Previsão hidrológica de longo-prazo baseada em componentes climáticas em bacias Hhdrográficas semiáridas/tropicais

Abstract

In a scenery of climate a environmental changes, assessments about hydrological regimes alterations in global and/or local scales coupled the climate variability conditions enable ponderations on the impacts and risks associate to water systems in a society increasingly dependent on the use of water, whether for basic activities, agriculture irrigation or hydropower generation. In this study we aimed was to evaluate the Long-term Precipitation Forecast in the Mundaú and Paraíba do Meio river basin, Northeast Brazil, through a stochastic model based on Artificial Neural Network (ANN) using climate indices as predictors. For this, precipitation data for the rainy season (May to July) from 1938 to 2008 were analysed and compared with climate indices such as Southern Oscillation Index - SOI, Northern Oscillation Index - NAO, Pacific Decadal Oscillation - PDO and Atlantic Multidecadal Oscillation - AMO, using some tools such as Principal Component Analysis (PCA), Cluster and Wavelets. It was highlighted that the local precipitation variability is associated with distinct climates and biomes in the region with the variability atmospheric modes that acting in that area. The SOI and NAO indices presented as precipitation predictors potential for the study area, with influences on the timescales interannual and decadal variability. The ANN highlights a satisfactory ability in predicting precipitation, with similarity fitting the data in steps of training, validation and testing on the model. These results could be a useful tool for assessing hydrological events, improving the water resources management in river basins.