Objective This paper integrates the three temperature model that eliminates impedance and impedance based dual source model to simulate and compare evapotranspiration (ET) and transpiration (T), achieving the fusion of different models and dynamic simulation of evapotranspiration.

Method Based on the existing two-source impedance model framework and the energy balance of dry soil and dry leaves, combined with numerical simulation technology, this study simulated the seasonal dynamics of reference temperature, and integrated it with the three-temperature model to simulate transpiration and soil evaporation.

Result (1) Compared with the eddy measured ET, the R² of ET estimated by the three-temperature model and the two-source model for the observations at Daman Station at the daily scale were 0.85 and 0.72, respectively, and R² of the simulated temperate grassland station at the hourly scale was 0.89, and the simulation effect was better at the hourly scale. The above research showed that the research framework provided theoretical support for the estimation of reference temperature in the three-temperature model, and better quantified the spatio-temporal dynamics of reference temperature. (2) Model scenario test analysis showed that the prediction results of T and ET of the two models were similar, provided that canopy available energy (such as net radiation) and ground soil available energy (such as net radiation and surface heat flux) were limited within the framework of energy balance. On the contrary, due to the mismatch of the input driving factors (temperature and radiation) in the dry and wet scenarios, the simulation accuracy of T and ET by the “three-temperature” model decreased significantly. Therefore, in the framework of energy balance, the effect of three-temperature model was better when the input temperature and radiation system match under dry and wet land conditions. (3) The reference temperature had certain diurnal and seasonal variations.

Conclusion In the daytime, the leaf or soil reference temperature is generally higher than the actual temperature, but lower than the actual temperature at night, showing a large daily variation. In terms of daily mean, the reference temperature is generally higher than the actual simulated temperature. The results of this study provide a promising new temperature gradient method and can be used to monitor canopy water stress and dynamics.