Deep Learning for Urban Flood Prediction: An LSTM Model Integrating Satellite Reanalysis and Historical Weather Data in Curitiba
Resumo
Urban floods pose increasing threats, causing significant human and economic losses. This research focuses on improving flood prevention using deep learning techniques together with satellite reanalysis data. We present the results of the implementation of an LSTM Neural Network with satellite reanalysis data to predict floods in the city of Curitiba, Brazil. To assess the benefits of the proposed model, we compare the results with datasets from previous related models. The results of each trained model are analyzed based on Receiver Operating Characteristic (ROC) curves. The results show the positive impact of the use of satellite data with respect to the area under the curve (AUC) metric.
Palavras-chave:
Urban flood prediction, LSTM, satellite reanalysis, deep learning, Curitiba, ERA5, ICARUS, machine learning, recurrent neural networks
Referências
Al-Rawas, G., Nikoo, M. R., Al-Wardy, M., and Etri, T. (2024). A critical review of emerging technologies for flash flood prediction: examining artificial intelligence, machine learning, internet of things, cloud computing, and robotics techniques. Water, 16(14):2069.
Batalini de Macedo, M., Mangukiya, N. K., Fava, M. C., Sharma, A., Fray da Silva, R., Agarwal, A., Razzolini, M. T., Mendiondo, E. M., Goel, N. K., Kurian, M., et al. (2024). Performance analysis of physically-based (hec-ras, caddies) and ai-based (lstm) flood models for two case studies. Proceedings of IAHS, 386:41–46.
CEMADEN (2024). Mapa Interativo — mapainterativo.cemaden.gov.br. [link]. Accessed: 2024-12-16.
de Sousa Araújo, A., Silva, A. R., and Zárate, L. E. (2022). Extreme precipitation prediction based on neural network model–a case study for southeastern brazil. Journal of Hydrology, 606:127454.
Fang, Z., Wang, Y., Peng, L., and Hong, H. (2021). Predicting flood susceptibility using lstm neural networks. Journal of Hydrology, 594:125734.
Fernandez, H. G. and Splendore, P. R. (2021). Sistema de Identificação Automática de Riscos Hidrometeorológicos com Retroalimentação e Reestruturação Autônoma da Infraestrutura de Comunicação.
Géron, A. (2022). Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow. ” O’Reilly Media, Inc.”.
Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., et al. (2020). The era5 global reanalysis. Quarterly journal of the royal meteorological society, 146(730):1999–2049.
Hochreiter, S. (1997). Long short-term memory. Neural Computation MIT-Press.
IPPUC (2022). Registros alagamentos. Accessed: 2022-11-07.
Le, X.-H., Ho, H. V., Lee, G., and Jung, S. (2019). Application of long short-term memory (lstm) neural network for flood forecasting. Water, 11(7):1387.
Maciel, E. (2025). Brazil faces huge surge in climate disasters amid poor prevention funding. [link]. Accessed: 2025-06-07.
Noboa, C. S., Pigatto, D., Buffon, E. M., and Gomes-Jr, L. (2024). Data analytics for a changing climate: Feature engineering for the forecast of hydrometeorological events. In Simpósio Brasileiro de Banco de Dados (SBBD), pages 715–721. SBC.
SEDEC (2024). Classificação e Codificação Brasileira de Desastres (COBRADE). [link]. Accessed: 2025-06-07.
USGS. What are the two main types of floods? [link]. Accessed: 2025-06-07.
Batalini de Macedo, M., Mangukiya, N. K., Fava, M. C., Sharma, A., Fray da Silva, R., Agarwal, A., Razzolini, M. T., Mendiondo, E. M., Goel, N. K., Kurian, M., et al. (2024). Performance analysis of physically-based (hec-ras, caddies) and ai-based (lstm) flood models for two case studies. Proceedings of IAHS, 386:41–46.
CEMADEN (2024). Mapa Interativo — mapainterativo.cemaden.gov.br. [link]. Accessed: 2024-12-16.
de Sousa Araújo, A., Silva, A. R., and Zárate, L. E. (2022). Extreme precipitation prediction based on neural network model–a case study for southeastern brazil. Journal of Hydrology, 606:127454.
Fang, Z., Wang, Y., Peng, L., and Hong, H. (2021). Predicting flood susceptibility using lstm neural networks. Journal of Hydrology, 594:125734.
Fernandez, H. G. and Splendore, P. R. (2021). Sistema de Identificação Automática de Riscos Hidrometeorológicos com Retroalimentação e Reestruturação Autônoma da Infraestrutura de Comunicação.
Géron, A. (2022). Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow. ” O’Reilly Media, Inc.”.
Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., et al. (2020). The era5 global reanalysis. Quarterly journal of the royal meteorological society, 146(730):1999–2049.
Hochreiter, S. (1997). Long short-term memory. Neural Computation MIT-Press.
IPPUC (2022). Registros alagamentos. Accessed: 2022-11-07.
Le, X.-H., Ho, H. V., Lee, G., and Jung, S. (2019). Application of long short-term memory (lstm) neural network for flood forecasting. Water, 11(7):1387.
Maciel, E. (2025). Brazil faces huge surge in climate disasters amid poor prevention funding. [link]. Accessed: 2025-06-07.
Noboa, C. S., Pigatto, D., Buffon, E. M., and Gomes-Jr, L. (2024). Data analytics for a changing climate: Feature engineering for the forecast of hydrometeorological events. In Simpósio Brasileiro de Banco de Dados (SBBD), pages 715–721. SBC.
SEDEC (2024). Classificação e Codificação Brasileira de Desastres (COBRADE). [link]. Accessed: 2025-06-07.
USGS. What are the two main types of floods? [link]. Accessed: 2025-06-07.
Publicado
29/09/2025
Como Citar
DOS SANTOS, Lucas Iuri; GOMES-JR, Luiz.
Deep Learning for Urban Flood Prediction: An LSTM Model Integrating Satellite Reanalysis and Historical Weather Data in Curitiba. In: SIMPÓSIO BRASILEIRO DE BANCO DE DADOS (SBBD), 40. , 2025, Fortaleza/CE.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2025
.
p. 984-990.
ISSN 2763-8979.
DOI: https://doi.org/10.5753/sbbd.2025.247630.
