The election made me do it.
Writing this essay sprung from listening to statistical polling analyst Nate Silver in the car the other day. He thought one of the many flaws in the presidential polls and their interpretation came from polling companies and media analysts failing to communicate the inherent uncertainties in polls that show a 3 or 4 point spread.
Relax; this is not going to be a political column.
But Silver triggered my recollection of numerous lectures I attended at past American Meteorological Society conferences on the necessity of communicating uncertainty. How much uncertainty is attached to any given weather forecast or extended range outlook?
In operational meteorology, we don’t deal with 3 or 4 percentage point spreads between possible forecast solutions and different models’ output. The uncertainty in projecting the behavior of the atmosphere over a small portion of a spinning globe, three-quarters of which is water covered, heated very unevenly by a thermonuclear furnace 93 million miles distant is generally higher than a single digit percentage figure.
However, we do have laws of physics and numerous equations to help us juggle the interactions. And we have powerful computer models and ensembles of those models to give us a range of solutions we can hopefully narrow down with continuing education, analysis, pattern recognition and a base of experience.
The uncertainties in forecasts vary widely from situation to situation. As I write this article, it’s a mostly sunny, breezy day. Last night when I worked at WKBW, the uncertainty for today and tomorrow in my mind was very small. It was a high-confidence forecast, and I didn’t have to do any on-air hedging.
I think most of you realize forecast confidence can drop off considerably in more complex scenarios.
An example: A few years ago we had an unseasonably cold storm system on a favorable track to bring us heavy precipitation fairly late in April. I knew if snow developed, it would be heaviest at high elevations, but the quandary was whether significant slushy snow would also fall at low elevations. With marginal temperatures near Buffalo and the lower elevations, it was an exceedingly difficult call. Most of us chose to go with marginally moderate to heavy snow even near the lakes. In fact, a cold rain was what resulted when our low level projected temperatures were just a couple of degrees too cold. On some of the high hills to the south, 15 to 18 inches of slushy snow fell on those thinly-populated locations. It was a lower-confidence forecast that turned out to be a bust for the large majority of our audience. And we all recall the October storm of 2006, which I wrote about recently.
In both of those forecasts that didn’t pan out very well, I know I was careful to express the uncertainties beforehand. Not all forecasts are created equal, or even close to equal in confidence.
It’s vital that we meteorologists bring the public the perspective on confidence and uncertainty, especially on tough calls and on seasonal outlooks. Most of you are busy and want a quick, firm answer to your forecast needs. The atmosphere has too many variables impacting its behavior to always supply a confident forecast.
As I’ve written before, the air around us isn’t Lego blocks that snap neatly together. Fluids, gases, uneven heating, oscillating interactions between the air and the sea – the list of variables goes on and on.
So I hope you’ll forgive us when it appears we’re hedging. Hedging can sometimes be the way to go.
I can give you a medical analogy where the stakes may be higher. Say you’ve had a worsening dull pain for a few weeks in your lower left torso and some nasty indigestion, so you go to see your doctor. After describing your vague symptoms for a minute or two, would you want your doctor to jump up and say, “I’ve GOT this!” Spit out a diagnosis and write a script? No tests, no consultations with radiologists or gastroenterologists? I don’t think so. What happened to scans, tests? Why would a highly trained doctor toss aside looking into uncertainties?
The same question should be asked of meteorologists – and, I suppose, pollsters.
Competent scientists are expected to express uncertainties and probabilities. That may be frustrating for some users, patients and the public. But it is quite literally the best we can do and should do.
Speaking of which, I’m expressing plenty of uncertainty at this point as to how much snow will fall and where it will fall this Sunday.