Investigating hydrological modeling uncertainties in the Mediterranean region by combining precipitation and soil moisture products

The study focus on five catchments in the Mediterranean Region distributed over Spain, France, Italy, Tunisia and Algeria. The study runs four lumped parameter hydrological models combining eight precipitation products and four soil moisture products. A Bayesian inference scheme is built to estimate posterior parameter distribution for any combination of hydrological model, precipitation and soil moisture. The simulated streamflows are evaluated using both Nash-Sutcliffe Efficiency criteria and multi-resolution analysis. The results from Bayesian inference combining streamflow and soil moisture is compared to the results when considering only streamflow as a benchmark. The results indicates that hydrological model performance through the Nash-Sutcliffe Efficiency criteria is more sensitive to the forcing precipitation product than the model structure. Also, forcing hydrological models with merged precipitation product brings better streamflow predictions than using satellite precipitation products. Regarding soil moisture accounting in hydrological modeling, the results show that including soil moisture in the parameter estimation can improve the predictive performance of hydrological models when the model is forced with satellite precipitation product. Also, soil moisture datasets derived from Sentinel-1 offer better consistency in hydrological modeling of river streamflow simulation. • HM performance is more sensitive to forcing precipitations than model structure. • CPC-GPM-ASCAT precipitation outperforms satellite precipitation products. • Considering soil moisture brings better model consistency. • Sentinel-1 derived soil moisture products outperforms other soil moisture products.