Abstract: Abstract: The accurate estimation of potential evapotranspiration (ET0) is an important content of hydrological cycle and flux cycle, which has an important theoretical and practical significance for effective use of agricultural water resources in the context of climate change. In order to acquire stable and reliable estimation method of evapotranspiration that need only a small number of climatic factors, an improved ET0 estimation method of bilinear surface regression model (BSRM) was used to calculate the daily evapotranspiration based on the observed meteorological data from 6 weather stations in the upper Yishu River watershed (34.37°-36.38°N, 117.40°-119.18°E). Three types of BSRM models were considered according to the difference in computing solar radiation in this study. In the first method, the relative insolation duration and the ET0 estimation was based on the calculated solar radiation, relative humidity and air temperature (BSRMn/N). In the second method, relative humidity, air temperature and solar radiation computed by relative insolation duration and extra-terrestrial radiation were used for ET0 estimation (BSRMRs). In the third method, three variables of solar radiation computed by Hargreaves-Allen equation, relative humidity and air temperature were used for ET0 estimation (BSRMt). Meanwhile, Penman-Monteith (P-M) equation was used to estimate the ET0 as a reference and comparison method. The precision of the 3 kinds of BSRM approaches was tested and evaluated based the standard of land surface potential ET0 calculated by the means of conversion coefficient and observed evapotranspiration. On the base of above studies, the influential factors of ET0 were further analyzed. The results showed that the model precision for ET0 estimation was highest in BSRMn/N, followed by BSRMRs and BSRMt. The BSRMn/N method has the minimum mean absolute of error and root mean square of error (0.48 and 0.64 mm). The P-M equation and BSRMn/N method could yield the reliable ET0 estimation, however the P-M equation could overestimate the results in daily ET0 between 2-6 mm. The general tendency of ET0 estimation from the 3 kinds of BSRM methods and P-M equation were consistent with that of the observation for monthly ET0, while the accuracy of ET0 from BSRMn/N was closer to the observation. Studying the impacts of major meteorological factors on ET0 showed that the BSRMn/N method had the least effect on the ET0, followed by BSRMRs and BSRMt method. The simulated ET0 from the BSRMn/N method was in accordance with the observed value at various relative insolation duration, relative humidity and air temperature. The underestimation and bias from BSRMt increased at the higher temperature, the lower relative insolation duration and relative humidity. The stronger wind speeded up the process of evapotranspiration. The estimated precision from the BSRMn/N model was slightly higher than the P-M model, and the former was little influenced by meteorological factors. The input meteorological variables were relatively less and convenient to obtain, which suggested that the proposed BSRMn/N model may become a stable and reliable alternative for routine daily evapotranspiration estimation in the study area.