Blizzard rages south, grading forecast model performance

They say all weather is local. Today we learned again what a difference 40 miles of "local" can make.

A major winter storm and blizzard pounded Minnesota communities just an hour south of the Twin Cities today as excepted. Travel is difficult to treacherous across southern Minnesota today.

Numerous crashes and spin-outs are reported. This semi vs. MNDOT plow incident is a good reminder to give the plows some extra space out there.

At one point semis could not make it up a hill on I-35 near Albert Lea.

Political Coverage Powered by You

Your gift today creates a more connected Minnesota. MPR News is your trusted resource for election coverage, reporting and breaking news. With your support, MPR News brings accessible, courageous journalism and authentic conversation to everyone - free of paywalls and barriers. Your gift makes a difference.

Twin Cities spared

The late shift in storm track spared the greater Twin Cities metro any significant snowfall. A razor sharp cutoff in snowfall set up in the far southeast metro Thursday night and Friday. The difference between accumulating snow and no snow at all? Literally just a few minutes drive down the road.

2-24-raddy1
Weather Underground

As much as 9" fell in Red Wing, a mere 40 miles southeast of St. Paul. Skiers were delighted.

Many of us forecasters were scrambling to update the forecast as it became clear last evening that the models were too far north with the northern edge of the snow zone in the Twin Cities.

Grades: Forecast models storm performance

No doubt about it, this storm track was one of the more difficult to get a handle on in recent memory. After a very good run of accuracy with multiple storms earlier in the snow season, many U.S. forecast models performed poorly in placing the heavy snow zone just 1-3 days before the storm arrived.

It's not everyday you see your local NWS office send a tweet like this one.

The variance in forecast model output was unusually erratic with this system. Here's are some letter some grades for various forecast models based on my assessment of how they handled this week's storm.

NOAA's NAM model:  F

NOAA's NAM 12-km resolution model snowfall output from Wednesday placed 14" to 20" snowfall totals in or near the Twin Cities, and 18" in St. Cloud. That's a total model fail.

Personally I never bought into the NAM solution, recognizing it has a northward bias. Here was my assessment of the NAM output on Wednesday.

“NOAA’s NAM model brings the heaviest snow band right into the heart of the metro. But many meteorologists note the NAM model often has a northward bias on storm track.”

NOAA's GFS model: C-

NOAA's premiere forecast model did better than the oft-maligned NAM, but left much to be desired compared the the European (ECMWF) and Canadian (GEM) products on this storm.

Early Tuesday the GFS still had heavy snow in the metro and points north.

Again, this is a total model fail. How is the NWS supposed to react as they approach the time frame to issue winter storm watches when this is their best output from U.S. forecast models?

By Thursday morning the GFS successfully grabbed hold of a more southerly storm track and the sharp snowfall cutoff on the storms northern edge. But the GFS still didn't move the snow far enough south to make for an accurate Twin Cities forecast.

While the GFS picked up on the southward shift, it was too late to provide a timely and 'actionable' forecast for most concerns in the Twin Cities.

Canadian GEM model: B

The GEM model from the Canadian Meteorological Centre did better with this storm track. The GEM consistently laid out a more southerly storm track, in contradiction to NOAA's best forecast models. The GEM looked like an outlier for much of the week, but grabbed hold of the notion of laying out the heavier snows south of MSP. Still, it produced a few inches in most of the Twin Cities, which was too high.

European model (ECMWF): A-

"King Euro" once again outperformed all models in keeping the heaviest snow south of the Twin Cities throughout last week. As you can see on the graphic below, the Euro projected 1.5 mm of liquid precipitation at MSP. That's .06" liquid, which would translate into about half an inch of snow.

2-23-euro
Euro snowfall output from Thursday for MSP Airport via Norwegian Met Institute.

To be fair, the Euro did also suggest a more northward shift with more snow in the Twin Cities at times last week. But overall the Euro performed better than all other forecast models.

It's a reminder to us as forecasters that the Euro, while not always the best with every situation, is the odds on favorite most days. There is statistical proof that the Euro is still the best weather model on the planet.

Many of us who are daily users of numerical weather forecasting products continue to ask the question: Why can't NOAA's U.S. forecast models perform as well as the European Model?

Stay tuned.