Correcting satellite-based precipitation products through SMOS soil moisture data assimilation in two land-surface models of different complexity: API and SURFEX

Abstract

Global rainfall information is useful for many applications. However, real-time versions of satellite-based rainfall products are known to contain errors. Recent studies have demonstrated how the information about rainfall intrinsically contained in soil moisture data can be utilised for improving rainfall estimates. That is, soil moisture dynamics are impacted for several days by the accumulated amount of rainfall following within a particular event. In this context, soil moisture data from the Soil Moisture Ocean Salinity (SMOS) satellite is used in this study to correct rainfall accumulation estimates provided by satellite-based real-time precipitation products such as CMORPH, TRMM-3B42RT or PERSIANN. An algorithm based on the SMOS measurements data assimilation is tested in two land-surface models of different complexity: a simple hydrological model (Antecedent Precipitation Index (API)) and a more sophisticated state-of-the-art land-surface model (SURFEX (Surface Externalisée)). We show how the assimilation technique, based on a particle filter method, generally leads to a significant improvement in rainfall estimates, with slightly better results for the simpler (and less computationally demanding) API model. This methodology has been evaluated for six years at ten sites around the world with different land use and climatological features. The results also show the limitations of the methodology in regions highly affected by mountainous terrain, forest or intense radio-frequency interference (RFI), which can notably affect the quality of the retrievals. The satisfactory results shown here invite the future operational application of the methodology in near-real time on a global scale.