Modelos chuva-vazão baseados em redes neurais artificiais para rios intermitentes no semiárido paraibano.

Abstract

This work aims to develop Artificial Neural Networks (ANN) models in order to estimate daily streamflows in Piancó River Basin, which is located in a semiarid region of Paraíba, Brazil. The basic principle of the proposed models consists of estimating daily streamflows based on past values of streamflow and precipitation. Rainfall-runoff models are essential for mitigating the impacts of climatic uncertainties and also for enabling an integrated knowledge of the regional climatology and hydrology. Semiarid areas suffer periodically from drought events and, paradoxically, with floods, that affect crops, rural villages, and poor neighborhoods of medium and small cities. This situation limits the social and economic development of that region. Extreme events can have their effects mitigated through the proper operation of existing reservoirs and, for that, a reliable prediction of streamflows is fundamental. The proposed ANN models were shown to be very efficient for the estimation of daily streamflows. The model is expected to be used in a drought and flood damage reduction system so that it can encourage the efficient management of existing water resources in that region and contribute to the minimization of social, economic and environmental conflicts.