Forecasts Have Done A Good Job Predicting Irma’s Shifting Path


Hurricane Irma made landfall this morning in the Florida Keys as a Category 4 hurricane with maximum sustained winds of 130 mph.3 But while Irma’s strength was forecast for days, the projected track of the storm did seem to wobble. As late as early Thursday morning, Irma was expected to travel along Florida’s eastern coast. Slowly, the forecast shifted west, from eastern to western Florida. The focus of government officials and media coverage shifted from Miami in the east to cities like Naples in the west. Such forecast movement may seem big, but it’s actually a normal hurricane forecasting error.

Hurricane forecasts have gotten much better over the past few decades, but they are far from perfect. Since the beginning of this decade, the average error for National Hurricane Center forecasts of the paths of tropical storms and depressions 48 hours before a storm passes a specific area has been greater than 80 miles. From 72 hours out, it’s been greater than 120 miles. As I’ve noted previously, 80 miles may not seem like a lot, but it’s greater than the distance between the western and eastern coasts of Florida at Miami.

In Irma’s case, the forecast error in the storm’s track was less than average. At 11 a.m. today, the storm was just north of the Florida Keys. On Thursday — three days earlier — the 11 a.m. forecast had projected Irma to be southeast of Miami, well to the northeast of the Keys, at that time on Sunday. That’s an error of about 100 miles, or about 20 miles less than the average error 72 hours out. By 11 a.m. Friday, the forecast was far more accurate, projecting that the storm would be just south of the western portion of Everglades National Park 48 hours later. That forecast was right about how far north the storm would be but put Irma about 30 miles east of where it was this morning.4 That’s less than half the average error this decade.

There were probably multiple reasons that forecasts were too far east. Ian Livingston, a forecaster for the Capital Weather Gang, a meteorology blog at The Washington Post, told me via email that the Bermuda High, a high-pressure system in the Atlantic Ocean that can keep hurricanes from drifting east out to sea, was a little stronger than forecast. That means the hurricane was forced a little more west than expected, he said. Livingston also noted that Hurricane Jose, currently in the Atlantic, was much more powerful than forecast and that may have pushed Irma farther west. He also pointed out that even a small difference in the timing of a disturbance in the mid-latitude westerly wind currents, which bring hurricanes from the tropics to the U.S., could have shifted the storm slightly.

Despite these shifts, Irma’s landfall in the continental U.S. was within the error cone of every single 5 p.m. forecast from the National Hurricane Center. (This cone is based on historical error rates.) In other words, all of these forecasts were within the expected range of error at their given time interval.

The case of Irma provides yet another example of why we need to think probabilistically about forecasting. The center line of a hurricane forecast cannot be taken as gospel. Western Florida was always in danger of bearing the brunt of Irma. The National Hurricane Center provides an error cone for a reason.