Google wants you to know that artificial intelligence is more than just a conversational chatbot. The DeepMind subsidiary unveiled a model in September that can identify genetic mutations that cause disease. Now he has created a tool that can predict where a hurricane will make landfall much earlier than existing systems.
Scientists involved in the project say this is a revolutionary invention. And according to a study published Tuesday in the journal The science, not surprising. GraphCast, a new model developed by Google DeepMind, was able to predict weather conditions up to 10 days in advancewith greater accuracy and much faster than the current gold standard.
GraphCast, for example, outperformed the European Center for Medium-Range Weather Forecasts (ECMWF) model by more than 90% of 1380 test sites. In forecasts for Earth’s troposphere, the lowest part of the atmosphere where most weather events occur, GraphCast outperformed the ECMWF model. in more than 99% of climate variables. Among them are rain and air temperature.
ECMWF has one of the largest supercomputing facilities and meteorological data archives in the world. It supports the work of programs such as Copernicus, funded by the European Union and a key source of tracking climate change. Therefore, for a tool to be presented as the best is a lot.
How do modern systems work?
Google’s DeepMind tool was able to predict where Hurricane Lee, a powerful event recorded in September, would hit in Canada. three days earlier than existing methods. The ability to warn earlier gives authorities and the public a key moment to be better prepared. Critical time to save lives.
A study published in October explains that Atlantic hurricanes are now more than twice as likely to intensify rapidly. Climate change is to blame. For example, Hurricane Lee went from Category 1 (with winds of 129 kilometers per hour) to Category 5 (249 kilometers per hour) in less than 24 hours. Therefore, it is extremely important to save time.
Traditional weather forecasting is based on real-time measurements of what is happening in the atmosphere. In the best cases, as the ECMWF team did, these measurements come from different parts of the world and from different instruments: satellites, buoys in the ocean, sensors on commercial aircraft.
ECMRWF’s Matthew Chantry reports. BBC that one of his predictions could account for about 10 million measurements. All this large amount of data is processed on some supercomputers to solve complex equations based on physics and various climate variables. Those located in this European center can perform up to a thousand billion calculations per second. And thus knowing what the probability is that a certain event will happen in the future.
However, such models require large computing resources. And, even with all their power, it can sometimes take them hours to dictate their predictions. Artificial intelligence speeds up analysis even with less energy.
Google tool’s contribution to hurricane forecasting
GraphCast uses machine learning to perform these calculations in less than a minute. Instead of physics-based equations, it uses four decades of historical weather data to forecast hurricanes and other events much faster.
Google’s DeepMind tool uses graph neural networks that They map the Earth’s surface at more than a million grid points. At each point, the model predicts temperature, wind speed and direction. Also average sea level pressure, humidity and other variables. In doing so, the neural network identifies patterns and predicts what will happen for each of these data points.
“GraphCast uses decades of historical weather data to learn the pattern of cause-and-effect relationships that drive the evolution of Earth’s climate from the present to the future,” the company explained in a blog post. Google DeepMind tool predicts hurricane path Lee was nine days early, and ECMRWF achieved it six days early.
But this tool does not replace the measurements of centers such as ECMWF, but rather complements them. “They go hand in hand,” Google DeepMind said. In fact, one of the important inputs that GraphCast uses is the measurements of that center. And the ECMWF team is already taking advantage of the new artificial intelligence system.
Peter Duben, head of Earth system modeling at ECMWF, said it was like Christmas when they introduced GraphCast to his team. “It showed that these models are so good that we can no longer avoid them.”said MIT Technology Review. Google DeepMind says it wants to do more than just predict climate: “By developing new tools and accelerating research, we hope artificial intelligence will help the global community solve our biggest environmental challenges.”
Source: Hiper Textual
