Application of Non-Dominated Sorting Genetic Algorithm in Calibration of HBV Rainfall-runoff Model: A Case Study of Tsengwen Reservoir Catchment in Southern Taiwan

Le Truong Vi

The objective of this study is to apply a multi-objective optimization algorithm for tuning parameters of the HBV rainfall-runoff model. This study selected the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) as optimization algorithm and examined various objective functions for investigating the performance of the HBV model in different flow situations (e.g., low flow and high flow). Two objective functions were chosen in this study: root mean squared error (RMSE) and mean absolute percentage error (MPE). Previous studies (e.g., Getahun and Van Laned, 2015) showed that the HBV might give bias estimates for low and high flow situations. Thus, the study proposed a season-dependent calibration strategy for further improving the biased estimates in different flow situations. The strategy is composed of two parts: (1) the RMSE-based objective function is used for wet seasons only (i.e., high flow situations); (2) the MPE-based objective function is used for dry seasons only (i.e., low flow situations). The preliminary results suggest that the proposed season-dependent strategy can improve results.

Keywords: multi-objective optimization algorithm, the HBV rainfall-runoff model, calibration strategy.