Caracterização das diferenças entre precipitações estimadas por satélite e obtidas por pluviômetros

Abstract

The monitoring of precipitation brings essential subsidies for the planning and operation of various sectors of society. However, due to its intermittence, an accurate representation of its spatiotemporal variability is extremely difficult. Satellite precipitation estimates are an alternative to overcome this objection. However, these have uncertainties associated with their estimates. In this work, we evaluated the performance of three satellite products (TMPA 3B42, CMORPHv1.0-CRT and PERSIANN) using categorical indices (POD, FAR and fBIAS) and continuous metrics (MAE and KGE') on daily, monthly and annual time scales on the Paranaíba river basin. Subsequently, techniques were applied to correct the bias of these estimates in order to approximate them with in situ observations. The results show that the products have good ability to identify rainfall events, but poor accuracy in classifying their intensities, and there is not a single product that performs superiorly in all analyzes performed. However, on average, the TMPA product outperforms the others on all time scales. The modified Kling-Gupta efficiency coefficient (KGE') proved to be a useful evaluation index because it decomposes the error into linear correlation, bias and variability components. The bias correction techniques evaluated were able to correct the first moment of satellite data distributions. The quantile mapping by nonparametric distributions, in particular, was also able to improve the second moment of the distributions. However, the mean absolute error remained virtually unchanged. It is concluded that, despite the continuous evolution of satellite precipitation estimates, its use in hydrological studies should be preceded by a careful analysis, guided by the objectives and relative importance of introducing new sources of uncertainty to the analysis.