A stepwise surrogate model for parameter calibration of the Variable Infiltration Capacity model: The case of the upper Brahmaputra, Tibet Plateau

Abstract

To alleviate the computational burden of parameter calibration of the Variable Infiltration Capacity (VIC) model, a stepwise surrogate model (SM) is developed based on AdaBoost. An SM first picks out the parameter sets in the range that the values of objective functions are close to the optimization objectives and then approximates the values of objective functions with these parameter sets. The ε-NSGA II (Nondominated Sorting Genetic Algorithm II) algorithm is used to search the optimal solutions of SM. The SM is tested with a case study in the upper Brahmaputra River basin, Tibet Plateau, China. The results show that the stepwise SM performed well with the rate of misclassification less than 2.56% in the global simulation step and the root mean square error less than 0.0056 in the local simulation step. With no large difference in the optimal solutions between VIC and the SM, the SM-based algorithm saves up to 90% time.