A novel hybrid dual-stream deep learning architecture integrating multi-state LSTM, CNN, and multi-head self-attention for water level and discharge prediction (Missouri River Basin, USA)
(2026) In Journal of Hydrology: Regional Studies 65.- Abstract
Study region: The Missouri River Basin, located between the states of Nebraska and Iowa, USA. Study focus: Water level (WL) and discharge (Q) are key hydrological variables, accurate prediction of which in both long-term and short-term (extreme events) scenarios is essential for water resources and flood risk management. We propose a novel hybrid deep learning architecture, CNN-NLSTM-SA, which integrates a Convolutional Neural Network (CNN) branch for extracting local features and a Multi-State LSTM (NLSTM) branch for capturing long-term temporal dependencies. NLSTM, with its child-parent structure, enhances memory propagation and mitigates vanishing and exploding gradients. The outputs of these two branches are fused through a... (More)
Study region: The Missouri River Basin, located between the states of Nebraska and Iowa, USA. Study focus: Water level (WL) and discharge (Q) are key hydrological variables, accurate prediction of which in both long-term and short-term (extreme events) scenarios is essential for water resources and flood risk management. We propose a novel hybrid deep learning architecture, CNN-NLSTM-SA, which integrates a Convolutional Neural Network (CNN) branch for extracting local features and a Multi-State LSTM (NLSTM) branch for capturing long-term temporal dependencies. NLSTM, with its child-parent structure, enhances memory propagation and mitigates vanishing and exploding gradients. The outputs of these two branches are fused through a multi-head Self-Attention (SA) mechanism, enabling the model to automatically emphasize the most informative representations. New hydrological insight: The proposed model is evaluated for both long-term and short-term forecasting scales. The long-term scenario leverages extensive historical data to provide a large set of training data, whereas the short-term scenario focuses on extreme events with limited training samples. To mimic real-world operational challenges in poorly gauged or data-scarce basins, the model is also tested under varying station-availability conditions using a Leave-n-Station-Out (LnSO) validation strategy. An ablation study comparing CNN-NLSTM-SA with several single- and dual-branch alternatives (CNN-LSTM-SA, LSTM-SA, CNN-SA, CNN-LSTM) shows the superior performance of the proposed architecture. Overall, CNN-NLSTM-SA demonstrates strong potential for WL and Q prediction in both data-rich and data-limited environments.
(Less)
- author
- Ahmadi Gharehtoragh, Milad ; Mehran, Ali ; Destouni, Georgia and Taheri Dehkordi, Alireza LU
- organization
- publishing date
- 2026-06
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Artificial Intelligence, Deep Learning, Environmental Modelling, Hydrology, Water Resources Management
- in
- Journal of Hydrology: Regional Studies
- volume
- 65
- article number
- 103523
- publisher
- Elsevier
- external identifiers
-
- scopus:105038423248
- ISSN
- 2214-5818
- DOI
- 10.1016/j.ejrh.2026.103523
- language
- English
- LU publication?
- yes
- additional info
- Publisher Copyright: © 2026 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license. http://creativecommons.org/licenses/by/4.0/
- id
- 27995b7e-7bbb-4b59-a570-a1a6ee886e3f
- date added to LUP
- 2026-07-14 15:10:34
- date last changed
- 2026-07-15 10:57:28
@article{27995b7e-7bbb-4b59-a570-a1a6ee886e3f,
abstract = {{<p>Study region: The Missouri River Basin, located between the states of Nebraska and Iowa, USA. Study focus: Water level (WL) and discharge (Q) are key hydrological variables, accurate prediction of which in both long-term and short-term (extreme events) scenarios is essential for water resources and flood risk management. We propose a novel hybrid deep learning architecture, CNN-NLSTM-SA, which integrates a Convolutional Neural Network (CNN) branch for extracting local features and a Multi-State LSTM (NLSTM) branch for capturing long-term temporal dependencies. NLSTM, with its child-parent structure, enhances memory propagation and mitigates vanishing and exploding gradients. The outputs of these two branches are fused through a multi-head Self-Attention (SA) mechanism, enabling the model to automatically emphasize the most informative representations. New hydrological insight: The proposed model is evaluated for both long-term and short-term forecasting scales. The long-term scenario leverages extensive historical data to provide a large set of training data, whereas the short-term scenario focuses on extreme events with limited training samples. To mimic real-world operational challenges in poorly gauged or data-scarce basins, the model is also tested under varying station-availability conditions using a Leave-n-Station-Out (LnSO) validation strategy. An ablation study comparing CNN-NLSTM-SA with several single- and dual-branch alternatives (CNN-LSTM-SA, LSTM-SA, CNN-SA, CNN-LSTM) shows the superior performance of the proposed architecture. Overall, CNN-NLSTM-SA demonstrates strong potential for WL and Q prediction in both data-rich and data-limited environments.</p>}},
author = {{Ahmadi Gharehtoragh, Milad and Mehran, Ali and Destouni, Georgia and Taheri Dehkordi, Alireza}},
issn = {{2214-5818}},
keywords = {{Artificial Intelligence; Deep Learning; Environmental Modelling; Hydrology; Water Resources Management}},
language = {{eng}},
publisher = {{Elsevier}},
series = {{Journal of Hydrology: Regional Studies}},
title = {{A novel hybrid dual-stream deep learning architecture integrating multi-state LSTM, CNN, and multi-head self-attention for water level and discharge prediction (Missouri River Basin, USA)}},
url = {{http://dx.doi.org/10.1016/j.ejrh.2026.103523}},
doi = {{10.1016/j.ejrh.2026.103523}},
volume = {{65}},
year = {{2026}},
}