Abstract: Evapotranspiration (ET) is composed of two separate processes-water loss to the atmosphere from soil surface by evaporation (E) and water loss to the atmosphere from plant canopy via transpiration (T). ET plays a key role in energy and water balance in agricultural system and is also a critical process in terrestrial hydrological cycle. Accurate estimation of ET is significant for improving water use efficiency and optimizing regional water use, particularly in arid and semi-arid regions. Although ET models have been an important tool in understanding the regulation of ET in ecological, agricultural and environmental sciences, the accuracy of the models is limited by aerodynamic and canopy resistance. Numerous models have been developed to integrate aerodynamic and canopy resistancese.g., Penman-Monteith (P-M) model in simulating the processes of response of ET, but many studies have suggested that the P-M model could produce large errors under partial or sparse canopy conditions because it treated plant canopy and soil surface as a single entity. Next, the dual-layered Shuttleworth-Wallace (S-W) model was developed to estimate ET under different conditions. In this model, the crop ET is divided into two components-latent heat flux from crop and that from soil. It has been tested by various surface conditions and widely used because of its good performance. In this study (which used maize data of three eddy covariance observations for the period from May through September 2012 in Heihe River Basin, an arid area in Northwestern China), two canopy resistance models coupled for maize. Two S-W models were coupled with canopy resistance models of maize taking or non-taking into account the effect of atmospheric CO₂. Then the whole maize growth period was divided into three stages, early, middle and late growth stages. Then maize ET on half hour scale was simulated using the two S-W models. The performances of the two S-W models were validated for three different growth stages using eddy covariance field-measured data. The results showed that simulated maize ET by the S-W model (which took into account the effect of atmospheric CO₂ at every growth stage of three different places) best agreed with field-measured eddy covariance data. Sensitivity analysis of the revised S-W model (taking into account the effect of atmospheric CO₂) showed that maize ET was more sensitive to canopy resistance (r_s^c) and aerodynamic resistance from canopy to reference surface height (r_a^a) at different growth stages. Therefore, it is very necessary to determine resistance parameters at different growth stages taking into account the effect of atmospheric CO₂ when calculating maize ET using the revised S-W model.