The snow plumes are in - now the fun begins
UPDATE: The 12Z plumes are in; Richmond has about the same range but the mean has increased to over 9 inches. The increase in snow by hour is about the same but the storm is slightly delayed. About half the ensembles are levelling off (snow stopping by the last forecast hour).
This morning, we're on the cusp of the NCEP SREF snow plumes being available for this weekend's snowstorm. This is one of my favorite model tools. It uses "ensemble" (or slightly different perturbations of) the SREF model, run at the same time (in this case, 09Z which was overnight), so it gives us a range, and a mean, which is the most likely amount, based on current information.
(Disclaimer: The official forecast can be found on AccuWeather.com and in this news story).
Because the window of the model only goes out to Friday night, we can't get totals yet and we have to concentrate on the heavy snow areas farther westward (i.e. earlier in the storm). Here's the total snow graph for Roanoke:
What this says is that the model predicts between zero and 18 inches of snow -- not that useful on face value, but the fact that 10 ensemble members are over 10 inches indicates a significant chance at a major snowstorm. The mean, however, is 8 inches, and if you were forced to make a one-number forecast from this model, that would have to be it. You would not, however, want to forecast more than 18 inches of snow for this time period because no ensemble member predicts that.
How are the models trending over time? This can be important. Above is the mean from the last four model runs. The last three have been at a higher angle (i.e. snow piles up faster) and the next run could be even higher. Not all the ensembles are leveling off during the last time period either, so the storm may not be over yet, on this graph.
Yes, the SREF is another model, but unlike the GFS or European, it is high resolution and is designed to be used at a local level and the snow algorithm was designed with an algorithm specifically for this kind of output.
Other models guess at snow amounts based on static snow ratios, which can be significantly inaccurate. The SPC says of these graphs that "Total Snow" is: "Cumulative snowfall forecast in inches. The SPC uses an equation [Snow-Liquid-Ratio SLR = 1000/(100+6T)] based on a paper by Boone and Etchevers (2001) to compute SLR, where T is the warmest saturated temperature (dewpoint depression < 2degC) in the lowest 300 mb. A cap is set at 40:1 for T at -12.5 degC (9.5degF) and colder."
Report a Typo
