Identifying User Types and Long-Stay Behavior on a National Forest Trail: An Integrated Analysis of GPS Trajectories and Geotagged Images

Increasing forest-trail use requires empirical identification of patterns that conventional surveys may miss, including prolonged stationary periods that may be associated with camping or backpacking. This study examined activity records from the Seonjaryeong section of Korea’s Daegwallyeong Forest Trail, combining GPS movement data with geotagged photographs linked at the activity-record level. Of 6505 GPS records collected in 2021, 3105 remained after cleaning. Stay point detection and K-means clustering identified three user types: short-distance (89.5%), long-distance (9.1%), and long-stay (1.4%). We also analyzed 1410 geolocated photographs using the Google Cloud Vision API. The long-stay type had an exceptionally long mean rest time (930 min) and distinct movement behavior. No significant difference in sex composition was detected, whereas age composition (with a small effect) and seasonal composition differed significantly among types. In the available photographs, forest and vegetation dominated object labels for all three types; second-ranked categories differed but rest on small, untested subsamples, two dominated by single contributors. The long-stay group, identified from GPS trajectories alone and defined by prolonged stationary time rather than verified on-site stays, represents a pattern current management systems do not explicitly address; the photographs add exploratory information on where some records stopped.