Abstract: In arid and semi-arid regions, groundwater evapotranspiration is one of the main discharges, therefore the accurate estimation of evapotranspiration is significant to groundwater resources assessments. The Loheide method is currently the most accurate method to estimate hourly groundwater evapotranspiration using diurnal water table fluctuations. Aiming to improve the detrend procedure, the linear and cubic polynomial detrend method were proposed to detrend water table data over 3~5 days. The performance of the modification was assessed quantitatively using a model. The results indicate that the linear and cubic polynomial detrend method improve the estimation to some extent for the cases that water table varies dramatically. The best method is the cubic polynomial detrend method that increases the estimation by 16% in comparison to the original detrend method. Compared to the method which uses one-day water table data to get the detrend fluctuation, the modified method also is characterized by the enhancement of the detrend efficiency.