Reference crop evapotranspiration (ETo) is a crucial component in calculating crop water requirements, and its accurate prediction is vital for effective agricultural water management and irrigation planning. Generally, the FAO Penman-Monteith 56 equation is recommended as the benchmark’s method for calculating Eto, but it requires extensive meteorological data—posing challenges in regions with sparse monitoring infrastructure. This review addresses a critical gap: the lack of systematic comparative analysis of machine learning (ML) methods for ETo estimation under data-limited conditions. We review 325 studies searched by Web of Science from 2001 to 2024, focusing on applications of machine learning models in ETo modeling and prediction. Then, this review evaluates these models regarding their characteristics, accuracy, and applicability, including artificial neural networks (ANN), support vector machines (SVM), ensemble learning (EL), and deep learning (DL). Crucially, EL models demonstrate superior stability and cost-effectiveness, with typical performance metrics of R2 > 0.95 and RMSE ranging from 0.1 to 0.6 mm·d−1. Notably, DL methods achieve the highest accuracy under conditions of data scarcity. Using only temperature data, they attain competitive performance (R2 = 0.81, RMSE = 0.56 mm·d−1). Additionally, we further synthesize optimal input variables, performance metrics, and domain-specific implementation guidelines. In summary, this study provides a comprehensive and up-to-date overview of machine learning methods for ETo modeling, thereby offering valuable insights for researchers in the field of evapotranspiration.
Nowadays, irrigation takes up about 35% of agricultural water consumption worldwide, and irrigation induced secondary soil salinizationsalinization affects the crop production and sustainable development of arid irrigation areas globally. However, the regular pattern of salt accumulation in the root zone during the cropping season and the contributions of its attribute factors are still unclear. Therefore, a 10-year monitoring was conducted in the Hetao Irrigation District to reveal the soil salt accumulation during the cropping season and to relate it to influential factors, including potential crop evapotranspiration, water input (field irrigation + precipitation) and water table depth. It was found that under the climate conditions and water-saving irrigation measures of the investigated 10-year period, (1) the salt accumulated during the cropping season could be effectively leached by autumn irrigation and the root zone soil could remain suitable for crop germination, (2) the cropping season water deficit (potential crop evapotranspiration − field irrigation − precipitation) showed strong correlation with the cropping season salt accumulation, and (3) maintaining the cropping season average water table depth larger than a critical depth (roughly 3 m) might be the most economical way to alleviate salt accumulation. Therefore, it is recommended to balance the salt leaching and the water table depth controlling in the future water-saving irrigation management practices.
Although the maximum carboxylation rate (Vcmax) is an important parameter to calculate the photosynthesis rate for the terrestrial biosphere models (TBMs), current models could not satisfactorily estimate the Vcmax of a crop because the Vcmax is always changing during crop growth period. In this study, the Breathing Earth System Simulator (BESS) and light response curve (LRC) were combined to invert the time-continuous Vm25 (Vcmax normalized to 25 °C) using eddy covariance measurements and remote sensing data in five maize sites. Based on the inversion results, we propose a Two-stage linear model using leaf age to estimate crop Vm25. The leaf age can be readily calculated from the date of emergence, which is usually recorded or can be readily calculated from the leaf area index (LAI), which can be readily obtained from high spatiotemporal resolution remote sensing images. The Vm25 used to calibrate and validate our model was inversely solved by combining the BESS and LRC and using eddy covariance measurements and remote sensing data in five maize sites. Our Two-stage linear model (R2 = 0.71–0.88, RMSE = 5.40–7.54 μmol m−2 s−1) performed better than the original BESS (R2 = 0.01–0.67, RMSE = 13.25–18.93 μmol m−2 s−1) at capturing the seasonal variation in the Vm25 of all of the five maize sites. Our Two-stage linear model can also significantly improve the accuracy of maize gross primary productivity (GPP) at all of the five sites. The GPP estimated using our Two-stage linear model (underestimated by 0.85% on average) is significantly better than that estimated by the original BESS model (underestimated by 12.60% on average). Overall, our main contributions are as follows: (1) by using the BESS model instead of the BEPS model coupled with the LRC, the inversion of Vm25 took into account the photosynthesis process of C4 plants; (2) the maximum value of Vm25 (i.e., PeakVm25) during the growth and development of maize was calibrated; and (3) by using leaf age as a predictor of Vm25, we proposed a Two-stage linear model to calculate Vm25, which improved the estimation accuracy of GPP.
25 September 2023 in Science of The Total Environment