Why It’s Hard to Forecast the Weather | National Geographic

People have short memories, and you're only as good as your last forecast. So, if you mess up a forecast, especially a high impact forecast, people will remember that. A 3-day forecast today is about as accurate as a 1-day forecast was in the 1970s. If you look at a one-day forecast, for example, of temperatures, we're within about 3 to 4° of being exactly right 90% of the time.

You have to, as an expert, give your best bet, but you also want to communicate that, especially with a high impact forecast, there's this range. There's this idea of Chaos Theory, which means at some point, there's a limit to how far out you can predict. Just because a random fluctuation somewhere can lead to big changes, which propagate and grow and get bigger and bigger as you get farther out in time.

There are several forecast models we use. There is an American model known as a GFS model, which is a Global forecast model that provides forecasts out to 16 days into the future. Then, the European Center for Medium-Range Forecasting has a model called the European model. If you look at statistics of how accurate the models are, the European model is the most accurate.

Some storms, the European model will do better; other storms, the GFS model will do better. Even though statistics might tell you that the European model is the best, you can't take it to the bank in every case. When you compare the jobs that weather forecasters do in terms of their batting average compared to other professions—whether it's diagnosing a disease as a medical professional, or whether it's doing economic forecasting, or whether it's an athlete who has a batting average of .333—you know, weather forecasting is actually pretty good when it comes to performance.