Comparison of Bio-Inspired Optimization-Based MLP Models for Monthly Reference Evapotranspiration Estimation in Southeastern Türkiye
Year: 2026
Type: article
Abstract: This study evaluates the performance of hybrid multilayer perceptron (MLP) models with six different biologically inspired meta-heuristic optimization algorithms for the estimation of monthly reference evapotranspiration (ET₀) in Southeast Anatolia, Türkiye. Using NASA POWER data (1984-2024) as main source of climatic data, model input variables for six provinces were determined by Recursive Feature Elimination (RFE) method. Moth-Flame Optimization (MFO), Dolphin Echolocation Optimization (DEO), Firefly Optimization (FOA), Grey Wolf Optimization (GWO), Golden Jackal Optimization (GJO) and Cuckoo Search Optimization (CSO) algorithms were used in the optimization of MLP models. Model performance was evaluated using the coefficient of determination (R²), Nash–Sutcliffe efficiency (NSE), root mean square error (RMSE), mean absolute error (MAE), RMSE-to-observation standard deviation ratio (RSR), and global performance index (GPI). The results show that DEO-MLP achieved the best overall performance among the investigated hybrid models, with mean values across the six stations of NSE = 0.970, R² = 0.972, MAE = 0.39, RSR = 0.17, and GPI = 0.33. The superior performance of DEO-MLP model was demonstrated by low error rates and high accuracy levels compared to the other models. In contrast, the MFO-MLP, CSO-MLP and GJO-MLP models exhibited lower accuracy levels. In general, the results demonstrate that the predictive performance of MLP models varies according to the meta-heuristic optimization algorithm employed, with DEO-MLP showing the best overall performance among the investigated hybrid models. This study is expected to contribute to the development of agricultural water management strategies in Southeast Anatolia and other regions characterized by similar climatic conditions and provides a practical modelling framework for estimating ET₀ when the availability of long-term ground-based meteorological observations is limited.
DOI: 10.36253/ijam-3853
Author: Mustafa Ozbuldu
Institution: Mustafa Kemal University
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: hybrid
OpenAlex ID: W7204572053