Posted: 6:50 pm Sunday, January 8th, 2017
By Kirk Mellish
Long-time listeners to WSB Radio 955 FM and AM 750 and/or long-time readers of my blogs the past 29 years (less for the web obviously) already know much of this as I explain it every winter.
Like predicting the future of the stock market or economy, or an election, or who will win the World Series or who will be in the Super Bowl and win it, forecasting is hard. What happened yesterday or what’s happening now is easy.
Much like when we show you hurricane track forecasts there is a RANGE of possible outcomes where the eye of the storm will hit or a “cone of uncertainty”.
Cone of uncertainty is a concept used in business and finance, construction, engineering, software development and other fields. You can google it.
The HUGE difference for weather is that unlike in other fields where the uncertainty will be reduced over time because they HAVE CONTROL over the VARIABLES, forecasters have NO control over the weather’s variability and chaos and no ability to reduce it. That’s why models are merely a numerical simulation of the future state of the atmosphere. There’s always a battle between these two guys in winter:
Sometimes we get this jerk:
And because of “The butterfly effect” (Lorenz Chaos Theory) you can google it, there will be times when a forecast can be dead wrong despite MODEL AGREEMENT:
In the case of a winter storm think of the cone of uncertainty as the range of snow amounts AS WELL AS the location of those amounts.
Say some location was in the bulls-eye of a 6-12 inch forecast. EVEN IF the forecast was ‘correct’, there would be places that got only 2 inches and some that got 16.
But what if you’re in a zone with a forecast of only a half an inch to 3 inches? Then the normal and expected margin of error means someone might get 5 or 6 and other locations zero.
And more rarely a forecast of 1-3 inches of snow may end up partly sunny (like the Joaquin example above). Or a forecast of partly sunny may end up with an inch of snow or ice.
Where I was right or where I was wrong there’s no hiding it. In most professions a mistake or miscalculation on the job is known only to you and the boss or the boss and those on the project. But your mistakes don’t land on the front page lol. Must be nice 😉 😉
Mother nature will always have the upper hand, as the old expression goes “God only knows”. In predicting the future you will always be off the mark to some degree and way off other times.
I am not even going to go into the ramifications of how very little of the real atmosphere we actually sample and that we do it with weather balloons (radiosonde) only twice a day early AM and early PM. That’s for another day.
Note that this is not meant to be an excuse just an explanation. Trust me I am just as frustrated as some of you that the forecast didn’t pan out better! Maybe more since I saw model projections of 10 inches at times and had a string of 13 hour days that were fun in a way but also stressful and fatiguing. Like anyone else I hate being wrong. I don’t even like being just partially correct. It makes me sad and angry and I don’t need the public’s thoughts to make me feel that way I assure you. It comes with the territory:
So what about THIS STORM SYSTEM?
It was what’s known as a “Miller A” system. Which is the best bet for Atlanta to have a decent shot at snow, you can google it.
The problem is Atlanta was always “right on the line” between cold enough aloft for snow/sleet or just barely warm enough for mostly rain. (this was shown in my forecast blog post for days-still online)
This is a screen shot from my blog Thursday/Friday:
In my forecast blog posts I showed the probability charts, or the “odds” of or “chance of” an inch of snow. If you go back and look at them you’ll see at no point was it ever 100% and in fact for areas near and south of I-20 it was just 50-50. Why people ignore this I’ll never know. Screenshot from my forecast blog:
In the end a couple of significant things happened that the numerical equations did not model correctly:
A) The low pressure storm tracked farther North than projected. A track around 100 miles or so south of the Gulf Coast is desirable for Atlanta snow. (with a mid-latitude cyclone the heaviest snow is found 62-186 miles to the left of the low track)
B) The more northerly track allowed a thin layer of warm air aloft known as the “warm nose” to be warmer longer than expected and more static rather than moving south sooner making rain more widespread and lasting longer than projected. Also allowing for more freezing rain I-20 south.
THERMODYNAMIC DIAGRAM UPPER-AIR SOUNDING LATE FRIDAY AT PEACHTREE CITY GA (see the warm nose straddling 850mb level):
C) The mountains of AL/GA again proved to be a barrier to allowing colder air to penetrate farther south soon enough to meet moisture and the limits of the models handling of terrain was as usual a problem.
D) The systems moisture ended up moving MUCH faster than modeling suggested so by the time it WAS cold enough aloft in more places the moisture was mostly gone. (the models showed still snowing at 8am Saturday, instead most of it was gone Friday Midnight-2am).
Snowflakes were reported half way to Macon and there was a trace of snow as far south as Fayette County reported by weather observers, but it didn’t last long.
The following is from North Carolina but applies to Atlanta and Birmingham as well:
The problem is when you’re on the line being wrong by just 1% in temp aloft can mean 100% off at the surface. I don’t think anyone wants a margin of error in their life or work that small but its what meteorologists face:
If it was easy everyone could do it and everyone would always be right.
Another screen shot from my blog Friday pm:
As you can see, there’s a lot of complexity beyond IS IT 32F OR COLDER? Here’s just ONE example of a computer snow output formula:
IF memory serves the warmest temp it has snowed at was in the 50s. Below is a Skew-T or thermodynamic diagram used by meteorologists to diagnose the radiosonde balloon results:
As far as model performance goes on this storm, none of them were very good. But toward the end a combo of the Canadian regional and global plus the ECMWF models were closest, and this time in the end the NAM thermal profiles proved best.
I didn’t get enough snow, but if I have work and health, family and friends then I have all I need.
Thanks to all of you for reading and listening to my reports. I appreciate it.