Performance evaluation of soft computing techniques for forecasting daily reference evapotranspiration

Abstract

Precise information on evapotranspiration aids in efficient irrigation scheduling and optimal crop production. In this study, soft computing tools, namely, artificial neural network (ANN) and k-nearest-neighbor (KNN) models, were evaluated by comparing with the Penman–Monteith (PM) model using climatic data from 1990 to 2020 of the Indian Agricultural Research Institute (IARI) farm, New Delhi, India. Results revealed that the ANN model with sigmoid activation function and the L-BFGS learning algorithm was selected as the best-performing model among 36 ANN models tested in this study. Among all four KNN models tested, the K4 KNN model was observed to be the best in forecasting daily ET0. Overall, the best ANN model (M11) outperformed the K4 KNN model with mean absolute error, mean squared error, r, mean absolute percentage error, and d values of 0.075, 0.018, 0.997, 2.76%, and 0.974, respectively, during the training period and 0.091, 0.053, 0.984, 3.16%, and 0.969, respectively, during the testing period. Sensitivity analysis of the best ANN model revealed that wind speed was the most influencing input variable compared to other weather parameters. Thus, the ANN model to forecast daily ET0 accurately for efficient irrigation scheduling of different crops grown in the study region can be recommended.

Reference evapotranspiration prediction using machine learning models: An empirical study from minimal climate data

“…For example, Rajput et al. (2023) evaluated ANN and KNN algorithms and compared them with the standard PM method using climatic data for New Delhi, India. The study found that the ANN model outperformed the KNN model in forecasting ET 0 when all input variables were considered.…”

“…Indian agriculture sector consumes approximately 90% of the total freshwater resources available in the country due to improved groundwater withdrawal infrastructures and poor on‐farm irrigation efficiency (Jain et al., 2019). Thus, irrigated agriculture has an important role in meeting the food need of the ever‐increasing population (Rajput et al., 2022). However, due to the rising demand for freshwater availability among different water use sectors, the freshwater availability for agriculture is projected to be reduced significantly (Kumari et al., 2022).…”

“…Additionally, the applicability of the empirical models requires estimating the equation coefficients in different climatic conditions for acceptable ET 0 estimation. These empirical models compute ET 0 considering only a few weather data, which may result in under‐ or over‐ET 0 estimation (Rajput et al., 2023b). Therefore, alternate models having better accuracy need to be developed and assessed in different climatic conditions.…”

“…Machine learning (ML) techniques have been employed to study various phenomena like evaporation (Kushwaha et al., 2021), reference evapotranspiration (Elbeltagi et al., 2022; Rajput et al., 2023b), water quality indices (Dimple et al., 2023; Kushwaha et al., 2023), among simulating and other phenomenon processes. In recent years, several research scholars have utilized ML models to estimate ET 0 (Bellido‐Jiménez et al., 2022; Chen et al., 2022; Farooque et al., 2022; Nagappan et al., 2020; Rajput et al., 2023c; Wu et al., 2019).…”

“…Rajput et al. (2023) evaluated ANN and KNN models for forecasting daily ET 0 and concluded that the ANN outscored the KNN model. The SVM and GPR models were used to predict ET 0 from limited climatic variables (Zouzou & Citakoglu, 2021).…”