Abstract: Evapotranspiration (ET) is the major component of the energy balance in the water cycle. The eco-hydrology can alter the canopy conductance and energy distribution. This study aims to clarify the influence of the environmental factors on the ET dynamics in the interaction and feedback between the hydrological and land surface. The research area was also taken as the Yanhe River Basin in the Loess Plateau. A dynamic optimal scheme of the canopy conductance was proposed to improve the ET model, in order to consider the dynamics of the land cover types. The spatiotemporal ET pattern was obtained under the ecological restoration from 2000 to 2023. The GeoDetector model was utilized to clarify the driving mechanisms of the multiple environmental factors on the ET. The results were summarized into three aspects. 1) The improved ET model significantly reduced the simulation error at the site scale. There was an excellent linear fitting relationship between the estimated ET by the water balance and the improved P-M_Zhang model. There were the R² = 0.97 (P < 0.01) and the RMSE = 35.1 mm/a, compared with the ET estimated by the water balance at the watershed scale. 2) The time variation of the annual ET displayed a non-significant upward trend from 2000 to 2020, but the upward trend was observed since 2020. The spatial evolution pattern of the ET significantly increased in most areas. There was an increasing trend from the west to the east, while a decreasing trend was found from the north to the south. The ET was reduced in the villages and towns, due to the sparse vegetation, hardened ground surface, and high albedo. There were the pathways and energy distributions of the ET. Hence, the improved ET model represented the heterogeneity of the ET fluxes under different surface cover types. The experiment was also in line with the actual situation. 3) The driving mechanism indicated that the q values (explanatory power) of FVC (fractional vegetation cover), P (precipitation), VPD (vapor pressure deficit), and LUCC (land use/cover change) were all greater than 0.10 (significance level of 0.01) in the three time periods of 2000-2010, 2011-2023 and 2000-2023. The main driving factors were determined on the spatiotemporal heterogeneity of the ET. Among them, the explanatory power of FVC was the greatest. SH (specific humidity) and POP (population) were gradually the key driving factors after 2010. The explanatory power of the interaction was greater than that of the single factor. The factor interaction was all nonlinear and double-factor enhancement. There was a more outstanding interaction of the FVC with other factors. The interactions between the FVC and VPD, VPD and P, and P and AT (air temperature) dominated the formation and heterogeneity of the ET. A full understanding was also obtained of the ET and its driving mechanism under various environments. The finding can also provide a strong reference for the ecological restoration, water resources, and regional ecological civilization in the Loess Plateau.