Publishers version - DTU Orbit

The empirical model functions rely on the assumption that wind speed increases logarithmically
with height above the sea surface. This is normally true if the atmospheric boundary
layer is neutrally stratified. Stable stratification would typically lead to an underestimation
and unstable stratification to an overestimation of the 10 m wind speed. Deviations from the
logarithmic wind profile are mostly found in near-shore areas where the atmospheric boundary
layer may be influenced by the land. GMFs can thus be expected to perform better over the
open ocean than in near-shore areas.
The CMOD4 and CMOD5 were tuned empirically using co-located observations of NRSC
and (mainly) ECMWF ocean wind vectors. On average the atmospheric stabilityof the marine
boundarylayerisnotneutral.Therefore,to provideneutral10mwindretrieval CMOD5.Nwas
developed (Hersbach, 2010). The differences of CMOD5 and CMOD5.N include the average
stability correction of 0.2 ms −1 and the addition of 0.5 ms −1 to compensate for the negative
bias of CMOD5.
In summary, the C-band GMFs conveniently available from scatterometry are suitable for
wind retrieval from SAR VV data. The largest uncertainty is related to the necessary a priori
wind direction input.
15.4 Beyond C-band VV
RADARSAT-1 only collects C-band HH data. There is no GMF from scatterometry for HH.
It is therefore relevant to use C-band VV GMF with an additional function. This function
is the so-called Polarization Ratio (PR) that relates the NRCS of HH to NRCS of VV as a
function of radar incident angle. The HH should basically be translated to VV through this
function. The first suggested relationship includes a coefficient (α) the value of which ranges
from around 0.6 to 1.0 (Elfouhaily, 1997; Thompson et al., 1998; Vachon & Dobson, 2000).
Later work has shown that PR is also a function of wind speed and direction (Mouche et al.,
2005; Zhang et al., 2011).
SEASAT, JERS-1 and ALOS PALSAR collect L-band data. Based on JERS-1 and local
wind data and based on ALOS PALSAR and observed scatterometer wind vectors, two Lband
GMFs for HH are established (Shimada et al., 2004; Isoguchi and Shimada, 2009). The
L-band GMFs appear less sensitive to moderate winds than C-band GMF. L-band could be
advantageous for wind speeds larger than 20 ms −1 as L-band may not saturate for the high
wind speeds as C- and X-band do.
TerraSAR-X, TanDEM-X and COSMO-SkyMed collect X-band data. Based on TerraSAR-
X and in-situ buoy data the X-band GMF XMOD2 is established (Ren et al., 2012). The
data set is sparse. Another XMOD function was developed by interpolating the well-known
coefficients at two neighbouring bands from the C-band and Ku-band GMFs (Thompson et
al., 2012).
ResearchonnewGMFsforX-,C-andL-bandcanbebasedonquad-poldatafromCOSMO-
SkyMed, RADARSAT2 and ALOS PALSAR, respectively. To establish and verify a new GMF
requires many co-located samples covering a broad variety of wind conditions. Cross-pol data
has a higher noise floor than co-pol data, thus cross-pol data cannot map low wind speeds.
One advantage of cross-pol data for wind retrieval is that the GMF does not depend upon
wind direction and moderate to high wind speeds may be mapped and be useful for observing
hurricanes (Zhang and Perrie, 2012).
15.5 Current practices in SAR wind retrieval
SAR data is usually distributed as raw data (Level 1). The user has to calibrate the SAR data
with (updated) calibration coefficients provided with the SAR product or separately. The
calibration error should not exceed 0.5 dB. SAR images are not projected to a geographical
coordinate system and the user has to do this. The correction for incidence angle across the
swath also has to be made by the user. It is usual to block-average several SAR cells to pixels
sizes around 500 m or more before wind retrieval. The block-averaging (multi-look) is done to
280 DTU Wind Energy-E-Report-0029(EN)