Hybrid GNN and temporal encoder models for flood forecasting in the Vu Gia-Thu Bon River Basin

Abstract

Flood forecasting in river networks requires models that can capture both spatial dependencies across river basins and the temporal dynamics of hydrological processes. This study proposes a hybrid approach combining Graph Neural Networks (GNNs) with temporal sequence models, such as Long Short-Term Memory (LSTM) networks and Transformer encoders, to forecast water levels in the Vu Gia-Thu Bon (VGTB) River Basin, a highly flood-prone catchment in Central Vietnam. The models are trained using multi-source historical datasets, including water level observations, rainfall records, and reservoir release data, and are evaluated for their ability to predict water levels several hours in advance. Two model variants are developed: (1) a GNN-LSTM model, which integrates GNN spatial learning with LSTM for temporal encoding, and (2) a GNN-Transformer model, which leverages a Transformer encoder to effectively capture long-range temporal dependencies. Results show that the hybrid GNN-temporal models significantly outperform a conventional LSTM model, with the GNN-Transformer model achieving a 20-30% reduction in root-mean-square error (RMSE). The GNN-LSTM model also demonstrates notable improvements, highlighting the importance of incorporating river network connectivity into flood forecasting. The study demonstrates that the hybrid models are particularly effective during major flood events, as they better capture rapid water level rises and the spatial propagation of floods. The findings suggest that integrating graph-based spatial learning with advanced temporal encoders offers a promising direction for improving flood forecasting accuracy, thereby contributing to more effective flood risk management and water resources planning.

Keywords

flood prediction / graph neural networks / LSTM / spatiotemporal modeling / transformer / Vu Gia-Thu Bon River Basin / water level forecasting

References

[1] Akkala, A., Boubrahimi, S. F., Hamdi, S. M., Hosseinzadeh, P., & Nassar, A. (2025). Spatio-temporal graph neural networks for streamflow prediction in the Upper Colorado Basin. Hydrology, 12(3), 60.

[2] Brunner, G. (2018). Benchmarking of the HEC-RAS two-dimensional hydraulic modeling capabilities (pp. 1-137). US Army Corps Eng. Davis.

[3] Bui, B. N. Q., P, T. T., Kantoush, S., Saber, M., Binh, D. V., Sumi, T., & Bui, D. D. (2022). Application of rainfall-runoff inundation (RRI) model for flood mapping at Vu Gia Thu Bon river basin in Vietnam. Proceeding IC-ENSURES 2022 Conjunction with ISTECC 2022, 122-126.

[4] Castangia, M., Grajales, L., Aliberti, A., Rossi, C., Macii, A., Macii, E., & Patti, E. (2023). Transformer neural networks for interpretable flood forecasting. Environmental Modelling & Software, 160, 105581.

[5] Costa, G. E. M., Menezes Filho, F. C. M., Canales, F. A., Fava, M. C., Brandão, A. R. A., & De Paes, R. P. (2023). Assessment of time series models for mean discharge modeling and forecasting in a sub-basin of the Paranaíba River, Brazil. Hydrology, 10(11), 208.

[6] Defferrard, M., Bresson, X., & Vandergheynst, P. (2016). Convolutional neural networks on graphs with fast localized spectral filtering. Advances in Neural Information Processing Systems, 29, 3844-3852.

[7] Donnelly, J., Daneshkhah, A., & Abolfathi, S. (2024). “Physics-informed neural networks as surrogate models of hydrodynamic simulators. Science of the Total Environment, 912, 168814.

[8] Duong, V. N., & Gourbesville, P. (2016). Model uncertainty in flood modelling. Case study at Vu Gia Thu Bon catchment-Vietnam. Procedia Engineering, 154, 450-458.

[9] Godara, N., Bruland, O., & Alfredsen, K. (2024). Comparison of two hydrodynamic models for their rain-on-grid technique to simulate flash floods in steep catchment. Frontiers in Water, 6, 1384205.

[10] Hoa, T. D., & Nam, D. H. (2019). River information management and early flood release in response to climate change in Vietnam. Vietnam Journal of Science, Technology and Engineering, 61(1), 92-96.

[11] Jia, X., Zwart, J., Sadler, J., Appling, A., Oliver, S., Markstrom, S., Willard, J., Xu, S., Steinbach, M., Read, J., & Kumar, V. (2021). Physics-guided recurrent graph model for predicting flow and temperature in river networks. Proceedings of the 2021 SIAM International Conference on Data Mining (SDM), 612-620.

[12] Kazadi, A., Doss-Gollin, J., Sebastian, A., & Silva, A. (2024). FloodGNN-GRU: A spatio-temporal graph neural network for flood prediction. Environmental Data Science 3, e21.

[13] Kipf, T. N. (2016). Semi-supervised classification with graph convolutional networks. In Proceedings of the 5th international conference on learning representations (pp. 17). ICLR.

[14] Kratzert, F., Klotz, D., Brenner, C., Schulz, K., & Herrnegger, M. (2018). Rainfall-runoff modelling using long short-term memory (LSTM) networks. Hydrology and Earth System Sciences, 22(no. 11), 6005-6022.

[15] Loi, N. K., Liem, N. D., Tu, L. H., Hong, N. T., Truong, C. D., Tram, V., Nhat, T. T., Anh, T. N., & Jeong, J. (2019). Automated procedure of real-time flood forecasting in Vu Gia-Thu Bon river basin, Vietnam by integrating SWAT and HEC-RAS models. Journal of Water and Climate Change, 10(3), 535-545.

[16] Luu, D. V., Doan, T. N. C., & Vo, N. D. (2020). Artificial neural network approach to flood forecasting in the Vu Gia-ThuBon catchment, Vietnam. International Conference on Computational Collective Intelligence, 616-625.

[17] Nga, T. T., Cong, V. H., & Hung, L. (2019). Assessing the impacts of climate change and reservoir operation on saltwater intrusion in the Vu Gia-Thu Bon River Basin. International Conference on Asian and Pacific Coasts, 1207-1212.

[18] Oruche, R., & O'Donncha, F. (2023). attention-based domain adaptation forecasting of streamflow in data-sparse regions. In Presented at the workshop on tackling climate change with machine learning. ICLR.

[19] Sabeti, R., Kjeldsen, T. R., Stamataki, I., & Rahi, B. (2024). Optimisation of hardware setups for time-efficient HEC-RAS simulations. Cambridge Prism. Water, 2, e13.

[20] Shi, J., Stebliankin, V., Wang, Z., Wang, S., & Narasimhan, G. (2023). Graph transformer network for flood forecasting with heterogeneous covariates. I-GUIDE Forum. 11, https://docs.lib.purdue.edu/iguide/2023/presentations/11

[21] Sit, M., Demiray, B., & Demir, I. (2021). Short-term hourly streamflow prediction with graph convolutional GRU networks. In Tackling climate change with machine learning workshop at ICML 2021.

[22] Tadesse, K. B., & Dinka, M. O. (2017). Application of SARIMA model to forecasting monthly flows in Waterval River, South Africa. Journal of Water and Land Development, 35(1), 229-236.

[23] Taghizadeh, M., Zandsalimi, Z., Nabian, M. A., Shafiee-Jood, M., & Alemazkoor, N. (2025). Interpretable physics-informed graph neural networks for flood forecasting. Computer-Aided Civil and Infrastructure Engineering, 40, 2629-2649.

[24] To, T. N., Vu, H. C., & Le, H. (2022). Impacts of reservoir operation and urbanization on flood inundation in The Vu Gia Thu Bon Basin, Vietnam. Water Supply, 22(4), 4656-4675.

[25] Vizi, Z., Batki, B., Rátki, L., Szalánczi, S., Fehérváry, I., Kozák, P., & Kiss, T. (2023). Water level prediction using long short-term memory neural network model for a lowland river: A case study on the Tisza River, Central Europe. Environmental Sciences Europe, 35(1), 92.

[26] Vu, M. T., Vo, N. D., Gourbesville, P., Raghavan, S. V., & Liong, S.-Y. (2017). Hydro-meteorological drought assessment under climate change impact over the Vu Gia-Thu Bon river basin, Vietnam. Hydrological Sciences Journal, 62(10), 1654-1668.

[27] Wang, Z., Si, Y., & Chu, H. (2022). “Daily streamflow prediction and uncertainty using a long short-term memory (LSTM) network coupled with bootstrap. Water Resources Management, 36(12), 4575-4590.

[28] Wu, Z., Pan, S., Long, G., Jiang, J., & Zhang, C. (2019). Graph wavenet for deep spatial-temporal graph modeling. In Proceedings of the 28th international joint conference on artificial intelligence (pp. 1907-1913).

[29] Yan, L., Shan, D., Zhu, X., Zheng, L., Zhang, H., Li, Y., Li, J., Hang, T., & Feng, J. (2026). Streamflow prediction of spatio-temporal graph neural network with feature enhancement fusion. Symmetry, 18(2), 240.

[30] Yu, B., Yin, H., & Zhu, Z. (2017). Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting. In Proceedings of the twenty-seventh international joint conference on artificial intelligence (pp. 3634-3640).

[31] Zhang, J., Zheng, Y., & Qi, D. (2017). Deep spatio-temporal residual networks for citywide crowd flows prediction. Proceedings of the AAAI Conference on Artificial Intelligence, vol. 31.

[32] Zhang, Z., Tian, W., Lu, C., Liao, Z., & Yuan, Z. (2024). “Graph neural network-based surrogate modelling for real-time hydraulic prediction of urban drainage networks. Water Research, 263, 122142.

[33] Zhao, Y., Li, H., Zhou, H., Attar, H. R., Pfaff, T., & Li, N. (2024). A review of graph neural network applications in mechanics-related domains. Artificial Intelligence Review, 57(11), 315.

RIGHTS & PERMISSIONS

2026 The Author(s). River published by Wiley-VCH GmbH on behalf of China Institute of Water Resources and Hydropower Research (IWHR).