On July 4th, 2024, the European Center for Medium-Range Weather Forecasts (ECMWF), a top European weather forecasting agency predicted that Hurricane Beryl, which was in the Caribbean at the time, would landfall in Mexico. However, a new artificial intelligence program predicted that the hurricane would landfall in Texas using only information on the earth’s atmosphere.
The ECMWF uses planes, buoys, spacecrafts, and room-sized supercomputers to predict the weather. This includes, but is not limited to, hurricanes. However, in this case, the A.I. program is limited to only predicting hurricane paths, and the parent company, DeepMind, owned by Google, named this experimental program GraphCast.
Hurricane Beryl proved A.I. right when it, unfortunately, wrecked havoc in Texas, killing 36 people and stopping power from reaching millions of people. The A.I. program predicted the path of the hurricane more accurately than humans in minutes and seconds, versus the hours it had taken before.
The complexity of weather patterns makes it extremely hard to create weather forecasts as accurate as the technology can. Weather conditions depend on how the earth tilts, spins, and whether it is daytime or nighttime. All of these factors can contribute to wind, rain, air pressure, and temperature in a certain area.
Weather forecasting has slightly improved due to supercomputers which have been doing the actual predicting before A.I. forecasting programs existed. However, the supercomputers need skilled modelers to create a virtual planet with millions of voids that have to be filled with current weather data. Even then, it would take hours for the computer to process the information, and in the meantime, the weather conditions may have change and the data would have to be updated.
“It’s a turning point,” said Maria Molina, a research meteorologist at the University of Maryland who studies A.I. programs for extreme-event prediction. “You don’t need a supercomputer to generate a forecast. You can do it on your laptop, which makes the science more accessible.”
The A.I. weather programs’ predictions are primarily made possible by machine learning, in which A.I. mimics human learning. This works since A.I. recognizes patterns remarkably well. In weather forecasts, the A.I. programs learn about the atmospheric forces of the earth, picks out patterns, and uses that information to accurately predict the weather.
Rémi Lam, the lead scientist of the A.I. program called GraphCast, conducted a series of tests in which GraphCast’s prediction beat the European Center for Medium-Range Weather Forecasts best models in accuracy more than 90 percent of the time. “If you know where a cyclone is going, that’s quite important,” he said. “It’s important for saving lives.”
DeepMind, the parent company of Graphcast, tested the program in a landfall prediction of Hurricane Lee, which, in September [which year?], was possibly headed towards New England or Canada. They discovered that Graphcast predicted landfall in Nova Scotia, Canada three days before the supercomputers did.
Even though A.I. programs can much more accurately predict weather in a timely manner, supercomputers also have strengths, which can complement A.I. As Dr. Molina of the University of Maryland states, “All models are wrong to some extent. [A.I.] might get the hurricane track right but what about rain, maximum winds and storm surge? There’re so many diverse impacts [that need to be assessed carefully].”
Jamie Rhome, the deputy director of the National Hurricane Center in Miami, agrees with Dr. Molina on using several “tools”. According to him, A.I. is “evolutionary rather than revolutionary” and believes supercomputers and humans will still be crucial in the development of A.I. weather forecasting. “Having a human at the table to apply situational awareness is one of the reasons we have such good accuracy,” he said.
“With A.I. coming on so quickly, many people see the human role as diminishing,” Mr. Rhome added. “But our forecasters are making big contributions. There’s still very much a strong human role.”
Link to article: https://www.nytimes.com/interactive/2024/07/29/science/ai-weather-forecast-hurricane.html
The ECMWF uses planes, buoys, spacecrafts, and room-sized supercomputers to predict the weather. This includes, but is not limited to, hurricanes. However, in this case, the A.I. program is limited to only predicting hurricane paths, and the parent company, DeepMind, owned by Google, named this experimental program GraphCast.
Hurricane Beryl proved A.I. right when it, unfortunately, wrecked havoc in Texas, killing 36 people and stopping power from reaching millions of people. The A.I. program predicted the path of the hurricane more accurately than humans in minutes and seconds, versus the hours it had taken before.
The complexity of weather patterns makes it extremely hard to create weather forecasts as accurate as the technology can. Weather conditions depend on how the earth tilts, spins, and whether it is daytime or nighttime. All of these factors can contribute to wind, rain, air pressure, and temperature in a certain area.
Weather forecasting has slightly improved due to supercomputers which have been doing the actual predicting before A.I. forecasting programs existed. However, the supercomputers need skilled modelers to create a virtual planet with millions of voids that have to be filled with current weather data. Even then, it would take hours for the computer to process the information, and in the meantime, the weather conditions may have change and the data would have to be updated.
“It’s a turning point,” said Maria Molina, a research meteorologist at the University of Maryland who studies A.I. programs for extreme-event prediction. “You don’t need a supercomputer to generate a forecast. You can do it on your laptop, which makes the science more accessible.”
The A.I. weather programs’ predictions are primarily made possible by machine learning, in which A.I. mimics human learning. This works since A.I. recognizes patterns remarkably well. In weather forecasts, the A.I. programs learn about the atmospheric forces of the earth, picks out patterns, and uses that information to accurately predict the weather.
Rémi Lam, the lead scientist of the A.I. program called GraphCast, conducted a series of tests in which GraphCast’s prediction beat the European Center for Medium-Range Weather Forecasts best models in accuracy more than 90 percent of the time. “If you know where a cyclone is going, that’s quite important,” he said. “It’s important for saving lives.”
DeepMind, the parent company of Graphcast, tested the program in a landfall prediction of Hurricane Lee, which, in September [which year?], was possibly headed towards New England or Canada. They discovered that Graphcast predicted landfall in Nova Scotia, Canada three days before the supercomputers did.
Even though A.I. programs can much more accurately predict weather in a timely manner, supercomputers also have strengths, which can complement A.I. As Dr. Molina of the University of Maryland states, “All models are wrong to some extent. [A.I.] might get the hurricane track right but what about rain, maximum winds and storm surge? There’re so many diverse impacts [that need to be assessed carefully].”
Jamie Rhome, the deputy director of the National Hurricane Center in Miami, agrees with Dr. Molina on using several “tools”. According to him, A.I. is “evolutionary rather than revolutionary” and believes supercomputers and humans will still be crucial in the development of A.I. weather forecasting. “Having a human at the table to apply situational awareness is one of the reasons we have such good accuracy,” he said.
“With A.I. coming on so quickly, many people see the human role as diminishing,” Mr. Rhome added. “But our forecasters are making big contributions. There’s still very much a strong human role.”
Link to article: https://www.nytimes.com/interactive/2024/07/29/science/ai-weather-forecast-hurricane.html