Open-source machine learning for weather nowcasting¶
MLCast is a collaborative effort bringing together meteorological services, research institutions, and academia across Europe to develop a unified Python package for AI-based nowcasting. It is an initiative of the E-AI WG6 (Nowcasting) of EUMETNET.
We connect the world of computer science and machine learning with meteorology and weather forecasting — particularly nowcasting, the short-range prediction of precipitation from recent radar observations.
The main Python package — ConvGRU ensemble nowcasting models, a configuration system, and a CLI / Python API for training.
A curated Intake catalog of mlcast-compliant source radar datasets, plus converters and a training-data sampler.
A tool that checks contributed datasets against the MLCast Zarr specification and common geospatial tools.
Our mission¶
We are an open-source community dedicated to advancing machine learning for weather nowcasting through collaborative development and shared resources.
Our collective gives institutes with GPU access the opportunity to train models, while those with limited resources can contribute to development, testing, and ultimately benefit from using these models.
Through equal ownership and transparent collaboration, we’re creating a unified Python package that makes state-of-the-art nowcasting accessible to everyone.
Get started¶
🚀 Get Started — install the package and run your first training
📖 Documentation — overview, architecture, and API reference
🔗 Quick Links — fast access to repos, datasets, and references
💻 GitHub — all MLCast community repositories
License¶
mlcast is dual-licensed under Apache-2.0 or BSD-3-Clause, at your option.