Intelligent optimization of Reference Evapotranspiration (ETo) for precision irrigation

Abstract

Reference Evapotranspiration (ETo) is the basis for the conservation of irrigation water for sustainable development in agriculture. The FAO-56 Penman-Monteith (FAO-56 PM) is the standard method of ETo determination that is complex due to the involvement of many parameters. Many machine learning approaches were proposed to simplify ETo determination with limited parameters. The existing machine learning-based solutions rely on a fixed number of limited parameters for the determination of ETo. The study proposed a solution to determine the raw ETo with only temperature and optional optimization of the raw ETo according to the availability of Relative Humidity (RH), Wind Speed (WS), and sunshine duration ratio (n/N). The study proposed an ensembled Artificial Neural Network (ANN) model (Model-1) to determine ETo according to available minimal environmental parameters. The environmental conditions from the year 2001–2021 of Pakistan are used to develop a dataset with the FAO-56 PM method. The unique attribute of the proposed solutions is the flexibility in the use of environmental parameters for ETo determination according to the FAO-56 PM method. The performance of the proposed ensembled model (Model-1) is compared against the individual ANN model with all parameters (Model-2) and with the individual ANN model with only temperature (Model-3). The predictions by the proposed solution from Model-1 are also compared against the FAO-56 PM method. It is observed that the proposed Model-1 is more accurate with 91.44% accuracy, as compared to Model-2 and Model-3. The predictions by the proposed Model-1 are more correlated with the ETo by FAO-56 PM with a Pearson correlation of 0.996 as compared to the ETo predictions by Model-2 and Model-3.