Long term variation of evapotranspiration and water balance based on upscaling eddy covariance observations over the temperate semi-arid grassland of China

As an important part of global semi-arid grassland, adequately understanding the spatio-temporal variability of evapotranspiration (ET) and water balance (WB, defined as precipitation-ET) over the temperate semi-arid grassland of China (TSGC) could advance our understanding of climate, hydrological and ecological processes over global semi-arid areas. Based on the largest number of in-situ ET measurements (13 flux towers) within the TSGC, we applied the support vector regression (SVR) method to develop a high-quality ET dataset for the TSGC from 1982 to 2015. The model performed well in validation against flux tower-measured data and comparison with water-balance derived ET. The annual ET ranged from 226 to 294 mm yr(-1) with an average of 250 mm yr(-1) and an insignificant trend during 1982-2015. Nevertheless, ET increased significantly during 1982-1998, and turn to decrease during 1999-2011. Regional WB, which approximately represents water availability, fluctuated without a significant trend for entire period. But the annual mean WB at 1999-2011 decreased by 26.1% compared with that before 1998. On average, ET was dominated by climatic factors, especially precipitation, rather than human activities. The reversal of ET trend after 1998 was predominantly explained by the slow increase of precipitation, the stalled increase of temperature, and the significantly decreasing radiation. The revegetation programs exacerbated the evaporative water consumption significantly in some areas after 1998, such as Yellow River Basin. The variation of WB was mainly driven by precipitation, whether in time or space. In addition, the spatio-temporal difference between our SVR ET dataset with four widely used global ET datasets (i.e., GLEAM, FLUXCOM, MODIS, and BESS) suggested that none of these four products can capture both the magnitude and variability of ET well in the TSGC, and local observations are of great importance.