Prediction of Reservoir Evaporation Using Feature Selection Techniques and a Hybrid ANFIS with Recent Metaheuristic Algorithms
Abstract Evaporation, a key factor affecting reservoir water storage, plays a crucial role in reservoir water budgets. This study modeled monthly evaporation from the Boukourdane reservoir (Algeria) using a hybrid Adaptive Neuro-Fuzzy Inference System (ANFIS) combined with four recent optimization algorithms: Black-winged Kite Algorithm (BKA), Exponential Distribution Optimizer (EDO), Driving Training-Based Optimization (DTBO), and Coulomb’s and Franklin’s Laws Inspired Algorithm (CFA). To determine the optimal input variables, three feature selection techniques—Principal Component Analysis (PCA), Gamma Test (GT), and Forward Selection (FS)—were applied. Model performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Correlation Coefficient (R). The best results were achieved using FS-selected inputs—minimum air temperature (Tmin), relative humidity (Hr), and maximum water temperature (Twmax)—with the ANFIS-CFA model outperforming the others. During the testing phase, ANFIS-CFA achieved RMSE = 1.07 mm, MAE = 0.87, and R = 0.87, indicating superior accuracy in evaporation prediction.
Authors
- Andrea Petroselli (ORCID: https://orcid.org/0000-0003-4943-0928)
- Mohammed Abdullahi Mu’azu
- Boudjerda Marouane
Institutions
- University of Jijel (DZ)
- Università degli Studi della Tuscia (IT)
- University of Hafr Al-Batin (SA)
Publication Details
- Journal
- Water Resources
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1134/s0097807826600841
- Primary Topic
- Hydrological Forecasting Using AI
- Type
- article
- Field-Weighted Citation Impact
- 0.00