Published: 2014-02-17 19:34:00
Updated: 2014-02-27 12:24:17
Posted February 17, 2014
Updated February 27, 2014
It is not often that we get two major, disruptive winter storms in the span of a couple of weeks, but here we are. Now that the snow and ice are all but gone, it is time to review some of the key elements of our forecast and compare them to how things played out.
We have always believed you deserve an honest, transparent assessment of the forecasts you depend on every day, and that is what we hope to provide. Evaluating our forecasts for high-impact events like this is key to ensuring we continually improve them over time.
Like the previous storm, this storm was not a surprise. We were mentioning the possibility as early as Friday of the week before, and our confidence in a forecast for wintry weather was high by Sunday evening:
As with all winter storms, there is plenty of uncertainty about how things will play out. Mike Moss jumped into some of the details about how we make winter weather forecasts in blog posts on Monday and Tuesday. Greg Fishel and I held a one-hour video Q&A on Tuesday evening, as well, to answer your questions and get more into the technical side of the forecast.
For some, this was a two-part storm system. Round One struck on Tuesday, and it played out almost exactly as expected. While the Triangle saw a few flurries and “novelty flakes” (a term I borrowed from a former intern, John Boyer, who is now a meteorologist at Fox 23 in Tulsa, Okla.), the main thrust was reserved for the Sandhills and southern coastal plain. Southern Wake County saw a dusting to an inch, but totals ramped up quickly south from there.
During that evening’s newscasts and the video chat, Greg Fishel described his reaction to the outcome as “tickled pink.” There were some higher totals than we forecast for areas between the WRAL viewing area and the coast, including along U.S. Highway 70; however, within our area, the snow fell almost exactly as forecast.
Wednesday’s storm was more widespread and overall, more impactful. The big forecast challenges were determining when it would arrive, when the precipitation would change from snow to a wintry mix, and how much we would see when it was finished. We began talking about the likelihood for a major winter storm on Sunday, and by Tuesday, the details were coming into focus. On Tuesday, I wrote a detailed blog post about what to expect on Wednesday and how to prepare. On TV, we used our expandable seven-day forecast to show you that we expected big problems on the roads as well as power outages:
We also took to Twitter and Facebook to alert you that this would be a big deal:
We always expected the storm to spread from south to north across the area, beginning Wednesday morning, and bringing precipitation to all of central North Carolina by the afternoon, if not before. Elizabeth Gardner was talking about this on TV and Facebook as early as Monday morning:
We continued to refine the timing as new model data came in:
While our FutureCast model suggested a slower timing, we were clear to frame that within the context that there was uncertainty about the exact timing, and that uncertainty should be factored in when making decisions:
If there was a shortcoming in our presentation of the forecast, it might have been that we were not more aggressive and consistent in emphasizing the potential for the precipitation to begin even sooner than noon, which was within the realm of possibilities. This might have prompted more folks to opt to stay home for the day or plan to leave work sooner.
That said, on Tuesday evening’s 11 p.m. newscast, Greg Fishel quite clearly showed that the precipitation would arrive in the Triangle around mid-day. Moreover – and more importantly – he emphasized that, unlike other snow events that start with flurries and gradually ramp up, the snow would come down steadily and heavily from the beginning. During our morning news on Wednesday, Elizabeth Gardner reiterated that point – and backed it up showing the snow on our Fayetteville Skycam. However, even knowing that the snow would come down quickly from the start, Greg said he was surprised at how quickly it began to stick to the roads.
Changeover and Amounts
As expected, the precipitation began as snow but changed over into a wintry mix during the afternoon and evening on Wednesday. We transitioned from our noon newscast into continuing coverage on the air, online and via social media to track the changeover. We also used our live cameras, Drive5, the DUALDoppler5000 and related technology to show you exactly where the changeover was happening and where it would move next.
Snow and Ice
The precipitation began and ended about on schedule, so the exact amounts would be determined by the eventual track and strength of the system. Those factors would determine how long the snow stuck around before it changed over to ice or even rain. Generally, a stronger or more westward-tracking storm than forecast would be able to tap into more warm air, causing an earlier changeover and resulting in less snow and more ice. A more east-leaning track or a weaker storm would have less warm air, a later changeover and more snow. Another factor was the temperature profile through the atmosphere, which would allow heavier precipitation to fall more as snow. Greg discussed this using an atmospheric sounding on WRAL’s 11 p.m. newscast Tuesday, showing how a shift of less than a degree for that layer could result in more or less snow at a given spot.
The combination of the isothermal structure of the atmosphere, the track of the low and the system’s overall strength gave us more snow and less ice across the board than forecast, and the band of highest ice totals set up about a county farther inland as well.
The pattern general location of heaver vs. lighter snow in our forecast was good, but the amounts at any one location were generally underdone by 1-2 inches, or about 0.1-0.2 inches of liquid water in any one spot. For example, we forecast totals in Wake County to range from 1 inch in the east to perhaps 4 inches in the far northwest; the county averaged 4-5 inches with a few pockets of 6 inches. Near Fayetteville, this under-forecast was more significant: We expected the snow to remain only briefly before changing over to ice. The snow lingered and added up to 4-5 inches.
The flip side, of course, is that more snow resulted in less ice. Again, the pattern and orientation of highest totals was good, but location was off a bit, and amounts were overdone. This was providential in the end because trees and power lines tend to do better with snow than the higher totals of ice that would have happened.
With the caveat about onset and road conditions noted above, our overarching message was that travel conditions would deteriorate rapidly, from south to north, during the day on Wednesday and that travel would remain dangerous, if possible at all, through Friday morning. This was spot on, and while, in retrospect, we could have been more emphatic about these details, the message was clear: Stay home Wednesday, if you can, and focus any necessary activities as early in the day as you could.
Power outages sneaking up a touch. Now 8k+ in Wake county. More than 140k across DEP's coverage area in NC/SC. pic.twitter.com/eGpGpyfNAq— Nate Johnson (@nsj) February 13, 2014
Regarding power outages, we expected that areas mainly south of the Triangle would experience “scattered” to “widespread” outages, and that farther north, outages would be more isolated. For the storm, there were on the order of 100,000 customers in North Carolina without power at some point (and each “customer” means a business or home, most of which have more than one person). Most of those power outages lasted 12 hours or less, as conditions were still such that crews could get out in the weather and address them. Because the area has a couple million people, I would consider what we saw to be on low end of “scattered.” The lower number of outages was a result of there being more snow and less ice overall.
Fortunately, the final act for this storm – some wraparound snow during the day on Thursday – was underwhelming. We expected this to be a late-morning to mid-afternoon shot of mostly snow, affecting northern areas more than southern areas. The timing was again spot on; however, the amounts we forecast were overdone. This kind of precipitation generally sets up north and west of the track of the upper-level low, and because the upper-level low tracked farther west of here than we expected, so too did the highest totals.
It is rare for our part of North Carolina to be struck with two significant winter storms in the span of less than three weeks. Fortunately, both storms – and more importantly, their impacts on our day-to-day lives – were well forecast. Greg and I discussed last week’s storm and felt we deserved an A- or B+ for the week, with the highest grade coming Tuesday, Thursday’s underperforming snow weighing down the class average, and Wednesday’s forecast in the middle. We have already noted a couple of things we will work on for the next winter storm, whether it is next week or next year. Your suggestions are welcome, too!
POST-SCRIPT: WHAT DID NOT HAPPEN
Let’s talk about computer models for a moment.
Model guidance is an important tool for meteorologists, however, it is only one tool of many! To make a forecast, meteorologists use a combination of weather observations, forecasting techniques, knowledge of the atmosphere, awareness of the limitations of the computer models, forecaster experience, and, yes, computer model guidance. Likewise, your doctor will also conduct a physical exam; take a medical history; evaluate results from various tests and other scans; confer with other specialists; and refer to their own medical knowledge, experience, and past cases before making a diagnosis and recommending treatment.
Much like you wouldn’t want your doctor to diagnose you with a serious condition on the basis of a single tool – and if they did, you would demand a second opinion – you don’t want your meteorologist to make a forecast based on a single model map.
And yet, leading up to this winter storm, people shared a number of snowfall maps from various computer models suggesting we would have a foot of snow or even more. (Some people even posted these maps on the WRAL Weather Facebook page!) Greg took a moment during Tuesday’s video Q&A to address this specifically. Getting hung up on one small part of one model run (and a part with known and well-documented limitations, at that) is akin to making a serious diagnosis based on a single X-ray. (A better, but still not fool-proof, approach is to use computer model ensembles as a forecasting aid. For this storm, Greg Carbin from the National Weather Service showed one particular ensemble that did well for us in Raleigh but was less accurate elsewhere for this same storm.)
As a general rule, we do show model graphics on TV, online and via social media from time to time, but only when they help us explain a forecast (or some aspect or sources of uncertainty you need to be aware of). Even when we do not show those model maps, rest assured we see all the same data that are being posted everywhere. We incorporate ALL of this information into the forecasts we share with you on WRAL and elsewhere.
As always, we will focus our energy and time on taking all of the tools and information at our disposal and producing the most accurate forecast we can, day in and day out.