Overview
Explore the cutting edge of global weather and cyclone forecasting with WeatherNext 2 (WN2), a powerful AI model developed by Google DeepMind and Google Research. This repository provides the code and pretrained models for WN2, building upon the foundations of its predecessors, GraphCast and GenCast. WeatherNext 2 delivers medium-range atmospheric forecasts with remarkable accuracy and efficiency, alongside specialized capabilities for tropical cyclone prediction.
Key Features
- Advanced Global Forecasting: WeatherNext 2 provides high-resolution medium-range weather predictions, enabling a deeper understanding of atmospheric dynamics.
- Cyclone Forecasting Breakthrough: The dedicated WeatherNext Cyclones models achieve state-of-the-art accuracy in forecasting tropical cyclone tracks and intensity.
- Operational Models: Access operational versions of WeatherNext 2 and WeatherNext Cyclones, trained on extensive datasets up to 2024, ready for real-world application.
- Flexible Deployment: Choose from various model resolutions, including high-resolution 0.25° (approx. 30km) for detailed forecasts and a lightweight 1° resolution "Mini" version for resource-constrained environments.
- Easy Access to Data: Directly consume daily WN2 model outputs through Google Cloud (Earth Engine, BigQuery, Vertex AI), WeatherLab, or the OpenMeteo API.
- Reproducible Research: The repository hosts code for prior models like GraphCast and GenCast, facilitating research and development.
- Comprehensive Documentation: Detailed guides, technical reports, and blog posts are available to help you understand and utilize the WeatherNext models.
Typical Use Cases
- Meteorological Research: Advance scientific understanding of atmospheric phenomena and improve forecasting methodologies.
- Disaster Preparedness: Enhance early warning systems for extreme weather events, particularly tropical cyclones.
- Climate Impact Analysis: Integrate high-fidelity weather forecasts into climate modeling and impact assessments.
- Operational Weather Services: Supplement existing numerical weather prediction (NWP) models with AI-driven insights for improved accuracy and speed.
- Academic Exploration: Experiment with state-of-the-art AI weather models for educational purposes and novel applications.
