Hurricane Joaquin has kind of been a forecasting nightmare. (NOAA)

Hurricane Joaquin has proven to be a difficult forecast — much more difficult than other recent East Coast hurricanes. A lot of the uncertainty stems from incomplete observations of the hurricane’s structure and low-confidence forecasts of the weather patterns over the Eastern Seaboard. But complicating matters is the National Hurricane Center’s cone of uncertainty itself, which actually has very little to do with the range of possible outcomes. In the case of Joaquin, the cone should be much wider than what is currently portrayed.

While it is true that track forecasts have improved dramatically in recent decades, predicting the future path of a hurricane remains an incredible challenge. Some storms are inherently easier to predict than others. Complicating factors like proximity to land, moderately strong wind shear, weak steering currents, and interactions with mid-latitude features still present major challenges to models and human forecasters alike.

To add even more complexity, the track and intensity depend on each other, so their uncertainty is compounded. In other words, if the future track were somehow known perfectly, there will still be uncertainty in the intensity forecast. But if there is large uncertainty in the track then the uncertainty in intensity is greatly magnified, and vice versa.


But surprisingly, the size of the National Hurricane Center’s cone of uncertainty — the cone that surrounds the main track line and gives a range of possibilities for the storm’s future path — is not determined by the specific forecast uncertainty for that particular storm. Instead, it represents an average track error over the past five years. As the forecasts get better each year, the cone shrinks, but it is also fixed in size for each storm that year, with no regard for whether the forecast is high- or low-confidence.

This is obviously a significant shortcoming of the current methodology — it has no relevance to the current atmospheric predictability — and some have criticized the National Hurricane Center for not adapting the cone for highly unpredictable situations like the current one with Joaquin.

“There is no rational explanation for why the National Hurricane Center product suite is so rigid it precludes incorporating readily available case dependent information that reflects reality much more so than simple statistics,” said Steve Tracton, a contributor to the Capital Weather Gang. “Especially in cases like this there is no need to await the long promised, but not delivered computer algorithms to do this.”

In Joaquin’s case, the track spread from model guidance is abnormally large — far larger than the cone of uncertainty. Below are some of the track forecasts from models available on Thursday morning. These are all skillful models, and they vary with each run.

The spread in possible tracks from west to east is extremely large. The outcomes in these tracks, which include high-resolution hurricane models as well as full global models, range from a sharp west turn and landfall in the Southeast United States, to a track far away from the East Coast with no impact at all.

But the range of these possible outcomes — even to only include the most-trusted and well-performing models — are not represented by the National Hurricane Center’s static cone of uncertainty.

“The two computer models that have traditionally performed best [the GFS and the European] BOTH go outside of the cone — one to the right and one to the left,” Weather Channel meteorologist Bryan Norcross wrote on Facebook. “The cone splits the difference.”

“The bottom line,” Norcross added, “we need to imagine the cone being significantly wider until we have more certainty in the forecast.”


Track forecasts from various global and regional models (left), the ECMWF 50-member ensemble (center), and the 20-member GFS ensemble (right). (UAlbany)

A dynamic cone of uncertainty that adjusts and morphs to encompass a wider range of possible solutions would be very useful. The realistic uncertainty in position at 72 hours, for example, probably won’t be a perfect circle for every storm all season long. It would be an elongated ellipse, perhaps stretched parallel to the track if the speed is more uncertain, or stretched perpendicular to the track if the directional steering is less certain.

So far, the official track forecasts for Hurricane Joaquin have not performed well. Joaquin has drifted to the southwest almost since it formed, while many models and and the National Hurricane Center, forecasted a turn to the west and north. All along Joaquin’s path thus far, the official forecast track has been adjusted over and over as the storm drifted farther and farther away from where it was forecast to be.


Observed track of 11L (thick black dots), and all NHC forecasts (colored lines). (Brian McNoldy)

Unfortunately, the track hasn’t been the only forecast struggle with the storm. Hurricane Joaquin’s rapid intensification has been incredible. While Joaquin’s origins as an upper-level low pressure system can be traced back about two weeks, it was only three days ago that it was declared a tropical depression. Just Wednesday morning it was still a tropical storm, and now it is a strengthening Category 3 hurricane.

These “Bahama Busters,” or storms that explosively intensify over or near the Bahamas, are notorious because the deep warm water there combined with the proximity to land make them extremely dangerous.

Because rapid intensification is so poorly understood, the official intensity forecasts from the National Hurricane Center have also been worse than average. Each forecast suggested Joaquin would be far weaker than what it actually turned out to be.

Joaquin’s entire existence has been a game of forecast catch-up.


Observed intensity of 11L (thick black dots), and all NHC forecasts (colored lines). (Brian McNoldy)

The lesson is this: do not take a particular forecast too literally. The National Hurricane Center is required to make a single deterministic 5-day forecast for every storm every six hours. That is not an easy task with millions of people eagerly awaiting each update. They are quick to acknowledge that sometimes a situation is less predictable, and uncertainty is greater than average. But so far, such uncertainty is not incorporated in their graphics or the fixed-size cone.