Abstract: Potential evapotranspiration depends greatly on local climate and vegetation distribution conditions, and is important in studying the cropland and basin hydrological circles. The current study provides a detailed comparison of the performances among three dual-source evapotranspiration models, including the layer model, the patch model and the hybrid model, in estimating and partitioning potential evaporation and potential transpiration under different hypothetical vegetation distribution conditions. The layer model ignores the difference of energy fluxes between under- and inter-canopy soil; while the patch approach assumes a full radiation loading for both the canopy and inter-canopy soil and ignores the evaporation from under-canopy soil surfaces; The hybrid model is a combination of the layer and patch models, and adopts the layer approach to partition available energy between canopy and soil and uses the patch approach to calculate energy fluxes. As a result, both under- and inter- canopy soil evaporation were estimated and distinguished in the hybrid model. In simulation scenarios, the height of vegetation was assumed to be 5 m with canopy leaf area index of 2 and minimum stomatal resistance of 170 s/m. The bulk surface leaf area index (LAI) varied from 0.5 to 5, and fractional vegetation coverage (Fr) varied from 10% to 100%. The vegetation clumpy patterns were quantified by fixing LAI while varying Fr. The climate data was obtained from the Linhe meteorological station located in an arid region in central Inner Mongolia of North China. The results indicated that both the patch and hybrid model performed reasonably well in estimating potential evapotranspiration under homogeneous vegetation distribution conditions. However, the layer model tended to overestimate potential evapotranspiration, as it generally gave higher potential evaporation estimates. The overestimation in potential evapotranspiration by the layer model was increased with the increase of Fr and the decrease of LAI. On the contrary, the patch model had a tendency to underestimate potential evaporation, especially with high Fr and low LAI. For heterogeneous vegetation distribution conditions, potential evapotranspiration estimated from the layer model was generally higher than that given by the patch and hybrid model, particularly with low Fr. Potential evaporation (potential transpiration) from the layer model increases (decreases) with the increase of LAI. However, both variables from the layer model did not change with changes of Fr. In contrast, potential transpiration estimated from the patch and hybrid model was increased with the increase of both LAI and Fr. Potential evaporation from the patch model was increased with the increase of Fr, but kept relative constant under various LAI conditions, while potential evaporation from the hybrid model was increased with the decrease of both Fr and LAI. The above results suggest that the layer model may give reasonable potential transpiration estimates over homogeneous vegetated surfaces, while the patch model is more suitable for surfaces with lower fractional and clumped vegetation cover. By contrast, the hybrid model performs better than both the layer model and the patch model, which can be used to estimate potential evaporation and potential transpiration partitioning for a wide range of surfaces with different vegetation distribution patterns.