AI has been consistently streamlining older workflows into cheaper, faster, and more efficient processes. Nvidia is now bringing that transformation to meteorology.
AI-powered weather forecasting promises to deliver predictions faster and more cheaply than traditional supercomputer-dependent models, potentially transforming who can generate custom forecasts, as the cost of entry is much lower, according to Nvidia. The advantage: more people can understand the planet’s atmospheric conditions and participate in weather and climate prediction.
As a result, it launched Earth-2, an open stack of AI models and tools, including inference libraries and frameworks, that allow developers to bring together separate weather and climate AI capabilities, such as generating atmospheric conditions and predicting weather weeks in advance, to accelerate all forecasting stages, according to Nvidia.
The new models launched include:
- Earth-3 Medium Range
- Earth-2 Nowcasting
- Earth-2 Global Data Assimilation
These models join:
- Earth-2 CorrDiff
- Earth-2 FourCastNet3
Earth-2 can be leveraged across the industry for AI weather tools, energy forecasting, and grid operations, as well as for financial risk firms. For example, Nvidia shared that Taiwan’s Central Weather Administration, The Weather Company, and S&P Global Energy are already using Earth-2.
Ultimately, Nvidia highlights that accurate weather forecasting can help decision makers across industries, including agriculture, energy, and public health, make better-informed decisions that can ultimately save lives.