Lecture 5. The Logarithmic Sublayer and Surface Roughness

Atm S 547 Boundary Layer Meteorology Bretherton
Lecture 5. The logarithmic sublayer and surface roughness

In this lecture…
• Similarity theory for the logarithmic sublayer.
• Characterization of different land and water surfaces for surface flux parameterization
Monin-Obukhov scaling in the surface layer
Near a solid boundary, in the `surface layer', vertical fluxes are transported primarily by
eddies with a lengthscale much smaller than in the center of the BL. A very successful similarity
theory is based on dimensional reasoning (Monin and Obukhov, 1954). It postulates that near
any given flat surface, the vertical gradients of the mean wind and thermodynamic profiles
should be determined purely by the height z above the surface (which limits the vertical size of
the eddies that carry the turbulent fluxes) and the surface fluxes which drive turbulence:
1/ 4
⎛ 2 2
⎞
1. Surface mom. flux. often expressed as friction velocity u* = ⎜ u′w′ 0 + v ′w′ 0 ⎟ .
⎝ ⎠
2. Surface buoyancy flux B0 = w′ b′ 0 .
In the ABL, a typical u* might be 0.3 m s-1. A typical range of B0 would be -3×10-4 m2s3
(nighttime) to 1.5×10-2 m2s-3 (midday), corresponding to a virtual heat flux of -10 W m-2 at night
and 500 W m-2 at midday. One can construct from these fluxes the
Obukhov length L = -u*3/kB0 (5.1)
Here k = 0.4 is the von Karman constant, whose physical significance we'll discuss shortly. L
is positive for stable BLs and negative for unstable BLs. The example values above give L = 200
m (nighttime) and -5 m (midday).
The logarithmic sublayer (Garratt, p. 41)
At height z, the characteristic eddy size, velocity, and buoyancy scale with z, u* and B0/u*. If
the buoyant acceleration acts over the eddy height, it would contribute a vertical velocity (zδb)1/2
= (z B0/u*)1/2. If z < |L|, this buoyancy driven contribution to the vertical velocity is much smaller
than the shear-driven inertial velocity scale u*, so buoyancy will not significantly affect the
eddies. In this case, the mean wind shear will depend only on u* and z. From now on, we will
use an unprimed variable to refer an ensemble mean (dropping the overline) unless otherwise
stated. We will also rotate our coordinates so +x is the near-surface mean wind direction and the
near-surface momentum flux is - u′w′ 0 i. By dimensional reasoning,
du/dz = u*/kz (z << |L|) (5.2)
This can also be viewed in terms of an eddy viscosity consistent with mixing length theory:
u′w′ 0 = -u*2 = -u*kz du/dz = - Kmdu/dz
Km ∝ (u*)(kz) = (eddy velocity)(eddy length) (5.3)
The von Karman constant k is the empirically determined constant of proportionality in (5.2).
Integrating (5.2), we get the logarithmic velocity profile law:
u(z)/u* = k-1 ln(z/z0) (z << |L|) (5.4)
5.1
The constant of integration z0 depends on the surface and is called the roughness length. The
figure below shows measured near-surface velocity profiles from the Wangara experiment that
compare very well with the predicted log-layer structure and are consistent with z0 = 5 mm.
Roughness length of various surfaces (Garratt, Ch. 4)

The roughness length is loosely related to the typical height of closely spaced surface obstacles,
often called roughness elements (e. g. water waves, trees, buildings, blades of grass). It depends
on the distribution as well as the height hc of roughness elements (see figure below), but as a rule
of thumb,
z0 ~ 0.1hc . (5.5)
5.2
5.3
z0 varies greatly depending on the surface, but a typical overall value for land surfaces is z0 = 0.1
m (see table on previous page). In rare circumstances, the surface is so smooth that the viscous
sublayer is deeper than roughness elements, whence
z0 ~ 0.1ν/u* ~ 0.015 mm for u* = 0.1 m s-1 . (5.6)
Near the surface, the log profile fits best if z is offset by a zero-plane displacement d0 which lies
between 0 and hc, and is typically around 0.7 hc:
u(z)/u* = k-1 ln([z-d0]/ z0) (z << |L|) (5.7)
Roughness of Water Surfaces (Garratt p. 97-100)
The roughness of a water surface depends on wind speed and the spectrum of waves. A strong
wind blowing from S to N across the SR 520 bridge shows the importance of fetch on wave
spectrum. On the south side, large waves will be crashing onto the bridge deck. On the N side,
the water surface will be nearly smooth except for short wavelength ripples (`cats paws')
associated with wind gusts. As one looks further N from the bridge, one sees chop, then further
downwind, longer waves begin to build. It can take a fetch of 100 km for the wave spectrum to
reach the steady state or fully developed sea assumed by most formulas for surface roughness.
It is thought that much of the wind stress is associated with boundary layer separation at sharp
wave crests of breaking waves or whitecaps, which start forming at wind speeds of 5 m s-1 and
cover most of the ocean surface at wind speeds of 15 m s-1 or more.
For wind speeds below 2.5 m s-1, the water surface is approximately aerodynamically
smooth, and the viscous formula for z0 applies. For intermediate wind speeds, the flow is
aerodynamically smooth over some parts of the water surface but rough around and in the lee of
the breaking whitecaps, and for wind speeds above 10 m s-1 it is fully rough. For rough flow,
Charnock (1955) suggested that z0 should depend only on the surface stress on the ocean and the
gravitational restoring force, i. e., u* and g, leading to Charnock's formula:
z0 = ac u*2/g, (ac = 0.016 ±20% from empirical measurements). (5.8)
This formula appears reasonably accurate for 10 m wind speeds of 4-50 m s-1. For 10 m wind
speeds of 5-10 m s-1, this gives roughness lengths of 0.1 - 1 mm, much less than almost any land
surface. Even the heavy seas under in a tropical storm have a roughness length less than mown
grass! This is because (a) the large waves move along with the wind, and (b) drag seems to
mainly be due to the vertical displacements involved directly in breaking, rather than by the
much larger amplitude long swell. The result is that near-surface wind speeds tend to be much
higher over the ocean, while surface drag tends to be smaller over the ocean than over land
surfaces.
Snow and Sand Surfaces (Garratt, p. 87-88)
The roughness of sand or snow surfaces also increases of wind speed, apparently due to
suspension of increasing numbers of particles. Charnock's dimensional argument again applies,
and remarkably, the same ac appears to work well, though now the minimum z0 is larger
(typically at least 0.05 mm), associated with the roughness of the underlying solid surface.
5.4
Bulk Aerodynamic Drag Formula (Garratt, p. 100-101)

Suppose that a wind measurement is taken at a standard reference level zR within the log
layer (A typical shipboard height of zR = 10 m is often used for ocean measurements). Then
(ignoring zero-plane displacement for simplicity), u(zR) = u* k-1 ln(zR/z0). The bulk aerodynamic
formula relates the surface stress ρ0 u′w′ 0 to the reference wind speed in terms of a neutral drag
coefficient CDN which depends on surface roughness:
- ρ0 u′w′ 0 = ρ0u*2 = ρ0CDNu2(zR), (5.9)
where CDN = k2/{ln(zR/z0)}2 (5.10)
The N, for ‘neutral’, in the suffix is to remind us that this formula only applies if when zR << |L|.
At typical reference heights (2 m or 10 m), this requires fairly neutrally stratified conditions, as
usually observed over the oceans but less reliably over land. For zR = 10 m and z0 = 0.1 m, CDN =
8×10-3.
Over the water, CDN is a function of surface roughness u* and hence implicitly of wind speed.
While Charnock's formula gives an awkward transcendental equation to solve for CDN in terms of
u(zR), a good approximation using mean 10 m wind speed u10 is:
CDN = (0.75 + 0.067u10) ×10-3 (water, neutrally stratified BL) (5.11)
5.5
Heat and Moisture Transfer in Neutral Conditions

Let a be a scalar (θ, q, etc.) transported by the turbulence. In the log-layer, we again might
hope for a flux-gradient relation of the form
w′ a′ = -Kada/dz, Ka = kazu* (5.12)
The nondimensional constant ka need not equal the von Karman constant k, since momentum
perturbations of fluid parcels are affected by eddy-induced pressure gradients, while scalars are
not. However, empirical measurements do suggest that ka = k in a neutral BL. A scale for
turbulent perturbations a¥ in the log layer is:
a* = w′ a′ 0 /u* (5.13)
Since the flux is approximately equal to its surface value throughout the surface layer,
da/dz = - w′ a′ 0 /(kzu*) = -a*/kz (5.14)
a(z) - a0 = - a*/k ln(z/za) (5.15)
This has the same logarithmic form as the velocity profile, but the scalar roughness length za
need not be (and usually isn't) the same as z0. In fact, it is often much smaller, because pressure
(form) drag on roughness elements helps transfer momentum between the interfacial (viscous)
sublayer around roughness elements to the inertial sublayer. No corresponding nonadvective
transfer mechanism exists for scalars, so they will be transferred less efficiently out of the
interfacial layer (za < z0) unless their molecular diffusivity is much larger than that of heat.
Note that temperature T is not an adiabatically conserved scalar; to use (5.15) you must
adiabatically correct the temperature measured at height z to the surface (z = 0) by adding gz/cp
(this is equivalent to using dry static energy s = cpT + gz as your conserved scalar).
The surface humidity over a water surface depends on whether it is fresh or salty, since salt
(which make up 2% of the molecules in sea water) does not evaporate, corresponding reducing
the vapor pressure over salt water. Also, one must use the skin temperature, which may differ
slightly from the bulk temperature measured below the water surface:
⎧⎪ 1, freshwater lakes
Over water: q0 = qsat ( p,Tskin )⋅ ⎨ (5.15a)
⎩⎪ 0.981, ocean
(5.15) can be converted into a bulk aerodynamic formula like (5.10), but the transfer
coefficient is different:
ρ0 w′ a′ 0 = ρ0CaNu(zR){a0 - a(zR)}, (5.16)
CaN = k2/{ln(zR/z0)ln(zR/za)} (5.17)
For most land surfaces, the heat and moisture scaling lengths zH and zq are 10-30% as large as z0,
resulting in a typical CHN of 0.7-0.95CDN . For water surfaces, the heat and moisture coefficients
are comparable to CDN for 10 m winds of 7 m s-1 or less, but remain around 1.1-1.3×10-3 (see
figures below) rather than increasing as wind speed increases. This corresponds to heat and
moisture scaling lengths appropriate for laminar flow even at high wind speeds. For instance,
ECMWF uses zH, zq = (0.4, 0.62)ν/u* following Brutsaert (1982).
5.6
Bulk aerodynamic formulas are quite accurate as long as (i) an appropriate transfer
coefficient is used for the advected quantity, the reference height, and the BL stability, and (ii)
Temporal variability of the mean wind speed or air-sea differences are adequately sampled. The
figure below shows comparisons between direct (eddy-correlation) measurements of moisture
flux in nearly neutrally stratified BLs over ocean surfaces compared with a bulk formula with
constant CqN = 1.32×10-3. In individual cases, discrepancies of up to 50% are seen (which are as
likely due to sampling scatter in the measured fluxes as to actual problems with the bulk
formula), but the overall trend is well captured. Due to this type of scatter, no two sources
exactly agree on the appropriate formulas to use, though all usually agree within about 10-20%.
5.7

Lecture 5. The Logarithmic Sublayer and Surface Roughness

Uploaded by

Copyright:

Available Formats

Lecture 5. The Logarithmic Sublayer and Surface Roughness

Uploaded by

Document Information

Original Title

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

Lecture 5. The Logarithmic Sublayer and Surface Roughness

Uploaded by

Copyright:

Available Formats

Atm S 547 Boundary Layer Meteorology Bretherton

Lecture 5. The logarithmic sublayer and surface roughness

Roughness length of various surfaces (Garratt, Ch. 4)

Bulk Aerodynamic Drag Formula (Garratt, p. 100-101)

Heat and Moisture Transfer in Neutral Conditions

You might also like