Predictive Modeling of ETO Across Iranian Climates: An ANN and Hybrid Approach

Abstract

Abstract Evapotranspiration is an essential component of the hydraulic cycle and is a crucial topic for water resource management. This study focuses on modeling daily reference crop evapotranspiration using an artificial neural network and a genetic algorithm in three stations: Ahvaz (dry climate), Saghez (semi-humid and dry climate), and Sardasht (humid climate). The study uses daily meteorological data from 2011 to 2020, including temperature, humidity, sunshine hours, and wind speed. The PMO-56 method is chosen as the modeling objective. The modeling process is investigated under various data scarcity conditions to determine the superior model in each scenario. The comparison of results between different models is based on Root Mean Squared Error (RMSE) and correlation coefficient (R). The results show that the best outcomes are achieved using four input parameters and the whale optimization algorithm approach. Also, ANN-WOA31, ANN-WOA21, and ANN-WOA11 models had the highest estimates in the Ahvaz station in different input conditions. ANN-WOA31, ANN-WOA24, and ANN-WOA11 models were the best in the Saghez station, and ANN-WOA33, ANN-WOA21, and ANN-WOA11 models performed best in Sardasht station with different input combinations. Furthermore, it is observed that with only the temperature parameter and using either an artificial neural network, genetic algorithm, or whale optimization algorithm method, good results can be obtained in all three stations. Following this, humidity and solar radiation can significantly influence the results, while wind speed alone does not substantially impact. Additionally, in all cases, the whale optimization algorithm consistently outperforms the other models in producing better results.