AI Is Going To Replace The Weatherman
After decades of research, a rise in highly educated scientists, and evolving efforts to predict dangerous storms, the batting average of weather forecasts remains poor. According to the National Environmental Satellite, Data, and Information Service division of the NOAA, 10-day forecasts are only right half of the time. The figure is better for seven-day forecasts at 80%
Today’s weather forecast technology is based on three things. The NOAA’s Geostationary Environmental Operational Satellite-R (GOES-R), which operates 20,000 miles above the Earth. These follow the Earth’s orbit. The next is Polar-Orbiting Satellites, which orbit 500 miles above the surface and move in circles from pole to pole 14 times a day. Finally, NOAA’s Deep Space Climate Observatory (DSCOVR) rotates one million miles above the Earth, where they track the sun’s energy and “makes observations about ozone and aerosols in Earth’s atmosphere.”
Among the information these systems gather are data on factors that may cause hurricanes, wildfires, changes in air quality, and drought. These also drive weather forecast models, most often used by the Global Forecast System (GFS) in the U.S. and the European Centre for Medium-Range Weather Forecasts (ECMWF). The forecasts from these two are often different, though they can be close. The systems have notable weaknesses. They are poor at local forecasts. They are also poor at accurately predicting the actions and direction of powerful storms.
Meteorologists face an AI threat to their jobs, as is common in many other industries today. AI may do a better job at their jobs.
The threat comes from three megatech companies and a number of start-ups. The large ones include Google DeepMind WeatherNext 2, NVIDIA: Earth-2, and Microsoft: Aurora from its Azure AI Foundry Labs. Each claims it is better, mid-range (10 days to 15 days), and local forecasts. Another feature is the ability to forecast weather within a one-mile radius. Each says it is better than current methods at forecasting the growing strength and paths of the most powerful storms. Finally, each can offer these features at much higher speeds than humans can.
Among start-ups, most offer broad gains over current methods. Each claims it works much faster than humans. And some are aimed at niche markets. For example, Atmo focuses on military uses. Brightband is “making AI weather forecasting tools available to all, to help humanity adapt to increasingly extreme weather.” Some use the tools from Google. Several are backed by the world’s largest venture capitalists. And most rely on endorsements from major universities and think tanks. The University of Chicago Institute for Climate and Sustainable Growth recently wrote, “Artificial intelligence models can produce weather forecasts up to 100,000 times faster than traditional systems.” If true, the effects on the industry will be staggering. And. Most of the human work will be unnecessary.
There is already a long list of industries that will be downsized because of AI. These range from law firms and investment banking to clerks at major retailers’ stores. It appears the weather forecasting industry will be, or is already, on the list
