Thursday, November 20, 2014

Snow Is Coming, But How Much?

I've been distracted by Buffalo's snowmaggedon the past couple of days.  Incredibly, they are getting more again today.  Never has so much powder gone to waste.  If only Mother Nature could have arranged for the Adirondacks, especially the high peaks, to be located more strategically relative to Lakes Erie and Ontario.

Of course, there is a place where Mother Nature has fortuitously aligned the weather, topography, and geography for great skiing and that's the Cottonwood Canyons.  And, the good news is that we are going to be getting some beginning this weekend.  The storm will be warm on Saturday with a passing warm front.  Following the cold frontal passage late Saturday, temperatures will fall and we are looking at what will likely be an active northwesterly flow period through Monday Night and possibly Tuesday.

The image below shows the 1200 UTC GFS forecast accumulated water-equivalent precipitation through 0000 UTC 25 Nov (6 PM Monday MST).  One can see the moisture corridor extending across the Pacific Northwest with precipitation maxima over the Wasatch–Teton ranges and the western Colorado Rockies.

Source: NCEP
If we extract a time series of accumulated precipitation for Salt Lake City from the GFS we see some healthy numbers, with a total of about 1.2 inches by late Monday.

There are, however, at least three critical factors to consider when interpreting such a forecast:

  1. The GFS does not adequately resolve the topography of the western U.S.  As a result, we typically have to adjust these precipitation amounts for topographic effects.  This is typically done in ad hoc fashion.  For example, the standard approach for estimating the total precipitation in upper Little Cottonwood is to multiply the GFS forecast for Salt Lake City by 2 or 2.5.  Using the latter would be a phat 3.6 inches of water, which would be marvelous for base building.
  2. For snowfall amount, we need to also forecast the snow-to-liquid ratio, or how much snow we'll get out of each inch of water.  There is no model that does this today, although we do have some algorithms that we can apply to model output to get a pretty good estimate for the upper Cottonwoods.
  3. Can I trust this forecast?  The atmosphere is quite chaotic.  A slight shift in the position of the jet stream and moisture plume in a situation like this could leave us either high and dry (north shift) or right in the thick of it (south shift).
Through the National Weather Service Collaborative Science, Technology and Applied Research (CSTAR) program, I've been working with Trevor Alcott to develop approaches for dealing with these issues.  More correctly, Trevor has been coming up with great ideas and implementing them, while I sit back, pat him on the back, and otherwise get the hell out of the way.  

The approach uses forecasts from the North American Ensemble Forecast System (NAEFS), which includes the Canadian ensemble system and the U.S. Global Ensemble Forecast System (GEFS).  These low-resolution forecasts are then downscaled using climatological precipitation –altitude relationships to generate high-resolution precipitation forecasts.  Then we apply a snow-to-liquid algorithm to estimate snowfall.  We do this to all members of the ensemble.  This allows us to:
  1. Calculate a mean of all the members, which over the long run is typically a better forecast than that produced by a single modeling system such as the GFS
  2. Evaluate the range of possibilities for a given period, including probabilities of exceeding specific thresholds.
Now, I'm going to show you these results, which are literally a day old.  I need to emphasize that this is an extremely experimental system, so don't plan your powder days or backcountry travel based on it.  The system has not been tested or calibrated, so right now, this is just great eye candy, but hopefully over time we can assess reliability and make refinements. I'm going to hold off on the snowfall estimates as I'm not sure we're even in the ballpark on those yet.  

The ensemble mean water equivalent forecast below is from last nights NAEFS and covers the period through 0000 UTC 25 November (5 PM Monday).  It shows a strip > 3 inches along most of the high Wasatch with a maximum of just over 5" on Ben Lomond Peak.  Structurally, this looks reasonable, but will it be skillful?  Time will tell.  I am skeptical of the Ben Lomond max.  Since we are using climatological precipitation-altitude relationships, they don't account for flow directions.  Typically Ben Lomond doesn't do well in northwesterly flow, so perhaps it will fall short.  Elsewhere perhaps we have a shot, but the storm will need to be productive.  If this does verify, we'll be in good shape for Thanksgiving skiing.

With an ensemble, however, we can take a look at many many forecasts to get a handle on the range of possibilities, as well as probabilities. Below is what we call a plume diagram and it shows the total accumulated precipitation at Alta-Collins from all the ensemble members (Canadian model light green, GEFS dark green) and the mean (thick green line). 

You can see that the mean gives almost 4 inches of water for Alta-Collins.  What a godsend that would be.  The range, however, is quite large.  The driest members produce around 1.5 inches of precipitation, and the wettest over 7 inches.  That's a lot of spread.  The Canadian model is in general wetter than the GEFS.  This appears to be due to the Canadian simply being more productive than the GFS since their large-scale forecasts are similar.  Canadians do like snow, eh? Of course, there is a member of the GEFS that goes for five inches.  

 Now, if this were a reliable and well calibrated system, one might draw an analogy of that plume to rolling two dice.  The most likely outcome is a 7 when you roll two dice.  In the case of the forecast above, the most likely outcome is probably something near the mean.  The extreme low and high precipitation amounts are more like rolling snake eyes or two sixes.  Less likely, but still possible outcomes.  Surely you see why it can be foolhardy to put specific numbers on winter storms many days in advance.  

Of course, we don't know yet how reliable and calibrated this system is, which is why you should continue to listen to official forecasts and not mad scientists like me.  I thought, however, that it might be interesting to share and perhaps we'll get some of this running on in the near future.  Putting all this together, it looks like we're going to see a good storm cycle through Monday, but we'll have to see how it all comes together in the coming days.  


  1. That's exciting and interesting work! ... and a very exciting forecast. Maybe I will be able to put my rock skis away sooner this season. :)

  2. I just found this blog this fall - fascinating, I love it. Posts like this are gold. Just dropping in to say thanks. Fingers crossed for some double-sixes on the snow dice this weekend!

  3. I just found this site, it looks excellent. Love the mix of science interpretation and skiier snow-jonesing :) Would be great to see a follow-up to this post after the storm passes to compare predicted to actual, what happened and why. Glad to have another source of insight and snow information while I sit in my office and wait for when I can get on the slopes!

    1. ditto. i'd like to see how over time the data speaks to the reliability of these forecasts. i'm always looking to get an edge on pow turns. so thanks for putting your work in the public forum.