Análise e previsão de secas no norte de Minas Gerais utilizando machine learning

Abstract

The Brazilian semi-arid region is one of the areas most impacted by climate change. This study investigated drought events that occurred between January 1961 and July 2020 in 25 municipalities in the northern mesoregion of Minas Gerais, which is part of the Brazilian semiarid region, and implemented a drought prediction model for 5 selected municipalities using a Recurrent Neural Network (RNN). The Standardized Precipitation and Evapotranspiration Index (SPEI) on a 12-month scale (SPEI-12) was used to verify the duration, magnitude, intensity, and relative frequency of droughts, with daily precipitation and potential evapotranspiration data. The results showed an increase in the frequency of dry periods in all 25 municipalities analyzed since 1990. Additionally, land use and land cover analysis between 1990 and 2020 revealed significant changes, such as the growth of forestry (116.08%), agriculture and pasture (25.45%), urban areas and other non-vegetated areas (103.76%), and mining (250%). Concurrently, there was a reduction in forest formation (-14.01%), savanna (9.34%), and river and lake areas (-29.36%). These land use changes may have contributed to the observed increase in drought frequency. In the drought prediction analysis using a recurrent neural network, it was found that, although all 5 selected municipalities showed good prediction performances for SPEI-12, São João da Ponte stood out as the municipality with the best performance. The presentation of the best performance in drought forecasts represents a significant result, as it can serve as a useful tool for managers in implementing preventive strategies against the impacts of drought, including water conservation measures and more effective agricultural practices and proper land management. The study also showed that the model's effectiveness may vary according to the physical and climatic characteristics of the municipalities analyzed. These results can contribute to SDG 6 - "Clean Water and Sanitation," emphasizing the management and regulation of water resources to ensure water security in drought-prone regions through actions such as advanced planning for the development of contingency plans, reducing negative impacts on agricultural activities, energy generation, and other sectors dependent on water availability.