The Earth-2 Weather Analytics Blueprint is a reference implementation of a geospatial data analysis service built from key components of NVIDIA's Earth-2 platform. It provides a starting point to build your own weather-related workflows accelerated by AI and Omniverse. The blueprint demonstrates three integrated components:
- NVIDIA Data Federation Mesh (DFM) for orchestrating distributed data processing workflows.
- NVIDIA Earth2Studio for running AI-driven weather inference workflows inside DFM pipelines.
- Earth-2 Command Center (E2CC), an NVIDIA Omniverse Kit application for high-fidelity visualization of geospatial data.
How the blueprint fits together: clients submit pipelines (graphs of operations) to a federation. The DFM runs them on sites and returns the results.
The Data Federation Mesh (DFM) is a programmable framework for orchestrating data processing across distributed sites. Each site is a group of services and resources in one location. DFM acts as "glue code as a service" as it coordinates where work runs and how data flows. It brings compute to the data by running pipeline steps close to where data lives to reduce latency, bandwidth, and cost, and helps keep data within desired security boundaries. Multiple sites form a federation and expose a single, coherent API so clients can submit pipelines without knowing the underlying topology.
This blueprint demonstrates two workflows for defining and submitting pipelines to the DFM:
- Jupyter notebook: For Python developers. Connect to DFM, build a pipeline from the federation’s operations API, execute it on DFM sites, and pull results back into the notebook for analysis.
- Earth-2 Command Center (E2CC): The same pattern from an Omniverse Kit app. Define and run pipelines from within a digital twin environment and visualize results on an interactive 3D globe.
In both cases, the client (notebook or E2CC) sends a pipeline to the federation, DFM runs it and returns results.
Pipeline operations on each site are implemented by adapters, which are plugin-like components that perform the actual work. In this blueprint, some of those adapters use the NVIDIA Earth2Studio toolkit under the hood. Earth2Studio is a comprehensive toolkit for AI weather inference workflows. It provides a unified API for weather data sources, NVIDIA's in-house weather models, as well as third-party models. When a client submits a weather pipeline, the site's adapters call into Earth2Studio to load data and run inference.
In practice, DFM adapters are much more general and not limited to weather. They can implement any operation a federation needs. This blueprint uses Earth2Studio because it focuses on AI weather workflows.
To keep the setup simple, this blueprint runs in DFM Proof of Concept (POC) mode, where all sites run on a single machine. This allows you to try the full flow without deploying a distributed federation.
In production, a federation can span multiple, independently managed sites on different machines, regions, clouds, or on-premises data centers. Each site administrator controls what their site offers and DFM orchestrates execution across sites, assigns operations to capable sites, and manages data flow. The DFM provides one API over a distributed, heterogeneous mesh while keeping data and control where each organization intends.
This repository is specifically designed for developers. Follow these steps to clone the repository:
git lfs install
git clone git@github.com:NVIDIA-Omniverse-blueprints/earth2-weather-analytics.git
cd earth2-weather-analyticsUse the blueprint guide to build and deploy the blueprint on your machine.
The Earth-2 Weather Analytics Blueprint is provided under the Omniverse License Agreement. Refer to LICENSE.md for the full license text.
The NVIDIA Earth-2 Weather Analytics Blueprint is shared as a reference and is provided "as is". The security of the production environment is the responsibility of the end users deploying it. When deploying in a production environment, have security experts review any potential risks and threats. In particular:
- Define the trust boundaries.
- Implement logging and monitoring capabilities.
- Secure the communication channels.
- Integrate AuthN and AuthZ with appropriate access controls.
- Keep the deployment up to date.
- Ensure the containers and source code are secure and free of known vulnerabilities.