Predicting Colorado’s volatile weather has long challenged even seasoned forecasters. Now, meteorologists say artificial intelligence is reshaping how they anticipate the state’s rapid swings, a shift they discussed Friday during the fourth annual Colorado SunFest at the University of Denver.
The real advantage of AI isn’t necessarily more accurate forecasts, but speed — allowing meteorologists to run far more model simulations than was previously possible, said Joel Gratz, founding meteorologist of OpenSnow.
“Instead of running that forecast one time, five times, 10 times, you run it 50 times, 100 times, eventually you run it 1,000 times and see what those probabilities are,” Gratz said during a panel discussion with local forecasters.
9News meteorologist Chris Bianchi said as he forecast one of Colorado’s driest winters on record, marked by historically low snowpack and a late-March heat wave, he watched AI models improve in real time, delivering sharper guidance than traditional forecasts.
“I now look at the AI model over the numerical models, the old models — and that is new as of probably two months ago,” Bianchi said.
Within five years, Bianchi expects AI to significantly improve short- and medium-range forecasts, increasing the reliability of five-day outlooks and beyond.
“I think within five years, you’re gonna see pretty darn good AI-driven forecasts up to two weeks out,” he said. “I think it’s pretty bold — I’m not sure I’m quite ready to say that on TV yet.”
One forecaster on Gratz’s team trained an AI model to predict specific weather events, like tornadoes and thunderstorms, using 40 years of tornado reports alongside data on lightning, hail and wind, he said.
“We couldn’t do this four or five years ago, let alone 30 years, but now, we have a system that updates every two minutes looking at radar satellite models,” Gratz said. “All this data that can show what’s the chance of lightning two minutes from now or what’s the chance of a tornado two minutes from now, an hour from now, and even four days from now.”
Access to decades of open historical data from the National Oceanic and Atmospheric Administration is key to making that possible, he added.
For veteran meteorologist Mike Nelson, the rise of AI is the latest step in decades of rapid change in forecasting technology, but not a replacement for the people behind it.
Even as models grow more sophisticated, Nelson hopes meteorologists remain at the center of the process, translating complex data into clear, actionable information for the public.
That role, he added, extends beyond day-to-day forecasts to helping viewers understand the difference between short-term weather and long-term climate trends.
“I was an advocate for this in the ’80s and the ’90s. I really began to push it harder with the birth of my grandchildren, knowing that for the impact it will have on many of us, it will be important — for my 13-, 10- and 8-year-old grandchildren it is going to be vitally important,” said Nelson, formerly with Denver 7 and 9News.
“We have to do this. This should be the top priority.”