Meetings: Documents

Aquarius’ CAP Ocean Surface Salinity and Wind Products and Their Applications to Water Cycle Research
[16-Apr-13] Yueh, S., Tang, W., Fore, A., Hayashi, A., Lee, T., Lagerloef, G., Bindlish, R., Jackson, T., Murty, V., and Papa, F.
Presented at the 2013 SMOS-Aquarius Science Workshop
Aquarius is a combined passive/active L-band microwave instrument developed to map the sea surface salinity (SSS) field from space. This paper describes Aquarius' version-2.0 (V2.0) Combined Active-Passive (CAP) retrieval algorithm for simultaneous retrieval of surface salinity and wind. The CAP V2.0 product is available through the JPL Physical Oceanography Distributed Active Archive Center.
The CAP V2.0 algorithm includes updates to the geophysical model functions (GMF) and water dielectric constant to remove systematic salinity retrieval bias with respect to significant wave height (SWH) and sea surface temperature (SST). We use more than one year of Aquarius and NOAA WaveWatch3, SSMI/S, and NCEP wind matchup data to explicitly parameterize the L-band backscatter and excess surface emissivity as a function of wind speed, wind direction and SWH. The empirical GMF is found not only effective in removing the systematic bias, but also in reducing the Root-Mean-Square-Error (RMSE) of SSS retrieval.
The other key update is the water dielectric constant model. We find that at low SST the retrieval using the Klein and Swift model will produce a small positive bias, while the use of Meissner and Wentz model will result in a small negative bias. We use a linear combination of these two water dielectric constant models for retrieval. Both systematic bias and RMSE of retrieved SSS are reduced with respect to SST and wind speed.
The previous cost function for the CAP algorithm uses the Aquarius data only. However we find that the wind speed is less accurate at crosswind direction due to the reduced sensitivity of L-band backscatter along that direction. A less accurate wind speed estimate in turn negatively impacts the salinity retrieval. We therefore include the NCEP wind as a priori by adding another term in the cost function to improve the retrieval at crosswind direction.
We assess the accuracy of CAP V2.0 wind by performing comparison with SSMI/S wind and triple collocation. The root-mean-square (RMS) difference with respect to the SSMI/S wind speed is about 1 m/s. The triple collocation using the SSMI/S, CAP and ECMWF winds indicates that the accuracy of CAP wind speed is about 0.7 m/s, essentially the same as that of SSMI/S wind speed and less than the 0.85 m/s error for the ECMWF.
We estimate the accuracy of CAP's monthly average salinity by comparison with HYCOM's model salinity. We compute the mean and standard deviation of the differences between CAP and HYCOM salinities on 1x1 degree latitude and longitude grid. We find that the standard deviation is in the range of 0.1 to 0.2 psu for many bins globally, but could be larger than 0.3 psu at high latitude and in the tropics. The average of RMSD is computed for every 10 degrees latitude bands, and is found to be in the range of 0.2 to 0.25 psu between +/- 40 degree latitudes. If the HYCOM error is 0.17 psu, the corresponding Aquarius CAP's error will be between 0.1 and 0.2 psu.
We finally demonstrate the application of Aquarius data to water cycle research by examining the time series of salinity in the Bay of Bengal and Aquarius soil moisture, GRACE mass and river discharge observations in the Indian subcontinent. We find the expected time delay of about 1 to 2 months between the changes of water mass on land and salinity in the Bay of Bengal.

View Document (PDF, 6.3 MB)