Monday, December 4, 2023

Looking Back

Mother Nature gave us the storm we needed.  Water and snow totals per this morning's Utah Avalanche Center are 3.38-4.66"/40-50" in the upper Cottonwoods and 1.80-2.58"/20-30" on the Park City Ridgeline.  Big totals as well in the northern and southern Wasatch.  All northern and central Wasatch SNOTEL sites are at or above the 1991-2020 median for the day except Thaynes Canyon which sits at 86%.  Sorry Wasatch Backers.  

I thought we would take a look back at a few forecasts for the event, starting with the ensembles.  I'll just use the ones for Alta Collins that I included in prior blog posts and will use the range of water equivalent and snow reported for the upper Cottonwoods for verification.  The Alta-Collins numbers would be at the top end of this range.  

The 3.38-4.66" of water was in the upper 12-22% of water equivalent forecasts produced by the downscaled NAEFS from 00Z 29 November.  For snow, however, we were much closer to the mean. 

For the SREF, the 3.38-4.66" of water was in the upper 15-40% of members (and roughly at the mean for the ARW members...this is the wetter of the two forecast models used for the SREF).  Similar to the NAEFS, for snow, we were near or just below the mean. 

So, feel fortunate that for water, which is most critical for building base, we came in on the high end of model forecasts.  It won't always work out like this.  

For snow, it's clear that the simple scheme that we use for the snow-to-liquid ratio in the ensembles was off. It had the right idea for trend (see lower right panels below), but was consistently too high on Saturday night and Sunday (roughly from 00Z 3 Dec to 00Z 4 Dec).  A big reason for this is we are still using a somewhat ancient snow-to-liquid ratio algorithm that does not consider wind.  We haven't had the time for an upgrade unfortunately, which is not quite a simple as it sounds.  

My impression is that the HRRR forecasts were generally quite good for this storm.  The forecast from 12Z 1 Dec (5 AM MST Friday) called for the passage of three troughs (which happened, modulating precipitation rates) and a total of just over 2" of water and almost 30" of snow.  The Utah avalanche center report from Sunday morning had storm totals to that point of 2.43" of water and 36" of snow, so this is only slightly underdone.  
In this case, the snowfall differential is due to a slight under prediction of water equivalent rather than snow-to-liquid ratio.  The Little Cottonwood products use a newer snow-to-liquid ratio algorithm. I was generally happy with its performance through this storm, although those who were in the field can perhaps quibble some more.  

The HRRR initialized at 12Z 3 December (5 AM Sunday) also seemed to verify quite well.  It generated just over 1.2" of water and 10" of snow.  This is consistent with a mean snow-to-liquid ratio of 8.3:1.   The numbers from Alta-Collins show 1.25" of water and 9" of snow for this period, yielding a snow-to-liquid ratio of 7.2:1. 

In the weather forecasting business, you can't do much better than that.  

If one were to grumble about the forecasts for this period, it would perhaps be the lack of confidence or agreement in the ensembles when the storm was still a few days out.  I don't see that necessarily as a negative.  The reality is, as discussed in the post Active Pattern with a Lot of Possibilities, that it was a difficult pattern to nail down.  The European model, which is the best performing individual model in the world, probably produced the worst forecast of this event.  If you had bought into that hook-line-and-sinker, you would have had a major underforecast.  If anything, this event shows the value in using all the available models and ensembles to anticipate the full range of possibilities. 

1 comment:

  1. Thanks for all of the time and effort you put into providing us with nice forecast model output data to look at. The forecasts are getting better every year as are the charts, images, and graphs you provide.