North and west of the Beltway, our forecast for today’s storm of 2-5 inches of snow was spot on. But inside the Beltway and to the south and east, it failed for reasons that may sound familiar.
In short, temperatures inside the Beltway and to the south and east were just a degree or two too warm for snow to accumulate and precipitation was modestly lighter than we expected.
We’ve seen these circumstances before – the most recent example being 9 months ago on March 6, aka the Snowquester debacle. Of course, today’s (half) bust was relatively minor by comparison. Whereas we forecast 6-10″ in Snowquester and the underachieving areas got 0-1 inch, this time we forecast 2-5″, and the the snow-challenged areas got 0-2 inches.
Relative to Snowquester – in which we became overconfident a dump of snow was coming – we took baby steps with this forecast in that we clearly laid out the bust scenario, assigning it a 20-25 percent likelihood. As late as 10:45 p.m. last night, CWG’s Ian Livingston wrote:
“…there is one scenario in which … temperatures don’t cool down enough and we end up with some sleet and closer to an inch of snow.”
Also, in an update yesterday afternoon, Wes Junker discussed about five different ways this forecast could wrong.
Despite these qualifiers, we can’t be satisfied that we forecast the possibility of disruptive snow when that failed to occur over a sizable chunk of the region – with consequences for people’s daily routines. When we make a similar mistake more than once, we need to try to identify the root cause. In my view, it relates to an over-reliance on models.
When examining the models yesterday, we took sub-freezing temperatures during the storm for granted because they all simulated that. But, in reality, temperatures were 32-36 degrees when the snow and other precipitation types fell and simultaneously melted away in the bust zone.
Around 10 p.m. last night, I noticed temperatures were running warmer than forecast and got a little nervous, but figured heavy precipitation would bring them down (through dynamic and evaporational cooling). But, as the precipitation was only moderate, it only brought them down part of the way – not enough for snow to substantially accumulate.
What I take home from this is that we have to be a little a bit more skeptical of the models and take clues from what’s happening outside our windows.
Over the years, I’ve now seen snow fall short of expectations repeatedly in the city because temperatures were a little too warm preceding the storm. The lesson here is that unless temperatures are subfreezing before the first flake falls, we have to be very cautious in accumulation forecasts. [The exception to this rule should only be when there is a clear and present stream of cold air feeding into the storm (in this case there wasn’t) or in the rare case we have very high confidence the precipitation will be heavy enough to overcome the lack of cold air (this was the case during Commutageddon, January 26, 2011). And we then have to be skilled in identifying these exceptions.]
We’ve made in progress in how we communicate the range of snow possibilities, now we need to get better at refining what’s most likely to happen.
A quick note about what we did well
In the spirit of giving ourselves a little credit, we did correctly predict the onset and finish time for the storm. And, as I mentioned up top, our 2-5″ inch forecast verified well for areas north and west of the Beltway.
Also, our forecast of 2-5″ was actually among the, if not the, most conservative of the different forecasts from various outlets (National Weather Service and TV stations) and thus, relatively speaking, among the most accurate. While we mentioned totals of 5″ or more as a low percentage possibility (a 25 or 30 percent chance), we were correctly skeptical and made that clear.
We welcome your feedback on our forecast, and how we communicated it…