Characterizing Off-Highway Road Use with Remote-Sensing, Social Media and Crowd-Sourced Data: An Application to Grizzly Bear ( Ursus Arctos ) Habitat

• Published in: Remote Sensing
• Main Authors: Sean P. Kearney, Terrence A. Larsen, Tristan R. H. Goodbody, Nicholas C. Coops, Gordon B. Stenhouse
• Format: Article in Journal/Newspaper
• Language: English
• Published: MDPI AG 2021
• Subjects: spatial ecology; geotagged social media data; circuit theory; road ecology; road use; telemetry; Science; Q; Ursus arctos
• Online Access: https://doi.org/10.3390/rs13132547; https://doaj.org/article/314096ef331a460f9c81efb36feb7d41

Characterizing roads is important for conservation since the relationship between road use and ecological impact can vary across species. However, road use is challenging to monitor due to limited data and high spatial-temporal variability, especially for unpaved roads, which often coincide with critical habitats. In this study, we developed and evaluated two methods to characterize off-highway road use across a large management area of grizzly bear ( Ursus arctos ) habitat using: (1) a ‘network-based’ approach to connect human activity hotspots identified from social media posts and remotely detected disturbances and (2) an ‘image-based’ approach, in which we modeled road surface conditions and travel speed from high spatial resolution satellite imagery trained with crowd-sourced smartphone data. To assess the differences between these approaches and their utility for characterizing roads in the context of habitat integrity, we evaluated how behavioural patterns of global positioning system (GPS)-collared grizzly bears were related to road use characterized by these methods compared to (a) assuming all roads have equal human activity and (b) using a ‘reference’ road classification from a government database. The network- and image-based methods showed similar patterns of road use and grizzly bear response compared to the reference, and all three revealed nocturnal behaviour near high-use roads and better predicted grizzly bear habitat selection compared to assuming all roads had equal human activity. The network- and image-based methods show promise as cost-effective approaches to characterize road use for conservation applications where data is not available.