Evapotranspiration (ET) is one of the most important fluxes in terrestrial ecosystems that regulate atmosphere -hydrosphere-biosphere interactions. Several studies have suggested that global ET has significantly increased in the past several decades, and that such increase has exhibited big spatial variability, but there are few detailed studies on the spatio-temporal change in ET over China. Here, we developed a high-resolution data oriented monthly ET product in China between 1982 and 2015 by integrating remote-sensing and the eddy covariance technique observed ET data in a machine learning approach (model tree ensemble, MTE). We showed that the mean annual ET over China is 552 +/- 14 nun yr(-1), which is comparable to the estimate from a MTE-derived product based on water balance, but is larger than that from both previous MTh-derived global product and process-based land surface models. ET in China significantly increased with a rate of 10.7 mm yr(-1) per decade over the past 30 years (p < 0.05). The largest increases in ET ( > 60 mm yr(-1) per decade) occurred in the eastern periphery of Sichuan, southern Taiwan, and central China, which was attributed to the increases in temperature and solar radiation, as well as the enhanced vegetation productivity. About 22% of the area showed a decreasing trend in ET, mainly in parts of southeastern, southwestern, and northeastern China. The regional decrease in ET was likely due to decreasing precipitation and/or vegetation browning. Although our finding of the significant increase in China's ET at the country scale is supported by five different ET products, there are still less agreement on the change in ET at the regional scale among different ET products.