The Thursday to Saturday morning storm was huge! Within the zone of our backcountry patrol area, Eldora reported a storm total of 30”, while near our patrol zone, Loveland reported a storm total of 39”, Winter Park reported a storm total of 30”, and Ski Granby Ranch reported a storm total of 33”. It’s extraordinarily unusual for Granby Ranch to wring more precipitation out of a storm than Winter Park, which speaks to the lift (which causes snow) having been primarily fueled by a jet streak. (Berthoud Pass’s Snotel reported 2.5” of SWE.)
As I’ve repeatedly mentioned in prior posts, models struggle to predict snow totals from jet streaks. But the word “struggle” seems to woefully underestimate how far off the models were coming into this storm. Please be patient with me for the rest of this post, if you feel like reading further, as I dive into what the models predicted, how far off they were, and what lessons I can learn from this storm.
Let’s compare what our patrol zone got, roughly 30” of snow, to what the models were predicting. One week before the storm started, the European Model was predicting 10”, the Canadian Model was predicting 8”, and the American Model was predicting 4”. Two days before the storm started, The European Model was calling for 17”, the WRF Model was calling for 16”, the Canadian Model was calling for 9.5”, and the American Model was calling for 8”.
Even once the storm started, the models were calling woefully low numbers for the second half of the storm. From Friday 6 am to Saturday 6 am, Eldora reported 18”. When we compare that 18” from the second half or the storm to the model’s predictions of the second half of the storm (once the storm had already begun), we can see how far off they were. The models underpredicted this second half of the storm, forecasting: 8” per HRRR, 7” per the Canadian Model, 6” per the European Model, 5” per WRF Model and American Model, and 4” per the NAM Model.
In my posts I rarely talk about the ensemble forecasts. However, for a storm that overproduced this much, I thought taking a retrospective look at the ensemble models would be interesting. The most user friendly ensemble forecast numbers come from the University of Utah’s downscale of the NAEFS ensemble model (basically the American and Canadian ensemble models combined). You may already be familiar with this, as Joel Gratz’ wonderful Open Snow weather report often inserts pictures of these ensemble runs into their weather posts. I rarely talk about this ensemble forecast as the University of Utah doesn’t provide numbers for any location within our backcountry patrol zone.
However, as it’s interesting to compare the ensemble runs with what actually happened, let’s compare the ensemble runs for Loveland Ski Area with what happened at Loveland Ski Area (i.e., 39”). The ensemble run is basically running the American and Canadian Models 42 times with slightly changed parameters to try to get a sense of all possibilities, and the most likely possibilities. Of the 42 members, a mere 18 hours before the storm started, the lowest member (one of the Canadian ones) of the 42 member run was calling for a storm total of 2”. The highest member (one of the American ones) was calling for a storm total of 28”. Right as the storm started, the ensemble still hadn’t changed much, with the lowest member (again one of the Canadian ones) calling for a storm total of 5”, and the highest member (again one of the American ones) calling for a storm total of 28”. In other words, the actual storm total of 39” wasn’t even within the most optimistic run of the 42 runs.
So, what are the lessons from this storm that I can take away if any, beyond to emphasize the well-known unpredictability of jet streaks?
All models failed on this storm. And, although the models differed, it’s hard to say any of them failed more or less spectacularly. In our patrol zone, the global model that did the worst was the American Model and the regional model that did the worst was the NAM Model. However, at Loveland Ski Area, the Canadian Model did far worse than the American Model, so it’s hard say the American Model was the biggest failure overall. Back to our patrol zone, while the European Model was wildly off before the storm, it was less wildly off than any of the other models. However, once the storm was underway, the European Model was further off than the HRRR or the Canadian Models.
If any lesson on reading the various models can be taken from this storm, it is that there’s no magic bullet model. I tend to gravitate towards the WRF Model. Many forecasters gravitate towards the European Model. Regardless, of any forecasters own preferences when it comes to models, sometimes every single model (including every single ensemble run) can be wildly off.
On the other hand, at least every major global model was predicting this storm a week out, even if they all woefully underpredicted it both a week out and while it was happening.
Stepping back from the model focus, are there other lessons for me? Snow is produced in five ways – orographic lifting, low pressure, fronts, convection, and jet streaks. The vast majority of snow in Colorado is a result of orographic lifting. So, as I try to predict snow (which I’ve been doing for years for myself in hunting powder), I tend to focus most of my attention on orographic lifting. But this last storm, fueled almost entirely by a jet streak, is a good reminder that focusing on orographic lifting should not take my attention away from the other producers of snow . This is especially true of jet streaks.
If anyone has made it this far in the post, thanks for your patience, and feel free to reach out to me to share your thoughts. To quote the great Winston Churchill, “success is the result of making many mistakes, and learning from the experience.” I’m committed to learning from the experience when the forecasts are this wildly off, even if true success is too great a goal to be obtainable with current technology when discussing weather. Weather, after all, is the ultimate chaotic system.
-Jordan (Saturday morning)
Note: Unless otherwise noted, all forecasts are for 10,000’ in exposed areas. References to American Model are the American (GFS) Model. References to the Canadian Model are the Canadian (GDPS) Model. References to the WRF Model are the CAIC WRF Hi-Res Model. References to the European Model are the European (ECMWF) Model.