Why can't scientists accurately predict the weather?

Yeah, we know it's hard to keep that sunny smile on your face when you're juggling all the variables that go into an accurate weather forecast.
Yeah, we know it's hard to keep that sunny smile on your face when you're juggling all the variables that go into an accurate weather forecast.
George Doyle/Getty Images

While most scientists are revered for making sense of our complex universe (Einstein is practically a hero), meteorologists often face ridicule. How can we put a person on the moon or foretell planetary alignments years in advance, yet still fail to put together accurate weather forecasts?

First, to give credit where credit is due: Weather forecasters have improved their game significantly over the last 20 years. The three-day forecasts they deliver today are better than the one-day forecasts they delivered 20 years ago. They're also much better equipped to provide advanced warnings of severe weather, doubling the lead times for tornado warnings and giving people an extra 40 minutes to escape flash floods.

Modern meteorologists wouldn't be nearly so accurate without numerical forecasting, which uses mathematical equations to predict the weather. Such forecasting requires powerful computers and lots of observational data collected from land, sea and air. A single weather station would never be able to collect so much information. Instead, thousands of stations across the globe are linked and their data pooled. Some of these stations -- ground-based wind gauges (what meteorologists call anemometers), rain collectors and temperature sensors -- resemble those used by amateur weather watchers. Others lie far out at sea, strapped to buoys. And still others travel on commercial airliners or shipping vessels, collecting weather data as passengers and goods are moved from point A to point B. Finally, weather satellites and balloons provide information from the upper regions of the atmosphere. Satellites photograph Earth's weather from their orbit in space, while balloons monitor upper-air data over a particular location.

Collectively, all of these sensors and gauges produce more than 1 million weather-related observations every day. A normal computer -- the kind you buy at your local electronics store -- would choke on all of this data. Luckily, meteorologists can rely on supercomputers, crazy-fast machines that perform millions of calculations per second. In the United States, these computers are housed at the National Centers for Environmental Prediction (NCEP), located in Camp Springs, Md. There, weather observations stream into a supercomputer's brain, which uses complex mathematical models to predict how, based on the incoming data, weather conditions might change over time. The computer's output form the basis of almost every forecast broadcast on radio and television channels across America.