Uso de parâmetros estatísticos para a classificação de regiões homogêneas de temperatura do ar em Santa Catarina

Abstract

Time series of hourly temperature from 146 weather stations located in Santa Catarina State – South Brazil were used to show that a compact data representation, using probability density functions (PDF) parameters, could be useful to classify homogeneous air temperature areas. The normal distribution fitted well the 146 weather stations temperature time series, presenting a median value of 0.9721 for the Pearson correlation coefficient. The means and standard deviations obtained by adjusting the Gaussian functions for the 146 stations were used as input parameters for two different classifiers: hierarchical and k-means. Both classifiers separated Santa Catarina's weather stations into four distinct groups. These groups had direct relationship with altitude ranges and with the influence of sea. The classification of weather stations in different homogeneous groups was useful to identify climatic behaviors of hourly temperatures. In addition to the characterization of the climate itself, this classification can be useful as a support for the validation of numerical weather forecast models, and for the identification of abnormal temperature time series in a regional spatial context.