Google has presented a new artificial intelligence model that improves the prediction of tropical cyclones, one of the most dangerous meteorological phenomena that exist. In the United States alone, hurricanes and typhons have charged hundreds of thousands of lives in the last 50 years and have caused economic losses worth 1.4 billion dollars.
Collaboratively developed between Google Deepmind and Google Research, this work includes an experimental model for a better prognosis of the arrival of cyclones through AI, as well as an interactive web page where Google's meteorological models can be visualized. Within the initiative, an association with the National Hurricane Center for the United States is also framed.
Experimental model against typhoons and hurricanes
Cyclones prediction is essential to prepare the population and minimize damage, but it is a challenge, since it is difficult to be precise. Typhones and hurricanes are very sensitive to the slightest differences in atmospheric conditions, and the data are scarce, so it is not always easy to succeed in prediction.
However, thanks to the use of AI, Google has created an experimental model that improves the ability to forecast the arrival and behavior of cyclones, which can result in considerable improvement in communities' safety.
This AI model is able to predict, with precision, the formation, trajectory, intensity and size of a cyclone. Google AI generates up to 50 possible scenarios up to 15 days in advance, which allows you to prepare for the different scenarios and take the appropriate prevention measures.
According to the Cooperative Institute for Atmosphere Research (CIRA), the Google model shows a comparable ability or greater than the best models based on operational physics on the trajectory and intensity of cyclones.
Interactive website
As part of this project, Google has also launched a public prior view of a new interactive website called Weather Lab. It shows the prediction of live cyclones, as well as historical evolutions of different meteorological models of AI.
To provide this information, both the experimental model of Google cyclones, Weathernext Graph Weathernext Gen, together with models based on physics from the European Center for Medium -term weather forecasts (CEPMPM) are used.
Collaboration with the National Hurricane Center
In addition, Google has also established a collaboration with the National Hurricane Center (NHC), which evaluates the risk of cyclones in the Atlantic and Eastern Pacific basins, to scientifically validate the approach and results of its experimental model of AI.
NHC meteorologists already use Google's experimental models for the prediction of hurricanes and typhoons, along with other models and observations based on physics.
Google has also collaborated with CIRA, the United Kingdom's weather service, the University of Tokyo, Weathernews Inc. of Japan and other experts to improve their predictive cyclones models.
Leave a Reply