Data assimilation of surface soil moisture, temperature, and evapotranspiration estimates in a SVAT model over irrigated areas in semi-arid regions: what’s best to constraint evapotranspiration predictions?

Abstract

This study presents a strategy to improve the evapotranspiration estimates in semi arid areas using data assimilation in a SVAT (Soil Vegetation Atmosphere Transfer) modeling, the ISBA scheme (Interaction Soil Biosphere Atmosphere). In the perspective to use remote sensing products, the overall objective of this work is to identify the best combination of data (surface soil moisture / surface temperature / evapotranspiration), the temporal repetitiveness of acquisition (daily / tri-daily / weekly / bi-monthly / monthly) and the kind of data assimilation technique (two dimensional variational method / Extended Kalman filter) to constraint evapotranspiration predictions. Within this preliminary study, synthetic data referring to a wheat crops experimental site located in the Haouz Plain, part of the Tensift basin near Marrakesh in Morocco have been used (from January to May 2003). The results show that in order to improve the evapotranspiration through the analysis of the root zone soil moisture, the surface soil moisture is the most informative observation to use in the assimilation process (roughly 40% improvement in evapotranspiration RMSE). Combinations of observations improve the results but not significantly (few % improvement in evapotranspiration RMSE). Assimilation is very efficient for short assimilation windows. It is also shown that the propagation of the background error matrix done through the Extended Kalman filter doesn’t represent a significant added value with regards to the constant matrix used with two dimensional variational method.