Multiobjective calibration and sensitivity of a distributed land surface water and energy balance model

The feasibility of using spatially distributed information to improve the predictive ability of a spatially distributed land surface water and energy balance model (LSM) was explored at the U.S. Department of Agriculture Agricultural Research Service (USDA‐ARS) Walnut Gulch Experimental Watershed in southeastern Arizona. The inclusion of spatially variable soil and vegetation information produced unrealistic simulations that were inconsistent with observations, which was likely an artifact of both discretely assigning a single set of parameters to a given area and inadequate knowledge of spatially varying parameter values. Because some of the model parameters were not measured or are abstract quantities a multiobjective least squares strategy was used to find catchment averaged parameter values that minimize the prediction error of latent heat flux, soil heat flux, and surface soil moisture. This resulted in a substantial improvement in the model's spatially distributed performance and yielded valuable insights into the interaction and optimal selection of model parameters.