Sim
It’s a complex climate problem – and it would take a very powerful computer to be able to really model the Earth System and its future with any sensitivity.
I recently heard Mike Pritchard, director of research at Nvidia and professor at UC Irvine, talk about this process and how it works.
“The physics spans 10 orders of magnitude in space-time,” he said, citing research problems such as figuring out whether a cloud particle favors the gathering of water vapor around it.
“If you want to simulate the planet hundreds of times, test many ‘what if’ scenarios of the future, unfortunately, even with the most powerful supercomputers, you can’t handle all that complexity,” he said. “Meanwhile, humanity’s questions about the future climate are too broad for simulation technology.”
For a specific example of looking directly at the problems, he talked about his commute from San Diego to Irvine and seeing a certain kind of cloud outside the window.
“It looks like a gray band on the horizon,” he said. “We call it the marine layer. If it breaks out on a day at the beach, you’ll be upset because it makes you cold. But what matters is that it’s the tip of a huge sheet of low clouds that you’ll see outside your window from the plane, halfway through the flight from San Diego to Hawaii, and that cloud reflects a lot of energy off the planet , keeping it cooler than it would otherwise be. So if it dissolves … that will enhance global warming … but if it thickens, which it could, that will moderate it. And that’s a multi-trillion dollar uncertainty. And it’s a simulation problem. We know that these clouds need very high resolution to simulate that we cannot afford to develop yet in climate simulation.”
A Cavalcade of Systems
Pritchard also mentions the word “ensemble”, which is often used in machine learning to talk about using more than one model at a time or crowdsourcing the outputs of different LLMs, but has a different meaning in weather forecasting.
“You don’t predict one hurricane,” he said, “you predict hundreds of hurricanes. You hope for the best and plan for the worst… card-carrying atmosphere scientists at the University of Washington take these AI weather models that were trained on the chaos of the real atmosphere, which is very noisy and messy, and after the fact, looking at them and asking them if they have learned physics by doing such things.”
Pitchard talked about how this works with technology and built a record of evidence for the ability of AI models to help us predict the weather.
New Software and Technology
As an example, Pritchard mentions the ability of Nvidia AI tools such as Modulus, Earth2Studio that enable the research, development and validation of such AI predictive models.
The company, which has risen to the top of the pack on the US stock market, actually has several research streams and partnerships with the atmospheric science community. These are released on open source domain and here are some of the prominent models:
StormCast Research – demonstrates an artificial intelligence generation model that mimics atmospheric dynamics, examining mesoscale weather phenomena and making predictions (paper).
CorrDiff– this is another artificial intelligence generation model that creates high-resolution weather forecasts (paper). One can learn and explore more about AI downscaling pretrained Corrdiff here.
FourCastNet– this model achieves 25 km resolution weather forecast for places around the world with Spherical neural Fourier operatorsrecently calibrated for huge sets. One can learn and explore more about medium range global forecasts using pre-trained Fourcastnet here.
Earth-2 platform is a digital dual cloud platform that helps businesses leverage these AI advances and accelerate traditional numerical simulations to reduce the computational bottleneck of climate and weather simulations. By combining these advances with advances in computer graphics such as RTX rendering technology, we can create digital twins of Earth’s climate and weather to help scientists explore, analyze and explain the complexity of weather phenomena, especially in context of climate change.
More about Climate Work
Pritchard also talked about advancing optimal variance in large AI weather forecasts and referred to new work that brings more detail to the emerging science of using AI to simulate low-probability, high-impact climate extremes. This, he said, will give climate risk modelers new tools to help us understand and protect ourselves from extreme weather events.
Map – Source: (other options can easily be exported there)
Back and Forward
Here’s another aspect of what Pritchard talked about in terms of useful AI models. He described traditional climate IT processes as “going to an Oracle” – large simulators create large datasets, he suggested, which users then need to mine to help inform what-if scenarios and questions about the future climate. AI predictions, he added, can be run forward and backward, which will help users more easily understand what might have changed given a different initial input.
“We may be entering a future where we can understand our influence on the future more easily, without having to experience all the bottlenecks of conventional simulation,” he said.
The Power of Twins
In closing, Pritchard also talked about the concept of digital twinning as applied to the largest single object we have in our world – the world itself.
“I think the really important examples of interactivity are chains and cascades of AI digital twins,… So you can imagine a future that evolves towards climate AI digital twins, combined with extreme weather (event) AI digital twins. »
Nodding to current research and what everyone is doing around this very complex problem, Pritchard gives us food for thought about how to deal with the climate of our time with technology that goes far beyond big data sets. Stay tuned for more from the recent AI and planet events here in Boston.
