Predicting water and fertilizer losses: A machine learning approach for center pivot fertigation management

Wind drift and evaporation losses (WDELs) of both water and fertilizer significantly impact the performance of center pivot irrigation systems. This study evaluated the influence of meteorological factors, fertilizer concentrations, and sprinkler types on application uniformity, accuracy, and WDELs during fertigation. Experiments were conducted using a center pivot system equipped with three common low-pressure sprinklers (Nelson R3000, D3000, and Komet KPT) and fertilizer concentrations (0.1 −0.3%) under varying conditions of wind speed, temperature, humidity, and solar radiation. Moreover, four machine-learning models were employed to evaluate and predict the fertilization solution wind drift and evaporation losses (FS-WDELs) and the fertilizer wind drift and evaporation losses (F-WDELs) using sensitive meteorological and operational parameters. The results indicated that the R3000 sprinkler minimized radial fluctuations in water and fertilizer distribution, achieving the highest distribution uniformity (uniformity coefficients for total fertilizer solution and fertilizer application were 91.0% and 89.4%, respectively). Correlation analysis identified wind speed and solar radiation as the primary factors influencing FS-WDELs and F-WDELs. The magnitude of WDELs followed the order D3000 > KPT > R3000, demonstrating that R3000 was the most resistant to WDELs. Among the machine-learning models, support vector regression (SVR) exhibited superior accuracy in estimating WDELs (R2 = 0.95, RMSE = 1.7%). Consequently, the R3000 sprinkler and SVR model are recommended for optimizing fertigation management and enhancing water and fertilizer use efficiency. These findings offer critical insights for selecting operational parameters in center pivot irrigation systems.