I called that snapshot out on our department WX list this morning trying to figure out what was going on. There have been some issues in the last few days with the system doing the post-processing, and a few of the post-processed files (though not the actual forecast) got screwed up. I clarified a few points on the 12km and 4km snow output today with Cliff, and thought I'd share them here.
So as those of you who poke around the forecast model output may have noticed, there are two fields for snow: "model-snow" and "snow." Model snow is the precip that the WRF model categorizes as snow, namely precip at or above the model freezing level. In the real world, precip continues to fall as snow as much as 1000ft below the freezing level, as it takes some time for the flakes to melt in the warmer air, and depending on the ground temp accumulations can still occur. To try and account for this, the output precip fields are post processed, and precip up to 1000ft below the freezing level is recategorized as snow, scaled by some empirical liquid water to snow accumulation factor, and added on to the amount the model considers snow. This will catch lowland snow that might hit the hilltops in the metro area when the temperatures are borderline and the model prematurely melts precip.
The error in those couple of outputs was the result of the freezing level field being corrupted at a few specific times by some failure that is still being diagnosed. The corrupted files put the freezing level right at the surface throughout the domain, meaning that all precip was considered snow, even though freezing levels will in reality be around 4000-3500ft. The "snow" field was then recomputed assuming the freezing level at sea level, resulting in spuriously high snow accumulations at a few specific times. Note that the "model-snow" field never displayed a glitch, as the actual forecast and output was not corrupted-- this only occurred in the additional post-processing to create the "snow" field. Sadly, this event will not be putting three feet across the crest in a three hours stretch... pity that, really.
In short, if you want a conservative estimate of snowfall, use the "model-snow" fields. Glancing at the difference between "model-snow" and "snow" can help show where some marginal accumulations might occur and potentially play havoc with a commute, but for skiing purposes, "model-snow" is probably safer.
Thanks for the info about the problems, and also the explanation of model-snow versus snow. I had been wondering what the exact difference was since the model-snow graphics started being included last year.
