This week’s global NVIDIA GTC, GPU Technology Conference, included a discussion about using accelerated computing to drive environmental sustainability initiatives and climate action.
During his keynote address, Jensen Huang, the company’s founder and CEO, said, “The generative AI revolution is here. But what else can we generate? What else can we learn? One of the things we would love to learn is climate. We would love to learn extreme weather. We would love to learn how we can predict future weather at regional scales, at sufficiently high resolution such that we can keep people out of harm’s way before harm comes.”
To do this, NVIDIA has developed the Earth-2 platform, which consists of four independent services. AI-based machine learning models like NVIDIA’s FourCastNet and generative super-resolution models like NIVIDIA’s CorrDiff. Cloud-based GPU-accelerated simulation. Serving weather and climate data from different sources, including static archives, AI-based inference results, or numerical simulation results. And, cloud-based interactive visualization of weather and climate data.
Current AI weather forecast models are limited to 25-km/15.5-mile resolution. CorrDiff reduces that resolution to 2 km/1.2 miles at 1,000 times the speed and 3,000 times the energy efficiency of conventional models. Combining CorrDiff’s generative AI with the speed and accuracy of FourCastNet, hundreds and potentially thousands of kilometer-scale regional weather forecasts can be explored to provide a clear picture of the impact of extreme weather events.
While Earth-2’s new cloud application programming interfaces (API) allow virtually any user to create AI-powered, interactive, high-resolution weather simulations, three startup companies, in particular, have taken advantage of the platform’s potential to pioneer climate AI advancements.
Tomorrow.io: Accurate and actionable weather insights
Boston-based Tomorrow.io’s constellation of high-tech low-Earth-orbit satellites, radars and other sensors capture high-accuracy global weather measurements. AI and machine learning models are applied to the dataset to translate it into actionable, weather-related insights for countries, businesses, and individuals.
Tomorrow.io uses Earth-2 AI forecast models to conduct observing-system simulation experiments, OSSEs, to identify the optimal configurations of their satellites and other instruments to improve weather forecasting.
ClimaSens: Extreme weather risk and response
ClimaSens, based in Melbourne, Australia, and New York, uses advanced AI models to combine historical, real-time, and future climate and weather information. Gathering petabytes of global climate, weather and geospatial data, ClimaSens provides in-depth, multi-hazard climate risk assessment and insights tailored to help regional and local communities plan, prepare and respond to climate extremes.
Its upcoming flood risk analysis model, FloodSens, which is currently in beta, was developed using Earth-2 APIs and FourCastNet to generate physically accurate representations of future weather conditions and models for assessing the probabilities of low-likelihood, high-impact flooding events.
North.io: Oceans of data
Helping to map the world’s largest carbon sink – oceans – using data-collecting autonomous underwater vehicles (AUV), German company North.io uses Earth-2 APIs to develop AI weather forecasts for operational planning, system management and risk assessment for its underwater fleet. Combining precise weather modeling with autonomous systems reduces safety risks for those operating in rough offshore environments.
The huge amounts of data collected by the AUVs are analyzed before being made easily accessible, shareable, and understandable via North.io’s TrueOcean platform.
Another key component of Earth-2 cloud APIs is NVIDIA Omniverse, a platform that enables the development of Universal Scene Description (OpenUSD)-based 3D workflows and applications. The Weather Company, a global leader in weather data forecasting and insights, plans to integrate its meteorological data and Weatherverse tools with Omniverse, enabling customers to better understand and visualize the impact of weather conditions for the first time.
Climate action is just one aspect of NVIDIA’s goal of accelerating computing in a sustainable way that maximizes energy efficiency and minimizes environmental impact. Jensen Huang’s full GTC keynote speech can be viewed in the video below.
Sources: NVIDIA, NVIDIA Newsroom
