Abstract: To improve the model simulation accuracy of crop response function to water, the modified model of cotton crop response to water was established based on the experimental data obtained in 2006 and 2008. Parameters of cotton crop response to water were solved with nonlinear programming by interval period division in growing period of cotton. The relationship of parameters and the number of periods was analyzed. The model of crop response to water was modified by introducing the number of periods into cumulative function of moisture sensitive index, and was compared with the existing model of crop response to water. The results showed that the relative error of yield simulation by modified model was reduced with number of periods increasing. When the number of periods was more than 11, the average of relative error and the biggest value were respectively reduced to less than 7% and 15%. Compared with existing model, the number of parameters was not increased and the simulation accuracy was improved for modified model. It shows that the modified model perform better for yield simulation with any period number and can accurately reflect the change of yield by the water stress time.