A major news story unfolded over the weekend as the Southeastern US got slammed with a snowstorm that dropped uncommon snow totals over the area, causing widespread travel disruptions. This region of the country is not accustomed to snowstorms of this scale and many municipalities were not prepared for it. Making matters worse, there was a major forecast bust in this storm, which shared key characteristics with a similar forecast bust that led to a high impact snowstorm hitting NYC a few weeks ago on November 15th (and may have prompted the ouster of the director of NYC Office of Emergency Management). For example, Richmond, VA had a forecast going into Sunday for only 1″ of accumulating snow, but in fact received 11.5″ when all was said and done – a near record-breaking storm.

Below, I’ll provide a “post-mortem” analysis of why forecasters missed the mark so badly in this case. The overall lesson here underscores the difficulty of forecasting snow when temperatures are expected to be hovering close to freezing, especially in coastal storms where the precipitation gradient can be quite sharp.

Dry air at the outset of the storm

Soundings from KWAL (Wallops Island NASA Launch Facility, which we can use as a reasonable proxy for areas in Virginia heavily impacted by snow) at the outset of this storm showed very dry air at the low levels of the atmosphere. This is indicated by the large gap between dew points (green line) and the environmental temperature (red line) on the Skew-T diagram below.

Since I think most people reading this are probably not familiar with Skew-Ts, let me provide a brief exposition. These charts are densely packed with data and can be difficult to read. To orient yourself, know that the y axis on these represents pressure levels from the surface (~1000 mb) all the way up to almost the very limit of the atmosphere at 100 mb. Pressure levels are also related to altitude, though this relationship is not linear because it depends on temperature. The x axis on these charts shows temperature in degrees Celsius. However, note that the lines of temperature are actually slanted at a 45 degree angle and not straight up. The dotted blue line to the right marks the 0 degree mark, critical for determining whether precipitation is frozen or not.

So back to the Skew-T at hand – notice that above the 700 mb layer, the dew point (green) and environmental temperature (red) lines were essentially overlapping. This indicates a layer of air that’s reached saturation since by definition, dew point is the temperature to which the air would need to be cooled to be saturated. When you see a thick layer of dew points and temperatures meeting, it generally indicates ongoing precipitation (thinner layers like this can indicate clouds). In this case, what’s happening is that precipitation is falling from about 400 mb down, but from 700 mb and below, the air is very dry.

With this set up in place, we have excellent conditions for evaporational cooling. As precipitation from above starts to saturate the layers below (some of the precipitation evaporates into the dry layer), the temperature actually cools because evaporation is a phase change of water that requires an input of energy (heat). This is exactly the same mechanism that occurs when you exercise and sweat, or when you step out of a shower (even a cold one) and feel cooler. The net effect of the evaporational cooling in this case, like in the storm that hit NYC in November, was to keep environmental temperatures below freezing for longer than expected (shifting the red environmental temperature line to the left on a Skew-T), allowing snow to fall and accumulate for a longer period as well.

The issue for forecasters here, and for NYC on November 15, was that the models were not all in agreement about how dry the low levels of the atmosphere would be at the outset of the storm. Forecasters are trained not to rely solely on just one model’s depiction of upcoming events, even though in this case, some models had what turned out to be a much more accurate take on dry air. As we’ve seen, the difference of a degree or two when temperatures in the atmosphere are close to the freezing line can have serious consequences for tangible weather impacts.

Frontogenesis and mesoscale (localized) banding

When coastal storms form off the East Coast during the winter, the temperature differential between the warmer air south of the storm’s core and the colder air to the north can lead to frontogenesis, which is the process of the formation of a frontal boundary. In these storms, the result is a coastal front. During this process, a mesoscale circulation forms as atmospheric dynamics attempt to restore equilibrium between cold and warm airmasses. This circulation can greatly enhance lift, a critical ingredient for heavy precipitation, as well as helping cool the air columns. For coastal storms during the winter, the result of strong frontogenesis is the development of narrow, but intense localized bands of heavy precipitation. The difference between an area impacted by a band like this can easily be more than 0.50″ of liquid equivalent, which if you convert to snow using a standard 10:1 snow-to-liquid ratio is 5″! The trouble with these mesoscale features, as is the case with thunderstorms, is that even the most advanced forecast models do not have sufficient resolution to accurately capture features on these scales. That means it’s often difficult to know for certain if/where/when one of these bands sets up and for how long – a critical, high impact detail that can make or break any forecast.

As it happened, with this storm, stronger frontogenesis than forecast took shape. The North American Model (NAM) actually had a pretty good handle on this, but as with the NYC storm, forecasters didn’t put all their eggs in one basket and side with this solution.

Cold air damming

Along the Eastern Seaboard, certain orientations of high pressure systems can lead to an effect known as cold air damming. This occurs when high pressure centers of Canadian origin set up northeast of the mid-Atlantic and Southeast. Anti-cyclonic clockwise flow around these highs brings cold air around the core of this high into the East Coast with easterly winds. At some point, these winds start to hit the eastern flank of the Appalachian mountains. Because cold air has higher density, the mountains provide an effective barrier to the westward (and upward) progress of this cold air. This then leads the air to gradually turn to the left (south) and progress further south than would otherwise be possible without the cold air damming effect. This is visible from the following surface analysis where you can see surface isobars linked to the high pressure center “sagging” south along the eastern edge of the Appalachians. This phenomenon can provide a critical shot of cold air in advance of a storm that can tip the balance from a rain event to a snow/mixed/frozen event. Forecasters probably did have a decent handle on this, but I mention it because it would have helped in keeping cold air in place prior to and during the beginning of the event.

What are some takeaways from this?

Given that this scenario has unfolded twice this season, a key takeaway for forecasters should be to have heightened awareness of snowfall totals exceeding model consensus when one or more of those models is indicating the possibility for both strong frontogenesis with a coastal storm like this and very dry air preceding such a storm. Ideally, forecasters and emergency managers should be in close communication about probabilities of exceeding forecast totals as soon as evidence and observations show a colder scenario unfolding. If possible, these details should be passed on to the general public by highlighting the uncertainty that exists and probabilities, even if they’re not high, of exceeding forecast totals dramatically. Municipalities should have a fallback plan for fast mobilization of personnel and equipment for snow removal in the event that a forecast bust of this magnitude starts to look more likely during the early onset of a storm when we can verify things like dew points, and observe trends of mesoscale bands on radar.