Google develops AI model for forecasting tropical cyclones – News

Google LLC today detailed an artificial intelligence model that can forecast the path and intensity of tropical cyclones days in advance.

According to the company, the algorithm was developed through a collaboration between its Google Research and DeepMind units. It’s available through a newly launched website called Weather Lab.

Usually, scientists rely on algorithms known as physics-based weather prediction models to forecast cyclones. Such algorithms generate projections by running a large number of data points through differential equations. According to Google, this approach has certain limitations that can complicate cyclone forecasting efforts. 

“In physics-based cyclone prediction, the approximations required to meet operational demands mean it’s difficult for a single model to excel at predicting both a cyclone’s track and its intensity,” the company’s researchers explained in a blog post. “This is because a cyclone’s track is governed by vast atmospheric steering currents, whereas a cyclone’s intensity depends on complex turbulent processes within and around its compact core.”

According to Google, its newly detailed AI model doesn’t share that limitation. The company claims that the algorithm can predict both the track and intensity of a cyclone with “state-of-the-art accuracy.” Moreover, its forecasts encompass other details such as the manner in which a cyclone forms, its size and shape.

Google trained the model using two datasets. The first described the path, intensity and other key properties of nearly 5,000 cyclones from the past 45 years. The other dataset included information about past weather conditions that was distilled from millions of observations.

During internal tests, Google successfully used the algorithm to predict the paths of four recent cyclones. For two of the storms, the model generated accurate forecasts nearly a week ahead of time. It can predict storms up to 15 days in advance by “generating 50 possible scenarios,” Google’s researchers wrote.

The company has made the AI accessible to researchers through a new website called Weather Lab. The model is available alongside two years’ worth of historical forecasts, as well as data from traditional physics-based weather prediction algorithms.

“Weather Lab users can explore and compare the predictions from various AI and physics-based models,” the researchers detailed. “When read together, these predictions can help weather agencies and emergency service experts better anticipate a cyclone’s path and intensity.”

The rollout of Google’s new cyclone prediction algorithm comes two days after Nvidia Corp. released a foundation model called cBottle. According to the chipmaker, the AI can simulate global climate with kilometer resolution. It has already been adopted by multiple research institutions including the Max-Planck-Institute for Meteorology and the Allen Institute for AI.

Image: Google