Exploring Deep Learning Techniques for Seasonal Prediction of Autumn Precipitation in South America

• Matheus Corrêa Domingos INPE
• Valdivino Alexandre de Santiago Júnior INPE
• Juliana Aparecida Anochi INPE
• Elcio Hideiti Shiguemori IEAv

DOI: https://doi.org/10.5753/wcama.2025.8543

Resumo

A previsão de precipitação sazonal é essencial para o planejamento climático e a redução de riscos de desastres naturais. Este estudo utilizou Convolutional Neural Network 1D (CNN 1D), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU) e Graph Convolutional Long Short-Term Memory (GConvLSTM) para prever a precipitação do outono de 2023. CNN 1D teve o melhor desempenho (MSE: 1.8827, Coeficiente de Correlação: 0.9137), destacando-se em precisão, mas perdendo detalhes espaciais. GConvLSTM capturou melhor os padrões espaciais, enquanto LSTM e GRU tiveram desempenho inferior. Apesar da superioridade da CNN 1D, a integração de componentes espaciais, como no GConvLSTM, poderia melhorar as previsões.

Referências

Anochi, J. A., de Almeida, V. A., and de Campos Velho, H. F. (2021). Machine learning for climate precipitation prediction modeling over south america. Remote Sensing, 13(13).

Anochi, J. A. and Shimizu, M. H. (2024). Precipitation forecasting and drought monitoring in south america using a machine learning approach. Meteorology, 4(1):1.

Chung, J., Gulcehre, C., Cho, K., and Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555.

CPTEC (1988). Eta model documentation. [link]. Accessed March 6, 2025.

DataCamp (2023). Machine learning and deep learning tutorial. Accessed on 06-Mar-2025.

Hersbach, H. et al. (2023). Era5 hourly data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). Data set, available online.

Holton, J. R. and Hakim, G. J. (2013). An introduction to dynamic meteorology, volume 88. Academic press.

INSTITUTO NACIONAL DE METEOROLOGIA (2024). Prognóstico climático de outono: Nota técnica conjunta inmet/inpe. INMET; INPE, Brasília, DF; São José dos Campos, SP. Disponível em: [link]. Accessed on: 20 Jan 2025.

Kiranyaz, S., Avcı, O., Abdeljaber, O., Ince, T., Gabbouj, M., and Inman, D. J. (2019). 1d convolutional neural networks and applications: A survey. ArXiv, abs/1905.03554.

Kirtman, B. P., Min, D., Infanti, J. M., Kinter III, J. L., Paolino, D. A., Zhang, Q., Van Den Dool, H., Saha, S., Mendez, M. P., Becker, E., et al. (2014). The north american multimodel ensemble: phase-1 seasonal-to-interannual prediction; phase-2 toward developing intraseasonal prediction. Bulletin of the American Meteorological Society, 95(4):585–601.

Monego, V. S., Anochi, J. A., and de Campos Velho, H. F. (2022). South america seasonal precipitation prediction by gradient-boosting machine-learning approach. Atmosphere, 13(2):243.

Reboita, M. S., Dias, C. G., Dutra, L. M. M., Rocha, R. P. d., and Llopart, M. (2018). Previsão climática sazonal para o brasil obtida através de modelos climáticos globais e regional. Revista Brasileira de Meteorologia, 33:207–224.

Seo, Y., Defferrard, M., Vandergheynst, P., and Bresson, X. (2016). Structured sequence modeling with graph convolutional recurrent networks.

Service, C. C. C. (2023). Era5 hourly data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). Accessed on 06-Mar-2025.

Staudemeyer, R. C. and Morris, E. R. (2019). Understanding lstm–a tutorial into long short-term memory recurrent neural networks. arXiv preprint arXiv:1909.09586.

Torres, F. L. R., Kuki, C. A. C., Reboita, M. S., Lima, L. M. M., Lima, J. W. M., and Queiroz, A. R. d. (2024). Refining seasonal precipitation forecast in brazil using simple data-driven techniques and climate indices. Revista Brasileira de Meteorologia, 39:e39240052.

Yang, J., Xiang, Y., Sun, J., and Xu, X. (2022). Multi-model ensemble prediction of summer precipitation in china based on machine learning algorithms. Atmosphere, 13(9).