The Grain for Green Project in China’s Loess Plateau has significantly increased the vegetation coverage (M) since it was implemented in 1999. Accurately modelling evapotranspiration (ET) and attributing its changes are important for assessing the hydrological effects of revegetation in this area. Budyko-based models have been widely used to estimate ET whereby the controlling parameter (ω) captures the effects of land surface conditions and climate seasonality. Although the effects of climate seasonality on ET variation have been theoretically discussed, its important role in ω remains further investigation. An improved climate seasonality and asynchrony index (SAI) was thus used to reflect the seasonality and asynchrony of water and energy distribution in this study. Then ω was extended to M and SAI at grid scale to model annual ET by linking Fu equation in China’s Loess Plateau for the period 1981–2012. Further, the whole study period was split into two sub-periods at the year of 1999, and then the complementary method was used to quantify the contributions of precipitation (P), potential evapotranspiration (E₀), M and SAI changes to ET variation between the two sub-periods. The results showed that ET increased by 5.1 mm/yr after 1999. ET is most sensitive to changes in P, followed by M, E₀, and SAI. However, increasing M dominated the overall increase in ET, outweighing the effects of decreasing P and increasing SAI. Because SAI accounted for almost a third of total ET change, the impacts of climate seasonality cannot be ignored in ET simulation and attribution analysis.