蒸散（Evapotranspiration,ET）是生态系统水循环中的重要一环,决定了生态系统水分和热量传输。从区域尺度对蒸散及其蒸腾（Transpiration,T）和蒸发（Evaporation,E）组分进行量化,认识环境因素对其的影响机制,有助于合理利用、分配水资源,为研究气候变化对区域生态系统水文循环的影响提供参考。基于生态系统生产力模拟（Boreal ecosystem productivity simulator, BEPS）模型,验证模型在研究区域的适用性,量化1981—2018年内蒙古半干旱区的ET及其组分的变化情况,并对其进行归因分析。结果表明：经不同数据验证,BEPS模型计算结果能够精确反应研究区域ET及其组分的分布情况和变化趋势。1981—2018年研究区草地、农田和森林多年平均ET分别为278.22 mm、362.50 mm和308.81 mm。E、T和ET多年呈显著上升趋势,上升速率分别为0.42 mm·a^-1、0.63 mm·a^-1和1.05 mm·a^-1。ET与T在全区域内空间分布格局相似,与E相反,ET年际波动主要受到T年际波动的影响。综合影响因子的变化和E、T、ET对因子的敏感性,研究区域草地和农田T和ET以及森林的ET主要受到饱和水汽压差（VPD）和平均气温（TEMP）变化的控制。农田和森林归一化植被指数（Normalized difference vegetable index, NDVI）都呈减小趋势,但森林环境T对NDVI的变化更加敏感,因此负贡献更大。

Evapotranspiration (ET) is the most important flux term in the ecosystem water budget because it is related to water and heat exchange between ecosystems and the atmosphere. Quantifying the magnitude and dynamics of ET and its components include evaporation (E) and transpiration (T) at regional scales and deriving a mechanistic understanding of the underline regulations is essential for managing water resources under the changing climate. On the basis of a mechanistic ecosystem model (BEPS), we estimated the magnitudes of ET and its components in the semi-arid region (40.16°-46.77°N and 110.33°-119.91°E) of Inner Mongolia, China from 1981 to 2018. Compared with different data, the BEPS could provide accurate estimates for changes in ET and its components by validating with ground measurements. The total average annual ET of grassland, cropland, and forests in the study area from 1981 to 2018 was 278.22 mm, 362.50 mm, and 308.81 mm, respectively. Over the study period, E, T, and ET increased at a rate of 0.42 mm·a^-1, 0.63 mm·a^-1, and 1.05 mm·a^-1, respectively. Across the study region, we found that the spatial distribution patterns of T was opposite to that of E and similar to that of ET. From the interannual variation standpoint, the interannual fluctuations in ET were primarily influenced by those in T. Considering the changes in influencing factors and the sensitivities of the three fluxes, we concluded that T and ET of the grasslands and croplands and ET of forests were mainly controlled by vapor pressure deficit and air temperature. Although both cropland and forest normalized difference vegetation index (NDVI) showed decreasing trends, forest environmental T was more sensitive to changes in NDVI, thus receiving more negative influence than cropland.