Document Type : Original Article
Authors
1
Ph.D., Civil Engineering, Shahrood University of Technology, Shahrood, Iran
2
Assistant Professor, School of Engineering, Damghan University, Damghan, Iran
3
MSc Student, Civil Engineering, Shahrood University of Technology, Shahrood, Iran
4
Ph.D. Student, Civil Engineering, Shahrood University of Technology, Shahrood, Iran
10.22044/jhwe.2025.16384.1070
Abstract
Accurate prediction of evaporation is critical for effective water resource management, particularly in arid and semi-arid regions. This research evaluates the performance of five machine learning algorithms Decision Tree, K-Nearest Neighbors, Support Vector Regression, Random Forest, and Artificial Neural Network in estimating monthly evaporation rates using meteorological data collected at Shahrood Synoptic Station from 1992 to April 2025. The dataset includes key climatic parameters such as average temperature, wind speed, precipitation, and relative humidity. Model performance was assessed through four metrics: Mean Absolute Error, Coefficient of Determination, Kling-Gupta Efficiency, and Average Absolute Relative Deviation. Results indicate that the Random Forest model outperformed all others, achieving the lowest MAE of 19.94 mm, highest KGE of 0.973, and lowest AARD of 0.521, reflecting superior accuracy and stability. The Artificial Neural Network model also demonstrated strong predictive capability, closely followed by Support Vector Regression, while simpler models like Decision Tree and K-Nearest Neighbors showed comparatively weaker performance due to their limited ability to capture complex evaporation dynamics. Temporal analysis revealed that all models effectively captured seasonal evaporation patterns, with Random Forest and Artificial Neural Network most accurately tracing peak and trough fluctuations. The results demonstrate that machine learning models possess strong predictive accuracy for evaporation estimation and offer a reliable approach for assessing evaporation and water loss.
Keywords
)