Scientists from Google’s artificial intelligence arm, Google DeepMind, announced this week the creation of a new model that is much more accurate and faster when forecasting the weather than traditional methods, GraphCast. In one minute, the tool forecasts the weather for the next 10 days, based on meteorological data collected in real time.
In this scenario, Google DeepMind’s AI model represents a huge leap forward in weather forecasting, which is fundamental to our society. Going beyond the individual sphere, such as whether or not to use an umbrella for work, understanding climate potential is essential in agriculture and helps authorities prepare against extreme events, such as storms, hurricanes and floods.
Google artificial intelligence for weather
In a study published in the magazine Sciences, the scientists responsible highlight the level of accuracy of the new tool. Through testing, the authors concluded that “GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1,380 survey targets.”
In fact, the model was more accurate than the world’s main traditional three- to ten-day weather forecasting system, the High Resolution Forecasting System (HRES), adopted by the European Center for Medium-Range Weather Forecasts (ECMWF).
The interesting thing is that GraphCast would not exist without HRES, since the basis of training was the forecasts made and climate data acquired by ECMWF over the last 40 years.
Fastest weather forecast
One of the most notable features of Google DeepMind’s AI model is that it delivers weather forecasts in less than 60 seconds, using a single computer. This is unthinkable with current methods.
For example, ECMWF weather forecasts require the use of supercomputers – plural – that process equations based on scientific knowledge of atmospheric physics. The whole process can take hours and requires a lot of energy.
Forecasting extreme weather events
Additionally, GraphCast can provide early warning of severe weather events, such as tornado formation and flood risk. It is also able to predict the occurrence of heat waves like the one affecting Brazil right now. This can help save lives by increasing the time between when one of these events occurs and when you prepare for it.
“Coffee trailblazer. Social media ninja. Unapologetic web guru. Friendly music fan. Alcohol fanatic.”