The focus of the present work is the time and spatial variability of precipitation in eastern Amazon (Pará, Maranhão and Tocantins states) during 1982 to 2015 period. The application of the cluster analysis allowed the identification of four homogeneous regions of precipitation over region, with the rainy (CHU+) and dry (CHU-) periods occurring in different ways throughout the months of the year, featuring the pronounced climate variability of the region. The key-areas of the patterns of sea surface temperature (SST) over Pacific and Atlantic Oceans were selected through significant correlations, in order to build simple and multiple regression models to generate seasonal forecasts of precipitation in regions homogeneous during the CHU + and CHU- regimes. The results of the models indicated that the lag0 predictions (SST predictors in the same period of occurrence of CHU + and CHU-) are better than the lagged forecasts (SST predictors of the month, two-months and three-months preceding the CHU + and CHU-), regardless of the model is simple (only one TSM area) or multiple (two, three or four areas of SST).
|
