Resumo:
In Brazil, the number of rainfall stations on the surface is still low and is not enough to
represent the spatio-temporal variability of rainfall throughout the territory. Therefore,
a good option to overcome this lack of data is the use of satellite precipitation
estimates. Therefore, this study aimed to validate, through statistical methods, the daily
precipitation estimated by the Global Precipitation Measurement (GPM) satellite for the
Sapucaí River Basin, using four methodologies: point by pixel - original data, without
going through no processing; pixel to point – debiased satellite data; pixel by pixel –
interpolated station data; average in the area – average in the sub-areas of Alto, Medio
and Baixo Sapucaí. The data used in the validation came from rainfall stations of the
National Agency for Water and Basic Sanitation (ANA). The results found for the point to-pixel methodology for the complete series, summer and winter, indicated the
existence of a moderate average correlation, with errors varying from magnitude of 3
to 10 mm day-1 and with an average tendency to overestimate the GPM in in relation
to ANA data. The same occurred for the pixel-to-point method, where the results were
very similar, in addition, a correction in the precipitation frequency was observed
through this methodology. The pixel-by-pixel method presented higher statistical
results, compared to the previous ones, and even increased the surface data
coverage. However, the averaging methodology in the area was the one that obtained
the best results: high correlation, coefficient of determination above 60% and lower
magnitude of errors. Overall, in both cases the GPM showed an average tendency to
overestimate the ANA data, but it proved to be viable for use in studies, mainly as a
way to meet the need for data in places without records.