Using remotely sensed data to map Joshua Tree distributions at Naval Air Weapons Station China Lake, California, 2018

Species distribution models (SDMs) that are derived through inference have been used to provide important insights toward species distributions. Their inferences can be robust in relation to known presences, but SDMs have error rates that cannot be quantified with certainty. For large plant species with unique signatures and in sparsely vegetated habitats, object-oriented satellite image interpretation provides a useful alternative to the more commonly used SDM approach. We tested visual image interpretation techniques in a pilot project to map the distribution of the Joshua tree (Yucca brevifolia), an arborescent succulent plant endemic to the Mojave Desert of North America. Naval Air Weapons Station China Lake (NAWS–CL) required assistance in mapping the distribution of Joshua trees across the 4,715 square kilometer (km²) military installation in support of their national security mission. Joshua trees were present on 1,307 1-km² cells in the species distribution model, or 27.7 percent of the military installation. This increases the published range of Joshua trees at NAWS–CL by 90 percent and corrects for two stands of Joshua trees that were previously identified but do not exist. Remotely sensed satellite data in combination with ground surveys of Joshua trees produced a more accurate distribution map at a 1-kilometer resolution than did previous SDMs based on correlative modeling (area under the curve [AUC] 0.9064 versus 0.5848, respectively). Ancillary comparison with light detection and ranging (lidar) data indicated that satellite and lidar data were equally successful with slightly different sources of error, but that using them in combination produced the best results.