Automatic Optimization of Drainage Basin Water Balance Model Parameters Using Machine Learning

Abstract

The study by Numajiri [1] well reproduced the seasonal variation of river runoff by a monthly basin water balance model using a distributed tank model. The search for the optimal parameters is important for the construction of this distributed tank model. However, the method is not easy. When Sugawara [2] invented the tank model, Sugawara gave up the automatic search for the optimum parameters, citing one of the reasons that the computing power of the computer takes time. Numajiri [1] attempted automatic optimization by the conditional brute force attack method using a personal computer with improved performance. The purpose of this study is to improve the efficiency of optimal parameter exploration by using a machine learning library.