Combining the ensemble mean and bias correction approaches to reduce the uncertainty in hillslope-scale soil moisture simulation

The ROSETTA model has routinely been applied to predict the soil hydraulic properties for simulating the water flow at the hillslope scale. However, the uncertainties in water flow simulations are substantial due to the soil heterogeneity and ROSETTA model structure. In order to reduce these uncertainties, this study used the HYDRUS-2D and ensemble mean to simulate soil moisture based on the outputs of all candidate models. In addition, the bias correction techniques (including linear bias correction (LBC) and cumulative distribution function (CDF) matching) were also applied to improve the prediction of soil moisture. A total of 320 days of observed soil moisture data at two depths (10 and 30 cm) in the upper and lower slope positions were adopted to evaluate the performances of different bias correction methods results showed that the uncertainty in hillslope-scale soil moisture simulation due to the ROSETTA model structure was more important than that due to the soil heterogeneity. The CDF matching-based nonlinear bias correction approach was generally better than the LBC in reducing the uncertainty in soil moisture simulation. Combining the ensemble mean and CDF matching was a viable approach to improve the accuracy of the numerical model for simulating the hillslope-scale soil moisture variations.