Calibração automática de modelos tipo chuva-vazão utilizando técnicas de suavização: uma aplicação ao modelo SMAP-II

Abstract

Parameter calibration is a critical, but essential step in Hydrologic Modelling. In this way, some optimization algorithms for automatic calibration have been applied to conceptual Rainfall-Runoff modelling, but several problems have been detected, which have motivated many researches. This work examines the possibility of using Quasi-Newton type (second derivative-based) optimization algorithm for calibration of conceptual Rainfall-Runoff models. The technique developed presents a previous structural modification of a typical model to obtain the objective function explicit in terms of parameters. The threshold structures that implicate in "if’s" structures, generators of discontinuites, were replaced by smoothing functions thereby preserving the conceptual intregrity of the model. Application to some syntetic cases were included and considerations were done about real cases. The results indicate that a significant progress was achieved in the area of automatic calibration of conceptual Rainfall-Runoff models. A few problems were identified and can be explored in future researches.