Application of multi-objective genetic algorithm on parameter optimization of DHSVM: A case study in shihmen reservoir catchment

This study aims to utilize a multi-objective genetic algorithm to searching optimize parameters of Distributed Hydrology-Soil-Vegetation Model (DHSVM) that lacking detail soil and plant parameters for model simulation instead of using default parameters. The runoff simulations of Shihmen reservoir by using optimize parameters and default parameters are discussed to identify the improvement of this study. DHSVM is a physicsbased distributed hydrological model that are required input data of elevation, boundary, flow direction, soil type, soil depth, vegetation, channel network to describe the physical phenomenon of catchment. However, the soil and vegetation parameters are too many to collect completely from observation data. Thus, the default setting of parameters are adopted in most researches. To improve this weakness of using default setting, a fast and elitist multi-objective genetic algorithm: NSGA-II (Non-dominated Sorting Genetic Algorithm II) is used to searching the best settings of soil and vegetation parameter of DHSVM. The optimization only focused on streamflow sensitive parameters such as porosity, lateral saturated hydraulic conductivity, maximum infiltration, field capacity, and the exponential decrease rate of lateral saturated hydraulic conductivity with soil depth. A design on the optimization is illustrated in this study by defining the encoding method, devising the conflict fitness value function. The simulation of using optimize parameters shows better performance than using default settings not only in typhoon events but also in long term period. The Nash-Sutcliffe efficiency coefficient increases from 0.6 to 0.8, and the ratio of volume increases from 0.56 to 0.81. The result shows that the genetic algorithm is feasible in optimizing parameters of the DHSVM model.