Nvidia is known for its graphics cards and virtual worlds. But their latest project is intended for the real world and aims to better prepare us for climate catastrophes
With their Earth 2 API, organizations and governments around the world will be able to create fast and accurate weather forecasts. As part of their newly announced CUDA-X microservices, the new cloud API will enable virtually any user to create so-called AI-powered emulations
Nvidia announced this in a press release on March 18, 2024. The aim is to mitigate the global damage caused by extreme weather events due to climate change.
Nvidia’s technology as a global weatherman against disasters
Climate disasters are now normal – historic droughts, catastrophic hurricanes and intergenerational flooding appear in the news with alarming frequency. The Earth-2 Cloud APIs are designed to help us better prepare for extreme weather events – and inspire us to act to mitigate them.
This is what Jensen Huang, founder and CEO of NVIDIA, said in the above press release We are talking about AI-supported simulations that users can feed with the relevant weather data.
How it works: According to Nvidia, its AI is intended to accelerate the provision of interactive, high-resolution simulations. This includes, for example, the atmosphere, local cloud cover, typhoons and turbulence
Governments or organizations around the world can, in simple terms, query location-specific weather data thanks to a generative AI model called CorrDiff , which creates a physical simulation in the form of images from the fed weather data.
This is the difference to previous forecasts according to Nvidia:
The advantage of an AI-supported weather forecast: According to Nvidia, the AI is able to generate 12.5 times higher resolution images. It also works 1000 times faster and 3000 times more energy-efficient than previous, conventional models
According to Nvidia, the AI model has learned to combine several sources of information and thus make more precise predictions. Nvidia writes about this:
In combination with proprietary data from companies in the $20 billion climate engineering industry, Earth-2 application programming interfaces help users deliver alerts and updated forecasts in seconds, compared to minutes or hours with traditional CPU-driven modeling.
Earth-2 in detail: Earth-2 includes the physical simulation of numerical models such as ICON and IFS, neural network models such as FourCastNet, GraphCast and Deep Learning Weather Prediction (DLWP) via NVIDIA Modulus as well as data fusion and visualization with NVIDIA Omniverse.
The technology behind it: Earth-2 runs on NVIDIA DGX GH200, HGX H100 and OVX super computers.
Taiwan is already planning permanent use
The Central Weather Bureau of Taiwan is already planning to use the Earth-2 Cloud API. The country is repeatedly hit by strong typhoons and hopes to be able to evacuate earlier and better.
Since 2000, Taiwan has been hit by a total of 136 typhoons. The use of AI-supported forecasting therefore makes perfect sense here. It remains to be seen how smoothly the global implementation will work. In principle, however, it is a sensible idea to use artificial intelligence to better assess the weather situation.
