Performance evaluation of SMOS sea surface salinity observations in retrieving salinity profiles

Abstract

The Soil Moisture and Ocean Salinity (SMOS) remotely sensed sea surface salinity (SSS) observations, on a global scale and with various resolutions, have been applied as inputs in three representative retrieval techniques. The purpose is to evaluate the SSS observations' performance, concerning the accuracy of the salinity (S) profiles they retrieve, the spatial scales they can effectively resolve, and the way their errors are projected at depth. The SMOS SSS errors are small in the tropics and large in the extratropics. The spatial resolution, rather than temporal one, is an important factor for the accuracy of retrieved S. In the upper ∼ 100 m mixed layer, on average, the root mean square (rms) of retrieved S is larger than, equivalent to, or smaller than the signal rms, with a SSS resolution of 1/4°, 1/2°, 1 °, respectively, in the tropics, and always larger elsewhere. This can be explained by the SSS-S regression coefficients, which are uniformly large in the mixed layer and thus project vertically most of the surface error features. Meanwhile, the projection generally reduces the mesoscale structures in the 1/4° SSS fields. Below the mixed layer, mainly influenced by the sea level anomaly instead, the regions with high-accuracy retrieved S displace to higher latitudes. The SSS error patterns disappear, except in the southern oceans where the abnormally large SSS-S regression coefficients project SSS errors down to 400 ∼ 500 m. © 2013. The authors.