Comparative insights into independent and hybrid modeling strategies for effective river water level prediction and management
| dc.contributor.author | Ibrahim, Osama Ragab | |
| dc.contributor.author | Vafaei, Aynaz | |
| dc.contributor.author | Ansari, Sanam | |
| dc.contributor.author | Abdi, Erfan | |
| dc.contributor.author | Sifaei, Maryam | |
| dc.contributor.author | Jafari Mohammadi, Seyed Mahdi | |
| dc.date.accessioned | 2026-01-27T13:57:31Z | |
| dc.date.available | 2026-01-27T13:57:31Z | |
| dc.date.issued | 2025-08-26 | |
| dc.department | Diğer | |
| dc.description.abstract | Accurate river water level prediction is essential for flood risk mitigation, efficient water allocation, and ecosystem protec- tion. It also enables assessment of climate change impacts on hydrological regimes, providing critical data for adaptation strategies. This study applies advanced forecasting models to two major rivers—the Mississippi and the Danube—charac- terized by distinct climatic and hydrological profiles, over 17 years (2008–2025). A rigorous data preprocessing protocol was implemented, involving outlier removal and imputation of missing values to ensure data integrity. For monthly water level prediction, two standalone models—deep neural network (DNN) and convolution neural network (CNN)—were used alongside two hybrid architectures: CNN with long short-term memory (CNN-LSTM) and CNN with bidirectional LSTM (CNN-BiLSTM). Input features included mean temperature, precipitation, relative humidity, and three lagged water level values. Model performance was evaluated using visual diagnostics and three statistical metrics: root mean square error (RMSE), Pearson’s correlation coefficient (r), and Nash–Sutcliffe efficiency (NSE). The CNN-BiLSTM model exhibited superior predictive capability. At the Danube station, it achieved an RMSE of 0.47 m, r of 0.94, and NSE of 0.88. At the Mississippi station, it yielded an RMSE of 0.93 m, r of 0.96, and NSE of 0.95. These results highlight the model’s robustness across diverse hydrological settings. The validation results show consistent performance and reliability. This modeling framework provides valuable support for water resources management, enabling informed planning and decision-making in these two basins. | |
| dc.identifier.citation | Ibrahim, O. R., Vafaei, A., Ansari, S., Abdi, E., Sifaei, M., & Jafari Mohammadi, S. M. (2025). Comparative insights into independent and hybrid modeling strategies for effective river water level prediction and management. Modeling Earth Systems and Environment, 11, 415. https://doi.org/10.1007/s40808-025-02611-9 | |
| dc.identifier.doi | 10.1007/s40808-025-02611-9 | |
| dc.identifier.scopusquality | Q3 | |
| dc.identifier.uri | https://doi.org/10.1007/s40808-025-02611-9 | |
| dc.identifier.uri | https://hdl.handle.net/20.500.12695/3955 | |
| dc.identifier.volume | 11 | |
| dc.identifier.wosquality | Q3 | |
| dc.indekslendigikaynak | Scopus | |
| dc.institutionauthor | Ibrahim, Osama Ragab | |
| dc.language.iso | en | |
| dc.publisher | Springer Nature Switzerland AG | |
| dc.relation.ispartof | Modeling Earth Systems and Environment | |
| dc.relation.publicationcategory | Gazete Makalesi - Uluslararası | |
| dc.rights | info:eu-repo/semantics/closedAccess | |
| dc.subject | Flood disaster · CNN-BiLSTM · Rivers water level · Analytical models · Hybrid model | |
| dc.title | Comparative insights into independent and hybrid modeling strategies for effective river water level prediction and management | |
| dc.title.alternative | Hybrid Deep Learning Models for River Water Level Prediction | |
| dc.type | Article |
