Tell Me Again How It Didn't Snow Anywhere
UPDATE 3/12/2013: A couple dozen AccuWeather forecasters had a great discussion today on what went wrong with the forecast in Philly & D.C. In hindsight, we identified some key meteorological points that were clues to the eventual storm outcome (I apologize for some of the meteorological terms here, unfortunately I don't have time to explain them all).
BEFORE THE STORM:
- This was a very unusual situation and the storm didn't follow the rules & the storm was not predictable 3-4 days out
- The boundary layer temps were never low enough, yes they could have been overcome by upward motion but it wasn't there either
- Looking at temp profiles would have confirmed that temps didn't support heavy snow accum. in DC
- There was little PVA (an upper-level measure of dynamics that produce precipitation) forecast
- When in the Plains (Mon.), QPF was overestimated by models for East
- The NAM's QPF (quantitative precip forecast) is almost always too high (time and time again)
- That said, NAM was only model saying heavy snow for Boston area, which happened
DAY OF THE STORM:
- There's a thin line between giving time for things to verify (don't call it 'til it's over) and admitting that meteorological things are going wrong
- The day of the storm, the snow area started shrinking, not growing, after 10:30 AM ET
- The closed low was further south than predicted, and moving east, no chance of cold air after that
- The line of storms in the DelMarVa may have robbed the northern areas of moisture
- The surface low was not intensifying after it formed; it was nearly vertically stacked (bad for heavy precip)
- The warm Tues. eve. 850mb temp push helped create the snow hole over DC
- Note that the 850 temps were no lower than -5, and it was raining when they were -5; there was no cold air coming in
- The Euro snowfall hinted (last minute) at a snow hole very close to heavy snow near DC
- The 4KM NAM (and later, radar) was showing heavy snow near PIT, although it ended up further north
- The storm was way off the coast, you wouldn't normally see it bringing heavy snow inland, but the piece of energy (which the GFS predicted first) coming from the north helped
- Underperformance in mid-Atlantic may have caused meteorologists to lower their Boston amounts (psychological, not meteorological)
UPDATE 3/11/2013: Thanks for (most) of the comments below. Read both Henry's blog and Elliot's blog today to see an example of what we meteorologists deal with. For next week's storm, the models are predicting two different solutions. Here's another look at the D.C. snow hole from last week's storm - note the higher amounts in every direction!
UPDATE: Here is a smoothed out, better looking version of the actual snowfall (bottom) versus our forecast (top). Even though the keys are different, you can see that the east and northern fringe was much smaller than expected (although this image exaggerates the area near D.C., which did get *some* snow), and the southwest Virginia area underperformed, while the 10"+ area overperformed in the north and in Pennsylvania.
UPDATE: Capital Weather Gang has their forecast fail explanation online. They also have an excellent article which all detractors should read:
- Your responsibilities as a weather forecast user
Also, we just added this image from Twitter to our photo story. It illustrates the stark contrast observed with this storm.
UPDATE: I realize I haven't explained why, meteorologically, eastern Washington D.C. and the Philly area got less snow than expected; there are a couple of excellent blog posts out there on that, including WJLA, this blog, Bob Ryan, and our own Joe Lundberg. The best explanation may come from the guys who live in the "D.C. Snow Hole" (Capital Weather Gang) but they haven't delivered yet (as of 1 PM) on a promise to explain yesterday's forecast.
I'm defending AccuWeather (and other meteorologists with similar forecasts) today against attacks on the Internet from those who didn't get "enough" snow yesterday. Was the forecast perfect? No, but if you look at the maps below, our forecast (two days out nonetheless) was mostly accurate. This is NOT what a "busted" forecast looks like, folks:
The highest amounts by state were (drumroll please): 24 inches at Franklin, W.Va.; 20.3" at Fishersville, Va.; 13" at Swanton, Md.; and a foot at multiple locations in Pennsylvania.
First, let's talk about what went wrong. Winter storm forecasts (especially in March) are always inaccurate around the edges due to the difficulty of predicting exact storm position or where the rain / snow line will be. But let's overlay a high-population map (in red below). The "problem" this time around is that some of the most populated areas (like D.C., Baltimore and Philadelphia) were progged to get a few inches of snow, and received less than 3, and in some cases, no snow accumulation at all. This map, showing high East Coast population vs. our original forecast, shows why the vitriol on social media was so bad this time.
Note that our worst forecast was for southwestern Virginia (where we predicted 6-12 inches and they got nothing) but we're not hearing from the population there, because it's very small. Meanwhile, D.C. people are eating us alive. But did we really bust in D.C.? Most of the city got 1 inch of snow or less, but reports of 2 inches and 3.5 inches fell just over the river from the nation's capital. Considering our forecast was 3-6 inches there, I don't consider 3.5 inches a "bust" for a city that isn't even 10 miles across.
CHALLENGE: Can you tell the difference between 0.1" and 0.5" of rain in the summer? Most of the public can't, yet that's the equivalent of what happened yesterday and these are some of the complaints we're hearing about. Snowstorm forecast inaccuracies are magnified.
The idea, which people are perpetuating in social media, that it didn't snow at all, or only snowed on mountaintops, is ludicrous. Some of the comments are simply from Internet trolls and I won't even address them. However, comments like this (on my Facebook Page) infuriate me:
A simple look at the map above shows that snow fell over an area close to what was forecast -- and in fact, the heaviest snow covered a larger area than forecast. I'm not making this stuff up -- these are observations of snowfall from NWS spotters & COOP reporters. Here's the map if you don't believe me.*
What of Fredericksburg, VA, where we predicted 8 and they got 9 inches? Or Richmond, VA where we said 2.5 and they got 2.0? How about Frederick, MD where the map said 6.5 inches but they got 5. All of these were excellent forecasts.
The problem with the Internet (if you haven't noticed) is that people can make outrageous claims with no proof. Take this comment for example:
Look again at the map. AccuWeather did not predict that anyone would get 18 inches of snow, for starters. Secondly, the area of 10-12 inches or more is actually larger than forecast, so actually more people got 18 inches or more than were expected to. Here's a list of 24 reports of 18 inches or more, from eight counties in Virginia and West Virginia.
Here's a similar falsehood:
Nobody ever forecast 6-12 inches for southeast Pennsylvania. We did overestimate the northeastward extent of the light snow; our forecast map above shows 1-3 where none fell (towards Philly) but 3-5 inches fell near York and Gettysburg, validating our forecast there.
Even the jokes are incorrect -- although this is amusing, a prediction of 1 to 3 inches does not a blizzard make. This description of how weather gets weird in Newark, N.J., however, is amusing, although exaggerated -- and could apply to almost any city in the mid-Atlantic and New England, which is what makes it funny.
I'm not saying this forecast was perfect, but I think it was good overall, and I think we need to be realistic when we're talking in Social Media. If you think we were wrong, then state that, but if you're going to say something outlandish, please offer up some proof. Do your research and look at the maps, don't trust your memory.
And finally, let's drop the "meteorologists are always wrong" comments. It's well-documented that short-range weather forecasts are 90% to 95% accurate; 5-day forecasts are range from 65 to 75%, depending on how you measure it, and all weather providers are within a few percent (if you can tell the difference between 45 and 47 degrees, congratulations, I can't). The reason it may not seem meteorologists aren't that accurate is because of perception of statistics. I bet you never remember when we say it's going to be partly cloudy and it's partly cloudy. You only remember when we're wrong.
*Yes, the keys don't match exactly, but I made the colors about the same, and I did remove "zero" amounts that were in areas that got over 6 inches, because they were bad reports that just weren't meteorologically possible. I did not alter the edges of the graphic, other than to recolor it.
NOTE: Older comments are no longer available due to a Facebook Comments change, but you can click below to view a screenshot of all of them.
