Ecological vulnerability evaluation is a prerequisite to conducting ecological restoration. The common ecological vulnerability evaluation methods have some applicability problems in arid pasturing areas: 1) the evaluation indicators are numerous, and the data are usually difficult to obtain; 2) an administrative region, such as a city or a country, is often used as the evaluation unit, which is too large and lacks specific significance. Commonly used data, such as meteorological data, soil data, land-use data, and water resource development and utilization data, as well as the simulation results in the SWAT model, are used in this research to build an ecological vulnerability evaluation of arid pasturing areas based on the SWAT model. Ecological vulnerability of arid pasturing areas has been divided into ecological vulnerability of the natural ecosystem and ecological vulnerability of the human-nature coupling system. Drought index, vegetation coverage, soil bulk density, and evapotranspiration capacity of soil water are used to depict ecological vulnerability of natural ecosystem; evenness index, contagion index, ecological suitability index, and water utilization modulus are used to depict the human-nature coupling system. Firstly, the SWAT model has been used to divide sub-basins and hydrological simulations by the combination of observed data and simulated data, and the association of the statistic data with spatial land type. All indicators have been set based on GIS, linear piecewise membership functions have been adopted to calculate index values, and analytic hierarchy process (AHP) is adopted to determine the weighs of criteria layers and index layer. The ecological vulnerabilities of the research area are divided into five levels. This evaluation model was used in the Aibugai Basin (which is located in arid pasturing areas in Northwest China) to evaluate the change in ecological vulnerability caused by land-use variation between 1990 and 2010, and 17 sub-basins were divided according to demand. The results showed that, during a period of 20 years, both the entire research area and each sub-basin went through a CIEV (comprehensive index of ecological vulnerability) decrease to some degree. Sub-basin 9, 10, 11, 12, and 15 developed a moderate vulnerability level from a slight vulnerability level in 1990, and sub-basin 16 and17 have changed to a sever vulnerability level from a moderate vulnerability level in 1990. The CIEV of the entire research area decreased to 0.59 in 2010 from 0.70 in 1990. The northwestern part of the research area had better ecological stability, and the degradation in ecological vulnerability levels took place mainly in the agro-pastoral ecotone located in the southeast of the research area. With the expanding demands of grain and cattle, agricultural areas have gradually expanded northward in the past 20 years. Declines in grassland areas and overdeveloped water resources caused by farmland expansion were the main reasons that caused the ecological degradation in the research area. Closely tied to the relationship with water, grass and cattle in arid pasturing areas to establish indicators, this evaluation model effectively reflected the aggravation of ecological vulnerability caused by cropland expansion in the research area from 1990 to 2010.