Spatio-temporal patterns of evapotranspiration based on upscaling eddy covariance measurements in the dryland of the North China Plain

Accurate evapotranspiration (ET) estimation is important for understanding hydrological cycle and water resources management in the cropland. Based on eight flux sites within the North China Plain (NCP) and the surrounding area, which were integrated together for the first time, we applied support vector regression method to develop ET dataset for the cropland in NCP from 1982 to 2015 with 1/12 degrees spatial resolution and eight-day temporal interval. The mean annual ET over cropland within NCP was 575 mm yr(-1) with a non-significant increasing trend. In wheat season, the seasonal ET increased significantly at the rate of 1.28 mm yr(-2) with a mean value of 238 mm yr(-1); in maize season, the mean value of seasonal ET was 221 mm yr(-1) with a decreasing trend of -0.95 mm yr(-2). The comparatively high mean annual ET values occurred primarily in the well irrigated area near the Taihang Mountains, irrigation districts along the Yellow River and the humid region. The attribution analysis which combined multiple regression with the first difference method revealed that human activities dominated ET trend at annual scale as well as in wheat season, with relative contributions of 52% and 56%, respectively. While in maize season, ET trend was mostly controlled by climate change with relative contribution of 77%. Among the climate factors, in wheat season, the significantly increasing air temperature was the principal climatic cause of ET increase which was partly offset by the significantly decreasing wind speed; while in maize season, the significant ET decline was primarily resulted from the significantly decreasing net radiation and wind speed. Compared with our ET dataset, the four widely used global or national ET datasets significantly underestimated ET in wheat season, especially in the extensively irrigated areas, suggesting that local observations are crucial for regional studies.