Hurricane Irma, a storm name likely to be retired, has been a challenging, exhausting, and perplexing storm to forecast for. There are so many characteristics of this storm that are textbook yet out of the ordinary at the same time. Being located in Central Florida just north of Orlando, this storm was close to home, warranting every bit of my intention. I thought to include some of my observations of this storm since its infancy, reaction to certain debates in the weather community, and lessons to be learned from Irma.
I Like Ike
Irma was a classic Cape Verde storm that formed off the African coast as a tropical wave. But as it got past the islands, strong high pressure to its north forced the storm to the west-southwest — which is unusual. Some in the meteorological community (which I include as meteorologists and weather enthusiasts) pointed out similarities to Ike. If one was to draw Ike’s path along Irma’s southwest movement, it would suggest an out-to-sea route. Many of the global models and ensembles suites were in support of the “out-to-sea” route as well; unfortunately, this would not be the case.
Ensembles, the use of multiple computer runs to get an ensemble mean, is an excellent tool to use when forecasting any high-impact weather event — whether it be hurricanes or blizzards. As Irma approached closer to island chains, I heard some commentary suggesting ensembles have a “windshield-wiper” effect when predicting a hurricane’s track as do the global models. Some respectable meteorologists do not look at ensembles for this reason. I agree that if one was to analyze ensembles solely the mean of tracks, they would see the windshield-wiper effect since the mean track would change with every run. However, the goal of ensembles is to identify the level of uncertainty at given forecast point. As long as the actual track of a storm is within the spread of ensemble members, then the ensembles successfully identified the level of uncertainty — even if the storm tracked along the most outlying solution. In the case of this American ensemble model run, known as the GEFS, Irma (marked by the thick black line) tracked within the spread of ensembles until Day 5. Even though the ensemble mean (the thin black line with square boxes) stayed off the east coast of Florida, the ensemble spread correctly identified the west coast risk. To be fair, it failed to include Irma’s actual track across southwest Georgia and into Alabama. The GEFS has a known bias of not have a wide-enough ensemble spread. The European ensemble suite, known as the EPS, has too much of a spread.
The UKMET, arguably one of the more underappreciated models, is considered to have some of the best physics in weather modeling. It is the second most accurate model in the world, just behind the European (ECMWF) but ahead of the American (GFS). I first took note of it in the days leading up to Hurricane Matthew. Despite being a western outlier, it was very consistent run-to-run on the idea of taking Matthew up the eastern coast of Florida. No, it was not correct on Matthew landfalling into Florida, but it was pretty close to its actual track. Overtime, models did adjust to its solution, resulting in a big “win” for the UKMET. I do want to emphasize it correctly and consistently forecasted Matthew from five days out — remarkable for hurricane forecasting.
Ironically, a similar pattern began to appear when forecasting Hurricane Irma. It consistently forecasted Irma to travel towards Cuba, making a later turn to the north and taking it through the Florida Keys, specifically just east of Key West. From there it would make landfall near or just east of Naples traveling across the spine of Florida. Yes, Irma didn’t technically make landfall in Cuba, nor did it cross Florida into the Atlantic, but it caught on to the westward trend in model guidance well ahead of the other guidance. The overall idea of Cuba interaction, Florida Keys landfall, and Central Florida impacts, were well identified by the UKMET — from four days out.
The Charley Maneuver
Hurricane Charley was notorious for making a turn to the east much sooner than forecast, hitting the Fort Myers area instead of Tampa, later causing impacts to the Orlando Metro area. Irma must have gotten some ideas from him — after making landfall in the keys, Irma was forecast to ride along the western coast of Florida. Instead, Irma continued on a due north path, making a second landfall just east of Naples. It would continue on its northern route well inland just west of Lake Okeechobee. Local meteorologists were beginning to extrapolate a course just to west of Orlando if the northern route continued. Thankfully, the storm made its gradual westward turn near Polk County, FL (located southwest of Orlando), avoiding both the Orlando and Tampa metropolitan areas. I can testify it was not comfortable watching Irma head straight for me before it begun its turn.
Irma is a storm that will be studied by current and future meteorology students. Its classic “buzzsaw” appearance, its rare west-southwest movement in the Atlantic, and the high degree of impact on the State of Florida, will make Irma a notorious storm for the 2017 Atlantic Hurricane Season, let alone Hurricane Harvey. As time progresses, meteorology will continually get better, but there is still a long ways to go for accurate Day 3 forecasts — even a 24 hour forecast by the National Hurricane Center has an average track error of 50 miles.
And at the time of this writing, we are only half way through the hurricane season.
— Vogt WeatherWatcher