Twitter can be used as an effective tool to predict air quality levels in areas affected by wildfires, a recent study suggests, and the social media app could potentially aid in assisting with rescue and relief efforts.
The study, a result of efforts from two U.S. Forest Service scientists, sorted through over 39,000 tweets specifically referencing the 15 wildfires which wreaked the most havoc on California during the summer of 2015 or including the words “wildfire” or “smoke.”
The study aimed to determine whether data could be crowdsourced and used to estimate air quality impacts from smoke in areas where air monitoring stations might not be present.
“Smoke from wildfires is a huge concern and the most robust ways of measuring air quality impact … comes from physical monitoring stations that are maintained by the EPA,” said Sonya Sachdeva, a computational social scientist with the U.S. Forest Service and a lead researcher on the study. “Those monitoring stations ... can’t be everywhere.
Sign Up and Save
Get six months of free digital access to The Sacramento Bee
Tweets were “geocoded,” or association with a location based on metadata looking at where the tweet was sent from or where the tweeter is located. These tweets were then assigned to a local air quality monitoring station in an area nearby.
Using air pollution levels reported by the Environmental Protection Agency on the day the tweet was posted, the tweets were linked to air quality data.
A previous study was conducted using 700 tweets about the 2014 King Fire. Both the 2014 study and the latest study corroborate findings that tweet frequency can be a good estimate of particulate matter 2.5.
PM 2.5 is a tiny air particle smaller than 2.5 micrometers and invisible to the human eye, released by vehicles and factories and present in wildfire smoke. Particles are so small, they can pass through the lungs and enter the bloodstream, posing significant health risks.
“People that were closest to the location of the fire … seemed to talk about distinct topics than people that were farther,” Sachdeva said in reference to findings from the King Fire study. “People that were closest to the fire were more likely to discuss those air quality impacts, more likely to discuss smoke being visible.”
Although some academic researchers are using social media to predict air quality, the use is not widespread among agencies.
In Shasta County, wildfires continuing to blaze are forcing residents out of their homes. John Waldrop, the air quality manager for the the county’s Department of Resource Management, said the department does not use social media for notifications.
Researchers hope to develop a streamlined technology to monitor the content of tweets which could potentially provide efficient, real-time air quality predictions.
“The more precise you can be in terms of the air quality for certain areas, the more targeted advice,” said Sarah McCaffrey, a research forester with the Forest Service and another lead researcher on the study. “That always makes it easier for people to respond and protect their health appropriately.”
With this technology, Sachdeva predicts the creation of heat maps showing the location of people requesting help via social media — “that’s where rescue efforts could be directed more systematically,” she said.
“Or you could help coordinate all the helping behavior,” McCaffrey said. “Connect people who need assistance to those who want to give assistance.”