Desenvolvimento de um modelo de previsão da seca com recurso à inteligência artificial: Instituto Nacional de Meteorologia (INAM)

Abstract

This research paper addresses the problem of meteorological droughts in Mozambique, focusing on the application of modern drought forecasting techniques and the use of the Standardized Precipitation Index (SPI) as an analysis methodology. Drought is a phenomenon that directly affects the lives of all living beings, limiting the availability of water and food, and is especially critical in the southern and central areas of the country, where the limited capacity for adequate climate monitoring exacerbates the impacts on communities and ecosystems. This research focuses on the difficulties faced by the National Meteorological Institute (INAM) in collecting comprehensive climate data, due to infrastructural limitations, and proposes the integration of Machine Learning technologies to improve forecast accuracy. To test this approach, an LSTM (Long Short-Term Memory) forecast model was developed using the SPI for intervals of 1, 3 and 6 months, with the aim of providing a clear view of the severity and duration of droughts. The practical implementation of these solutions faced certain challenges, such as the need for large volumes of historical data and the optimization of models to ensure a smaller margin of error in the results. However, the research provides a solid basis for modernizing climate forecasting practices in Mozambique.