Error Covariance Matrices Characterization in the Ocean Salinity Retrieval Cost Function within the SMOS Mission

Abstract

he interests of the scientific community working on the Soil Moisture and Ocean Salinity (SMOS) ocean salinity level 2 processor definition are currently focused on improving the performance of the retrieval algorithm, which is based on an iterative procedure where a cost function relating models, measurements, and auxiliary data is minimized. For this reason, most of the effort is currently focused on the analysis and the optimization of the cost function.Within this framework, this study represents a contribution to the assessment of one of the pending issues in the definition of the cost function: the optimal weight to be given to the radiometric measurements with respect to the weight given to the background geophysical terms.A whole month of brightness temperature acquisitions have been simulated by means of the SMOS-End-to-End Performance Simulator. The level 2 retrieval has been performed using the Universitat Politècnica de Catalunya (UPC) level 2 processor simulator using four different configurations, namely, the direct covariance matrices, the two cost functions currently described in the SMOS literature, and, finally, a new weight (the so-called effective number of measurement).Results show that not even the proposed weight properly drives the minimization, and that the current cost function has to be modified in order to avoid the introduction of artifacts in the retrieval procedure. The calculation of the brightness temperature misfit covariance matrices reveals the presence of very complex patterns, and the inclusion of those in the cost function strongly modifies the retrieval performance. Worse but more Gaussian results are obtained, pointing out the need for a more accurate modeling of the correlation between brightness temperature misfits, in order to ensure a proper balancing with the relative weights to be given to the geophysical terms.