* Snow totals from the National Weather Service *

After we forecast a coating, the metro region got a surprise 1-3″ of cotton-candy-like snow this morning (officially 2.8″ at Reagan National Airport). A very localized, difficult to anticipate area of moderate snow developed, that just happened to focus on the D.C. area and nowhere else.

I’m not going to beat ourselves up over this forecast, as we couldn’t have done much better given available tools (and no one else did, including the National Weather Service).

Models didn’t capture it

This was simply a very localized event that models could not reasonably simulate until the snow had effectively started, if even then. The NAM model, run at 7 a.m. this morning (shown below), simulated 0.05″ of liquid equivalent precipitation for D.C. – which at most – would’ve suggested 0.5″ of snow. The GFS model simulated 0.04″ of liquid equivalent precipitation. The European model predicted similar amounts.

NAM model simulation of today’s snow at 7 a.m. this morning (StormVistaWxModels.com)

While the models clearly underplayed the amount of snow which fell in the D.C. area, they were right in most other areas – north of Baltimore and south of Fredericksburg, which got coating amounts or less. A narrow, very localized heavy snow band just happened to set up right over the D.C. area that the models and human forecasters simply weren’t good enough to capture.

(Around 7 or 8 a.m. this morning, high resolution short-range models – not shown – did begin hinting this event might be bigger than suggested by the other models.)

Radar didn’t help

Radar offered few clues we were going to get dumped on. There wasn’t some massive plume of precipitation coming at us from the west. Rather, the snow literally developed right over top of us this morning. In fact, around 7 a.m., it had started to snow around Baltimore and in central Virginia, but there was a big snow hole over the D.C. area.  It then rapidly filled in after around 8:30 a.m. with little warning, while the snow to the north and south faded away.

Radar at 6:52 a.m. this morning showed D.C. split (RadarScope)

Atmospheric dynamics were there

Although the conventional tools we use to forecast offered little indication of the snow boom, critics could point to the presence of a jet streak – or river of strong winds at high altitudes – which can sometimes can enhance snowfall rates when it coincides with an incoming disturbance. Yes- that was there and perhaps we should’ve mentioned the possibility snow could overachieve in light of it. But again, when jet streaks intensify snow, it typically happens on a very small scale which is impossible to anticipate far ahead of time and models have trouble pinpointing.

NAM model shows a jet streak – or region of fast high altitude winds – which coincided with area of heavier snowfall. While the model showed this jet streak, it did not simulate a lot of snow underneath it. (StormVistaWxModels.com)

Forecast wasn’t catastrophic

While we clearly missed on the amount of snow which fell in the region, it wasn’t terribly consequential in terms of the impacts on people’s lives. For the record, this is what our forecast issued at 5 a.m. this morning (prepared overnight) said:

Clouds control our skies today with mainly light snow in the mid-to-late morning hours. The snow could mix with or even change to light rain before ending by around noon or 1pm in the city, mainly south and east of the District. A coating or so is possible, mainly on grassy surfaces, with the best chance north and west of the city.

Yes, it snowed more than we thought and was heaviest south of town rather than north. But we got the timing essentially right and its impact did not significantly exceed what we forecast since most of the accumulation was on grassy areas. Roads, aside from some slush, were not hazardous for the most part.

The bottom line is that forecast busts (or “booms”) like this – where the forecast for a coating of snow turns into 1-3 inches – are difficult to avoid given the state of weather forecasting models. Until models improve to the point they can capture or resolve very localized features, they’ll remain a possibility.