Hybrid neural network for daily streamflow projection on a sub-seasonal timescale

Everton Santos Castro; Carlos Eduardo Sousa Lima; Antônio Duarte Marcos Junior; Cleiton da Silva Silveira; Francisco das Chagas Vasconcelos Junior

doi:10.5327/Z2176-94782495

Autores

Everton Santos Castro Universidade Federal do Ceará (UFC) - Brasil https://orcid.org/0000-0003-2544-4874
Carlos Eduardo Sousa Lima Universidade Federal do Ceará (UFC) - Brasil https://orcid.org/0000-0003-2199-3382
Antônio Duarte Marcos Junior Universidade Federal do Ceará (UFC) - Brasil https://orcid.org/0000-0001-7015-1355
Cleiton da Silva Silveira Universidade Federal do Ceará (UFC) - Brasil https://orcid.org/0000-0003-3303-5157
Francisco das Chagas Vasconcelos Junior Fundação Cearense de Meteorologia e Recursos Hídricos (FUNCEME) - Brasil https://orcid.org/0000-0002-1558-8383

DOI:

https://doi.org/10.5327/Z2176-94782495

Palavras-chave:

Previsão de vazão, Rede Neural Artificial, Wavelets, Previsão subsazonal

Resumo

Previsões de variáveis hidrológicas são de extrema importância para a gestão dos recursos hídricos em regiões particularmente vulneráveis aos efeitos das mudanças climáticas e variabilidade climática. Nesse contexto, o presente trabalho objetivou o desenvolvimento de uma rede neural híbrida capaz de produzir previsões diárias de vazão em escala subsazonal. Essa rede neural híbrida consiste em uma Rede Neural Artificial com entradas pré-processadas pela transformada Wavelet (RNAW). A RNAW foi testada nas bacias hidrográficas dos reservatórios de Três Marias, Sobradinho e Retiro Baixo, todas localizadas na Bacia Hidrográfica do rio São Francisco (BHSF). Os resultados obtidos mostram que a RNAW apresentou alta precisão nas previsões de curto prazo (7 a 28 dias); entretanto, para previsões de longo prazo (35 e 42 dias), houve uma queda significativa no desempenho, especialmente durante os períodos de transição para a estação chuvosa e nos meses secos. A comparação entre as métricas de desempenho das previsões da RNAW e dos modelos operacionais do Operador Nacional do Sistema Elétrico (ONS) para as bacias de Três Marias, Sobradinho e Retiro Baixo mostrou que a RNAW superou todos esses modelos. Os resultados obtidos indicam que a RNAW é uma ferramenta promissora para lidar com os complexos e dinâmicos desafios da hidrologia, conferindo-lhe grande valor no processo de tomada de decisão sobre a gestão dos recursos hídricos.

Downloads

Não há dados estatísticos.

Referências

Agarwal, S.; Roy, P.; Choudhury, P.; Debbarma, N., 2022. Estudo comparativo sobre a previsão de vazão de rios usando GMNN e GMNN baseado em wavelet. Journal of Water and Climate Change, v. 13 (9), 3323-3337. https://doi.org/10.2166/wcc.2022.226.

Akujuobi, C.M., 2022. Wavelets and wavelet transform systems and their applications. Springer International Publishing. Berlin/Heidelberg, Germany.

Al-Juboori, A.M., 2023. Prediction of hydrological drought in semi-arid regions using a novel hybrid model. Water Resources Management, v. 37 (9), 3657-3669. https://doi.org/ 10.1007/s11269-023-03520-1.

Arsenault, R.; Martel, J.L.; Brunet, F.; Brissette, F; Mai, J., 2023. Continuous streamflow prediction in ungauged basins: long short-term memory neural networks clearly outperform traditional hydrological models. Hydrology and Earth System Sciences, v. 27 (1), 139-157. https://doi.org/10.5194/hess-27-139-2023.

Bellier, J.; Whitin, B., Scheuerer, M.; Brown, J.; Hamill, T.M., 2023. A multi-temporal-scale modulation mechanism for the postprocessing of precipitation ensemble forecasts: Benefits for streamflow forecasting. Journal of Hydrometeorology, v. 24 (4), 659-673. https://doi.org/10.1175/JHM-D-22-0119.1.

Benedetto, J.J.; Michael, W.F., 2021. Wavelets: mathematics and applications. CRC Press, [S.l.].

Bettencourt, P.; Fulgêncio, C.; Grade, M.; Alcobia, S.; Monteiro, J.P.; Oliveira, R.; Leitão, J.C.; Leitão, P.C.; Fernandes, P.A.; Sousa, S.; Brites, S.; Fernandes, J.; Simões, J.; Scarton, M.; Santiago, E.; Aguiar, R.; Giffoni, M.; Melo, F.; Paes, A.C., 2016. Plano de recursos hídricos da bacia hidrográfica do rio São Francisco. Recursos Hidricos, v. 37 (1), 73-80. https://doi.org/10.5894/rh37n1-cti3.

Bevacqua, E.; Zappa, G.; Lehner, F.; Zscheischler, J., 2022. Precipitation trends determine future occurrences of compound hot–dry events. Nature Climate Change, v. 12 (4), 350-355. https://doi.org/10.1038/s41558-022-01309-5.

Ekwueme, B.N., 2024. Deep neural network modeling of river discharge in a tropical humid watershed. Earth Science Informatics, v. 17 (2), 1161-1177. https://doi.org/10.1007/s12145-023-01219-w.

Ferreira, N.C.; Chou, S.C.; Dereczynski, C., 2023. Evaluation of subseasonal precipitation simulations for the Sao Francisco River Basin, Brazil. Climate, v. 11 (11), 213. https://doi.org/10.3390/cli11110213.

Freire, P.K.M.M.; Santos, C.A.G.; Silva, G.B.L., 2019. Analysis of the use of discrete wavelet transforms coupled with ANN for short-term streamflow forecasting. Applied Soft Computing, Elsevier, v. 80, 494-505. https://doi.org/10.1016/j.asoc.2019.04.024.

Höge, M.; Scheidegger, A.; Baity-Jesi, M.; Albert, C.; Fenicia, F., 2022. Improving hydrologic models for predictions and process understanding using neural ODEs. Hydrology and Earth System Sciences, v. 26 (19), 5085-5102. https://doi.org/10.5194/hess-26-5085-2022.

Hunt, K.M.; Matthews, G.R.; Pappenberger, F.; Prudhomme, C., 2022. Using a long short-term memory (LSTM) neural network to boost river streamflow forecasts over the western United States. Hydrology and Earth System Sciences, v. 26 (21), 5449-5472. https://doi.org/10.5194/hess-26-5449-2022.

Hurkmans R.T,; Van Den Hurk, B.A; Schmeits, M.; Wetterhall, F. Pechlivanidis, I.G., 2023. Seasonal streamflow forecasting for fresh water reservoir management in the Netherlands: An assessment of multiple prediction systems. Journal of hydrometeorology, v. 24 (7), 1275-1290. https://doi.org/10.1175/JHM-D-22-0107.1

Khazaeiathar, M.; Schmalz, B., 2025. Advancing Flood Forecasting With Wavelet‐LSTM: The Role of Nonlinearity in Discharge Prediction. Journal of Flood Risk Management, v. 18 (4), e 70148. https://doi.org/10.1111/jfr3.70148.

Li, W.; Pan, B.; Xia, J.; Duan, Q., 2022. Convolutional neural network-based statistical post-processing of ensemble precipitation forecasts. Journal of Hydrology, v. 605, 127301. https://doi.org/10.1016/j.jhydrol.2021.127301.

Li, X.; Xu, W.; Ren, M.; Jiang, Y.; Fu, G., 2022. Hybrid CNN-LSTM models for river flow prediction. Water Supply, v. 22 (5), 4902-4919. https://doi.org/10.2166/ws.2022.170.

Maceira, M.; Damázio, J.; Ghirardi, A.; Dantas, H., 1999. Periodic arma models applied to weekly streamflow forecasts. In: IEEE. Power Tech Budapest 99. Abstract Records. (Cat. No. 99EX376). [S.l.], p. 86.

Nearing, G.; Cohen, D.; Dube, V.; Gauch, M.; Gilon, O.; Harrigan, S.; Matias, Y., 2024. Global prediction of extreme floods in ungauged watersheds. Nature, v. 627 (8004), 559-563. https://doi.org/10.1038/s41586-024-07145-1.

Operador Nacional do Sistema Elétrico (ONS), 2015. ONS Relatório Anual de Avaliação das Previsões de Vazões – 2015. [S.l.], ONS.

Ougahi, J.H; Rowan, J.S., 2025. Enhanced streamflow forecasting using hybrid modelling integrating glacio-hydrological outputs, deep learning and wavelet transformation. Scientific Reports, v. 15 (1), 2762. https://doi.org/10.1038/s41598-025-87187-1.

Paiva, L.F.G.; Acioli, G.C.L., 2007. Previsão de vazões com o modelo cpins – cálculo e previsão de vazões naturais e incrementais a sobradinho. In: Anais do I Workshop de previsão de vazões. ONS, Rio de Janeiro.

Quedi, E.; Fan, F.; Siqueira, V.; Collischonn, W.; Paiva, R.; Petry, I.; Gama, C.; Silveira, R.; Paranhos, C.; Freitas, C., 2024. Sub-seasonal streamflow forecasts for hydropower dams in the Brazilian Eletrical Interconnected System. RBRH, v. 29, e7. https://doi.org/10.1590/2318-0331.292420230109.

Ribeiro, S.C., 2017. Caracterização geoambiental da sub-bacia do rio salgado na mesorregião sul cearense–parte i–clima e arcabouço geológico. Geoconexões, v. 1, 4-16. https://doi.org/10.15628/geoconexoes.2017.6290.

Santana, A.S.; Santos, G.R., 2020. Impactos da seca de 2012-2017 na região semiárida do Nordeste: notas sobre a abordagem de dados quantitativos e conclusões qualitativas. Boletim Regional, Urbano e Ambiental (BRUA), v. 22 (22), 119-129. https://doi.org/10.38116/brua22art9.

Santos, A.L.L., 2022. Previsão de vazão afluente da UHE-Tucuruí por redes neurais recorrentes LSTM. Orientador: Raphael Barros Teixeira. 2022. [9], 44 f. Undergraduate Thesis (Bachelor’s Degree in Electrical Engineering) – Faculdade de Engenharia Elétrica, Campus Universitário de Tucuruí, Universidade Federal do Pará, Tucuruí. Retrieved 2025-10-20, from https://bdm.ufpa.br/handle/prefix/4578.

Saraiva, S.V.; Carvalho, F.O.; Santos, C.A.G.; Barreto, L.C.; Freire, P.K.M.M., 2021. Daily streamflow forecasting in Sobradinho Reservoir using machine learning models coupled with wavelet transform and bootstrapping. Applied soft computing, v. 102 (107081), 107081. https://doi.org/10.1016/j.asoc.2021.107081.

Sharma, P.; Singh, S.; Sharma, S.D., 2022. Artificial neural network approach for hydrologic river flow time series forecasting. Agricultural Research, v. 11 (3), 465-476. https://doi.org/10.1007/s40003-021-00585-5.

Sherstinsky, A., 2020. Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. Physica D: Nonlinear Phenomena, v. 404, 132306. https://doi.org/10.1016/j.physd.2019.132306.

Silva, M.V.M.; Silveira, C.S.; Costa, J.M.F.; Martins, E.S.P.R; Júnior, F.C.V, 2021. Projection of climate change and consumptive demands projections impacts on hydropower generation in the São Francisco River Basin, Brazil. Water, v. 13 (3), 332. https://doi.org/10.3390/w13030332.

Silva, M.V.M; Silveira, C.S; Costa, J.M.F; Martins, E.S.P.R; Júnior, F.C.V., 2020. Projections of climate change in streamflow and affluent natural energy in the Brazilian hydroelectric sector of CORDEX models. RBRH, v. 25. https://doi.org/10.1590/2318-0331.252020200020.

Souza, T.S.D.; Nascimento, P.D.S., 2022. Sequential climatic water balance applied in the hydrographic region of Paraguaçu, in Brazil. Sustainable Water Resources Management, v. 8 (1), 18. https://doi.org/10.1007/s40899-022-00608-1.

Tian, B.; Kong, X.; Wu, G., 2024. The application of the LSTM neural networks on the hydrology forecast. In: Proceedings of the 2024 6th International Conference on Pattern Recognition and Intelligent Systems. pp. 93-97. https://doi.org/10.1145/3689218.3689233.

Tuğrul, T.; Hinis M.A., 2025. Performance enhancement of models through discrete wavelet transform for streamflow forecasting in Çarşamba River, Türkiye. Journal of Water and Climate Change, v. 16 (2), 736-756. https://doi.org/10.2166/wcc.2025.709.

Wilks, D.S., 2011. Statistical methods in the atmospheric sciences. Academic Press, [S.l.]. v. 100.

Yilmaz, M., Tosunoğlu, F., Kaplan, N.H., Üneş, F., Hanay, Y.S., 2022. Predicting monthly streamflow using artificial neural networks and wavelet neural networks models. Modeling Earth Systems and Environment, v. 8 (4), 5547-5563. https://doi.org/10.1007/s40808-022-01403-9.

Zanial, W.N.C.W., Malek, M.B.A., Reba, M.N.M., Zaini, N., Ahmed, A.N., Sherif, M.; Elshafie, A., 2023. River flow prediction based on improved machine learning method: Cuckoo Search-Artificial Neural Network. Applied Water Science, v. 13 (1), 28. https://doi.org/10.1007/s13201-022-01830-0.

Zhao, X.; Wang, H.; Bai, M.; Xu, Y.; Dong, S.; Rao, H.; Ming, W., 2024. A comprehensive review of methods for hydrological forecasting based on deep learning. Water, v. 16 (10), 1407. https://doi.org/10.3390/w16101407.