On the Comparison of the Global Surface Soil Moisture product and Land Surface Modeling

Thanks to its large spatio-temporal coverage, the new ESA CCI multi-instruments dataset offers a good opportunity to assess and improve land surface models parametrization. In this study, the ESA CCI surface soil moisture (SSM) combined product (v2.2) has been compared to the simulated top first layers of the ORCHIDEE LSM (the continental part of the IPSL earth system model), in order to evaluate its potential of improvements with data assimilation techniques. The ambition of the work was to develop a comprehensive comparison methodology by analyzing simultaneously the temporal and spatial structures of both datasets. We analyzed the SSM synoptic, seasonal, and inter-annual variations by decomposing the signals into fast and slow components. ORCHIDEE was shown to adequately reproduce the observed SSM dynamics in terms of temporal correlation. However, these correlation scores are supposed to be strongly influenced by SSM seasonal variability and the quality of the model input forcing. Autocorrelation and spectral analyses brought out disagreements in the temporal inertia of the upper soil moisture reservoirs. By linking our results to land cover maps, we found that ORCHIDEE is more dependent on rainfall events compared to the observations in regions with sparse vegetation cover. These differences might be due to a wrong partition of rainfall between soil evaporation, transpiration, runoff and drainage in ORCHIDEE. To refine this analysis, a single value decomposition (SVD) of the co-variability between rainfall provided by WFDEI and soil moisture was pursued over Central Europe and South Africa. It showed that spatio-temporal co-varying patterns between ORCHIDEE and rainfall and the ESA-CCI product and rainfall are in relatively good agreement. However, the leading SVD pattern, which exhibits a strong annual cycle and explains the same portion of covariance for both datasets, explains a much larger fraction of variance for ORCHIDEE than for the ESA-CCI product. These results highlight that the role of other surface variables presenting a strong seasonal variability (like vegetation cover, possibly irrigation) is not accounted for similarly in both the model and the product, and that further work is needed to explore these discrepancies.