News

Ask Matt ... What forecast model works best

By Matt Matteson, Published: January 22, 2022

Q. What’s the difference between the European and American model in predicting weather? I often see these models presented by our local weather people but they never explain which is more accurate.

To sort all this out, I contacted the chief meteorologist at each of our four local TV networks. OK, yes I was snubbed by Jason, Chris, Christy and even Kendra. Pressing on, I did reach Paul Speranza, WTZQ’s weather forecaster. Speranza, who has been in the “weather business” for over fifty years, shared that unlike the American model, the European model is not a free service and that there are over fifty different models which explains the “spaghetti strings” often seen in hurricane forecasts. Speranza said that for tropical storms, the American model tends to predict storm tracks further away from the Florida coast than the European model.
The American model is officially called the Global Forecast System (GFS) model and is created and operated by the National Weather Service. The GFS generates data for dozens of conditions such as temperature, wind, precipitation, soil moisture and ozone. It all gets loaded into a supercomputer and maps are generated.
Headquartered in England, the European model is a partnership of 34 nations. I found a good comparison on the Global Weather & Climate Center website. The American model predicts conditions 16 days in advance while the European model forecasts only 10 days in advance but with superior computing capabilities. Both models employ some of the fastest computers in the world. “The European model was able to accurately predict when Hurricane Sandy (2012) would turn into the northeastern section of the United States before the American model could,” the model’s website says. Hurricane Sandy’s damage in the U.S. alone was estimated at $65 billion.
Yet the European model muffed the prediction for Hurricane Florence in 2018. A story that ran in the Washington Post (8/9/18) noted that “the European model wavered and the American model had Florence going out to sea but not the National Hurricane Center’s own forecasts. Five days out, the Miami-based Center predicted a Carolina landing that ended up being just two miles off target.” Florence, which made landfall near Wilmington, N.C., packed 137 mph winds and carried an 11-foot storm surge. It caused flooding as far inland as Chapel Hill and 43 deaths.
I went to UNC-Asheville’s Department of Atmospheric Sciences to get more answers to the battle of the weather models. “Both models are high quality models and there are only subtle differences between those two,” said Dr. Christopher Hennon, who chairs the department. “The American model is more of a snapshot in time whereas the European model employs trends,” he said. “Forecasters have to look at all models, not just one. At the core, every model operates the same way. It gauges the state of the atmosphere, applies the physical laws of nature, and makes a prediction.”
More recently, we experienced Tropical Storm Fred, which crossed the Florida Panhandle before it arrived in Western North Carolina on August 17, 2021. I asked my UNC-Asheville source about the warnings. “Once it came ashore, Fred was a slow moving system,” said Hennon. “The the rainfall of 6 to 12 inches was accurately forecast by the Weather Prediction Center.” Among other things, this center, located in Maryland, forecasts heavy rains and snow and potential flooding and posts the information on the internet for public use. Tropical Storm Fred would dump 14 inches of rain on Haywood County and cause six deaths and $20 million in damage and caused widespread damage to crops in the French Broad Valley in Henderson County.
Incidentally, after Speranza nailed winter storm Izzy’s snow, we asked him which model he used. He predicted 10-15 inches of snow on during his WTZQ weather forecast days before the National Weather Service and local TV broadcasters. A forecaster for 52 years, Speranza has been on the air locally for 28 years.
“That was on Tuesday that I put numbers out,” he said. “The American model kept showing that we were in for a major snow event. So I looked at that and I looked at that and I said to (radio host) Mark (Warwick) in the morning, ‘You know what? It’s Tuesday morning and if I’m a forecaster I’m not waiting for other people to dillydally around. So I put that out there — 10 to 15 inches if we stayed all snow. If we ended up with a little bit of sleet, which we did, I ended up with an even foot here (on Long John Mountain).
“We had 9 inches, we had a 10, we had 11, we had a couple of 12s — it was right in the range. We said 9 to 13 if we had the sleet, we said 10-15 if we didn’t get the sleet,” he said. “But the European model really did a lousy job. For the storm that came, the European model was only printing out 1-2 inches of snow. That was the whole storm. It was going to come up to the Carolinas and then go out to sea. The American model said, no,” and accurately predicted the timing and accumulation. “You have to look at all of them (but) the American model was king out there of forecasting.”
Here’s what Speranza said on WTZQ at noon Wednesday: “I think you’re not going to be going places on Sunday. I think you’re going to be staying home and looking out your window. The southern part of the county may encounter a little bit of sleet mixed in — just depends on the track of that storm. But overall, more snow than anything else. … Ten to 15 inches of snow is what I’m looking for.”

* * * * *

Send questions to askmattm@gmail.com.