As the election season grinds on political forecasters line up to predict the outcome of the November elections. Even with seemingly unlimited voter polling tools election prediction accuracy isn’t getting much better.
The financial types also fail more than succeed in forecasting the future. How many saw the crash coming in 2008 and predicted the severity and lasting impact of the market failure? Not many. And predicting when and where earthquakes will occur is essentially a guess.
But weather forecasters have made real progress. According to a report in the Sunday New York Times a high temperature forecast made three days in advance missed by an average of 6 degrees in 1972. Now the accuracy average for that same forecast is plus or minus 3 degrees cutting the prediction error in half.
Hurricane track forecasting has improved even more. The average error in the forecast of where a hurricane would strike land was 350 miles 25 years ago, but now the average error is about 100 miles when forecast the same three days in advance. The Times also points out that in 1940 the average American had a 1 in 400,000 chance of being killed by lightning. Now those odds are 1 in 11 million, much of that risk reduction due to greatly improved weather forecasting.
Most of the weather forecast improvements have come from better, and more comprehensive raw data collection, and from more powerful computers to crunch that raw data. Automated observation systems on the ground feed a steady stream of atmospheric information such as air pressure, temperature, dew point and wind direction and velocity into the forecasting system. Ground based sensors can “see” the wind aloft and measure its direction and speed. And many airplanes—mostly cargo jets—carry systems that automatically report air pressure, temperature, wind and even moisture from the flight levels.
Computer solutions are, of course, absolute, but there are no absolutes in weather so human forecasters still intervene. Even though the weather service uses some of the fastest and most powerful computers on earth, human analysis and modification of a computer generated forecast improves accuracy by at least 10 percent, and often as much as 25 percent.
Weather forecasters have also enhanced the usefulness of their product by embracing chaos theory, the theory that dynamic situations like the global atmosphere are so complicated that tiny, unknowable events, can make dramatic changes. That’s why you seldom see 100 percent chance of rain in a forecast, or why a hurricane forecast track is always a cone of probability, not a single line.
But when it comes to aviation forecasts, we don’t get that same “chaos theory” range of probability. Look at the TAF for an airport and you see predictions for wind direction and speed, cloud height, cloud cover, visibility and type of weather phenomenon expected. Occasionally a TAF will contain a probability for showers, for example, but more likely is the VCSH note that means “showers will be in the vicinity of the airport.” Interesting, but VCSH, or VCTS (thunderstorms in the vicinity) don’t give you a glimpse of how likely the forecaster thinks the showers or thunderstorms are to form, or to move over the airport.
Area forecasts that pilots flying VFR rely on most are even more absolute, and thus less usable. An FA (area forecast) lists expected cloud coverage and height, possible precip, and restrictions to visibility. But there is no mention of probability of the forecast being accurate. Instead the aviation forecasters cover their tails by using terms like “scattered” or “broken” as a way of saying “maybe”.
Look at the forecast for the general public covering the same area and you will see a percent probability of the forecast being accurate. You probably won’t see 50 percent chance listed very often because the weather types see that as weaseling out with a 50-50 coin flip. But you will see 20 or 30 percent which shows low probability, or low confidence, or both. And you will see 70, 80 and sometimes 90 percent probability in a general forecast which is really going out on a limb by the forecaster.
Apparently in aviation forecasts we just get the worst case scenario with no probability for us to judge the forecasters confidence. When you see the TAF predict the ceiling to be 600 feet, for example, we don’t know if that is 90 percent likely, or 20 percent likely in the forecasters opinion. Only that 600 feet is possible, or maybe likely, or could happen, but we don’t know.
So if I have one wish from aviation weather forecasters it is that they would treat us the way they do the general public. Tell us how confident they are by listing a probability of accuracy as a percentage the same way general forecasts do. Armed with that information we could make better preflight decisions, or more useful decisions in flight, about the weather. I do believe forecasts are more accurate than even a few years ago, but a scorecard would be nice.