Quick Snowfall Verification

Well, it's past mid-day Saturday in the upper midwest now.  So how did all of those ensemble predictions (and my interpretation of them) do in terms of their snow forecasts?

By the regional radar mosaic right now, the timing seems to be about right--the snow bands are pushing out of northern Illinois and down into central Illinois right now.

Fig 1 -- Base reflectivity radar mosaic for the Ohio River valley for 1908Z, Dec. 4, 2010.  From http://www.weather.gov/radar.

So at least that seemed to verify.  But what about the snowfall amounts?

One thing I've found in looking at snowfall reports over the years is that it's difficult to get a good accurate summary, particularly as an event is occurring or immediately after it.  Part of this is because it's far more difficult (and more expensive) to measure snow depth automatically and remotely transmit it.  Heated rain gauges melt the snow to give a snow-water equivalent, but that doesn't tell us much about the depth of the snow pack.  There are some automated sensors that do it--one way is by using what amounts to a flat scale sitting on the ground that measures the change in weight on the scale as snow accumulates and then, using the density of ice and surface area of the scale, calculates a snow depth.  Another way is to use a mini radar-gun-like device mounted on top of a pole and pointed at the ground.  As the snow accumulates on the ground, the radar gun (or laser) will measure the change in distance between it and the "surface" of the snow.  From this, you can get a depth measurement.

For the most part, though, snow depths and rates are measured by hand.  While it may be as simple as sticking a ruler in the snow and seeing how deep it is (and often those are the reports that we get), the National Weather Service (and associated groups like CoCoRaHS) have certain guidelines for measuring snow.  These include things like staying away from buildings and using a "snowboard" underneath the snow to measure from instead of just measuring over grass.

So, anyhow, back to our question--how did snowfall amounts match with our ensemble predictions (which, if you recall, had generally large areas of 2.5 inches of snow but some areas of possibly 5 inches--assuming a 10:1 snow-liquid ratio)?  There are many places that offer snowfall reports.  First, let's check the National Climatic Data Center--NCDC:

Fig 2 -- NCDC One-day snowfall reports as of 1500Z, Dec. 4, 2010.  From the NCDC website.

Those are not a whole lot of observations in our area of interest.  In fact, that's really very few.  There are a few 0-1 inch observation dots along that swath, though in the Mississippi River valley there are a couple that get up to 1-2 and 2-3 inches.  Of course, this has a cutoff time of 1500Z and it's now past 1930Z--that's another four and a half hours since the above image that hasn't been accounted for.  But, even so, surely we can do better than this map.

Let's try NOAA's National Operational Hydrologic Remote Sensing Center.

Fig 3 -- NOHRSC Snow Depth map as of 06Z, Dec 4, 2010.  From the NOHRSC website.

Now this is a snow depth map in general--so places that had snow before (like anywhere in the northern half of Minnesota and points north) will have snow depths shown here, but from previous snowfall.  However, that swath from southeastern Minnesota down through the Mississippi Valley and into northern Illinois is all new.  But look at the timing of this map--I'm assuming that the "06" after the date on it means that this was current as of 06Z--midnight last night?  So today's snowfall amounts since midnight aren't even included.  Even so--there's a big swath along the Mississippi River that they say has between 2-3.9 inches--within our range of estimate from the models.  But that was just through midnight--clearly a lot has happened since then.

It has been my experience that when you want good snowfall amounts in as close to real-time as possible, the place to go is the local weather service office page.  In this case, the LaCrosse, Wisconsin, office is right in the middle of our snow swath.  So let's see what they say.

Fig 4 -- Map of estimated snowfall amounts for a 24 hour period ending 8AM CST, Dec 4, 2010.  From the NWS LaCrosse website.

Now this is the kind of map I like to see.  Much higher resolution, though I'm not sure exactly how they compiled this map.  This shows a swath of amounts of 6-8 inches from extreme northwestern Illinois through southeastern Minnesota and 8-12 inches in parts of southeastern Minnesota.  Yikes--far more snow than we were thinking.  And this is just through 8AM local time--more snow could have accumulated since then, particularly in the southern half of this swath.  So wow--my quick snowfall amount interpretation from these model ensembles yesterday was a bit low.  But the geographical extent of the snow and the timing of the snow looks like it was nearly spot on from the ensemble analyses.  Perhaps slightly further southwest than I personally was thinking, but in general--very, very close.

If you don't quite believe this map, the NWS LaCrosse office has put together a webpage giving a brief overview of this event including the snowfall observations they have received.  Yes, those numbers do match their graphic--and there are some pretty impressive numbers, particularly from southeastern Minnesota.  There are also a whole lot of observations for them to work with--much more than the scattered few dots that NCDC had.  There just seems to be a trickle-up effect for those observations to get to the big national analysis centers.  That's why I usually like to go straight to the source.

This still leaves the question of why the snowfall amounts that I forecast were so low. The weather service advisories did slowly increase their forecast snowfall amounts as this system approached and I think they ended up covering it rather well.  Back on Wednesday of this week, the Milwaukee forecast office specifically stated in their forecast discussion that they were using an 11:1 snow-liquid ratio to calculate their snowfall amounts--similar to the 10:1 ratio I was using.  Things obviously changed since then.  Like I said in my last blog, when I get time I'm planning to look into the different ways we compute snowfall depth from liquid water values.  So look for a blog post on that coming up sometime soon.

Also, for any people out there with a desire to become more actively involved in the weather community, the CoCoRaHS organization I mentioned (COmmunity COllaborative RAin, Hail and Snow) is a network that allows the everyday citizen to use their rain gauges, snowboards and rulers to submit reports of rainfall, snowfall, and hail size at their location.  That's how the NWS gets a whole lot of their reports--from your average everyday people who have a healthy interest in the weather.  Feel free to explore their website if you want more information about how to join.