They tell us that weather forecasting is seeing dramatic improvements thanks to new supercomputers capable of making hundreds of trillions of calculations per second. The latest computers in the arsenal of the National Weather Service are now more than twice as fast as they were just a few years ago, enabling far more accurate forecasts further out in time.
I don’t know about you, but it sometime seems like weather forecasting is becoming worse and not better over time. For instance, yesterday’s forecast for my part of the country called for partly cloudy skies with a slight chance of a stray afternoon thunderstorm. Instead, it rained buckets for most of the day under low ceilings and IFR visibility.
If the computers are so good, how can the forecasters get it so wrong?
According to NOAA, weather forecasting has indeed become much better over time, despite the odd off forecast. With the latest supercomputers, they will get even better. Much better, in fact.
“These upgrades are a game-changer for the entire public and private weather industry,” says National Weather Service Director Louis W. Uccellini, Ph.D. “In addition to the benefits to our own forecasters and products, we will provide our private sector partners with better information to empower them to enhance their services.”
Nicknamed Tide, the new NOAA weather supercomputer in Reston, Virginia, operates at a speed of 213 teraflops. What does that mean exactly? One teraflop equals one trillion operations per second. So Tide is capable of performing 213 trillion calculations a second, up from about 90 trillion operations in previous generation NOAA computers.
That’s impressive. The increase in computing horsepower will lead to a “quantum leap” in forecasting capability, NOAA ensures us. Still, humans must interpret the computer models. And the models are inherently bad at predicting small-scale weather phenomena, so humans must pull in satellite and radar data as well as surface observations to make forecasts. Sometimes humans get it wrong. As a pilot, you understand that.
We pilots also need to realize that weather can be random. It was a meteorologist (chaos theory pioneer Edward Norton Lorenz) who coined the term “butterfly effect” when he noted that a slight change in model data input can change a model output completely. Has weather forecasting improved in our lifetimes? Undeniably, yes. Will it ever be a perfect science? Even with future super-duper-uber-computers, the answer is no.
Remember that the next time you log on to your favorite weather website or talk with a briefer before heading out on a flight. And make sure you always have a Plan B for when the computers and human forecasters have an off day.
We welcome your comments on flyingmag.com. In order to maintain a respectful environment, we ask that all comments be on-topic, respectful and spam-free. All comments made here are public and may be republished by Flying.
