The lower limit temperature in the crop water stress index (CWSI) model refers to the canopy temperature ( T c ) or the canopy-air temperature differences ( dT ) under well-watered conditions, which has significant impacts on the accuracy of the model in quantifying plant water status. At present, the direct estimation of lower limit temperature based on data-driven method has been successfully used in crops, but its applicability has not been tes-ted in forest ecosystems. We collected continuously and synchronously T c and meteorological data in a Quercus variabilis plantation at the southern foot of Taihang Mountain to evaluate the feasibility of multiple linear regression model and BP neural network model for estimating the lower limit temperature and the accuracy of the CWSI indicating water status of the plantation. The results showed that, in the forest ecosystem without irrigation conditions, the lower limit temperature could be obtained by setting soil moisture as saturation in the multiple linear regression mo-del and the BP neural network model with soil water content, wind speed, net radiation, vapor pressure deficit and air temperature as input parameters. Combining the lower limit temperature and the upper limit temperature determined by the theoretical equation to normalize the measured T c and dT could realize the non-destructive, rapid, and automatic diagnosis of the water status of Q. variabilis plantation. Among them, the CWSI obtained by combining the lower limit temperature determined by the dT under well-watered condition calculated by the BP neural network model and the upper limit temperature was the most suitable for accurate monitoring water status of the plantation. The coefficient of determination, root mean square error, and index of agreement between the calculated CWSI and measured CWSI were 0.81, 0.08, and 0.90, respectively. This study could provide a reference method for efficient and accurate monitoring of forest ecosystem water status.