Sistema informatizado de mineração para obtenção de dados referentes a ocorrências de tempestades de granizo na região sul e estado de São Paulo

Abstract

Hailstorms can cause damage in various regions of the world. The southern region of Brazil is identified as one of the areas with the highest incidence of destructive hail. The damage caused by destructive hail is not limited to rural areas. It also affects urban areas, with vehicles being one of the main items damaged. Some hailstorms are so disastrous that they force municipalities to declare a state of emergency or public calamity. The historical relationship between hail damage and climate change is mentioned by many authors, with the tri-border region between Paraná, Santa Catarina, and Argentina identified as a location of high activity related to hail events and potentially favorable atmosphere for storm formation. In addition, some projections indicate a possible worsening of the damage caused by hailstorms. Prior knowledge of these events can help prepare for possible emergencies, but their detection is a very complex task. The advancement of technology has allowed for the creation/improvement of various electronic devices, such as sensors, new types of data storage, and the improvement of processing equipment, resulting in an increase in the production of various types of data, including meteorological data. Along with this, data mining techniques and algorithms for finding frequent patterns in a data set can be essential tools for improving the analysis of hailstorms and other extreme weather events. With the aim of analyzing rules of associations of meteorological events, this study used pattern mining techniques on data that were collected in an automated manner on three different platforms: REDEMET, IPMET/Unesp, and S2ID. The method used to perform automated collection was using web scraping techniques. Subsequently, the collected data was processed and saved in a MongoDB database. After a significant amount of centralized data from the three selected platforms, this data was prepared for use with the FP-Growth algorithm, responsible for extracting frequent patterns in data sets, resulting in the creation of association rule tables. At the end of the study, a dashboard was developed to present the collected data and the association rules between extreme weather events and hail. After analyzing the rules and based on the results obtained, it was found that centralizing the data is essential to verify the association of hail with other atmospheric events. Thus, the developed database can help in predicting future hail events using artificial intelligence models.