Evapotranspiration partitioning and its interannual variability over a winter wheat-summer maize rotation system in the North China Plain

Partitioning of evapotranspiration (ET) is important for understanding surface-atmosphere interactions, hydrological cycle, and plant water use strategy. Here, we applied seven widely used eddy covariance (EC)-based methods and 15-year EC measurements in a winter wheat-summer maize rotation cropland in North China Plain to partition ET into soil evaporation (E) and plant transpiration (T). Then the two-stage theory of bare soil evaporation was employed to evaluate these partitioning methods. This innovative evaluation approach is particularly suitable for this kind of ecosystem, requires no direct measurements of ET components, and avoids spatial mismatching between the source areas of reference values and EC-based ET. Combining the two-stage theory and the meta-analysis, we found that among seven partitioning methods, only the Transpiration Estimation Algorithm (TEA) (hereafter, N18 method) utilizing a machine learning approach not only simulated the dynamics of 14-day transpiration fraction of ET (T/ET) quite well, but also yielded reliable 14-day and mean growing season magnitudes of T/ET for both crops. Furthermore, we developed a new partitioning method based on the stomatal slope parameter in the optimality-based unified stomatal optimization (USO) model. The newly developed method showed similar performances with N18 method on partitioning ET of both crops at our site. By using the N18 method and the newly developed method, we revealed that the multi-year mean growing season T/ET (+/- standard deviation) was 0.72 +/- 0.03 and 0.77 +/- 0.04 for maize and wheat, respectively. For the interannual variability of ET and its components, only T of maize increased significantly during 2005-2019 at our site. Moreover, it was found that E had the highest interannual variability followed by T and then T/ET for both crops.

Automated summary

Partitioning of evapotranspiration (ET) is crucial for understanding surface-atmosphere interactions and plant water use strategy. In this study, seven different partitioning methods were compared, with the machine learning method TEA showing the best performance in simulating crop transpiration and soil evaporation. A new method based on the USO model also demonstrated good partitioning results. The study also revealed the interannual variability of crop T/ET, as well as the differences in ET components.