NOAA Privacy Policy | NWS Disclaimer
NOAA logo Main WW3 logo National Weather Service logo

ERS2 Data Used To Test NWW3

 Point of contact: NCEP.EMC.waves@NOAA.gov
Last update: September 01, 2006
ERS2 FD data are used to validate the wind speeds and wave heights from NWW3. NCEP exclusively uses (reprocessed) FD products as only those are available in time to be used in operational models. For NWW3, this strategy is also used, in particular with an eye towards the intended future use of the data in assimilation schemes.
 
The altimeter produces estimates of wind speed and wave height along the ground track of the satellite every second (6.7 km spacing). To obtain wave observation at a spatial scale more representative for NWW3, these observations are averaged over 10 s (67 km). The data is quality controlled by removing data with unrealistic radar backscatter for open water, unrealistic retrieved wind speeds and wave heights, and large variability of the observation both within the 1 s observation and the 10 s averages. The scatterometer produces a wide swath of observations offset from the satellite ground track with an effective resolution of approximately 50 km. These data are used to validate driving wind fields only, and are used in its native resolution and with the included quality control checks. Only the ESA wind speed is considered.
 
To address the quality of the altimeter and scatterometer data, the observations as described above have been collocated with the deep ocean buoy observations, using a maximum collocation distance of 100 km and a maximum time difference of 30 min. for the altimeter and 50 km for the scatterometer (see also Tolman 1998a). For the scatterometer only the closest observation within the swath is used. This implies a mean collocation distance of roughly 50 km for the altimeter and 15 km for the scatterometer. All buoy wind data have been converted to 10 m height assuming neutral stability. Collocation data has been obtained for a full year starting with March 1997 and ending with February 1998, resulting in 1725, 1530 and 3163 collocations for the altimeter wave height , altimeter wind speed, and scatterometer wind speed, respectively.

Figure 1 shows the collocation results for the altimeter and buoy wave heights. For higher wave heights, a linear correction is needed (as has been shown in many publications). The correction is obtained by `inverse' regression (see Tolman 1998b in the references ) as

Hs = 0.03 m + 1.09 HFD

For lower wave heights a nonlinear correction is needed because the altimeter is unable to produce zero wave heights. An ad-hoc quadratic correction is constructed which results in Hs = 0 for a retrieved wave height of HFD = 0.6 m, and which fits with constant value and derivative to the above linear correction at a retrieved wave height of HFD = 2.0 m.
 
Figure 2 shows the corrected ERS2 FD wave height as used for the validation of NWW3. This wave height could be considered to be of at least as good a quality as the buoy wave heights, as most of the scatter in this figure can be attributed to sampling errors in the buoy wave height and to collocation errors in space and time (see Tolman 1998b).

Figure 3 shows the collocation results for the altimeter and buoy wind speeds. These data are stratified with the nondimensional wave height

H* = 3.33 g Hs U 10-2 ,

because altimeter wind data are known to be contaminated by the development stage of the wave field. H* = 1 roughly corresponds to fully developed seas, whereas an larger nondimensional wave height corresponds to swell conditions. This figure suggest a systematic dependence of the altimeter wind speed retrievals on the nondimensional wave height. This is confirmed by Fig. 4, which shows the corresponding regression lines. The retrieved wind speed clearly increases with increasing nondimensional wave height, although the dependence appears smaller than in Tolman (1998b) for ERS1. This might be explained by the fact that the latter study used two swell-dominated observations sites (32302 and 41018) which are presently no longer available. Furthermore, the latter study used northern hemisphere winter data only.
 
Figure 5 shows the mean and random error of the altimeter wind speed as a function of the (buoy) wind speed. These results have been obtained with and error-corrected bin-averaging technique as described in Tolman (1998a). The blue line presents the estimated of the random observation error as needed in such an analysis. The dashed red line represents and error-corrected linear regression line. The altimeter shows a small low bias for most wind speeds. A statistically stable and robust bias correction is obtained by using the linear regression line (dashed red line).

U10 = 0.34 m/s + 1.01 U a ,

where Ua is the altimeter wind speed and U10 is the corrected 10 m wind speed. To assure that the lowest possible wind speed remains 0 m/s, a quadratic correction with no effect for Ua = 0 m/s, and joining the linear correction with continuous value and derivative for Ua = 2.5 m/s is used. Figure 6 shows the collocation data set after this correction.
 
Several additional remarks need to be made on the (corrected) altimeter wind speeds :

  • The altimeter wind speed retrievals saturate around 21 m/s, and therefore cannot be used to assess extreme wind conditions.
  • The contamination of the wind speed retrieval by wave cannot be removed satisfactorily (see Tolman 1988b). These data are therefore suspect in a validation study of wave models.

Figure 7 shows the collocation results for the scatterometer and buoy wind speeds, also stratified with the nondimensional wave height. Note that the altimeter wind speeds show a somewhat discontinuous character due to the fairly course round-off in the data as obtained by NCEP. Furthermore, no wind speeds below 2 m/s are found. Figure 8 shows the regression lines for the stratified data sets. These data are of poorer quality than the corresponding ERS1 data (e.g., Tolman 1998b), and show only a moderate dependence on the development stage of the wave field. The apparent dependence appears to be at least partially due to the artificial low-wind-speed cutoff in the data. It therefore appears safe to assume that the scatterometer wind speeds are not significantly influenced by the local wave conditions.
 
Figure 9 shows the mean and random errors of the scatterometer wind, similar to Fig. 5 for the altimeter. The altimeter wind speeds are systematically biased low for high wind speeds. Again using the linear regression (dashed red line), the corrected wind speed becomes

U10 = -0.72 m/s + 1.15 U s ,

where Us is the scatterometer wind speed and U10 is the corrected 10 m wind speed. As with the altimeter, a nonlinear correction is applied for U s < 2.5 m/s. Figure 10 shows the collocation data after this correction.
 
Several additional remarks need to be made on the (corrected) scatterometer wind speeds :

  • The scatterometer wind speed saturates at high wind speeds similar to the altimeter wind speeds.
  • The ERS2 scatterometer wind speeds are of significantly poorer quality than the ERS1 wind speeds. This will make the use of altimeter wind speeds for case studies more difficult. The mean wind speeds of larger numbers of observations, however, are nevertheless useful, and appear not to be influenced significantly by the local wave conditions.

main wave model page | wave model validation page