Matheus Corrêa Domingos, a graduate student representing INPE, successfully presented his master's proposal titled "AI-based Approaches for Seasonal Precipitation Prediction Using Machine Learning and Deep Learning Techniques." The presentation tackled one of the major challenges in climate modeling: precipitation forecasting, which is complicated by the highly non-linear and chaotic nature of atmospheric processes involved in cloud formation and development.

Despite advances in Numerical Weather Prediction (NWP) models, which rely on differential equations and complex computational simulations, significant limitations remain in the accuracy of predictions, particularly in seasonal scales. In this context, the use of artificial intelligence (AI), especially machine learning (ML) and deep learning (DL), has shown great potential for precipitation forecasting. However, more robust investigations, considering different ML and DL approaches and utilizing diverse datasets, are still needed to determine the most effective approach for seasonal predictions.

The goal of this master's dissertation is to contribute to seasonal precipitation prediction via ML and DL. To achieve this, the research proposes a method that investigates ML approaches such as Extreme Gradient Boosting (XGBoost) and Random Forests, alongside DL models such as Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Bidirectional Long Short-Term Memory (BI-LSTM), One-Dimensional Convolutional Neural Network (CNN1D), Graph Convolutional Long Short-Term Memory (GConvLSTM), and Graph Convolutional Gated Recurrent Unit (GConvGRU). Historical reanalysis data from ERA5 and observed precipitation data from the Global Precipitation Climatology Project (GPCP) for the year 2023 were used to support the predictions.

As part of the experimentation, the ML and DL techniques were compared with numerical models such as the Brazilian Global Atmospheric Model (BAM) and North American Multi-Model Ensemble (NMME). Preliminary results showed that the ML and DL models outperformed the numerical weather and climate models, particularly during transitional seasons. These findings align with existing literature, which highlights the success of AI in weather and climate forecasting.

The defense was chaired by Dr. Alan James Peixoto Calheiros, with the guidance of Dr. Valdivino Alexandre de Santiago Júnior and Dr. Juliana Aparecida Anochi, all from INPE, and internal member Dr. Leonardo Bacelar Lima Santos from INPE. Their valuable feedback contributed significantly to the success of the defense.