Aplicação de técnicas de mineração de dados para análise climática no estado do Tocantins

Abstract

Climatic changes are phenomena with a high rate of variability that, eventually, can cause extreme events, such as droughts or intense rains. These events can have an impact on socioeconomics, agriculture in a region and even the loss of human lives. Therefore, having knowledge about the frequencies, behavior and climate changes is extremely important to carry out a plan that seeks to minimize damage caused by the intense climate. Currently, climate information is being collected and stored in large databases. Knowledge extraction can be done through computational intelligence such as, for example, the Knowledge Discovery in Databases process, which has a set of data mining techniques capable of extracting useful patterns. Based on this context, this work aims to apply the Knowledge Discovery process to climatic data from the state of Tocantins. For this, historical time series from the National Institute of Meteorology was used. Initially, the grouping task was applied to define homogeneous climatic zones for the first twenty years (1980 to 1999) and for the second twenty years (2000 to 2019), both with the same number of zones. To analyze the frequency and climatic behavioral variations, a joint research between the clusters was adopted to analyze the variability and the individual research at different scales (annual, monthly and daily), carried out using the Quantis and Analysis Technique of Seasonality on the precipitation variable. The results showed that the techniques of data mining and statistical analysis are suitable for studies on behavior and climatic variations, reinforced that, through this research, the main decades were identified with record of extreme events in Tocantins.