A Multi-Tier Hybrid Machine and Deep Learning Ensemble Framework for Water-Level Prediction: A Case Study of the Indus Basin River

Aqsa Saleem; Hafiza Mamona Nazir; Muhammad Zubair

doi:10.64060/JASRv2i32

Authors

Aqsa Saleem Department of Statistics, University of Sargodha, Sargodha 40100, Pakistan Author https://orcid.org/0009-0001-2591-9000
Hafiza Mamona Nazir Department of Statistics, University of Sargodha, Sargodha 40100, Pakistan Author https://orcid.org/0000-0002-0998-6369
Muhammad Zubair Department of Statistics, University of Sargodha, Sargodha 40100, Pakistan Author

DOI:

https://doi.org/10.64060/JASRv2i32

Keywords:

Flood forecasting, River level prediction, Indus Basin, Hybrid machine learning, Deep learning, Ensemble learning, Hydrological modeling, Water resources management

Abstract

Accurate river water-level forecasting supports flood risk mitigation and water-resource management in the Indus Basin System, despite increased hydrological variability. This study proposes a hybrid multi-tier ensemble framework for daily river-level forecasting that incorporates 13 heterogeneous machine learning (ML) and deep learning (DL) models from the tree-based, kernel-based, neural network, and deep learning families, which are combined using Simple Averaging, Inverse-RMSE, Adaptive Inverse-RMSE weighting, and stacking-based strategies. Lag-based feature engineering and chronological TimeSeriesSplit validation maintained temporal relationships while preventing information leakage. Support Vector Regression (SVR) and Extra Trees (ET) outperformed ML models, while GRU and LSTM dominated DL architectures. The proposed hybrid ensemble outperformed the persistence benchmark, with an RMSE of 1.7621 m and R^2 of 0.3012. Diebold-Mariano statistical testing verified the comparative forecasting ability with a rigorous statistical significance assessment. Furthermore, the findings show that predicting accuracy increased only up to an ideal ensemble size, after which additional models yielded declining benefits. Overall, the findings show that model diversity and complementary learning characteristics have a greater impact on forecasting performance than increasing ensemble size or using increasingly complex weighting strategies, resulting in a strong framework for operational river water-level forecasting in data-scarce hydrological environments.

References

Abdoulhalik, A., & Ahmed, A. A. (2024). A comparative analysis of advanced machine learning techniques for river streamflow time-series forecasting. Sustainability, 16(10), 4005. https://doi.org/10.3390/su16104005

Agaj, T., Budka, A., Janicka, E., & Bytyqi, V. (2024). Using ARIMA and ETS models for forecasting water level changes for sustainable environmental management. Scientific Reports, 14(1), 22444. https://doi.org/10.1038/s41598-024-73405-9

Ahmad, S. (1994). Irrigation and water management in the Indus Basin of Pakistan: A country report. In Irrigation performance and evaluation for sustainable agricultural development (pp. 190–211).

Al-Mansori, N. J. H., Al-Zubaidi, L. S. A., & Ashoor, A. Z. L. S. (2020). One-dimensional hydrodynamic modeling of the Euphrates River and prediction of hydraulic parameters. Civil Engineering Journal, 6(6), 1074–1090. https://doi.org/10.28991/cej-2020-03091532

Al-Sulttani, A. O., Al-Mukhtar, M., Roomi, A. B., Farooque, A. A., Khedher, K. M., & Yaseen, Z. M. (2021). Proposition of new ensemble data-intelligence models for surface water quality prediction. IEEE Access, 9, 108527–108541. https://doi.org/10.1109/ACCESS.2021.3102118

Aricò, C., Filianoti, P., Sinagra, M., & Tucciarelli, T. (2016). The FLO diffusive 1D–2D model for simulation of river flooding. Water, 8(5), 200. https://doi.org/10.3390/w8050200

Asefa, T., Kemblowski, M., McKee, M., & Khalil, A. (2006). Multi-time scale stream flow predictions: The support vector machines approach. Journal of Hydrology, 318(1–4), 7–16. https://doi.org/10.1016/j.jhydrol.2005.06.001

Basharat, M., & Rizvi, S. (2016). Irrigation and drainage efforts in Indus Basin: A review of past, present and future requirements.

Bengio, Y., Simard, P., & Frasconi, P. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on Neural Networks, 5(2), 157–166. https://doi.org/10.1109/72.279181

Chowdhury, M. A. H., Goni, M. O., Gazi, M. U., Ahmed, M. S., Jubayer, M. F., Debnath, P., & Sarker, M. A. R. (2025). Meta-learning Model for Low-Data River Water Quality Analysis in The Northeastern Region of Bangladesh.

Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv. https://doi.org/10.48550/arXiv.1412.3555

Diebold, F. X., & Mariano, R. S. (1995). Comparing predictive accuracy. Journal of Business & Economic Statistics, 13(3), 253–263. https://doi.org/10.1080/07350015.1995.10524599

Ebtehaj, I., & Bonakdari, H. (2022). A reliable hybrid outlier robust non-tuned rapid machine learning model for multi-step ahead flood forecasting in Quebec, Canada. Journal of Hydrology, 614, 128592. https://doi.org/10.1016/j.jhydrol.2022.128592

Elhanafy, H., & Copeland, G. J. M. (2007). Flash floods simulation using Saint Venant equations. Proceedings of the International Conference on Aerospace Sciences and Aviation Technology, 12, 1–14.

Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. The Annals of Statistics, 29(5), 1189–1232. https://doi.org/10.1214/aos/1013203451

Ghimire, G. R., & Krajewski, W. F. (2020). Exploring persistence in streamflow forecasting. JAWRA Journal of the American Water Resources Association, 56(3), 542–550. https://doi.org/10.1111/1752-1688.12821

Kader, M. Y. A., Badé, R., & Saley, B. (2020). Study of the 1D Saint-Venant equations and application to the simulation of a flood problem. Journal of Applied Mathematics and Physics, 8(7), 1193–1206. https://doi.org/10.4236/jamp.2020.87100

Kedam, N., Tiwari, D. K., Kumar, V., Khedher, K. M., & Salem, M. A. (2024). River stream flow prediction through advanced machine learning models for enhanced accuracy. Results in Engineering, 22, 102215. https://doi.org/10.1016/j.rineng.2024.102215

Khan, U., Jamshed, R., Tahir, A. A., et al. (2025). Anticipating future hydrological changes in the northern river basins of Pakistan: Insights from the snowmelt runoff model and an improved snow cover data. Water, 17(14), 2104. https://doi.org/10.3390/w17142104

Khozani, Z. S., Precht, E., & Ionita, M. (2025). Weekly streamflow forecasting of the Rhine River based on machine learning approaches. Natural Hazards, 121(4), 4135–4153. https://doi.org/10.1007/s11069-024-06962-x

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(11), 6005–6022. https://doi.org/10.5194/hess-22-6005-2018

Kratzert, F., Klotz, D., Shalev, G., Klambauer, G., Hochreiter, S., & Nearing, G. S. (2019). Towards learning universal, regional, and local hydrological behaviors via machine learning applied to large-sample datasets. Hydrology and Earth System Sciences, 23(12), 5089–5110. https://doi.org/10.5194/hess-23-5089-2019

Kumar, V., Kedam, N., Sharma, K. V., Mehta, D. J., & Caloiero, T. (2023). Advanced machine learning techniques to improve hydrological prediction: A comparative analysis of streamflow prediction models. Water, 15(14), 2572. https://doi.org/10.3390/w15142572

Li, Y., Liang, Z., Hu, Y., Li, B., Xu, B., & Wang, D. (2020). A multi-model integration method for monthly streamflow prediction: Modified stacking ensemble strategy. Journal of Hydroinformatics, 22(2), 310–326. https://doi.org/10.2166/hydro.2019.092

Lutz, A. F., Immerzeel, W. W., Kraaijenbrink, P. D. A., Shrestha, A. B., & Bierkens, M. F. P. (2016). Climate change impacts on the upper Indus hydrology: Sources, shifts, and extremes. PLOS ONE, 11(11), e0165630. https://doi.org/10.1371/journal.pone.0165630

Parasar, P., Moral, P., Srivastava, A., Krishna, A. P., Majumdar, S., Bhattacharjee, R., Mishra, A. P., Mustafi, D., Rathore, V. S., Sharma, R., & Mustafi, A. (2025). Integrating genetic algorithms and machine learning for spatiotemporal groundwater potential zoning in fractured aquifers. Journal of Hydrology: Regional Studies, 62, 102800. https://doi.org/10.1016/j.ejrh.2025.102800

Shamseldin, A. Y. (1997). Application of a neural network technique to rainfall–runoff modelling. Journal of Hydrology, 199(3–4), 272–294. https://doi.org/10.1016/S0022-1694(96)03330-6

Sun, X., Zhang, H., Wang, J., Shi, C., Hua, D., & Li, J. (2022). Ensemble streamflow forecasting based on variational mode decomposition and long short-term memory. Scientific Reports, 12(1), 518. https://doi.org/10.1038/s41598-021-03725-7

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. https://doi.org/10.1515/jwld-2017-0088

Workneh, H., & Jha, M. (2025). Utilizing hybrid deep learning models for streamflow prediction. Water, 17(13), 1913. https://doi.org/10.3390/w17131913

Xu, Y., Hu, C., Wu, Q., Li, Z., Jian, S., & Chen, Y. (2021). Application of temporal convolutional network for flood forecasting. Hydrology Research, 52(6), 1455–1468. https://doi.org/10.2166/nh.2021.050