Técnicas de minera??o de dados para análise da precipita??o pluvial decenal no Rio Grande do Sul

摘要

The objective of this study was to analyze the spatio-temporal rainfall in Rio Grande do Sul, between 1987-1996 and 1997-2006 decades by using techniques of data mining. The historical series were acquired from the hydrological information Hidroweb system. The methodology used was based on the CRISP-DM (Cross Industry Standard Process for Data Mining) model. First, homogeneous precipitation areas for the 1987-1996 and 1997-2006 decades were defined. Subsequently, by the overlapping of the clusters obtained from both periods, six common areas were defined ranging from A to F. The changes in the volume of precipitation were evaluated annual seasonal and monthly. The results indicated significant increases (20 to 240 mm) in annual precipitation in all areas, except in area A. In the seasonal analysis, the variations were random, and in the spring all areas showed significant increase (44 to 142 mm). In the monthly analysis, the results revealed a reduction occurred in January in all areas, except in area E. In the other months, the changes were random. The results showed that among the decades there has been a change in the volume of rainfall in all timescales analyzed.