Rapid City Snow 2/24 -2/25/2020

Rapid City saw a substantial snow event that spanned from February 24 through February 25.

The National Weather Service in Rapid City was calling from 3-6 inches of snow in town at 2:11 am on 2/24/2020. They noted in their 2:32 am forecast discussion that the potential for upslope snow in Rapid City was increasing and that the moisture was higher than previously forecast, both of which could bring higher snowfall totals. Their forecast discussion at 1:49 pm from 2/24/2020 mentioned that the snowfall forecast for Rapid City "will be tricky, as it will depend greatly on where upslope enhancement occurs." They went on to say, "the Froude number, along with lack of inversion in the soundings, suggest that the upslope enhancement may produce more snowfall over the central Black Hills, rather than the Rapid City area." At 3:22 pm, the NWS upped their snowfall forecast for Rapid City to 5-7 inches but it was obvious that plenty of uncertainty remained in the forecast.

Various weather models were showing snowfall totals as given below (1200 UTC model output from 2/24/2020 with 36 hour forecast snowfall totals given for both 10:1 snow and Kuchera snow ratios; ranges due to snow gradient across Rapid City).

Model: 10:1 ratio, Kuchera ratio
GFS: 3-5 inches, 5-9 inches
HRRR: 7-9 inches, 8-10 inches
NAM: 5-7 inches, 7-10 inches
ECMWF: 5-7 inches, Kuchera N/A
NAM 3km: 7-10 inches, 8-14 inches

The 18 UTC model output for a 30-hr forecast (to compare with the previous 36 hour forecast):

Model: 10:1 ratio, Kuchera ratio
GFS: 4-7 inches, 6-9 inches
HRRR: 4-7 inches, 6-9 inches
NAM: 5-7 inches, 7-11 inches
ECMWF: N/A
NAM 3km: 5-8 inches, 6-10 inches

All considered, there was fairly good consistency among the weather models for a forecast snow total of more than 4 inches but generally less that 9 inches when considering a 10:1 snowfall ratio. The Kuchera method takes a different approach using information on the thermodynamic profile to get a snowfall ratio that can change with time. As snowfall during colder temperatures can bring snowfall ratios on the order of 20:1 or 30:1, using the Kuchera method can give some idea on how much snow to expect if a non-10:1 ratio is possible. This Kuchera method for this event gave snowfall totals generally from 6 to 10 inches., with ratios ranging from 10:1 to 16:1.

My forecast for snowfall in Rapid City ranged from 4-9 inches. I was assuming that there would be a fairly tight gradient of snowfall across the city due to localized upslope enhancement. I was figuring northern/northwestern Rapid City was going to see 6-9 inches while southern Rapid City down through Hermosa was in the 4-6 inch range. There were caveats to the forecast, as the NWS had mentioned, but more than 10 inches of snow from an Alberta Clipper-type event seemed unlikely.

What happened was that we got ~50% more snow than expected. The map below is as of 11:30 am on 2/25/2020. Note that some stations reported their snowfall totals at earlier times and have not updated as of this writing. Find current snowfall information here: https://www.weather.gov/source/crh/snowmap.html?sid=unr

Why?

Obviously this is a tough question to answer. But I've got a few ideas:

1. Convective snow showers were present over the region. This likely led to higher snowfall totals in those areas beyond what was forecast in the models due to differing dynamic processes at play. 
2. Additional mid-level moisture. Water vapor imagery continued to show high levels of mid-level moisture advecting into the region through the snow event.  
3. Localized mesoscale banding. There were several bands of precipitation that set up during this snowfall event. Although the models were consistently showing some banding, the precise location where they occurred may not have been properly forecasted. The HRRR model was showing one of these bands east of town through several model runs. Radar imagery from overnight shows a band of heavy snow directly over Rapid City for 5-6 hours in the early AM. Miles matter but the weather models are not capable of perfectly resolving where these bands will set up.  
4. A system slow down. The main forcing for the snowfall appeared to stay over Rapid City longer than what had been forecast by the models. Snowfall totals are determined by: the rate of snowfall x the duration of the event.  

Our weather models are a diverse and valuable set of tools but we must remember that they are only numerical representations of the atmosphere. They provide meteorologists with exceptional guidance, but they are, by nature, imperfect.

Next time I see a similar scenario like this where Rapid City will see a period of northeasterly upslope winds and the potential for convective snowfall, I will likely lean towards the higher-end of the model guidance for snowfall accumulation.