Abstract: Soil moisture is an important factor for grassland vegetation growth and its measurement is a critical task for drought monitoring systems. In order to monitor moisture in Xilingol grassland, monthly evapotranspiration (ET) data from MODIS MOD16A2 and measured soil moisture data from ground monitoring stations were collected. Correlation and regression analyses were employed to establish a retrieval model for soil volumetric moisture (SVM) based on the evapotranspiration deficit index (ETDI): SVM=-48.851×ETDI+54.669. The root mean square error (RMSE) of this model was 3.27%. This model can be used to retrieve soil moisture at regional scale. The thresholds of drought grades were established based on the national standard and used to analyze drought dynamics in Xilingol grassland over the past 15 years (2000-2014). The results showed that SVM fluctuated in 14% of the meadow steppe, approximating normal levels except for the slight droughts in 2007 and 2009. SVM fluctuated in 11% of the typical steppe and the sandy vegetation areas, as well as in the Xilingol grassland as a whole, approaching slight drought conditions except for the moderate droughts recorded in 2007 and 2009. SVM fluctuated in 8% of desert steppe, approaching severe drought conditions except for the moderate droughts recorded in 2002, 2003 and 2012. On average, over the past 15 years some 66% of the Xilingol grassland has experienced drought conditions, though to varying degrees. Non-drought and severe drought areas increased, while slight and moderate drought areas decreased, but significant tests indicate that none of the changes were significant (P>0.05).