Thanks for the great work! I have a question regarding the API: Is it possible to specify a particular "Model Run" (or initialization time)? I would like to retrieve forecast data from a specific run rather than the latest updated forecast provided by default.
Hi, we're developing a new API endpoint that will provide the full forecast horizon from any historical weather model run. We began collecting data in September 2025 and are in the process of backfilling data from the past 2–3 years. Unfortunately, only a limited number of national weather services offer accessible archives. At the moment, we are downloading ECMWF IFS HRES data for all of 2025, and the next step is to begin downloading GFS data.
I have a couple of questions regarding the self-hosted version:
1.GitHub Update: Has the code for this new endpoint been merged into the latest release on GitHub?
2.Local Usage: If I deploy the service locally, does it support retrieving these specific past runs out-of-the-box? (e.g., does it fetch the historical run data from S3 automatically, or does it require additional configuration?)
Yes, the entire source code is available on GitHub. OpenMeteo is developed in public and all changes are immediately visible to everyone.
We only have a limited amount of data we can publish through an AWS open-data sponsorship. We only keep the last couple of weeks of full horizon weather model runs. Data in form of continuous time-series is retained indefinitely. See https://github.com/open-meteo/open-data
We are working on an internal storage system to keep the full forecast horizon of weather model for the past years, but no details are available yet.
Regarding the local deployment: I noticed that a standard request works fine, but querying a specific run (e.g., &run=2026-01-20T00:00) returns null, even for recent dates.
Since you mentioned these recent runs are available on AWS Open Data for a limited time, could you share the specific command or configuration needed to sync these raw model runs locally?
My current sync process seems to only build the continuous time-series. Do I need to target a specific S3 prefix or use a different flag with the openmeteo-api sync tool to enable the &run= endpoint?"
Hi. ECMWF IFS model outputs are licensed under open-data CC-BY-4.0 but this doesn't mean they're freely downloadable. Only the lower resolution 0.25° datasets with 3-hourly data are available for free. The high-resolution native O1280 HRES 9 km model output with 1-hourly data requires a service charge. We pay for access to ECMWF data and are permitted to redistribute it including HRES, seasonal forecasts (9-13 months) and sub-seasonal forecasts (EC46 46 days). More information about the ECMWF service charge is available here: https://confluence.ecmwf.int/display/DAC/Service+Charges%3A+From+01+July+2024
This is great news! thank you for the update. I see we can access the data from your api, do you have a place where we can get the grib to extract raster for doing images ?
Hi, we redistribute data in spatial form on AWS [1]. IFS HRES is in its native O1280 grid which most plotting libraries cannot work with. We have some code examples to convert O1280 to 0.1° grids [2]
Hi, for now we won’t be adding upper-level variables from IFS HRES. At 850/500 hPa, the benefit of higher-resolution data is quite limited, and the existing IFS 0.25° upper-air data should be sufficient.
Thanks, Patrick. You're right that higher resolution doesn’t bring much benefit for upper-air data. I was mainly thinking about having access to the data without the additional time delay that comes with the IFS 0.25° dataset.
I have one question: will the historical forecasts also be available through the Historical Forecast API? If so, from what date will they be available?
I noticed that in the Forecast API, IFS HRES includes the variable Wind Speed at 10m, 80m, 120m, and 180m, but in the Historical Forecast this variable is only available at 10m. Will the other altitudes be included in the Historical Forecast API in the future?
Thanks for the great work! I have a question regarding the API: Is it possible to specify a particular "Model Run" (or initialization time)? I would like to retrieve forecast data from a specific run rather than the latest updated forecast provided by default.
Hi, we're developing a new API endpoint that will provide the full forecast horizon from any historical weather model run. We began collecting data in September 2025 and are in the process of backfilling data from the past 2–3 years. Unfortunately, only a limited number of national weather services offer accessible archives. At the moment, we are downloading ECMWF IFS HRES data for all of 2025, and the next step is to begin downloading GFS data.
You can test the API endpoint here: https://open-meteo.com/en/docs/single-runs-api. Please note that there might be smaller changes to the API
Hi Patrick, thanks for the info!
I have a couple of questions regarding the self-hosted version:
1.GitHub Update: Has the code for this new endpoint been merged into the latest release on GitHub?
2.Local Usage: If I deploy the service locally, does it support retrieving these specific past runs out-of-the-box? (e.g., does it fetch the historical run data from S3 automatically, or does it require additional configuration?)
Yes, the entire source code is available on GitHub. OpenMeteo is developed in public and all changes are immediately visible to everyone.
We only have a limited amount of data we can publish through an AWS open-data sponsorship. We only keep the last couple of weeks of full horizon weather model runs. Data in form of continuous time-series is retained indefinitely. See https://github.com/open-meteo/open-data
We are working on an internal storage system to keep the full forecast horizon of weather model for the past years, but no details are available yet.
"Thanks for the explanation, Patrick.
Regarding the local deployment: I noticed that a standard request works fine, but querying a specific run (e.g., &run=2026-01-20T00:00) returns null, even for recent dates.
Since you mentioned these recent runs are available on AWS Open Data for a limited time, could you share the specific command or configuration needed to sync these raw model runs locally?
My current sync process seems to only build the continuous time-series. Do I need to target a specific S3 prefix or use a different flag with the openmeteo-api sync tool to enable the &run= endpoint?"
Fantastic news for the open-data community and everyone working with real-time climate information.
The release of ECMWF IFS HRES at 9 km marks a huge step toward truly open, high-resolution forecasting.
Congratulations, Patrick, for making it accessible to all through Open-Meteo — a great achievement for the weather and climate ecosystem.
dear @Patrick , for me It's unclear whether access to the ECMWF IFS HRES 9km data is truly open directly on ECMWF.
Do you access it because you have a license, or can anyone access it? If anyone can access it, is it possible to find out the route?
Hi. ECMWF IFS model outputs are licensed under open-data CC-BY-4.0 but this doesn't mean they're freely downloadable. Only the lower resolution 0.25° datasets with 3-hourly data are available for free. The high-resolution native O1280 HRES 9 km model output with 1-hourly data requires a service charge. We pay for access to ECMWF data and are permitted to redistribute it including HRES, seasonal forecasts (9-13 months) and sub-seasonal forecasts (EC46 46 days). More information about the ECMWF service charge is available here: https://confluence.ecmwf.int/display/DAC/Service+Charges%3A+From+01+July+2024
Where can I get the original 0.1 degree gridded data?
Windy on suicide watch
Is PBLH available? I see it was there for one day but no longer.
Hi, PBLH is available again!
This is great news! thank you for the update. I see we can access the data from your api, do you have a place where we can get the grib to extract raster for doing images ?
Hi, we redistribute data in spatial form on AWS [1]. IFS HRES is in its native O1280 grid which most plotting libraries cannot work with. We have some code examples to convert O1280 to 0.1° grids [2]
[1] https://openmeteo.s3.amazonaws.com/index.html#data_spatial/ecmwf_ifs/
[2] https://github.com/open-meteo/python-omfiles/blob/main/examples/regrid_ecmwf_ifs_hres_gaussian_O1280_to_0.1_degree.py
Would it be possible to include some (common) upper-air variables, e.g. 850/500 hPa temperature and geopotential? Thanks for the great service!
Hi, for now we won’t be adding upper-level variables from IFS HRES. At 850/500 hPa, the benefit of higher-resolution data is quite limited, and the existing IFS 0.25° upper-air data should be sufficient.
Thanks, Patrick. You're right that higher resolution doesn’t bring much benefit for upper-air data. I was mainly thinking about having access to the data without the additional time delay that comes with the IFS 0.25° dataset.
Excellent news!
I have one question: will the historical forecasts also be available through the Historical Forecast API? If so, from what date will they be available?
Thank you very much in advance!
Hi, archived IFS HRES data is already available from 2018 onwards. See here: https://open-meteo.com/en/docs/historical-forecast-api?models=ecmwf_ifs&start_date=2025-01-01. We also store the full forecast horizon for individual model runs and working on an API to access it, but its not ready yet ;-)
Thank you for your answer!
I noticed that in the Forecast API, IFS HRES includes the variable Wind Speed at 10m, 80m, 120m, and 180m, but in the Historical Forecast this variable is only available at 10m. Will the other altitudes be included in the Historical Forecast API in the future?
Wind level 100m is available. Level 200m will be in the future. You can overwrite the &hourly= parameter in the URL to get 100m wind: https://open-meteo.com/en/docs/historical-forecast-api?models=ecmwf_ifs&start_date=2025-09-28&hourly=wind_speed_100m,wind_speed_200m&end_date=2025-10-01
very interesting, thank you! I noticed GFS global also has wind_speed 100m