Throughout the winter season, time and time again, someone will publish a QPF x 10 model map (otherwise known as “snowfall”) and the meteorological community (TV, NWS, academic, and private sector mets, students, and weather enthusiasts alike) will throw their hands up: It’s too early! It hasn’t even reached land yet! Irresponsible!
But in the warm season, when NWS meteorologists have confidence in severe thunderstorms or flash flooding, outlooks are issued by national centers (SPC/WPC). SPC has perhaps the most well-known of the “outlook” stage products, the Convective Outlook. These outlooks are issued as far out as Day 8 when SPC have confidence widespread severe weather. They’re so popular the categories were recently revamped to add more detail in the short-range, acknowledging the wide-spanning use of the product. (For Day 4-8, now “15%” and “30%” risk areas are possible, previously only 30% was used; For Day 1-3 the categories expanded and names were adjusted “to bring better consistency to the risks communicated in SPC outlooks.”) For Flash Flooding, WPC issues a similar (but much less popular in my experience) Excessive Rainfall Outlook for Day 1-3.
Winter Weather Left out in the Cold?Winter Storm fatalities are vastly but knowingly under-counted by NWS (click to expand for details)
You might notice the “winter” category is very low on the 10-year average (light blue bars). More people have been killed by lightning compared to winter storms? No way! The reason is capturing these statistics – work done on a case-by-case basis at local National Weather Service offices – is extremely time-consuming and challenging. When it comes to winter weather, the NWS Storm Data directive (NWS 10-1605) declares winter weather deaths from vehicle accident are almost always “indirect” fatalities: “Fatalities and injuries due to motor vehicle accidents on slippery, rain, or ice-covered roads are indirect.” Indirect fatalities are not included in weather-related fatality statistics, thus the smaller number in the graph above.
While fatalities from car accidents are not captured in NOAA/NWS statistics, the U.S. Department of Transportation does capture this data in the National Highway Transportation Safety Administration’s Fatality Analysis Report System (FARS). Vox’s Sarah Frostenson recently reported on a pair of studies that reviewed the data and counted 31,098 car crash fatalities attributed to winter weather conditions in the past 36 years, or an average of 863/year nationwide. If these were included in Storm Data, Winter Weather would exceed every other category combined. While this comes at no surprise to anyone in the industry, it quantifies the fatal nature of winter storms.
While having a high impact on life (not to mention economic impacts), this phenomenon is without any formal product in the “outlook” phase beyond local (and inconsistent) mentions in local Hazardous Weather Outlooks, a great but often hidden product. (When no threat exists for Day 1 (today and tonight) this product is not highlighted on the local WWA map nor on the point-and-click forecast pages.)
WPC Launches Experimental Day 4-7 Probabilistic Winter Weather Outlook
For the 2015-2016 winter season, WPC launched an experimental probabilistic Winter Weather Outlook (more details), a single map for each of day 4-7 containing a range of binned probabilities (10-30%, 30-50%, 50-70%, 70-90%, and 90%+) for “probability of snow/sleet accumulating > 0.25″ liquid equivalent.” Here’s what the maps looked like for the 2016 east coast blizzard, what was generally a well-predicted event at the long range in my observations: (Note color changes at the end – they must’ve changed it at some point.)
And the verifying snowfall water equivalent (SWE) for the identical 24-hour period. (This isn’t exactly the same as actual reports, but close enough for this purpose.) Around the third-darkest shade of blue and above would verify the WPC maps above.
Not a bad job at all! Keep in mind this event was longer than 24 hours, so while it seems like a “miss” for areas of eastern PA and NJ, the snowfall fell on the other side of 12z. I think this product is a great first attempt at filling the need for a winter weather outlook for the “medium range,” but it’s not the only experiment going on.
WFO Sterling’s experimental Long Range Winter Storm Threat product
WFO Sterling (AKA Baltimore/Washington, DC, or simply “LWX”) is doing a very similar experiment (more details/live product) using threat categories and colors (green, yellow, orange, red, purple representing none, slight, enhanced, moderate, and high, the same as SPC categorical outlooks) and corresponding 3×3 “threat matrix” to convey confidence and potential impact for a “winter storm” at Day 4-7 range. This is displayed on a map for two areas of their County Warning Area (CWA) – west of the Blue Ridge & Catoctin Mountains and east of the Blue Ridge & Catoctin Mountains, including northern Virginia (NoVA), DC, and Baltimore. The documentation does not specifically define a “winter storm,” but given the Winter Storm Warning criteria is 4″ for most of their region outside of the higher terrain, it could be inferred as a confidence marker of at least 4″ of snowfall. It’s important to note this product focuses on potential impact, not snowfall amounts. For the same storm as above, here’s a collection of examples of this product taken from Sterling’s Facebook page.
As the storm approaches, the confidence level increased, reaching the highest level possible in both confidence and potential impact at Day 4, providing four days advanced notice for high confidence in a high-impact winter storm. It’s important to note that even before this, the 3×3 threat matrix highlighted the high (3 of 3) potential impact versus medium confidence (2 of 3).
Comparing These Attempts at a Winter Storm Outlook
While these two experiments are not completely separate (the WFO Sterling experiments discusses using the WPC experimental outlooks in production of their winter storm threat product), they both provide different information in different formats; the WPC product seems to be aimed at more technical users while WFO Sterling’s product is tuned to the needs of their local partners such as Emergency Managers. Do both have a reason to exist side by side? Comparing the two ideas:
|WPC's Winter Weather Outlook||WFO Sterling's Winter Storm Threat|
|Output||Probabilistic guidance for CONUS (0-100% in bins)||Categorical (5 categories)|
|Criteria||> 0.25" Liquid Equivalent of Snow/Sleet||Winter Storm Impact (subjective)|
|Intended audience||NWS forecasters, Emergency Managers, and other interested parties (advanced users)||Emergency Managers, School Officials, general public|
|Advantages||+ Easy to produce (mainly automated with human adjustment)|
+ Gives precise probability range
|+ Output categories easier to explain/convey
+ Simple one-word forecast summary for each day
+ Ability to depict both confidence and potential impact
+ Subjectivity gives forecaster some leeway in communicating threat
+ Focuses on impacts relative to geographic area (Eg. Impact difference of 2" of snow in Minnesota compared to Georgia)
|Disadvantages||- Criteria may not be well understood (0.25" liquid could be <1" to >6" of snowfall)|
- May be difficult to understand (is a 30% chance high?)
- Only one probability instead of a few/many categories like the usual Day 1-3 probabilistic guidance
- Does not take into account impacts nor climatology
|- May be more difficult to produce (due to subjective nature, but based on suite of automated guidance)
- Large geographic area has bust potential for storms with a sharp gradient
Thought’s on WPC’s Winter Storm Outlook
The output focusing on liquid-equivalent instead of snowfall is confusing and requires the user to also understand the expected snow-to-liquid ratio – not very user friendly for forecasters wanting simple guidance let alone a non-meteorologist who won’t know how to find this information. I like the idea as guidance, but using liquid-equivalent amounts is not as helpful as a specific amount (or a few amount categories) would be, especially for non-technical users.
Thoughts on WFO Sterling’s Winter Storm Threat
The simple categories focusing on impact are excellent. The use of a threat matrix may be challenging to understand, but I like the idea. The production could be difficult, but with proper training and experience I really think this is the direction NWS should pursue. The output maps are not very clearly presented in my opinion (more of a web display/GIS issue), but otherwise I think this is a great experiment and I’m very happy they’ve had at least one major event to really get some use out of it!
Taking Aim at a National Winter Storm/Threat Outlook
Ultimately I think the experiment at Sterling could and should be expanded to a national scale with some small modifications. The SPC model in which a national office produces the product with some local input could be imitated by WPC. One of the main challenges would be understanding local impacts – forecasting snowfall is difficult enough, but understanding impacts on a nationwide scale is very challenging. An answer to this challenge could be developing a threat index based on local climatology, essentially trying to equalize winter weather impact everywhere from the inter-mountain west to the Gulf Coast. (NWS meteorologists are actually already working towards this idea, though it is currently in a per-experimental phase.) The production of such a national outlook could be done by WPC using a similar approach as its current suite of winter weather products. To develop confidence/potential impact ratings could be more subjective, but using ensemble guidance, three maps could be produced: potential impact (based on amount of snow/sleet/ice), confidence (based on ensemble and run-to-run spread), and a final outlook category based on the potential impact and confidence maps. The first two maps could be developed and collaborated on internally, identical to how WPC produces its winter weather and QPF guidance. WFO’s in turn could then use the final categorical outlook in their communication, using consistent terminology nationwide. The potential impact and confidence maps could either be shown separately or displayed in a visual point-forecast display such as on the point-and-click pages or elsewhere. Enhanced or higher outlook categories could even be highlighted on the point-and-click icons with details in the wording, similar to the new WWA display.
A friendly reminder that this post as well as all of the content on this website, on my Twitter (@wxjoe) and on my Facebook page do not reflect the views of my employer. These are my opinions only.