IStock Photo 8309384 © Saso Novoselic

It’s Saturday morning, and Joe and Jane have plans: nine holes of golf with friends, then back to their house for a barbecue. But as they sip their coffee, news on the radio casts a shadow over their day: “Our forecast calls for mostly cloudy weather, with a 60% chance of showers and thunderstorms.”

Will Joe and Jane go to the movies instead? That depends on how well they understand the forecast. Some couples might assume that it will rain for 60% of the day, or that there’s a 60% chance (1 in 1.66 odds) that it will rain in their neighborhood. Many people would change plans in either case. The right answer, though, is that there is a 60% chance of measurable rain (at least 0.01 inches) falling sometime during the day at any given point in the forecast area, which could cover many square miles. Given those odds, Joe and Jane might well head for the links.

Although they aren’t usually explained in radio and TV weather reports, the National Weather Service (NWS) assigns specific probability ranges to the terms it uses to predict rain and snow. “Likely” means a 60 to 70% chance (1 in 1.66 to 1 in 1.42). A “chance of rain” means the odds are 30 to 50% (1 in 3.33 to 1 in 2), while a “slight chance” usually equals a probability of about 20% (1 in 5).

The vague-sounding terms that describe sky conditions (the mix of clouds and sun) also represent real numbers. When skies are “clear” or “sunny,” forecasters expect that less than one-eighth of the sky will be covered by opaque clouds. “Mostly sunny” means that one to two eighths will be covered; “partly cloudy” or “partly sunny” means that three or four eighths will be obscured; and when skies are “mostly cloudy,” five to seven eighths will be covered.

Forecasters develop these predictions with sophisticated computer models that analyze data from ground observation stations, weather balloons, and satellites. Some commercial sites (which use NWS data plus their own information) predict weather up to two weeks in advance, but forecasts more than three or four days ahead are not very reliable. That’s because weather patterns are dynamic interactions among many variables, including temperature, atmospheric pressure, humidity, and wind. Minor changes to one input, such as the speed of an advancing cold front, can cause major changes in outcomes—a phenomenon known as “the butterfly effect.”

Before radar, satellites, and supercomputers were invented, weather forecasting was more of an art than a science. Many conditions spawned popular sayings: some were nonsense, but others contained a grain of science. For example, “Red sky at morning, sailors take warning” means that the rising sun is reflecting off of dust particles and water vapor in the air, a sign that a storm may be moving in from the west (storms in North America usually track across the continent from west to east). “Red sky at night, sailor’s delight” signals that the setting sun is shining onto clouds that have already moved away from the western horizon, so the next day is likely to be clear.

To make things even simpler, assume that conditions will be the same tomorrow as they are today—the so-called persistence method—and the odds are better than 1 in 2 that you’ll be right. In one experiment, a Salt Lake City meteorologist found that the persistence method predicted temperature and the presence or absence of rain more accurately than climatology calculations or two commercial weather services.

The persistence method works best in areas where weather patterns are very stable, but it’s far from foolproof. Remember Harris Telemacher, the weatherman played by Steve Martin in the movie L.A. Story, who pre-records sunny forecasts so he won’t have to go to in to the studio on weekends? “This is L.A. What’s going to change?” Telemacher says to his producer. But then it pours on the network owner’s fishing trip and Telemacher gets fired, giving new meaning to the term “rained out.”