Open-Meteo

About Open-Meteo

Open-Meteo is a weather API providing forecast, historical, and archived model data through a simple JSON HTTP GET interface with no authentication required. It aggregates output from more than 30 weather models operated by national weather services including ECMWF, NOAA, DWD, Météo-France, JMA, KMA, KNMI, DMI, MeteoSwiss, UK Met Office, BOM, CMA, and GeoSphere Austria, with resolution ranging from 1-2 km for regional models to 9-11 km for global models. A best_match option auto-selects the highest-resolution model for a given location, or individual models can be selected directly. Global models typically update every 6 hours, while high-resolution regional models such as ICON-D2, HRRR, and AROME update every 1-3 hours; 15-minutely data is available for Central Europe and North America.

Beyond live forecasts, the service offers four historical data layers: ERA5 reanalysis from January 1940, a Historical Forecast archive from 2021 in the same format as the live API, a Previous Runs API for lead-time-stratified skill analysis from January 2024, and a Single Runs API with archived individual model runs by init time (ECMWF IFS HRES from March 2024) for ML post-processing without look-ahead bias. Additional APIs cover ensemble forecasts, climate change projections, marine weather, air quality, geocoding, elevation, and flood data. Output can be returned in JSON, CSV, or XLSX, and multiple locations can be queried in a single request using comma-separated coordinates.

The API is free without an API key or sign-up for non-commercial use up to 10,000 daily calls, requiring attribution under the CC BY 4.0 data licence; subscription plans exist for commercial use or higher call volumes, adding rate-limit increases and priority support. The server codebase is open source under AGPLv3 on GitHub, allowing self-hosting for unlimited API calls, which is suited to high-volume ML workloads or air-gapped environments.

Key features

• 30+ weather models from national weather services including ECMWF, NOAA, DWD, and Met Office, up to 1 km resolution
• ERA5 historical reanalysis data from 1940, hourly and spatially complete
• Historical Forecast archive from 2021 in the same format as the live Forecast API for bias-correction pipelines
• Archived individual model runs by init time for ML post-processing without look-ahead bias
• Simple JSON API via HTTP GET with no authentication required, plus CSV and XLSX output formats
• Open-source server code on GitHub (AGPLv3) that can be self-hosted for unlimited API calls

Frequently asked questions