But, considering how difficult this forecast was, we’re not beating ourselves up over it. Importantly, we correctly predicted when the precipitation would begin and stressed that the roads would be a mess. Our forecasts helped inform prudent decisions and keep people safe, which is the ultimate goal.
The most important thing not to forget is what you and your fellow meteorologists got right which was how dangerous the morning commute was going to be and it was. Jurisdictions deciding to close last night based on that information was the right call.— morbrem (@morbrem) February 18, 2021
For this event, the devil was in the details, which were incredibly challenging to pin down, as we explained in our forecasts.
The main forecast problem was whether more of the precipitation would fall as snow or as sleet. All of the computer models projected at least some accumulating snow before a changeover to sleet. But they were wrong.
Whether cloud layers generate snow or sleet depends critically on the layering of temperature in and below those clouds.
If a mild layer of air above 32 degrees intrudes at an altitude around 5,000 feet, it can partially melt snow that formed at higher altitudes and turn it into sleet. Models showed such a mild layer in their simulations but not until after several hours of snowfall Thursday morning.
But the mild layer arrived earlier than models projected. In fact, the routine weather balloon release at Dulles Airport revealed its presence Wednesday night. In our update at 9:30 p.m. Wednesday, recognizing that observations were warmer than models, we indicated there was a chance more sleet would fall than anticipated. Washington would be right at the dividing line between snow and sleet.
I didn’t think the forecast was bad. The last one I read before bed said pretty clearly “one degree 5,000 miles up will be the difference between inches of snow or accumulating sleet.” When I woke up, I thought “huh, guess it was one degree warmer.” 🤷🏽♀️— BON, Esq. 💛🐝 (@BON_Esq) February 18, 2021
Ultimately, instead of the 1 to 3 or 2 to 4 inches of snow and sleet we called for, most locations saw a heavy coating of up to 2 inches of mostly sleet, with just a little snow in the mix.
There are other aspects of the forecast you could quibble with:
- On balance, while there were some heavy bursts of sleet, sometimes mixing with snow, the precipitation was generally a little lighter than we anticipated. So, visibility was not as impaired as we thought it might be, and the wintry mix did not accumulate as rapidly.
- The steady frozen precipitation ended rapidly midday, whereas our initial expectation was that it wouldn’t taper off until at least midafternoon.
- Due to the overall lighter intensity and shorter duration of the precipitation, this event would rank more as a Category 2 or “disruptive” winter weather event on our winter storm impact scale, whereas we touted it as a Category 3 or “significant” winter storm.
We have seen a lot of comments on social media about how we’ve over-relied on models, which have been inflating snow potential. Here’s our take on that:
1) Models are by far the best tools we have for informing our weather predictions. They have taken massive strides in recent decades, and forecasts would be vastly inferior without them.
2) While we use models as guides, we also take into account their biases and limitations and our own knowledge of how the atmosphere works. For every event this winter, we have adjusted our snow forecasts based on their lowest projections, knowing that the setup for snowfall has been less than ideal in most cases.
This has been one of those winters where the Washington area just hasn’t been a great spot for snow given the prevailing storm track. We’ve tried to account for that, but probably not enough.
We can also recall winters in which snowfall has frequently exceeded model simulations. Sometimes certain weather patterns lend themselves toward snow booms and others toward snow busts. This winter has clearly been the latter.
As humbling as trying to predict snowfall has been this winter, if you look back at the balance of forecasts we’ve made and the commentary and explanatory material that has accompanied them, there haven’t really been any huge surprises. Compared with other forecasting teams in the region, our predicted amounts have almost always been among the lowest and closest to reality, even if missing the mark at times.
Still, we wish we could have done better. That’s especially the case with this event.
When we issued our winter outlook for this season, we wrote regarding tricky snow forecasts: “We think we’ll have our share of them this winter.” Too true.
We still have another four to six weeks in which it could realistically snow, and we will try to learn from what’s happened so far this year to adapt our forecasts accordingly.
We’ll also try to explore ways of more effectively communicating snowfall uncertainty to manage expectations. Of all winters, this one calls for that.
Jeffrey Halverson contributed to this report.