NVIDIA CorrDiff: Resolving Extreme Weather Events With Generative AI

Опубликовано 18 марта 2024, 22:05

NVIDIA's E2 Inference service is revolutionizing extreme weather forecasting by enabling users to combine global medium-range AI weather prediction models like NVIDIA’s FourCastNet with generative AI super-resolution / downscaling models like NVIDIA’s CorrDiff. This innovative approach enables the rapid exploration of numerous high-resolution weather forecast scenarios, providing decision-makers with a detailed understanding of a storm's potential impacts. By empowering proactive strategies, this groundbreaking technology aims to minimize the loss of life and property damage caused by extreme weather events.

Video Outline:

As the earth’s climate changes, AI-powered weather forecasting is allowing us to more accurately predict and track severe storms like Super-Typhoon Chanthu which caused widespread damage in Taiwan, the Philippines, China, and Japan in 2021. Current AI forecast models can accurately predict the track of storms, but they are limited to 25-kilometer resolution, which can miss important details. NVIDIA's CorrDiff is a revolutionary new generative model trained on high-resolution radar-assimilated WRF weather forecasts and ERA5 reanalysis data. Using CorrDiff, extreme events like Chanthu can be super-resolved from 25km to 2km resolution, with 1,000 times the speed and 3,000 times the energy efficiency of conventional weather models. By combining the speed and accuracy of NVIDIA's weather forecasting model FourCastNet and generative AI models like CorrDiff, we can explore hundreds or even thousands of kilometer-scale regional weather forecasts to provide a clear picture of the best, worst, and most likely impacts of a storm. This wealth of information can help minimize loss of life and property damage. Today, CorrDiff is optimized for Taiwan. But soon, generative super-sampling will be available as part of the NVIDIA Earth-2 Inference Service for many regions across the globe.

NVIDIA’s E2 Service consists of cloud-based building blocks enabling weather and climate practitioners to benefit from the latest advances in accelerated computing, AI, and visualization.

The platform consists of four independent services:

- Inference using AI-based models for global forecasting such as NVIDIA’s FourCastNet and regional downscaling using generative super-resolution models like NVIDIA’s CorrDiff

- Simulation: running GPU-accelerated process-based simulation in the cloud

#Earth2 #Inference #AI #Forecasting #Regional #CorrDiff #FourCastNet #DataFederation #Omniverse #E2 #SuperResolution #Downscaling #GTC24

- Data Federation: serving weather and climate data from different sources, including static archives, AI-based inference results, or numerical simulation results

- Visualization: cloud-based interactive visualization of weather/climate data

Key Points:

- Services of the E2 platform: inference, simulation, visualization, data federation.

- All services are optimized for NVIDIA AI Enterprise platform, running on heterogeneous clouds.

- Service APIs will be available after GTC which will enable ISVs to leverage these services within their offerings.

- We are working with the weather/climate community to bring their tools to the NV platform and make it accessible as services.