Assessing Forest Recreational Potential from Social Media Data and Remote Sensing Technologies Data

How to cite this study

Lingua, F., Coops, N.C. and Griess, V.C. 2023. Assessing forest recreational potential from social media data and remote sensing technologies data. Ecological Indicators 149, 110165.


This study uses Flickr data and remote sensing technologies to identify forest biometric and topographic data to map and estimate recreational potential of British Columbia’s provincial park system. Potential recreation and consumer surplus are mapped in Cypress Provincial Park using variables that influence visitors’ preferences for recreational activities. A machine learning approach called convolutional neural network (CNN) was used to filter Flickr images. The CNN was found to perform best in the classification of images depicting hiking, wildlife, skiing, and biking, but not as well for images depicting camping. The analysis also revealed that consumer surplus in Cypress Provincial Park was primarily concentrated in two hotspots.


This study is relevant to those interested in estimating recreational activities in large natural areas. This information can help inform tourism management and infrastructure planning. Researchers using a forest classification model may find it useful that in this study, the most important variable in improving the accuracy of type of forest recreation predictions was seasonality, followed by topographic variables, and then human impacts on the landscape, and finally forest biometrics variables.


This study is located in British Columbia’s provincial park system which covers approximately 68,000 km². The random forest classifier was used to assess recreational potential specifically in Cypress Provincial Park. Cypress Provincial Park is the most visited provincial park in BC and receives more than 1.5 million visitors each year.

Trail Type

This study covers forests protected by the BC Provincial Park System. BC parks have around 6,000 km of hiking trails and more than 10,000 vehicle-accessible campsites, with more than 20 million visitors annually.


The purpose of this study was to develop a methodology that integrates cultural ecosystem services into forest management plans using remote sensing and social media data.


  • The CNN classified recreation photos as: biking (2.1%), camping (3.5%), hiking (11.2%), wildlife viewing (4.5%), and skiing (5.9%). Climbing and water-related activities totaled 6.6%, but were excluded from the analysis.
  • The CNN performed well in the classification of images depicting hiking, wildlife, skiing, and biking but not as well for images depicting camping.
  • Forests characterized by low slopes favor camping and wildlife viewing, intermediate slopes favor biking, and steep slopes favor hiking.
  • In Cypress Provincial Park, hiking has the most recreational potential in spring, summer, and fall, while skiing has the most recreational potential in winter. Skiing has the highest recreational potential in the southeast portion of the park. Wildlife viewing, camping, and biking have overall low recreation potential in the park.
  • Most of the consumer surplus provided by Cypress Park’s forests is located in two hotspots – one in the southeast of the park and one in the central part of the park. This is where most of the park’s infrastructure is located. However, in most of the park there were no images posted, making it impossible to estimate consumer surplus in these locations.


Geotagged images in BC forested parks were downloaded from Flickr’s API and filtered by a convolutional neural network (CNN) based on the displayed recreational activities from January 1, 2005, through December 31, 2020. Using the geographic coordinates, each image’s topography, forest biometrics (canopy height, forest cover, volume, total biomass), and anthropic impacts (1 indicating no anthropic impact and 44 indicating maximum anthropic impact) were identified and analyzed using the Python package “rasterstats.” 64,670 photos were acquired. 

Using these variables, a random forest classifier model predicted where a recreational activity would occur in Cypress Park, generating a map displaying various probabilities. Maps displaying the recreational monetary value derived from areas in Cypress Park were produced using the random forest classifier. This consumer surplus was calculated using yearly visitation statistics by BC and data from Rosenberger et al. (2017) who reported consumer surplus in forest recreational activities in North America using a meta-analysis of 342 recreation economic studies.

Added to library on November 27, 2023