Achite et al. (2025) Performance enhancement of daily reservoir evaporation rate estimation models using stacking regression by discretization with AI methods

Identification

• Journal: Theoretical and Applied Climatology
• Year: 2025
• Date: 2025-09-16
• Authors: Mohammed Achite, Okan Mert Katipoğlu, Ahmed Elbeltagi, Nehal Elshaboury, Kusum Pandey, Somayeh Emami, Chen Yongguo
• DOI: 10.1007/s00704-025-05720-8

Research Groups

• Faculty of Nature and Life Sciences, Water and Environment Laboratory, Hassiba Benbouali University of Chlef, Algeria
• Georessources Environment and Natural Risks Laboratory, University of Oran 2 Mohamed Ben Ahmed, Algeria
• Department of Civil Engineering, Faculty of Engineering, Erzincan Binali Yıldırım University, Turkey
• Agricultural Engineering Department, Faculty of Agriculture, Mansoura University, Egypt
• Construction and Project Management Research Institute, Housing and Building National Research Centre, Egypt
• G. B. Pant National Institute of Himalayan Environment, Garhwal Regional Centre, Srinagar, India
• Department of Water Engineering, University of Tabriz, Iran
• School of Civil Engineering and Architecture, Zhejiang University of Science and Technology, China

Short Summary

This study developed an advanced machine learning framework based on Regression by Discretization (RD) and ensemble methods to accurately predict daily reservoir evaporation rates at the Sidi-M’Hamed Ben Aouda Dam Basin in Algeria. The RD-Bagging model demonstrated superior performance with high predictive accuracy and low bias, making it a reliable tool for water resource management.

Objective

• To develop a novel framework for daily reservoir evaporation prediction by applying Regression by Discretization (RD) as a standalone model and in combination with ensemble learners (Random Committee, Random Subspace, Bagging, and Random Forest).

Study Configuration

• Spatial Scale: Sidi-M’Hamed Ben Aouda Dam Basin, Wadi Mina Basin, northwest Algeria (geographically defined by 34°41’57” N to 35°35’27” N and 00°22’59” E to 01°09’02” E, covering an area of 4,900 km²).
• Temporal Scale: Daily data from January 2010 to December 2023.

Methodology and Data

• Models used: Regression by Discretization (RD), RD-Random Committee, RD-Bagging, RD-Random Subspace, RD-Random Forest.
• Data sources: Key meteorological variables (maximum temperature, minimum temperature, mean temperature, dew point temperature, relative humidity, solar radiation, surface pressure, precipitation, and wind speed) downloaded from the National Aeronautics and Space Administration (NASA) POWER data access viewer.

Main Results

• The RD-Bagging model exhibited the best performance during both testing and validation periods.
• Testing period: Correlation Coefficient (CC) = 0.7849, Mean Absolute Error (MAE) = 0.0139 m/d, Root Mean Square Error (RMSE) = 0.0172 m/d, Percent Bias (PBIAS) = -1.42%.
• Validation period: CC = 0.822, MAE = 0.0119 m/d, RMSE = 0.0154 m/d, PBIAS = -1.18%.
• Temperature-related parameters (mean, maximum, and minimum temperature) and solar radiation showed the strongest positive correlations with evaporation, while relative humidity showed a significant negative correlation.
• The RD-Bagging model was selected as the most robust and generalizable model, resistant to overfitting, based on its consistent low errors in new data (test data) and cross-validation.

Contributions

• Introduced a novel methodological framework by integrating regression discretization with ensemble learners for reservoir evaporation modeling.
• Enhanced model robustness and generalizability, addressing limitations of region-specific models.
• Demonstrated practical applicability in data-scarce environments, such as the Sidi-M’Hamed Ben Aouda dam basin.
• Provided a foundation for developing data-driven decision support systems in hydrological modeling and climate resilience planning.

Funding

This research did not receive any specific grant from public, commercial, or not-for-profit funding agencies.

Citation

@article{Achite2025Performance,
  author = {Achite, Mohammed and Katipoğlu, Okan Mert and Elbeltagi, Ahmed and Elshaboury, Nehal and Pandey, Kusum and Emami, Somayeh and Yongguo, Chen},
  title = {Performance enhancement of daily reservoir evaporation rate estimation models using stacking regression by discretization with AI methods},
  journal = {Theoretical and Applied Climatology},
  year = {2025},
  doi = {10.1007/s00704-025-05720-8},
  url = {https://doi.org/10.1007/s00704-025-05720-8}
}