Improvements of the most recent reprocessed SMOS soil moisture products

International audience; Soil Moisture and Ocean Salinity (SMOS) satellite has been providing soil moisture (SM) and ocean salinity (OS) retrievals at L-band for five years (2010-2014). During these five years, the SM retrieval algorithm, the L-MEB (L-Band Microwave Emission of the Biosphere; (Wigneron et al., 2007)) model has been progressively improved and hence results in different versions of the SMOS SM products. This study aims at evaluating the last improvement in the SM products of the most recent SMOS level 3 (SMOSL3) reprocessing (SMOSL3_New, Processor V2.72) vs. the earlier versions (SMOSL3_Old, retrieval algorithm processors V2.45-48). The new SM product (V2.72) includes improvements, but not limited to, in (i) Level 1 brightness temperature reprocessing including a better detection of Radio Frequency interferences (RFI) and (ii) the L-MEB SM retrieval model such as replacing the Dobson dielectric model (Dobson et al., 1985) by the Mironov dielectric mixing model (Mironov et al., 2009). METHOD The SMOSL3_New and SMOSL3_Old versions were compared against SM-DAS-2 SM data assimilation system produced by the European Center for Medium range Weather Forecasting (ECMWF) forecast model at the global scale. RESULTS Correlation, bias, Root Mean Square Difference (RMSD) and unbiased RMSD (unbRMSD) were used as performance criteria in this study. First results show that the SMOS SM estimates have been improved: (i) SMOSL3_New was closer to SM-DAS-2 over most of the globe-with the exception of arid regions-in terms of bias, RMSD, and unbRMSD (ii) SMOSL3_New was closer to SM-DAS-2 over Europe and India but comparable with SMOSL3_Old over most of the rest of the globe in terms of correlations. For instance, Fig. 1 shows the areas, in green, where SMOSL3_New has the lowest unbRMSD (66% of the considered pixels), and the areas, in red, where SMOSL3_Old has the lowest unbRMSD (34% of the considered pixels). The positive impact of introducing the improvements mentioned above in the new processing of SMOS datasets is clearly observed in the figure with the exception of deserts, particularly over sand dunes areas, where the Dobson model may be more accurate than the Mironov model. This study is in progress and will be extended using other existing satellites (ASCAT, AMSR-E, etc.) and in situ observations to precisely evaluate the improvements of the most recent SMOS soil moisture reprocessing.