Brandt's vole (Lasiopodomys brandtii) is a common rodent pest in the steppes of Inner Mongolia, China. To explore the potential mechanisms of Lasiopodomys brandtii occurrence, and develop effective survey techniques on a large-scale, a field experiment was conducted in the steppes of East Ujimqin Banner. Images were collected using an unmanned aircraft system and stitched from two different habitat plots of Brandt's vole. Each image was divided into 1hm² grids, and the numbers of overwintering colonies were identified. Environmental factors and geospatial data were extracted from remote sensing images and the correlation between these factors and colony density was analyzed using Pearson's test and a generalized additive model (GAM). The results showed that vegetation (0.1 ＜ NDVI ＜ 0.18) indices can indicate the selection of overwintering colonies in Brandt's vole, with voles preferring sites with sparse and low vegetation to build overwintering colonies. NDVI, especially in September and October, was found to be highly correlated with colony density. High density colonies were mainly found in areas with specific NDVI ranges from 0.25 to 0.35 in September, and there was a significant negative correlation between NDVI and colony density in October. Furthermore, high density colonies were concentrated on southeast slight gradients (0.3°-1.6°) with a relatively high altitude (895m-915m). These findings provide valuable insights for predicting the selection of overwintering colonies and managing Brandt's vole populations in spring.