Читать книгу Weather For Dummies - John D. Cox - Страница 16

High-powered computer models solving numerical weather prediction equations are a vast improvement in accuracy from the early days of forecasting, no doubt about it, but they are not perfect. As Ed Lorenz and the “Butterfly Effect” make clear, the atmosphere is inherently chaotic. The slightest error in defining the state of the atmosphere can lead to big effects down the road. You can’t get a 100 percent perfect rendering of the state of the atmosphere, and you can’t get a 100 percent perfect solution to the equations. What you get at the end of a single run of the numerical equations is not a forecast of absolute certainty, but rather a “best fit.”

Dealing with this perpetual state of uncertainty, weather scientists have developed another way to tackle the problem. Rather than running a single numerical model with a single set of initial conditions, they run more than one model, and they run them more than once and on different models — tweaking the incoming data ever-so-slightly to get a slightly different result. This method is known as ensemble forecasting, and it produces a broader range of possible weather outcomes. The output of ensemble forecasting produces a variety of forecasts with different probabilities — like the odds at a racetrack. Users are able to see which outcome is the most likely, the favorite, and which is the longshot, and place their bets accordingly.

It represents a big improvement: being able to measure the level of uncertainty in a forecast. An ensemble forecast with a wide trail of possibilities inspires less confidence than an ensemble with a narrow trail. Check out the forecast the next time a hurricane is threatening landfall somewhere. Most likely, you will see a band of approach much wider than the storm. The width of the band is the footprint of ensemble forecasting.

For forecasts beyond the 12-hour period, and especially for general “outlooks” of weather beyond the next three to five days, the mix of forecasting tools may change. The projections of a variety of computer models still are consulted, but the climatology of a place — its average weather for this time of year — is weighed more heavily. This approach focuses on the question, “What usually happens here this time of year?”