Airborne LiDAR and Aerial Imagery to Assess Potential Burrow Locations for the Desert Tortoise (Gopherus agassizii)

Abstract

The Southwestern United States desert serves as the host for several threatened and endangered species, one of which is the desert tortoise (Gopherus agassizii). The goal of this study was to develop a fine-scale, remote-sensing-based approach that indicates favorable burrow locations for G. agassizii in the Boulder City (Nevada) Conservation Easement area (35,500 ha). This was done by analyzing airborne LiDAR data (5–7 points/m2) and color imagery (four bands, 0.15-m resolution) and determining the percent vegetation cover; shrub height and area; Normalized Difference Vegetation Index (NDVI); and several geomorphic characteristics including slope, azimuth, and roughness. Other field data used herein include estimates of canopy area and species richness using 1271 line transects, and shrub height and canopy area using plant-specific measurements of ~200 plants. Larrea tridentata and Ambrosia dumosa shrubs were identified using an algorithm that obtained an optimum combination of NDVI and average reflectance of the four bands (IR, R, G, and B) from pixels in each image. The results, which identified more than 65 million shrubs across the study area, indicate that percent vegetation cover from aerial imagery across the site (13.92%) compared favorably (14.52%) to the estimate obtained from line transects, though the LiDAR method yielded shrub heights approximately 60% those of measured shrub heights. Landscape and plant properties were combined with known locations of tortoise burrows, as visually observed in 2014. Masks were created using roughness coefficient, slope percent, azimuth of burrow openings, elevation, and percent vegetation cover to isolate areas more likely to host burrows. Combined, the masks isolated 55% of the total survey area, which will help target future field surveys. Overall, the approach provides areas where tortoise burrows are more likely to be found, though additional ecological data would help refine the overall method.