Kernel smoothing dos dados de chuva no Nordeste

Abstract

Northeastern Brazil has great climatic adversity, is considered a very complex region, attracting the interest of scholars from around the world. The rainfall over this region is considered by seasonal behave more intensely on three internal zones of the region in different periods of the year, lasting three months, besides suffering heavily influenced by the incidence of El Niño, La Niña and other phenomena acting on the basins of the tropical Pacific and Atlantic oceans. In this work the technique was applied computational mathematics-interpolation Kernel Smoothing the data of rain on northeastern Brazil collected in the period from 1904 to 1998, from 2283 conventional weather stations located in all states of the Northeast. The calculations were performed on the GPU developed "Cluster Neumann" Program Graduate in Applied Statistics and Biometry, Department of Statistics and Informatics UFRPE through software "kernel" written in C language and Cuda. This tool allowed to do the interpolation of more than 26 million measurements of rainfall over the entire Northeast, allowing generate maps of rainfall intensity over the entire region, and make estimates in areas of missing data, and calculate statistics for precipitation Northeast in general scope and seasonal. According to the interpolations made could be detected among the studied period, the driest years and wettest, the spatial distribution of rainfall in each month as well as the characteristic of rainfall in times of El Niño and La Niña.