Desagregação especial da precipitação simulada por modelos atmosféricos no Nordeste do Brasil.

Abstract

This dissertation was motivated by the importance of long-range precipitation forecasting for regional water resources management, and by recent research that showed good precipitation predictability by an Atmospheric General Circulation Model for a region in north-eastern Brazil. For applying these forecasts in water resources, the problem of difference of scales, spatial and temporal, between the atmospheric model and the drainage basins has been dealt with the so-called downscaling techniques. This work aimed to improve downscaling of seasonal and monthly precipitation, for a drainage basin in the semiarid north-eastern Brazil, using an Artificial Neural Network (ANN) technique. The seasonal downscaling was improved in comparison with a previous study for the same basin and, both in seasonal and monthly time scales, high correlations between modelled and observed data were obtained, thus indicating that the ANN technique is adequate for downscaling.