Airborne laser altimetry and multispectral imagery for modeling Golden‐cheeked Warbler (Setophaga chrysoparia) density

Robust models of wildlife population size, spatial distribution, and habitat relationships are needed to more effectively monitor endangered species and prioritize habitat conservation efforts. Remotely sensed data such as airborne laser altimetry (LiDAR) and digital color infrared (CIR) aerial photography combined with well-designed field studies can help fill these information voids. We used point count-based distance sampling survey data and LiDAR-fused CIR aerial photography to model density of the Golden-cheeked Warbler (Setophaga chrysoparia), an endangered songbird, on the 10 000-ha Balcones Canyonlands National Wildlife Refuge (BCNWR). We developed a novel set of candidate models to explain Golden-cheeked Warbler detection probability and density using habitat covariates characterizing vegetation structure, composition, and complexity as well as habitat fragmentation, topography, and human infrastructure. We had the most model support for covariates calculated using focal means representing a 3.2 ha territory size (100 m radius) vs. 1.8 and 7.0 ha territory sizes. Detection probability decreased with canopy cover and increased with topographic roughness. Golden-cheeked Warbler density increased with canopy cover, was highest at a 7:3 ratio of Ashe juniper (Juniperus ashei) to broadleaf tree canopy cover, and decreased with global solar radiation. Predicted warbler densities using 3 min point counts were similar to six estimates from independently collected warbler territory mapping on BCNWR with a mean difference of 6% and a Root Mean Squared Error of 1.88 males/40 ha. The total population size for BCNWR was estimated at 884 Golden-cheeked Warbler males (95% CI 662, 1206) and predicted densities across the refuge ranged from 0.0 to 0.50 male warblers per ha. On the basis of observed habitat relationships, we defined high quality habitat as having at least 60% canopy cover with Ashe juniper comprising 50-90% of the canopy. We estimated 48% of the area at BCNWR managed for Golden-cheeked Warblers was in high quality habitat conditions and identified patches within the lower habitat quality areas (14% of warbler management areas) that had the greatest potential to become high quality habitat with management. Our approach combined robust wildlife surveys with highly scalable remotely sensed data to examine habitat relationships, estimate population size, and identify existing areas of high quality habitat. This method can be applied to other species of conservation interest and can be used with multiple years of remotely sensed data to assess changes in habitat at local to regional scales.