流出モデル評価への情報量規準の導入について

Abstract

Phisically-based runoff models are generally classified into two groupes: distributed parametermodel (DPM) and lumped parameter model (LPM). In this paper, we present a framework toevaluate LPM's, while a DPM is regarded as a prototype and used to simulate discharge datacontaining only random errors.Akaike's information criterion (AIC) is incorporated into the evaluation framework; so thatour method can be used to determine the order of a hierarchical model and to compare varioustypes of models.Tank cascade model (TC), Sugawara-type tank model (TANK) and Kimura's storage functionmodel (KSFM) are evaluated for several storm-flood data simulated. Though KSFM givesminimum AIC values for each data, the values of the model parameters vary with the flood.The framework presented here is important to evaluate LPM's and establish a proper LPM.