An open‐source image classifier for characterizing recreational activities across landscapes

Snapshots of Nature-Based Recreation Across Rural Landscapes: Insights from Geotagged Photographs in the Northeastern United States

Nature-based recreation is of increasing economic importance to rural communities transitioning away from traditional natural resource extraction. Rural areas are rich in cultural ecosystem services (CES) that function as essential public goods, providing benefits to livelihoods and fostering conservation of large landscapes. Geotagged photographs from social media platforms capture detailed information on public use of natural areas that can be useful for stakeholders interested in promoting resilient social-ecological systems, but applications of such "big data" are often limited in either spatiotemporal or thematic scope. We integrated multiple aspects of crowdsourced image data to better understand human-environment interactions, focusing on the Northern Forest, a culturally and ecologically significant region of the northeastern United States. Images for the region were mined from Flickr, a crowdsourcing image platform, from 2012 to 2017 and assigned themes via a neural network-cluster analysis pipeline while geographic drivers of nature-based visitor engagement were assessed with random forest models to predict use of CES across the Northern Forest. Daily and seasonal patterns in photography were broadly consistent throughout the region, whereas annual photography trends were more variable. Eleven core themes were identified in images, with 70% of photographs depicting activities related to CES. Visitor engagement with nature was greater near roads and shorelines, and at higher elevations, reflecting tensions between accessibility and aesthetics for recreation. Automated image classification tools can rapidly extract relevant information from crowdsourced photography for exploring human-environment interactions, but researchers should consider multiple spatial and temporal scales in CES assessments of large social-ecological systems.