This repository is a simple, human-readable catalog of publicly available forecast models and the observational data that feeds them.
Its primary focus is numerical weather prediction (NWP) and related atmospheric forecast systems whose data can be freely downloaded by anyone. It also includes a smaller catalog of operational observation systems with openly distributed raw data — weather satellites and radar networks — since these are foundational inputs to the models documented here.
The goal is to make it easy to discover:
- which forecast models exist
- who produces them
- what regions they cover
- where their data can be downloaded
- and which weather satellites and radar networks provide raw data that can be freely accessed
No meteorology background is required to use this repository.
- Numerical Weather Prediction (NWP) models (deterministic)
- Ensemble forecast models (global and regional)
- Nowcasting systems (0–6 h prediction by observation extrapolation, seamless extrapolation–NWP blending, or machine learning)
- Wave forecast models (deterministic and ensemble)
- Ocean physics models (temperature, salinity, currents, sea level, sea ice — distinct from wave models)
- Storm surge and coastal water level models (deterministic and ensemble)
- Tropical cyclone / hurricane models
- Air quality and atmospheric composition models (deterministic and ensemble)
- Long-range forecast systems (sub-seasonal, seasonal, and interannual prediction)
- Hydrological models (
models/hydrological_models/) — river discharge, runoff, and flood forecast systems. These carry two geometries: a land surface grid where water is generated and a routing network where it moves. Reach-based vector output on a river network is admitted as a documented exception to the gridded-data rule, paralleling the station time series exception for storm surge. - Fire danger models (
models/fire_danger_models/) — fire danger index and fire behaviour forecast systems. Named for danger ratings rather than fire weather, because the outputs are indices computed from weather rather than weather fields, and because smoke dispersion models live undermodels/air_quality_models/. - Space weather models (
models/space_weather_models/) — Earth-referenced upper atmosphere forecast systems: ionosphere, thermosphere, aurora, and ground-level geomagnetic and geoelectric effects. Heliospheric models are out of scope: they grid a volume in heliocentric coordinates in which Earth is a single moving point, and produce no terrestrial grid. - AI-based and hybrid physics–AI forecast systems
- Operational weather satellites with openly distributed raw data (Level 1 calibrated radiances, Level 2 retrievals)
- Operational weather radar networks and composites with openly distributed gridded data (e.g., reflectivity mosaics and radar-derived precipitation estimation); rendered radar imagery and viewer-only web loops remain out of scope
- Free and publicly downloadable data, including:
- Data with no account required (e.g., NOAA NOMADS, AWS Open Data buckets)
- Data requiring free registration or a free API key (e.g., EUMETSAT Data Store, KNMI Data Platform, Météo-France public API)
- Data distributed under free open licenses (e.g., Creative Commons Attribution 4.0, Etalab Open Licence, Copernicus Marine Service licence, NOAA public domain)
- Models provided by government, publicly funded, or military institutions that distribute output openly
- Global, regional, coastal, and storm-centered forecast systems
- Experimental and pre-operational systems that are publicly distributed by their operating agency. These are clearly labelled as experimental in their entries (see
STATUS.mdfor the current list)
- Models, satellite data, or radar data behind paywalls, commercial licensing, or usage-fee structures
- Data that requires approval gates, vetted-user programs, or commercial-use restrictions beyond simple registration
- Models or observation systems that only provide maps or viewers without downloadable data — including nowcasting products distributed only as web radar loops, and rendered radar imagery or tile services (e.g. RIDGE-style viewers) where no gridded feed is offered
- Research-only or one-off model runs that are not distributed by an operating agency
- Climate reanalysis and other historical-only datasets (e.g., ERA5, CFSR) — note that forward-looking long-range forecast systems (sub-seasonal, seasonal, and interannual prediction) are in scope and are documented under
models/climate_models/; only reanalyses and historical-only products are excluded - Commercial weather satellites and Earth observation missions outside the operational weather/atmospheric scope
- Sun-referenced space weather models — heliospheric systems such as WSA-Enlil grid a spherical volume in heliocentric coordinates from roughly 0.1 to 2 AU, in which Earth is a single moving point. Only Earth-referenced terrestrial-effect models are documented, under
models/space_weather_models/
A note on free registration: some operators (notably EUMETSAT and KNMI) require a free account to download data. These are not paywalls — there is no fee, no approval gate, and no commercial-use restriction. Entries clearly note when registration is required so users know what to expect.
A note on experimental systems: some entries document systems their operators explicitly label as experimental. These are included when the operating agency distributes the data through an operational or near-operational channel (e.g., ECCC's MSC datamart, NOAA NOMADS). Pure research code, one-off runs, and systems without ongoing public distribution are not included.
Forecast models are organized by:
- Model type (weather, ensemble, nowcasting, wave, storm surge, ocean, tropical cyclone, air quality, hydrological, fire danger, space weather, long-range)
- Geographic scope (global vs regional)
- Country or organization of origin
Ensembles are filed by phenomenon, not by ensemble status. Wave, storm surge, and air quality ensembles live alongside their deterministic siblings under
models/wave_models/,models/storm_surge_models/, andmodels/air_quality_models/, cross-linked in both directions.models/ensemble_models/is reserved for ensembles of the atmospheric state itself — GEFS, IFS ENS, ICON-EPS and similar.
Seasonal forecasts are filed by phenomenon too, not by forecast range. A seasonal river discharge or fire danger forecast is an impact model driven by a long-range atmospheric system, standing in the same relation to that system as a wave model does to its driving NWP model. These live under
models/hydrological_models/andmodels/fire_danger_models/beside their medium-range siblings, cross-linked upstream to the driving system's entry.models/climate_models/is reserved for long-range forecasts of the atmospheric or coupled state itself, including downscalings of it.
Each model has its own file describing:
-
what the model is
-
who runs it
-
what area it covers
-
basic resolution and update information
-
where to download the data
-
Satellites (
observations/satellites/) — organized by orbit type (geostationary vs polar-orbiting) and country or organization of origin. Each entry covers a constellation or series (e.g., GOES-R Series rather than separate GOES-16 and GOES-19 entries). -
Radar (
observations/radar/) — operational weather-radar networks and composites, organized by country or organization of origin. Each entry covers a network or composite product (e.g., MRMS, OPERA) rather than individual radars.
These observation entries follow the same general format as the model entries. See observations/satellites/README.md for the index of documented satellite systems.
In addition to the directory structure, several index files make specific slices easier to discover:
AI_MODELS.md— index of all AI-based and hybrid physics–AI forecast systems documented in the repository, organized by how AI is used (standalone deterministic, standalone ensemble, hybrid, input-consumer)STATUS.md— tracker for upcoming implementations, scheduled retirements, experimental systems, version upgrades, and format changesCOPERNICUS.md— index of all products distributed through the Copernicus programme, spanning the Marine Service, Atmosphere Monitoring Service, Climate Change Service, and Emergency Management Service, organized by operator and coverageUFS.md— index of all NOAA models that are part of, being consolidated into, or being retired by the Unified Forecast System, with the full programme context and rollout sequencingobservations/satellites/README.md— index of operational weather satellites with openly downloadable raw data, with quick-reference tables for access tiersobservations/radar/README.md— index of operational weather-radar networks and composites with openly downloadable gridded data, with quick-reference tables for access tiers and endpoints
These exist because the country-and-type directory structure doesn't surface every useful view on its own. Users arriving with "show me all the AI models" or "what's being retired this year" or "which satellites have no-account-needed S3 access" can go straight to the relevant index.
Information about forecast models, weather satellites, and radar networks is often:
- scattered across many websites
- written for experts only
- difficult to compare or discover
This repository aims to provide a clear, neutral, and beginner-friendly index of what is publicly available.
This repository is maintained as independent research. Everything stays free and openly accessible — no paywalls, accounts, ads, or affiliate links.
If you've found the catalog useful in your work or research, you can optionally support ongoing maintenance through any of the following:
Sponsorship helps support the time this takes: tracking operational changes, verifying sources, documenting upgrades and retirements, and adding new systems as they become publicly available. Even small contributions are meaningful and appreciated — but the catalog will always be free for everyone.
This repository is for information and discovery only. Always consult the official provider's documentation for:
- data usage terms
- licensing
- operational or safety-critical use
I try to document each entry as accurately as possible based on official product user manuals, technical documentation, agency publications, and peer-reviewed references. However, information in this repository may be incomplete, out of date, or inaccurate. This is particularly true for:
- Systems with limited English-language documentation
- Systems operated by agencies that do not publish detailed technical specifications publicly
- Recent operational upgrades that may not yet be reflected in published documentation
- Data distribution details (access URLs, file formats, update schedules) that change without announcement
- Operational role assignments for satellite constellations, which rotate over time as new spacecraft commission and older ones retire
Operational forecast and observation systems evolve continuously. Resolution changes, physics upgrades, ensemble size changes, instrument decommissioning, and distribution channel changes can happen with little public notice, and the gap between documented and operational configurations can be months or years for some systems.
If you spot an error, know of a more recent version of a model's documentation, or have corrections to contribute, please see the Contributions section — corrections and updates are genuinely welcome and are the main way this catalogue stays useful over time.
Corrections, additions, and improvements are welcome.
If you add a model, satellite, or radar entry, please ensure:
- the data is genuinely free and downloadable (free registration is acceptable; paywalls and approval gates are not)
- links point to official sources
- descriptions remain simple, factual, and non-promotional
- the entry uses one of the templates in
templates/as a starting point