Nvidia ‘Climate in a Bottle' opens a view into Earth's future. What will we do with it?

Mint

a day ago

Nvidia has unveiled a new generative foundation model that it says enables simulations of Earth's global climate with an unprecedented level of resolution. As is so often the case with powerful new technology, however, the question is what else humans will do with it.
The company expects that climate researchers will build on top of its new AI-powered model to make climate predictions that focus on five-kilometer areas. Previous leading-edge global climate models typically don't drill below 25 to 100 kilometers.
Researchers using the new model may be able to predict conditions decades into the future with a new level of precision, providing information that could help efforts to mitigate climate change or its effects.
A 5-kilometer resolution may help capture vertical movements of air in the lower atmosphere that can lead to certain kinds of thunderstorms, for example, and that might be missed with other models.
And to the extent that high-resolution near-term forecasts are more accurate, the accuracy of longer-term climate forecasts will improve in turn, because the accuracy of such predictions compounds over time.
The model, branded by Nvidia as cBottle for 'Climate in a Bottle," compresses the scale of Earth observation data 3,000 times and transforms it into ultra-high-resolution, queryable and interactive climate simulations, according to Dion Harris, senior director of high-performance computing and AI factory solutions at Nvidia. It was trained on high-resolution physical climate simulations and estimates of observed atmospheric states over the past 50 years.
It will take years, of course, to know just how accurate the model's long-term predictions turn out to be.
Nvidia says it has run tests on historic data to confirm that cBottle would have predicted the climate that actually eventually followed. The model also has a promising pedigree as part of Nvidia's Earth-2 platform, which is used to create digital twins of the planet. Nvidia introduced it four years ago to advance weather forecasting and climate science.
Scientific research institutions and policymakers, including the Alan Turing Institute of AI and the Max Planck Institute of Meteorology, are actively exploring the new model, Nvidia said Tuesday at the ISC 2025 computing conference in Hamburg.
Bjorn Stevens, director of the Planck Institute, said it 'represents a transformative leap in our ability to understand, predict and adapt to the world around us."
'By harnessing Nvidia's advanced AI and accelerated computing, we're building a digital twin of the planet," Stevens added, 'marking a new era where climate science becomes accessible and actionable for all, enabling informed decisions that safeguard our collective future."
It's easy to imagine other decisions based on cBottle simulations, though, that would be more parochial in nature.
Will home insurance companies withdraw from additional markets now, for example, because simulations using cBottle predict growing flood or fire risks 10 years out? What happens to property values and property tax revenue in areas like those?
And what might governments of the world do in response to climate forecasts on a five-kilometer scale years in the future?
Predicting conditions that could lead to food or water shortages might enable officials and residents to better prepare. But great powers' jockeying over the Arctic Circle might equally intensify if they believe they know exactly where and how many new sea lanes will open.
The Earth-2 platform is in various states of deployment at weather agencies from National Oceanic and Atmospheric Administration in the U.S. to G42, an Abu Dhabi-based holding company focused on AI, and the National Science and Technology Center for Disaster Reduction in Taiwan.
Spire Global, a provider of data analytics in areas such as climate and global security, has used Earth-2 to help improve its weather forecasts by three orders of magnitude with regards to speed and cost over the last three or four years, according to Peter Platzer, co-founder and executive chairman. The company, which gathers data from its fleet of low-Earth-orbit satellites, built its own technology on top of Earth-2, he said. It was an early adopter of Earth-2.
Tasks that once required eight hours to complete can now be executed in three minutes or less, Platzer added. 'The dramatic acceleration of processing power is the massive thing," he said.
Along with time, money has been the other major constraint on more-powerful forecasts. The cost of running a high-resolution weather simulation every hour for a full year is $3 million if you do it the traditional way, on CPUs, according to Nvidia. The company said its CorrDiff generative AI model running on its GPUs can perform that task for $60,000.
For Platzer, the savings mean weather predictions of 15 or 45 days went from being not useful to valuable, according to Platzer. That is 'far beyond what was previously thought possible," he said.
The new models still don't provide 100% certainty. They don't state that the weather or climate will be one thing or another, but rather provide probabilities for certain outcomes.
If you watched those percentage-based predictions about who would win the 2016 presidential election between Hillary Clinton and Donald Trump, you know what that means. Even a very high chance isn't the same as a sure thing.
That uncertainty might give Climate in a Bottle-based model users appropriate humility and caution when considering how to react to a projection about three decades in the future.
Or it might lead them to make a decision with big implications and costs, emboldened by the imprimatur of advanced AI, only to have bet on an outcome that never arrives.
Write to Steven Rosenbush at [e-mail]