Developing heuristic multi-criteria auto calibration method for continuous HEC-HMS in snow-affected catchment

Abstract

Listen

Translate article

ABSTRACTTwo search methods are currently available for calibrating the HEC-HMS program: the Univariate Gradient (UG) and the Nelder Mead (NM) methods; both having poor performance in continuous models with snowmelt module. In this paper, a novel approach using Melody Search algorithm (MeS) is presented for auto-calibration of HEC-HMS and compared to Self-adaptive Global Harmony Search (SGHS) and Genetic Algorithms (GA). The results show that the suggested approach is able to overcome the poor performance of aforementioned methods yielding more enhanced solutions. While NM and UG reach a Nash Sutcliffe Index (NSE) of −0.33 and 0.105, GA, SGHS and MeS have been able to reach to 0.526, 0.635, and 0.674 respectively. In addition, the improving power of algorithms was tested by starting from the final solution found by the other methods. It was noticed that in all cases MeS and to some degree SGHS are able to substantially improve the solutions found by either UG or NM. Interestingly, while SGHS was able to slightly improve the solution of MeS algorithm, MeS made a substantial improvement on the solution reached by the SGHS. Finally, the novel MeS based multi-objective algorithm proposed here, i.e. NSMeS, and Non-dominated Sorting Genetic Algorithm-II (NSGA-II) are compared with different combinations of objective functions where NSMeS demonstrated better performance.