Computer Aided Numerical Methods for Hydrological Model Calibration: An Overview and Recent Development

Abstract

In this paper, the computer aided numerical method for hydrological model calibration is reviewed. The content includes review of the watershed hydrological models (data-driven model, conceptual model, and distributed model), review of the model calibration methods (manual calibration, single-objective automatic calibration, multi-objective automatic calibration, objective function, termination criterion, and data utilized for calibration), and review of the parallel computing accelerated model calibration (multi-node computer cluster, multi-core CPU, many-core GPU, and heterogeneous parallel computing). Recent development and the state-of-the-art are also analyzed. Three conclusions can be drawn: (1) Nowadays, different types of hydrological models have their own application fields and perform very well. Distributed hydrological model becomes the development direction and has a good future. (2) Computer aided automatic hydrological model calibration method has become the mainstream. Single-objective optimization method such as SCE-UA and multi-objective optimization method such as NSGA-II are very suitable to the model parameter calibration. (3) Heterogeneous parallel computing technology is the most powerful acceleration method for the hydrological model parameter calibration. However, researches about the acceleration of SCE-UA and NSGA-II based on heterogeneous parallel computing technique is rare and should be focused in the future.