Nvidia unveils AI weather forecasting advance

• AI-based forecasting is more efficient, can be done much faster, and generally doesn't require supercomputers.

Driving the news: A study on the model, known as "NVIDIA CorrDiff," and its capabilities was published Monday in the journal Communications Earth & Environment.

• The company says the model can capture extreme weather events, including ones that depart from historical records. This is important given climate change-related extreme weather trends.
• Nvidia claims CorrDiff is 500 times faster and up to 10,000 times more energy-efficient than traditional high-resolution weather prediction models.

The big picture: The model is one in a series of AI forecasting advances during the past few years that are poised to remake how weather and climate hazards are predicted.

• Nvidia may be best known for its chip-making business. But it, along with other companies such as Microsoft and Google, have used the data-heavy challenge of predicting weather to explore and prove the superiority of artificial intelligence.
• Traditional forecast methods rely on slower, expensive and energy-intensive physics-based computer models. To generate forecasts, these models use hundreds of equations describing how the air, land and ocean function.
• Such calculations are done using supercomputers and take hours to run, limiting the number of model projections available each day.

AI forecasting methods train models based on how the atmosphere behaved when certain data parameters were present.

• The models gradually simulate the physics of the atmosphere through learning from observational data, AI forecasting experts have told Axios.

The intrigue: The new model is a generative AI model that allows for local-level forecasts of wind, temperature and precipitation type and amounts, the company said in a statement.

• The model takes coarse weather data from another simulation and zooms in on specific regions to reveal new details through generative AI, in a new way of conducting a process known as "downscaling."

What they're saying: Mike Pritchard, Nvidia's head of climate simulation research, described the model as "a significant development."

• "It proves that generative AI models have a very powerful property — they can simultaneously super-resolve multiple weather variables and synthesize new channels from information only indirectly related, received at much coarser spatial resolution," he told Axios via email.
• Pritchard said the new model learns to forecast directly from the data, rather than via physics equations.

Zoom out: Forecasters at the National Weather Service, in the media and elsewhere increasingly are using AI models to augment traditional, physics-based forecast tools.

• This helps them learn their biases and capabilities as the science develops.
• The European Centre for Medium-Range Weather Forecasting is now running its own AI model, while Google and other tech companies have made theirs available too.
• Most weather experts see AI as adding to and improving current tools used for forecasting, rather than replacing present-day models and human forecasters.

Yes, but: Such expectations could change as AI-based forecasting rapidly evolves and could yield some surprise breakthroughs.