When severe storms hit, many people reach for their phones and cameras to capture images and videos of what’s happening around them.
And in Canada, storm chasers and organizations like Environment and Climate Change Canada (ECCC) rely partly on the public when tracking severe weather activity.
But in recent years, people have begun fabricating weather stories using photo editing software to modify images, creating photos and videos using AI (artificial intelligence) and even lying about the date and time a photo was taken.
“We have awesome satellites and radars, but they can only tell us so much,” Crawford Luke, a meteorologist with ECCC, told CBC.
Luke said they especially rely on reports and photos from the public for thunderstorms and heavy snow storms.
“We might issue a warning saying we’ve had this report of hail this size or trees down,” Luke said. “And then if we find out later that wasn’t true, well, now we’ve issued that warning like a false alarm.”
He explained that some images aimed at deceiving ECCC don’t even have a touch of editing. Luke used an example of a tornado image that was submitted in Ontario.
“It was a real photo of a real tornado, but it was just from a different place,” Luke said. “It was sort of sent in as [if this] just happened in southwestern Ontario.”
Before confirming the tornado had happened, the organization tracked the image back to a newspaper based in Texas. Luke said it’s just one instance of deception during severe weather events.
Credibility affected
Jenny Hagan is a storm chaser and severe weather specialist in Saskatchewan. Tracking down storms as they happen and documenting them for organizations like ECCC is a big part of her job.
Hagan said an influx of false weather reports and images is beginning to affect the credibility of everyone’s work in the field.
“It ruins the credibility of real work,” Hagan said. “Storm chasers will spend years honing their skills and capturing these real moments. AI kind of diminishes the value of those efforts.”
“When we’re looking at educating the public on storms and storm safety, this fabricated content can really hurt that accuracy of storm chasing.”
As someone who is very active on weather social media pages, Hagan said fake content is a trend that’s beginning to take off.
“Everybody’s chasing a little bit of attention,” Hagan said. “It does happen quite frequently, I think, between AI images and images shared under false pretences. I see those a couple times a week at least nowadays.”
She pointed out that even credible storm chasers like herself will edit storm photos, like removing a power pole obstructing the view of cloud formations or adjusting brightness if the camera didn’t capture a storm the same way our eyes do.
But editing a severe weather image in any way can cause confusion for storm chasers and the public.
Last year, an image that circulated on social media sparked confusion in the Saskatchewan storm world, Hagan said. A photo depicting a tornado and how it changed over the course of a storm got confused for being an image of six tornadoes touching down in the province at once.
Hagan isn’t completely writing off the use of AI in her field. She said new storm chasers often use it to break down complicated weather models or to track the movement of certain storms.
Hagan said as false severe weather images become more common, some tips for determining if an image is real include looking for bold pops of colour or unrealistic cloud formations, and using critical thinking to determine if the weather event could plausibly happen.
Erosion of public trust
As wildfires rage in Canada and extreme events become more common, the public relies on weather reports now more than ever.
“This is a time when people really need validated and verified information because their property and their lives depend on it,” said Manjula Selvarajah, a technology columnist for CBC.
Selvarajah said one of the easiest ways to detect a false image is by looking at its metadata: the time, date and exact location the photo was taken. However, she said now even that is something that can easily be manipulated.
“Now that becomes meaningless, because if you look at what used to be important information, the picture, the moving image and the audio could be generated by AI,” Selvarajah said. “So that level of proof is now gone. I think that’s going to be a real loss.”
Typically, the human eye can detect images that seem faulty. Selvarajah pointed out that severe weather conditions aren’t something that most people are accustomed to, making it difficult to distinguish the truth.
As the ability to modify images continues to become more accessible, Selvarajah warned it will only become more difficult to determine which images are credible.
“I think at a certain point, we’re actually not going to be able to tell what’s real,” she said. “We just lost this really powerful tool for storytelling and transmitting information.”
