Friday, January 19, 2018

Probabilistic Snowfall Forecasting

For decades, snowfall forecasts have typically involved the issuance of a range of accumulation amounts, typically (but not always) based on a factor of two.  For example, 3-6 inches, 4-8 inches, etc.

I have no idea why.  Perhaps it is a convenience thing.  Maybe people like it that way.  I don't know what that range even means.  Does it represent the middle 50% of possible outcomes, with a 25% chance of more and a 25% chance of less?  Does it represent the middle 80%?  Why always use a factor of two?  Sometimes the range needs to be bigger, especially in longer range forecasts. 

And then there is my favorite, "higher amounts in favored locations."  What the hell does that really mean and how do you verify it? 

There was a time when snowfall forecasting was truly guesswork, but things are changing.  Computer models are now capable or will soon be capable of simulating smaller storm details, what meteorologists call "cloud scale."  Ensembles can be used to better estimate the future outcomes.  There remains much work to do, but there is great potential to dramatically improve snowfall forecasting. 

The National Weather Service is now producing experimental probabilistic snowfall forecasts, and they are available at  They provide much richer information about storm potential than a simple range.  For example, ,aps are provided showing the likelihood of snowfall above several thresholds, an example of which is the probability of 6" of snow or more for the period from 5 AM today to 5 PM Sunday, shown below. 

Source: NWS
They also provide a table with snow amount potentials and probabilities of snow within certain ranges, shown below, as well as above certain thresholds. 

Source: NWS
Readers of this blog are snow lovers.  Start perusing these forecasts and provide feedback through the links on the page. 

Now, to clarify some of my scattered comments to yesterday's post about the situation on Saturday.  Although we have a front pushing through tonight, it is a very slow mover.  As a result, this is not a frontal passage in which we quickly get into deep, unstable, northwesterly flow on Saturday morning. 

This is evident in the NAM time-height section for Alta below.  The front at Alta is a late arriver (light blue line), in this case moving through at or just after 0600 UTC (11 PM MST tonight).  Then, look at the winds behind the front on Saturday (circled).  They are NNW at low levels, but NNE near 700 mb (10,000 ft) and then SSW at 600 mb. 

This reflects the slow movement of the front through the area. 

If we look at the sounding for 1800 UTC (11 AM MST) Saturday morning, we see the low level northerly flow, but note how the winds shift to NNE and then SSW with height.  The temperature and dewpoint traces show a sharp inversion just above 700-mb, or 10,000 feet.  
This is not a recipe for our classic northwesterly instability snow showers over Alta for two reasons.  First, the flow direction isn't right.  Second, the instability is too shallow.  

However, if you look at the sounding for 0000 UTC (5 PM MST) tomorrow afternoon, the low level flow is NNW through a deeper layer, although a capping inversion remains based just below 600 mb.  This is closer to what is needed for the NW instability showers, but the capping inversion height is right on the edge of what I would like to see.  Tough to say if it's high enough that Alta can benefit, or just a bit too low so that the mid and lower canyons and east bench do better.  
And that's just one model run.  There are variations in the timing of these changes, wind directions with height, etc., if one looks at other models.  

All of this illustrates what a complex mess this is for Saturday and why probabilistic forecasting is necessary.  The good news is there's enough going on that in the end, this will be a decent storm for the mountains and even the mountain valleys after snow levels lower today and this evening.  


  1. As someone who works pretty regularly with stats in my job, those NWS forecasts are way better...thanks for the link.

  2. "Readers of this blog are snow lovers."

    Yes. Yes I am one.

    The low end snow totals always make me frustrated, because there have been days where Provo is on the news for hitting 12 inches, and I'm in Ogden raking leaves. Or vice versa, we had a couple of of 15 inch storms last year, and Provo got very little snow those days.

    My rule of thumb, if the low end prediction is around 50% less than the high end, it's going to be wrong. A city along the Wasatch Front always ends up with just a trace even if the reported minimum is 2 inches.

    1. Many storms feature large contrasts in snowfall along the Wasatch Front. This is more than just the mountains getting more and the valleys getting less. There can be dramatic variations within the Salt Lake Valley, Utah Valley, etc. It is really not possible to provide a single range that completely covers this variability, or you would hear forecasts of "0-12 inches" all the time. There are storm periods where that is the range of snowfall in the Salt Lake Valley, for example.

      This brings up an issue that I did not have time to discuss in this post and that is if we can produce very detailed and reliable probabilities, how does one then provide useful information to the public and other users? This is an issue not only for the NWS, but anyone involved in weather communication through broadcast and online media.

    2. I know it's hard. I'm on the weatherman's side on this. :)

      I just get overly excited when somebody promises me 2-3 inches of snow minimum, and then I get angry when only a trace falls. In a way it reminds me of the November 2010 years back. Everyone heard "First blizzard in Utah in X years" and got excited. Then the storm blew in, dumped just a few inches, and was out. People didn't realize blizzard meant a windy storm, not a high precipitation snow storm.

      I don't have an answer how to communicate the range. I just don't want to get let down with lack of snow.

  3. Looking at tonight's forecast, I was just telling my wife that I had no idea what the range meant -- in rigorous terms. Nice to see some probabilities. A CDF for the desired location would be nice to have so that I could quickly visualize the cumulative probability.

  4. When was the last time we had a heavy snow warning in the Wasatch?

    1. This came up in a post earlier this week. The National Weather Service no longer issues Heavy Snow Warnings. They simply issue winter storm warnings.