Skip to content
SCIENCE

Yeah, the forecast for Hurricane Dorian is a mess

It's as if the world's best model just threw up its arms in despair.

Person holding a box with office supplies
Story text
Hurricane-track forecasting has gotten pretty good over the last couple of decades—so much so that some hurricane scientists believe we are close to reaching a limit on predictability. However, there are still outlier storms that are difficult to forecast, and Hurricane Dorian is definitely one of those cases. This uncertainly is only going to amplify the misery of Floridians seeking to prepare for or evacuate from what is likely a major hurricane bearing down on the state. In this case, the problem may be further worsened by premature confidence in where Dorian will go: as of Thursday evening, the one thing we do know about this storm is that we don't know where it's going to go. Spaghetti plots of forecast tracks are commonly shared online, both in news stories and on social media, and they're sometimes useful. With Dorian, a plot like the one below from Thursday evening, which shows the official forecast track from the National Hurricane Center (in black) as well as other models, is actually very deceiving.
From the plot above, if we dismiss TCLP and CLP5 (which you should, because they are not really models at all but persistence and trajectory tracks), there is a nice clustering of models bringing Dorian to the Central Florida coast on Monday. The agreement is quite good. But don't believe it. To get a better idea of where Dorian is likely to go over the next week or so, a better practice is to look at ensemble forecasts from the global models. They do a good job of capturing the range of outcomes based upon slightly different initial conditions. For this article, we'll look at the Thursday 12z ensemble output from the European model. (In this case, 12z means the models were initialized at 12:00 UTC Thursday and finalized about six hours later).
What is important to note about this plot is that it only shows the track of Dorian out to 120 hours, so that means its position as of Tuesday at 8am ET. Notice that there is a remarkable variance in the location of the "center" of Dorian in these roughly four dozen ensemble members. Yes, a reasonable amount of the ensemble members bring Dorian to the coast between Sunday and Tuesday, but some are far, far away. In some scenarios, Dorian turns north before even reaching Florida. This uncertainty meshes with what we're actually observing about the atmospheric conditions that will help steer Dorian—namely that the steering currents around the Bahamas are weak. Eventually, Dorian is going to turn northward and traverse around the western periphery of a ridge of high pressure over the Atlantic, but it's difficult to say when that turn will come. In the meantime, the storm is going to move slowly. The 5pm ET Thursday track forecast from the National Hurricane Center shown at the top of this post—which is absolutely the authority for such things—captures this ambiguity. It encompasses the entire Florida peninsula within the five-day cone of uncertainty. All of this is singularly unhelpful to the residents of the Sunshine State, who are staring down at the possibility of a major hurricane making landfall somewhere along the peninsula in three to four days. The time for preparations and evacuations is now, but it is hard to offer much in the way of specificity about where the worst storm surge, damaging winds, and inland rainfall will occur. Certainly a slow-moving storm will make the latter variable, rainfall, worse. For what it's worth, the European model provides ensemble tracks for the storm 10 days out. If we look at the 240-hour tracks, the confusion only increases. Weather enthusiast Jack Sillin captured this nicely on Twitter this afternoon, highlighting several of the variables in what remains a complicated forecast. So yeah, Dorian's forecast is a mess. Beware of anyone who tries to tell you otherwise over at least the next day or so.