Applicability evaluation of the nonparametric approach for estimating evapotranspiration on irrigated farmland in the North China Plain

Evapotranspiration (ET), generated by the evaporation of water from a natural surface into the atmosphere, is an important component of the water cycle and surface energy balance system, which is a fundamental information for agriculture, water resource management, and climate change research. In recent decades, the estimation of ET or latent energy (LE, which is the amount of heat required for ET) has remained one of the most challenging problems for researchers. A nonparametric approach for estimating ET may avoid the complex parameterization process and reduce the calculation uncertainties; therefore, it has broad application prospects. However, a more detailed applicability evaluation of the nonparametric approach in different regions or ecosystems is needed, as most of the current studies on the application of this approach focus on arid basins, with few applicability analysis reports focusing on irrigated farmland in sub-humid areas. In this study, the eddy covariance data modified by the energy residual closed correction method in three irrigated farmland stations (Luancheng Agroecosystem Experimental Station, Chinese Academy of Sciences; Yucheng Comprehensive Experimental Station, Chinese Academy of Sciences; and Guantao Experimental Station, Beijing Normal University) in the North China Plain were used as references, and the applicability of the nonparametric approach for estimating LE on irrigated farmland at different time scales (daily scale and 30 min scale) and seasons in the three stations was evaluated. The results showed that the nonparametric approach had reliable and robust performance for different crop types and time scales on irrigated farmland in the North China Plain. The estimated LE could ideally reflect seasonal and intraday variations, but these values were generally underestimated. Furthermore, the bias, coefficient of determination, root mean square error, and Nash-Sutcliffe coefficient at the daily (and 30 min) scale were −16.18 to −12.88 W∙m⁻² (−13.30 to −7.68 W∙m⁻²), 0.80 to 0.83 (0.88), 22.45 to 31.06 W∙m⁻² (39.22 to 42.15 W∙m⁻²), and 0.66 to 0.75 (0.86 to 0.87), respectively. The nonparametric approach significantly underestimated the latent heat flux when the water supply was sufficient or when crops were growing vigorously; moreover, this approach slightly or not underestimated the latent heat flux when the water supply was insufficient or when crops were sparse. In addition, the response of the nonparametric approach to irrigation activities was not considered sufficiently, and further improvement to the model structure may be required to improve the simulation accuracy of irrigated farmland. Finally, the parameter sensitivity of the nonparametric approach in irrigated farmland in the North China Plain, from high to low, was air temperature, surface temperature, net radiation, and soil heat flux, but the influence of soil heat flux can be ignored. Ultimately, this study not only provides a reference for the improvement of the nonparametric ET estimation approach but also helps further the understanding of ET fundamental theory.