Avaliação de um modelo estocástico de erro multidimensional aplicado a estimativas de precipitação por satélite

Abstract

One of the most important applications of satellite rainfall estimates is the hydrological modeling in basins where the conventional and real time rain gauges networks are inadequate in term of the spatial and temporal resolution. This study discuss the performance of the multidimensional stochastic error model (SREM2D), which simulates an ensemble of daily precipitation fields with the same statistical patterns (spread) as the differences of satellite precipitation fields and rain gauges of a longer data series. Most models treat errors only in one dimension, without recognizing that rainfall is a time and space intermittent process. The SREMD2 model characterize the spatial and temporal structure, and the stail variability of errors, of rainfall estimates. This study assess SREM2D simulations results for several rainfall estimates algorithms in the Tocantins-Araguaia river basin. Results show that the ensemble derived from the SREM2D model reduced bias, of the satellite precipitation estimation algorithms mainly for basin with drainage area higher than 12000 km2