Abstract: Groundwater evapotranspiration (ET_G) is widely estimated using methods based on diurnal water table fluctuations. However, these methods fail to account for the recharge of capillary water by groundwater at night, which results in an underestimation of ET_G due to the underestimated lateral recharge of groundwater. This study proposes a modified method that takes into account the capillary water recovery rate at night. HYDRUS-1D, a one-dimensional saturated unsaturated flow model, and field measurements are used to evaluate the modified method. Results demonstrate that the relative errors of the Loheide method are all larger than 45% under various soil conditions based on the model. The errors rise as the water table depth and real ET_G both rise. The modified method significantly improves estimation accuracy, with an error of less than 4%. The hour-scale ET_G estimated by the modified method shows a stronger association with potential evapotranspiration in the validation of the measured data. According to the above results, the modified method considerably raises the hour-scale ET_G calculation accuracy. It can also be applied successfully to hour-scale ET_G calculations in semi-arid areas, offering technical support for quantitative studies on the connection between groundwater and plant ecology.