Simulating Temporal And Spatial Variation Of Evapotranspiration Over The North China Plain

North China Plain?NCP?is one of the main grain production bases in China,however,it suffers from severe water shortages and groundwater overexploitation.Evapotranspiration?ET?is the crucial process in water cycle,especially in semi-humid area.Therefore,the accurate estimation of regional ET and the precise analysis of its spatial-temporal pattern are important for understanding the water cycle and maintaining sustainable food production.This study applied a machine learning algorithm,Support Vector Regression?SVR?,to develop ET dataset for cropland over NCP from 1982 to 2015 with 1/12°spatial resolution and eight-day temporal frequency based on eight flux-sites within NCP and the surrounding area.The SVR model performed well in cross validation experiments with mean-square-errors less than 0.63mmday-1 and R² greater than 0.75.It proved that the model was competent for regional ET estimation.According to the estimates,temporally,the mean annual ET over cropland within NCP was 576?25mm yr^-1 during 1982-2015 with a non-significant increasing trend.In wheat season,the mean annual ET was 238?19.1mmyr^-11 which increased significantly at the rate of 1.28mmyr^-1;in maize season,the annual mean was 221?16.3 mmyr^-1 with a decreasing trend of-0.95 mmyr^-2.Spatially,the northern area with little irrigation and the area adjacent to Bohai Sea showed the lowest ET,followed by the Tai Mountain and Shandong Peninsula.The comparative high ET occurred in the well irrigated area near Taihang Mountains and the Yellow irrigation district in semi-humid region.Besides,the largest ET occurred in the southern humid area.As for the spatial distribution of ET trend,at annual scale,the most area in NCP showed non-significant trend,except for the southern humid area and the piedmont plain.In wheat season,ET increased significantly over 70.8%of the study area;in maize season,ET decreased significantly over 68.6%of the study area.Based on the attribution analysis which combined multiple regression with the first difference method,the contributions of climate change and human activities were separated.The result showed that,at annual scale,human activities controlled ET trend with actual and relative contributions of 1.42 mmyr^-2 and 52%,respectively;in wheat seaon,the ET trend was also dominated by human activities as with actual and relative contributions of 0.83 mmyr^-2 and 56%,respectively;while in maize season,ET trend was totally controlled by the climate change with relative and actual contributions of 77%and-1.11 mmyr^-2,respectively.As for the dominant climate factors of ET trend,the result showed that,at annual scale,net radiation and wind speed were dominant climate factors;in wheat season,the significant increasing air temperature and non-significant increasing net radiation were the principal cause of ET increase;while in maize season,the significant ET decline was primarily resulted from the significant decreasing net radiation and wind speed.