The Day I Thought I Misled the President of the United States: A Visualization Tragicomedy
On August 27, 2019, the National Hurricane Center (NHC) predicted that tropical storm Dorian would become a hurricane and that Florida was in its probable path:
A few months before, I had finished writing a new book about how visualizations can mislead and deceive. One of its chapters explains how to interpret hazard risk maps such as the ones produced by the NHC and its parent organization, the National Weather Service.
A few days later, on August 31, the NHC predicted that Dorian would be a major hurricane, and the possible path of its center could run anywhere over Florida, including Miami, where I live. My family and I started to prepare by stocking food and water, filling up gas tanks, and closing up window shutters. If you live in Florida, these actions become second nature after a couple of hurricane seasons.
Fortunately, by the morning of Sunday, September 1, Dorian’s cone had moved Eastward. This didn’t mean we were out of danger, but the probability of being badly hit by the hurricane decreased:
That same morning the President of the United States sent a now much-maligned tweet mentioning states that he thought would be hit “(much) harder than anticipated” by Dorian. Alabama was among them:
But then something clicked in my brain: I looked at the times of President Trump’s tweets: he sent one at 8:51 a.m. and a second version at 10:51 a.m. I also remembered that he wakes up early every morning and spends a few hours watching TV or reading news publications, sometimes The New York Times:
That very same day, on September 1st, The New York Times Sunday edition published an article I wrote trying to teach how to read hurricane cone maps. You can see its online version here. The beautiful graphics are by Stuart Thompson and Tala Schlossberg:
In the article, I explained that the cone of uncertainty is essentially an abstract representation of many possible positions of a hurricane center for the following five days and that there’s always a chance that the actual path of the storm center will be outside of the cone.
Hurricanes are enormous things, so if the path of Dorian had ended up, say, on the West side of the cone of uncertainty, the winds, rain, and storm surge would still affect large parts of the continental United States. Even if you live outside a hurricane’s cone of uncertainty, if you’re near its boundary you should still take some precautions, prepare, and stay alert.
Go back to the second map in this article and you’ll notice that hurricane Dorian’s cone had touched a tiny portion of the Southeastern tip of Alabama on August 31. But on September 1, the cone had moved so far to the East that it was very unlikely to impact Alabama at all, as the Birmingham National Weather Service explained in a tweet:
I was worried for a few hours: Why did Trump make the mistake of including Alabama in his tweet? Maybe he hadn’t seen the early morning September 1 NWS/NHC updates or received a briefing? Had he read The New York Times instead, linked my article in his brain to the already dated August 31 maps he had seen, and then sent his tweet, alarming Alabamians? Had I confused or misled the President of the United States?
As it often happens, the reality was more mundane. Trump indeed hadn’t seen the NWS/NHC latest updates in the morning of September 1, so his tweet was careless. My conjecture—and I insist this is just conjecture—is that he was likely confused by the maps his advisors, who often aren’t expert meteorologists, showed him in the previous days. For instance, the wind probability products, which look like this:
As you may notice, there was a ~10% chance on the days before Trump’s tweet of Alabama receiving tropical storm winds. It never went higher than that on any forecast, but if you don’t know how to read these maps well, then confusion may ensue.
We can all make mistakes when dealing with complex visualizations, and I wish the President had simply issued a quick amendment and apologized after the Birmingham NWS office had called him out. But he didn’t. In the days that followed Trump doubled, tripled, quadrupled, quintupled, and sextupled down, insisting that his Alabama tweet was accurate. He even did a public briefing in the Oval office:
During that briefing, he showed an early cone of uncertainty, but with a noticeable addition:
What could have been just a silly, understandable, forgivable, and easily correctable oversight has led to a crisis that threatens to undermine the credibility of agencies that lives depend on, such as the National Weather Service and its National Hurricane Center. This is very serious, as cartographer Mark Monmonier has explained. Eroding institutions that are supposed to be neutral is appalling and should disqualify anyone from holding any kind of public office. If you disagree, please read Michael Lewis’s The Fifth Risk. It’s terrifying.
This tragicomedy has an absurd ending — as it should. One of my favorite Trump tweets from those days showcased the following “spaghetti” map:
Notice the caption at the bottom:
“If anything on this graphic causes confusion, ignore the entire product.” That’s perfect. So perfect, in fact, that one of my followers on Twitter, Chen Mingi, suggested I should use it as a title, maybe for an article or even for the next book I’ll write after How Charts Lie. I’m seriously considering it, and even designed a draft for a cover: