Can camera traps be used to differentiate species of North American flying squirrels?

Abstract

AbstractCamera traps are becoming an increasingly important tool to survey wildlife populations. However, the application of camera trapping for reliable species identification between nondistinctive, morphologically similar sympatric species is untested for most small mammals, including North American flying squirrels (Glaucomys spp.). Camera traps are a successful monitoring technique where flying squirrel species are allopatric, however there are zones of sympatry between Humboldt's flying squirrel (HFS, G. oregonensis) and northern flying squirrel (NFS, G. sabrinus) in the Pacific Northwest and NFS and southern flying squirrels (SFS, G. volans) in eastern North America. We used camera trap data collected during flying squirrel surveys in 2013–2020 at 59 sites in California, North Carolina, and Virginia, USA, to determine if a reliable method could be used to differentiate the species. With a subset of 100 high‐quality, independent capture events per species (50 of dorsal views, 50 of lateral views), we used body measurements and pelage characteristics to differentiate species using random forest classification models. Our models predicted species identification accuracy rates of 90.9% for dorsal views and 68.2% for lateral views. Species misclassification rates between HFS and NFS were 23.5% for dorsal views and 26.5% for lateral views, whereas misclassification rates between NFS and SFS were 16.6% for dorsal views and 5.7% for lateral views. Although misclassification rates were lower than we expected between NFS and SFS, we are cautious about recommending camera trapping as opposed to ultrasonic acoustics for North American flying squirrel species identification, especially in areas of species sympatry due to potential false positive rates and high numbers of unusable images. However, camera trap settings and white‐flash photography should be explored to determine their potential for improving flying squirrel species classification rates and the percentage of usable photos.