From Sept. 8-10, Jared Smith, the man behind the prolific Twitter account, sent out more than 200 tweets, including retweets of local government agencies, preparation advice, and forecasts for severe weather in the area. Then, when the effects of Irma reached Charleston on Sept. 11, Smith switched over into wall-to-wall coverage of the storm, racking up almost 500 tweets and retweets in just 24 hours. But while Smith worked to deliver continuous storm updates to his more than 38,500 followers, others were spreading rumors and outright lies.
Along with the false claims of sharks swimming through the flooded streets of Miami and permanent closures of Disney World attractions was the fabricated story of a USC student impaled by storm debris in Columbia. As Irma’s effects began to die down across South Carolina, Sarah Ellis, reporter with The State newspaper, tweeted “FYI, this is ACTUAL fake news. No such thing happened. But neat photoshop job, I guess?” along with a screenshot of the online article with the headline “USC Student Severed by Telephone Pole.”
The altered image was based as an article posted on The State’s website detailing a downed tree that displaced four students in an apartment building owned by Gov. Henry McMaster. Other than the change in headline, the beginning of the fictitious article was identical to the original until it claimed that a student had been declared dead after a loose tree limb penetrated her torso and pierced her aorta.
FYI, this is ACTUAL fake news. No such thing happened. But neat photoshop job, I guess? pic.twitter.com/y9uLKPQES9— Sarah Ellis (@ellissk04) September 11, 2017
According to Ellis, who shared a byline on the original story, the fake was sent to her by a source who had received the article from someone else. Oddly enough, the name of Ellis’ co-author on the piece had been switched out with the name “Chrisitan Bruder,” who if he exists isn’t part of The State’s newsroom.
With anonymous individuals taking such lengthy steps to spread misinformation online, Smith of @chswx says his first focus when reporting on severe weather events is simple — verify, verify, verify.
“I have a policy with extreme events such as tornadoes or significant wind damage that I will not rebroadcast it until I have a few independent and verifiable sources confirming it. I’m doubly skeptical if there are no pictures. Make sure that any pictures submitted over social [media] are authentic and not sourced from prior events. (This happens a lot; see the shark-on-the-interstate photo that often goes around with each tropical system.) Google Image Search and TinEye are two great tools to do just that,” says Smith.
Smith also recommends that those using social media to track and report on dangerous weather avoid polluting hashtags with automated tweets, check timestamps — especially on Facebook — and never post or share fake photos. Even if these images start out as a joke, fake photos only slow media outlets and government agencies trying desperately to take stock of a rapidly developing situation while people are in danger.
When gathering information on an approaching storm, Smith keeps a close eye on messages coming from the National Weather Service (NWS). Coupled with his knowledge of the area and experience with past storms, Smith does his best to distill that information down to what is relevant to his audience across Berkeley, Charleston, and Dorchester counties.
“The key part is ensuring that the messaging is consistent. Inconsistent messaging is confusing and could even be life-threatening,” says Smith. “Up to and during an event, I do a lot of temperature-taking with other outlets and NWS to make sure the proper emphasis is on the right things. We are fortunate that the Charleston media market does a good job coalescing around a weather message; there are many other markets in severe weather regions that do not do as good of a job, creating confusion and uncertainty where it’s not needed.”
Now that practically all broadcast meteorologists and the National Weather Service, have Twitter accounts, anyone with questions about potentially misleading information they find online can easily ask a trusted source. Smith is more than happy to dispel any rumors, as well. The important thing to remember is to consider the source of any suspicious news. According to Smith, the only official forecasts for tropical systems in particular come from the National Hurricane Center.
For Smith, the key to not raising the alarm too soon on the threat of severe weather is understanding the state of the science and working within those limitations to craft a message appropriately. Beyond seven days from any potential impact, forecasting specific weather for a certain location becomes much less skillful. For tropical systems, the National Hurricane Center doesn’t forecast beyond five days, according to Smith.
- Jared Smith has manned @chswx since 2008
“With Irma, I handled things a little differently than I normally do. In a typical scenario, I usually only begin reporting on a tropical system in earnest when Charleston reaches the cone. In this particular case, though, there was a combination of a general nervousness in the community, the spread of irresponsible rumors on social media in the wake of Harvey, as well as concerning trends in the global weather model ensemble which led me to begin reporting a lot sooner than I usually do,” he says.
Smith adds, “If people want to track every storm out there, that’s what @NHC_Atlantic is for. If a long-range trend in a model stirs up panic (such as an app calling for snow 9-10 days out), I’ll jump in, as well. Sometimes I feel like the most important part of running @chswx is being Snopes for weather.”
While Twitter has become a popular source to quickly disseminate information of a rapidly changing weather event, researchers have also began evaluating how the social media platform can be used to gather data on flooding and storm evacuations. Using grant funding from the 2015 SCFloods Research Initiative, researchers at USC have completed two recent studies, with the first using Twitter to quickly map areas at risk of flooding through the use of Twitter.
With floods as the leading cause of natural disaster losses in the United States and more than 75 percent of declared federal disasters related to floods, Zhenlong Li, Cuizhen Wang, and Diansheng Guo of USC’s Department of Geography and Christopher Emrich of the University of Central Florida developed a template for using the geolocation information available from tweets in Columbia and Charleston during the 2015 floods to chart the possibility of flooding in specific areas in real time.
Comparing the number of tweets sent from inundated areas, the research team discovered that almost 10 percent of tweets from within an inundated area were related to floods. That number dropped to 6.9 percent approximately 1 kilometer from the flood area and 5.6 percent of tweets sent 2 kilometers from the flood area.
“The preliminary results showed that the model output provided a consistent and comparable estimation of the flood situation across the whole study area. Such a map, which can be generated in near real time, is useful for improving situational awareness during or right after the flooding event,” researchers wrote. “This is of particular importance when social media (and/or stream gauges) is the only data available during the floods. We believe that the proposed approach provides a valuable means for building a near real-time flood impact assessment system leveraging social media, allowing emergency responders to make quicker and better decisions and responses with improved situational awareness.”
In another study, published in July, researchers from USC’s Hazards and Vulnerability Research Institute and Department of Geography again focused on Twitter data, this time as it relates to evacuation compliance during Hurricane Matthew in 2016.
As Matthew approached the South Carolina coast, then-Gov. Nikki Haley ordered an evacuation of the area on Oct. 4 to start the following day. Using metadata from tweets, the USC research team were able to estimate the number of Twitter users who left the coast, the approximate duration of their evacuation, and where they sought shelter.
Among Twitter users in Charleston, Dorchester, and Berkeley counties, 50 percent evacuated the area during the Hurricane Matthew. Of all the Twitter users who evacuated the coast, 46 percent remained in South Carolina, while 18 percent were tracked to North Carolina and 9 percent fled to Georgia.
Going against the official recommendations of the governor, coastal evacuees began returning home on Oct. 9, just one day after Matthew made landfall near McClellanville, S.C. On this date, 11 percent of evacuated Twitter users made their way back to the coast, with rates increasing over the following days.
“While the approach does have its limitations ... we do have a more robust measure of when Twitter users likely left, returned, and where they went,” wrote researchers, comparing their methods of evaluating evacuation rates to traditional questionnaires. “So the trade-off is the immediacy and relative accuracy of the timing of the evacuation and the destination of the evacuees versus the detailed motivation behind such behavioral responses assessed months later.”