Reducing Bias in Flash Drought Forecasts by Optimizing Parameters in Noah-MP Multiple Parameterization Schemes

Noah-MP is a land surface model that provides multiple parameterization schemes with different levels of complexity. Application of this model in forecasting flash drought requires a basic understanding of various parameterization schemes in representing soil moisture, transpiration, and runoff. One of the challenges is to quantify the uncertainties in model structure (parameterization schemes) and to determine optimal parameters related to each of the schemes. This study uses Noah-MP to take account of three parameterization schemes for soil moisture–stomatal resistance relationships and four schemes for runoff and groundwater. The related parameters in the resultant 12 combinations of parameterization schemes are calibrated against observations from six selected FLUXNET and Asia-FLUX sites in China. The evolutionary multiobjective optimization method ε-NSGAII (Epsilon Nondominated Sorted Genetic Algorithm-II) is applied to optimize parameters in each combination of parameterization schemes. Features of the simulated flash droughts are analyzed based on soil moisture and evapotranspiration. The results reveal important but complementary relationships between the soil moisture–stomatal resistance schemes and runoff–groundwater schemes in predicting flash drought. The overall performance can be enhanced by choosing suitable schemes and parameter calibration, which indicates that the optimization of parameters makes up the biases due to the deficiency in model structure.