This paper uses the Community Land Model (CLM3.0) coupled with a modified Dynamic Global Vegetation Model (DGVM)to investigate the impact of vegetation interannual variability on global annual evapotranspiration. Two sets of off-line numerical experiments are designed. In the default experiment, the modified CLM-DGVM is run 600 years, cycling driven by the observed atmospheric forcing data of 1950-1999. In this simulation, vegetation has interannual variation. It is denoted as DGVM simulation, and the results from the last 50 years are analyzed. For comparison, a 50-year CLM only simulation (i.e., without coupling to DGVM) is performed, initiated by the states at 550th year of the DGVM simulation and forced by the same atmospheric forcing data. The vegetation parameters such as fractional coverage (FC) and daily leaf area index (LAI) are taken from the climatology values derived from the last 50 years of DGVM simulation, and hence they have no interannual variations. It is denoted as the CLM simulation. Results show that: (1) FC interannual variability is largest over grassland and is smallest over forest regions. The grassland and shrubland have relatively large LAI interannual variability. Over regions dominated by deciduous trees, LAI interannual variability is greater in spring and autumn than other seasons due to the emergence and senescence of leaves. (2) Over most forest regions, e.g., the Amazon region, central of Africa, southeast of the United States, Europe, and southeast of China, vegetation interannual variabilities lead to the increasing in total evapotranspiration, which is resulted from changes of its three components, that is, increasing in ground evaporation and decreasing in the canopy evaporation and transpiration. Over shrubland and grassland, changes in evapotranspiration and its components are roughly opposite as in the forest regions. (3) Over low latitude, interannual variabilities of LAI in different seasons lead to strongly seasonal variations of evapotranspiration as well as its components. Over mid-latitude of North Hemisphere, the timing of largest changes of evapotranspiration shift from March to June as the latitude increases. All three components of evapotranspiration exhibit the same spatial and temporal variability. (4) Larger interannual variability of FC and LAI lead to larger differences in both evapotranspiration and its components. Particularly, when FC interannual variability is over 30% or LAI interannual variability is over1.6m²/m², the differences in ground evaporation and canopy evaporation are negative, while the difference in transpiration is positive, and as a result, the total evapotranspiration decreases. For a case study, 50 year variation of the evapotranspiration and its components in a single grid cell (71°W,18°N) are analyzed. During the years when this area is dominated by grass, ground evaporation decreases and canopy evaporation and transpiration increase, while in other years when it is dominated by bare soil, the three components of evapotranspiration change oppositely as the former one. These results imply that vegetation interannual variabilities induce different responses among the partitions of evapotranspirations, which then alter the total evapotranspiration. This conclusion is especially important in the semiarid grassland and shrubland areas where climate interannual variability is relative large, and the ecosystem is fragile, which is easily affected by climate change and environment.