Snow amounts on Sunday, February 19 (left) compared to CWG’s forecast for snow amounts on that day, made on February 17 at 11 p.m. (right)

Looking back at the forecasts, we first mentioned Sunday storm potential on Monday morning, February 13. I wrote: “there’s a slight chance a storm develops to the south. So we’ll keep you posted if that looks like it could be a real threat.”

Here’s how are our forecast evolved after that:

* Monday afternoon, I showed some of the long range models simulating the weekend storm, noting: “The usual uncertainties are all present about how strong a storm will form, the track the storm will take, and whether the air will be cold enough for snow. But the situations bears watching.” I also noted 6-day out model forecasts are an “eternity.”

* Matt Rogers wrote Tuesday: “But Sunday into Monday (what is it about President’s Day around here?) sees a storm system passing to our south (and turning up the coast?). It may come close enough to our area to offer clouds and precipitation Sunday into early Monday. I prefer the more tranquil scenario...” Wes Junker posted an analysis later that day, concluding: “The most we can now say about this weekend is there is a chance of rain or snow, but it’s even more likely the entire system stays far enough to our south that we remain dry.”

* On Wednesday, we said the storm could go either way - as a hit or a miss. I wrote in the afternoon, “The potential exists for a major East Coast storm this weekend, but it’s just as likely the storm passes to our south, harmlessly out to sea.”

* By Thursday, we began to think the storm scenario was becoming more likely. Matt Rogers wrote: “Question marks begin for Saturday night and Sunday as various models argue different scenarios, but they are trending stormier.” Thursday afternoon, we issued a Snow Lover’s Crystal Ball, indicating a 30-50% chance of 1” of snow. On Thursday night, after models - almost universally tracked the storm close enough to bring significant precipitation , I wrote: “So it looks like the necessary moisture and cold air may be coming together for accumulating snow Sunday. But - with the storm still 60-70 hours away - significant fluctuations to the forecast are possible.”

* Friday was the day of a little weather prediction whiplash. In the morning and afternoon, we were slightly favoring the storm scenario but cautioned the storm could be a “boom or bust”. Wes Junker described the situation as “a nearly impossible winter weather forecast” and explained all of the complexities. In the afternoon, I wrote 1-4” was most likely but that less than 1” (especially north) and more than 4” (especially southwest) were about as equally likely. But Friday night, when model consensus showed the storm missing us, I backtracked, concluding “we now favor the low scenario which is for less than 1” and possibly no snow Sunday in much of the metro region.” That’s when we issued our first call accumulation map forecast, which we ended up sticking with the entire weekend.

Throughout the weekend, we emphasized little or no snow was most likely, but that a small shift in the storm - of say 45-60 miles - could bring accumulating snow into the metro region. And that’s how close the accumulating ended up being. Fredericksburg got about 1-2” of snow.

The hardest forecasts (for both human and computer models) are always those where a given region is on the edge of the precipitation...because the most minor shifts in the storm track can be the difference between cloudy skies and meaningful snow. These situations REQUIRE forecasters to pay very close attention to model trends and make the necessary adjustments to the forecast. It’s not going to be possible to just issue one forecast and stick with it. If you happen to get it right doing that, it’s pure luck, not an indication of skill.

One other point: adapting a forecast to changes in the models is not “model hugging” but a recognition that models are the best tools we have for trying to figure out where the edge will be. Some forecasters on Facebook got burned by trying to outsmart the models. While forecaster experience, understanding of the weather pattern, and intuition can help in overcoming individual model biases - they will not beat model consensus - which is what we had by Friday night (more or less).

Bottom line: This was a hard forecast - as we were on the edge of getting snow or no snow in model simulations all week long. We tried to communicate that and make the right adjustments. While we necessarily had to flip flop a little based on trends in the model consensus (which flip flopped), we ultimately made the right adjustment and the right forecast.