Verification of a Blowing Snow Model and Applications for Blizzard Forecasting
Jeff Makowski, Thomas Grafenauer, Dave Kellenbenz, Greg Gust
National Weather Service – Grand Forks
Outline
Marginal vs. Real BlizzardsCanadian Blowing Snow Model
What is it?Is it useful?
Marginal vs. Real Blizzards Real blizzard = Widespread zero visibility for a long enough
duration
Usually with several inches of falling snow
Shuts down most if not all activities/commerce/transportation
Marginal blizzard = Areas of zero visibility for a long enough duration
Usually with very little falling snow
Rural vs urban areas - cities may not be affected
Challenges:
Easier when heavy snow is predicted, but:
Events with little to no falling snow, difficult to forecast the differences between marginal and real blizzards
Events with little to no falling snow, difficult to forecast the differences between marginal blizzards and winter weather advisories for blowing snow
Is it possible to forecast these differences?
Is there accurate guidance available that would assist in the forecast process, and could help collaboration?
Help from Environment Canada
Baggaley, D. G., and J. M. Hanesiak, 2005: An Empirical Blowing Snow Forecast Technique for the Canadian Arctic and Prairie Provinces. Wea. Forecasting, 20, 51-62.
What is the Canadian (Baggaley) Blowing Snow Model? Based on a robust set of observations from Canadian
Prairie stations
Simplifies the complexities related to forecasting blowing snow
Inputs: SnowRate, Temperature, WindSpeed, Snow Age
Outputs: Probability, Low End Wind Threshold (Patchy), High End Wind Threshold (Definite)
Probability = Probability that the visibility due to blowing snow will be 1/2sm or less
Needs a snow density model (How much snow is available to blow around?) – FUTURE WORK
Very Brief Literature Review
Created a series of charts that summarize the proportion of times where the combinations of wind speed, temperature, and snow age gave blowing snow visibility reductions of a given threshold.
This method will not always give a deterministic answer, but rather a statistical likelihood.
SnowAge 1-2 Hours
SnowAge 3-5 Hours
SnowAge 6-11 Hours
SnowAge 12-23 Hours
SnowAge 24-47 Hours
SnowAge 48+ Hours
Tips – From Dave Baggaley
We generally want to see some big numbers, for several hours.
Probabilities around 50% = “Blowing snow at times" or perhaps just limited to vulnerable areas.
Probabilities 80+% = Straight blowing snow forecast with the understanding that there will be variability through the period.
Probabilities 100% = Unbroken <1/4 mile visibilities.
Can this Model provide useful Guidance?
If yes…forecasting the differences between ‘real’ and marginal blizzards may be possible (or the difference between marginal blizzards and advisories).
Research Results (so far…)
Looked at each of the 10 verified 2013-2014 winter season blizzards within FGF CWA
For each blizzard: Selected the most severe hour
Determined the Blowing Snow Model output for selected sites (KDVL, KJMS, KGFK, KFAR, KHCO, KBJI, KPKD)
Observed data
Model data (NAM, GFS, ECMWF, MOSGuide, SREF)
Attempted to define a marginal blizzard Compare Blowing Snow Model results
Computed MOS wind speed biases at each forecast hour
Defining a Marginal Blizzard
The difference between a marginal blizzard and a ‘real’ blizzard depends on two factors: Coverage of low visibility
Duration of that low visibility
Downloaded ASOS/AWOS observations from each blizzard event
Defining a Marginal Event
Developed Python scripts to read the observations, and calculate coverage and duration values at different visibility thresholds (2sm, 1sm, 3/4sm, 1/2sm, 1/4sm)
Snow
Snow
SnowSmall Area
SnowSmall Area
Classify Blizzards by Coverage and Duration – Related to Impacts
Real blizzard with snow – March 31st
Real blizzard no snow – Jan. 26th
Real/marginal blizzard – Dec. 28th and Jan. 16th
Marginal blizzard – Jan. 22nd and Feb. 13th
Marginal/no blizzard – Jan. 3rd, Feb. 26th,
and March 5th
Not used March 21 (Very small area)
Some Preliminary ResultsBlowing Snow Model probabilities
(based on observed data)
Jan 26th and March 31st
Probability = 92%
Dec. 28th and Jan. 16th Probability = 69%
Jan 22nd and Feb 13th Probability = 55%
Jan 3rd, Feb 26th and March 5th Probability = 29%
Note: If 6-hr snowfall was less than 1 inch, used the NoSnow
probability
Observed Coverage vs. BLSN Model Probabilities
Model Biases
Blowing Snow Model - Model biases
Inputted model T, Wind, SnowAmt into the Blowing Snow Model, and then compared that value to the observed Blowing Snow Model value (with falling snow).
Used a recent model run
NAM12
GFS40
ECMWF
MOSGuide
SREF
MOS Guidance Biases – Winter 2013/14 Blizzard Events
Takeaways - Conclusion
Canadian Blowing Snow Model shows usefulness:
Coverage indicator of low visibilities.
Output could potentially provide better shift to shift, and office to office consistency:
<50% Probability = Lower impact marginal blizzard or advisory
50% to 70% Probability = Lower impact marginal blizzard
70% to 90% Probability = High impact marginal blizzard
>90% Probability = “Real” blizzard
All information could be used in some sort of a program to give a probability based on known biases (especially wind).
Potential for a MOS Guidance Bias Smarttool (used during winter cold air advection events)?
Need to look at more cases…