Читать книгу Weather For Dummies - John D. Cox - Страница 19

Computer models, or software programs, of numerical weather prediction divide the atmosphere over the surface of the Earth and above it into imaginary individual blocks, or gridpoints. Different models use different techniques to force changes in their virtual atmospheres. Some are better at depicting the progress of one type of weather change, and some are better at others.

Each of these separate blocks of air has features, such as temperature, pressure, wind, and humidity, with their own values. The computers take the data of all these values from all these blocks of air and apply a set of equations that represent basic physical laws. (No, I know you must be deeply disappointed, but the only equation you are going to find in this book is 1 Weather For Dummies = 0 equations.)

Anyway, the output of this incredibly intensive process of computation produces a new value for each key weather feature at each gridpoint. But each complete computation moves the atmosphere forward into the future only a few minutes.

And so the whole process starts all over again. The computer begins analyzing the data in each block with the new values it has forecast, nudging the weather system forward a few minutes at a time until the desired length of weather forecast is achieved. The experts say that to produce a weather forecast of a few days, modern computer models may need to complete more than one trillion calculations (1,000,000,000,000). So just how super are these computers? Scientists figure they are about six million times more powerful than the average desktop computer.

The forecast method known as Numerical Weather Prediction has led to some pretty impressive improvements in recent years, but nobody expects computer models to make individual human forecasters obsolete anytime soon. Even with all that computing power, the numerical prediction models are not detailed enough for every weather forecasting purpose. For example, the models don’t very accurately account for the effects on weather of local landscape features such as mountains or lakes. And thunderstorms, which look pretty big when they’re coming at you, are too small and too short-lived even to show up on the big numerical models.

And another thing about all this computing power and these state-of-the-art weather forecasting models — did I mention that they are not always accurate?