT the input perturbations should be set in radiance_ad%bt(:).
For PC-RTTOV (see section 8.8 for details), the input perturbation must be specified in the pccomp_ad/k structure: the
member to which the perturbation applies depends on the setting of opts%rt_ir%pc%addradrec and
opts%rt_all%switchrad. If reconstructed radiances are not required (opts%rt_ir%pc%addradrec is false), then the
input perturbation should be specified in pccomp_ad/k%pcscores. If reconstructed radiances are required, the input
perturbation should be specified in pccomp_ad/k%total_pccomp if opts%rt_all%switchrad is false, or
pccomp_ad/k%bt_pccomp if opts%rt_all%switchrad is true. Before calling the PC-RTTOV AD/K models, all
elements of the radiance_ad/k structure should be initialised to zero.
57
Version : 1.3
Date : 05/03/2019
The PC-RTTOV AD and K models differ slightly in the output: for the AD model the output adjoint is in the
profiles_ad(:) array as for standard RTTOV.
For the K model, profiles_k(:) contains the Jacobians for the PC predictor channel set. These are always in terms of
radiances (not brightness temperatures), regardless of the setting of opts%rt_all%switchrad.
If opts%rt_ir%pc%addradrec is false the PC Jacobians are output in the profiles_k_pc(:) array. This must have size
equal to the number of PC scores multiplied by the number of profiles. These are gradients of PC scores with respect to
the profile variables and in this case the setting of opts%rt_all%switchrad has no impact on the units of the Jacobians.
Alternatively if opts%rt_ir%pc%addradrec is true the PC Jacobians are output in the profiles_k_rec(:) array which
must have size equal to the number of reconstructed radiance channels multiplied by the number of profiles. In this case
the setting of opts%rt_all%switchrad determines the units of the output (radiances or brightness temperatures).
Note that only one of profiles_k_pc(:) or profiles_k_rec(:) should be supplied to rttov_k depending on the setting of
opts%rt_ir%pc%addradrec.
7.10. Multi-threaded execution

If RTTOV is compiled with OpenMP (see section 5) then substantial gains in execution time can be made by calling
RTTOV through the parallel interface. The interfaces to the parallel subroutines are almost identical to rttov_direct,
rttov_tl, rttov_ad and rttov_k: the subroutines are named rttov_parallel_direct, rttov_parallel_tl and so on. The
only difference is a final optional parameter named nthreads which specifies the number of threads to use. For
RTTOV-SCATT there are similar rttov_parallel_scatt, rttov_parallel_scatt_tl, and rttov_parallel_scatt_ad
interfaces.
Each thread may be assigned simulations for multiple channels (possibly across multiple profiles), but the smallest unit
of computation for a single thread is a simulation for one channel for one profile. Therefore to make use of N threads,
you must be simulating at least N individual channel radiances (for one or more profiles) and to obtain optimal
performance you should usually be simulating many channels and/or profiles with each call to the parallel interface. For
PC-RTTOV each thread is assigned at least one profile so you must call the parallel interface for at least N profiles to
make use of N threads for PC simulations. The same applies to polarimetric instruments such as Windsat and COWVR.
If the surface emissivity and/or reflectance arguments are omitted in a call to the parallel interface for a simulation in
which they are mandatory, the code will run using calcemis and/or calcrefl set to true. This is in contrast to the standard
RTTOV interfaces which instead fail with an error message.
7.11. RTTOV performance

This section provides a rough guide to the relative cost of various RTTOV options and simulation types. It is difficult to
be precise as execution times vary greatly depending on the compiler, compilation flags, computer architecture, the
combination of RTTOV options specified at run-time, and the configuration of the input profile. Some measurements of
execution time (in particular comparisons against the previous version of RTTOV) can be found in the RTTOV v12
Performance Test Logs available on the RTTOV web site.
Coefficients, trace gases, solar radiation

Simulations with coefficient files supporting more optional trace gases (related to the version of the optical depth
predictors) will be slower than those with fewer optional gases. Simulations using coefficients based on v7 predictors
are fastest and those using v9 predictors are slowest. MW simulations are generally faster than IR simulations (all other
things being equal) as there are no optional trace gases in the MW simulations. Simulation of visible/near-IR channels
(with solar radiation) is more expensive than for IR channels (no solar radiation). Simulations using coefficient files
based on a larger number of levels (e.g. 101L) run slower than those with coefficients on fewer levels (e.g. 54L).
For MW simulations with CLW the run-time with the Liebe is only a little longer than without CLW. The TKC option
is slightly more expensive than Liebe. However the Rosenkranz option is at least 50% more expensive than Liebe due
to the relative complexity of the parameterisation. If you have more input pressure levels than coefficient levels it is
generally more expensive to run the CLW calculation on the input levels, but this is nevertheless recommended.
58
Version : 1.3
Date : 05/03/2019
Number of input levels, interpolation

Execution time increases with the number of levels in the input profile. Using the RTTOV interpolator increases run-
time. The computational cost depends on the selected interpolation mode: the modes can be ranked from fastest to
slowest as 2, 3, 1, 5, 4. For the direct model and for the TL/AD/K with the lgradp option set to false, the variation in
cost between the modes is not very large. If lgradp is true (so that pressure is an active variable in the TL/AD/K) the
variation in the cost betweens modes is larger. For the K model the execution cost of the interpolator vs no interpolation
is much larger with lgradp true. Some measurements of the performance impact of interpolation vs no interpolation can
be found in the RTTOV v12 Performance Test Log available on the RTTOV web site.
Surface emissivity and reflectance

Use of the internal surface emissivity and BRDF models for sea surfaces is generally more expensive than providing
input values for surface emissivity and reflectance. The computational cost of ISEM is insignificant, and the cost of
IREMIS is very small. For FASTEM, computation cost is modest and generally increases with the FASTEM version
number (FASTEM-6 being the most expensive). The cost of TESSEM2 is approximately half that of FASTEM-6. Use
of the Lambertian surface option also increases run-time slightly. The sea surface BRDF model (either option) is
expensive due to its complexity, though it underwent some optimisation for v12.2. The cost of either model is
comparable to that of the clear-sky radiative transfer calculation itself though the new Elfouhaily option is slightly faster
than the old JONSWAP one.
Cloud/aerosol scattering
The simple cloud scheme is run “for free” with visible/IR clear-sky simulations. The cost of visible/IR scattering
simulations increases with the number of layers containing scattering particles in the input profile. For aerosol
simulations with Chou-scaling, the extra cost over the clear-sky case results mainly from the computation of the layer
aerosol optical depths: the integration of the radiative transfer equation is the same as the clear-sky case. For cloud
simulations it is both the layer optical depth calculations and also the additional cloud columns which are generated by
the cloud overlap scheme which increase run-time. The radiative transfer equation is integrated for every cloud column
and the number of columns can be up to twice the number of cloudy layers.
The new DOM multiple scattering solver is, broadly speaking, at least one or two orders of magnitude more expensive
than Chou-scaling. However there are various factors which affect the speed of the DOM algorithm significantly (the
input cloud profile, the number of DOM streams, etc). In particular, the cost of DOM simulations grows rapidly as the
number of DOM streams increases, and as the number of layers containing scattering materials increases. The second of
these factors particularly affects the solver for solar radiances: solar DOM simulations run significantly faster with
fewer scattering layers.
Aerosol/cloud scattering simulations with explicit optical properties are slightly faster than those which use the pre-
defined optical properties. For DOM simulations the effect of this becomes less noticeable as the cost of the radiative
transfer calculation is larger. Using the Baran ice scheme is slightly more expensive than using the SSEC ice properties
for Chou-scaling. It is significantly more expensive when using DOM and/or for solar simulations because the phase
function needs to be explicitly reconstructed for the Baran scheme which is a relatively expensive calculation.
The MFASIS direct model is broadly comparable in speed to DOM with 4 streams, although as the number of cloudy
layers increases DOM quickly becomes significantly more expensive while the impact on MFASIS run-times is much
more modest. For the TL model and especially the AD and K models, MFASIS is much faster than DOM.
RTTOV-SCATT is, broadly speaking, a few times slower than an otherwise equivalent (clear-sky) RTTOV simulation.
PC-RTTOV/HTFRTC
For simulating a full spectrum of a hyperspectral sounder, the PC-based models are substantially faster than standard
RTTOV. The PC-RTTOV K model has known performance issues, particularly with reconstructed radiances: these will
be addressed in a future release. The run-time of the HTFRTC model is broadly comparable to PC-RTTOV for
equivalent simulations, but as the number of variable trace gases increases, PC-RTTOV becomes relatively faster. The
new simple cloud option and (in particular) the new overcast option for HTFRTC both significantly increase the run-
time.
59
Version : 1.3
Date : 05/03/2019
7.12. Summary of steps for running RTTOV v12

You need to ensure the following in your program which calls RTTOV:
 Create an instance of the rttov_options type and set any of the member flags as required. The full list of options is
given in Annex O.
 Optional: you may wish to call rttov_errorhandling to specify the logical unit to use for error messages.
 An instance of the rttov_coefs type should be populated by calling rttov_read_coefs.
 Optional: you may wish to call rttov_user_options_checkinput to ensure the selected options are consistent with
the coefficient file.
 Allocate the input/output structures to RTTOV with the number of channels and profiles you want to run with (e.g.
by calling rttov_alloc_direct/tl/ad/k, see Annex D). If the emissivity and/or BRDF atlases are required these
should be initialised now (see Annexes F and G). See src/test/example_*.F90 for example code detailing the input
variables required for RTTOV.
 Initialise the profile structure with your atmospheric profile. This is shown in Table 12 and listed in Annex O.
 Optional: you may wish to check the input profiles for unphysical or out-of-specification values before calling
RTTOV. This can be achieved using the rttov_user_profile_checkinput subroutine (Annex N). If this is used,
opts%config%do_checkinput can be set to false. You may also find the rttov_print_profile subroutine (Annex
N) useful for debugging purposes: this prints out the contents of a single profile structure.
 Initialise the chanprof array with the channel and profile indexes as described in section 6.4.
 You may give a surface emissivity value in emissivity(:)%emis_in for each radiance calculation (for example
from one of the land surface emissivity atlases), in which case you must set calcemis(j) to false for the desired
channels. Alternatively you may let the code compute it by the use of the internal emissivity models by setting the
appropriate calcemis(j) to true.
 For solar calculations you may supply a surface BRDF value in reflectance(:)%refl_in for each radiance
calculation (for example from the land surface BRDF atlas), in which case you must set calcrefl(j) to false for the
desired channels. This is recommended for land surfaces. Alternatively you may let RTTOV compute/select BRDF
values internally by setting the appropriate calcrefl(j) to true.
 Call RTTOV (rttov_direct) with the input/output variables and with the coefficient structure corresponding to the
instrument you want to simulate.
 When all RTTOV calls are made, then you should free memory by de-allocating the various input and output
structures with the rttov_dealloc_coef , rttov_alloc_direct/tl/ad/k (see Annexes C and D). If the emissivity and or
BRDF atlases were initialised, they should also be de-allocated now (see Annexes F and G).
 Most user-level RTTOV subroutines return an error status. This variable should be tested after each call: non-zero
values indicate that an error occurred.
 If you want to run the cloud or aerosol options for visible/ IR sensors follow the guidance in sections 8.5/8.6.
 For microwave scattering calculations, the rttov_scatt routines are a level up from rttov_direct but they have a
similar calling structure and arrays to fill (see section 8.7 and Annex M). The example program supplied
example_rttovscatt.F90 can be used as an example and similar rules apply to calling rttov_direct.
60
Version : 1.3
Date : 05/03/2019
8. Details of specific RTTOV capabilities

8.1. Simulation of clear air radiances for infrared and microwave channels
This is the simplest mode of operation for RTTOV. The instructions given in section 7 describe the necessary steps for
carrying out IR and MW clear-sky simulations. A short summary is given here: these steps are common to the majority
of RTTOV simulations.
An instance of the RTTOV options structure is declared which is used to configure various aspects of the simulation.
The options are described in full in Annex O. The coefficient file for the instrument of interest is then read in. For IR
and MW simulations, v7, v8 or v9 predictor files may be used (see Table 4).
The atmospheric profile and surface parameters are stored in the RTTOV profile structure. Section 7.3 describes this in
detail. In the case of IR sensors the optional variable trace gases available depend on the coefficient file (see Table 4).
For MW sensors there are no optional trace gases, but cloud liquid water may be included as an absorbing medium. An
instance of the chanprof structure is populated with the profile and channel numbers to be simulated as described in
section 7.4. The input surface emissivity may be specified via the RTTOV emissivity structure or RTTOV can calculate
surface emissivities internally if required. See sections 7.5 and 8.4.
Finally RTTOV is invoked by calling one of the main subroutines (or their parallel counterparts for multi-threaded
execution via OpenMP): these are described fully in Annexes I, J, K and L. The outputs from RTTOV are described in
section 7.8.
8.2. Simulation of clear air radiances for visible and near-infrared channels
RTTOV can also simulate clear-sky radiances for visible and near-IR channels. To include solar radiation the logical
flag opts%rt_ir%addsolar must be set to true (see Annex O for full description of the options structure). With this
flag set, the contribution of solar radiation is included in all channels with wavelengths below 5µm. Note that for
channels below 3µm emission is ignored so the only contribution is solar radiation.
In addition to the profile variables required for IR and MW simulations, the solar zenith angle, and satellite and solar
azimuth angles must also be specified in the input profile structure in profiles(:)%sunzenangle, profiles(:)%azangle
and profiles(:)%sunazangle respectively. Note that solar radiation is only included if the solar zenith angle is less than
85o. Figure 4 illustrates the definition of azimuth angle.
To satellite/sun
North
Zenith
angle
East
Azimuth angle
Figure 4. Azimuth angles are defined as the angle in the clockwise direction between North and the projection of the
view-path or sun-surface path onto the horizontal plane at the surface.
61
Version : 1.3
Date : 05/03/2019
Solar simulations can only be performed with v9 predictor coefficient files (see Table 4). These files contain a
SOLAR_SPECTRUM section which includes a flag for each channel which is 0 for “thermal” channels (no significant
solar contribution), 2 for “solar” channels (no significant thermally emitted contribution) and 1 for channels with both
thermally emitted and solar contributions. These flags are read into the coefficients structure in
coefs%coef%ss_val_chn(:) which has one element per channel read from the coefficient file. It is important to note
that for the sake of efficiency no thermally emitted contributions are calculated for the solar channels (i.e. those with
wavelengths less than 3µm). Users have noted that a solar contribution due to sun glint can be detected even in window
channels in the thermal IR. By default these channels are considered thermal-only, but it is possible to include the solar
contribution for them by setting the element of coefs%coef%ss_val_chn(:) corresponding to the channel in question to
1 after the coefficient file has been read.
The output radiance structure contains arrays for satellite-seen reflectances. These are bi-directional reflectance factors
(BRFs) calculated as the ratio of the satellite seen radiance to that which would result if the surface was a perfect
Lambertian reflector. BRFs are only calculated for channels with a solar contribution. Similarly, the output brightness
temperature arrays will not be populated for channels with no thermally emitted contribution, and you should be careful
in interpreting the output brightness temperatures for channels with both solar and emitted contributions since these
may differ significantly from physical temperatures.
The SOLAR_SPECTRUM section in the RTTOV coefficient files also provides the top of atmosphere band-integrated
solar irradiance at 1 AU (astronomical unit) in units of mW/m2/cm-1. These values are read into the coefs%coef%
ss_solar_spectrum(:) array and are calculated using the AER solar source function software (based on Kurucz, 1992).
These irradiances are used for the solar calculations: it is possible to change the values in the coefficient file or directly
in the coefs%coef%ss_solar_spectrum(:) array after reading the coefficients if desired. If the profile(:)%date(:) array
has been populated with a date for the profile (year, month, day) then these are used to adjust the band solar irradiances
according to the Earth-sun distance. If the date(:) contains a date earlier than or equal to the default value (1/1/1950),
no adjustment is performed. Note the date is only used to adjust the solar spectrum value for the Earth-sun distance
according to the month and day of the year.
The clear-sky solar calculations comprise solar radiation reflected from the surface and, for channels with wavelengths
below 2µm, a parameterisation of atmospheric Rayleigh single-scattering. Details of the solar calculations can be found
in the RTTOV v11 Science and Validation Report.
For simulations including solar radiation, the surface bi-directional reflectance function (BRDF) must be specified (see
section 7.6). The BRDF is defined as the ratio of reflected radiance towards the satellite to incoming solar irradiance
(assuming the sun is treated as a point source). In the absence of an atmosphere the output BRF would be equal to the
surface BRDF multiplied by π.
For wind-roughened sea surfaces RTTOV has an internal BRDF model as noted in section 7.6 with two choices for the
wave spectrum parameterisation. To make use of this model additional profile variables should be set: the water type
(fresh or salt water) profiles(:)%watertype, and the 10m wind and wind fetch profiles(:)%s2m%u,
profiles(:)%s2m%v, profiles(:)%s2m%wfetc. See Annex O for details of these quantities in the profile structure and
section 7.6 for more details about the sea-surface reflectances RTTOV calculates.
If calcrefl(:) is set to true for land or sea-ice surfaces a default BRDF of 0.3/π or 0.8/π (respectively) is used for visible
and near-infrared channels: this is naturally very crude and is not generally recommended. If a land/sea-ice surface
emissivity has been supplied for the channel (for example if the channel has significant emitted and solar contributions)
then the BRDF is set to (1-emissivity)/π.
For solar-only channels RTTOV provides a land surface BRDF atlas (Vidot and Borbas, 2013) similar to the IR
emissivity atlas. This takes as input latitude, longitude, satellite and solar zenith and azimuth angles (via the profiles
structure), and provides BRDF values suitable for input to RTTOV. The BRDF atlas interface is described in detail in
Annex G, and when passing these values in to RTTOV the corresponding elements of calcrefl(:) should be set to false.
62
Version : 1.3
Date : 05/03/2019
8.3. Simple cloud

More complicated treatments of cloud and precipitation are available (see sections 8.5 and 8.7), but the simplest cloud
approach described here is applied for all clear-sky visible/IR simulations. Assuming black, opaque clouds at a single
level which fill the radiometer field of view the simulation of cloud affected radiances LCld(v,θ) is defined as:
( , ) =  Cld ( , ) B( ,T Cld ) +  Cld B( ,T)d

Cld 1
L (6)
where τCld (v,θ) is the cloud top to space transmittance and TCld the cloud top temperature specified by the cloud top
pressure in the input state vector. The emissivity of the cloud top is assumed to be unity which is a tolerable assumption
for optically thick water cloud at infrared radiances but not valid for optically thin cloud or any cloud at microwave
frequencies. For partially cloud filled fields of view, the cloudy radiance given by equation 6 is combined with the clear
sky radiance (equation 5) using the input fractional cloud cover N as follows:
L(v,  )  (1  N ) LClr (v,  )  NLCld (v,  ) (7)
This simple cloud scheme is activated by providing a cloud top pressure in profiles%ctp and a non-zero cloud fraction
in profiles%cfraction (with 1 indicating fully overcast). The relevant radiance outputs in the radiance structure are
clear(:) (clear radiance), cloudy(:) (radiance assuming 100% cloud fraction) and total(:) (cloudy radiance with
specified cloud fraction). The cloudy radiance is calculated by interpolating the overcast(:,:) radiances to the cloud top
pressure. Note that the simple cloud scheme is not applied if the full scattering calculations have been activated by
setting opts%rt_ir%addclouds to true.
This simplistic cloud treatment has also been implemented for visible/near-infrared channels with no thermally emitted
component. In this case the cloudy radiance LCld(v,θ, θsol), where θsol is the solar zenith angle, is calculated as the
reflected component of the solar radiation from the cloud top assuming the cloud is optically thick and acts as a
Lambertian reflector with a default BRDF of 0.7/π for wavelengths below 1µm and 0.6/π for wavelengths above 1µm.
If the reflectance(:)%refl_cloud_top array contains values greater than zero, these are used in preference to the default
BRDFs for the corresponding channels (see Annex O for a description of the rttov_reflectance structure). The cloudy
radiance includes the effects of clear-sky Rayleigh scattering above the cloud top. This is a crude treatment of cloud in
visible and near-infrared channels, but may be useful for qualitative applications such as simulated imagery, particularly
for visible channels. Note that refl_cloud_top is not an active variable in the TL/AD/K models.
NB No solar contribution is included in the simple cloudy radiances for channels which have significant thermally
emitted and solar contributions (for example, channels around 3.9µm) in accordance with the assumption that the cloud
top is perfectly emissive. The simple cloud scheme is never applied for MW sensors.
8.4. Definition of surface emissivity

Section 7.5 describes how to use the internal emissivity models provided by RTTOV and how to provide your own
input emissivities to RTTOV (including emissivities from the atlases). This section provides more details about the
various emissivity models.
MW emissivity models
RTTOV incorporates the FASTEM emissivity model for MW surface emissivities and the TESSEM2 model for sea
surface emissivities. There are 6 versions of the FASTEM sea surface emissivity model. FASTEM-6 is the generally
recommended option for most sensors, but TESSEM2 is recommended for Metop-SG ICI (and in general for channels
above 200GHz). The FASTEM options are summarised in Table 19.
The version of FASTEM to use (1 to 6) is set in the opts%rt_mw%fastem_version variable. TESSEM2 is selected by
setting this option to zero. Any value outside the range 0-6 will result in RTTOV returning an error.
The input profile variables used by FASTEM and TESSEM2 are given in section 7.5.
For FASTEM it is also important to define the polarisation status of each channel. This is specified in the instrument
coefficient file as a series of numbers in the FASTEM section and the polarisation IDs are defined in Table 20.
Different polarisations are defined for cross-track scanners, conical scanners and polarimetric instruments. In general
63
Version : 1.3
Date : 05/03/2019
you shouldn’t need to worry about this unless you specifically want to change the polarisation of a particular microwave
channel in which case you should edit the coefficient file to modify the relevant polarisation ID(s) or you can modify
the values directly in coefs%coef%fastem_polar(:).
FASTEM also provides emissivities for land and sea-ice surface types. This calculation uses the parameters defined in
profiles(:)%skin%fastem(1:5). Values for these parameters for different surface types are given in Table 21.
FASTEM-1 Fast emissivity model fitted to geometric optics model (English and Hewison, 1998).
FASTEM-2 As FASTEM-1 but with reflectivity dependence on atmospheric transmittance (Deblonde, 2000).
FASTEM-3 As FASTEM-2, but with azimuthal dependence (Liu and Weng, 2003).
FASTEM-4 Fast emissivity model as previous versions, but fitted to a new two-scale scattering model (Liu et al,
2011).
FASTEM-5 As FASTEM-4, but with a different foam coverage parametrisation and some fitting issues introduced
with FASTEM-4 now corrected (Bormann et al, 2012).
FASTEM-6 As FASTEM-5, but with an improved azimuthal dependence (Kazumori and English 2014 and RTTOV
(default) v12 Science and Validation Report).
NB No azimuthal dependence is simulated for the 3rd and 4th Stokes channels on Windsat.
Table 19. Summary of FASTEM versions.
FASTEM Pol_ID
Definition Examples of applicable sensors
in coef file
0 Average of vertical and horizontal polarisation ie 0.5(H+V) SSMIS
1 Nominal vertical at nadir rotating with view angle QV AMSU-A/B, MSU, MHS
2 Nominal horizontal at nadir rotating with view angle QH AMSU-A, MSU, MHS
3 Vertical V SSM/I, SSMIS, AMSR, Windsat
4 Horizontal H SSM/I, SSMIS, AMSR, Windsat
5 + 45 minus -45 (3rd stokes vector) S3 Windsat
6 Left circular - right circular (4th stokes vector) S4 Windsat
Table 20. Definition of polarisation status for FASTEM-2-6.
Surface type FASTEM parameters 1:5

Typical RTTOV default for land 3. 0, 5.0, 15.0, 0.1, 0.3
Summer land surface
Forest 1.7, 1.0, 163.0, 0.0, 0.5
Open grass 2.2, 1.3, 138.0, 0.0, 0.42
Bare soil 2.3, 1.9, 21.8, 0.0, 0.5
Winter surface type
Forest and snow 2.9, 3.4, 27.0, 0.0, 0.0
Deep dry snow 3.0, 24.0, 60.0, 0.1, 0.15
Frozen soil 117.8, 2.0, 0.19, 0.2 ,0.35
Sea ice
Grease ice 23.7, 7.7, 17.3, 0.0, 0.15
Baltic nilas 1.6, 3.3, 2.2, 0.0, 0.0
New ice (no snow) 2.9, 3.4, 27.0, 0.0, 0.0
New ice (snow) 2.2, 3.7, 122.0, 0.0, 0.15
Brash ice 3.0, 5.5, 183.0, 0.0, 0.0
Compact pack ice 2.0, 1700000.0, 49000000.0, 0.0, 0.0
Fast ice 1.5, 77.8, 703.0, 0.1, 0.35
Lake ice + snow 1.8, 67.1, 534.0, 0.1, 0.15
Multi-year ice 1.5, 85000.0, 4700000.0, 0.0, 0.0
Table 21. Values for FASTEM parameters for various land and sea-ice surface types.
Taken from English and Hewison (1998).
64
Version : 1.3
Date : 05/03/2019
IR emissivity models
The IR sea emissivity model is chosen using opts%rt_ir%ir_sea_emis_model. A value of 1 selects the ISEM model
(Sherlock, 1999) which was included in earlier RTTOV versions and a value of 2 (the default) selects the new IREMIS
model (RTTOV v12 Science and Validation Report). Section 7.5 lists the input profile variables used by each model. As
noted in section 7.5 PC-RTTOV uses its own sea surface emissivity model (see also section 8.8).
For land and sea-ice surface types RTTOV uses fixed emissivity values of 0.98 (land) and 0.99 (sea-ice) where
calcemis(:) is true.
Emissivity atlases
The IR and MW emissivity atlases are described in section 7.5. The atlases are accessed externally to RTTOV and the
emissivity values are passed in emissivity(:)%emis_in with the corresponding elements of calcemis(:) set to false. The
interfaces to the emissivity atlases are described in Annex F.
The CAMEL 2007 IR emissivity atlas (previously referred to as the CAMEL atlas) was new in RTTOV v12.1. This is
similar in many ways to the UWIRemis atlas. One practical difference is the atlas resolution: the UWIRemis atlas data
are at 0.1° resolution while the CAMEL atlas data are at 0.05 ° resolution (which means the CAMEL atlas requires
more memory). RTTOV v12.3 introduces a third option, the CAMEL climatology atlas. As described in Borbas and
Feltz (2019), this is derived from a multi-year climatology (2000-2016), provides emissivity standard deviations
computed from the climatology, and includes a more sophisticated treatment of snow. For the UWIRemis and CAMEL
2007 atlases, if the profile snow_fraction is greater than zero, the atlas emissivity value is linearly combined with a
fixed snow emissivity spectrum. In contrast, the CAMEL climatology atlas treats snow as follows:
 if the input snow_fraction = 1: if the atlas emissivity contains snow then the snow PC spectrum is used,
otherwise the fixed snow emissivity spectrum is used.
 if 0 < snow_fraction < 1: if the atlas emissivity contains snow then the atlas emissivity is returned without
modification, otherwise the atlas emissivity and fixed snow spectrum are linearly combined according to the
snow_fraction.
 if snow_fraction = 0: if the snow_correction flag is true (recommended, this is an argument to

rttov_get_emis) then if the atlas emissivity contains snow contamination the atlas will attempt to find a snow-
free PC spectrum to use. If it cannot find one (or if snow_corr is false) it simply returns the atlas emissivity.
The PC-RTTOV land/sea coefficients were trained using the UWIRemis atlas so it is recommended (though not strictly
mandatory) to use this atlas to provide land and sea-ice emissivities for PC-RTTOV.
For MW instruments, the TELSEM2 atlas and interpolator provides an updated version of the earlier TELSEM atlas
which includes climatological sea-ice and is valid for channels above 200GHz (i.e. for Metop-SG ICI). The CNRM
MW atlas has also been updated for RTTOV v12. This provides land surface emissivity values for AMSU-A, AMSU-B,
MHS, SSMI/S and ATMS for the years 2014 and 2015.
8.5. Simulation of visible and IR cloud-affected radiances

The simple cloud calculation (equation 7, section 8.3) can be used for single layer optically thick water clouds at mid-
infrared wavelengths but for more complex cloud types and/or multi-layer clouds RTTOV provides multiple scattering
calculations. RTTOV v12 provides independent options for treating the thermal emission and solar source terms in
visible and IR scattering simulations (note that a separate model is used for MW scattering – see section 8.7):
Discrete Ordinates Method (DOM) – thermal emission and solar source terms
The DOM solver is a pseudo-exact algorithm for solving the monochromatic radiative transfer equation. The
implementation of the DOM algorithm is described in the RTTOV v12 Science and Validation Report. The number of
DOM streams determines the accuracy of the simulation, but increasing the number of streams can dramatically
increase the run-time. The number of streams is set in opts%rt_ir%dom_nstreams: this should be an even integer
greater than 2 (odd values are rounded up). The default value is 8. This value should not exceed the number of Legendre
coefficients for the phase functions provided as input (see below). The cloud and aerosol coefficient files contain 128
coefficients for every phase function so this is the maximum number allowed when using these pre-defined particle
65
Version : 1.3
Date : 05/03/2019
types. If you supply your own optical properties (described below) there is no fixed upper limit. Experimentation
suggests that there is little benefit in using more than 16 DOM streams and that 8 or 12 should be fine for most
applications at least in the IR. You may wish to experiment to find the most suitable DOM parameter values for your
application.
(NB “DOM streams” are not to be confused with “cloud streams” discussed elsewhere which refer to cloud columns
defined by the cloud overlap scheme).
Important notes:
 By default RTTOV calculates the path of radiation through the atmosphere accounting for the curvature of the
Earth (and optionally also refraction). The DOM algorithm as implemented requires a strictly plane-parallel
atmospheric geometry. Therefore if the DOM solver is being used either for either thermal emission or
solar radiation calculations then the strict plane-parallel geometry is applied for all simulated radiances
(including clear-sky radiances). It is possible to turn on the strict plane-parallel geometry by setting
opts%rt_all%plane_parallel to true, but it is not necessary to do so when using DOM as RTTOV does this
automatically.
 When using DOM the surface is treated as a Lambertian reflector (this does not make use of the specularity
profile variable). When solving for thermal emission the albedo is calculated as 1-emissivity. When solving for
solar radiation the albedo is equal to the input or calculated (sunglint) BRDF multiplied by π: if this value
exceeds one the albedo is capped at one.
 The DOM algorithm is relatively slow and has relatively large memory requirements. The number of DOM
streams and the number of layers in the input profile have a direct impact on the speed. In addition, the more
layers containing scattering particles, the slower it runs. There are two parameters which can be used to
increase efficiency:
o opts%rt_ir%dom_accuracy: for solar radiation the azimuthal loop within the DOM algorithm will
exit early if the computed radiance increment is smaller than this fraction of the total radiance
(actually if this condition has been met twice). This is identical to the DISORT accuracy parameter.
The default is 0. (i.e. no truncation). This option should be used with some caution: it is recommended
that you carry out sensitivity tests.
o opts%rt_ir%dom_opdep_threshold: if this value is greater than zero, then the DOM solver
excludes any atmospheric layers below the level at which the absorption optical depth to space
exceeds this value. This typically only affects IR channels where there is significant atmospheric
absorption. A value of around 10. can decrease run-time noticeably with negligible impact on
radiances. The default is 0. (i.e. no layers are excluded).
In addition, use of opts%rt_ir%cldstr_threshold to exclude cloud streams (columns) with small weights can
have a significant impact on the run-time and memory requirements, subject to the caveats discussed below
regarding this option.
The optical properties of scattering particles required for this solver are:
 absorption and scattering coefficients
 at least opts%rt_ir%dom_nstreams coefficients of the Legendre decomposition of the phase function
 for solar radiation the full phase function is also required
Chou-scaling – thermal emission source term only
This is the fast multiple-scattering parameterisation for IR instruments available in previous versions of RTTOV and
described in the RTTOV v9 Science and Validation Report (available on the NWP SAF web site). The optical
properties of scattering particles required for this solver are:
 the “b” parameter (representing the fraction of back-scattered radiation) computed from the phase function
66
Version : 1.3
Date : 05/03/2019
Single-scattering – solar source term only
The solar single-scattering capability from previous versions of RTTOV for clouds and aerosols remains available as an
option, although it is not an accurate method especially in cases where scattering dominates over absorption. It is not
generally recommended. The optical properties of scattering particles required for this solver are:
 the full phase function
MFASIS – solar source term only
The MFASIS fast visible scattering parameterisation (Scheck et al, 2016) was introduced in RTTOV v12.2. MFASIS is
described in more detail in a separate section below.
Rayleigh scattering
As described in section 8.2 RTTOV includes a single-scattering approximation for atmospheric (molecular) Rayleigh
scattering. This is also applied in the case of cloud and aerosol scattering simulations. However when using the DOM
solver there is no interaction between the Rayleigh single-scattered radiation and the cloud or aerosol particles. This is
primarily done for reasons of efficiency: if Rayleigh scattering is included in the DOM multiple-scattering calculations
then every layer contains scattering particles which would result in extremely slow calculations (as noted above). In the
case of thick cloud, for example, this can result in an under-estimation of the TOA radiance in visible channels,
especially for larger satellite or solar zenith angles.
The scattering models for thermal emission and solar radiation are selected independently using the following options:
1=>DOM
Thermal emission opts%rt_ir%ir_scatt_model
2=>Chou-scaling (default)
1=>DOM (default)
Solar radiation opts%rt_ir%vis_scatt_model 2=>single-scattering
3=>MFASIS (cloud only)
RTTOV provides two methods of specifying the optical properties of the scattering particles.
Method 1 – use pre-defined optical properties: specify abundance profiles for the pre-defined particle types listed in
Table 24. For ice clouds there is a choice regarding the parameterisation of the optical properties of the ice particles.
Method 2 – provide optical properties explicitly: supply profiles of the scattering optical properties for each instrument
channel directly. This provides greater flexibility as you are not limited to the pre-defined particle types, but it is a
slightly more complicated way of calling RTTOV.
Note that all visible/IR scattering-related inputs are provided on layers rather than levels where layer i is bounded by
input levels i and i+1.
For both methods 1 and 2
Cloud scattering simulations are enabled by setting opts%rt_ir%addclouds to true. The choice between methods 1
and 2 is made by setting opts%rt_ir%user_cld_opt_param to false for method 1 (the default) or true for method 2.
For cloud simulations a cloud fraction profile must be specified in profiles(j)%cfrac(:) for each profile j. This specifies
the total fractional coverage of all cloud in each layer with 0 is no cloud and 1 is overcast: for method 1 the cloudy
fraction is assumed to contain a mixture of cloud types, each with their specified concentration (see below). The
remaining fraction of the layer is assumed to be clear.
When running the tangent linear (TL), adjoint (AD) or K models, you are advised to avoid specifying layers with a
cfrac equal to 1.0. Instead a value very close to 1.0 should be used (e.g. 0.999999). In addition you are advised not to
specify identical values of cfrac on adjacent layers. The reason for this is that the cfrac Jacobians are very sensitive to
perturbations in these cases (the direct model is not differentiable for fully overcast layers or where identical values of
cfrac are in adjacent layers). If this advice is not followed, RTTOV will make very small adjustments to the input cfrac
67
Version : 1.3
Date : 05/03/2019
profile in accordance with the above advice to ensure consistency between the direct, TL, AD and K models. These
adjustments are sufficiently small to not change the direct model radiances. These restrictions on the values in cfrac do
not apply when running the direct model alone.
As detailed in the RTTOV v9 Science and Validation Report, the computation of cloud affected radiances is performed
by dividing the computed atmospheric path into a number of independent cloud streams (or columns). The number of
cloud streams used for the scattering calculation is computed internally by RTTOV. It is possible to reduce the number
of cloud streams and save time and memory by considering only those cloud streams whose weight is larger than the
variable opts%rt_ir%cldstr_threshold. By setting cldstr_threshold to a negative number (or zero), all the cloud
streams will be processed (the default). This feature should be used with caution. Since the sum of the weights of all
cloud streams (including the clear one) must be equal to 1, if some cloud streams are excluded, the weight of the clear
stream has to be adjusted to a greater value. Consequently, if the value used for cldstr_threshold is too large, this can
result in a disproportionate weight of the clear stream with negative implications for the accuracy of the results. You
should select the value of cldstr_threshold in order to remove only the cloud streams with a very small weight. These
cloud streams have little impact on the total radiance and their exclusion can result in a sensible reduction of the
computational time required for cloud affected computations. Note however that by default cldstr_threshold is set to a
negative number and this is the recommended setting if running the TL, AD or K models to ensure the sensitivity to
cfrac is correctly computed in the TL, AD or Jacobian output.
Method 1 details – using pre-defined optical properties
As of RTTOV v12.3 there are two options for cloud liquid water optical properties and three options for ice cloud
optical properties. These are specified in profiles(:)%clw_scheme and profiles(:)%ice_scheme respectively.
Cloud liquid water – clw_scheme = 1 (OPAC)

These are the optical properties for 5 OPAC water cloud types available in previous versions of RTTOV (see Table 24).
These represent 5 different size distributions. The layer cloud liquid concentrations for one or more of the 5 types are
specified in profiles(:)%cloud(1:5,lay) for layer lay.
Cloud liquid water – clw_scheme = 2 (“Deff”)

These properties were new in v12.2 and are consistent with Mie cloud liquid water optical properties available with the
libRadtran package (Mayer and Kylling, 2005; http://www.libradtran.org). This is a single particle type with properties
parameterised in terms of particle effective diameter. Cloud liquid water concentration for layer lay can be specified in
any of profiles(:)%cloud(1:5,lay). Note that in this case all 5 columns are treated as one single particle type: RTTOV
will sum up the cloud concentrations in elements 1:5. In addition you must specify the cloud liquid water effective
diameter in microns in the profiles(:)%clwde(:) array. This value must be non-zero for all layers containing non-zero
CLW concentration, otherwise the layer will be treated as clear.
Cloud ice water – ice_scheme = 1 (SSEC/Baum)

These properties (Baum et al 2011) were introduced in RTTOV v12.1 and are parameterised in terms of ice effective
diameter. Layer ice cloud concentrations are input in profiles(:)%cloud(6,:). For ice particles RTTOV provides four
parameterisations of diameter in terms of temperature and ice water content (see Table 23). This is specified in
profiles(:)idg. Alternatively you can provide explicit values for the diameter in profiles(:)%icede(:) (in microns). For
layers containing cloud ice water the supplied icede(:) value will be used if non-zero, otherwise the selected
parameterisation will be used.
Cloud ice water – ice_scheme = 2 or 3 (Baran 2014, Baran 2018)

The Baran ice optical property parameterisation is described in Vidot et al (2015). The new 2018 parameterisation is
more spectrally consistent and as such is recommended over the 2014 parameterisation. As for theSSEC/Baum scheme
the ice cloud concentrations are input in profiles(:)%cloud(6,:). The Baran database has no explicit dependence on
particle size so the idg and icede(:) profile variables are ignored. Note that when using the DOM solver the Legendre
coefficients for the phase functions are calculated at run-time and as such the Baran scheme is slower than using the
SSEC ice properties with the DOM solver.
All of these properties are summarised in Tables 22 and 23. These tables also give valid ranges of profile variables
associated with the various options. If the input parameters exceed the given limits, RTTOV uses the limit value(s)
instead in the optical property calculations. The exception is IWC for the SSEC ice database: a warning is given by
rttov_user_profile_checkinput if the values are out of bounds, but the hard limits are not imposed by the code.
68
Version : 1.3
Date : 05/03/2019
Both sets of cloud liquid water optical properties and the SSEC/Baum optical properties are stored in coefficient files
whose names begin “sccldcoef_”. The Baran ice scheme optical properties are parameterised within the code. When
reading the RTTOV coefficient files (using the rttov_read_coefs subroutine – see section 7.2) the
opts%rt_ir%addclouds option must be true and the cloud coefficient file is read in at the same time as the optical
depth coefficient file. The cloud coefficients must be defined for the same set of channels as the optical depth
coefficients: there are separate cloud scattering coefficient files available (by default) in rtcoef_rttov12/cldaer_ir/ and
rtcoef_rttov12/cldaer_visir/ which should be used with the IR-only (v7 and v8 predictor optical depth coefficient files
and hyper-spectral sounder coefficient files) and with the visible+IR coefficient files respectively.
When populating the RTTOV profiles structure, the cloud fraction profile is specified in profiles(:)%cfrac(:) as
described above. The cloud liquid and ice water densities are specified in profiles(:)%cloud(1:6,lay) for layer lay as
described above. The units for liquid and ice water content are determined by the logical flag
profiles(:)%mmr_cldaer: if this is true (the default) the units are kg/kg, if the flag is false the units are g/m3 as in
previous RTTOV versions. Note that this setting affects units of both clouds and aerosols. All profiles passed into
RTTOV in a single call must use the same units.
As of RTTOV v12.3 the cloud concentration supplied to RTTOV may be a grid box average value which is often the
quantity provided by GCMs. In this case you must set the opts%rt_ir%grid_box_avg_cloud option to true. In
previous versions of RTTOV it was necessary to supply the concentration of cloud in the cloudy fraction of the layer:
this is equal to the grid box average concentration divided by the cloud fraction. In this case the
opts%rt_ir%grid_box_avg_cloud option must be set to false (the default).
The cloud concentration can be non-zero for multiple cloud types in any layer: all cloud types present are assumed to be
mixed together within the fraction of the layer specified by cfrac(:), with the remaining fraction of the layer assumed to
be clear. If either cfrac(j) or all cloud(1:6,j) are zero for layer j then there is no contribution to the simulated radiance
from cloud in this layer. An example input cloud and cloud fraction profile is given in file:
rttov_test/test_example.1/cld_prof.dat.
To compute cloudy radiances via method 1:

 Set opts%rt_ir%addclouds to true
 Set opts%rt_ir%user_cld_opt_param to false (this is the case by default)
 Ensure the cloud scattering coefficient file is read in the call to rttov_read_coefs (see Annex C). The naming
convention for these files is sccldcoef_msg_3_seviri.dat where SEVIRI on MSG-3 is the sensor in this case.
Remember that the cloud coefficient file must match the optical depth coefficient file (IR-only or visible+IR).
 Specify the units of cloud concentration in profiles(:)%mmr_cldaer (true by default for units of kg/kg).
 Populate the input profiles(:)%cloud(:,:) array with liquid or ice water/ice concentrations (grid box average
concentration if grid_box_avg_cloud is true, or concentraton of cloudy fraction of layer otherwise) in appropriate
units for each cloud type (the first index is the cloud type, the second is layer number).
 Populate the input profiles(:)%cfrac(:) array with the layer cloud fractions from 0-1.
 Specify the treatment of cloud liquid and ice particles according to the options in Tables 22 and 23.
An example program src/test/example_cld_file_fwd.F90 has been created which demonstrates these steps for cloud
scattering and can be used as a template for your own program.
OPAC CLW properties “Deff” CLW properties

Set profiles(:)%clw_scheme = 1 Set profiles(:)%clw_scheme = 2
Specify vertical profiles of Properties are available for one CLW type. Specify the vertical profile of
concentrations for one or more cloud concentration in, say, profiles(:)%cloud(1,:)
liquid water types (see Table 24 NB RTTOV will sum CLW concentrations in profiles(:)%cloud(1:5,lay) for
below) in each layer lay.
profiles(:)%cloud(1:5,:) Specify a vertical profile of CLW effective diameter in profiles(:)%clwde(:)
Min eff. diameter: 2µm
Max eff. diameter: 52µm
Table 22. Summary of cloud liquid water particle parameterisation options and associated parameter limits.
69
Version : 1.3
Date : 05/03/2019
SSEC ice properties Baran 2014 or 2018 ice schemes

Set profiles(:)%ice_scheme = 1 Set profiles(:)%ice_scheme = 2 or 3
Specify ice effective diameter in Specify parameterisation for effective No explicit dependence on ice
microns for each layer in diameter in effective diameter: icede(:) and idg
profiles(:)%icede(:) Set profiles(:)%idg are ignored.
1 = Ou and Liou (1995)
2 = Wyser (1998) (recommended)
3 = Boudala et al (2002)
4 = McFarquar et al (2003)
and ensure profiles(:)%icede(:) is zero.
Min eff. diameter = 10 µm Min T = 193.157 K
Max eff. diameter = 120 µm Max T = 273.127 K
Min IWC = 4.984E-5 g m-3 Min IWC = 6.0E-06 g m-3
-3
Max IWC = 0.1831 g m Max IWC = 1.969466 g m-3
Only effective diameter limits are imposed in the code.
Table 23. Summary of ice particle parameterisation options and associated parameter limits.
Column 1: Stratus Continental STCO

Column 2: Stratus Maritime STMA
Column 3: Cumulus Continental Clean CUCC
Column 4: Cumulus Continental Polluted CUCP
Column 5: Cumulus Maritime CUMA
Column 6: Ice cloud (all types despite the name “CIRR”) CIRR
Table 24. Cloud types available in RTTOV v12.
Method 2 details – providing optical properties explicitly
Instead of using the pre-defined optical properties you can input your own. In this case a vertical profile is provided for
each channel containing the optical property data for each layer. No cloud coefficient file us required and the
profiles(:)%cloud(:,:) array is not used, but profiles(:)%cfrac(:) must still be provided. The cloud optical property
profiles are input to RTTOV using the cld_opt_param argument (of derived type rttov_opt_param). This comprises
arrays for each scattering parameter required by RTTOV for each layer and channel. The “nchanprof” dimension
corresponds to the channels in the chanprof(:) array. Note that the properties specified apply specifically to the cloudy
fraction in each layer: the opts%rt_ir%grid_box_avg_cloud flag is ignored for method 2. The members of the
rttov_opt_param structure are:
 abs(:,:) : absorption coefficient (units: km-1), dimensions (nlayers, nchanprof). Always required.
 sca(:,:) : scattering coefficient (units: km-1), dimensions (nlayers, nchanprof) . Always required.
 bpr(:,:) : “b parameter” (no units), dimensions (nlayers, nchanprof). Only required by the Chou-scaling
parameterisation for channels with thermal emission (wavelength > 3µm). This is the fraction of back-scattered
radiation from each layer and is calculated from the phase function using a supplied subroutine (see below).
 nmom : number of Legendre coefficients specified for all phase functions. Only required by the DOM solver.
 legcoef(:,:,:) : coefficients of Legendre decomposition of phase function (no units), dimensions (1:nmom+1,
nlayers, nchanprof). Only required by the DOM solver (for thermal emission or solar radiation). These can be
computed using a supplied subroutine (see below). Note that the “zeroth” moment in legcoef(1,:,:) must always be
one and nmom coefficients are supplied in addition to this.
 phangle(:) : the angles over which the phase functions are defined (units: degrees), dimension (nphangle). This
should cover the full range of scattering angles monotonically from 0° to 180° inclusive. The angle grid does not
have to be evenly-spaced.
 pha(:,:,:) : azimuthally-averaged phase function, dimensions (nphangle, nlayers, nchanprof). Required by
scattering simulations for solar radiation and used for the calculation of the b parameter and Legendre coefficients.
Phase functions should be normalised such that the integral over all scattering angles is 4π. Given phase function P
defined over scattering angles  from 0 to  , and zenith angle  and azimuthal angle  we require
70
Version : 1.3
Date : 05/03/2019
  2
4   P()d   P( ) sin( )dd
0 0 0
so that:

1  12  P( )sin( )d
0
A variable cld_opt_param of type rttov_opt_param should be declared and then allocated by calling
rttov_alloc_opt_param (see Annex D). (This subroutine should also be called to deallocate the structure at the end of
your program). The various parameters required by the scattering model(s) being used, as described above, should be
specified. The phase functions for pre-defined particle types are defined over one of two angular arrays both of which
can be found in the rttov_const module as phangle_lores(1:208) and phangle_hires(1:498). Alternatively you can
define your own angle grid.
A subroutine is provided to calculate the b parameters from the phase functions. This is relatively slow and so is not
performed internally within RTTOV. You may find it beneficial to calculate the b parameters off-line and store them for
future use if this is practical for your application. To generate the values:
 call rttov_bpr_init to initialise some tables to speed up the calculation
 call rttov_bpr_calc to calculate the b parameters from the phase functions: this is called once for every phase
function (i.e. for each layer containing scattering particles and for each channel)
 call rttov_bpr_dealloc to release allocated memory
In addition the rttov_legcoef_calc subroutine is provided to calculate the Legendre coefficients for a phase function.
This is quite fast and you only need to calculate as many coefficients as DOM streams you will specify for your
simulations.
The interfaces for all these subroutines are described in Annex E.
Note that if you already have values for bpr and/or the Legendre coefficients and you are not running solar simulations
(i.e. opts%rt_ir%addsolar is false) then the phangle and pha arrays (i.e. phase angles and phase functions) do not
need to be populated for the call to RTTOV.
Finally, if solar scattering calculations are being performed, you must call the rttov_init_opt_param subroutine (see
Annex D) to precalculate some values related to the phase angles. This is not required if opts%rt_ir%addsolar is
false.
To compute cloudy radiances via method 2:

 Set opts%rt_ir%addclouds to true
 Set opts%rt_ir%user_cld_opt_param to true
 Declare a variable of type rttov_opt_param, for example cld_opt_param, and allocate its member arrays by
calling rttov_alloc_opt_param
 Populate cld_opt_param with absorption and scattering coefficients and, for solar radiation, the phase function and
phase angles
 Populate cld_opt_param with the b parameters (if using the Chou-scaling parameterisation) either from pre-
calculated data or by calling rttov_bpr_calc
 Populate cld_opt_param with the Legendre coefficients (if using the DOM solver) either from pre-calculated data
or by calling rttov_legcoef_calc
 If performing solar calculations call rttov_init_opt_param
 Populate the input profiles(:)%cfrac(:) array with cloud fraction from 0-1
 Pass the cld_opt_param argument into rttov_direct (or _tl/ad/k)
 Optionally pass cld_opt_param_tl/ad/k into rttov_tl/ad/k respectively (see section 7.9).
 When finished with RTTOV call rttov_alloc_opt_param again to deallocate cld_opt_param.
An example program src/test/example_cld_param_fwd.F90 has been created which demonstrates these steps for
cloud scattering and can be used as a template for your own programs.
71
Version : 1.3
Date : 05/03/2019
MFASIS fast visible scattering model
The MFASIS fast cloud scattering parameterisation (Scheck et al, 2016) has been incorporated into RTTOV. This is a
look-up table (LUT) based model which currently enables cloud simulations for visible channels (below ~1µm).
An example program src/test/example_cld_mfasis_fwd.F90 has been created which demonstrates a call to the
MFASIS direct model and can be used as a template for your own programs.
In RTTOV, MFASIS simulations are performed in the same way as other cloud simulations through calls to the
RTTOV direct/TL/AD/K models, but there are a few key differences. The MFASIS model does not use the RTTOV
optical depth parameterisation and therefore does not use the RTTOV interpolator. Optional trace gases are not enabled
in the MFASIS simulations. MFASIS simulations ignore the setting of opts%config%do_checkinput (see section
7.3): the input profile is not checked for invalid data. You can use the rttov_user_profile_checkinput subroutine, but
note that this tests all profile variables, including those which are not used by MFASIS (see below). MFASIS
simulations enforce the strict plane-parallel geometry (as DOM does). MFASIS simulations cannot be run with aerosols
and they are not compatible with the explicit optical property inputs. When running MFASIS simulations, radiances and
reflectances are only returned for channels supported in the MFASIS LUT file. Therefore you cannot run simulations
for IR and visible channels with MFASIS in the same call to RTTOV.
The input profile variables used by MFASIS are given in Table 25. All other profile inputs are ignored. MFASIS
assumes a Lambertian surface with albedo given by pi times the surface BRDF and this value is capped at one, as for
DOM. The surface BRDF options are the same as for other RTTOV calls.
MFASIS is selected by setting both opts%rt_ir%addclouds and opts%rt_ir%addsolar to true and setting
opts%rt_ir%vis_scatt_model = 3. You must set these options before the call to rttov_read_coefs to read in the
various coefficient files. You can change, for example, the selected solar scattering model later in your code if desired.
MFASIS requires a standard RTTOV rtcoef file, an sccldcoef cloud optical properties file, and an MFASIS cloud LUT
file for the sensor you wish to simulate. All three files are read in the same call to rttov_read_coefs (see Annex C). Due
to their size the MFASIS LUTs are provided in HDF5 format. There is no problem reading ASCII rtcoef and sccldcoef
files at the same time as an HDF5 MFASIS LUT. You must use the v9 predictor rtcoef files and the associated
sccldcoef cloud property files.
Profile structure variables Notes

gas_units Mandatory
p(:), t(:), q(:) Mandatory
s2m%p Mandatory
zenangle, azangle, sunzenangle, sunazangle Mandatory
skin%surftype Mandatory
elevation Mandatory
latitude Mandatory
s2m%u, s2m%v, s2m%wfetc, skin%watertype Used by solar sea BRDF model
mmr_cldaer Mandatory
clw_scheme Mandatory
clwde(:) Required if clw_scheme = 2
ice_scheme Mandatory
idg, icede(:) Required if ice_scheme = 1 (see Table 23)
cfrac(:) Mandatory
cloud(:,:) Mandatory
Table 25: profile variables used by MFASIS.
MFASIS LUTs are trained using specific cloud liquid and ice water optical properties, as determined by the
clw_scheme and ice_scheme profile variables (as described above). When running MFASIS simulations the
clw_scheme and ice_scheme variables must match those in the LUT. All MFASIS LUTs are currently trained using
ice_scheme = 1 (the SSEC/Baum properties). LUTs are available based on both the OPAC and new Deff CLW optical
properties, as indicated by the LUT filenames. After reading a LUT you can determine the properties used to train it
from the coefs%coef_mfasis_cld%clw_scheme and coefs%coef_mfasis_cld%ice_scheme variables.
72
Version : 1.3
Date : 05/03/2019
When running MFASIS you can specify any channel selection valid for the rtcoef file, but only channels supported by
the MFASIS LUT file will be simulated. Radiances for other channels will be zero on exit.
There are various conditions which can cause possible issues with MFASIS radiances. The most common ones related
to certain combinations of solar and viewing geometry which can result in questionable accuracy, or parameters which
lie beyond the range of values used for training the LUTs. If relevant, warning flags are set in the radiance%quality(:)
output array described in section 7.8. You should check whether or not the quality(:) array contains non-zero values
(indicating potential issues) after calling RTTOV. The following warning flags may be set by MFASIS simulations (the
constants referred to are defined in the rttov_const module):
 qflag_mfasis_zenangle: the sallite and solar zenith angles should each be less than mfasis_maxzenangle=85°.
 qflag_mfasis_sumzenangle: the sum of the satellite and solar zenith angles should be less than
mfasis_maxsumzenangle=150°.
 qflag_mfasis_loscaangle/qflag_mfasis_hiscaangle: the scattering angle (see below) must lie in the range
[mfasis_minscaangle, mfasis_maxscaangle] = [9°, 140°].
 qflag_mfasis_geometry_bounds: the scattering angle should lie within the range of values used for training
the LUT (see below).
 qflag_mfasis_opdpedia_bounds: this flag indicates that a total optical depth or effective diameter value (for
liquid or ice cloud) has exceeded the LUT bounds (see below). For the LUTs available on the RTTOV website
this flag does not necessarily indicate a problem if the optical depths exceed the maximum values used in
training as the liquid and ice water cloud optical depth maxima are large.
The scattering angle is computed as follows, with scat_angle=0° occurring when zen_sat and zen_sun are equal and
rel_azi=0° (rel_azi is the difference between the solar and satellite azimuth angles):
cos(scat_angle) = cos(zen_sat) * cos(zen_sun) + sin(zen_sat) * sin(zen_sun) * cos(rel_azi)
Table 26 lists the minimum and maximum values used in training the MFASIS LUTs available on the RTTOV website
at the time of release of v12.3. These ranges are given in the README_LUT section of the MFASIS LUT files. You
can print out information about an MFASIS LUT using the rttov_mfasis_lut_info.exe executable (see Annex A)
including the training bounds of each dimension. If a variable exceeds the bounds used in training MFASIS, the closest
value is used from the LUT: there is no extrapolation beyond the bounds of the training data. Note that as the surface
albedo is capped at one within RTTOV, this can never fall outside the bounds.
MFASIS variable Minimum Maximum

Scattering angle 9° 140°
Cloud liquid water total optical depth 0 1000
Cloud ice water total optical depth 0 300
Cloud liquid water total Deff 5µm 50µm
(Deff CLW optical properties only)*
Cloud ice water total Deff 10µm 120µm
Surface albedo 0 1
Table 26. Ranges of values used in training MFASIS LUTs available on the RTTOV website.
*For LUTs trained using the OPAC optical properties, the Deff value is fixed for each of the 5 OPAC particle types.
The LUT data are computed for these 5 specific Deff values so the LUT is never interpolated in this dimension and can
never give out-of-bounds errors.
8.6. Simulation of visible and IR aerosol-affected radiances

The aerosol-affected simulations are very similar to the cloud-affected simulations so you should read section 8.5 as
most of the information is relevant. The same scattering models are available for aerosols and clouds (except MFASIS)
and, as for clouds, there are two methods for running aerosol simulations.
For both methods 1 and 2
Aerosol scattering simulations are enabled by setting opts%rt_ir%addaerosl to true. The choice between methods 1
and 2 is made by setting opts%rt_ir%user_aer_opt_param to false for method 1 (the default) or true for method 2.
73
Version : 1.3
Date : 05/03/2019
Due to the way in which the aerosol-affected radiances are calculated, the “clear-sky”outputs in the rttov_radiance
structure (e.g. radiance%clear, radiance%bt_clear, radiance%refl_clear) include the effects of aerosol when
aerosols are included.
Method 1 – using pre-defined optical properties
RTTOV v12 now provides two sets of files containing aerosol optical properties. The pre-existing OPAC files define
thirteen types of aerosol particle. New files have been generated containing optical properties for nine species from the
CAMS aerosol model. These optical properties are contained in coefficient files whose names begin “scaercoef_”. The
new CAMS files contain “_cams” in the filenames. When reading the RTTOV coefficient files (using the
rttov_read_coefs subroutine – see section 7.2) the opts%rt_ir%addaerosl option must be true and the aerosol
coefficient file is read in at the same time as the optical depth coefficient file. The aerosol coefficients must be defined
for the same set of channels as the optical depth coefficients: there are separate aerosol scattering coefficient files
available (by default) in rtcoef_rttov12/cldaer_ir/ and rtcoef_rttov12/cldaer_visir/ which should be used with the IR-
only (v7 and v8 predictor optical depth coefficient files and hyper-spectral sounder coefficient files) and with the
visible+IR coefficient files respectively.
When populating the RTTOV profiles structure, the aerosol concentrations are specified in profiles(:)%aerosols(i,j)
where index j is the layer number and index i runs from 1-13 (OPAC) or 1-9 (CAMS) and represents each of the aerosol
particle types listed in Table 27. The units are determined by the logical flag profiles(:)%mmr_cldaer: if this is true
(the default) the units are kg/kg, if the flag is false the units are number density in cm -3 as in previous RTTOV versions.
Note that this setting affects units of both clouds and aerosols. All profiles passed into RTTOV in a single call must use
the same units. For most aerosol types (both OPAC and CAMS) the input is dry aerosol mass ratio. The only exception
are the CAMS sea salt types for which the input is mass ratio at 80% relative humidity. This is consistent with outputs
from the CAMS model.
The CAMS species are described in Bozzo et al (2017). For the OPAC aerosol files, particle types 1-11 are described in
detail in Matricardi (2005). The new volcanic ash particle type (number 12) uses a log-normal size distribution function
calculated for radii between 0.005 and 20µm with parameters derived from aircraft measurements of the 2010 Icelandic
eruption (Johnson et al, 2012). The refractive indices are from Pollack et al (1973). The optical parameters for the new
Asian dust particle type (number 13) are calculated using a linear combination of size distributions for the MINM,
MIAM and MICM aerosol particles for radii between 0.01 and 60µm: the weights were obtained by fitting to a particle
size distribution derived from sky radiometer measurements made at Dunhuang, China (Han et al, 2012). The refractive
indices are from Volz (1972, 1973).
RTTOV v12.3 provides a new tool rttov_make_scaercoef.exe described in Annex N which allows you to generate
custom aerosol optical property files for use with RTTOV. These can contain an arbitrary number of aerosol species: in
this case when the profiles structure is allocated, the second dimension of the profiles(:)%aerosols(:,:) array is the
number of aerosol species in your scaercoef file and you specify the species concentrations in the order in which they
are defined in the file.
You can define any mix of the various aerosol components/species defined in the aerosol property file. An example
input aerosol profile using the OPAC types is given in file: rttov_test/test_example.1/aer_prof.dat. RTTOV also
provides a subroutine to calculate various climatological aerosol profiles with pre-defined mixtures of OPAC
components (see the RTTOV v9 Science and Validation Report). The binary create_aer_clim_prof.exe (source:
src/other/create_aer_clim_prof.F90) (see Annex N) can be used to generate climatological aerosol profiles off-line.
The list of climatological compositions output by this program is also shown in Annex N. The file
data/prof_aerosl_cl.dat contains climatological aerosol profiles generated using the standard RTTOV 101 pressure
levels with T and q profiles taken from the file data/prof.dat for a latitude of zero, surface elevation of zero, surface
level 101 and scale factor 1.0. You may wish to re-run the executable to generate your own sets of aerosol profiles for
different input parameters, profiles or numbers of levels. Alternatively, the example program
src/test/example_aer_file_fwd.F90 gives an example of calling the rttov_aer_clim_prof.F90 subroutine (see Annex
N) to obtain climatological profiles at run-time.
74
Version : 1.3
Date : 05/03/2019
To compute aerosol-affected radiances via method 1:

 Set opts%rt_ir%addaerosl to true
 Set opts%rt_ir%user_aer_opt_param to false (this is the case by default)
 Ensure the aerosol scattering coefficient file is read in the call to rttov_read_coefs (see Annex C). The naming
convention for these files is scaercoef_msg_3_seviri.dat where SEVIRI on MSG-3 is the sensor in this case.
Remember that the aerosol coefficient file must match the optical depth coefficient file (IR-only or visible+IR).
 Specify the units of aerosol concentration in profiles(:)%mmr_cldaer (true by default for units of kg/kg).
 Populate the input profiles(:)%aerosol(:,:) array with mean layer aerosol concentration in appropriate units for each
aerosol type.
An example program src/test/example_aer_file_fwd.F90 has been created which demonstrates these steps for aerosol
scattering and can be used as a template for your own programs.
Column OPAC aerosol components CAMS aerosol species

1 Insoluble INSO Black Carbon BCAR
2 Water soluble WASO Dust, bin 1, 0.03-0.55 micron, ref. index: Woodward 2001 DUS1
3 Soot SOOT Dust, bin 2, 0.55-0.90 micron, ref. index: Woodward 2001 DUS2
4 Sea salt (acc mode) SSAM Dust, bin 3, 0.90-20.0 micron, ref. index: Woodward 2001 DUS3
5 Sea salt (coa mode) SSCM Ammonium sulphate SULP
6 Mineral (nuc mode) MINM Sea salt, bin 1, 0.03-0.5 micron SSA1
7 Mineral (acc mode) MIAM Sea salt, bin 2, 0.5-5.0 micron SSA2
8 Mineral (coa mode) MICM Sea salt, bin 3, 5.0-20.0 micron SSA3
9 Mineral transported MITR Hydrophilic organic matter OMAT
10 Sulphated droplets SUSO
11 Volcanic ash VOLA
12 New volcanic ash VAPO
13 Asian dust ASDU
Table 27. OPAC aerosol components and CAMS aerosol species.
Method 2 – providing optical properties explicitly
This is very similar to Method 2 for cloud profiles except there is no cloud fraction. You should read the description of
Method 2 above for clouds: the same parameters are required for the various scattering models. The optical property
data are input to RTTOV via the aer_opt_param argument of derived type rttov_opt_param.
To compute aerosol-affected radiances via method 2:

 Set opts%rt_ir%addaerosl to true
 Set opts%rt_ir%user_aer_opt_param to true
 Declare a variable of type rttov_opt_param, for example aer_opt_param, and allocate its member arrays by
calling rttov_alloc_opt_param
 Populate aer_opt_param with absorption and scattering coefficients and, for solar radiation, the phase function and
phase angles
 Populate aer_opt_param with the b parameters (if using the Chou-scaling parameterisation) either from pre-
calculated data or by calling rttov_bpr_calc
 Populate aer_opt_param with the Legendre coefficients (if using the DOM solver) either from pre-calculated data
or by calling rttov_legcoef_calc
 If performing solar calculations call rttov_init_opt_param
 Pass the aer_opt_param argument into rttov_direct (or _tl/ad/k)
 Optionally pass aer_opt_param_tl/ad/k into rttov_tl/ad/k respectively (see section 7.9).
 When finished with RTTOV call rttov_alloc_opt_param again to deallocate aer_opt_param.
An example program src/test/example_aer_param_fwd.F90 has been created which demonstrates these steps for
aerosol scattering and can be used as a template for your own programs.
75
Version : 1.3
Date : 05/03/2019
8.7. Simulation of microwave radiances scattered by cloud and precipitation

A separate interface, known as RTTOV-SCATT, is provided for simulating microwave radiances affected by cloud and
precipitation. The scattering effects of hydrometeors at microwave frequencies are computed using the delta-Eddington
approximation. Note that the cloud simulations described in section 8.5 are completely different: they are currently only
applicable to visible and infrared radiances. RTTOV-SCATT is described by Bauer et al. (2006) and the cloud overlap
is described in Geer et al. (2009a). Further information can also be found in the RTTOV v9, v10 and v11 Science and
Validation Reports.
The RTTOV-SCATT code calls the core RTTOV for the clear air part but adds the scattering effects from water/ice in
the profile. RTTOV-SCATT uses a two-independent column approximation, summarised by:
LTotal
B  (1  C ) LClear
B  CLRainy
B (8)
Here, C is the effective cloud fraction in the vertical profile and L is radiance or brightness temperature (depending on
the setting of opts_scatt%lradiance: it is recommended to set this flag to true so that the RT calculations are done in
radiance rather than brightness temperature). The clear-air RTTOV is called from within RTTOV-SCATT and returns
the radiance or brightness temperature of the clear sky column, LClear, and the profile of clear sky transmittances.
RTTOV-SCATT then computes the cloudy or rainy radiance or brightness temperature, LRainy, using the clear sky
transmittances provided by the core RTTOV, and lookup tables for Mie scattering properties. Finally, equation 8 is used
to linearly combine the two independent columns, producing the total radiance or brightness temperature LTotal.
RTTOV-SCATT is called via the subroutine interface rttov_scatt, which is quite different from that for the core
RTTOV, i.e. rttov_direct. The input profiles are the same as for the clear-sky RTTOV (e.g. rttov_profile; Table 12
and section 7.3) but additional information is required, principally hydrometeor profiles, supplied in
rttov_profile_cloud and listed in Table 28. The use_totalice logical variable in rttov_profile_cloud should be set to
false for separate ice and snow and to true for total ice. These two options are mutually exclusive.
The vertical level-layer discretisation used by RTTOV-SCATT is different to that used by the core RTTOV routines.
Within the rttov_profile_cloud structure the constituent and hydrometeor amounts are given on 'full' pressure levels,
and they apply to a domain bounded by 'half' pressure levels. Conventionally, the bottom half level is the surface (2m)
pressure and the top half level is the top of the atmosphere. Full pressure levels are those supplied in rttov_profile (in
profiles(:)%p(:)), but the half level pressures need to be supplied in rttov_profile_cloud. Figure 5 shows the
arrangement of full and half levels. Note that in common with the treatment of cloud variables, for doing the radiative
transfer in the cloudy column, RTTOV-SCATT treats profile variables in the rttov_profile structure as also
representing full pressure level quantities (rather than layer quantities, as in the rest of RTTOV). Within the
rttov_profile_cloud structure is the option to specify the rain and solid precipitation (sp) in units of [kg/kg] (the
default) rather than as a flux [kg/(m2)/s]. The input cloud concentrations must be the layer grid-box-average
concentration (as opposed to the concentration within the cloudy fraction of each layer): the grid-box-average is
commonly the value provided by GCMs.
Profile variable Contents

nlevels number of atmospheric levels, which should match that supplied in the other input profiles
use_totalice logical flag to switch between using separate ice and snow, or total ice hydrometeor types.
mmr_snowrain logical flag to set units for snow and rain: if true units are kg/kg (the default), if false units
are (kg/(m2)/s)
cfrac Optional: if opts_scatt%lusercfrac=true., supply the effective cloud fraction, C, here.
This is normally calculated internally in RTTOV-SCATT
ph(:) nlevels + 1 of half-level pressures (hPa)
cc(:) nlevels of cloud cover (0-1)
clw(:) nlevels of cloud liquid water (kg/kg)
ciw(:) nlevels of cloud ice water (kg/kg)
totalice(:) nlevels of total ice (kg/kg)
rain(:) nlevels of rain flux (units depend on mmr_snowrain)
sp(:) nlevels of solid precipitation flux (units depend on mmr_snowrain)
Table 28. RTTOV-SCATT profile variables for rttov_profile_cloud
76
Version : 1.3
Date : 05/03/2019
An example of an RTTOV-SCATT forward model call is provided in src/test/example_rttovscatt_fwd.F90. This

reads atmospheric profiles from a data file and writes the simulated brightness temperatures to an output file (this
should be called using rttov_test/run_example_rttovscatt_fwd.sh, see section 5.3).
Another example of calling RTTOV-SCATT including a Jacobian calculation is provided in

src/mw_scatt/example_rttovscatt.F90. The test programs rttovscatt_test.F90 (top level, driven by
rttov_test/test_rttovscatt.sh) and rttovscatt_test_one.F90 both in src/mw_scatt/ contain further examples of how to
use the tangent-linear, adjoint and K functionality.
A limited number RTTOV options are available in RTTOV-SCATT via the rttov_options_scatt type described in
Annex O. This also contains the lusercfrac variable which was an argument of the rttov_scatt subroutine in RTTOV
v10. If opts_scatt%lusercfrac = .true., you can supply your own effective cloud fraction (see Geer et al., 2009b). By
default this flag is false and the effective cloud fraction is calculated internally in RTTOV-SCATT. The optional
diagnostic output array cfrac is available from the direct call, should you want to look at the value computed internally.
If the effective cloud fraction is smaller than a certain threshold, then the cloudy radiance is ignored and only the clear-
sky radiance is returned. Prior to RTTOV v12.3 this threshold was hard-coded at 0.05, but in v12.3 it can be specified in
opts_scatt%cc_threshold (the default remains 0.05).
Support for all-sky dynamic emissivity retrievals was added in v12.2, following the method of Baordo and Geer (2016).
An optional output structure of the direct call to RTTOV-SCATT, the rttov_scatt_emis_retrieval_type contains the
relevant parameters for the emissivity retrieval (see Annex O); as with other interface structures there are helper
routines to allocate and deallocate it. To calculate these additional outputs, opts_scatt%lradiance must be set to true.
The function rttov_scatt_emis_retrieval is then available to retrieve the surface emissivity (see Annex N).
Figure 5. RTTOV-SCATT full and half levels, showing half and full level pressure (Ph,Pf) and examples of the variables
specified on full levels (e.g. q, T, but also cloud and hydrometeors)
RTTOV-SCATT optical properties
RTTOV-SCATT relies on both clear-sky coefficient files (e.g. rtcoef_noaa_15_amsua.dat) and precomputed tables of
scattering parameters (e.g. mietable_noaa_amsua.dat). These files are referred to as “Mietables”, but they also now
include optical properties for non-spherical particles. Due to their large size the Mie files for microwave sensors are not
supplied in the package, but are provided on the RTTOV web site for download. The executable
rttov_ascii2bin_scattcoef.exe may be used to convert ASCII Mie tables to binary format for faster performance. This
is described in Annex A.
You may also create your own Mie coefficient files (or re-generate those supplied on the RTTOV website) using the
UNIX shell script src/mw_scatt_coef/mie_table_generation.ksh. This script may need editing, for example to point to
the location of your RTTOV binaries. The Mie table generation is based around an input file channels.dat, examples of
which may be found in the same directory. These define the parameters for the calculations and the instruments and
channels for which to generate coefficients. The existing channels.dat files provide useful templates. See the associated
readme.txt in the src/mw_scatt_coef/ directory for full details. Modifying the Mie tables may be useful in some cases
but is not generally recommended.
77
Version : 1.3
Date : 05/03/2019
8.8. Simulation of hyperspectral IR sounder radiances using PC-RTTOV

A principal component (PC) based version of RTTOV is available for the simulation of the full spectrum of
hyperspectral IR sounders as PCs (Matricardi, 2010). The PC-based model uses polychromatic RTTOV radiances to
predict the principal component scores using a linear regression scheme. To invoke the PC calculations in RTTOV v12
a logical flag opts%rt_ir%pc%addpc is set to true. An example of calling PC-RTTOV is given in
src/test/example_pc_fwd.F90: this can be used as a model for your own code.
Coefficients
The PC-RTTOV coefficient filenames begin “pccoef_” and can be downloaded from the RTTOV web site. These must
be used alongside the RTTOV optical depth (rtcoef) coefficient files with which they were trained. PC-RTTOV is
trained used v9 predictor 101 level coefficients. At the time of release the PC coefficients were trained with the same
optical depth coefficients used to train the RTTOV v11 PC coefficients. These optical depth coefficient files have
“pcrttov_compat” in the filename. If an incompatible optical depth coefficient file is used an error will result. The PC
coefficient file is read in at the same time as the optical depth coefficient file in the call to rttov_read_coefs.
The PC-compatible optical depth coefficient files allow for the variation of O3, CO2, N2O, CO and CH4. RTTOV v12.2
introduced new PC coefficient files which allow all of these gases to vary (optionally) in the same was as for standard
(non-PC) RTTOV. PC coefficients which support all of these trace gas species have “trace” in the filename. Older PC-
RTTOV coefficients were only trained with variable ozone and when using those PC coefficient files any CO2, N2O,
CO and CH4 values specified in the RTTOV input state vector are ignored and instead the fixed profiles stored in the
PC coefficient file are used. This also means that the Jacobians calculated by the RTTOV K model are zero for these
trace gas species. PC-RTTOV has different regression limits for gases than the RTTOV optical depth coefficients, but
these are treated in the way as for standard RTTOV (see section 7.3). The PC regression limits can be found on the
RTTOV coefficients download page:
https://www.nwpsaf.eu/site/software/rttov/download/coefficients/coefficient-
download/#Reference_profiles_and_regression_limits
RTTOV v12.2 also introduced new PC coefficient files which support aerosol simulations using the OPAC aerosol
components: these have “aer” in the filename. A suitable OPAC aerosol property file should be read in when the
RTTOV optical depth and PC coefficient files are read by the call to rttov_read_coefs. The aerosol simulations are
carried out as described in section 8.6, but note that you must use the Chou-scaling parameterisation
(opts%rt_ir%ir_scatt_model = 2). The PC coefficients are trained using climatological combinations of the OPAC
aerosol components: this includes only indices 1-10 from the RTTOV aerosol property files (see Table 27). RTTOV
will ignore any input aerosol concentrations for the volcanic ash and Asian dust species. The minimum/maximum
aerosol concentrations (regression limits) used in training the PC profiles can be found on the RTTOV coefficients
download page. Note that in some layers for some species there was no variability in the aerosol concentrations (this
includes the sulphates component in all layers). RTTOV automatically uses the regression limit values for those layers.
For layers where the aerosols varied in the training RTTOV will clip the input aerosol profiles to the limits if the
apply_reg_limits option is true in a similar way as for gases (see section 7.3). If the aerosol regression limits were
exceeded in one of these layers the radiance%quality(:) flags have the qflag_pc_aer_reg_limits bit set for the PC
predictor channels corresponding to the relevant profile (see section 7.8).
The PC coefficients are trained with the RTTOV atmospheric refraction option turned on. Therefore refraction is
accounted for regardless of the setting of opts%rt_all%addrefrac. It should be noted that a number of the options
available in RTTOV were not enabled in the PC-RTTOV training and as such these options should not be used with
PC-RTTOV. These include the inclusion of the solar term in short-wave channels and the Lambertian surface option.
You can call the rttov_user_options_checkinput subroutine (Annex N) to check that your options are compatible with
PC-RTTOV.
PC coefficient files are available which have been trained over all surface types. These coefficient files have “landsea”
in the filename. It is strongly recommended to set calcemis(:) to true for sea profiles: PC-RTTOV is trained using its
own sea surface emissivity model and this is used automatically for sea profiles when calcemis(:) is true. It is also
recommended to use the UW IR land surface emissivity atlas (see section 7.5) for land and sea-ice surface types as this
was used to train the PC-RTTOV coefficients. However as the new PC-RTTOV training encompasses a wide range of
surface emissivities, the use of alternative physically realistic sources for surface emissivity should be acceptable. For
78
Version : 1.3
Date : 05/03/2019
this reason RTTOV v12 carries out no checks on how surface emissivity is specified for PC-RTTOV: you must ensure
that you do not use any sea-only PC coefficients which have just “sea” in the filename over land surfaces.
PC-RTTOV has been extended to carry out simulations that can include a NLTE correction (see section 8.10) to the
TOA radiances (Matricardi et al 2016). This option is available in PC coefficient files with “nlte” in the filename and
these files can be used to optionally include NLTE effects in the spectral region at 4.3 μm for solar zenith angles
between 0 and 90 degrees for the entire viewing geometry of IASI.
As new PC-RTTOV coefficient files are created enabling more options in the simulations (for example, NLTE,
additional trace gases, aerosols), the most recent coefficient file for a sensor can replace previous PC coefficient files.
The PC-RTTOV coefficient files contain the regression coefficients used to predict the PC scores and up to 400 (IASI,
AIRS) or 600 (IASI-NG) eigenvectors to reconstruct radiances from the PC scores: regression coefficients are stored for
the predictor sets listed in Table 29. The files contain the regression coefficients for the PC-RTTOV sea surface
emissivity model (activated by setting calcemis(:) to true over sea as noted above). In addition they store the fixed
profiles of CO2, N2O, CO and CH4 used in the LBLRTM computations on which PC-RTTOV is trained for older files,
or the minimum and maximum gas and (where supported) aerosol regression limits.
Coefficient file opts%rt_ir%pc%ipcbnd opts%rt_ir%pc%ipcreg Max npcscores

IASI – global + trace 1 => full spectrum 1, 2, 3 or 4 => 300, 400, 500 or 600 400
gases + NLTE (all channels) predictors respectively
IASI – global + trace 1 => full spectrum 1, 2, 3 or 4 => 300, 400, 500 or 600 400
gases + aerosols (all channels) predictors respectively
AIRS – global 1 => full spectrum 1, 2 or 3 => 200, 300 or 400 predictors 400
(all channels) respectively
IASI-NG – sea-only 1 => full spectrum 1, 2, 3 or 4 => 300, 400, 500 or 600 600
(all channels) predictors respectively
Table 29. Available options for band and predictor sets for PC coefficient files available
at the time of the RTTOV v12.3 release.
Predictor channel sets
In order to call PC-RTTOV a specific set of channels must be specified in the chanprof(:)%chan array for each profile
being simulated. The simulated RTTOV radiances for this set of channels comprise the predictors in the PC-RTTOV
regression. The size of the predictor channel set determines the number of channels being simulated per profile. The
predictor set is selected in opts%rt_ir%pc%ipcreg. Table 29 gives the valid values for the PC-RTTOV coefficients
available at the time of the RTTOV v12.3 release. By choosing a larger predictor set one trades reduced computational
efficiency for increased accuracy.
The specific channel list for a given predictor set can be obtained directly from the PC coefficient structure in
coefs%coef_pccomp%pcreg(i,j)%predictindex(:) for band i (see below - usually 1) and predictor set j. An alternative
way of obtaining the channel list is via the subroutine rttov_get_pc_predictindex (Annex H): an example of this can
be seen in src/test/example_pc_fwd.F90. RTTOV will report an error if the input channel list (in chanprof(:)%chan)
does not match the predictor channel list.
NB The predictor channel sets are unique to each new PC-RTTOV coefficient file for a given instrument so that, for
example, the new NLTE-compatible IASI coefficients require a different predictor channel set to the RTTOV v11 IASI
PC-RTTOV coefficients.
The variable opts%rt_ir%pc%ipcbnd provides a choice of carrying out calculations for limited spectral bands, but
currently the PC coefficient files available on the website contain information for the whole spectrum only (band 1) and
hence this variable must always be set to 1.
The number of simulated principal components can vary from 1 to 400 (IASI, AIRS) or 1 to 600 (IASI-NG). Prior to
RTTOV v12.3 this was defined when calling the rttov_alloc_pccomp subroutine (see Annex D) to allocate the pccomp
output structure (see below). Note that the value supplied here is the number of principal components to be calculated
for all profiles in the call to RTTOV. Again, example_pc_fwd.F90 provides a simple example. A typical choice is 200
79
Version : 1.3
Date : 05/03/2019
principal components and 500 predictors for IASI and 100 principal components and 300 predictors for AIRS. Again,
the number of simulated PC scores can be increased in order to improve accuracy at cost in computational efficiency.
RTTOV v12.3 provides a new option opts%rt_ir%pc%npcscores which can be used to specify the number of PC
scores to use in the simulations. If this value is larger than zero then it determines the number of PC scores for the
simulation. In this case you must ensure that the pccomp structure was allocated to be large enough for the number of
PC scores specified via the option. This allows you to specify the size of pccomp for the maximum number of profiles
you wish to simulate per call to RTTOV, but then to simulate fewer profiles in a given call (which may happen, for
example, in the last batch of profiles from a larger set): previously RTTOV would not allow this. The default behaviour
of v12.3 replicates v12.2 and earlier (npcscores defaults to -1 so that the number of PC scores is determined by the size
of pccomp).
PC-RTTOV outputs
The computed PC scores are stored in the rttov_pccomp structure (see Annex O) in the pcscores(:) array. It is possible
to reconstruct radiances from the PC scores by setting opts%rt_ir%pc%addradrec to true. In this case the total
number of reconstructed radiances for all profiles (or the maximum number of profiles being passed per call to
RTTOV) must be passed in the call to rttov_alloc_pccomp (Annex D) and the reconstructed channel list
channels_rec(:) must be supplied to rttov_direct (or to the TL, AD or K model; Annexes I, J, K, L). The reconstructed
radiances are also stored in the pccomp structure in the total_pccomp(:) array, and the corresponding brightness
temperatures are in bt_pccomp(:). The channels_rec(:) input argument to rttov_direct is mandatory if
opts%rt_ir%pc%addradrec is true.
It is technically possible to run PC-RTTOV using coefficient files containing a subset of channels (created using
rttov_conv_coef.exe, see Annex A) or to read in a subset of channels in the call to rttov_read_coefs (see Annex C).
However it is important to note that the subset of extracted channels must include the set of predictor channels being
used and also any channels for which reconstructed radiances are required. In addition, the channels specified in
chanprof(:)%chan and channels_rec(:) are always indexed starting from 1 into the set of channels read into RTTOV.
This means you must keep track of the remapped channel numbers. For this reason it is not advisable to run PC-RTTOV
with channel subsets unless you are confident in what you are doing.
8.9. Simulation of hyperspectral IR sounder radiances using HTFRTC

HTFRTC is a Principal Components-based fast radiative transfer model (Havemann et al, 2018). The HTFRTC model
was introduced into RTTOV in v12.1 as an external library. In RTTOV v12.2 the implementation was improved
significantly to be easier to compile, and more flexible and efficient to run. In RTTOV v12.3 the capability was further
developed to enable all RTTOV optional variable gases, to enable all RTTOV gas units options, to enable input/output
of surface emissivities, to compute overcast radiances and simple cloudy radiances (equivalent to those computed by
RTTOV - see section 8.3), and to support some additional RTTOV options. The implementation currently enables direct
and K-model clear-sky simulations over all surface types for hyperspectral IR sounders.
The core HTFRTC code is now included within the RTTOV package and is compiled with RTTOV which means that
HTFRTC does not have to be downloaded and compiled separately. The HTFRTC coefficient files are in NetCDF
format and so RTTOV must be compiled against the NetCDF v4 library (see section 5.2).
HTFRTC simulations are enabled by setting the opts%htfrtc_opts%htfrtc option to true. When running HTFRTC
simulations the usual RTTOV rtcoef coefficient files are not required. HTFRTC simulations require two input files: one
is a static coefficient file (htfrtc_coef_static.nc) that is required for all HTFRTC simulations, and the other is specific to
the sensor being simulated (e.g. htfrtc_coef_sensor_metop_iasi.nc). Different static files are required for RTTOV v12.2
and v12.3. HTFRTC coefficients are available on the RTTOV coefficient downloads page (see section 5.1). Both
coefficient files are read by calling the rttov_read_coefs_htfrtc subroutine which is used instead of rttov_read_coefs
for HTFRTC simulations (see Annex C). The coefficient data are read into the rttov_coefs structure just as for other
RTTOV simulations. HTFRTC computes reconstructed radiances (if requested) for all channels read from the
coefficient file. By default all instrument channels are read in, but you can specify a subset of channels via the optional
channels_rec argument to rttov_read_coefs_htfrtc.
80
Version : 1.3
Date : 05/03/2019
HTFRTC makes use of only a limited number of RTTOV options and profile variables. The number of PC scores
required is specified in opts%htftrtc_opts%n_pc_in. The maximum number allowed by the HTFRTC input files is
300. If you want reconstructed radiances these are enabled by setting opts%htftrtc_opts%reconstruct to true. As
noted above radiances are calculated for all channels read from the coefficients file (i.e. the channels_rec argument to
rttov_direct and rttov_k is ignored). In RTTOV v12.3, all RTTOV optional trace gases are supported and ozone is
now optional. If a gas is not supplied, the RTTOV background profile is used. For any gas which is supplied you must
set the corresponding “gas_data” option to true, just as for RTTOV: for example, if you supply an ozone profile you
must set opts%rt_ir%ozone_data option to true. The only other RTTOV options used by HTFRTC are
opts%rt_all%use_q2m, opts%rt_all%switchrad, opts%rt_all%do_lambertian and opts%rt_all%plane_parallel:
all other options are ignored. In the case of the do_lambertian option, HTFRTC uses a fixed angle for the downwelling
radiance and makes use of the specularity skin profile variable in the same way as RTTOV.
In RTTOV v12.3, HTFRTC can optionally compute simple cloudy radiances for a single grey cloud at a given pressure
using the profile ctp and cfraction variables. This is enabled by setting opts%htfrtc_opts%simple_cloud to true.
Similarly, the new capability to output overcast radiances for each layer (equivalent to the RTTOV overcast radiance
output) is enabled by setting opts%htfrtc_opts%overcast to true. These options can be used together or alone, but
note that they increase the run-time. If you wish to generate these outputs, you must also set the options to true before
you call rttov_alloc_pccomp to ensure the relevant output arrays are allocated (though you can subsequently turn them
off for standard “clear-sky” simulations if required) - see below.
The profile variables used by HTFRTC are listed in Table 30. Profiles can be input on any set of pressure levels:
profiles do not need to have the same pressure levels, but all profiles must be on the same number of levels (as for
RTTOV).
Profile structure variables Notes

gas_units Mandatory - these affect all gas input units just as for RTTOV.
p(:), t(:), q(:) Mandatory
o3(:), co2(:), n2o(:), co(:), so2(:) Optional - set corresponding opts%rt_ir%*_data to true
s2m%p, s2m%t Mandatory
s2m%q Used if opts%rt_all%use_q2m is true
s2m%u, s2m%v Used by emissivity calculation, mandatory for sea profiles
skin%t Mandatory
skin%surftype Mandatory
skin%specularity Used if opts%rt_all%do_lambertian is true.
zenangle Mandatory
elevation Mandatory
ctp, cfraction Used if opts%htfrtc_opts%simple_cloud is true.
Table 30: profile variables used by HTFRTC.
A number of mandatory arguments to rttov_direct and rttov_k are not used by HTFRTC. These arguments must be
passed into the subroutine calls, but the structures do not need to be allocated. In particular, the chanprof(:) argument is
not used: HTFRTC computes PC scores (and optionally reconstructed radiances) for all input profiles. In order to
prevent compilation issues you can allocate a chanprof(:) array of size 0 and pass this into RTTOV.
The HTFRTC outputs are contained in the pccomp structure (see Annex O). When allocating this structure (by calling
rttov_alloc_pccomp – see Annex D) you must specify the number of PC scores (as for PC-RTTOV) and, if required,
the number of reconstructed radiances (which is the same as the number of channels read from the coefficients, as
described above). HTFRTC does not write to the radiance or transmission data structures so there is no need to
allocate these structures as for standard RTTOV calls. The same applies to the equivalent K model variables for rttov_k
calls. In RTTOV v12.3 the pccomp structure now contains a number of additional output arrays for HTFRTC which are
intended to mirror the radiance structure outputs. In particular there are “clear” and “total” outputs for PC scores,
radiances and BTs which contain the clear-sky and simple-cloudy simulated outputs if the simple_cloud option is
enabled or otherwise are identical. In addition the 100% cloudy PC scores and radiances are output for the simple cloud
scheme, and overcast PC scores and radiances if the overcast option is true. The pcscores output remains for backward
compatibility, and this is always equal to clear_pcscores. See Annex O for the full description of the pccomp structure.
As of RTTOV v12.3 HTFRTC provides more flexibility for surface emissivities. HTFRTC can still provide surface
emissivities using the PC-RTTOV model for sea surfaces and the fixed RTTOV values for land and sea-ice (0.98 and
81
Version : 1.3
Date : 05/03/2019
0.99 respectively). The emissivity (and emissivity_k) inputs are now optional: if supplied then they will return the
emissivities used/calculated by HTFRTC in emissivity(:)%emis_out and the emissivity Jacobians calculated by
HTFRTC in emissivity_k(:)%emis_in. You can also provide input emissivities to HTFRTC by passing the calcemis
argument and providing input emissivities in emissivity(:)%emis_in. This works in a similar manner to RTTOV, but it
is important to note that the input/output emissivities for HTFRTC are specified for the reconstructed radiance channels.
This means that these inputs/outputs are only available if opts%htftrtc_opts%reconstruct is true and the size of the
calcemis, emissivity and emissivity_k arrays is the number of reconstructed channels multiplied by the number of
profiles. The emissivity argument is mandatory if calcemis is supplied, and calcemis must be false for all reconstructed
radiance channels corresponding to a given profile for the input emissivities to be used - otherwise, if any member of
calcemis for a profile is true, then HTFRTC computes the emissivities internally for that profile. Input emissivities are
linearly interpolated internally from the reconstructed radiance channel wavenumbers to the HTFRTC centroid
frequencies for use the the RT calculation.
It is also possible to use the RTTOV emissivity atlases with HTFRTC: the atlas interface is the same as if calling it for
standard RTTOV simulations, including optionally initialising for use with a single instrument. The only difference is
that the chanprof argument to the rttov_get_emis subroutine is not used because emissivities are returned for all
reconstructed radiance channels in the initialised HTFRTC coefficients in the rttov_coefs structure for each profile.
When running Jacobian simulations using rttov_k you should initialise the pccomp_k structure to zero (for example
using the rttov_init_pccomp subroutine – see Annex D). HTFRTC uses an input perturbation of 1. in every channel.
The Jacobians are calculated in terms of PC score if opts%htftrtc_opts%reconstruct is false. Otherwise they are
calculated in terms of radiance if opts%rt_all%switchrad is false or in terms of BT if opts%rt_all%switchrad is true
(this is similar to PC-RTTOV).
NB It is important to note that you must supply both profiles_k_pc and profiles_k_rec arguments to rttov_k when
computing Jacobians for reconstructed radiances as HTFRTC always calculated the Jacobians for PC scores.
HTFRTC may be called via the RTTOV parallel interface in exactly the same way as for RTTOV.
Example code demonstrating a forward model HTFRTC call can be seen in src/test/example_htfrtc_fwd.F90 and this
also includes code using the emissivity atlas. This is similar to the other example code, and can be run from within the
rttov_test/ directory using the run_example_htfrtc_fwd.F90 script in the same way that the other example scripts are
run.
8.10. Inclusion of non-local thermodynamic equilibrium effects

RTTOV v12 is able to estimate NLTE effects above altitudes of ~40 km in the CO2 ν3 band (around 4.3 µm). Here,
local thermodynamic equilibrium breaks down due to the absorbtion of the strong solar radiation field. The NLTE
model has been updated since RTTOV v11 and at the time of release coefficients for the new model are only available
for IASI. NLTE coefficients for other hyperspectral sounders can be requested via the NWP SAF helpdesk.
To invoke the NLTE correction it is necessary to:

 Use an optical depth (rtcoef) coefficient file that contains the NLTE coefficients for the regression.
 Set opts%rt_ir%do_nlte_correction = .true.
As noted in section 8.8, PC-RTTOV coefficients are now available which can include the NLTE correction. To invoke
the NLTE correction for PC-RTTOV it is necessary to:
 Use an optical depth (rtcoef) coefficient file that contains the NLTE coefficients for the regression and is
compatible with PC-RTTOV (see section 8.8). This allows the addition of a NLTE correction to the PC-
RTTOV predictors.
 Use a PC-RTTOV coefficient file that has been trained including NLTE effects.
 Set opts%rt_ir%do_nlte_correction = .true.
For both classical RTTOV and PC-RTTOV the correction is valid (and is applied to channels) between 2200 cm−1 –
2400 cm−1 which corresponds to:
 IASI channels 6221 – 7021 (801 channels)
82
Version : 1.3
Date : 05/03/2019
The radiance correction scheme is documented in Matricardi et al (2016) and consists of eight predictors. These
predictors consist of various combinations of the solar zenith angle, θ sol the sensor zenith angle, θsat and the average
kinetic temperature in two broad atmospheric layers above ~51 hPa (i.e. the average temperature between 0.005 hPa and
~0.2 hPa, T1 and the average temperature between ~0.3 hPa and ~51 hPa, T2). The predictors p are shown in Table 31.
Predictor number Predictor

1 constant
2 cos(θsol)
3 (cos(θsol))0.5
4 cos(θsol) sec(θsat)
5 (cos(θsol) sec(θsat))2
6 cos(θsol) T1
7 cos(θsol) T2
8 sec(θsat) T1
9 sec(θsat) T2
Table 31: the predictors used in the PC-RTTOV NLTE algorithm
The radiance correction, Rch

NLTE
, is added to the LTE TOA radiance to give the NLTE TOA radiance, thus,
RchNLTE  RchLTE  RchNLTE , (9)

where R
NLTE
ch is written as:
9
RchNLTE   cch , j p j (10)
j 1
Here cch , j are the regression coefficients.

You are responsible for ensuring that input values for the solar zenith angle are in range (i.e. >0° and < 90°). For any
particular profile, RTTOV will not calculate the correction if the supplied solar zenith angle lies out of range and will
extrapolate the correction if the supplied satellite zenith angle is out of range (as long as the solar zenith angle is valid).
Note that the NLTE correction is distinct from the solar simulation capability and as such it is not necessary to set
opts%rt_ir%addsolar to true for the NLTE correction to be applied. The NLTE correction may be used in conjunction
with IR scattering calculations, but you should be aware that the NLTE correction is added to the final LTE
cloudy/aerosol-affected radiances and so the NLTE radiation does not interact with the scattering particles.
A more detailed discussion of the science and the impact of including the new NLTE correction is available in the
RTTOV v 12 Science and Validation Report.
8.11. Option to treat surface as a Lambertian reflector

The reflection of downward radiation over snow or multi-year sea-ice is better characterised by assuming a Lambertian
approximation rather than the specular reflection which is the default in RTTOV. This is particularly relevant for
microwave sounders as sea-ice and snow have high reflectances at these frequencies. True Lambertian reflection
requires an integral over a range of angles to cover the hemisphere which would prove difficult (and costly) in the
RTTOV framework. Fortunately Matzler (2005) has developed an approximation as a function of optical depth
providing a fixed angle of ~55º to be used for the downward radiation. The option of Lambertian reflection is now
possible in the MW and the IR. To invoke Lambertian reflection set the opts%rt_all%do_lambertian flag to true. By
default it is false. Significant changes (up to 10K) will be seen in microwave window channels with the largest
differences for nadir views. More details are provided in the RTTOV v11 Science and Validation Report.
RTTOV v12.3 includes a new option (opts%rt_all%lambertian_fixed_angle) to enable use of the parameterisation
(Guedj et al, 2010) of the downwelling angle in terms of the total atmospheric optical depth: if false (the default), the
fixed angle of 55º is used, otherwise if true the parameterisation for the angle is used.
83
Version : 1.3
Date : 05/03/2019
In addition RTTOV v12.3 introduces a new profile variable: profiles(:)%skin%specularity. Most surfaces are not
truly Lambertian or truly specular, but lie somewhere between the two. This variable is active when the do_lambertian
option is true, and specifies the weighting applied when linearly combining the downwelling specular and Lambertian
radiances. The valid range of values is 0 (fully Lambertian) to 1 (fully specular - the same as if do_lambertian is false).
NB For profiles/channels where an internal sea surface emissivity model is being used (i.e. FASTEM or TESSEM2 in
the MW and ISEM or IREMIS in the IR) the Lambertian option is not valid and so is not applied for these channels.
When activated the Lambertian option is applied for sea surfaces where calcemis(:) is false and for land and sea-ice
surfaces regardless of calcemis(:).
8.12. Zeeman effect for SSMIS and AMSU-A

For microwave sensors that have high peaking weighting functions in the mesosphere such as SSMIS, channels close to
lines of molecular oxygen may be significantly affected by the redistribution of line intensity through Zeeman splitting
as described in the RTTOV v10 Science and Validation Report. The absorption for the affected channels will depend on
the strength and orientation of the magnetic field. You must specify two input variables for the geomagnetic field in the
rttov_profiles structure, these being the magnitude, Be, of the field and the cosine, cosbk, of the angle between the
field vector and the viewing path considered. For SSMIS, values will be available with the satellite data stream, and will
therefore already match the geographical location and orientation of the viewing path. For AMSU-A, this is not the
case, but the values may be obtained from a pre-computed look-up table. For instance, the rttov_zutility module
provided in the src/other/ directory may be used to provide values (see Annex N). For a normal run where the Zeeman
effect is not computed the variables can be set to any value, including zero, but if the Zeeman effect is to be calculated,
Be should lie in the range 0.2-0.7 gauss as this covers the range of values over which the Zeeman coefficients were
trained. In particular, Be must not be set to zero when calling RTTOV with a Zeeman coefficient file: if the Zeeman
effect is not important for an application a non-Zeeman coefficient file should be used instead.
A ‘Zeeman’ coefficient file will have the Zeeman flag set to unity in the ‘Fast Model Variables’ section and “zeeman”
in the filename. To include the Zeeman effect for a given sensor, you must run RTTOV v12 with a Zeeman coefficient
file.
As with all RTTOV simulations your input profile pressure levels should reach sufficiently high that they span the
weighting functions of the channels being simulated (i.e. sufficiently high that there is insignificant absorption above
the top-most input pressure level). If this condition is not met there can be significant errors resulting from the treatment
of emission from the region of atmosphere above the top input level. After interpolating the calculated optical depth
profile onto the input pressure levels, RTTOV by default sets the optical depth of the top-most level to zero. This
behaviour is intended to mitigate the case where there is significant absorption above the top level: in this case the top
layer is effectively stretched to reach the space boundary and the effects of emission and absorption from the region of
atmosphere above the top level are included in the integration of the radiative transfer equation. However without an
accurate representation of the temperature of the atmosphere above the top level the emission term will be in error to
some degree. Note that this feature has negligible impact if the top input profile level is sufficiently high because in that
case the interpolated optical depth at the top level will be close to zero anyway. This feature is enabled by setting
opts%interpolation%spacetop to true which is the case by default. If this flag is set to false the interpolated optical
depth at the top level is not modified: when the radiative transfer equation is integrated, the absorption due to the
atmosphere above the top input level is accounted for, but emission from this region of atmosphere is omitted entirely.
In general there is no need or benefit to set this flag to false. Note that the spacetop flag affects all simulations, not only
Zeeman-affected simulations.
For SSMIS channels 19-22 which are affected by Zeeman splitting, the brightness temperature in channel 20 may be
altered by as much as 10 K – the change in column absorption will shift the channel weighting function, but the effect
of this will actually depend on the temperature profile. When you run RTTOV with a non-Zeeman coefficient file, the
mixed gas prediction scheme will be based on the usual set of predictors. However, when a Zeeman coefficient file is
used, the mixed gas scheme will incorporate additional predictors used for the high peaking channels. In the optical
depth calculation for channels 1-18 and 23-24 (non-Zeeman), contributions from the additional predictors will be
nullified by zero coefficients. In contrast, for channels 19-22 (Zeeman), it is only the contributions from the new
predictors that contribute.
For AMSU-A, only channel 14 is affected. This channel, while dominated by oxygen absorption, sounds lower down in
the atmosphere than the Zeeman channels of SSMIS, and it is also located further from the oxygen line centres. The
84
Version : 1.3
Date : 05/03/2019
impact is therefore much smaller (~0.5K). If you run with a non-Zeeman coefficient file, all channels will use the usual
set of mixed gas predictors and the Zeeman effect will not be represented in channel 14. If a Zeeman coefficient file is
used, then a small set of additional predictors will be included. These will contribute for channel 14 but will be nullified
for the other channels by zero coefficients.
8.13. Simulation of SSU radiances

For SSU, which uses pressure modulated gas cells to define the channels, RTTOV v12 allows you to take some account
of inadvertent cell pressure changes that may have occurred over the lifetime of the instrument.
SSU coefficients were updated for RTTOV v10 using more recent molecular spectroscopy (LBLRTMv12.0), more
variable gases (now O3 as well as H2O, CO2) and more stratospheric levels (now 51). RTTOV v11 provided an
additional set of coefficients with the label ‘pmcshift’ in the file name which are based on the same nominal set of cell
pressures, and differ only in having a new ‘PRESSURE_MODULATED_CELL’ section.
When simulating SSU with these pressure modulated cell (PMC) shift coefficients, you must now provide a preferred
set of gas cell pressures (in hPa), one for each channel. These may be different from those in the nominal set and
should be assigned to the channel array coefs%coef%pmc_ppmc. This should be done after the coefficient file has
been read using the rttov_read_coefs subroutine. It is also mandatory to supply an input CO2 profile when using the
PMC shift coefficients and this also feeds into the cell pressure scheme.
Note that the SSU coefficient files do not support the new IREMIS sea surface emissivity model so you must set
opts%rt_ir%ir_sea_emis_model to 1 to select ISEM otherwise RTTOV will fail with an error. In practice this makes
no difference to the simulations as SSU is not sensitive to the surface.
8.14. Simulations with variable SO2

RTTOV v12 introduces new optical depth coefficient files which allow for variable SO 2. SO2 is treated in the same way
as other gases. You should set the opts%rt_ir%so2_data flag to true to indicate you are supplying SO2 profiles and the
input profile data are specified in profiles(:)%so2(:). The SO2 optical depth prediction has been trained using a
selection of profiles which cover both “clean” (low-SO2) and “volcanic” (high-SO2) atmospheres. If you do not supply
an SO2 profile with an SO2-enabled coefficient file then RTTOV uses a “clean” background SO2 profile.
The SO2-enabled coefficients show slightly larger errors (as measured by comparisons to the line-by-line model used to
train RTTOV) in the spectral regions where SO2 is active for “clean” SO2 profiles. For this reason coefficients are
available for all trace gases with and without variable SO 2. SO2-enabled coefficients have been generated for
hyperspectral sounders and are identified by “_so2” in the filename: they can be downloaded from the RTTOV
coefficients web page. More detailed information about the new coefficients is available in the RTTOV v12 Science
and Validation Report.
85
Version : 1.3
Date : 05/03/2019
9. Limitations of RTTOV v12

There are a number of scientific limitations of RTTOV v12 you should be aware of. The main ones are listed here:
 RTTOV v12 only simulates top of atmosphere radiances from a nadir or off-nadir view which intersects with the
Earth’s surface (i.e. no limb paths or upward viewing paths).
 RTTOV v12 only allows for water vapour, ozone, carbon dioxide, nitrous oxide, methane, carbon monoxide and
sulphur dioxide to be variable gases with all others included in the mixed gases transmittance calculation.
 RTTOV v12 can only simulate radiances for instruments for which a coefficient file has been generated. The
instruments currently supported are listed in Table 3. Only sensors with channels at wavelengths greater than 0.4
microns can be simulated with RTTOV v12.
 The accuracy of simulations for very broad channels (e.g. SEVIRI channel 4 at 3.9 microns) is poor with significant
biases noted (~1-2K) (see e.g. Brunel and Turner, 2003). This is the case for all versions of RTTOV. A work around is
to use Planck weighted coefficient files (which are now standard for all sensors where this is a problem) resulting in
much lower biases. Whether coefficients are Planck-weighted can be determined by examining the
PLANCK_WEIGHTED section in the coefficient file (if it is not present, there are no Planck-weighted channels).
 PC-RTTOV computations are limited by the configuration of the coefficient training. More information is given in
section 8.8.
 HTFRTC: TL and AD models are not implemented.
10.Reporting and known bugs for RTTOV v12

Bug reports or other comments/feedback can be submitted via the NWP SAF helpdesk: http://nwpsaf.eu/site/helpdesk/.
Select RTTOV as the “department” and include the following information:
- RTTOV version number (i.e. 12.3)
- Platform and operating system you are running the code on (e.g. Linux PC, IBM, Cray)
- Compiler used (e.g. gfortran, ifort, pgf90, etc) and compilation flags
- Classification of report as: serious, cosmetic or improvement
- Report of problem including any input /output files the SAF can use to reproduce the problem
Once the problem has been analysed it will be posted on the RTTOV web site
(http://nwpsaf.eu/site/software/rttov/rttov-v12/code-updates/) with a description of the fix if appropriate. There is also a
RTTOV v12 email list where major bugs are announced which you can subscribe to by contacting the help desk. When
you register to download RTTOV you will be automatically included on this list unless you indicate otherwise.
The known issues in RTTOV v12.3 are given below. Any further problems and corrections will be provided via the
RTTOV v12 web page as they become known:
http://nwpsaf.eu/site/software/rttov/rttov-v12/code-updates/
i. On the NEC architecture for RTTOV v10 some subroutines could not be compiled with optimisations. See the
nec-meteofrance file in the build/arch/ directory. The internal RTTOV interpolator was found to run
relatively slowly on the NEC when compared to other platforms. Due to unavailability of a NEC platform
RTTOV v12 has not been tested on this architecture, but similar issues may be expected.
ii. RTTOV v12 compiled with NAG Fortran v6.0 (build 1032) fails when running the test suite. This appears to
be a compiler bug related to reading nested structures in a namelist. Since the example programs and tests
outside of the rttov_test.pl test harness run OK it is believed that RTTOV is compatible with this compiler,
but thorough testing has not been undertaken. NAG v6.1 does not have this problem.
11.Frequently asked questions

This section has now been put on the RTTOV v12 web site to allow updating:
http://nwpsaf.eu/site/software/rttov/documentation/faq/
86
Version : 1.3
Date : 05/03/2019
12.Glossary
AD RTTOV Adjoint model
AMSU Advanced Microwave Sounding Unit
ATMS Advanced Technology Microwave Sounder
ATOVS Advanced TIROS Operational Vertical Sounder
B parameter/BPR Back-scattering parameter for IR scattering simulations
BRDF Bi-directional Reflectance Distribution Function
BRF Bi-directional Reflectance Factor
BT Brightness Temperature
CAMS Copernicus Atmosphere Monitoring Service
CLW Cloud Liquid Water
CNRM Centre National de Recherches Météorologiques
DOM Discrete Ordinates Method
ECMWF European Centre for Medium-Range Weather Forecasts
EUMETSAT European Organisation for the Exploitation of Meteorological Satellites
FASTEM MW surface emissivity model
GCM Global Circulation Model
HDF5 Hierarchical Data Format version 5
HTFRTC Havemann-Taylor/HyperspecTral Fast Radiative Transfer Code, a PC-based fast RT model
IR Infrared
IREMIS New physically-based IR sea surface emissivity model
ISEM Older IR sea surface emissivity model, depends only on zenith angle
IWC Ice Water Content
K RTTOV Jacobian model
LBLRTM Line-By-Line Radiative Transfer Model used to generate RTTOV coefficients for visible/IR
sensors
Liebe-89 MPM Line-by-line model used to generate RTTOV coefficients for MW sensors
MFASIS a Method for FAst Satellite Image Simulation, a fast visible cloudy radiance solver
MODIS Moderate resolution Imaging Spectroradiometer
MHS Microwave Humidity Sounder
MW Microwave
NIR Near-Infrared (see VIS below)
NLTE Non Local Thermodynamic Equilibrium
NWP Numerical Weather Prediction
OPAC Optical Properties of Aerosols and Clouds
OpenMP Application Programming Interface supporting multi-platform shared-memory parallel
programming
PC Principal Components
PMC Pressure Modulated Cell
PW Planck-Weighted
RTTOV Radiative Transfer for TOVS
RTTOV-SCATT RTTOV interface for MW cloud and hydrometeor scattering simulations
TELSEM2 A Tool to Estimate Land Surface Emissivities from Microwaves to Millimetres
TESSEM2 A Tool to Estimate Sea Surface Emissivities from Microwaves to Millimetres
TL RTTOV Tangent Linear model
SAF Satellite Applications Facility
SEVIRI Spinning Enhanced Visible and Infrared Imager
SSMI/S Special Sensor Microwave Imager/Sounder
SSU Stratospheric Sounding Unit
TIROS Television Infrared Observation Satellite
TOVS TIROS Operational Vertical Sounder
VIS Visible (here used synonymously with VIS/NIR)
87
Version : 1.3
Date : 05/03/2019
13.References
Aires F., C. Prigent, F. Bernado, C. Jiménez, R. Saunders, and P. Brunel, 2010. A tool to estimate Land Surface
Emissivities at Microwaves frequencies (TELSEM) for use in numerical weather preduction. Q.J.R. Meteorol. Soc. 137,
690-699. doi: 10.1002/qj.803.
Baordo, F. and Geer, A.J., 2016. Assimilation of SSMIS humidity‐sounding channels in all‐sky conditions over land
using a dynamic emissivity retrieval. Q.J.R. Meteorol. Soc., 142, 2854-2866.
Bauer P., E. Moreau, F. Chevallier, and U. O'Keeffe 2006 Multiple-scattering microwave radiative transfer for data
assimilation applications. Q.J.R. Meteorol. Soc. 132, 1259-1281.
Baum, B. A., P. Yang, A. J. Heymsfield, C. Schmitt, Y. Xie, A. Bansemer, Y. X. Hu, and Z. Zhang, 2011.
Improvements to shortwave bulk scattering and absorption models for the remote sensing of ice clouds. J. Appl. Meteor.
Clim., 50, 1037-1056.
Bormann, N, A. Geer, S. English, 2012. Evaluation of the microwave ocean surface emissivity model FASTEM-5 in the
IFS. ECMWF Technical Memorandum 667.
Borbas, E. E. and B. C. Ruston, 2010. The RTTOV UWiremis IR land surface emissivity module. NWP SAF report.
https://www.nwpsaf.eu/publications/vs_reports/nwpsaf-mo-vs-042.pdf
Borbas, E, 2014. The RTTOV UWiremis module Investigation into the angular dependence of IR surface emissivity.
NWP SAF report. https://www.nwpsaf.eu/publications/vs_reports/nwpsaf-mo-vs-050.pdf
Borbas E. and Feltz M., 2019: Updating the CAMEL surface emissivity atlas for RTTOV. NWP SAF report NWPSAF-
MO-VS-058. https://www.nwpsaf.eu/publications/vs_reports/nwpsaf-mo-vs-058.pdf
Boudala, F.S., Isaac, G.A., Fu, Q., and Cober, S.G., 2002: Parameterization of effective ice particle size for high latitude
clouds. Int. J. Climatol., 22, 1267-1284.
Bozzo, A., Remy, S., Benedetti, A., Flemming, J., Bechtold, P., Rodwell, M.J., Morcrette, J.J., 2017: Implementation of
a CAMS-based aerosol climatology in the IFS. ECMWF Technical Memorandum 801
Brunel, P. and S. Turner 2003 On the use of Planck-weighted transmittances in RTTOV presented at the 13th
International TOVS Study Conference, Ste Adele, Canada 29 Oct – 4 Nov 2003.
http://cimss.ssec.wisc.edu/itwg/itsc/itsc13/thursday/brunel_poster.pdf
Chen, Y., Y. Han, P. van Delst, F. Weng, 2013: Assessment of Shortwave Infrared Sea Surface Reflection and Nonlocal
Thermodynamic Equilibrium Effects in the Community Radiative Transfer Model Using IASI Data. J. Atmos. Oceanic
Technol., 30, 2152–2160. doi: http://dx.doi.org/10.1175/JTECH-D-12-00267.1
Clark, R.N., Swayze, G.A., Wise, R., Livo, E., Hoefen, T., Kokaly, R., Sutley, S.J., 2007, USGS digital spectral library
splib06a: U.S. Geological Survey, Digital Data Series 231, http://speclab.cr.usgs.gov/spectral.lib06.
Deblonde, G., 2000. Evaluation of FASTEM and FASTEM2, NWP SAF report. NWPSAF-MO-VS 1. Available here:
https://www.nwpsaf.eu/site/download/documentation/rtm/papers/evalfastems.pdf
https://www.nwpsaf.eu/site/download/documentation/rtm/papers/evalfastemsfigs.pdf
Elfouhaily, T., Chapron, B., Katsaros, K., and Vandemark, D., 1997: A unified directional spectrum for long and short
wind-driven waves, J. Geophys. Res., 102(C7), 15781-15796, doi:10.1029/97JC00467.
English, S.J. and T.J. Hewison, 1998: A fast generic microwave emissivity model, Proceedings of SPIE, 3503,
Microwave Remote Sounding of the Environment, Eds. T Hayasaka, D.L. Wu, Y. Jin and J. Jiang, 288-300
Eyre J. R. 1991 A fast radiative transfer model for satellite sounding systems. ECMWF Technical Memorandum 176
88
Version : 1.3
Date : 05/03/2019
Geer A.J., P. Bauer and C. W. O’Dell, 2009a: A revised cloud overlap scheme for fast microwave radiative transfer in
rain and cloud, J. App. Met. Clim., 48, 2257–2270
Geer, A.J., R.M. Forbes and P. Bauer, 2009b: Cloud and precipitation overlap in simplified scattering radiative transfer,
EUMETSAT/ECMWF Fellowship Programme Research Report no. 18. http://www.ecmwf.int/en/research/publications
Guedj S., Karbou F., Rabier F., and Bouchard A., 2010: Toward a Better Modeling of Surface Emissivity to Improve
AMSU Data Assimilation Over Antarctica. IEEE Trans. Geosci. Remote Sensing 48 4
https://doi.org/10.1109/TGRS.2009.2036254
Han, H.-J., B.-J. Sohn, H.-L. Huang, E. Weisz, R. Saunders, and T. Takamura, 2012. An improved radiance simulation
for hyperspectral infrared remote sensing of Asian dust, J. Geophys. Res., 117, D09211, doi:10.1029/2012JD017466.
Hasselmann, K. et al. (1973) Measurements of wind-wave growth and swell during the Joint North Sea Wave Project
(JONSWAP) Dtsch.Hydrogr.Z., 12, 95 pp.
Havemann, S., J.-C. Thelen, J. P. Taylor and R. C. Harlow, 2018. The Havemann-Taylor Fast Radiative Transfer Code
(HT-FRTC): a multipurpose code based on Principal Components. JQSRT, 220, 180-192.
https://doi.org/10.1016/j.jqsrt.2018.09.008
Hocking, J., 2014: Interpolation methods in the RTTOV fast radiative transfer model. Met Office Forecasting Research
Technical Report 590. Available here: http://www.metoffice.gov.uk/binaries/content/assets/mohippo/pdf/i/k/frtr590.pdf
Johnson, B., K. Turnbull, P. Brown, R. Burgess, J. Dorsey, A. J. Baran, H. Webster, J. Haywood, R. Cotton, Z.
Ulanowski, E. Hesse, A. Woolley, and P. Rosenberg, 2012: In situ observations of volcanic ash clouds from the FAAM
aircraft during the eruption of Eyjafjallajökull in 2010, J. Geophys. Res., 117, D00U24, doi:10.1029/2011JD016760.
Karbou, F., E.Gérard, and F. Rabier, 2006: Microwave land emissivity and skin temperature for AMSU-A and -B
assimilation over land, Q.J.R. Meteorol. Soc. 132, No. 620, Part A, pp. 2333-2355(23), doi :10.1256/qj.05.216
Karbou, F., E. Gérard, and F. Rabier, 2010: Global 4DVAR assimilation and forecast experiments using AMSU
observations over land. Part I: Impacts of various land surface emissivity parameterizations. Wea. Forecasting, 25, 5–
19.
Kazumori, M. and S.J. English, 2015: Use of the ocean surface wind direction signal in microwave radiance
assimilation, Q.J.R. Meteorol. Soc. 141, No. 689, Part B, pp 1354–1375, doi: 10.1002/qj.2445
Liebe, H.J., 1989: An atmospheric millimeter-wave propagation model. Int. J. of Infrared and Millimeter Waves, 10, 6.
Liu Q. and F. Weng, 2003. Retrieval of sea surface wind vector from simulated satellite microwave polarimetric
measurements. Radio Sci. 38, 8078, doi: 10.1029/2002RS002729.
Liu, Q., F. Weng, and S. English, 2011: An Improved Fast Microwave Water Emissivity Model, IEEE Geosci. Remote
Sensing, 49 1238 – 1250. doi: 10.1109/TGRS.2010.2064779.
Matricardi, M., F. Chevallier and S. Tjemkes 2001 An improved general fast radiative transfer model for the
assimilation of radiance observations. ECMWF Technical Memorandum 345.
http://www.ecmwf.int/en/research/publications
Matricardi, M. 2003 RTIASI-4, a new version of the ECMWF fast radiative transfer model for the infrared atmospheric
sounding interferometer. ECMWF Technical Memorandum 425. http://www.ecmwf.int/en/research/publications
Matricardi, M., 2005 The inclusion of aerosols and clouds in RTIASI, the ECMWF fast radiative transfer model for the
Infrared Atmospheric Sounding Interferometer. ECMWF Technical Memorandum 474.
Matricardi, M. 2008 The generation of RTTOV regression coefficients for IASI and AIRS using a new profile training
set and a new line-by-line database. ECMWF Technical Memorandum 564.
89
Version : 1.3
Date : 05/03/2019
Matricardi, M. 2010 A principal component based version of the RTTOV fast radiative transfer model. ECMWF
Technical Memorandum 617.
Matricardi M., López Puertas M., and Funke B. 2018: Modelling of nonlocal thermodynamic equilibrium effects in the
principal component based version of the RTTOV fast radiative transfer model. JGR Atmos. 123, 11, 5741-5761
https://doi.org/10.1029/2018JD028657
Matzler, C. 2005 On the determination of surface emissivity from Satellite observations. Geoscience and Remote
Sensing Letters, 2, 160-163.
Mayer, B. and Kylling, A., 2005. Technical note: The libRadtran software package for radiative transfer calclations -
description and examples of use, Atmos. Chem. Phys., 5, 1855-1877.
McFarquar, G.M., Iacobellis, S. & Somerville, R.C.J., 2003 : SCM simulations of tropical ice clouds using
observationally based parameterizations of microphysics. J. Clim., 16, 1643-1664.
Ou, S. & Liou, K.-N., 1995: Ice microphysics and climatic temperature feedback. Atmos. Res., 35, 127-138.
Pollack, J. B., O. B. Toon, and B. N. Khare (1973), Optical properties of some terrestrial rocks and minerals, Icarus, 19,
372–389, doi:10.1016/0019-1035(73)90115-2.
Prigent, C., Aires, F., Wang, D., Fox, S. and Harlow, C., 2016: Sea surface emissivity parameterization from
microwaves to millimeter waves. Q.J.R. Meteorol. Soc. Accepted Author Manuscript. doi:10.1002/qj.2953
Rochon, Y., L. Garand, D.S. Turner and S. Polavarapu. 2007: Jacobian mapping between vertical co-ordinate systems
in data assimilation. Q.J.R. Meteorol. Soc. 133 1547-1558.
Rosenkranz, P.W., 2015: A Model for the Complex Dielectric Constant of Supercooled Liquid Water at Microwave
Frequencies. IEEE Trans. on Geosci. and Remote Sensing, 53, 3, 1387-1393.
Saunders R.W., M. Matricardi and P. Brunel 1999 An Improved Fast Radiative Transfer Model for Assimilation of
Satellite Radiance Observations. Q.J.R. Meteorol. Soc. 125, 1407-1425.
Saunders, R., Hocking, J., Turner, E., Rayer, P., Rundle, D., Brunel, P., Vidot, J., Roquet, P., Matricardi, M., Geer, A.,
Bormann, N., and Lupu, C., 2018: An update on the RTTOV fast radiative transfer model (currently at version 12),
Geosci. Model Dev., 11, 2717-2737, https://doi.org/10.5194/gmd-11-2717-2018
Scheck, L., P. Frèrebeau, R. Buras-Schnell, B. Mayer, 2016: A fast radiative transfer method for the simulation of
visible satellite imagery. JQSRT, 175, 54-67.
Sherlock, V. 1999 ISEM-6: Infrared Surface Emissivity Model for RTTOV-6. NWP SAF report.
https://www.nwpsaf.eu/site/download/documentation/rtm/papers/isem6.pdf
Turner, D.D., Kneifel, S., and Cadeddu, M.P. 2016: An Improved Liquid Water Absorption Model at Microwave
Frequencies for Supercooled Liquid Water Clouds. J. Atmos. Oceanic Tech., 33, 33-44, doi: 10.1175/JTECH-D-15-
0074.1
Vidot, J. and E. Borbas, 2013: Land surface VIS/NIR BRDF atlas for RTTOV-11: Model and Validation against
SEVIRI Land SAF Albedo product. Q.J.R. Meteorol. Soc. 140, 2186–2196, doi: 10.1002/qj.2288
Vidot, J., A. J. Baran, and P. Brunel, 2015: A new ice cloud parameterization for infrared radiative transfer simulation
of cloudy radiances: Evaluation and optimization with IIR observations and ice cloud profile retrieval products. J.
Geophys. Res. Atmos., 120, 6937–6951. doi: 10.1002/2015JD023462.
Volz, F. E. (1972), Infrared refractive index of atmospheric aerosol substances, Appl. Opt., 11, 755–759,
doi:10.1364/AO.11.000755.
90
Version : 1.3
Date : 05/03/2019
Volz, F. E. (1973), Infrared optical constants of ammonium sulfate, Sahara dust, volcanic pumice, and flyash, Appl.
Opt., 12, 564–568, doi:10.1364/ AO.12.000564.
Wang, D., C. Prigent, L. Kilic, S. Fox, R. C. Harlow, C. Jimenez, F. Aires, C. Grassotti, and F. Karbou, 2017. Surface
emissivity at microwaves to millimeter waves over Polar Regions: parameterization and evaluation with aircraft
experiments. J. Atmos. and Oceanic Tech. https://doi.org/10.1175/JTECH-D-16-0188.1
Wu, X. and W.L. Smith, Emissivity of rough sea surface for 8-13\mum: modeling and validation, Appl. Opt. 36, 1-11
Wyser, K., 1998: The effective radius in ice clouds. J. Clim., 11, 1793-1802.
91
Version : 1.3
Date : 05/03/2019
14.Annexes
Annex A - Coefficient information and conversion tools
1. RTTOV_COEF_INFO.EXE
The program rttov_coef_info.exe (located in the bin/ directory) can be used to display information about any given
rtcoef_ coefficient file. This is particularly useful for determining the contents of binary or HDF5 coefficient files.
The usage is as follows:

$ rttov_coef_info.exe --coef ... -–format FORMATTED|UNFORMATTED|HDF5 --verbose
RTTOV will usually determine format of the coefficient file automatically so the --format argument is not
generally required. The --verbose option prints out additional per-channel information so the amount of output is
greatly increased for hyperspectral sounders.
Argument Description
--coef Input coefficient file.
--format Format of coefficient file (optional)
--verbose Include additional per-channel information (optional).
2. RTTOV_MFASIS LUT_INFO.EXE
The program rttov_mfasis_lut_info.exe (located in the bin/ directory) can be used to display information about a given
HDF5 MFASIS LUT file. This is particularly useful for determining the cloud liquid and ice water schemes used for
training the LUT and for finding out the range of values for which the LUT is trained in each dimension (see section
8.5).

$ rttov_coef_info.exe –mfasis_lut ... --verbose
The executable only reads files in HDF5 format. The --verbose option additionally prints out the “README_LUT”
section of the file which may contain additional information.
--mfasis_lut Input MFASIS LUT file.
--verbose Include additional per-channel information (optional).
3. RTTOV_CONV_COEF_.EXE
The program rttov_conv_coef.exe (located in the bin/ directory of the RTTOV build) is used to convert coefficient
files between ASCII, Fortran unformatted (binary) and HDF5 formats and to create coefficient files for subsets of
channels. A help message can be displayed as follows:
$ rttov_conv_coef.exe --help

$ rttov_conv_coef.exe \
--format-in FORMATTED|UNFORMATTED|HDF5 \
--format-out FORMATTED|UNFORMATTED|HDF5 \
--channels 1 2 3 4 5 ... \
--coef-in ... --scaer-in ... --sccld-in ... --pccoef-in ... \
--coef-out ... --scaer-out ... --sccld-out ... --pccoef-out ... \
--hdf5-reals32 --force-single --all-in-one --compress
92
Version : 1.3
Date : 05/03/2019
--format-in FORMATTED | Format of input coefficient file(s). FORMATTED=ASCII;
UNFORMATTED | HDF5 UNFORMATTED=binary, HDF5 only applicable if RTTOV
compiled with HDF5 capability (optional).
--format-out FORMATTED | Format of output coefficient files(s).
UNFORMATTED | HDF5
--channels 1 2 3 4 5 ... List of channels to extract (optional)
--coef-in/--coef-out Input/output RTTOV coefficient file (output file optional).
--scaer-in/--scaer-out Input/output aerosol scattering coefficient file (optional).
--sccld-in/--sccld-out Input/output cloud scattering coefficient file (optional).
--pccoef-in/--pccoef-out Input/output Principal Components coefficient file (optional).
--hdf5-reals32 If present store ALL reals in HDF5 32 bits, default is the RTTOV
real Kind. Only applicable to HDF5 output (optional).
--force-single If present store real 2D and larger arrays in HDF5 32 bits, default
is the RTTOV default real kind. Only applicable to HDF5 output
(optional).
--all-in-one If present write all coefs (optical depth, scattering, PC) to a single
output file. Only applicable to HDF5 output (optional).
--compress If present use HDF5 internal GZIP compression, only useful for
hi-res sounders. Only applicable to HDF5 output (optional).
Most arguments are optional, though both --coef-in and --format-out must be specified at least. RTTOV will
try to determine the input file format if this is not specified. Note that in all cases an optical depth coefficient file
(rtcoef_*) will be created: if --coef-out is not specified this file is written to the directory containing the input
coefficient file.
As described in section 7.4, when you extract some subset of n channels to a new coefficient file the channels will then
be identified by the indices 1 to n in RTTOV and not by the original channel numbers. If you are carrying out PC
calculations the channels that RTTOV must simulate are prescribed by the PC predictor selection chosen as described in
section 8.8. In this case, if you wish to create a smaller coefficient file for use with these simulations, it is strongly
recommended that you extract specifically the set of channels used as predictors for the PC calculations you require: if
you extract some other subset of channels (which must in any case be a superset of the necessary PC predictor channels)
it will become complicated to manage the channel numbering correctly.
Example 1 – convert an HDF5 file to binary format:
$ rttov_conv_coef.exe --format-out unformatted --coef-in rtcoef_eos_2_airs.H5

--coef-out rtcoef_eos_2_airs.bin
Example 2 – extract a subset of channels to a compressed HDF5 file:
$ rttov_conv_coef.exe --format-out hdf5 --coef-in rtcoef_eos_2_airs.H5 \

--coef-out rtcoef_eos_2_airs_subset.H5 -–compress --channels 10 20 30 ...
93
Version : 1.3
Date : 05/03/2019
4. RTTOV_ASCII2BIN_SCATTCOEF.EXE
The program rttov_ascii2bin_scattcoef.exe (located in the bin/ directory) can be used to convert ASCII Mie table files
to binary format.

$ rttov_ascii2bin_scattcoef.exe --coef-in ... -–coef-out ...
--coef-in Input ASCII RTTOV-SCATT Mie table file.
--coef-out Output binary format Mie table file.
5. RTTOV11_CONV_COEF_11TO12.EXE
The program rttov11_conv_coef_11to12.exe (located in the bin/ directory) can be used to convert v10/v11-format
coefficient files to v12 format. This works with ASCII and HDF5 files and the output file is in the same format as the
input file.
The NLTE coefficients in v11 files are not compatible with the v12 NLTE model so these are not copied to the output
file. The output v12-format file will not include coefficients for the IREMIS sea surface emissivity model so you must
select ISEM as the emissivity model for IR sensors when running RTTOV (you can manually copy the IREMIS section
from an original v12 coefficient file for the same sensor to the converted file).

$ rttov11_conv_coef_11to12.exe --coef-in ... -–coef-out ...
--coef-in Input ASCII or HDF5 v10/v11 RTTOV coefficient file.
--coef-out Output v12-compatible RTTOV coefficient file.
6. RTTOV11_CONV_COEF_12TO11.EXE
The program rttov11_conv_coef_12to11.exe (located in the bin/ directory) can be used to convert v12-format
coefficient files to v10/v11 format. This works with ASCII and HDF5 files and the output file is in the same format as
the input file.
The NLTE coefficients in v12 files are not compatible with the v11 NLTE model so these are not copied to the output
file. Similarly the IREMIS emissivity model is not implemented in v11 so these coefficients are not copied either.

$ rttov11_conv_coef_12to11.exe --coef-in ... -–coef-out ...
--coef-in Input ASCII or HDF5 v12 RTTOV coefficient file.
--coef-out Output v10/v11-compatible RTTOV coefficient file.
94
Version : 1.3
Date : 05/03/2019
Annex B – RTTOV_ERRORHANDLING interface

call rttov_errorhandling (err_unit)
rttov_errorhandling may optionally be called at any time to set the Fortran file unit number to which output error
messages are written. The default value is the one given in the rttov_const module (currently 0). On most platforms the
standard error is 0, but for HP it is 7. You should set the value according to your system. If no call is made, it is the
same as calling the routine with the default values.
Type In/Out Variable Description

Integer Intent(in) err_unit Logical error unit
95
Version : 1.3
Date : 05/03/2019
Annex C – Coefficient allocation and deallocation subroutines

1. RTTOV_READ_COEFS interface
call rttov_read_coefs (
err,
coefs,
opts,
channels,
channels_rec,
form_coef,
form_scaer,
form_sccld,
form_pccoef,
file_coef,
file_scaer,
file_sccld,
file_pccoef,
file_id_coef,
file_id_scaer,
file_id_sccld,
file_id_pccoef,
instrument,
path)
This subroutine is used to read the coefficient file(s). Only the arguments relevant to the required coefficient files are
necessary. The format arguments may be one of “formatted”, “unformatted” or “hdf5” (the latter only if the code was
compiled with HDF5 capability). The routine will attempt to determine the format of input coefficient files
automatically if this is not supplied. Note that the input files (optical depth, cloud, aerosol, MFASIS, PC) do not need
to be in the same format. This means, for example, you can read ASCII optical depth coefficient and cloud property
files and an HDF5 MFASIS LUT in the same call.
You can supply the path and filename of coefficient files or the logical unit of the file if it has already been opened.
These are the recommended methods for reading coefficient files. If neither of these arguments are supplied, the
instrument argument is mandatory so that the routine can use it to construct the coefficient filename(s). The satellite
and instrument IDs are listed in Tables 2 and 3. By default the coefficient file(s) must be in the current directory, but
you can specify an alternative directory via the path argument.
Important notes:
If the channels(:) argument is supplied to extract data for n channels from the coefficient file(s), then within RTTOV
these channels are referred to using the indices 1...n rather than the original channel numbers. In particular, these new
indices should be used when populating the chanprof(:)%chan array.
The settings in the options structure must be consistent with the coefficients being read. When reading visible/IR cloud
coefficient files opts%rt_ir%addclouds must be true and opts%rt_ir%user_cld_opt_param must be false. Similarly
when reading aerosol coefficient files opts%rt_ir%addclouds must be true and opts%rt_ir%user_cld_opt_param
must be false. When reading PC-RTTOV coefficient files opts%rt_ir%pc%addpc must be true and
opts%rt_ir%pc%addradrec must be true if you want to calculate reconstruct radiances..
When reading PC-RTTOV coefficient files, if you specify the channels(:) argument then this channel list must be a
super-set of the channel list which forms the predictor set for the selected PC regression configuration.
When calling RTTOV-SCATT you must read all channels from the coefficient file (i.e. do not use the channels(:)
argument). The RTTOV-SCATT Mietable file is read using the rttov_read_scattcoefs subroutine (see below).
96
Version : 1.3
Date : 05/03/2019

Integer Intent(out) err Return code
Type(rttov_coefs) Intent(out) coefs RTTOV coefficients structure.
Type(rttov_options) Intent(in) opts RTTOV options structure.
Integer Intent(in), channels(:) List of channels to extract.
optional
Integer Intent(in), channels_rec(:) List of channels for which to calculate radiances from
optional PC scores (only applicable if both
opts%rt_ir%pc%addpc and
opts%rt_ir%pc%addradrec are true).
Character Intent(in), form_coef Format of RTTOV coefficient file: should be either
optional “unformatted” (binary), “formatted” (ASCII) or
“hdf5” (only applicable if RTTOV has been compiled
with HDF5 functionality).
Character Intent(in), form_scaer Format of visible/IR aerosol scattering coefficient file.
optional
Character Intent(in), form_sccld Format of visible/IR cloud scattering coefficient file.
optional
Character Intent(in), form_pccoef Format of Principal Components coefficient file.
optional
Character Intent(in), file_coef Name of RTTOV coefficient file.
optional
Character Intent(in), file_scaer Name of visible/IR aerosol scattering coefficient file.
optional
Character Intent(in), file_sccld Name of visible/IR cloud scattering coefficient file.
optional
Character Intent(in), file_pccoef Name of Principal Components coefficient file.
optional
Integer Intent(in), file_id_coef Logical unit of pre-opened RTTOV coefficient file.
optional
Integer Intent(in), file_id_scaer Logical unit of visible/IR aerosol scattering
optional coefficient file.
Integer Intent(in), file_id_sccld Logical unit of visible/IR cloud scattering coefficient
optional file.
Integer Intent(in), file_id_pccoef Logical unit of Principal Components coefficient file.
optional
Integer Intent(in), instrument(3) platform ID, satellite ID, instrument ID (see Tables
optional 2/3). If no filename is supplied, the instrument
argument is used to construct the coefficient file
name(s).
Character Intent(in), path Used with instrument argument, directory containing
optional coefficient files (if omitted coefficient files are
assumed to be in the current directory).
97
Version : 1.3
Date : 05/03/2019
2. RTTOV_READ_COEFS_HTFRTC interface
call rttov_read_coefs_htfrtc (err, coefs, fname_coef, fname_sensor,
channels_rec)
This subroutine is used to read the input files for the HTFRTC model. HTFRTC simulations require two input files: the
static file (specified in fname_coef) is required for all simulations. In addition a sensor-specific file is required
(specified in fname_sensor). The channels for which reconstructed radiances are required are specified when the
HTFRTC files are read in: if channels_rec is omitted then reconstructed radiances are calculated for all sensor channels.
Otherwise you can specify a channel list here. Note that you cannot specify a channel list when calling
RTTOV/HTFRTC (unlike for PC-RTTOV).

Type(rttov_coefs) Intent(inout) coefs RTTOV coefficients structure.
Character Intent(in) fname_coef Path to HTFRTC static coefficient file.
Character Intent(in) fname_sensor Path to HTFRTC sensor-specific coefficient file.
Integer Intent(in), channels_rec(:) List of channels for which to calculate radiances from
optional PC scores. If omitted HTFRTC simulations will
calculate reconstructed radiances for all sensor
channels if opts%htfrtc%reconstructed is true.
3. RTTOV_DEALLOC_COEFS interface
call rttov_dealloc_coefs (err, coefs)
rttov_dealloc_coefs is called to de-allocate the memory for the coefs structure.

Type(rttov_coefs) Intent(inout) coefs RTTOV coefficients structure.
4. RTTOV_READ_SCATTCOEFFS interface
call rttov_read_scattcoeffs (err, opts_scatt, coef_rttov, coef_scatt,
file_id, file_coef, path)
This subroutine is called to read the RTTOV-SCATT Mietable coefficients file into the rttov_scatt_coef structure. The
filename (including full path) of the Mietable file may be specified in file_coef. Alternatively it is possible to open the
Mietable file before calling this subroutine and to pass the logical unit into rttov_read_scattcoeffs. If neither optional
argument is specified the filename of the Mietable file is constructed from the name of the platform and instrument in
the RTTOV coefficient file. In this case the file is assumed to be in the current directory unless you specify an
alternative directory in the path argument.
This subroutine can read both ASCII Mietable files (which have file extension “.dat”) and binary formatted Mietable
files created with rttov_ascii2bin_scattcoef.exe described in Annex A (which have file extension “.bin”).
98
Version : 1.3
Date : 05/03/2019

Type( rttov_options_scatt) Intent(in) opts_scatt RTTOV-SCATT options structure.
Type(rttov_coefs) Intent(in) coef_rttov RTTOV coefficients structure.
Type(rttov_scatt_coef) Intent(inout) coef_scatt RTTOV Mietable coefficients structure.
Integer Intent(in), file_id Logical unit of pre-opened RTTOV-SCATT
optional Mietable file.
Character Intent(in), file_coef Name of RTTOV-SCATT Mietable file.
optional
Character Intent(in), path If the file_id and file_coef arguments are not
optional supplied you can specify the directory containing the
Mietable file via this argument, otherwise RTTOV
assumes the file is in the current directory.
5. RTTOV_DEALLOC_SCATTCOEFFS interface
call rttov_dealloc_scattcoeffs (coef_scatt)
rttov_dealloc_scattcoeffs is called to de-allocate the memory for the RTTOV-SCATT Mietable structure.

Type(rttov_scatt_coef) Intent(inout) coef_scatt RTTOV Mietable coefficients structure.
99
Version : 1.3
Date : 05/03/2019
Annex D – RTTOV allocation/deallocation and initialisation subroutines

RTTOV provides four subroutines which may be used for allocating all input arrays and structures required by the
direct, TL, AD and K models. Alternatively it is possible to call individual subroutines to allocate each input and output
structure individually. Additionally there are subroutines which can be called to initialise the structures which can be
useful to ensure data from a previous RTTOV call does not interfere with a subsequent call.
1. RTTOV_ALLOC_DIRECT interface
call rttov_alloc_direct (err, asw, nprofiles, nchanprof, nlevels, chanprof,
opts, profiles, coefs, transmission, radiance,
radiance2, calcemis, emissivity, calcrefl, reflectance,
aer_maxnmom, aer_nphangle, aer_opt_param, cld_maxnmom,
cld_nphangle, cld_opt_param, traj, npcscores,
nchannels_rec, pccomp, channels_rec, init)
rttov_alloc_direct may be called to allocate or de-allocate any or all input arrays and structures for the RTTOV direct
model (with the exception of the cloudy profile structure for RTTOV-SCATT). It may be convenient to call a single
allocation subroutine rather than multiple subroutines. The order of the arguments closely mirrors the arguments for
rttov_direct.
It is important to note that array arguments passed to this subroutine (e.g. chanprof, profiles, calcemis) must be declared
as pointers. The arrays themselves are (de)allocated by the call to this subroutine as well as any members within the
derived types.
Note also that due to the large number of optional arguments it is common that named arguments must be supplied.
100
Version : 1.3
Date : 05/03/2019

Integer Intent(in) asw Switch (1=allocate; 0=deallocate)
Integer Intent(in) nprofiles Number of profiles per call to RTTOV
Integer Intent(in) nchanprof Number of channels simulated per call to RTTOV
(size of chanprof array)
Integer Intent(in) nlevels Number of profile levels
Type(rttov_chanprof) Pointer, chanprof(:) chanprof array pointer: (de)allocated after call
optional
Type(rttov_options) Intent(in) opts RTTOV options structure
Type(rttov_profile) Pointer, profiles(:) Profiles structure array pointer to be (de)allocated:
optional the profiles array itself will be (de)allocated as
well as the member arrays.
Type(rttov_coefs) Intent(in) coefs RTTOV coefficient structure
Type(rttov_transmission) Intent(inout), transmission Transmission structure to be (de)allocated
optional
Type(rttov_radiance) Intent(inout), radiance Radiance structure to be (de)allocated
optional
Type(rttov_radiance2) Intent(inout), radiance2 Secondary radiance structure to be (de)allocated
optional
Logical Pointer, calcemis(:) calcemis array pointer: (de)allocated after call
optional
Type(rttov_emissivity) Pointer, emissivity(:) Surface emissivity array pointer: (de)allocated
optional after call
Logical Pointer, calcrefl(:) calcrefl array pointer: (de)allocated after call
optional
Type(rttov_reflectance) Pointer, reflectance(:) Surface reflectance array pointer: (de)allocated
optional after call
Integer Intent(in), aer_maxnmom Maximum number of coefficients in Legendre
optional expansion of aerosol phase functions.
Integer Intent(in), aer_nphangle Number of phase angles over which aerosol phase
optional functions are to be defined.
Type(rttov_opt_param) Intent(inout), aer_opt_param Aerosol optical parameters structure to be
optional (de)allocated.
Integer Intent(in), cld_maxnmom Maximum number of coefficients in Legendre
optional expansion of cloud phase functions.
Integer Intent(in), cld_nphangle Number of phase angles over which cloud phase
optional functions are to be defined.
Type(rttov_opt_param) Intent(inout), cld_opt_param Cloud optical parameters structure to be
Type(rttov_traj) Intent(inout), traj Trajectory structure
optional
Integer Intent(in), npcscores Number of principal components to calculate for
optional each profile multiplied by the number of profiles.
Integer Intent(in), nchannels_rec The number of channels for which reconstructed
optional radiances are required multiplied by the number of
profiles. Only needed if opts%rt_ir%pc%addpc
and opts%rt_ir%pc%addradrec are true.
Type(rttov_pccomp) Intent(inout), pccomp pccomp structure to be (de)allocated
optional
Integer Pointer, channels_rec(:) Array pointer for list of channel numbers for
optional which to reconstruct radiances
Logical Intent(in), init Initialise the newly allocated structures
optional
101
Version : 1.3
Date : 05/03/2019
2. RTTOV_ALLOC_TL interface
call rttov_alloc_tl (err, asw, nprofiles, nchanprof, nlevels, chanprof, opts,
profiles, profiles_tl, coefs, transmission,
transmission_tl, radiance, radiance_tl, calcemis,
emissivity, emissivity_tl, calcrefl, reflectance,
reflectance_tl, aer_maxnmom, aer_nphangle, aer_opt_param,
aer_opt_param_tl, cld_maxnmom, cld_nphangle, cld_opt_param,
cld_opt_param_tl, traj, traj_tl, npcscores, nchannels_rec,
pccomp, pccomp_tl, channels_rec, init)
rttov_alloc_tl may be called to allocate or de-allocate any or all input arrays and structures for the RTTOV TL model
(with the exception of the cloudy profile structure for RTTOV-SCATT). It may be convenient to call a single allocation
subroutine rather than multiple subroutines. The order of the arguments closely mirrors the arguments for rttov_tl.
derived types.
Most of the arguments are common to rttov_alloc_direct: only the additional arguments are listed in the following
table.

Type(rttov_profile) Pointer, profiles_tl(:) Profiles TL structure array pointer to be
optional (de)allocated: the profiles_tl array itself will be
(de)allocated as well as the member arrays.
Type(rttov_transmission) Intent(inout), transmission_tl Transmission TL structure to be (de)allocated
optional
Type(rttov_radiance) Intent(inout), radiance_tl Radiance TL structure to be (de)allocated
optional
Type(rttov_emissivity) Pointer, emissivity_tl(:) Surface emissivity TL array pointer: (de)allocated
optional after call
Type(rttov_reflectance) Pointer, reflectance_tl(:) Surface reflectance TL array pointer: (de)allocated
optional after call
Type(rttov_opt_param) Intent(inout), aer_opt_param_tl Aerosol optical parameter TL structure to be
Type(rttov_opt_param) Intent(inout), cld_opt_param_tl Cloud optical parameter TL structure to be
Type(rttov_traj) Intent(inout), traj_tl Trajectory TL structure
optional
Type(rttov_pccomp) Intent(inout), pccomp_tl pccomp TL structure to be (de)allocated
optional
102
Version : 1.3
Date : 05/03/2019
3. RTTOV_ALLOC_AD interface
call rttov_alloc_ad (err, asw, nprofiles, nchanprof, nlevels, chanprof, opts,
profiles, profiles_ad, coefs, transmission,
transmission_ad, radiance, radiance_ad, calcemis,
emissivity, emissivity_ad, calcrefl, reflectance,
reflectance_ad, aer_maxnmom, aer_nphangle, aer_opt_param,
aer_opt_param_ad, cld_maxnmom, cld_nphangle, cld_opt_param,
cld_opt_param_ad, traj, traj_ad, npcscores, nchannels_rec,
pccomp, pccomp_ad, channels_rec, init)
rttov_alloc_ad may be called to allocate or de-allocate any or all input arrays and structures for the RTTOV AD model
subroutine rather than multiple subroutines. The order of the arguments closely mirrors the arguments for rttov_ad.
derived types.
table.

Type(rttov_profile) Pointer, profiles_ad(:) Profiles AD structure array pointer to be
optional (de)allocated: the profiles_ad array itself will be
Type(rttov_transmission) Intent(inout), transmission_ad Transmission AD structure to be (de)allocated
optional
Type(rttov_radiance) Intent(inout), radiance_ad Radiance AD structure to be (de)allocated
optional
Type(rttov_emissivity) Pointer, emissivity_ad(:) Surface emissivity AD array pointer: (de)allocated
optional after call
Type(rttov_reflectance) Pointer, reflectance_ad(:) Surface reflectance AD array pointer:
optional (de)allocated after call
Type(rttov_opt_param) Intent(inout), aer_opt_param_ad Aerosol optical parameter AD structure to be
Type(rttov_opt_param) Intent(inout), cld_opt_param_ad Cloud optical parameter AD structure to be
Type(rttov_traj) Intent(inout), traj_ad Trajectory AD structure
optional
Type(rttov_pccomp) Intent(inout), pccomp_ad pccomp AD structure to be (de)allocated
optional
103
Version : 1.3
Date : 05/03/2019
4. RTTOV_ALLOC_K interface
call rttov_alloc_k (err, asw, nprofiles, nchanprof, nlevels, chanprof, opts,
profiles, profiles_k, coefs, transmission,
transmission_k, radiance, radiance_k, calcemis,
emissivity, emissivity_k, calcrefl, reflectance,
reflectance_k, aer_maxnmom, aer_nphangle, aer_opt_param,
aer_opt_param_k, cld_maxnmom, cld_nphangle, cld_opt_param,
cld_opt_param_k, traj, traj_k, npcscores, nchannels_rec,
pccomp, pccomp_k, profiles_k_pc, profiles_k_rec,
channels_rec, init)
rttov_alloc_k may be called to allocate or de-allocate any or all input arrays and structures for the RTTOV K model
subroutine rather than multiple subroutines. The order of the arguments closely mirrors the arguments for rttov_k.
derived types.
table.

Type(rttov_profile) Pointer, profiles_k(:) Profiles K structure array pointer to be
optional (de)allocated: the profiles_k array itself will be
Type(rttov_transmission) Intent(inout), transmission_k Transmission K structure to be (de)allocated
optional
Type(rttov_radiance) Intent(inout), radiance_k Radiance K structure to be (de)allocated
optional
Type(rttov_emissivity) Pointer, emissivity_k(:) Surface emissivity K array pointer: (de)allocated
optional after call
Type(rttov_reflectance) Pointer, reflectance_k(:) Surface reflectance K array pointer: (de)allocated
optional after call
Type(rttov_opt_param) Intent(inout), aer_opt_param_k Aerosol optical parameter K structure to be
Type(rttov_opt_param) Intent(inout), cld_opt_param_k Cloud optical parameter K structure to be
Type(rttov_traj) Intent(inout), traj_k Trajectory K structure
optional
Type(rttov_pccomp) Intent(inout), pccomp_k pccomp K structure to be (de)allocated
optional
Type(rttov_profile) Pointer, profiles_k_pc(:) PC score Jacobian structure array pointer to be
optional (de)allocated: the profiles_k_pc array itself will be
Type(rttov_profile) Pointer, profiles_k_rec(:) PC reconstructued radiance Jacobian structure
optional array pointer to be (de)allocated: the
profiles_rec_k array itself will be (de)allocated as
well as the member arrays.
104
Version : 1.3
Date : 05/03/2019
5. RTTOV_ALLOC_PROF interface
call rttov_alloc_prof (err, nprofiles, profiles, nlevels,
opts, asw, coefs, init)
rttov_alloc_prof is called to allocate or de-allocate the memory for the profiles structure. When allocating
profiles_k(:) for the K model remember that the size of the array is nchanprof. The profiles(:) array must have been
allocated before calling this subroutine.

Integer Intent(in) nprofiles Number of profiles per call to RTTOV
Type(rttov_profile) Intent(inout) profiles(nprofiles) Array of profiles structures to be (de)allocated
Type(rttov_coefs) Intent(in), coefs RTTOV coefficient structure. This is mandatory if
optional either opts%rt_ir%addclouds or
opts%rt_ir%addaerosl are true, otherwise it may be
omitted.
Logical Intent(in), init Additionally initialise profiles structure
optional
6. RTTOV_INIT_PROF interface
call rttov_init_prof (profiles, p)
rttov_init_prof is used to initialise a previously allocated profiles structure. This is particularly useful when calling the
adjoint or K models as the profiles_ad/profiles_k structures must be initialised before each call.

Type(rttov_profile) Intent(inout) profiles(nprof) Profiles structure to be initialised
Real Intent(in), p(nlevels) Optional pressure profile with which to initialise
optional profiles(:)%p(:) (this argument is not commonly used)
7. RTTOV_ALLOC_RAD interface
call rttov_alloc_rad (err, nchanprof, radiance, nlevels, asw, radiance2, init)
rttov_alloc_rad is called to allocate or de-allocate the memory for the radiance structure and optionally for the
radiance2 structure.

Integer Intent(in) nchanprof Total number of channels being simulated
Type(rttov_radiance) Intent(inout) radiance Radiance structure to be (de)allocated
Type(rttov_radiance2) Intent(inout), radiance2 Secondary radiance structure to be created
optional
Logical Intent(in), init Additionally initialise radiance structure(s)
optional
105
Version : 1.3
Date : 05/03/2019
8. RTTOV_INIT_RAD interface
call rttov_init_rad (radiance, radiance2)
rttov_init_rad is used to initialise a previously allocated radiance structure, and optionally also a radiance2 structure.

Type(rttov_radiance) Intent(inout) radiance Radiance structure to be initialised
Type(rttov_radiance2) Intent(inout), radiance2 Secondary radiance structure to be initialised
optional
9. RTTOV_ALLOC_TRANSMISSION interface
call rttov_alloc_transmission (err, transmission, nlevels, nchanprof, asw, init)
rttov_alloc_transmission is called to allocate or de-allocate the memory for the transmission structure.

Type(rttov_transmission) Intent(inout) transmission Transmission structure to be (de)allocated
Logical Intent(in), init Additionally initialise transmission structure
optional
10. RTTOV_INIT_TRANSMISSION interface

call rttov_init_transmission (transmission)
rttov_init_transmission is used to initialise a previously allocated transmission structure.

Type(rttov_transmission) Intent(inout) transmission Transmission structure to be initialised
106
Version : 1.3
Date : 05/03/2019
11. RTTOV_ALLOC_PCCOMP interface

call rttov_alloc_pccomp (err, pccomp, npcscores, asw, init, nchannels_rec, &
opts, nlevels)
rttov_alloc_pccomp is called to allocate or de-allocate the memory for the pccomp structure. This is only required for
PC-RTTOV or HTFRTC simulations.

Type(rttov_pccomp) Intent(inout) pccomp pccomp structure to be (de)allocated
Integer Intent(in) npcscores Number of principal components to calculate for each
profile multiplied by the number of profiles.
Logical Intent(in), init Additionally initialise pccomp structure
optional
Integer Intent(in), nchannels_rec The number of channels for which reconstructed
optional radiances are required multiplied by the number of
profiles. Only needed if opts%rt_ir%pc%addradrec
(PC-RTTOV) or opts%htfrtc_opts%reconstruct
(HTFRTC) is true.
Type(rttov_options) Intent(in), opts RTTOV options structure, required for HTFRTC
optional simulations in which case opts%htfrtc_opts%htfrtc
must be true, and the overcast and/or simple_cloud
options must be true if they will be requested in the
simulations.
Integer Intent(in), nlevels Number of levels in profile structure, required for
optional HTFRTC only if opts%htfrtc_opts%overcast is true.
12. RTTOV_INIT_PCCOMP interface

call rttov_init_pccomp (pccomp)
rttov_init_pccomp is used to initialise a previously allocated pccomp structure.

Type(rttov_pccomp) Intent(inout) pccomp pccomp structure to be initialised
107
Version : 1.3
Date : 05/03/2019
13. RTTOV_ALLOC_OPT_PARAM interface

call rttov_alloc_opt_param (err, opts, opt_param, nchanprof,
nlevels, nmom, nphangle, asw, init)
rttov_alloc_opt_param is called to allocate or de-allocate the memory for the rttov_opt_param structure (for aerosol
or cloud optical parameters).

Type(rttov_opt_param) Intent(inout) opt_param Optical parameters structure to be (de)allocated
Integer Intent(in) nmom Maximum number of coefficients in Legendre
expansions of phase functions. Can be zero if DOM
scattering solver is not used.
Integer Intent(in) nphangle Number of phase angles over which phase functions
are to be defined. Phase functions are required for
solar scattering simulations. They may also be
required for non-solar simulations to calculate the
Legendre coefficients or the “bpr” parameters.
Logical Intent(in), init If true, zero the newly allocated structure.
optional
14. RTTOV_INIT_OPT_PARAM interface

call rttov_init_opt_param (err, opts, opt_param)
rttov_init_opt_param is called to initialise phase angle variables in the rttov_opt_param structure (for aerosol or
cloud optical parameters). NB This is only required if opts%rt_ir%addsolar is true.
It can also be used to initialise an optical property structure to zero using the zero_only argument: this is useful when
calling the AD/K models.

Type(rttov_options) Intent(in), opts RTTOV options structure (mandatory if the
optional zero_only argument is omitted or is false).
Type(rttov_opt_param) Intent(inout) opt_param Optical parameters structure to be initialised.
Logical Intent(in) zero_only If true just initialise the structure to zero: use this to
zero AD/K optical property structures. Set this to
false (or omit it) for the direct model structure.
108
Version : 1.3
Date : 05/03/2019
15. RTTOV_ALLOC_TRAJ interface

call rttov_alloc_traj (err, nprofiles, nchanprof, opts, nlevels, coefs,
asw, traj, traj_tl, traj_ad, traj_k)
When calling RTTOV a number of temporary structures are required. These are stored in an encompassing “trajectory”
structure. To avoid allocating and deallocating these data structures inside every call to RTTOV this subroutine may be
used to create trajectory structures so they can then be passed as arguments to every call to the RTTOV core routines
(direct, TL, AD or K). Whether this is useful depends on the architecture/compiler: it has been seen to improve
performance in some cases on Linux platforms so you may wish to consider this. In no case should it be detrimental to
performance. Note that the trajectory structures cannot be used with the RTTOV parallel interfaces.
NB When passing the directory structure to RTTOV, the coefs variable MUST be declated with the TARGET
attribute.
Once all radiances have been computed, you should call rttov_alloc_traj again with asw set to zero to deallocate the
trajectory structure(s).

Integer Intent(out) err Return code.
Integer Intent(in) nprofiles Number of profiles per call to RTTOV.
Type(rttov_coefs) Intent(in) coefs RTTOV coefficient structure.
Type(rttov_traj) Intent(inout), traj, traj_tl, traj_ad, traj_k Trajectory structures: a structure for each
optional core RTTOV routine can be (de)allocated
with a single call to rttov_alloc_traj.
109
Version : 1.3
Date : 05/03/2019
16. RTTOV_ALLOC_SCATT_PROF interface

call rttov_alloc_scatt_prof (err, nprof, cld_profiles, nlev, use_totalice, asw,
init, mmr_snowrain)
rttov_alloc_scatt_prof is called to allocate or de-allocate the memory for the cld_profiles structure which contains
cloud profile information for RTTOV-SCATT (the MW scattering model).

Integer Intent(in) nprof Number of profiles per call to RTTOV.
Type(rttov_profile_cloud) Intent(inout) cld_profiles(nprof) Array of RTTOV-SCATT cloud profile
structures to be (de)allocated.
Type(rttov_options) Intent(in) nlev Number of levels in input profiles.
Logical Intent(in) use_totalice Choose separate ciw and snow, or totalice
Logical Intent(in) init Additionally initialise cld_profiles structure
Logical Intent(in), mmr_snowrain Units of snow and rain input units:
optional False => kg/m2/s; True => kg/kg (default)
17. RTTOV_INIT_SCATT_PROF interface

call rttov_init_scatt_prof (cld_profiles)
rttov_init_scatt_prof is used to initialise the cld_profiles structure. This is particularly useful when calling the adjoint
or K models as the cld_profiles_ad/cld_profiles_k structures must be initialised before each call.

Type(rttov_profile_cloud) Intent(inout) cld_profiles(nprof) RTTOV-SCATT cloud profile structure to
be initialised.
18. RTTOV_ALLOC_EMIS_RET_TERMS interface

call rttov_alloc_emis_ret_terms (err, nchanprof, emis_retrieval_terms, asw)
rttov_alloc_emis_ret_terms is called to allocate or de-allocate the memory for the emissivity retrieval terms structure
which contains radiance and transmittance outputs from RTTOV-SCATT which can be used for all-sky emissivity
retrievals. This is an optional argument to RTTOV-SCATT.

Integer Intent(in) nchanprof Total number of channels being
simulated
Type(rttov_scatt_emis_retrieval_type) Intent(inout) emis_retrieval_terms Emissivity retrieval terms structure
to be (de)allocated.
110
Version : 1.3
Date : 05/03/2019
19. RTTOV_SCATT_SETUPINDEX interface

call rttov_scatt_setupindex (nprofiles, n_chan, coef, nchannels, chanprof,
frequencies, lchannel_subset)
Before calling RTTOV-SCATT you should make a call to rttov_scatt_setupindex which initialises the chanprof and
the frequencies input arrays to RTTOV-SCATT. The frequencies array contains the indices into the Mietable
coefficients for each channel to be simulated.
The n_chan argument must be the number of channels in the RTTOV coefficient file. The chanprof(:) array should
have size equal to the total number of channels to simulate over all profiles. If you want to simulate every channel for
every profile then the size of chanprof(:) is nprofiles*n_chan and the lchannel_subset(:,:) argument is not required.
However, if you do not require certain channels to be simulated (either for every profile or for some profiles) then you
can indicate the channels to be simulated for each profile by setting the corresponding elements of lchannel_subset(:,:)
to .TRUE. In this case the chanprof(:) array must have size equal to the number of .TRUE. elements in
lchannel_subset(:,:).

Integer Intent(in) nprofiles Number of profiles.
Integer Intent(in) n_chan Number of channels in coefficient
structure.
Type(rttov_coefs) Intent(in) coef_rttov RTTOV coefficients structure.
Integer Intent(in) nchannels Total number of channels being simulated.
Type(rttov_chanprof) Intent(out) chanprof(nchannels) RTTOV chanprof structure.
Integer Intent(out) frequencies(nchannels) Frequency number for each channel being
simulated.
Logical Intent(in), lchannel_subset(nprofiles, Logical array to specify which channels to
optional n_chan) simulate for which profiles.
111
Version : 1.3
Date : 05/03/2019
Annex E – Optical parameter calculation subroutines

1. RTTOV_BPR_INIT interface
call rttov_bpr_init (err, phangle)
rttov_bpr_init is called before calculating b parameters from phase functions using rttov_bpr_calc: this subroutine
prepares some tables to speed up the calculations.

Real Intent(in) phangle(:) Array of angles over which phase functions are
defined.
2. RTTOV_BPR_CALC interface
call rttov_bpr_calc (err, pha, phangle, bpr, nthreads)
rttov_bpr_calc is used to calculate a single b parameter given a phase function. Note that this is a relatively slow
calculation and as such is intended to be called “off-line” rather than within performance-critical code. If you compile
RTTOV with OpenMP this subroutine can exploit multiple cores/CPUs: choose the number of threads with the
nthreads argument.

Real Intent(in) pha (:) Phase function.
Real Intent(in) phangle(:) Array of angles over which phase function pha is
defined.
Real Intent(out) bpr Calculated b parameter.
Integer Intent(in), nthreads Use multiple threads, requires RTTOV to have been
optional compiled with OpenMP, default = 1.
3. RTTOV_BPR_DEALLOC interface
call rttov_bpr_dealloc (err)
rttov_bpr_dealloc is called after b parameters have been calculated to deallocate memory.

112
Version : 1.3
Date : 05/03/2019
4. RTTOV_LEGCOEF_CALC interface
call rttov_legcoef_calc (err, pha, phangle, nmom, legcoef, ngauss, q, w)
rttov_legcoef_calc is used to calculate the Legendre expansion of a given phase function. The nmom parameter need
only equal the largest value of opts%rt_ir%dom_nstreams with which RTTOV will be called. The coefficients are
calculated using a 1000 point Gaussian quadrature by default: you can specify an alternative quadrature size or, for
greater efficiency (if making many calls to this subroutine), you can pre-compute a quadrature and the associated
weights and pass these in.

Real Intent(in) pha (:) Phase function.
Real Intent(in) phangle(:) Array of angles over which phase function pha is
defined.
Integer Intent(in) nmom Number of Legendre coefficients to calculate
(excluding first coefficient which is always unity).
Real Intent(inout) legcoef(:) Calculated Legendre coefficients, array must be at
least nmom+1 in size.
Integer Intent(in), ngauss Choose size of Gaussian quadrature to use in
optional calculation (the default is 1000 points).
Real Intent(in), q(:) You can supply pre-computed quadrature points
optional which may be faster if calling this subroutine many
times.
Real Intent(in), w(SIZE(q)) Weights corresponding to quadrature q.
optional
113
Version : 1.3
Date : 05/03/2019
Annex F – Emissivity atlas subroutines

This section describes the interface to the atlas subroutines. More information about using the atlases is provided in
section 7.5. As of RTTOV v12 it is possible to use the TELSEM2 atlas without compiling RTTOV with any external
library dependencies because the TELSEM2 atlas data are contained in ASCII files. The other atlas data are provided in
HDF5 files and as such RTTOV must be compiled with HDF5 support (see section 5.2).
1. RTTOV_SETUP_EMIS_ATLAS interface
call rttov_setup_emis_atlas (err, opts, imonth, atlas_type, atlas,
atlas_id, path, coefs, ir_atlas_read_std,
ir_atlas_ang_corr, year)
The atlases provide monthly climatologies so you must provide the month (imonth) for which to load atlas data. By
default the atlas files are assumed to be in the current directory, but you can optionally specify the path to the directory
containing the atlas data files. The type of atlas (MW or IR) is selected by the atlas_type argument (1=>MW, 2=>IR).
The atlas argument is used to hold the loaded atlas data. In contrast to previous RTTOV versions you can initialise data
for multiple atlases, months and instruments simultaneously: the only limitation is the available memory. The atlas
argument is of derived type rttov_emis_atlas_data which is defined in the module mod_emis_atlas_data. You can
declare as many instances of this type as you require and load the data from any atlas for any month and instrument into
each one by calling this subroutine for each atlas variable.
The atlas_id argument is used to select between the available atlases: if omitted the default IR or MW atlas is selected.
The valid IDs are as follows:
IR instruments MW instruments
 UWIRemis – atlas_id = 1 (default)  TELSEM2 – atlas_id = 1 (default)
 CAMEL 2007 atlas – atlas_id = 2  CNRM MW atlas – atlas_id = 2
 CAMEL climatology atlas – atlas_id = 3
Once initialised the TELSEM2 atlas data can be used with any MW instrument: in this case the optional coefs argument
is ignored. The CNRM MW atlas is always initialised for a specific instrument and the initialised atlas data can only be
used with that particular instrument (AMSU-A, AMSU-B, MHS, SSMI/S or ATMS). In this case the coefs argument is
mandatory and must contain the coefficients for a compatible sensor.
For the IR atlases the coefs argument is optional. If omitted the loaded atlas data can be used with any IR sensor. If the
coefs argument is supplied the atlas data are loaded for use with that specific instrument. This makes it much faster to
obtain emissivities from the atlas, but the loaded atlas data can only be used with that specific instrument. If you are
only using the atlas for one IR instrument it is recommended to supply the coefs argument.
In addition the IR atlases can optionally include a zenith angle correction which is activated when reading the data by
setting ir_atlas_ang_corr to true. The IR atlases can also optionally return emissivity data: this must be requested at
setup by setting ir_atlas_read_std to true, but note that this increases the atlas memory requirements significantly.
Note that if you do not require the standard deviations you do not need to download the standard deviation atlas data
files.
The IR atlases can be used for HTFRTC in the same way as for standard RTTOV calls. The differences are that the
emissivity(:), emis_std(:) and emis_flag(:) arrays have size equal to the number of reconstructed radiance channels
multiplied by the number of profiles, and the chanprof argument to rttov_get_emis is not used (see section 8.9).
114
Version : 1.3
Date : 05/03/2019

Integer Intent(in) imonth Month (1-12) of atlas data to be loaded.
Integer Intent(in) atlas_type Specifies whether to load a MW or IR
atlas: 1 => MW, 2 => IR.
Type(rttov_emis_atlas_data) Intent(inout) atlas Structure to hold the loaded atlas data.
(from mod_rttov_emis_atlas)
Integer Intent(in), atlas_id ID of atlas to use (see above); defaults to
optional 1 for both IR and MW.
Character Intent(in), path Path of directory containing emissivity
optional atlas data files. Defaults to the current
directory.
Type(rttov_coefs) Intent(in), coefs RTTOV instrument coefficients structure.
optional This is mandatory for the CNRM MW
atlas and is ignored by TELSEM2. For the
UW and CAMEL IR atlases, if this
argument is passed they will be initialised
specifically for use with this instrument:
the atlases are then faster to access.
Logical Intent(in), ir_atlas_read_std If true, initialise IR atlas error dataset. If
optional the errors (standard deviations) are not
required, this should be set to false: the
atlas look-ups will be faster and the atlas
will require much less memory. Default is
false. Not applicable to MW atlases.
Logical Intent(in), ir_atlas_ang _corr If true, initialise IR atlas so that zenith
optional angle correction will be applied. This
requires the additional angular correction
data files. Default is false. Not applicable
to MW atlases.
Integer Intent(in), year This only applies to the CNRM MW atlas:
optional data are available for multiple years as
indicated in the atlas filenames. This
argument selects the year (default=2015).
2. RTTOV_GET_EMIS interface
The rttov_get_emis subroutine takes the profiles(:) array as input: Table 15 lists the profile variables used by each
atlas. Although there is only one routine to access the IR and MW atlases, each atlas has a number of options which are
unique to it. Therefore the rttov_get_emis subroutine will be described separately for the IR and MW atlases,
discussing just those arguments relevant in each case. If an argument is supplied which is not applicable to the given
atlas, a warning message is printed and the argument is ignored. The output emissivity(:) array can be used as input to
rttov_direct, rttov_tl, rttov_ad and rttov_k via emissivity(:)%emis_in with the corresponding elements of
calcemis(:) set to false. However it is very important to check the emissivities that are returned to ensure they are
positive values: the atlases may return negative values if they do not have data for the given surface type or lat/lon
location. You may wish to set calcemis(:) to true where the returned emissivities are negative, for example.
In the discussions below, nchanprof is the size of the chanprof(:) array, nprof is the size of the profiles(:) array, and
nchan is the largest number of channels computed for any given profile.
115
Version : 1.3
Date : 05/03/2019
Returning IR emissivities (UWIRemis, CAMEL 2007 and CAMEL climatology atlases):

call rttov_get_emis (err, opts, chanprof, profiles, coefs, atlas,
emissivity, emis_std, emis_flag, snow_correction)
If the zenith angle correction was selected when calling rttov_setup_emis_atlas then the zenangle and sunzenangle
profile members are required: sunzenangle is used to determine whether the day or night bias correction should be
applied so it only needs to be less than 85º for day or greater than 85º for night (the exact value is not important). It
returns emissivity values for both land and sea-ice surface types, and, over land, returns a linear combination of the land
surface emissivity and a snow emissivity weighted according to the snow fraction. If the snow fraction is zero, just the
land surface emissivities are returned. See section 8.4 for more information about the treatment of snow in the CAMEL
climatology atlas. The IR atlases can optionally return the emissivity standard deviations: the CAMEL 2007 atlas can be
used with the UWIRemis standard deviation dataset, while the CAMEL climatology atlas has a new standard deviation
dataset derived from the multi-year climatology at higher spatial resolution. In addition, each surface point in the atlas
has an associated flag which may be returned. You may wish to use this flag as a form of quality control (see the IR
atlas documentation Borbas et al., 2010 and the RTTOV v12 Science and Validation Report). Note that there is only
one flag per profile (ie per location). However the emis_flag(:) output array is of size nchanprof to be consistent with
the emissivity(:) array. If the coefs argument was passed to rttov_setup_emis_atlas you must ensure the coefs
argument here is compatible with the atlas argument.

Type(rttov_chanprof) Intent(in) chanprof(:) Chanprof structure.
Type(rttov_profile) Intent(in) profiles(:) Profiles structure.
Type(rttov_coefs) Intent(in) coefs RTTOV instrument coefficient structure.
Type(rttov_emis_atlas_data) Intent(in) atlas Atlas data previously loaded via
(from mod_rttov_emis_atlas) rttov_emis_atlas_setup.
Real Intent(out) emissivity(nchanprof) Emissivity values.
Real Intent(out), emis_std(nchanprof) Emissivity errors (standard deviations).
optional
Integer Intent(out), emis_flag(nchanprof) Emissivity atlas flags.
optional
Logical Intent(in), snow_correction CAMEL climatology atlas only: if true
optional then atlas attempts to find snow-free
spectra when profile snow_fraction is
zero, default true (recommended).
Returning MW emissivities (TELSEM2):

call rttov_get_emis (err, opts, chanprof, profiles, coefs, atlas,
emissivity, emis_std, emis_cov, resolution)
TELSEM2 consists of both an atlas and an interpolator which carries out interpolation of emissivity values in frequency
and in space. The atlas has a nominal spatial resolution of 0.25°. If the resolution argument is supplied and is larger
than 0.25, the emissivity values returned are integrated over the atlas grid according to the specified resolution. As with
the IR atlases, the TELSEM MW atlas can optionally return estimated emissivity errors. It can also optionally return
emissivity covariance matrices: this provides emissivity covariances among all channels for each profile.
116
Version : 1.3
Date : 05/03/2019

Real Intent(out), emis_std(nchanprof) Emissivity errors (standard deviations).
optional
Real Intent(out), emis_cov(nprof, nchan, Emissivity covariances.
optional nchan)
Real Intent(in), resolution Return emissivities at user-defined
optional resolution. Units are degrees
latitude/longitude. The default (i.e.
nominal atlas) resolution is 0.25°.
Returning MW emissivities (CNRM MW atlas):

call rttov_get_emis (err, opts, chanprof, profiles, coefs, atlas, emissivity)
The CNRM atlas has emissivity datasets only for specific instruments and so does not need to carry out interpolation in
frequency to instrument channels. It does not provide an estimate of emissivity error. The coefs argument must be
compatible with the atlas data.

3. RTTOV_DEALLOCATE_EMIS_ATLAS interface
call rttov_deallocate_emis_atlas (atlas)
rttov_deallocate_emis_atlas is called to de-allocate the memory for an emissivity atlas data structure. This should be
called for every loaded atlas structure.

Type(rttov_emis_atlas_data) Intent(inout) atlas Atlas data structure to deallocate.
(from mod_rttov_emis_atlas)
117
Version : 1.3
Date : 05/03/2019
Annex G – BRDF atlas subroutines

This section describes the interface to the atlas subroutines. More information about using the atlas is provided in
section 7.6. The atlas data are provided in HDF5 files and as such RTTOV must be compiled with HDF5 support (see
section 5.2).
1. RTTOV_SETUP_BRDF_ATLAS interface
call rttov_setup_brdf_atlas (err, opts, imonth, atlas, atlas_id, path, coefs)
The atlas provides monthly climatology so you must provide the month (imonth) for which to load atlas data. By
default the atlas files are assumed to be in the current directory, but you can optionally specify the path to the directory
containing the atlas data files.
The atlas argument is used to hold the loaded atlas data. In contrast to previous RTTOV versions you can initialise data
for multiple months and instruments simultaneously: the only limitation is the available memory. The atlas argument is
of derived type rttov_brdf_atlas_data which is defined in the module mod_brdf_atlas_data.
The atlas_id argument is used to select between the available atlases: currently there is only one BRDF atlas
(atlas_id=1) so this argument can be omitted.
The coefs argument is optional. If omitted the loaded atlas data can be used with any visible/IR sensor. If the coefs
argument is supplied the atlas data are loaded for use with that specific instrument. This makes it much faster to obtain
BRDFs from the atlas, but the loaded atlas data can only be used with that specific instrument. If you are only using the
atlas for one visible/IR instrument it is recommended to supply the coefs argument.

Integer Intent(in) imonth Month (1-12) of atlas data to be
loaded.
Type(rttov_brdf_atlas_data) Intent(inout) atlas Structure to hold the loaded atlas data.
(from mod_rttov_brdf_atlas)
Integer Intent(in), atlas_id ID of atlas to use. Currently there is
optional only one atlas, so this argument can be
omitted.
Character Intent(in), path Path of directory containing BRDF
optional atlas data files. Defaults to the current
directory.
Type(rttov_coefs) Intent(in), coefs RTTOV instrument coefficients
optional structure. If this argument is passed the
atlas will be initialised specifically for
use with this instrument: the atlas is
then faster to access.
2. RTTOV_GET_BRDF interface
The rttov_get_brdf subroutine takes the profiles(:) array as input: the routine uses the latitude, longitude, zenangle,
azangle, sunzenangle, sunazangle, skin%surftype and skin%watertype members of each profile. The output brdf(:)
array can be used as input to rttov_direct, rttov_tl, rttov_ad and rttov_k via reflectance(:)%refl_in with the
corresponding elements of calcrefl(:) set to false. However it is very important to check the BRDFs that are returned to
ensure they are positive values: the atlas may return negative values if it does not have data for the given surface type or
lat/lon location. You may wish to set calcrefl(:) to true where the returned BRDFs are negative, for example. In the
discussions below, nchanprof is the size of the chanprof(:) array.
118
Version : 1.3
Date : 05/03/2019
call rttov_get_brdf (err, opts, chanprof, profiles, coefs, atlas,

brdf, brdf_flag, bh_albedo)
The BRDF atlas returns BRDF values for over land and also over sea, but it does not take sun glint into account (set
calcrefl to true to use RTTOV’s sea surface reflectance model instead). The BRDF atlas can optionally return the bi-
hemispherical (black-sky) albedo. In addition, each surface point in the atlas has an associated flag which may be
returned. You may wish to use this flag as a form of quality control (see the BRDF atlas documentation in the RTTOV
v11 Science and Validation Report). Note that there is only one flag per profile (ie per location). However, the
brdf_flag(:) output array is of size nchanprof to be consistent with the brdf(:) array. If the coefs argument was passed
to rttov_setup_brdf_atlas you must ensure the coefs argument here is compatible with the atlas argument.

Type(rttov_brdf_atlas_data) Intent(in) atlas Atlas data previously loaded via
(from mod_rttov_brdf_atlas) rttov_brdf_atlas_setup.
Real Intent(out) brdf (nchanprof) BRDF values.
Integer Intent(out), brdf_flag(nchanprof) BRDF atlas flags.
optional
Real Intent(out), bh_albedo(nchanprof) Bi-hemispherical (black-sky) albedo.
optional
3. RTTOV_DEALLOCATE_BRDF_ATLAS interface
call rttov_deallocate_brdf_atlas (atlas)
rttov_deallocate_brdf_atlas is called to de-allocate the memory for a BRDF atlas data structure. This should be called
for every loaded atlas structure.

Type(rttov_brdf_atlas_data) Intent(inout) atlas Atlas data structure to deallocate.
(from mod_rttov_brdf_atlas)
119
Version : 1.3
Date : 05/03/2019
Annex H – RTTOV_GET_PC_PREDICTINDEX interface

call rttov_get_pc_predictindex(err, opts, predictindex, form_pccoef,
file_pccoef, file_id_pccoef, instrument)
This subroutine can be found in the src/coef_io/ directory. It may be used to obtain the indices for the specified set of
Principal Components regression channels, which depends on the setting of opts%rt_ir%pc%ipcbnd and
opts%rt_ir%pc%ipcreg. Note that the regression channel set is available through the coef_pccomp structure after the
PC coefficient file has been read (by calling rttov_read_coefs – see Annex C). This is demonstrated in
example_pc_fwd.F90. The rttov_get_pc_predictindex routine may be used outside of RTTOV.
Before calling the subroutine, opts%rt_ir%pc%ipcbnd must be set to a value between 1 and 3 (IASI clear-sky) or to 1
(IASI cloudy, AIRS) and opts%rt_ir%pc%ipcreg must be set to a value between 1 and 3 for AIRS or 1 and 4 for
IASI, corresponding to the predictor channel set required (see Section 8.8). Either the filename, the logical unit of a pre-
opened coefficient file, or the instrument ID triplet should be supplied.

Integer, Pointer Intent(out) predictindex(:) The output channel list. This array is
allocated with the appropriate dimension
within the routine.
Character Intent(in), form_pccoef Format of PC coefficient file: should be
optional either “unformatted” (binary) or
“formatted” (ASCII) or “hdf5”.
Character Intent(in), file_pccoef Filename of PC coefficient file.
optional
Integer Intent(in), file_id_pccoef Logical unit of pre-opened PC coefficient
optional file.
Integer Intent(in), instrument(3) platform id; satellite id, instrument id (see
optional Tables 2/3). If no filename is supplied, the
instrument argument is used to construct
the coefficient file name.
An example of how this routine might be used is included in comments in example_pc_fwd.F90. The executable
rttov_test_get_pc_predictindex can be used to test the above subroutine:
$ rttov_test_get_pc_predictindex.exe --pccoef-in ... --ipcbnd ... --ipcreg ...
where the arguments are the PC coefficient file name and the indices of the PC band and regression set respectively.
120
Version : 1.3
Date : 05/03/2019
Annex I – RTTOV_DIRECT interface

call rttov_direct (errorstatus, chanprof, opts, profiles, coefs, transmission,
radiance, radiance2, calcemis, emissivity, calcrefl,
reflectance, aer_opt_param, cld_opt_param, traj, pccomp,
channels_rec)
rttov_direct should be called for each instrument required. Each call carries out computations for the profiles in the
array profiles(:). The transmission structure contains the calculated atmospheric transmittances and the radiance (and
optionally radiance2) contain the calculated radiances, brightness temperatures and reflectances (see Annex O for
information on the output structures).
The chanprof(:) array must be set up with channel and profile indexes as illustrated in Table 13. For Principal
Components calculations (see section 8.8) the chanprof(:) array must be set up for the PC predictor channels. Therefore
chanprof(:)%chan should be populated with the channel list defined by opts%rt_ir%pc%ipcbnd and
opts%rt_ir%pc%ipcreg for each profile being simulated in the call to rttov_direct.
For most simulations the calcemis(:) and emissivity(:) arguments must be supplied. These are used to pass surface
emissivities into and out from RTTOV and to specify whether RTTOV should calculate emissivities internally. See
sections 7.5 and 8.4 for more information about surface emissivity.
For solar simulations the calcrefl(:) and reflectance(:) arguments must be supplied. These relate to the surface BRDF
and are analogous to the corresponding emissivity arrays. See sections 7.6 and 8.2 for more information about surface
reflectance.
Note that if no solar calculations are required the calcrefl and reflectance arguments may be omitted. Likewise, if
carrying out simulations only for channels with no significant thermally emitted contribution (wavelengths < 3µm) the
calcemis and emissivity arguments may be omitted.
For IR scattering calculations for which you wish to supply the scattering optical parameters directly, the
aer_opt_param (for aerosols) and/or the cld_opt_param (for clouds) arguments should be supplied as described in
sections 8.5 and 8.6.
The “trajectory” structure traj may be (de)allocated outside of RTTOV and passed in to improve performance on some
architectures as described in section 7.7.
The files src/test/example_fwd.F90 and src/test/example_pc_fwd.F90 provide examples of running rttov_direct for
standard RTTOV and PC-RTTOV. The files src/test/example_aer_file_fwd.F90 and
src/test/example_cld_file_fwd.F90 provide examples of running rttov_direct for IR aerosol and cloud scattering using
coefficient files. The files src/test/example_aer_param_fwd.F90 and src/test/example_cld_param_fwd.F90 provide
examples of running rttov_direct for IR aerosol and cloud scattering where the scattering optical parameters are
supplied to RTTOV explicitly via aer_opt_param/cld_opt_param.
In the following table nchanprof is the size of the chanprof(:) array (i.e. the total number of radiances to compute),
and nprof is the size of the profiles(:) array (i.e. the number of profiles to process in each call to RTTOV).
nchannelsrec is the number of reconstructed radiances required per profile for Principal Components calculations.
The rttov_parallel_direct subroutine has the same arguments as rttov_direct plus an optional final argument nthreads
to specify the number of threads. Note that the trajectory argument cannot be used by the parallel routine.
There are three additional optional arguments to rttov_direct (traj_sta, traj_dyn and lbl_check) which are not intended
for use in typical calls.
121
Version : 1.3
Date : 05/03/2019

Integer Intent(out) errorstatus Return code.
Type(rttov_chanprof) Intent(in) chanprof(nchanprof) Chanprof structure.
Type(rttov_profile) Intent(in) profiles(nprof) Profiles structure.
Type(rttov_transmission) Intent(inout) transmission Output transmittances (0-1).
Type(rttov_radiance) Intent(inout) radiance Output radiances (mW/cm-1/sr/m2, degK,
and BRF/unitless).
Type(rttov_radiance2) Intent(inout), radiance2 Secondary output radiances (mW/cm-
1
optional /sr/m2).
Logical Intent(in), calcemis(nchanprof) .true. if RTTOV should calculate/select
optional surface emissivity internally or .false. if
user is supplying an emissivity value
(e.g. from an atlas).
Type(rttov_emissivity) Intent(inout), emissivity(nchanprof) Input/output emissivities.
optional
Logical Intent(in), calcrefl(nchanprof) .true. if RTTOV should calculate/select
optional surface BRDF internally or .false. if user
is supplying a BRDF value (e.g. from the
atlas).
Type(rttov_reflectance) Intent(in), reflectance(nchanprof) Input/output reflectances.
optional
Type(rttov_opt_param) Intent(in), aer_opt_param Aerosol optical parameter input profiles
optional if opts%rt_ir%addaerosl and
opts%rt_ir%user_aer_opt_param are
true.
Type(rttov_opt_param) Intent(in), cld_opt_param Cloud optical parameter input profiles if
optional opts%rt_ir%addclouds and
opts%rt_ir%user_cld_opt_param are
true.
Type(rttov_traj) Intent(inout), traj Trajectory structure to hold temporary
optional data: may improve performance on some
architectures for certain types of
simulation.
Type(rttov_pccomp) Intent(inout), pccomp Structure to hold output PC scores and
optional reconstructed radiances.
Integer Intent(in), channels_rec(nchannelsrec) Channels for which to compute
optional reconstructed radiances if
opts%rt_ir%pc%addradrec is .true.
122
Version : 1.3
Date : 05/03/2019
Annex J – RTTOV_K interface

call rttov_k (errorstatus, chanprof, opts, profiles, profiles_k, coefs,
transmission, transmission_k, radiance, radiance_k,
calcemis, emissivity, emissivity_k, calcrefl, reflectance,
reflectance_k, aer_opt_param, aer_opt_param_k, cld_opt_param,
cld_opt_param_k, traj, traj_k, pccomp, pccomp_k, profiles_k_pc,
profiles_k_rec, channels_rec)
The file src/test/example_k.F90 provides an example of calling the RTTOV K model. rttov_k should be
called for each instrument required. Each call carries out computations for the profiles in the array profiles(:). rttov_k
calculates the Jacobian for the RTTOV direct model, the output being written to profiles_k. As such, the profiles_k
array should have the same size as the chanprof array (i.e. the total number of radiances being computed). The
emissivity_k(:)%emis_in and reflectance_k(:)%refl_in arrays contain the Jacobians for the surface emissivity and
BRDF respectively.
See section 7.9 for more information about the RTTOV K model. The file src/test/example_k.F90 provides a simple
example of calling the RTTOV K model for clear-sky simulations.
Note that in RTTOV v9 the radiance_k argument was optional, but in RTTOV v12 this argument is mandatory. To
obtain the same behaviour as the default in RTTOV v9 you should set either radiance_k%total(:) or
radiance_k%bt(:) to 1.0, depending on the setting of opts%rt_all%switchrad.
All K arguments should be initialised to zero before calling rttov_k. The only exception is the array (or arrays) in
which the input perturbations are specified in radiance_k (or pccomp_k for PC-RTTOV).
Note that the emissivity and reflectance arguments are optional for rttov_k in exactly the same way as for rttov_direct.
In the following table, nchanprof is the size of the chanprof(:) array, and nprof is the size of the profiles(:) array.
npcscores is the number of PC scores requested per profile, and nchannelsrec is the number of reconstructed radiances
required per profile for Principal Components calculations.
The rttov_parallel_k subroutine has the same arguments as rttov_k plus an optional final argument nthreads to
specify the number of threads. Note that the trajectory argument cannot be used by the parallel routine.
123
Version : 1.3
Date : 05/03/2019

Type(rttov_profile) Intent(inout) profiles_k(nchanprof) Jacobian on profile variables.
Type(rttov_transmission) Intent(inout) transmission_k Jacobian of transmittances.
Type(rttov_radiance) Intent(inout) radiance Direct model output radiances (mW/cm-
1
/sr/m2, degK, BRF/unitless).
Type(rttov_radiance) Intent(inout) radiance_k Input radiance perturbations (see text).
Logical Intent(in), calcemis(nchanprof) True if RTTOV should calculate/select
optional surface emissivity internally or false if user is
supplying an emissivity value.
optional
Type(rttov_emissivity) Intent(inout), emissivity_k(nchanprof) Jacobian on surface emissivity.
optional
Logical Intent(in), calcrefl(nchanprof) True if RTTOV should calculate/select
optional surface BRDF internally or false if user is
supplying a BRDF value.
Type(rttov_reflectance) Intent(inout), reflectance(nchanprof) Input/output BRDFs.
optional
Type(rttov_reflectance) Intent(inout), reflectance_k(nchanprof) Jacobian on surface BRDF.
optional
Type(rttov_opt_param) Intent(in), aer_opt_param Aerosol optical parameter input profiles if
optional opts%rt_ir%addaerosl and
opts%rt_ir%user_aer_opt_param are true.
Type(rttov_opt_param) Intent(inout), aer_opt_param_k Jacobian of aerosol optical properties.
optional Optional even if aer_opt_param argument is
present.
opts%rt_ir%user_cld_opt_param are true.
Type(rttov_opt_param) Intent(inout), cld_opt_param_k Jacobian of cloud optical properties. Optional
optional even if cld_opt_param argument is present.
Type(rttov_traj) Intent(inout), traj Trajectory structure to hold temporary data.
optional
Type(rttov_traj) Intent(inout), traj_k Trajectory structure to hold temporary data.
optional
Type(rttov_pccomp) Intent(inout). pccomp Structure to hold output PC scores and
Type(rttov_pccomp) Intent(inout). pccomp_k Input PC or reconstructed radiance
optional perturbations (see text).
Type(rttov_profile) Intent(inout). profiles_k_pc(npcscores* Jacobian on principal component scores. Only
optional nprofiles) required if opts%rt_ir%pc%addradrec is
.false.
Type(rttov_profile) Intent(inout). profiles_k_rec(nchannelsrec* Jacobian on profile variables for
optional nprofiles) reconstructed radiance channels. Only
required if opts%rt_ir%pc%addradrec is
.true.
Integer Intent(in), channels_rec(nchannelsrec) Channels for which to compute reconstructed
optional radiances if opts%rt_ir%pc%addradrec is
.true.
124
Version : 1.3
Date : 05/03/2019
Annex K – RTTOV_TL interface

call rttov_tl (errorstatus, chanprof, opts, profiles, profiles_tl, coefs,
transmission, transmission_tl, radiance, radiance_tl,
calcemis, emissivity, emissivity_tl, calcrefl, reflectance,
reflectance_tl, aer_opt_param, aer_opt_param_tl, cld_opt_param,
cld_opt_param_tl, traj, traj_tl, pccomp, pccomp_tl, channels_rec)
rttov_tl should be called for each instrument required. Each call carries out computations for the profiles in the array
profiles(:). rttov_tl calculates the tangent linear of the RTTOV direct model given a profile perturbation in
profiles_tl(:).
See section 7.9 for more information about the RTTOV TL model.
Note that the emissivity and reflectance arguments are optional for rttov_tl in exactly the same way as for rttov_direct.
The rttov_parallel_tl subroutine has the same arguments as rttov_tl plus an optional final argument nthreads to
125
Version : 1.3
Date : 05/03/2019

Type(rttov_profile) Intent(inout) profiles_tl(nprof) Input profile variable increments.
Type(rttov_transmission) Intent(inout) transmission_tl TL of transmittances.
Type(rttov_radiance) Intent(inout) radiance Direct model output radiances (mW/cm-
1
/sr/m2, degK, BRF/unitless).
Type(rttov_radiance) Intent(inout) radiance_tl TL of radiances.
optional surface emissivity internally or .false. if
user is supplying an emissivity value
optional
Type(rttov_emissivity) Intent(inout), emissivity_tl(nchanprof) Input/output surface emissivity TL.
optional
optional surface BRDF internally or .false. if
user is supplying a BRDF value (e.g.
from the atlas).
optional
Type(rttov_reflectance) Intent(inout), reflectance_tl(nchanprof) Input/output surface BRDF TL.
optional
Type(rttov_opt_param) Intent(in), aer_opt_param Aerosol optical parameter input profiles
optional if opts%rt_ir%addaerosl and
opts%rt_ir%user_aer_opt_param are
true.
Type(rttov_opt_param) Intent(inout), aer_opt_param_tl Input aerosol optical property
optional increments. Optional even if
aer_opt_param argument is present.
opts%rt_ir%user_cld_opt_param are
true.
Type(rttov_opt_param) Intent(inout), cld_opt_param_tl Input cloud optical property increments.
optional Optional even if cld_opt_param
argument is present.
optional data.
Type(rttov_traj) Intent(inout), traj_tl Trajectory structure to hold temporary
optional data.
Type(rttov_pccomp) Intent(inout), pccomp_tl TL of principal components.
optional
126
Version : 1.3
Date : 05/03/2019
Annex L – RTTOV_AD interface

call rttov_ad (errorstatus, chanprof, opts, profiles, profiles_ad, coefs,
transmission, transmission_ad, radiance, radiance_ad,
calcemis, emissivity, emissivity_ad, calcrefl, reflectance,
reflectance_ad, aer_opt_param, aer_opt_param_ad, cld_opt_param,
cld_opt_param_ad, traj, traj_ad, pccomp, pccomp_ad, channels_rec)
rttov_ad should be called for each instrument required. Each call carries out computations for the profiles in the array
profiles(:). rttov_ad calculates the adjoint of the RTTOV direct model, the output being written to profiles_ad. The
emissivity_ad(:)%emis_in and reflectance_ad(:)%refl_in arrays contain the adjoint for the surface emissivity and
surface BRDF respectively.
All AD arguments should be initialised to zero before calling rttov_ad. The only exception is the array (or arrays) in
which the input perturbations are specified in radiance_ad.
See section 7.9 for more information about the RTTOV AD model.
Note that the emissivity and reflectance arguments are optional for rttov_ad in exactly the same way as for
rttov_direct.
The rttov_parallel_ad subroutine has the same arguments as rttov_ad plus an optional final argument nthreads to
127
Version : 1.3
Date : 05/03/2019

Type(rttov_profile) Intent(inout) profiles_ad(nprof) AD on profile variables.
Type(rttov_transmission) Intent(inout) transmission_ad AD of transmittances.
Type(rttov_radiance) Intent(inout) radiance Direct model output radiances
(mW/cm-1/sr/m2, degK, BRF/unitless).
Type(rttov_radiance) Intent(inout) radiance_ad Input radiance perturbations (see text).
optional surface emissivity internally or .false.
if user is supplying an emissivity value
optional
Type(rttov_emissivity) Intent(inout), emissivity_ad(nchanprof) AD on surface emissivity.
optional
optional surface BRDF internally or .false. if
user is supplying a BRDF value (e.g.
from the atlas).
optional
Type(rttov_reflectance) Intent(inout), reflectance_ad(nchanprof) AD on surface BRDF.
optional
Type(rttov_opt_param) Intent(in), aer_opt_param Aerosol optical parameter input
optional profiles if opts%rt_ir%addaerosl and
opts%rt_ir%user_aer_opt_param
are true.
Type(rttov_opt_param) Intent(inout), aer_opt_param_ad AD on aerosol optical properties.
optional Optional even if aer_opt_param
Type(rttov_opt_param) Intent(in), cld_opt_param Cloud optical parameter input profiles
optional if opts%rt_ir%addclouds and
opts%rt_ir%user_cld_opt_param
are true.
Type(rttov_opt_param) Intent(inout), cld_opt_param_ad AD on cloud optical properties.
optional Optional even if cld_opt_param
optional data.
Type(rttov_traj) Intent(inout), traj_ad Trajectory structure to hold temporary
optional data.
Type(rttov_pccomp) Intent(inout), pccomp_ad Input PC or reconstructed radiance
optional perturbations (see text).
128
Version : 1.3
Date : 05/03/2019
Annex M – RTTOV_SCATT interface

call rttov_scatt (errorstatus, opts_scatt, nlevels, chanprof, frequencies,
profiles, cld_profiles, coef_rttov, coef_scatt, calcemis,
emissivity, radiance, cfrac, emis_retrieval_terms)
rttov_scatt should be called for each instrument required. Each call carries out computations for the profiles defined in
the arrays profiles(:) and cld_profiles(:). See src/mw_scatt/example_rttovscatt.F90 and
src/test/example_rttovscatt_fwd.F90 for examples of calling rttov_scatt.
In RTTOV v12 it is possible to control a limited number of the internal RTTOV options via the opts_scatt parameter
(see Annex O). The lusercfrac option has been moved into this structure. Also note that the optional switch to revert to
the “old cloud scheme” has been removed.
Surface emissivities may now be supplied via the emissivity input variable of type rttov_emissivity (see Annex O) as
for rttov_direct. This also passes the emissivities calculated by RTTOV back out via the emis_out member.
The frequencies(:) and chanprof(:) arrays should be populated by calling rttov_scatt_setupindex (Annex D) before
calling rttov_scatt.
The rttov_parallel_scatt subroutine has the same arguments as rttov_scatt plus an optional final argument nthreads to
specify the number of threads.

Type(rttov_options_scatt) Intent(in) opts_scatt Options to control aspects of
RTTOV.
Integer Intent(in) nlevels Number of input profile levels.
Integer Intent(in) frequencies(nchanprof) Frequency indices.
Type(rttov_profile_cloud) Intent(in) cld_profiles(nprof) Cloud profile structure.
Type(rttov_coefs) Intent(in) coef_rttov RTTOV coefficient structure.
Type(rttov_scatt_coef) Intent(in) coef_scatt RTTOV_SCATT coefficient
structure.
Logical Intent(in) calcemis(nchanprof) .true. if RTTOV should calculate
surface emissivity or .false. if
user is supplying an emissivity
value (e.g. from the atlas).
Type(rttov_emissivity) Intent(inout) emissivity(nchanprof) Input/output emissivities.
Type(rttov_radiance) Intent(inout) radiance Output radiances (mW/cm-
1
/sr/m2 & degK).
Real Intent(out), cfrac(nprof) Effective cloud fractions actually
optional used (diagnostic).
Type(rttov_scatt_emis_retrieval_type) Intent(inout), emis_retrieval_terms Output structure containing
optional radiance and transmittance data
which can be used for all-sky
emissivity retrievals.
129
Version : 1.3
Date : 05/03/2019
call rttov_scatt_tl (errorstatus, opts_scatt, nlevels, chanprof, frequencies,

emissivity, profiles_tl, cld_profiles_tl, emissivity_tl,
radiance, radiance_tl)
rttov_scatt_tl should be called for each instrument required. Each call carries out computations for the profiles defined
in the arrays profiles(:) and cld_profiles(:). rttov_scatt_tl calculates the tangent linear of the RTTOV-SCATT direct
model given a profile perturbation in profiles_tl(:) and cld_profiles_tl(:). The emissivity TL values are treated as for
rttov_tl. See section 7.9 for more information about the RTTOV TL model.
See the description of the rttov_scatt subroutine above for information on common arguments.
The rttov_parallel_scatt_tl subroutine has the same arguments as rttov_scatt_tl plus an optional final argument
nthreads to specify the number of threads.

Type(rttov_options_scatt) Intent(in) opts_scatt Options to control aspects of RTTOV.
structure.
surface emissivity or .false. if user is
supplying an emissivity value (e.g.
from the atlas).
Type(rttov_profile) Intent(in) profiles_tl(nprof) Input profile variable increments.
Type(rttov_profile_cloud) Intent(in) cld_profiles_tl(nprof) Input cloud profile variable
increments.
Type(rttov_emissivity) Intent(inout) emissivity_tl(nchanprof) Input/output surface emissivity TL.
Type(rttov_radiance) Intent(inout) radiance Output radiances (mW/cm-1/sr/m2 &
degK).
Type(rttov_radiance) Intent(inout) radiance_tl TL of radiances.
130
Version : 1.3
Date : 05/03/2019
call rttov_scatt_ad (errorstatus, opts_scatt, nlevels, chanprof, frequencies,

emissivity, profiles_ad, cld_profiles_ad, emissivity_ad,
radiance, radiance_ad)
rttov_scatt_ad should be called for each instrument required. Each call carries out computations for the profiles
defined in the arrays profiles(:) and cld_profiles(:).
rttov_scatt_ad calculates the adjoint or the Jacobian of the RTTOV direct model, the output being written to
profiles_ad and cld_profiles_ad. In order to run the adjoint model the profiles_ad(:) and cld_profiles_ad(:) arrays
should be the same size as the profiles(:) array (i.e. the number of profiles being simulated). To run the Jacobian model
the profiles_ad(:) and cld_profiles_ad(:) arrays should be the same size as the chanprof(:) array (i.e. the total number
of channels being simulated over all profiles). In either case profiles_ad(:) and cld_profiles_ad(:) must be the same
size.
The input perturbation to the RTTOV-SCATT AD/K models is always in brightness temperature and should usually be
supplied in the radiance_ad%bt(:) array. All other AD/K arguments should be initialised to zero before calling
rttov_scatt_ad. The emissivity_ad argument is the same as for rttov_ad/rttov_k. See section 7.9 for more information
about the RTTOV AD and K models.
See the description of the rttov_scatt subroutine above for information on common arguments.
The rttov_parallel_scatt_ad subroutine has the same arguments as rttov_scatt_ad plus an optional final argument
nthreads to specify the number of threads.

Type(rttov_options_scatt) Intent(in) opts_scatt Options to control aspects of RTTOV.
structure.
surface emissivity or .false. if user is
supplying an emissivity value (e.g.
from the atlas).
Type(rttov_profile) Intent(inout) profiles_ad(nprof) OR AD or Jacobian on profile variables.
profiles_ad(nchanprof)
Type(rttov_profile_cloud) Intent(inout) cld_profiles_ad(nprof) OR AD or Jacobian on cloud profile
cld_profiles_ad(nchanprof) variables.
Type(rttov_emissivity) Intent(inout) emissivity_ad(nchanprof) AD or Jacobian on surface emissivity.
Type(rttov_radiance) Intent(inout) radiance Output radiances (mW/cm-1/sr/m2 &
degK).
Type(rttov_radiance) Intent(inout) radiance_ad Input BT perturbations (see text).
131
Version : 1.3
Date : 05/03/2019
Annex N – RTTOV Utility routines

1. RTTOV_USER_OPTIONS_CHECKINPUT interface
call rttov_user_options_checkinput(err, opts, coefs)
This subroutine checks that the input options are consistent with one another and with the supplied coefficient structure.
The routine will set err to errorstatus_fatal and print an error message if any inconsistencies are found. This can be
useful for debugging problems.
The source can be found in the src/main/ directory.

Type(rttov_options) Intent(in) opts Options structure.
Type(rttov_coefs) Intent(in) coefs RTTOV coefficients structure.
2. RTTOV_USER_PROFILE_CHECKINPUT interface
call rttov_user_profile_checkinput(err, opts, coefs, prof, aer_opt_param,
cld_opt_param, reg_limits_exceeded)
This subroutine checks profiles on user levels against the regression limits. You may wish to use this to check profiles
for unphysical or out-of-specification values before calling RTTOV. If you are supplying scattering optical parameters
for aerosol or cloud simulations, you can optionally pass in these structures as well. In this case,
opts%config%do_checkinput may be set to false so that input profiles are not checked again within the call to
RTTOV.
Only one profile is checked at a time. The value of err is errorstatus_fatal if any unphysical values are found in the
profile. If the additional logical flag reg_limits_exceeded is passed in, it is set to true if any profile variable exceeds the
RTTOV regression limits. See section 7.3 for details on the implications of this. Comparisons are made using the values
on the nearest coefficient pressure level below the input profile pressure level.
The source can be found in the src/main/ directory.

Type(rttov_coefs) Intent(in) coefs RTTOV coefficients structure.
Type(rttov_profile) Intent(in) prof Profile structure.
Type(rttov_opt_param) Intent(in), aer_opt_param Aerosol optical parameter profiles.
optional
Type(rttov_opt_param) Intent(in), cld_opt_param Cloud optical parameter profiles.
optional
Logical Intent(out), reg_limits_exceeded Set to true if any profile variable
optional exceeds the RTTOV regression limits
132
Version : 1.3
Date : 05/03/2019
3. RTTOV_PRINT_OPTS interface
call rttov_print_opts (opts, lu, text)
This subroutine prints out the contents of the options structure to the selected logical unit (or to error_unit if the lu
argument is omitted). The ability to see the option values being input to RTTOV can be useful for debugging problems.
The source can be found in the src/other/ directory.

Integer Intent(in), lu Logical unit for output (defaults to
optional error_unit).
Character Intent(in), text Additional text to print out.
optional
4. RTTOV_PRINT_OPTS_SCATT interface
call rttov_print_opts_scatt (opts_scatt, lu, text)
This subroutine prints out the contents of the RTTOV-SCATT options structure to the selected logical unit (or to
error_unit if the lu argument is omitted). The ability to see the option values being input to RTTOV-SCATT can be

Type(rttov_options_scatt) Intent(in) opts_scatt Options structure.
optional
5. RTTOV_PRINT_INFO interface
call rttov_print_info (coefs, lu, text)
This subroutine prints out some information about RTTOV (the library version number, and largest integer and real
values allowed). If a coefficient structure is supplied, information about this is also displayed. Output is printed to the
supplied logical unit (or to error_unit if the lu argument is omitted). The ability to see information about the
coefficient file being input to RTTOV can be useful for debugging problems.

Type(rttov_coefs) Intent(in), coefs Coefficients structure.
optional
optional
133
Version : 1.3
Date : 05/03/2019
6. RTTOV_PRINT_PROFILE interface
call rttov_print_profile (profile, lu, text)
This subroutine prints out the contents of the profile structure to the selected logical unit (or to error_unit if the lu
argument is omitted). The ability to see the profile values being input to RTTOV can be useful for debugging problems.

Type(rttov_profile) Intent(in) profile Profile structure.
optional
7. RTTOV_PRINT_CLD_PROFILE interface
call rttov_print_cld_profile (cld_profile, lu, text)
This subroutine prints out the contents of the RTTOV-SCATT cloudy profile structure to the selected logical unit (or to
error_unit if the lu argument is omitted). The ability to see the profile values being input to RTTOV-SCATT can be

Type(rttov_profile_cloud) Intent(in) cld_profile Cloud profile structure.
optional
8. RTTOV_PRINT_RADIANCE_QUALITY interface
call rttov_print_radiance_quality (quality, lu, text)
This subroutine prints out a human-readable interpretation of a single radiance%quality output. A non-zero value
indicates a warning flag has been set for the given output radiance. The output is written to the selected logical unit (or
to error_unit if the lu argument is omitted). This is intended for debugging purposes.

Integer Intent(in) quality Radiance quality output.
optional
134
Version : 1.3
Date : 05/03/2019
9. CREATE_AER_CLIM_PROF.EXE
This executable may be used to generate profiles on the layers defined by an arbitrary set of pressure levels for the first
10 aerosol types in the RTTOV OPAC aerosol property files (i.e. excluding the volcanic ash and Asian dust types) for
10 different climatological compositions. The source code can be found in src/other/.
The routine requires files in the current directory specifying the pressure profile (plevs.dat), and the temperature and
water vapour profiles (prof.dat). The units of water vapour are ppmv. Example input files can be found in the data/
directory of the RTTOV distribution.
It is perhaps easiest to run create_aer_clim_prof.exe from the data/ directory. The program prompts you for a latitude,
an elevation (km), the level of the surface (where the lowest level is 1 and the top of the atmosphere is nlev), and a scale
factor. The calculated profiles are multiplied by the scale factor.
The output is written to the file prof_aerosl_cl.dat in the current directory and consists of 10 columns, one for each of
the RTTOV aerosol types (excluding volcanic ash). Each column contains 10 consecutive profiles for the following
climatological compositions:
1 Continental clean
2 Continental average
3 Continental polluted
4 Urban
5 Desert
6 Maritime clean
7 Maritime polluted
8 Maritime tropical
9 Arctic
10 Antarctic
Note the units are number densities (cm-3) so you must set the RTTOV profiles(:)%mmr_cldaer variable to false when
passing these profiles intro RTTOV.
10. RTTOV_AER_CLIM_PROF
This subroutine is called by create_aer_clim_prof.exe to generate the aerosol profiles. You may wish to call this
subroutine directly in your own code: this is done in src/test/example_aer_file_fwd.F90.
call rttov_aer_clim_prof (p, t, q, levsurf, latitude, elevation, &

scalefactor, aerprof)
135
Version : 1.3
Date : 05/03/2019

Real Intent(in) p(nlevels) Pressure profile (hPa).
Real Intent(in) t(nlevels) Temperature profile (K).
Real Intent(in) q(nlevels) Water vapour profile (units: see
gas_units argument).
Integer Intent(in) gas_units Units for water vapour: 0=>ppmv
over dry air; 1=>kg/kg over moist air;
2=>ppmv over moist air
Logical Intent(in) mmr_aer Output aerosol units: true => kg/kg,
false => cm-3
Integer Intent(in) levsurf Index of level nearest the surface.
Real Intent(in) latitude Latitude of profile (degrees).
Real Intent(in) elevation Elevation of profile (km)
Real Intent(in) scalefactor Output profiles are multiplied by this
scale factor.
Real Intent(out) aerprof(nlayers,10,13) Output aerosol number densities on
layers for each of the 10 compositions
listed above, and for each of the 13
RTTOV OPAC aerosol types (units
determined by mmr_aer argument)
11. RTTOV_ZUTILITY
The module src/other/rttov_zutility.F90 may be used to obtain values for the magnetic field strength and orientation
for use with the Zeeman coefficient files.
The look-up table (LUT) must first be loaded with the following function:
errorstatus = load_bfield_lut(filename_LUT)
One LUT is supplied with RTTOV v11 in data/Be_LUT.2007.txt. The LUT is stored in the rttov_zutility module in
the array BField. To return the field strength and orientation there are three options:
call compute_bfield(latitude, longitude,

Bx, By, Bz, Be)
call compute_bfield(latitude, longitude, sensor_zenang, sensor_aziang,

Be, cos_bkang, cos_baziang)
call compute_bfield(latitude, longitude, sensor_zenang, sensor_relative_aziang,

Julian_day, utc_time, Be, cos_bkang, cos_baziang)
Finally, a subroutine is available to return the field orientation:
call compute_kb_angles(Bx, By, Bz, sensor_zenang, sensor_aziang,

cos_bkang, cos_baziang)
The arguments to these routines are detailed in the table below.
136
Version : 1.3
Date : 05/03/2019

Real Intent(in) latitude Latitude of location (-90 to +90)
Real Intent(in) longitude Longitude of location (accepts 0 to
360 or -180 to 180)
Real Intent(in) sensor_zenang Sensor zenith angle
Real Intent(in) sensor_aziang Sensor azimuth angle (0 to 360,
North=0, positive clockwise)
Real Intent(in) sensor_relative_aziang Solar azimuth angle minus sensor
azimuth angle
Integer Intent(in) Julian_day Julian day 1=Jan 1, 365=Dec 31 (366
leap year)
Real Intent(in) utc_time Universal time 0.00-23.999 (GMT, Z
time)
Real Intent(out) Bx, By, Bz Magnetic field components: east, north
and zenith (positive upwards)
respectively. Units: Gauss.
Real Intent(out) Be Magnetic field strength. Units: Gauss.
Real Intent(out) cos_bkang Cosine of the angle between the
magnetic field Be vector and the wave
propagation direction k.
Real Intent(out) cos_baziang Cosine of the azimuth angle of the Be
vector in the (v, h, k) coordinates
system, where v, h and k comprise a
right-hand orthogonal system, similar
to the (x, y, z) Cartesian coordinates.
The h vector is normal to the plane
containing the k and z vectors, where
k points to the wave propagation
direction and z points to the zenith.
h = (z cross k)/|z cross k|. The azimuth
angle is the angle on the (v, h) plane
from the positive v axis to the
projected line of the Be vector on this
plane, positive counter-clockwise.
137
Version : 1.3
Date : 05/03/2019
12. RTTOV_OBS_TO_PC.EXE
The program rttov_obs_to_pc.exe (located in the bin/ directory, source code is in src/other/rttov_obs_to_pc.F90)
demonstrates how to convert radiance observations into PC-space which is necessary, for example, in applications
involving the assimilation of PCs.
The usage of the example executable is as follows:

$ rttov_obs_to_pc.exe \
--rtcoef_file ... \
--pccoef_file ... \
--obs_file ... \
--ipcbnd ... \
--ipcreg ... \
--npcscores ... \
--obs_bt
This program reads in a set of observations from the specified “obs_file” and calculates the required number of PC
scores based on the specified input coefficient files and the regression band and predictor sets (see section 8.8 for
information on PC simulations). PC scores are written to standard out.
You would typically need to modify the source code to suit your own application.
--rtcoef_file Optical depth coefficient file.
--pccoef_file PC coefficient file.
--obs_file ASCII file containing white-space-separated observation
values. Units are Kelvin if --obs_bt is present, otherwise
mW/m-2/sr-1/cm-1.
--ipcbnd PC band to use.
--ipcreg PC regression predictor set to use.
--npcscores The number of PC scores to calculate.
--obs_bt Optional: if present, input observations are BTs, otherwise
they are radiances (see --obs_file).
13. RTTOV_SCATT_EMIS_RETRIEVAL
After calling rttov_scatt with the optional emis_retrieval_terms argument, this subroutine can be used to carry out the
emissivity retrieval calculation given the corresponding observed brightness temperatures.
call rttov_scatt_emis_retrieval(chanprof, coef, emis_terms, obs_tb, land_emis)

Type(rttov_chanprof) Intent(in) chanprof(nchanprof) RTTOV chanprof structure as used
in call to rttov_scatt.
Type(rttov_coef) Intent(in) coefs RTTOV coefficients structure.
Type(rttov_scatt_emis_retrieval_type) Intent(in) emis_terms Emissivity retrieval terms structure
populated by a call to rttov_scatt.
Real Intent(in) obs_tb(nchanprof) Observed brightness temperatures
corresponding to simulated
brightness temperatures (K).
Real Intent(out) land_emis(nchanprof) Retrieved emissivities.
138
Version : 1.3
Date : 05/03/2019
14. RTTOV_MAKE_SCAERCOEF.EXE
This executable can be used to generate custom aerosol optical property files (scaercoef) containing an arbitrary number
of species for visible/IR scattering simulations. The primary input data required are refractive indices over a range of
wavelengths covering the sensor channels and the size distribution. Aerosols may be hydrophobic or hydrophilic: in the
latter case, the refractive indices and size distribution data must be supplied for each of several (user-specified) relative
humidity values. The executable is controlled by a namelist file which defines the optical properties and specifies paths
to files containing the refractive index and size distribution data in simple ASCII format. The tool is fully described in a
separate user guide: docs/readme_rttov_make_scaercoef.txt and there are example input files in the data/ directory.
15. RTTOV_SCALE_REF_GAS_PROF
This subroutine is intended to make it easy to pass scaled copies of the RTTOV reference trace gas profiles into
RTTOV simulations. A typical use case would be in simulating an older instrument for which you do not have an
explicit CO2 profile, but for which the RTTOV reference profile is inappropriate because the concentrations are
contemporary (~400ppmv). You first set up your simulation to specify all trace gases you wish to include: for example,
by setting opts%rt_ir%co2_data to true. You then populate the profiles structure with all of the profile data as usual
except for the CO2 profiles. Finally, before the call to RTTOV, you call rttov_scale_ref_gas_prof passing in the
profiles(:) and, for example, the maximum CO2 concentration in ppmv. The subroutine populates the co2(:) member of
each element of profiles(:) with the RTTOV reference (background) CO2 profile scaled to have the maximum CO2
value you specified. The subroutine may be used to populate one or more of the optional trace gas species in a single
call, and the scale factors are determined either by providing maximum values in ppmv over dry air, or by providing
column-integrated amounts in kg/m^2 (or, for ozone only, a column-integrated amount in Dobson units). The column-
integration is calculated as the sum of the layer delta-pressure multiplied by the layer average gas concentration in
kg/kg over moist air divided by the acceleration due to gravity. The subroutine handles any value of gas_units in the
input profiles. The coefficient file must support the variable gases being requested via the arguments. This subroutine
can also be used with HTFRTC in exactly the same way as for standard RTTOV.
call rttov_scale_ref_gas_prof(err, coefs, profiles, &

o3_col_int, o3_col_int_du, o3_max_ppmv, &
co2_col_int, co2_max_ppmv, &
n2o_col_int, n2o_max_ppmv, &
co_col_int, co_max_ppmv, &
ch4_col_int, ch4_max_ppmv, &
so2_col_int, so2_max_ppmv)

Type(rttov_coef) Intent(in) coefs RTTOV coefficients structure.
Type(rttov_profiles) Intent(inout) profiles(nprofiles) RTTOV profiles structures populated with all data except
the gas profiles in question.
Real Intent(in), o3_col_int Specify one of these (column integrated amount kg/m^2,
optional o3_col_int_du column integrated amount in Dobson units, or maxmimum
o3_max_ppmv amount in ppmv over air) to populate profiles with
suitably-scaled copies of the RTTOV reference ozone
profile.
Real Intent(in), co2_col_int Specify one of these (column integrated amount kg/m^2,
optional co2_max_ppmv or maxmimum amount in ppmv over air) to populate
profiles with suitably-scaled copies of the RTTOV
reference CO2 profile.
Real Intent(in), n2o_col_int Specify one of these (column integrated amount kg/m^2,
optional n2o_max_ppmv or maxmimum amount in ppmv over air) to populate
profiles with suitably-scaled copies of the RTTOV
reference N2O profile.
Real Intent(in), *_col_int … similarly for the other trace gases …
optional *_max_ppmv
139
Version : 1.3
Date : 05/03/2019
Annex O – RTTOV v12 derived types

RTTOV’s derived types are defined in the module rttov_types.F90. The derived types which are required in your code
which calls RTTOV are described here.
Options structure
The rttov_options structure holds switches which configure various aspects of RTTOV. The first step in running
RTTOV is to declare an instance of this structure and to set the members to appropriate values. The options are held in
several sub-types which group them by function.
Type Variable Description

General configuration options: opts % config
Logical opts%config%do_checkinput If true checks whether input profiles are within both absolute and
regression limits. If false no check is performed (default = true)
Logical opts%config%apply_reg_limits If true input profiles outside the limits specified in the coefficient
files are reset to the min/max. If false such profiles will generate
warning messages unless verbose is false (default = false).
Logical opts%config%verbose If false only messages for fatal errors are output (default = true).
Logical opts%config%fix_hgpl If false the input surface elevation is assigned to the pressure
level at or immediately below the input 2m (surface) pressure. If
true the surface elevation is assigned to the specified surface
pressure (default = false).
General radiative transfer options: opts % rt_all

Logical opts%rt_all%switchrad Determines input perturbation for AD/K routines: if true
radiance_ad/k % bt is used for channels with wavelengths >
3µm, otherwise radiance_ad/k % total is used (default = false)
Logical opts%rt_all%do_lambertian If true activate treatment of surface as Lambertian instead of
specular reflector for downwelling emitted radiance (default =
false). The specularity profile variable determines how specular
or Lambertian the surface is.
Logical opts%rt_all%lambertian_fixed_ angle If true the Lambertian downwelling radiance is computed using a
fixed effective angle. If false the angle is derived from a
parameterisation computed from the total atmospheric optical
depth (default = true). Only relevant if do_lambertian is true.
Logical opts%rt_all%rad_down_lin_tau If true the downwelling atmospheric emission is computed using
the linear-in-tau approximation for the Planck source term. If
false a simpler layer average source term is used which is slightly
faster and has negligible impact on radiances (default = true,
recommended = false). Note that the Lambertian downwelling
radiances always use the layer average source term.
Logical opts%rt_all%use_q2m If true activate use of surface humidity (default = true)
Logical opts%rt_all%addrefrac If true RTTOV calculations account for atmospheric refraction
(default = false)
Logical opts%rt_all%plane_parallel If true treat atmosphere as strictly plane-parallel (no curvature or
refraction); automatically activated for all simulated radiances if
the DOM scattering solver is used (default = false)
MW-only radiative transfer options: opts % rt_mw

Logical opts%rt_mw%apply_band_correction Selects whether to apply or ignore any band correction
coefficients present in the coefficient file. These only have a
significant impact on radiances, not brightness temperatures
(default = true, recommended = true).
Logical opts%rt_mw%clw_data If true user is supplying cloud liquid water profiles (default =
false). NB This applies to “clear-sky” simulations only: the cloud
140
Version : 1.3
Date : 05/03/2019
is treated as a purely absorbing medium. For full scattering use

RTTOV-SCATT instead.
Logical opts%rt_mw%clw_calc_on_coef_lev If true the CLW optical depths are computed on coefficient
levels, if false they are computed on input levels (default = true,
recommended = false)
Integer opts%rt_mw%clw_scheme Choose CLW permittivity parameterisation: 1=>Liebe (1989),
2=>Rosekranz (2015), 3=>Turner,Kneifel,Cadeddu(2016)
(default = 1/Liebe, recommended = 2/Rosenkranz)
Real opts%rt_mw%clw_cloud_top Lower pressure limit for MW CLW calculations (hPa): any CLW
in levels with pressures smaller than this value is ignored (default
= 322 hPa).
Integer opts%rt_mw%fastem_version Select the MW sea surface emissivity model to use. Valid values
are 0-6; values 1-6 specify FASTEM-1 to FASTEM-6; value 0
specifies TESSEM2 (default = 6).
Logical opts%rt_mw%supply_foam_fraction If true use the foam fraction value specified in
profiles(:)%skin%foam_fraction in FASTEM (default = false).
Does not apply to TESSEM2.
Visible/IR-only radiative transfer options: opts % rt_ir

Logical opts%rt_ir%ozone_data If true user is supplying ozone profiles (default = false)
Logical opts%rt_ir%co2_data If true user is supplying CO2 profiles (default = false)
Logical opts%rt_ir%n2o_data If true user is supplying N2O profiles (default = false)
Logical opts%rt_ir%co_data If true user is supplying CO profiles (default = false)
Logical opts%rt_ir%ch4_data If true user is supplying CH4 profiles (default = false)
Logical opts%rt_ir%so2_data If true user is supplying SO2 profiles (default = false)
Logical opts%rt_ir%addsolar If true enable solar calculations for solar-affected channels
(default = false)
Logical opts%rt_ir%rayleigh_single_scatt If false disables the Rayleigh single-scattering calculation for
visible channels when addsolar is true. Note that Rayleigh
extinction is included in the LBL simulations used to train the
optical depth coefficients so this is always included (default =
true, not recommended to set this to false without good reason).
Logical opts%rt_ir%do_nlte_correction If true includes non-LTE bias correction for hi-res sounders
(default = false) NB this is independent of addsolar.
Integer opts%rt_ir%solar_sea_brdf_model Select the solar sea BRDF model: 1=>JONSWAP, 2=>Elfouhaily
et al. (default = 1, recommended = 2).
Integer opts%rt_ir%ir_sea_emis_model Select the IR sea surface emissivity model to use.
Valid values: 1 => ISEM; 2 => IREMIS (default and
recommended = 2).
Logical opts%rt_ir%addaerosl If true account for scattering due to aerosols (default = false)
Logical opts%rt_ir%addclouds If true account for scattering due to clouds (default = false)
Logical opts%rt_ir%user_aer_opt_param If true and addaerosl is true the user specifies the aerosol
scattering optical parameters instead of supplying number density
profiles for pre-defined particle types (default = false).
Logical opts%rt_ir%user_cld_opt_param If true and addclouds is true the user specifies the cloud
scattering optical parameters instead of supplying concentration
profiles for pre-defined particle types (default = false).
Logical opts%rt_ir%grid_box_avg_cloud If true input cloud concentrations should represent grid box
average values (i.e. not accounting for cloud fraction). If false the
cloud concentration should represent the value for the cloudy
fraction of the layer (i.e. the grid box average divided by the layer
cloud fraction). Only applies when addclouds is true and
user_cld_opt_param is false (default = false)
Real opts%rt_ir%cldstr_threshold Threshold for cloud stream (column) weights to compute for
cloud scattering calculations (default = -1); recommended to be
set negative when calling TL, AD or K models. Only applies
when addclouds is true.
141
Version : 1.3
Date : 05/03/2019
Logical opts%rt_ir%cldstr_simple If true, a simplified cloud stream approach is used with one clear
and one cloudy column. The “effective” cloud fraction for the
profile is taken as the maximum input cloud fraction in the
profile layers from the ToA down to the pressure specified by
cldstr_low_cloud_top. Use with caution: this is intended for
mid- to upper-tropospheric channels (default = false). Only
applies when addclouds is true.
Real opts%rt_ir%cldstr_low_cloud_top Only applies when cldstr_simple is true: cloud fraction is
computed as maximum in layers from ToA down to the specified
pressure in hPa (default = 750 hPa). Use with caution.
Integer opts%rt_ir%ir_scatt_model Scattering model to use for thermal emission source term: 1 =>
DOM; 2 => Chou-scaling (default = 2); only applies when
addclouds or addaerosl is true.
Integer opts%rt_ir%vis_scatt_model Scattering model to use for solar source term:
1 => DOM; 2 => single-scattering (default = 1); only applies
when addclouds or addaerosl is true and addsolar is true.
Integer opts%rt_ir%dom_nstreams Number of streams (discrete ordinates) to use with DOM
scattering solver. Must be >=2 and even (default = 8); only
applies when addclouds or addaerosl is true and DOM is
selected as a scattering solver.
Real opts%rt_ir%dom_accuracy Parameter to determine convergence criterion for DOM
azimuthal loop for solar scattering simulations. If zero or less the
loop never exits early (default = 0.); only applies when
addclouds or addaerosl is true and DOM is selected as a
scattering solver.
Real opts%rt_ir%dom_opdep_threshold DOM ignores layers below this total level-to-space absorption
optical depth. If zero or less all layers are treated in solver
(default = 0.); only applies when addclouds or addaerosl is true
and DOM is selected as a scattering solver.
Principal Components-only radiative transfer options: opts % rt_ir % pc

Logical opts%rt_ir%pc%addpc If true carry out Principal Components calculations (default =
false).
Integer opts%rt_ir%pc%npcscores Specifies the number of PC scores to use in PC-RTTOV
calculations. If <=0 the number of PC scores is determined from
the size of the arrays in the pccomp structure. The advantage of
using this option is that you can allocate pccomp for the
maximum number of profiles you want to pass to RTTOV, but
then call RTTOV with any number of profiles up to this
maximum (which was not possible before this option was
introduced in RTTOV v12.3) (default = -1).
Logical opts%rt_ir%pc%addradrec If true the PC calculations will return reconstructed radiances as
well as the PC scores (default = false)
Integer opts%rt_ir%pc%ipcbnd The index of the PC spectral band: this should be 1 for PC
coefficients current as of v12.3
Integer opts%rt_ir%pc%ipcreg The index of the required set of PC predictors: 1-4 for IASI/IASI-
NG (see table 29 for the number of predictors), and 1-3 for AIRS
(for 200, 300 and 400 predictors).
Options related to interpolation and the vertical grid: opts % interpolation

Logical opts%interpolation%addinterp If true input profiles may be supplied on user-defined levels, and
internal interpolation is used (default = false).
Integer opts%interpolation%interp_mode Set the interpolation mode (see Table 9 in section 7.3). Valid
values are 1-5 (default = 1).
Logical opts%interpolation%reg_limit_extrap Extrapolate input profiles up to top coefficient level maintaining
relative values with respect to regression limits – see section 7.3
(default = false, recommended = true); only applies if addinterp
142
Version : 1.3
Date : 05/03/2019
is true.
Logical opts%interpolation%lgradp Allow TL/AD/K of user pressure levels (default = false); only
applies if addinterp is true.
Logical opts%interpolation%spacetop If true treat user’s model top as space boundary (default = true, it
is not recommended to set this false).
Options related to HTFRTC: opts % htfrtc_opts

Logical opts%htfrtc_opts%htfrtc Set to true to call HTFRTC (default = false)
Integer opts%htfrtc_opts%n_pc_in The number of Princpal Components to calculate. If <= 0 then
HTFRTC uses the maximum number of PC scores allowed by the
coefficient file (default = -1)
Logical opts%htfrtc_opts%reconstruct Set to true to enable computation of reconstructed radiances
(default = false)
Logical opts%htfrtc_opts%simple_cloud Enable simple cloud scheme (using cfraction and ctp profile
variables) in HTFRTC simulations (default = false)
Logical opts%htfrtc_opts%overcast Enable calculation of overcast radiances (analogous to the
RTTOV overcast radiance outputs) in HTFRTC simulations
(default = false)
Options structure for RTTOV_SCATT

The rttov_options_scatt type defined in rttov_types.F90 gives control over some of the internal RTTOV options for
MW scattering calculations. Note that RTTOV_SCATT purposefully does not make all core radiative transfer options
available.

Logical opts_scatt%config%do_checkinput If true, checks whether input profiles are within both absolute and
regression limits. If false, no check is performed (default = true).
Logical opts_scatt%config%apply_reg_limits If true, input profiles outside the limits specified in the coefficient
files are reset to the min/max. If false, such profiles will generate
warning messages unless verbose is false (default = false).
Logical opts_scatt%config%verbose If false, only messages for fatal errors are output (default = true)
Logical opts_scatt%config%fix_hgpl If false the input surface elevation is assigned to the pressure level
at or immediately below the input 2m (surface) pressure. If true the
surface elevation is assigned to the specified surface pressure
(default = false).
Logical opts_scatt%rad_down_lin_tau Sets the opts%rt_all%rad_down_lin_tau option for the call to
RTTOV (see above, default = true, recommended = false).
Logical opts_scatt%apply_band_correction Selects whether to apply or ignore any band correction coefficients
present in the coefficient file. These only have a significant impact
on radiances, not brightness temperatures (default = true,
recommended = true).
Integer opts_scatt%interp_mode Set the interpolation mode (see Table 9 in section 7.3). Valid
values are 1-5 (default = 1).
Logical opts_scatt%lgradp Enable TL/AD/K of user pressure levels in the call to RTTOV
(default = false).
Integer opts_scatt%reg_limit_extrap Extrapolate input profiles up to top coefficient level maintaining
relative values with respect to regression limits – see section 7.3
(default = false, recommended = true).
Integer opts_scatt%fastem_version Select the MW sea surface emissivity model to use. Valid
values are 0-6; values 1-6 specify FASTEM-1 to FASTEM-6;
value 0 specifies TESSEM2 (default = 6).
Logical opts_scatt %supply_foam_fraction If true use the foam fraction value specified in
profiles(:)%skin%foam_fraction in FASTEM (does not
apply to TESSEM2).
Logical opts_scatt%use_q2m If true, activate use of surface humidity (default = true)
Logical opts_scatt%lradiance If true do RT calculation in radiances instead of brightness
143
Version : 1.3
Date : 05/03/2019
temperatures (default = false, recommended = true).

Logical opts_scatt%lusercfrac If true, you should supply supply the effective cloud fraction in
the rttov_profile_cloud (see below). If false, this is calculated
internally in RTTOV-SCATT (default = false).
Real opts_scatt%cc_threshold If the effective cloud fraction for the profile is below this value, it
is ignored and the simulation is clear-sky. The value must be in
the range 0-1 (default = 0.05).
Profile structure
The rttov_profile structure is composed of the atmospheric part and two other structures for 2 meter air and skin
surface. If you are not able to provide ozone, CO2, etc profiles the flags ozone_data, co2_data and so on in the options
structure should be set to false.

Surface skin – type rttov_skin
Integer surftype 0=land, 1=sea, 2=sea-ice
Integer watertype 0=fresh water, 1=ocean water. Used only for sea surface solar BRDF model
and BRDF atlas.
Real t Radiative skin temperature (K)
Real salinity Practical salinity unit %o – FASTEM-4/5/6 only.
Real fastem(1:5) Land/sea-ice surface parameters for FASTEM.
Real soil_moisture Soil moisture (m3/m3) – not currently used.
Real snow_fraction Surface snow coverage fraction (0-1). Used only by IR emissivity atlas.
Real foam_fraction Foam fraction (0-1) to use in FASTEM if
opts%rt_mw%supply_foam_fraction is true. By default FASTEM
calculates the foam fraction internally.
Real specularity When opts%rt_all%do_lambertian is true, the Lambertian and specular
downwelling radiances are linearly combined with the specularity as the
weight: a value of zero implies fully Lambertian, a value of one implies
fully specular (valid range is 0-1).
Surface 2m – type rttov_s2m

Real t Temperature (K)
Real q Water vapour (units as per gas_units) – only used if opts%rt_all%use_q2m
is true.
Real o Ozone (units as per gas_units) – not currently used.
Real p Surface pressure (hPa)
Real u U 10m wind component (m/s) – only used by IREMIS, FASTEM and
TESSEM2 sea surface emissivity models and sea surface solar BRDF
model.
Real v V 10m wind component (m/s) – only used by IREMIS, FASTEM and
TESSEM2 sea surface emissivity models and sea surface solar BRDF
model.
Real wfetc Wind fetch (m) (length of water over which the wind has blown, typical
value 100000m for open ocean) – only used by sea surface solar BRDF
model.
Atmospheric profile – type rttov_profile

Character id User may give text ID to each profile.
(len=128)
Integer date(3) Year, month, day – used by solar calculations to adjust the solar irradiance
based on the time of year.
Integer time(3) Hour, minute, second – not currently used.
Integer nlevels Number of atmospheric levels –automatically populated when the profile
structure is allocated.
Integer nlayers Number of atmospheric layers (i.e. nlevels-1) – automatically populated
144
Version : 1.3
Date : 05/03/2019
when the profile structure is allocated.

Integer gas_units Units for gas abundances: (must be the same for all profiles)
2 => ppmv over moist air (default)
1=> kg/kg over moist air
0 (or less) => ppmv over dry air
Logical mmr_cldaer Units for clouds and aerosols: (must be the same for all profiles)
True => kg/kg for cloud and aerosol (default)
False => g/m3 for cloud, cm-3 for aerosol
Real p(nlevels) Pressure (hPa)
Real t(nlevels) Temperature (K)
Real q(nlevels) Water vapour (units as per gas_units)
Real o3(nlevels) Ozone (units as per gas_units)
Real co2(nlevels) CO2 (units as per gas_units)
Real n2o(nlevels) N2O (units as per gas_units)
Real co(nlevels) CO (units as per gas_units)
Real ch4(nlevels) CH4 (units as per gas_units)
Real so2(nlevels) SO2 (units as per gas_units)
Real clw(nlevels) Cloud liquid water (kg/kg) – MW only, treats cloud as absorbing medium
only in “clear-sky”; not used with RTTOV-SCATT.
Real aerosols(naertyp,nlayers) Aerosols (units as per mmr_cldaer switch) – visible/IR only. Note: the
naertyp is the number of number of aerosol types is defined in the scaercoef scattering coefficient
aerosol types defined in file. Not used if opts%rt_ir%user_aer_opt_param is true.
the scaercoef file
Real cloud(ncldtyp,nlayers) Cloud water/ice (units as per mmr_cldaer switch) – visible/IR only. Note:
ncldtyp=6: five water the number of water cloud types is defined in the sccldcoef scattering
cloud types and ice cloud coefficient file. Not used if opts%rt_ir%user_cld_opt_param is true.
Real cfrac(nlayers) Cloud fractional cover (0-1) – visible/IR only.
Real clwde(nlayers) Cloud liquid water particle effective diameter (microns) – this is mandatory
if clw_scheme=2. Not used if opts%rt_ir%user_cld_opt_param is true or
if the OPAC clw scheme is used.
Real icede(nlayers) Ice particle effective diameter (microns) – this is optional, where non-zero
this value is used in preference to the parameterisation specified by idg. Not
used if opts%rt_ir%user_cld_opt_param is true or if the Baran ice
scheme is used.
Integer clw_scheme Optical properties for cloud liquid water: 1=OPAC; 2=”Deff” scheme. This
is not used if opts%rt_ir%user_cld_opt_param is true.
Integer ice_scheme Optical properties for ice crystals: 1=SSEC; 2=Baran scheme. This is not
used if opts%rt_ir%user_cld_opt_param is true.
Integer idg Scheme for ice water content to effective diameter, Dg 1=Ou and Liou;
2=Wyser et al (recommended); 3=Boudala et al; 4=McFarquhar et al. This
is not used if opts%rt_ir%user_cld_opt_param is true or if the Baran ice
scheme is used or in layers where profiles(:)%icede(:)>0.
Real zenangle Local satellite zenith angle (degrees), maximum valid value depends on
coefficient file (see Table 4).
Real azangle Local satellite azimuth angle (0-360º; measured clockwise, east=90º, see
Figure 4) –used in solar simulations and by FASTEM.
Real sunzenangle Local solar zenith angle (degrees), solar radiation only included up to 85º.
Real sunazangle Local solar azimuth angle (0-360º; measured clockwise, east=90 º, see
Figure 4)
Real elevation Elevation (km)
Real latitude Latitude (deg) -90º to +90º
Real longitude Longitude (deg) 0-360º – used only by emissivity and BRDF atlases.
Real Be Earth magnetic field strength (Gauss) – Zeeman only.
Real cosbk Cosine of the angle between the Earth magnetic field and wave propagation
direction – Zeeman only.
Real ctp Cloud top pressure (hPa) – simple cloud scheme only.
Real cfraction Cloud fraction (0 - 1), 1 for 100% cloud cover – simple cloud scheme only.
145
Version : 1.3
Date : 05/03/2019
Profile structure for RTTOV_SCATT cloud/precipitation

The rttov_profile_cloud defined in rttov_types.F90 is for the RTTOV_SCATT microwave scattering calculations.
Type Profile variable Contents

Integer nlevels number of atmospheric levels, which should match that supplied in the other input
profiles – automatically populated when the profile structure is allocated.
Logical use_totalice logical flag to switch between using separate ice and snow, or total ice hydrometeor
types – automatically populated when the profile structure is allocated.
Logical mmr_snowrain logical flag to select units for snow and rain hydrometeors: True => kg/kg (default);
False => kg/m2/s – automatically populated when the profile structure is allocated.
Real cfrac Optional: if opts_scatt%lusercfrac=true., supply the effective cloud fraction, C,
here. This is normally calculated internally in RTTOV-SCATT
Real ph(:) nlevels+1 of half-level pressures (hPa)
Real cc(:) nlevels of cloud cover (0-1)
Real clw(:) nlevels of cloud liquid water (kg/kg)
Real ciw(:) nlevels of cloud ice water (kg/kg)
Real totalice(:) nlevels of total ice (kg/kg)
Real rain(:) nlevels of rain flux (units depend on mmr_snowrain)
Real sp(:) nlevels of solid precipitation flux (units depend on mmr_snowrain)
Optical parameter structure

The rttov_opt_param structure is used to specify profiles of optical parameters for each channel for aerosol and cloud
scattering simulations if the opts%rt_ir%user_aer_opt_param and/or the opts%rt_ir%user_cld_opt_param flags
are true.

Real abs(nlayers,nchanprof) Absorption coefficients (km-1). For each layer one value is
provided for every channel being simulated corresponding to the
chanprof structure.
Real sca(nlayers,nchanprof) Scattering coefficients (km-1)
Real bpr(nlayers,nchanprof) “b” parameters (no units): represents the fraction of backscattered
radiation at each layer. The rttov_bpr_calc subroutine is
provided to calculate these from the phase functions. Required by
the Chou-scaling scattering model.
Real phangle(:) Angles over which the phase function is defined (degrees).
Real pha(nphangle,nlayers,nchanprof) Phase functions (no units, must integrate to 4π over all scattering
angles): used for calculating the b parameter for all channels, and
used in solar scattering calculations.
Integer nmom Number of Legendre coefficients for phase function expansions
(excluding the first coefficient which is always unity); required
by the DOM scattering solver, must be greater than or equal to
opts%rt_ir%dom_nstreams.
Real legcoef(1:nmom+1,nlayers,nchanprof) Phase function Legendre coefficients. First coefficient is always
1. The rttov_legcoef_calc subroutine is provided to calculate the
coefficients from the phase functions. Required by the DOM
scattering solver.
146
Version : 1.3
Date : 05/03/2019
Chanprof structure
The rttov_chanprof structure is used to specify the channel and profile indices for each call to RTTOV. An array
should be declared of size equal to the total number of radiances to be computed (across all channels and profiles). Each
element of the chanprof(:) array provides a channel index and a profile index. The array should be ordered so that all
channels for the first profile are listed, followed by all channels for the second profile, and so on (see Table 13).

Integer chan Channel index.
Integer prof Profile index.
Emissivity structure
The rttov_emissivity structure is used to pass emissivity values into RTTOV (where calcemis is false) and to return the
emissivity values used by RTTOV. An array should be declared of size equal to the total number of radiances to be
computed (across all channels and profiles).

Real emis_in Input emissivity (only used if corresponding element of calcemis is false).
Real emis_out Output emissivity (same as emis_in where calcemis is false).
Reflectance structure
The rttov_reflectance structure is used to pass BRDF values into RTTOV (where calcrefl is false) and to return the
BRDF values used by RTTOV. An array should be declared of size equal to the total number of radiances to be
computed (across all channels and profiles). Note that refl_cloud_top is not an active variable in the TL/AD/K models.

Real refl_in Input BRDF (only used if corresponding element of calcrefl is false).
Real refl_out Output BRDF (same as refl_in where calcrefl is false).
Real refl_cloud_top Optionally specify BRDFs for cloud tops in the simple cloud scheme. Set to
zero to use internal defaults.
147
Version : 1.3
Date : 05/03/2019
Radiance structure
The rttov_radiance structure is composed of the output radiances in units of mW/cm-1/sr/m2 and the output brightness
temperatures in degK for each channel. Single element arrays are of size nchanprof (i.e. the size of the chanprof(:)
array), and arrays of 2 dimensions are of size (nlayers, nchanprof), where nlayers = nlevels - 1. Radiances, brightness
temperatures and reflectances (for solar calculations) are computed.

Radiances - units of mW/cm-1/sr/m2
Real clear(nchanprof) Clear sky top of atmosphere radiance output for each channel. This
includes aerosol scattering if opts%rt_ir%addaerosl is true.
Real total(nchanprof) Clear+cloudy top of atmosphere radiance for given cloud top pressure and
fraction for each channel for simple cloud scheme or fully cloudy radiance
if opts%rt_ir%addclouds is true.
Real cloudy(nchanprof) Cloudy top of atmosphere radiance for 100% fraction for each channel at
given cloud top pressure for simple cloud scheme or same as total if
opts%rt_ir%addclouds is true.
Real overcast(nlayer,nchanprof) Level to space overcast radiance at the level bounding the bottom of each
layer. For the layer j containing the surface (or the layer immediately
above the pressure level on which the surface lies), overcast(j,:) contains
the overcast radiance as if the cloud was at the surface pressure and 2m
temperature. For IR channels (wavelengths > 3µm) this assumes an
opaque black cloud. For solar channels with no thermally emitted
component (wavelengths < 3µm), this consists of reflected solar radiation
according to assumptions described in section 8.3. This is not calculated
for PC-RTTOV or IR aerosol simulations or scattering simulations using
the DOM solver.
Brightness temperatures – units of deg K.
Real bt(nchanprof) BT equivalent to total (clear+cloudy) top of atmosphere radiance output
for each channel
Real bt_clear(nchanprof) BT equivalent to clear top of atmosphere radiance output for
each channel
Bi-directional reflectance factors (BRFs) – unitless
Real refl(nchanprof) Reflectance (BRF) equivalent to total (clear+cloudy) top of
Real refl_clear(nchanprof) Reflectance (BRF) equivalent to clear top of atmosphere
radiance output for each channel
Additional outputs
Logical plane_parallel This flag is set to true if RTTOV used strict plane-parallel
geometry. This is automatically enforced for certain types of
simulations.
Integer quality(nchanprof) This is a bit mask which flags possible issues with simulated
radiances (see section 7.8 for the bit positions and their
descriptions). If zero there are no problems.
Secondary radiance structure

The rttov_radiance2 structure holds the “secondary” output radiances in units of mW/cm-1/sr/m2. These are only
calculated within the direct model, and are not calculated for PC or IR aerosol simulations or scattering simulations
using the DOM solver. They do not contain any solar contributions. One-dimensional arrays are of size nchanprof (i.e.
the size of the chanprof(:) array), and arrays of 2 dimensions are of size (nlayers, nchanprof), where nlayers =
nlevels - 1.
Similarly to rad%overcast, for the layer containing the surface (or the layer immediately above the pressure level on
which the surface lies), the radiance quantities defined on layers (up, down and surf) contain the radiances from/to the
surface pressure. For surf, this is evaluated at the surface skin temperature.
148
Version : 1.3
Date : 05/03/2019

Radiances - units of mW/cm-1/sr/m2
Real upclear(nchanprof) Clear sky upwelling radiance at top of atmosphere including surface
emission term, but omitting downwelling reflected radiance term.
Real dnclear(nchanprof) Clear sky downwelling radiance at surface.
Real refldnclear(nchanprof) Reflected clear sky downwelling radiance contribution to top of
atmosphere radiance.
Real up(nlayer,nchanprof) Summed upwelling atmospheric emission term at top of atmosphere for
layers down to the level bounding the bottom of each layer.
Real down(nlayer,nchanprof) Summed downwelling atmospheric emission term at bottom of layer for
layers down to the level bounding the bottom of each layer.
Real surf(nlayer,nchanprof) Radiance emitted by a black cloud at the level bounding the bottom of
each layer; for the surface layer this is evaluated for the surface skin
temperature.
Transmission structure
The rttov_transmission structure contains output transmittances. The transmittances are unitless and lie in the interval
[0, 1]. There are separate outputs for “thermal” IR channels and solar-affected channels for the surface-satellite
transmittances. This is because the solar optical depth calculation uses the “effective” path length (the combined path
length along the sun-surface-satellite path) while the thermal optical depths are calculated using the path length along
the surface-satellite path. In addition, some channels use Planck-weighted coefficients for the thermal component
calculations and for these channels it is necessary also to calculate non-Planck-weighted transmittances for the solar
calculations. Also note that for IR aerosol simulations the output transmittances are those which result from the Chou
scaling scheme (if used), not the clear-sky transmittances.

Real tau_total(nchanprof) Transmittance from surface to top of atmosphere (TOA) along
the satellite view path. Only populated for channels with a
Real tau_levels(nlevel, nchanprof) Transmittance from each standard pressure level to TOA along
the satellite view path. Only populated for channels with a
Real tausun_total_path2(nchanprof) Transmittance for combined sun-surface-satellite path. Only
populated for solar-affected channels.
Real tausun_levels_path2(nlevel, nchanprof) Transmittance from TOA to each standard pressure level to
TOA along combined sun-surface-satellite path. Only
populated for solar-affected channels.
Real tausun_total_path1(nchanprof) Transmittance from surface to TOA along the satellite view
path. Only populated for solar-affected channels.
Real tausun_levels_path1(nlevel, nchanprof) Transmittance from each standard pressure level to TOA along
the satellite view path. Only populated for solar-affected
channels.
RTTOV-SCATT emissivity retrieval terms structure

The rttov_scatt_emissivity_retrieval_type structure holds radiance and transmittance outputs from RTTOV-SCATT
which can be used for dynamic all-sky emissivity retrievals. The radiances are in units of mW/cm-1/sr/m2. These are
only calculated within the RTTOV-SCATT direct model. All arrays are of size nchanprof (i.e. the size of the
chanprof(:) array).

Real cfrac(nchanprof) RTTOV-SCATT effective cloud fraction (Tallsky = cfrac * Tcld + (1-
cfrac) * Tclr
Real bsfc(nchanprof) Surface blackbody Planck radiance.
Real tau_cld(nchanprof) Along-path transmittance, surface to space (cloudy column).
149
Version : 1.3
Date : 05/03/2019
Real up_cld(nchanprof) TOA upwelling cloudy radiance or TB from atmosphere (not inc. surface
emission or reflection).
Real down_cld(nchanprof) Surface downwelling cloudy radiance or TB (inc. cosmic term).
Real tau_clr(nchanprof) Along-path transmittance, surface to space (clear column).
Real up_clr(nchanprof) TOA upwelling clear radiance or TB from atmosphere (not inc. surface
emission or reflection).
Real down_clr(nchanprof) Surface downwelling clear radiance or TB (inc. cosmic term).
PCCOMP structure
The rttov_pccomp structure contains the results of the Principal Component calculations for PC-RTTOV or HTFRTC.
You must specify the options you require for PC-RTTOV or HTFRTC in the options structure before calling the
rttov_alloc_pccomp subroutine to ensure the correct arrays are allocated. The “*pcscores” members have size equal to
the number of principal components multiplied by the number of profiles. The output radiance and BT arrays are only
allocated if opts%rt_ir%pc%addradrec or opts%htfrtc_opts%reconstruct is true. These are sized according to the
number of channels for which reconstructed radiances are required multiplied by the number of profiles. Unless
otherwise indicated, arrays are allocated for both PC-RTTOV and HTFRTC.

Real pcscores(npcscores) Computed PC scores, equal to clear_pcscores for HTFRTC.
Real total_pcscores(npcscores) Computed PC scores for cloudy radiances, equal to clear_pcscores if
simple_cloud is false (HTFRTC only)
Real clear_pcscores(npcscores) Computed PC scores for clear-sky radiances (HTFRTC only)
Real cloudy_pcscores(npcscores) Computed PC scores for 100% cloudy radiances. Only allocated if
simple_cloud is true. (HTFRTC only).
Real overcast_pcscores(nlayers, Computed PC scores for level to space overcast radiances at the level
npcscores) bounding the bottom of each layer. For the layer j containing the
surface (or the layer immediately above the pressure level on which
the surface lies), overcast(j,:) contains the overcast radiance as if the
cloud was at the surface pressure and 2m temperature. Only allocated
if overcast is true (HTFRTC only).
Real total_pccomp(nchannels_rec) Radiances constructed using pcscores for PC-RTTOV.
Radiances constructed using total_pcscores for HTFRTC.
Real clear_pccomp(nchannels_rec) Clear-sky radiances constructed using clear_pcscores (HTFRTC
only).
Real cloudy_pccomp(nchannels_rec) 100% cloudy radiances constructed using cloudy_pcscores, only
allocated if simple_cloud is true (HTFRTC only)
Real overcast_pccomp(nlayers, Overcast radiances constructed using overcast_pcscores, only
nchannels_rec) allocated if overcast is true (HTFRTC only)
Real bt_pccomp(nchannels_rec) BTs equivalent to total_pccomp reconstructed radiances.
Real bt_clear_pccomp(nchannels_rec) Clear-sky BTs equivalent to clear_pccomp reconstructed radiances
(HTFRTC only).
End of User Guide
150

Users Guide Rttov12 v1.3

Uploaded by

Copyright:

Available Formats

Users Guide Rttov12 v1.3

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Users Guide Rttov12 v1.3

Uploaded by

Copyright:

Available Formats

RTTOV v12 Users Guide Doc ID : NWPSAF-MO-UD-037

RTTOV v12 Users Guide

James Hocking, Peter Rayer, David Rundle and Roger Saunders

Copyright 2019, EUMETSAT, All Rights Reserved.

1. INTRODUCTION AND SCOPE ................................................................................................ 5

2. OVERVIEW OF RTTOV V12 .................................................................................................... 5

3. SUMMARY OF RTTOV CONCEPTS .................................................................................... 12

4. CHANGES FROM PREVIOUS VERSIONS .......................................................................... 14

10. REPORTING AND KNOWN BUGS FOR RTTOV V12 ....................................................... 86

11. FREQUENTLY ASKED QUESTIONS ................................................................................... 86

12. GLOSSARY ................................................................................................................................ 87

13. REFERENCES ........................................................................................................................... 88

14. ANNEXES ................................................................................................................................... 92

1. RTTOV_SETUP_EMIS_ATLAS interface ........................................................................ 114

1. Introduction and Scope

2. Overview of RTTOV v12

δy  H(x0 )δx (2)

B( ,T) (5)

Level Pressure Tmax Tmin Qmax Qmin O3max O3min O3Ref

Platform RTTOV ID Sat ID range Platform RTTOV ID Sat ID range

Sensor Chans => Sensor Chans =>

MODIS 13 (20-25, 27-36) => 1-16 1-36 => 1-36

ASTER 74 10-14 => 1-5 (1-2,3N,3B,4-14) => 1-15

3. Summary of RTTOV concepts

Important notes on v7/v8 vs v9 predictor coefficients:

v7 predictors v8 predictors v9 predictors

4. Changes from previous versions

4.1. Changes from RTTOV v12.2

Main new/updated features in v12.3

Clear-sky/general simulation updates:

Visible/IR scattering updates: (sections 8.5 and 8.6)

HTFRTC updates (see section 8.9):

Additional changes since v12.1

The CrIS FSR instrument ID has changed from 27 to 28.

Other new options

4.2. Changes from RTTOV v12.1

Main new/updated features in v12.2

Clear-sky/general simulation updates:

RTTOV-SCATT (section 8.7):

Visible/IR scattering simulations:

Principal Components simulations:

Additional changes since v12.1

Changes in behaviour and interfaces

New outputs and options

Principal Components simulations

4.3. Changes from RTTOV v11

Changes to RTTOV derived type names

Some derived types have been renamed to improve consistency:

Changes to the rttov_options structure

The opts%rt_mw%do_lambertian and opts%rt_ir%do_lambertian options have been removed. The

Changes to the rttov_profile structure

Changes to rttov_opt_param structure (visible/IR scattering optical properties)

Changes to subroutine interfaces

List of subroutines with interface changes:

Changes to emissivity and BRDF atlases

Other changes to behaviour and outputs

Additional capabilities of RTTOV v12 over RTTOV v11

Visible/IR scattering updates (sections 8.5 and 8.6):

Other science updates:

5. FORTRAN-90 UNIX/LINUX installation and testing