Compared with more comprehensive physical algorithms such as the Penman–Monteith model, the Priestley–Taylor model is widely used in estimating evapotranspiration for its robust ability to capture evapotranspiration and simplicity of use. The key point in successfully using the Priestley–Taylor model is to find a proper Priestley–Taylor coefficient, which is variable under different environmental conditions. Based on evapotranspiration partition and plant physiological limitation, this study developed a new model for estimating the Priestley–Taylor coefficient incorporating the effects of three easily obtainable parameters such as leaf area index (LAI), air temperature, and mulch fraction. Meanwhile, the effects of plastic film on the estimation of net radiation and soil heat flux were fully considered. The reliability of the modified Priestley–Taylor model was testified using observed cotton evapotranspiration from eddy covariance in two growing seasons, with high coefficients of determination of 0.86 and 0.81 in 2013 and 2014, respectively. Then, the modified model was further validated by estimating cotton evapotranspiration under three fractions of mulch cover: 0%, 60%, and 100%. The estimated values agreed well with the measured values via water balance analysis. It can be found that seasonal variation of the modified Priestley–Taylor coefficient showed a more reasonable pattern compared with the original coefficient of 1.26. Sensitivity analysis showed that the modified Priestley–Taylor coefficient was more sensitive to LAI than to air temperature. Overall, the modified model has much higher accuracy and could be used for evapotranspiration estimation under plastic mulch condition.