Mid- to Long-Term Runoff Forecasting in the Huaihe River Basin Using a Spatiotemporal Graph Neural Network Coupled with LSTM
Main Article Content
Keywords
spatiotemporal graph neural network, Huaihe River Basin, mid- to long-term runoff forecasting, spatial connectivity, complex river-network basin
Abstract
To address the difficulty of simultaneously characterizing runoff temporal variation and spatial connectivity among stations under the complex river network conditions of the Huaihe River Basin, a spatiotemporal graph neural network model integrating temporal feature extraction and graph-based spatial modeling was developed for mid- to long-term runoff forecasting. Based on multi-station daily precipitation, air temperature, and runoff data from 2011 to 2020 in the Huaihe River Basin, data preprocessing was conducted, including missing-value imputation, outlier detection, standardization, and basin graph construction, and an adjacency matrix was established according to the upstream–downstream hydrological connectivity among stations. The model employed LSTM to extract dynamic features from historical sequences and graph convolution to capture spatial dependencies within the basin. Compared with LSTM, K-nearest neighbor, support vector regression, decision tree, and AdaBoost models, the proposed model achieved better overall predictive performance on the basin-wide test set, with an average R² of 0.915, average MAE of 30.019, and average RMSE of 50.243, demonstrating strong runoff process fitting ability and good error control at the basin scale. The results indicate that incorporating basin spatial connectivity can effectively improve the accuracy of mid- to long-term runoff forecasting and provide methodological support for water resources regulation and flood-control decision-making in complex river basins.
References
- [1] Chen, J., Xu, Q., & Cao, D. X. et al. (2024). A medium- to long-term runoff forecasting model based on multi-factor and multi-model ensemble. Advances in Water Science, 35(3), 408-419.
- [2] Sun, Z. L., Liu, Y. L., & Zhang, J. Y. et al. (2023). Research progress and prospects of medium- to long-term runoff forecasting. Water Resources Protection, 39(2), 1-11.
- [3] Su, H. D., Jia, Y. W., & Ni, G. H. et al. (2018). Application of machine learning in runoff forecasting. China Rural Water and Hydropower, (6), 40-43.
- [4] Hu, L. Y., Fu, X. L., & Jiang, X. L. et al. (2024). Runoff forecasting based on three machine learning methods: LSTM, RF, and SVR. Hydrology, 44(5), 17-24.
- [5] Xie, S., Huang, Y. F., & Li, T. J. et al. (2018). Medium- to long-term runoff forecasting using coupled LASSO regression and support vector regression. Journal of Basic Science and Engineering, 26(4), 709-722.
- [6] Li, L. J., Wang, Y. T., & Hu, Q. F. et al. (2020). Long-term reservoir runoff forecasting based on random forest and support vector machine. Journal of Water Resources and Water Engineering, (4), 33-40.
- [7] Zou, H. M., & Zhu, C. T. (2024). Comparative study on medium- to long-term reservoir inflow runoff prediction based on LSTM and BP neural network. Hydrology, 44(4), 27-31+37.
- [8] Zheng, K. F., Ma, X. H., & Cui, G. T. et al. (2026). Runoff forecasting in the Beiliu River based on SAO-optimized LSTM model. Water Resources and Hydropower Engineering (Chinese and English), 57(2), 83-94.
- [9] Kratzert, F., Klotz, D., Brenner, C., et al. (2018). Rainfall-runoff modelling using Long Short-Term Memory (LSTM) networks. Hydrology and Earth System Sciences, 22, 6005-6022.
- [10] Wang, M. Y., Wang, E. Z., & Luo, H. Q. et al. (2025). Graph neural network model for runoff forecasting in small and medium-sized basins: A case study of the Shaxi River Basin in Fujian. Journal of Hydroelectric Engineering, 44(6), 50-61.
- [11] Yang, S., Zhang, Y., & Zhang, Z. (2023). Runoff prediction based on dynamic spatiotemporal graph neural network. Water, 15(13), Article 2463.
- [12] Liu, Y., Hou, G., Huang, F., et al. (2022). Directed graph deep neural network for multi-step daily streamflow forecasting. Journal of Hydrology, 607, Article 127515.
- [13] Xue, L., & Zhu, Y. (2025). Dynamic hydrological flow prediction with self-iterative spatiotemporal graph neural network: Modeling long- and short-period topological dynamics. Journal of Hydrology, 663, Article 134122.
- [14] Yuan, R., Cai, S., Liao, W., et al. (2021). Daily runoff forecasting using ensemble empirical mode decomposition and long short-term memory. Frontiers in Earth Science, 9, Article 621780.
- [15] Xu, B., Yang, F. G., & Li, Y. J. (2020). Application of two ensemble learning algorithms in medium- to long-term runoff forecasting. Water Power, 46(4), 21-24+34.
- [16] He, F., Zhang, H., Wan, Q., et al. (2023). Medium term streamflow prediction based on Bayesian model averaging using multiple machine learning models. Water, 15(8), Article 1548.
- [17] Kaur, S., & Chavan, S. R. (2025). Comparative analysis of deep learning and machine learning models for one-day-ahead streamflow forecasting in the Krishna River basin. Journal of Hydrology: Regional Studies, 60, Article 102549.
Article Sidebar
This work is licensed under a Creative Commons Attribution 4.0 International License.