WSET.com - ABC13What goes on in there? by Sean Sublette

Ever been in a trivia game where you get a first guess in your head, then you talk yourself out of it, only to discover your first guess was right?

Living it tonight.

Most of the meteorological community has known since Friday (21st) that there was going to be some type of winter issue today (26th).  Including me.

Of course, any time we (and I mean any meteorologist) mentions snow is in the forecast, the first question from the public is, "How much?"

It's not that we don't have any idea, but we get concerned that if we give out a certain range of numbers five days in advance, we will be locked into those numbers for the next five days. 

Things change, new data come in.  New observations, new simulations.  We saw this in a big way with the Boxing Day Blizzard in December.  There were wild forecast swings back and forth several days before the thing actually hit.  And I admit to falling into the trap of hanging on each computer simulation (model run).

So for this storm (26th), I made a conscious decision to walk away from the fatigue that comes from hanging over the computer and waiting for the next model run.  Instead, I decided to refocus on the overall large-scale pattern, and from there, determine what the most likely outcome would be 5 days in advance.  With that in mind, here's what I posted on my professional facebook page at 5:10 on Friday afternoon:

Sean Sublette - ABC13 ‎...so a couple (or a few) inches is still the best idea for now. Probably would not start until afternoon Tues. Updates over the weekend.

Sure enough, there were wild swings in the computer model data over the weekend, but there were no consistent and obvious trends in the data.  We look for trends in the models to see if they are converging to a specific weather scenario.  For example, the freezing line may be farther and farther in one direction in each successive model run. 

Didn't really see much of that over the weekend… until Sunday night when I took a little time to go over the latest data.  Here's the post from 11:26 Sunday night:

Sean Sublette - ABC13 A little more convergence to an idea for the storm this week. Something like this: Storm begins as light to moderate rain for several hours, then ends as a couple (perhaps a few) inches of snow. Time frame is a bit slower, so it would appear to start a couple of hours before dawn on Wednesday morning... ending Wednesday night.

In hindsight, that was a tremendous forecast for being more than 48 hours away from the onset of the precipitation.  I should have quit when I was ahead. 

But the drive to be as precise as possible remained.  No one wants to be wrong.  It's one thing to be wrong when all other meteorologists are wrong.  But to be out on your own and be wrong is painful.

On Monday, there appeared to be a trend in the models suggesting very cold air would advance into the state just as a final surge of precipitation would move through.  If the timing were right, we could legitimately get 6-8" of snow.  But it had to be just right. 

We knew the window for this storm was going to be small… probably only 4-6 hours of snow.  Could we get that much snow that fast?  The ground temperature would be above freezing, and it would still be wet from the morning rain.  In addition, the snow would start at least a couple of hours before sunset.  Those things all work against snow accumulations.

But the snowfall rates were expected to be very high.  Convective snow… similar to what we have in summer as thunderstorms.  That is why there was thunder and lightning when this thing came through.  That was not a surprise.  Although in retrospect, I should have given that aspect of the forecast more prominence.

Anyway, I figured there were just too many factors working against big snow, so on Monday night's forecast, I held my ground:  1-2" in Greater Lynchburg. 

I get up Tuesday morning and look at the trends in the data.  Sure enough, the colder air looked like it would be advancing more quickly.  I begin to hear things from other meteorologists like 3-6, 4-8, 5-9.  The raw model data supported those numbers, and to be embarrassingly honest, some peer pressure develops.  Are they seeing something that I am missing?

Those data trends are too substantial to ignore, so I make the decision on Monday afternoon to take us to 3-5.  While I know those much higher numbers are plausible, I have a hard time convincing myself that they are likely.

No big changes show up in the data late Tuesday night, so I decide to stick with 3-5, but it was not a confident forecast.  In a facebook note Monday afternoon, I tried something different: Forecasting with odds… Vegas Style:

For those who like numbers, here are the Lynchburg Snow Total Probabilities:

Less than 1".... 10%

1-3"... 25%

3-5"... 40%

5-8"... 20%

More than 8"... 5%

What this tells you is that in my mind, the odds were less that 50/50 that I think I will be right with the idea of 3-5.  I weighted the odds to the lower end, because I felt there was a better chance of this thing going belly-up rather than ballistic.

Doing the math tells you that even though I thought the odds of getting 3-5" were 40%, the odds of getting less than 3 inches were 35%.  Not much difference there.

Turns out that was the case.  Most of Greater Lynchburg got about one inch… give or take a few tenths.

Each storm is a learning experience.  This one had a lot to teach.  I'd be curious to hear your thoughts on that Vegas Style Forecast.