TRUTH ABOUT ‘SPAGHETTI’ MODELS: Hurricane prediction paths draw from different data, aren’t trying to do same thing

Divining a hurricane's path is far from simple: it includes regular feedings of 85 billion clues to supercomputers -- information on everything from the tiniest raindrop to the rowdiest thunderstorm, gentle sea breezes to tree-toppling gales, the heat of the ocean's surface to the very underside of space.

Numbers are crunched, and, voila, a "spaghetti" model is served.

Social media and four consecutive years of overachieving storms have given celebrity status to certain spaghetti model runs -- regularly pitting the Euro against a souped-up American model during dinner table chatter.

"I have a friend totally obsessed with these things, and his wife is always telling him to please not talk to me about the Euro model again," said Hugh Gladwin, a Florida International University researcher and anthropologist. "There is a desire for concreteness when people look at these, but it also feeds into phobias, and we just have to sift through it best we can."

The noise was no different with Hurricane Dorian as the colorful lines snaked toward Florida, some skewering Palm Beach County like a kebab, others scraping the storm along the coast. Twitter tiffs erupted over which model was better suited for the situation, and public shaming of those who dared post a single run was common.

Experts caution it's misleading to consistently tout one model over another because the whacked-out rainbows aren't always what they seem. Models have different purposes, and how they arrive at their end result is through a carefully chosen set of equations that differs depending on the programmer.

"If one model was best all the time, we'd quit running all the other models," said Brian McNoldy, a senior research associate at the University of Miami's Rosenstiel School of Marine and Atmospheric Science. "I generally wish the spaghetti plots weren't shown like they are because not all of those lines are equal."

Statistical models, including ones called the XTRP and CLP5, don't consider what's happening in the atmosphere at all, basing their routes solely on how past storms in the same place and time have acted.

A Canadian model, which can often be found on spaghetti model plots as CMC or CEMN, isn't given high ratings by James Franklin, the former chief of forecast operations at the National Hurricane Center.

"Tropical cyclones just aren't a problem they devote a tremendous amount of resources to," he said.

Deterministic models are a single run of a specific model and show up as a solid colored line on maps. Each model run is fattened with equations that take into account billions of atmospheric data points from all manner of measuring systems, including weather balloons, ocean buoys, ships at sea, backyard weather stations and satellites.

It can take several hours to run a single model. Some of the more prominent models include the Navy's NVGM, the Hurricane Weather Research and Forecast System's HWRF, the U.K. Met Office Global Model, or UKMET, and the Hurricane Multi-scale Ocean-coupled Non-hydrostatic model, which is better known as the HMON.

The popular Euro (ECMWF) and new American model, which is also called the Global Forecast System, or GFS, and shows up on spaghetti models as the AVNO, are two of the leaders of the deterministic models.

"The casual user might not know that HMON is a brand-new model under development and is going through a lot of growing pains, but it is plotted on there with everything else," Franklin said. "You can make some incorrect conclusions if you don't know the details."

A recent presidential dust-up that pitted President Donald Trump against National Weather Service forecasters included a spaghetti model map of Hurricane Dorian from the South Florida Water Management District's website.

Trump tweeted that Alabama was going to be hit by Dorian days after models had taken the state out of play and wind arrival time maps had only a tiny sliver of southeastern Alabama as having any chance of feeling tropical storm-force gales.

When an Alabama NWS office tweeted in response that the state would feel no effects from Dorian, the president tweeted a photo of a spaghetti model plot from four days prior that showed a spider web of lines that reached, in one case, all the way to Louisiana.

The map included the solid lines of the deterministic models, but also their ensemble runs, which can show up in lighter tones or in a different color. Ensembles allow forecasters to see a range of outcomes so they can better gauge uncertainty by tweaking initial conditions slightly to see the results.

A big jolt in model reaction is a sign that minor changes in the atmosphere can have a major change in the forecast. The GFS ensemble has 20 runs, while the Euro has 50.

"Anyone looking at these spaghetti models can latch onto any one they want and see the solution that they want to see or focus on whatever is closest to them," said Jonathan Belles, a digital meteorologist with Weather.com, an IBM business. "There's not going to be one model that wins the entire year. One model can perform very well on a certain storm but be garbage on the next one."

This year, the GFS (the new American model) got an upgrade equivalent to replacing a car's engine -- the first major boost in nearly 40 years that added the Finite Volume Cubed-Sphere dynamical core, or FV3.

It is being run on a supercomputer that got a multimillion-dollar brain boost after Superstorm Sandy devastated the Northeast in October 2012, exposing limitations of the GFS model. The GFS underperformed compared with the Euro, which correctly forecast Sandy's track toward the East Coast.

The GFS was outdone again during 2015's Category 5 Hurricane Joaquin. The GFS and European model disagreed on forecast track with the GFS showing the dangerous cyclone making landfall in the Mid-Atlantic region, while the European had it going out to sea.

The Euro won.

Thirty-three crew members aboard the cargo ship El Faro died when it sank in Hurricane Joaquin. The U.S. Coast Guard ultimately determined that the captain misjudged Joaquin's path and "failed to understand the severity of the situation."

"If you look at the averages over the past two to three years, the Euro's errors are lower, but they are all still in the same general ballpark," Franklin, the former chief of forecast operations at the National Hurricane Center, said. "But if you only consider the Euro and disregard the GFS, you will get yourself into trouble."

Consensus models -- when the forecaster takes top-performing models and averages them -- usually outperform the Euro, Franklin said.

Belles, of Weather.com, said the 2018 hurricane season ended with the Euro having the best track forecasts, with the GFS and HWRF not far behind.

"I would say, in general, the Euro does the best for track, not always, and that's why you can't just blankly use it across the board," McNoldy said.

The National Hurricane Center doesn't post spaghetti models on its website because its authorities think they confuse some users. Others get caught up in the individual forecast scenarios, many of which have little or no chance of being correct, said NHC spokesman and meteorologist Dennis Feltgen.

Track models also give no forecast for storm impacts or the intensity of the storm, which is best forecast by a different set of models or combo models, such as the HWRF.

Gladwin said it's unclear why some people fixate on track model plots even when they're not sure what all the scribbles mean.

"People want to feel sure about things. They want to think science is always right," Gladwin said. "When people are unsure about things, they look at the most recent event and grab onto statistics or numbers."

Or spaghetti.

___

(c)2019 The News-Journal, Daytona Beach, Fla.

Visit The News-Journal, Daytona Beach, Fla. at www.news-journalonline.com

Distributed by Tribune Content Agency, LLC.