A few days ago, some of the "deterministic" global forecast models, namely the GFS and the ECMWF HRES, forecast a truly spectacular cold surge for the continental United States this coming week. There were a few forecast runs that brought the cold surge directly into the mountain west. Below is an example from the ECMWF HRES initialized on 0000 UTC 13 December, which produced a 240 hour forecast [valid 0000 UTC 23 December (5 PM MST Thursday) with a spectacularly low 700-mb (10,000 ft) temperature of -34˚C (-29˚F) over southern Wyoming and -30˚C over Salt Lake City.
How insane is cold air like that for Salt Lake City? A 700-mb temperature of -30˚C has only been observed once in the upper-air sounding record for northern Utah, which began in 1950.
However, today we run ensembles of forecasts to get a handle on the uncertainty in the forecast, and other members of those ensembles had different ideas. Some didn't bring the cold surge into northern Utah.
It is relatively easy to view forecasts from the GFS or ECMWF HRES out to 10 or more days online. The GFS and ECMWF HRES are higher resolution than the members of their respective ensembles (known as the GEFS and EPS, respectively) and there may be some validity to weighting them a little more heavily in a short range forecast, but there is little validity to doing so in a medium range forecast. After a few days, the higher resolution GFS or ECMWF HRES aren't significantly more accurate than the lower resolution members of their respective ensembles.
This is because large scale forecast uncertainty trumps resolution at long lead times. The development of high and low pressure systems and waves in the jet stream is too uncertain to place great faith in a single high resolution forecast model. Going all in on the GFS or the ECMWF HRES at day 10 makes little sense.
Now let's forecast to today. The ECMWF is still forecasting a potent cold surge, but at 0000 UTC 23 December (5 PM Friday), which corresponds to the same time presented above, the coldest air is located farther east and centered over Nebraska and the high plains.
Northern Utah, however, is in the strong temperature gradient on the southern and western edge of the cold surge, and this is still a very very difficult forecast situation. If cold air shifts a bit eastward or westward, it will make a HUGE difference in our temperatures. HUGE.
The National Weather Service "National Blend of Models" or NBM is a forecast system that pulls in data from the various global ensembles and produces not one but many forecasts based on those members. This allows one to look at the range of possibilities, or what one might call a distribution. Below are box-and-whisker plots of the maximum and minimum temperature forecasts for Salt Lake City based on the NBM forecasts from 6 AM this morning. Prior to the 21st, the boxes are relatively squat. Then on the 21st they get very tall. This continues through the 23rd, when the get squat again.
In the box and whisker plots above, the middle half of the forecasts lie within the box. The fancy word for this is interquartile range. Whiskers then extend above and below the box and those cover the middle 80% of the forecasts. Essentially, this is a visualization of how large the range of forecasts is, and that range is very large from the 21st to the 23rd because we are on the edge of the cold surge.
If the cold surge remains to the east or just barely trickles in, the highs on Thursday could be around 40 or even higher since there are a few forecasts that are above even the top whisker. If the cold surge pushes west and we get deep in the cold air, highs could be a very nasty 15˚F or colder. Minimum temperatures also show a huge range during this period. Basically, almost anything could happen (within reasonable limits).
For maximum and minimum temperature, the NBM suggests that the period covering Thursday and Friday is the most difficult to predict of the next 10 days. Although we think of uncertainty as increasing with increasing lead time, for Salt Lake City, the uncertainty actually goes down again in this instance because of the extreme sensitivity of what will happen with the cold surge later this week.
We will see where all of this ends up. Right now there are a lot of possibilities that Mother Nature could roll with her weather dice.
I was pondering if one of your data points in the HRRR was Ben Lomond peak or Trail, but quickly realized that was likely lost in the model resolution. But it got me wondering if you have ever studied the North Fork microclimate with a super hires model imbedded in a lower res model, or if that is even possible. We do such things in mechanics using variable meshes, or by initializing high res models with boundary conditions from lower res models.
ReplyDeleteThat point is essentially the Ben Lomond Peak snotel site, but you are right that the model terrain does not match reality (although at 2.8 km grid spacing, the HRRR is not awful). I haven't done any work looking at North Fork.
Delete