Experts from the National Hurricane Center provide a map every few hours that illustrates their best guess as to where the storm will be at several time points over the next five days. You can draw a line connecting these points and most view this path as THE track for the storm. The experts aren’t nearly as convinced of its accuracy and, in recognition of this, they surround these points with a “cone of uncertainty.”

They’re wise to do so, since usually the prediction changes as more information becomes available. This annoys some because, focusing only on that line, they’ve made their preparations and now find it was unnecessary. They’ll say of the hurricane experts, “They don’t know what they’re doing.”

In actual fact, I believe they do know what they are doing. It’s just that we don’t know what they’re doing. The cone of uncertainty is an attempt to clarify the imprecision of prediction. Most of us have learned to accept the line may not be exact, and that the cone indicates possible deviations. But is that the whole story? I have not seen anyone properly interpret it for the general public, and I feel this is where the experts fail. So I decided to look into it. While far from obtaining a full understanding, here is what I discovered.

Let’s start with a simple idea. Suppose a car is traveling 30 miles per hour for two hours. It covers 60 miles. If it moves 40 miles per hour it covers 80 miles. How far it goes depends on the idea we call “speed.” Speed is what’s known as a

*variable*and our tiny problem has only a single variable. That makes it easy to analyze. If I tell you the speed in miles per hour, you can tell me how far the car goes in two hours. Simply multiply the speed by two. We can be pretty sure our answer is correct. But what happens if the car speeds up and slows down, slips on icy roads, faces wind resistance, is carrying extra weight in the trunk? Now figuring out how far it goes isn’t nearly so simple, because there are many more variables to consider.

It’s even worse with hurricanes. I shudder to think of the number of variables that affect the location and intensity of a storm, a few of which are global winds and their strength and direction, water temperature and its variations, current speed of the storm, the presence of high and low pressure systems, the earth’s rotation, the jet stream, wind shear, land interaction, presence or absence of the Bermuda high, and history of past storms. To make it worse, most of these variables are continually changing with time and they’re interacting with each other in ways that can greatly complicate things. This means it’s really hard to analyze, and to obtain an accurate prediction. How could one not expect imprecision?

It takes supercomputers to take into account all the changing variables. A computer program that produces a track for the storm is called a

*model*and it’s effectiveness depends on whatever variables it is fed and how it is programmed to analyze those variables and their relationships. There are at least 13 different models in use and the various tracks can be seen at https://web.uwm.edu/hurricane-models/models/al052019.png. These all differ because they consider different variables or interpret them in different ways. Indeed, the predictions can vary widely from one model to the next.

What it all means is that predictions can’t be perfect and that gets us to the cone. How is it formed and what does it really signify?

As a first step in a new track prediction, the current location of the storm is noted. Then best estimates are made for where it will be at 12, 24, 36, 48, 72, 96, and 120 hours in the future, and the points are marked on a map. Of course, these positions may not be correct and acceptance of this gives rise to the cone. For each of these seven estimated spots, a circle is drawn centered at the position that indicates possible deviation from the path. The radius of the circle depends on the time. For example, it’s about 30 miles at the 12-hour point, 78 miles at the 48-hour point, and 228 miles at the 120-hour (five-day) point. No surprise the circles get bigger as the time ahead increases since there is more uncertainty then. Finally, all the circles are enclosed in a surrounding curve and that becomes the cone of uncertainty.

So how are the sizes of the circles determined? Intuitively I would have thought they’d be large enough so any deviations from the path would fit in the portion of the cone surrounding that circle. Afraid not! The radius depends on past errors observed over years of predicting. A circle is just big enough so that two-thirds of all past errors in prediction for that distance would fit in the surrounding curve. Only two-thirds! So that’s saying that portion of the curve will contain the true path, the path the storm actually follows—two times out of every three! That is, the path will go outside the curve in the area of the circle one-third of the time. The cone of uncertainty is even more uncertain than it first seems. In fact, suppose you take a cone at any given instant and you never make a new one. Then you plot what the hurricane does over the next five days. Its actual path will stay in that same cone only 60% to 70% of the time! That’s why the cone is constantly updated.

I understand the frustration with changing predictions. I worry that those displeased will wind up not trusting and thus ignoring future warnings, as many have threatened. I wish people could understand the true use of the cone and its limitations.

And I wish they could see it is far better to heed warnings that were later unnecessary than to ignore them and suffer the serious consequences if the warnings turn out to be correct. It’s advice I intend to heed even though, as I post this, Dorian, which has devastated the lives of so many, seems to have given us a break.