The Soil Moisture and Ocean Salinity (SMOS) mission launched in November 2009 will provide, for the first time, satellite observations of sea surface salinity (SSS). At level 3 (L3) of the SMOS processing chain, the large amount of SSS data obtained by the satellite will be summarized in gridded products with the aim of synthesizing the information and reducing the error of individual SSS observations. In this paper, we present the algorithm adopted by the CP34 SMOS processing center to generate the SMOS L3 products and discuss the choices adopted. The algorithm is based on optimal statistical interpolation. This method needs the following: 1) the prescription of a background field; 2) a prefiltering procedure to reduce the data set size; 3) the definition of a suitable correlation model; and 4) the characterization of the observational error statistics. For the present initial stage, a monthly climatology is chosen as the best background field. The spatiotemporal correlations between the departures from the climatology are described using a bivariate Gaussian function. The correlation model parameters are obtained by fitting the function to the realistic ocean model data. The sensitivity experiments show that an accurate correlation model that permits local variations in the correlation parameters is the best option. The observational error statistics (bias, variance, and correlation) are addressed from the results of the SMOS level-2 processor simulator. Finally, several sensitivity experiments show that a bad prescription of observational errors in the L3 algorithm does result in a dramatic impact on the generation of L3 products.

Peer Reviewed