Utilizing Remote Sensing and GIS to Detect Prairie Dog Colonies

Abstract

The locations of black-tailed prairie dog (Cynomys ludovicianus [Ord]) colonies on a 550-km2 study site in northeastern Wyoming, United States, were estimated using 3 remote sensing methods: raw satellite imagery (Landsat 7 ETM+), enhanced satellite imagery (integration of imagery with thematic layers via a Geographic Information System), and aerial reconnaissance (observations taken from a small plane). A supervised classification of the raw satellite imagery yielded an overall accuracy of 64.4%, relative to ground-truthed locations of prairie dog colonies. The enhanced satellite imagery, resulting from a filtering of the data based on an index derived from the sum of weighted ecological factors associated with prairie dog colonies (slopes, land cover, soil, and ‘‘greenness’’ via the Normalized Difference Vegetation Index) yielded an overall accuracy of 69.2%. The aerial reconnaissance method provided 65.1% accuracy. The highest rate of false positives resulted from the aerial reconnaissance method (39.9%). The highest rate of false negatives resulted from the raw satellite imagery (60.0%), a value that was markedly reduced via the enhancement with ecological data from thematic layers (45.8%). Given the accuracy, interpretability of results, repeatability, objectivity, cost, and safety, the enhanced satellite imagery method is the recommended approach to large-scale detection of black-tailed prairie dog colonies. If a greater accuracy is required, this method can be employed as a coarse filter to narrow the scale and scope of a more costly and laborious fine-scale analysis effectively.