Clean lakes get more visitors than dirty ones, says a University of Minnesota study published earlier this month.
That’s rather intuitive, said the study’s lead researcher Bonnie Keeler, a scientist at the Institute on the Environment.
But the method to get to this conclusion wasn’t.
Forgoing an expensive, time-consuming survey to come to that finding, researchers compared the locations tagged on Flickr photos taken at Minnesota and Iowa lakes with their water quality.
The study is part of a growing trend toward research that intertwines social media and science. And though sites like Flickr, Twitter and Instagram have made research like Keeler’s easier, the sites’ users don’t always accurately represent a population.
Keeler’s team collected almost 42,000 photos taken at about 1,000 lakes to map them and discover which were most popular.
Researchers also found that bigger lakes and ones with boat launches got visitors who were willing to travel farther to get to clean lakes.
The researchers used Flickr because it was the easiest photo sharing site from which to pull photos, Keeler said. But the fact that researchers used a single website may have skewed the data, she said, as Flickr users trend to be female and have high education levels.
“In general, the best work will make an effort to ensure the social media sample used reflects the public,” said Brent Hecht, a University computer science assistant professor who researches social media.
He said some sites aren’t good representations of populations because their users tend to exclude rural populations.
“Twitter is not racially representative,” Hecht said. “Social media sites tend to be urban-oriented.”
Some age groups also tend to be excluded from data collected from social media.
“If I’m a company and I want to see what people think of my product, let me do some analysis on Twitter,” Hecht said. “That’ll give me what some subset of the public thinks of the product, but it won’t, for instance, give me what 70-year-olds and up think of the product.”
But there are upsides to using social media instead of a survey. Surveys are subject to respondents’ memories and biases, Keeler said, so social media sites can be more accurate.
In late January, Carlson School of Management assistant professor Jason Chan released a study indicating a relationship between Craigslist and the prevalence of HIV.
Chan said the trend will only increase as more social media apps like Tinder and Grindr are developed.
“Apps like those are more pervasive to the trend,” he said.
Chan and another researcher looked at the rate of HIV in the central United States before and after Craigslist became popular in those areas. They found the virus was more prevalent after the website allowed people to post hookup advertisements.
Hecht said research like that of Chan and Keeler is becoming more common.
Location data from social media sites, like the data Keeler collected, may inform regional planning and resource management in the future, she said.
“We think there’s a lot of potential to use social media to explore a whole host of questions about where people go,” Keeler said.