Note ... this is a slightly modified re-post of a guest blog here.
I’m among the first to complain about people offering their opinions about what “the public” wants from weather forecasts, rather than collecting evidence through a process of literally asking a representative sample of people. However, the latter is not something easily done. “The Public” is not a homogeneous block of people with equal needs and expectations. Rather, it’s quite diverse and it’s not obvious to me even how to go about collecting a sample that might be accepted as representative (by those whose expertise is in doing such surveys). There are some social scientists who have such expertise, I’m sure. I might even know some of them.
Nevertheless, I’m going to go ahead and offer my unvalidated opinion regarding this issue, anyway. I’m working with the notion that “the public” in this context excludes all meteorologists and those who already are adept at using weather forecasts effectively. My perception is that most people don’t pay much attention to the weather most of the time, and know little or nothing about how it works, or what we meteorologists can claim legitimately to know about the atmosphere. When they hear a forecast, if they think it might actually matter to them on a particular day (for whatever reason), they want the forecast to be perfect so their lives will be spared (if hazardous weather is possible) and/or they won’t be seriously inconvenienced by the weather as they go about their business.
Regrettably, forecasters never know with absolute certainty exactly what’s going to happen – high uncertainty typically is present on a day when the weather is changing rapidly. I’m not going to go into a long-winded discussion of the sources for weather forecast uncertainty, but they generally arise from the fact that the weather evolves from some starting structural state to some other state according to atmospheric physics that we know only imperfectly. We don’t even know the starting point with absolute accuracy. It’s sort of like putting together a complex itinerary for a trip, where we don’t know exactly where we’re starting from, and we have incomplete and imperfect knowledge of how the transportation system operates. We will almost certainly wind up in a different place than what our original destination was thought to be, although in the case of weather forecasting, it usually turns out we come fairly close most of the time, despite being forced to use incomplete information.
Wanting forecasts to be perfect is natural and very understandable. We think our own lives are too complex to be completely and accurately predictable, but if we can rely on the weather forecasts to be perfect, it makes our decision-making a lot easier. Re-schedule that picnic if it’s going to rain. Water your garden if it’s going to stay sunny and dry. Go to the pharmacy to refill your prescription before the heavy snow flies. In fact, this is just what's happening on most days as a result of the existing imperfect forecasting systems we use – people can and do make use of our forecasts for just this sort of decision-making despite the imperfections of the forecasts. If someone makes a bad decision and everything goes bad for them because of the weather, they can always blame the damned forecaster! Some surveys I’ve seen make it clear that many in the public know and understand our forecasts aren’t perfect, but still some people become upset when the weather doesn’t follow precisely what they heard in the forecast(s). Note that in the real world, one thing forecasters do is to update their forecasts based on new weather information. Hopefully, it won’t come as a surprise to most people that our forecasts get worse, the farther ahead they are predicting. Conversely, we improve as the “lead time” gets shorter. Don’t expect the forecast for weather a week in advance to have the same level of accuracy as one 12 hours in advance!
When the forecasts are changing frequently as a result of new information, this is usually because of large uncertainties on that day. Not all days are equally difficult to forecast, of course; our forecast uncertainty is not a constant. In fact, our uncertainty is also not perfectly predictable!
Let me tell a personal anecdote that I’ve used often to illustrate the value of knowing and using the uncertainty information in a weather forecast. Some years ago, on a fall football weekend here in Norman, there was a slow-moving, strong front in the OKC area (about 20 miles north of Norman). On the south side of that front, skies were mostly clear and temperatures were expected to rise into the mid-70s (in deg F) in southerly winds, while on the north side of that front, skies were overcast with low clouds and rain with temperatures in the upper 30s or so, and a strong northerly wind. It was about equally likely the front would stay north of Norman or push a few miles south of Norman by mid-day (around the time the game kicked off). The forecaster didn’t have the option of saying that the weather that day had about a 50% chance of either option, so the forecaster was forced to make a choice. As it turned out, the forecast decision that morning was for warm and sunny, whereas the real weather turned out to be miserably cold and rainy. Tens of thousands of football fans were caught in summer clothing because they accepted the forecast, and they were not happy! Since I understood the situation, I dressed for the warm option, but carried cold weather rain gear in my backpack. It was a simple matter to prepare for both possible outcomes! I’ve often told this story and then asked the audience: “Would you prefer to be offered the whole story of the forecast, including the uncertainty, or do you just want the forecast without any uncertainty information?” I almost never get anyone who chooses the latter option! Is that surprising to anyone? Nevertheless, many people just want to know what’s going to happen, even though most of them understand the science doesn’t allow them to have absolute certainty.
Every forecast that doesn’t include uncertainty information is tantamount to withholding critical information from the public! And the public needs to accept some responsibility to learn how to use that uncertainty for their own purposes – they have to set their own thresholds regarding uncertainty. If the worst thing that could happen to you is getting a little wet, you can accept more uncertainty than if you stand to lose your life if some hazardous weather potential exists. Unfortunately, low uncertainty, highly confident forecasts are just not possible in some situations. We can’t predict precisely the path and intensity of a tornado, so a tornado warning generally always has relatively high uncertainty. The same can be true for deciding just when and where winter storm weather will occur. From a meteorological standpoint, getting the heavy snow band to within 50-100 km of its eventual location is an excellent forecast. But that might mean the difference between heavy snow mostly in rural areas versus in a major metropolitan area. Expecting that forecast to be perfect is just asking to be frustrated. People can want a perfect forecast, but people in hell want a glass of water, too. Are they going to get it? Nope. Likewise for perfect weather forecasts.
C’mon people! You know we can’t make forecasts with absolute certainty, so why keep complaining when it turns out we can’t make perfect forecasts? The forecasts have been improving steadily, and are much better than we were even 10 years ago. The public is being well-served, as I see it. Where we have a problem is communicating our uncertainty and the public is remiss in not working very hard in trying to learn how to use any uncertainty information we do provide. It would be nice to figure out this bottleneck. Sadly, I have no easy solutions to offer.