抄録
In this paper, the particle swarm optimization (PSO) is applied into automatic parameter calibration process of a distributed runoff model. As distributed runoff models require long simulation time compared with general optimization problems, the number of particles and repeat computation times should be selected property. We conducted sensitivity experiments for the number of particles and found that the PSO has to be applied in following conditions: i) to set the number of particles more than 100 in the case of calibrating about five parameters, ii) to conduct repeat computations about 25 times. Analyzed river discharge using identified parameters shows good agreement with the observed one.
