Investigating the performance of artificial rabbit optimization hybrid algorithm (ANN-ARO) in forecasting reference evapotranspiration with limited climatic parameters

Reference evapotranspiration (ETo) is considered one of the important variables in hydrology and agricultural science and is a determining factor in water resources management. This study investigates a hybrid model of an artificial neural network with an artificial rabbit optimization algorithm (ANN-ARO) for daily modeling of ETo with limited meteorological parameters. It compares it with other hybrid methods, i.e. ANN with a particle optimization algorithm (ANN-PSO). ANN with genetic algorithm (ANN-GA) and five different data mining models. These models were evaluated using long-term daily climate data from 2000 to 2023 in two climates. The investigated stations included Birjand (with a desert climate) and Mashhad (with a cold semi-arid climate). The statistical comparison showed that considering all climatic parameters, the hybrid ANN-ARO model in Mashhad city with R2=0.9986 and MSE=0.0001 and in Birjand city with R2=0.9986 and MSE=0.0001 gave better estimates than other methods. In addition, the ANN_ARO optimization algorithm has the best estimation with "temperature" and "relative humidity" by considering the minimum meteorological parameter, and also by considering two and three input parameters, it performs better than other methods. In general, nature-inspired optimization algorithms are powerful tools to enhance the performance of ANN in ETo simulation. According to the results, the ANN-ARO model is highly recommended for estimating ETo in similar climate regions with limited climate data. This study proposes powerful models for accurate estimation of ETo with limited inputs in arid and semi-arid climates, which provide practical implications for the development of precision agriculture.