Euro weather model outshines U.S. version in forecasting our big snow event
Warning: This blog contains a lot of weather-geeky language and charts. May require cup of coffee to wade through...
Now that we have the "final score" of how much snow fell around here I thought it would be useful to see how our main forecast models did in predicting the storm. As you remember, they got a big, fat F on the Oct. 15 windstorm, but fared a much better grade this time around (whew!).
Although they weren't perfect.
First, let's summarize some of those "final scores": Seattle officially had 7.1" at Sea-Tac; while about 1-5" fell in Seattle proper depending on elevation (1" city; 3-5" hilltops); 2" fell in Everett, 2-4" in SW Snohomish County, about 4-8" in Bellevue proper, 6-8" in the South Sound areas like Renton and Federal Way, 9-12" in the Eastside foothills, 11-15" in the SE King/Pierce County foothills, 5-7" in the Tacoma area, and about 4" in the Olympia area and 4-5" in Kitsap County. Much less fell in the far North Sound.
There are a lot of models but I'm going to focus on two of the main ones: The Euro (Run by a joint venture by a number of European countries) and the American WRF-GFS. The Euro model actually picked up on the snow as early as last Tuesday with a sudden prediction of several inches of snow around the region, including 14 inches in Seattle. (This shows an ensemble run where the model is run 51 times with a minor tweak, and each model's results are given at the top with the overall average given in tabular form at the bottom. The blue columns represent the main model's decision (14" by Monday evening) while the green bars are the average of the 51 other models it ran (going much lower -- closer to 6")
That snow forecast went away the next model run (whew!) and the models were quiet for a couple of days, but then the snow forecast for Monday returned on Thursday and Friday:
The Euro model (the pink/blue one on the left) had 11" at Sea-Tac, 7" at Boeing Field and 12" in Renton and in the SE foothills. It turned out to be way overdone on the I-5 corridor, but was looking pretty close to what the foothills ended up with. The American (GFS) model (The blue/purple/red/green one) was predicting nearly a foot of snow in South Seattle and Tacoma and Olympia. Also overdone, but five days out, properly announced a heavy snow was likely.
At this point, we local meteorologists certainly had our ears piqued but wanted to see a lot more evidence of consistent models runs before we run out with those kind of snow forecasts.
The Euro's best forecast actually came Saturday morning, when it predicted 7" at Sea-Tac Airport -- just about right on bulls-eye. What would be the GFS's best forecast came out Saturday evening:
That correctly had less to the north, more to the south and had 6-8" in the South Sound, 2-4" in the North Sound and pushing 10" in the foothills. It overdid Kitsap by about 2-1 margin and has quite a bit more in the heart of the city than what fell, but then again you're talking about a data point every 4 km (2.5 miles) so it doesn't pick up on the terrain as much. Some spots like West Seattle hill did get over 5 inches.
Late Friday into Saturday was when we finally decided we had enough evidence of consistent snow forecasts to start actually writing about it.
But Friday and Saturday, you'll notice the forecasts had been a bit conservative versus what the models had been painting. We had been going with a "most likely scenario" of 1-3 inches in the Seattle with hints of more. This was in part because this event was going to be really on the edge for temperatures and had significant bust potential, despite the models' insistence there would be more.
Well, more evidence came with those gaudy numbers again on Saturday --- The Euro was all aboard but the GFS took a pause and stepped the forecast back. But both were on board again with the Saturday night runs and with time getting closer, we started ramping up the alerts. When some of the overnight Saturday higher resolution model runs that only go out 48 hours started painting 4-5" in Seattle too, adding a final bit of consensus that there was going to be a significant snow in Western Washington, were now on board, and the National Weather Service hoisted Winter Storm Warnings for the much of the region Sunday morning.
What many might not know was...Sunday morning, the Euro model backed off. (Eeeek!) All of a sudden, the Euro which had a range of 8-11" for Seattle over its last five runs, now "only" had 4" with less in the surrounding area. Was the cold air not coming? We had been playing to conservative and then the moment we pull the trigger, everything falls apart? It was an anxious time for sure. The GFS sill had some reassurance but was now down to 2-4" in the city as well. But the higher resolution models were still staying "go".
But Sunday evening as we could monitor the rain changing to snow, we felt better. The hourly models runs that only go out 15 hours were starting to confirm most areas would be cold enough for snow, and the final Sunday evening models were back to a pretty significant snow across the region.
Here's a recap of the Euro model performance, showing its forecast for Sea-Tac Airport over each run since late Friday:
Yes, this is a bit difficult to grasp but it's showing accumulated snow with time from left to right. From top to bottom are each of the last model runs. I marked some of the vertical columns as A, B and C -- marking predicted snow by time frame. The A column was predicted total snow by 10 p.m. Sunday, the B column was by 4 a.m. Monday and the C column was by 10 a.m. Monday. So "C" was essentially the storm total prediction.
The top bar was the Friday night model run that predicted 4" in Seattle by 10 a.m. Monday, Starting Saturday morning (row 3, the "96" hour row), the Euro predicted over its next five model runs Sea-Tac would have 8, 7, 7, 7, and 9 inches by 10 a.m. Monday. As we saw that was near perfection! (Remember it was 7.1" that was the actual total). The only issue was that the model incorrectly predicted Seattle would get another 1-4 inches during the day Monday. It did snow some more, but had warmed up just a touch to where it wasn't really sticking or adding to the tally. Still, an impressive show!
What about others city forecasts by the Euro for total snowfall by 10 a.m. Monday?
Some caveats: The Euro went on to predict Olympia to get several more inches of snow after 10 a.m. and had 8-13" by the end of the day. Oops. Tacoma was the same way; it had about a foot of snow in the city by the end. So the Euro did reasonably well, especially for what to have expected for the Monday morning commute. But way overdid what was expected of the Monday evening commute.
As for GFS, I don't have raw numerical output for their snow totals, but here are the maps:
And here are the estimated forecasted snow totals for certain cities, just gleaning off the map for last 6 runs (snowfall in inches):
The UW high-resolution map only runs out a four day forecast for the GFS so I didn't include the lower-resolution model forecasts from Thursday in the chart to be consistent, but they did have about 4-6" in Puget Sound. But then Friday began a string of big under-forecasts. It seemed to figure it out on Saturday evening...then backed off a bit. It still did OK on areas close on the I-5 corridor but way underestimated the 8-16" that fell east of I-5.
So overall, the Euro seemed to be quite the winner again. As UW Professor Cliff Mass has written before, the U.S. models badly lag behind the performance of the Euro model Some efforts are being done to catch up, but in the interim, it appears the Euro is still doing quite well.