As a meteorologist, the holiday season is often filled with gripes and light-hearted ribbing about how I can be employed in a profession where I can provide the public with wrong information and still retain my job.
“Those lousy weathermen don’t know what they’re talking about!” said one of my aunts. “Just this past summer the TV said ‘twenty percent chance of showers and thunderstorms’ and then it rained on my picnic!”
Well, my dear aunt, the rain your picnic is the reason why the weatherman said there would be a twenty percent chance of showers and thunderstorms. Technically, the forecast was correct. If the weatherman had forecasted a zero percent chance of any precipitation, then that would be a poor forecast.
The major problem concerning the public misconception of the true quality of weather forecasts is that the overwhelming tendency is to remember what is considered a failed forecast, and not what is considered a correct forecast. In general, weather forecasts provide very useful information that the public can use to go ahead and plan their days successfully, but those times when it rains on their picnic are remembered much more readily than the days when everyone in their neighborhood was outside mowing their lawns on a correctly forecasted sunny day.
That being said, I would hedge a bet that there will never be a perfect weather forecast in my lifetime. And as far as I am concerned, I have a lot of living left to do. Why would I say such a thing, when I just spent the last minute or so saying that the public has a distorted view of reality? Quite simply, the atmosphere that surrounds our planet is an extremely complex system that evolves based on the solutions of mathematical equations that cannot be solved completely with current computing power.
You might say, “Well the National Weather Service has huge supercomputers that can run complicated models in a matter of minutes! Why can’t they use these to predict the weather in my back yard?”
This begs for an explanation of what a numerical weather model really is, and what it can provide a meteorologist who may be making a forecast for your back yard.
First, let’s imagine that we want to make a weather forecast for a location somewhere in the continental United States. Picture this map in your mind. Now put a big rectangular box around it. We will call this the model “domain.” A very fast, and very sophisticated computer will create a forecast for this area. Now one thing to take into consideration is that the majority of numerical weather models make calculations of numerous equations that describe the state of the atmosphere at various points on a large grid. A state of the art numerical weather model might use a grid with squares that are 10 km on a side. So now take your rectangle and draw evenly spaced lines, 10 km apart across the entire continental United States. That is the horizontal grid. But that flat surface only represents one level of the atmosphere. We need to replicate that layer a bunch of times and stack them at pre-determined distances apart up through our troposphere (and into the lower stratosphere) because that is where our weather occurs. Now, if you have pictured this correctly, you have a three-dimensional box, full of crisscrossing lines. For sake of simplicity, we can now say that the numerical weather model has to make a calculation at every location where these lines intersect, based on the information at that location, and at surrounding locations. Oh yeah, and it also has to do this every few seconds.
Sound simple enough? Now let us imagine that we are zooming in to the box that surrounds your back yard, which just happens to be perfectly centered in the box. The numerical weather model is now making calculations at the corners of the box. But what is going on in your back yard? What the model “sees” is at the corners. Everything in the middle of the box must be filled either based on averages of the values at the corners of the box, or by assumptions called “parameterizations” of what will be going on inside the box. This is not a trivial matter. The state of the atmosphere is determined by processes that occur at scales as small as the individual molecules of which is it comprised, not to mention your house, your trees, and that patch of blacktop where kids are playing basketball and hopscotch. The wind will flow around and over your house, changing the wind speed and direction. It will then get slowed down by the trees. As it moves over the blacktop, the air gets heated and becomes buoyant, which causes convection. The model can assume what might be going on inside this box, but it cannot be certain unless everything is explicitly defined within the box.
This disconnect is where the power of computers is overestimated. In order to predict the true state of the atmosphere, the ideal solution would be to have a direct numerical simulation, which can model the state of fluids (yes, our atmosphere is a fluid) down to the molecular level. This can be done for modeling fluids in a laboratory, but the enormous scale of the atmosphere makes this task impossible to complete in a timely fashion with the technology of today.
“Ah, you said ‘in a timely fashion,’ so that means it can be done!” I’m glad you are actually paying attention. Another major problem is the time it takes to accomplish these tasks. The National Weather Service offices, or private forecasting companies for that matter, need to disseminate the best possible forecast in the shortest time possible. After all, if you turn on your TV and the weatherman tells you that he will know what the weather did an hour ago when he checks his fancy computer model thirty minutes from now, that is not going to help you make a decision about whether or not to carry an umbrella for the rain that may or may not be falling when you leave work later in the day.
In order to provide the public with the forecasts they desire when the public wants them, concessions must be made when determining the configuration of a numerical weather model. As computer power increases, fewer and fewer concessions will need to be made, as the weather model will be available to meteorologists and the public more quickly. Until these concessions are eliminated (as I previously wagered, it will probably not be during my lifetime) there will always be “uncertainty” in weather forecasts.
Until there is no uncertainty in a weather forecast (doubtful, but I will hold out hope), weathermen will continue to give the public the best possible forecasts, based on current, state of the art numerical weather models, and their own personal experience as a weather forecaster. The key to changing public opinion about the quality of weather forecasts is communication and education about how to use the information that a weatherman is delivering. The majority of the time, the person providing you with your weekend weather outlook will be correct if that information is interpreted in the manner in which it was intended. This, above all, should be remembered the next time an isolated shower puts a damper on your barbecuing festivities.
