Recent advances in integrated hydrological and soil-vegetation-atmosphere transfer (SVAT)

modelling have led to improved water resource management practices, greater crop production, and better flood forecasting systems. However, uncertainty is inherent in all numerical models ultimately leading to imperfect model forecasts. It remains a crucial challenge to account for system uncertainty, so as to provide model outputs accompanied by a quantified confidence interval. Properly characterizing and reducing uncertainty opens-up the opportunity for risk-based decision-making and more effective emergency and disaster management.

The objective of this study is to develop and investigate methods to reduce hydrological model uncertainty by using supplementary data sources. The data is used either for model calibration or for model updating using data assimilation. Satellite estimates of soil moisture and surface temperature are explored in a multi-objective calibration experiment to optimize the parameters in a SVAT model in the Sahel. The two satellite derived variables were effective at constraining most land-surface and soil parameters. A data assimilation framework is developed and implemented with an integrated hydrological and tested by assimilating synthetic hydraulic head observations in a catchment in Denmark. Assimilation led to a substantial reduction of model prediction error, and better model forecasts. Also, a new assimilation scheme is developed to downscale and bias-correct coarse satellite derived soil moisture estimates. The approach successfully corrected soil moisture at the land surface, but with modest improvements in deeper soil layers.