Measuring Recreational Visitation with Crowd-Sourced Photographs

Thesis (Master's)--University of Washington, 2015In the age of big data and internet social media sites, there are myriad opportunities for researchers to leverage crowd-sourced online data for research purposes. One such opportunity is the possibility of using photos posted online to infer information about visitors to recreational areas. In this paper, I assess the validity of using data from photos posted on the website Flickr to infer information about visitors to National Parks in the Western United States. By comparing the photo data to statistics published by the National Park Service (NPS), I examine the relationship between the two datasets and if used properly, whether the photos can yield similar information to that provided by the Park Service. I address two aspects of visitor use data: the number of people visiting a park and where those visitors reside. Together, researchers can use these two datasets in conducting travel cost studies to estimate the economic recreational value of a site. Using multiple regression analysis, I build a statistical model whereby I can infer the count of visitors to a park in a given month through the number of photos posted online. Overall, I find that a one percent increase in “photo-user-days” corresponds to a 0.65 percent increase in NPS visitation, holding all else constant. To evaluate the validity of using photos to infer visitor’s home origins, I compare information from photo-taker’s profiles to survey data provided by the NPS. I find that the photos give an accurate big-picture view of visitor origins and moreover, provide an accurate view of the distance that visitors live from the park. Using findings from my analyses, I estimate the travel costs incurred by in-state visitors to reach Mount Rainier National Park in 2012. Using data from the NPS, I find in-state visitors incur an average cost of $61.38 per person and a total cost of 36.31 million US dollars across all visitors. Using the photo data, I estimate the average cost incurred is $58.90 per person, with a total cost of 37.20 million US dollars (95% CI 29.32-45.01 million). This finding shows that the photo data can produce a strikingly similar economic estimate to those obtained by more traditional methods. The ability to accurately infer visitor use information with photos has the potential to reduce costs for land managers and opens new opportunities for research on the recreational value of lands. As visitor use information is expensive and time-consuming to obtain, this method has significant potential to save time and money for the Park Service. In addition, other land managers and researchers can apply this method to evaluate visitation to remote or hard-to-measure locations that currently lack detailed visitor use data. Overall, this work represents a crucial first step in evaluating a new and exciting means to measure recreational use value