Will your weather forecasts soon be done without humans?

A listener recently asked how Artificial Intelligence or AI is involved in weather forecasting. That is a great and timely question.

The field of AI is growing rapidly involving many areas such as web searches, automation, data analysis, and much more. Weather forecasts are yet another area where the use of AI is swiftly moving forward.

Last week, the National Oceanic and Atmospheric Administration (NOAA) Administrator Rick Spinrad gave a presentation on this topic at the Amazon AWS Reinvent conference in Las Vegas. The bottom line is that AI is playing a greater role in the assimilation of weather data at sea, on land, and up into space, and integrating that data in fine-tuning weather forecast guidance.

MyNorthwest Weather: Winds up to 100 mph in the Hanukkah Eve storm

He said forecasters have been able to make a 7-year advancement in forecast capability in a matter of months using AI.

A local example of the use of AI was during the recent bomb cyclone event. Data collected as the storm was rapidly developing well offshore was integrated into the forecast guidance called the High-Resolution Rapid Refresh, which is updated each hour.

The result was fine-tuning the short-term forecast information that produced greater accuracy for the location of wind speeds, particularly the strongest winds as the storm evolved. Power utilities, emergency managers, and many others used that enhanced weather information to better prepare and respond to the wind storm.

Administrator Spinrad highlighted NOAA’s future plans, including incorporating AI into more NOAA applications. He said at the end of the day, what we are going to do is improve lives, livelihoods and lifestyles.

History: Winds up to 100 mph in the Hanukkah Eve storm

In the recent journal of Nature, they reported on a new AI tool named GenCast from DeepMind, a Google company. For decades, weather forecasts have been limited in accuracy going further out in time, such as 10 or 15 days, thanks to the chaotic nature of the globe’s atmosphere. DeepMind appears to have cracked through that time-limit barrier.

GenCast used a huge amount of archived global weather data from 1979 to 2018 to train the system on the world’s weather behavior, including its interaction with complex terrain such as that found in the Pacific Northwest. The results thus far have found a greater accuracy at longer range of time such as 15 days, and generating forecast output more quickly than current technology.

Going forward, AI is going to play a far greater role in weather forecasts. Local forecasters will still be needed to issue weather and river flood warning information and provide interpretation for the public. Yet the use AI in the weather forecast process is going to provide greater socioeconomic benefits for the public in the years ahead.

Ted Buehner is the KIRO Newsradio meteorologist. You can read more of Ted’s stories here and follow him on X.