How to cite this study
Wartmann, F.M., Baer, M.F., Hegetschweiler, K.T., Fischer, C., Hunziker, M. and Purves, R.S. 2021. Assessing the potential of social media for estimating recreational use of urban and peri-urban forests. Urban Forestry & Urban Greening 64(2021),127261.
This study analyzes data from Twitter, Flickr, and Instagram to assess visitation at 10 urban and peri-urban Swiss National Forest Inventory sites and evaluates recreational models on a national scale. Social media text was analyzed and a Flickr-based model and Twitter-based model were compared with a pre-existing potential recreation demand (PRD) model to estimate recreational use. Both the Flickr and Twitter models correlated with the existing PRD model; the Instagram model was not evaluated due to data limitations.
This study is relevant to researchers interested in using social media data to estimate visitation. A limitation of this study is that most social media posts were referenced in nearby urban areas or at forest edges, and fewer in specifically forested areas, especially on smaller mapping scales. This suggests that social media data may not be well-suited for informing specific management in smaller forest recreation areas due to limited data availability.
Study sites were located across Switzerland.
The 10 study sites (1.4 km x 1.4 km) from the Swiss National Forest Inventory (NFI) had different forest compositions, including mixed forest, deciduous forest, and coniferous forest, and were either urban, peri-urban, or rural. All sites showed a high recreational use potential, according to the potential recreational demand (PRD) model used by the Swiss NFI.
The purpose of this study was to evaluate the potential of social media data for assessing recreation in urban and peri-urban forests in Switzerland. The research for this paper was financially supported by the Swiss Federal Office for the Environment (FOEN).
- Though the sample sizes from the three social media platforms are different, the datasets are all significantly correlated. There was a higher concentration of social media data around urban centers and attractive places for tourists, while less popular alpine locations had consistently less data across all platforms.
- In the three sites analyzed for Instagram text data, the top three terms describing landscape elements, qualities, or activities were: Zurich Uetliberg – “winter,” “snow,” “cars”; Neuchâtel – “beautiful,” “nature,” “lake”; and Locarno – “autumn,” “lake,” “view.”
- Models based on Flickr and Twitter data correlated with the existing PRD model. Instagram data was excluded from this analysis since the researchers were unable to collect data at the national scale.
- The recreation model based on Flickr determined recreational use to be higher in areas with lower population densities and lower in areas with high touristic infrastructure compared to the PRD model. This revealed that the Flickr model provided a better estimate of actual use in areas where census data is low but many people still visit.
- Calibrating the Flickr and Twitter models and PRD model for forested cells only (rather than all of Switzerland) resulted in a slightly stronger correlation between the two models for estimates of recreation use. This is likely because forest plots do not contain residential areas, which typically have high counts of social media data and therefore higher discrepancies between social media data counts and potential recreation demand.
The data availability between Twitter, Flickr, and Instagram were analyzed in 10 National Forest Inventory sites in Switzerland. The data was gathered between October 2017 and January 2018. 2,590 geotagged photos were obtained from Flickr, 4,106 from Twitter, and 208,285 from Instagram. Since the National Forest Inventories sites are not always near recreational infrastructure, radiuses of 250, 500, 1000, 2000, 3000, 4000, and 5000 m were used to include recreational activities. National Forest Inventories (NFIs) offer county-wide information on forests.
Differences in text data from Instagram were analyzed across three peri-urban forest sites in three different language areas (German-speaking Zürich Uetliberg, French-speaking Neuchâtel, and Italian-speaking Locarno). Posts within a radius of 5000 m that were within the forest polygons were extracted. Extraction was in English to allow comparison across all three platforms. The most frequently listed words were displayed in word clouds. A model of recreation was calculated based on data from Twitter and Flickr and compared to an existing model of potential recreation demand that was based on census data and accessibility (the PRD model) for all of Switzerland. The same cells as the PDR model were used, based on the 1.4 km spacing of the NFI. A regression model was developed for both platforms. Instagram data was not used to create a model since the researchers were unable to collect data at the national scale and instead, a text analysis was conducted.
Added to library on November 27, 2023