Uncertainty assessment and sensitivity analysis of soil moisture based on model parameter errors – Results from four regions in China

Abstract

Model parameter errors are an important cause of uncertainty in soil moisture simulation. In this study, a conditional nonlinear optimal perturbation related to parameter (CNOP-P) approach and a sophisticated land surface model (the Common Land Model, CoLM) are employed in four regions in China to explore extent of uncertainty in soil moisture simulations due to model parameter errors. The CNOP-P approach facilitates calculation of the upper bounds of uncertainty due to parameter errors and investigation of the nonlinear effects of parameter combination on uncertainties in simulation and prediction. The range of uncertainty for simulated soil moisture was found to be from 0.04 to 0.58 m3 m−3. Based on the CNOP-P approach, a new approach is applied to explore a relatively sensitive and important parameter combination for soil moisture simulations and predictions. It is found that the relatively sensitive parameter combination is region- and season-dependent. Furthermore, the results show that simulation of soil moisture could be improved if the errors in these important parameter combinations are reduced. In four study regions, the average extent of improvement (61.6%) in simulating soil moisture using the new approach based on the CNOP-P is larger than that (53.4%) using the one-at-a-time (OAT) approach. These results indicate that simulation and prediction of soil moisture is improved by considering the nonlinear effects of important physical parameter combinations. In addition, the new approach based on the CNOP-P is found to be an effective method to discern the nonlinear effects of important physical parameter combinations on numerical simulation and prediction.