Based on experiments in the field, a comparison of modified referenced evapotranspiration model such as Penman model and Priestley Taylor model, actual evapotranspiration model such as Penmen Menteithin model for closed canopy and Shuttleworth Wallace model for sparse canopy is made. Observed data of soil moisture in two years of different climatic patterns is used to simulate the variation of soil moisture. The results show that in dry years the fitness of four models is better than that in rainy years, especially for dynamical models. In rainy years, when maize plants fully grow, the fitness of Shuttleworth Wallace model is better than that of Penman Menteith model, but in dry years when leaf area is bigger, the fitness of Penman model is better. Comparatively, fitness of modified referenced evapotranspiration models is stable owing to those models including parameters of soil moisture. There are many parameters in dynamical models among which some parameters are not observed. Therefore, when maize plant is small and the moisture content in surfaces of both leaves and soil varied frequently, the fitness of dynamical models is unstable. The results also show that the fitness of variation of soil moisture content simulated by using environmental factors is reasonable in a certain time scale and a big accumulated error will appear on a long time scale over one month.