Calibration of WetSpa model using NSGA-II and PSO multi-objective optimization algorithms

Conceptual rainfall-runoff (RR) models, aiming at predicting stream flow from the knowledge of precipitation over a catchment, evapotranspiration, tempreture, and topography of the basin, have become basic and effective tools for flow regime simulation. Calibration of RR models, e.g. WetSpa which has been developed in Belgium, is a process in which parameter adjustment are made so as to match the dynamic behaviour of the RR model to the observed behaviour of the catchment. This research presents an application of the Non-dominated Sorting Genetic Algorithm II (NSGA-II) and Particle Swarm Optimization (PSO) for multi-objective calibration of WetSpa in Karoon river basin, Iran to optimize 11 global parameters of the WetSpa model. The objective functions are Nash–Sutcliffe and logarithmic Nash–Sutcliffe efficiencies in order to improve the model's performance. Results showed that the evolutionary NSGA-II and PSO algorithms are capable of locating optimal parameter sets in the search space. The measured correlation coefficient in the calibration process was 0.69 and 0.71 for the NSGA-II and PSO algorithms, respectively. Moreover RMSE values were calculated as 119.8 and 152.3 m3/s for the algorithms. The WetSpa model then was applied for a period of 1-year flood simulation in the basin and the results were analysed. Finally a sensitivity analysis was conducted on the global parameters in which the surface runoff coefficient was the most sensitive parameter with more than 40% influence on the results.