Incorporating Shrub and Snag Specific LiDAR Data into GAP Wildlife Models

Abstract The U.S. Geological Survey’s Gap Analysis Program (hereafter, GAP) is a nationally based program that uses land cover, vertebrate distributions, and land ownership to identify locations where gaps in conservation coverage exist, and GAP products are commonly used by government agencies, nongovernmental organizations, and private citizens. The GAP land-cover designations are based on satellite-derived data, and although these data are widely available, these data do not capture the 3-dimensional vegetation architecture that may be important in describing vertebrate distributions. To date, no studies have examined how the inclusion of snag- or shrub-specific Light Detection and Ranging (LiDAR) data might influence GAP model performance. The objectives of this paper were 1) to assess the performance of the National GAP models and Northwest GAP models with independently collected field data, and 2) to assess whether the inclusion of 3-dimensional vegetation data from LiDAR improved the performance of National GAP and Northwest GAP models. We included only two parameters from the LiDAR data: presence or absence of shrubs and presence or absence of snags ≥25 cm diameter at breast height. We surveyed for birds at>150 points in a 20,000-ha coniferous forest in northern Idaho and used data for eight shrub- and cavity-nesting species for validation purposes. On a guild level, National GAP models performed only marginally better than Northwest GAP models in correct classification rate, and LiDAR data did not improve vertebrate distribution models. At the scale used in this study, GAP models had poor predictive power and this is important for managers interested in using GAP models for species distributions at scales similar to ours, such as a small park or preserve <200 km2 in size. Additionally, because the inclusion of LiDAR data did not consistently affect the performance of GAP models, future studies might consider whether LiDAR data affect GAP model performance by examining 1) different spatial scales, 2) different LiDAR metrics, and/or 3) species-specific habitat relationships not currently available in GAP models.