Applying Novel Visitation Models using Diverse Social Media to Understand Recreation Change after Wildfire and Site Closure

How to cite this study

White, E.M., Winder, S.G. and Wood, S.A. 2023. Applying novel visitation models using diverse social media to understand recreation change after wildfire and site closure. Society & Natural Resources 36(1): 58-75.


This study analyzes how visitation in the Columbia River Gorge of Oregon and Washington changed as a result of the 2017 Eagle Creek Fire. A visitation model is constructed by incorporating data from social media to measure visitation patterns and used to look for a substitution effect on nearby recreation sites after the fire.


This study is relevant to those interested in understanding how recreational patterns can change after a natural disturbance and how social media can be used to improve these estimates. Social media can be especially useful at locations where on-site data is inconsistent or unavailable.


This study is located on the Columbia River Gorge in Oregon and Washington.

Trail Type

The Columbia River Gorge has a wide variety of trails, developed recreation sites, campgrounds, waterfalls, and wilderness areas. Among the 14 study sites impacted by the 2017 Eagle Creek fire, eight were located within the fire while the remaining six were located adjacent to it. All six sites reopened within a year after the fire whereas the sites within the burned area remained closed for the duration of this two-year study. Most of the sites were singletrack non-motorized trails.


The purpose of this study was to use social media data to measure visitation levels before and after the Eagle Creek Fire in the Columbia River Gorge. This work was funded under a USDA Forest Service Pacific Northwest Research Station Joint Venture Agreement.


  • Using social media data, seasonal data, and precipitation data, the model described more than 75% of the observed variability in weekly visitation at sites where the model was parameterized in the Mount Baker-Snoqualmie National Forest.
  • Visitation increased in the years before the fire at the high-use sites and medium-use sites. The rate of increase varied between sites, ranging from 800 additional visits per year (Site 33 Coyote Wall) and 29,000 additional visits per year (site 7 Oneonta Gorge)
  • Total visits to all 41 sites declined in year 1 after the fire (Sept 2017- Aug 2018). Visitation in year 1 was 60% of that observed in the previous year. Visitation in year 2 after the fire (Sept 2018 – August 2019) increased but remained below the visitation levels observed in the year before the fire. 
  • If the fire had not occurred, 481,000 visits were predicted in year 1 and 581,000 were predicted in year 2 at the sites that were closed after the fire. Once the fire occurred, the sites that reopened in the first year experienced only 4,000 visitations, demonstrating a loss of 434,000 potential visits. Visitation at study sites that remained open after the fire continued to grow in the years following the fire.
  • In general, visitation at sites that were open after the fire continued their pattern of year-to-year increases in visitation, consistent with pre-fire trends. There was little evidence of displacement of visits. 
  • Visitation rebounded to pre-fire trends at all but one (Site 8) of the six periphery sites that reopened after the fire. Access to Site 8 (Horsetail Falls), although allowed for 11 months in Year 2, was limited to a relatively short (1.4 km) two-waterfall loop. Proximity to a burned area did not directly result in a reduction in visitation.


A total of 41 sites were selected in the Columbia River Gorge, including sites affected by the Eagle Creek Fire and unaffected sites that likely served as substitutes. To choose sites that were likely alternatives to the sites affected by the fire, site selection was guided by input from local USFS recreation managers, guidebooks, and several public lists of recreation sites. Visitation was estimated by using the model structure detailed in Wood et al. 2020, which quantifies the relationship between on-site counts of visitors, precipitation, holiday, seasonality, and volume of social media shared by visitors to each site in a multiple linear regression. The model was parameterized with data collected from 2016 and 2018 at 27 recreation sites in the Mount Baker-Snoqualmie National Forest (MBSNF). Flickr, Twitter, Instagram, and All Trails were used to estimate the number of visits to each study site between September 2014 and August 2019. For all social media data, the number of user days (unique social media users who post each day) were calculated per week, per site. Finally, a linear model was used to estimate what visitation would have been without the fire.

Added to library on November 27, 2023