The National Oceanic and Atmospheric Administration (NOAA) has been promoting the fact it “nailed” hurricane Irene’s track four days out. On both of its home page and Environmental Visualization Laboratory website, it features an animation of Irene threading the needle of the National Hurricane Center (NHC) track issued at 11 p.m. ET on Tuesday, August 23.

Animation showing Hurricane Irene’s actual track using satellite imagery and NHC’s forecast track issued at 11 p.m. ET on August 23.

It’s true that NHC correctly predicted Irene’s track at that particular point in time. But what about subsequent times? What about previous times? And what about its intensity?

Let’s take a look at NOAA’s forecast tracks after the 11 p.m. track on 8/23.

Animation showing NHC’s forecast track for Hurricane Irene between 11 p.m. on August 23 and 1 a.m. on August 27.

Notice the track shifted east of the actual track on Wednesday, August 24, so the skill of the three-day track forecast actually regressed from the four-day forecast. This had important implications for areas on the western periphery of the track guidance where it appeared a glancing blow was becoming more likely (when it wasn’t). But then the track guidance shifted west again on Thursday, August 25. From that point forward (within 48 hours), NHC had the track more or less right.

The point I’m trying to make is that the NOAA PR machine should not lull us into the sense that we can rely on the exact track given in a 96-hour forecast. In fact, four days out, the NHC was prudently reminding the public that track errors average about 200 miles. The fact NHC had it right at the point in time was more fortuitous rather than an indication that the science has advanced to the point the track can be “nailed”at that range.

And track errors are even larger prior to four days preceding landfall as the NHC tracks showing Irene hitting south Florida indicate.

Animation showing NHC’s forecast track for Hurricane Irene between 7 p.m. on August 21 and 5 a.m. on August 24.

Having said all of that, NHC track forecasts have come a long way since 1990, improving by over 50%. Meteorologist Jeff Masters wrote on his blog today:

NHC director Bill Read stated in a interview this week that had Hurricane Irene come along before the recent improvements in track forecasting, hurricane warnings would have been issued for the entire Florida, Georgia, and South Carolina coasts. At an average cost of $1 million per mile of coast over-warned, this would have cost over $700 million. We can credit the investments made in hurricane research, improved satellites, and better computer models for the majority of this improvement.

While the track forecast was excellent four days out, so-so three days out, and then pretty much spot-on the rest of the way, the intensity forecast was not as good. NHC predicted Irene would make landfall as a Category 3 storm, whereas it actually came ashore at Category 1. The landfall intensity forecast contributed to some of the hype (justified or not) in terms of storm surge and wind impacts from the Outer Banks into coastal New England. Had the intensity forecast been better, perhaps more of the media attention would have appropriately focused on the inland flooding threat rather than the coastal impacts. And perhaps fewer people would have had to be evacuated.

A New York Times piece from yesterday provides a nice overview of the intensity forecasting problem and efforts to address it. Here’s an excerpt:

“With intensity, we just haven’t moved off square zero,” Dr. Marks said. Forecasting a storm’s strength requires knowing the fine details of its structure — the inter nal organization and movement that can affect whether it gains energy or loses it — and then plugging those details into an accurate computer model.

Scientists have struggled to do that. They often overestimate strength, which can lead to griping about overpreparedness, as it has with Irene. But they have sometimes underestimated a storm’s power, too, as with Hurricane Charley in 2004. And it is far worse to be underprepared for a major storm.

We’ll have more on the intensity prediction problem in future posts.

My main point point though is that hurricane track forecasts are not characteristically perfect four days out and NOAA should be cautious in how it spins a specific forecast at that range.

UPDATE: The New York Times’ Andy Revkin has a nice piece on the NHC Irene forecast: Federal hurricane forecasters did their job