For Entertainment Purposes Only–Part III

What a winter!  Enough snow to make Western ski areas envious has blanketed the Mid-

Atlantic region.  The U.S. government was shut down–not that anybody noticed–for several days.  And there is undoubtedly more nastiness to come.  It is, after all, only mid-February.

So what better time to press on with my series on extended-range weather forecasting.  Below is Part III, drawing on material from my unpublished manuscript INSIDE THE WEATHER CHANNEL.  You can find Part II here

Oh, and if you’d like to see another take on this winter’s weather (a bit tongue-in-cheek) take a look at my February 15 blog for The Weather Channel.

FOR ENTERTAINMENT PURPOSES ONLY–Part III

Extended-range Forecasts

NEGATIVE SKILL

If nothing else, I suspect you’re beginning to understand why weather forecasting beyond a few days into the future is so darned difficult.  But don’t take my word for it.  Consider a study from a group of  Texas A&M researchers on the reliability of The Weather Channel’s probability of precipitation forecasts.

Reliability has to do with the validity of the percentage chance of precipitation predicted, not whether it was a good yes-or-no forecast.  That is, if a prediction calls for a 40 percent chance of precipitation, does it indeed rain or snow on 40 percent of the days for which such a forecast was issued?  If the answer is yes, that’s a reliable prediction.

It turns out that the predictions are pretty good out through about three days.  But after that, uh oh.  What the study discovered was that although there remains positive skill in forecasts out through seven days, it decreases steadily each day.  Skill is measured against climatology.  In other words, a prediction is evaluated against whether it gives you better information than just knowing that six days from now, for example, records indicate rain will fall on four out of every ten days.

After seven days, skill tumbles into an abyss.  Eight days in the future, there is virtually none.  What that means is that forecasts at that range are no better than climatology.  But it gets worse.  Nine and ten days out there is actually negative skill–or at least there was during the period the study was performed from November 2004 to January 2006.  Negative skill means forecasts are worse than relying on climatology.  Saying it another way, you’d be better off checking to see how often it’s rained or snowed historically nine or ten days in the future than casting your lot with a prediction.

It turns out the negative skill present during the time frame of the study was a computerized system shortcoming that has since been addressed.  Still, there remains little positive skill in precipitation forecasting that goes out beyond a week.  With climatology considerations now built into the system, you will never, past the seven-day mark, see a probability of precipitation prediction that exceeds 60 percent.

Let me quickly add that temperature forecasting is much more reliable than attempting to divine precipitation in the extended range.  But even thermal soothsaying can go astray as the length of the view grows.  Forecasters, in fact, really don’t like pushing their credibility beyond about five days.  Mike Seidel, a veteran on-camera meteorologist and field reporter at The Weather Channel has his own view of extended range forecasting: “In most cases, for entertainment purposes only,” he says.

Over the years, I’ve defined my personal limits and have developed what I call the Bernard Three-Day Theorem.  Simply, it’s this: don’t make any plans based on a forecast beyond three days.

SO WHY DO IT?

There’s an obvious question here, of course: If extended-range forecasting is so difficult and fraught with so many pitfalls, why do it?

Simple answer: because management said so.  Television, you see, is a copycat industry.  Once something new or different that seems to work hits the air, every network and every local channel–not wanting to appear as if they’re a backwater operation–jumps onto the bandwagon. In the meteorological field, you may remember that terrible 3-D weather depiction that was offered up for awhile.  It was totally useless and uninformative.  I certainly never knew what I was looking at or what it meant.  But it was new and different, swoopy and zoomy, and must have appealed to somebody.  Once one station had it, every station had to, useless or not.

So it was with seven- to ten-day outlooks.  Once one station or network put them on the air, every operation had to, including The Weather Channel.  Meteorologists at the channel fought it, arguing their credibility would be stretched by providing outlooks beyond five days.  But no matter.  The channel had to “keep up with the Joneses.”

Ultimately, even the position I held at the channel, a sort of briefer-communicator-coordinator function, was tasked with producing a seven-day outlook every Monday.  This was an outlook for internal use only.  I’m not sure who all it went to, but apparently it was used to gauge how large The Weather Channel’s audience would be during the up-coming week.  That is, would there be a weather event (or events) that would send people flocking to the network?

I, or whoever handled the shift–usually me, since I worked virtually every Monday–had to issue a more detailed outlook than was available by merely looking at seven-days worth of graphics generated from the Global Forecast Center’s predictions.  These graphics, the ones you see on TV, are very generalized.  For instance, it’s unusual beyond the three- or four-day point to see “rain” or “snow” forecast; usually you’ll just see rain or snow “showers.”  This is driven by the inherent uncertainty (and concomitant lower probability of precipitation predictions) as outlooks reach further into the future.

Never, on the extend-range graphics, are there any indications of major storms or severe weather outbreaks.  Yet these were precisely the types of events I was expected to foresee.  To do this, I had to become my own little Global Forecast Center, examining up to a half-dozen different models, studying the ensembles and reading the extended-range forecast discussions issued by the NWS’s Hydrometeorological Prediction Center.

After doing all of this, there were still more than a few times I didn’t have an inkling on Monday what the weather pattern was going to be on Sunday.  As a long-time friend and colleague of mine, David Spiegler, an American Meteorological Society Fellow, has often reminded me: “Meteorology is a humbling profession.”

Photo: Snow cover in an Atlanta suburb, February 13, 2010.
He didn’t deliver the feet of snow with which he has buried the Mid-Atlantic region this season, but Old Man Winter finally slapped the Deep South this past weekend.