Add example to read .om file with xarray, subset by coordinates and time, and export to NetCDF/GeoTIFF

[reneignacio]

Hello,

It would be very helpful to have an example (script or notebook) that demonstrates how to:

1. Read a .om file using OmFilePyReader and convert it into an xarray.DataArray or xarray.Dataset.
2. Select a region of interest (ROI) using geographical coordinates (lat/lon).
3. Filter by a date or date range using the file’s temporal metadata.
4. Visualize the selected region using xarray.plot or matplotlib.
5. Export the subset to NetCDF or GeoTIFF.

Each weather model (e.g., ICON, IFS, etc.) uses a different grid configuration. Apparently, there are auxiliary metadata files to interpret these grids, but in my case, I haven’t been able to locate them or figure out how to use them. Ideally, the library should allow users to specify the model as a parameter, or detect it automatically, so that .om files can be opened without requiring additional files.

Thanks for considering this suggestion and congratulations on all the work behind Open-Meteo!

[terraputix]

This is definitely on the to-do-list! Beware that for the processed om-files there aren't even auxiliary metadata files to convert between geographic and grid coordinates at the moment. Therefore, some logic would be needed implementing this conversion.

I will try to prepare an initial version/example for this tomorrow!

[reneignacio]

This is definitely on the to-do-list! Beware that for the processed om-files there aren't even auxiliary metadata files to convert between geographic and grid coordinates at the moment. Therefore, some logic would be needed implementing this conversion.

I will try to prepare an initial version/example for this tomorrow!

Excellent! I was confused about that, thank you very much for the clarification. I’ll stay tuned for the example when it’s ready. I really appreciate the work you're doing.

[terraputix]

I added an example script which demonstrates how you can read data from different models using geographic coordinates and natural dates in this PR: #32. Feel free to have a look if that is of any use already.

There is additional logic required for other domains (so far I only implemented dwd_icon_d2, ecmwf_ifs025 and meteofrance_arpege_europe), so I will further iterate on this.
Currently I did not implement any kind of box selection, so (for now) one would need to iterate over the method findPointXy to find all the grid coordinates contained in a specific ROI.

Note that all domains currently implemented are using a RegularLatLonGrid (a simple regular latitude longitude grid as used by open-meteo), but some domains are defined on other grids such as RotatedLatLon or Stereographic. Implementing these different projection grids will be a bit of work.

The alternative approach would be to have the latitude and longitude arrays exposed in a static-folder, such as dwd_icon_d2/static. This would involve the overhead of always accessing these files and would require some nearest neighbor search algorithm to efficiently find the points of interest.

[reneignacio]

I added an example script which demonstrates how you can read data from different models using geographic coordinates and natural dates in this PR: #32. Feel free to have a look if that is of any use already.

There is additional logic required for other domains (so far I only implemented dwd_icon_d2, ecmwf_ifs025 and meteofrance_arpege_europe), so I will further iterate on this. Currently I did not implement any kind of box selection, so (for now) one would need to iterate over the method findPointXy to find all the grid coordinates contained in a specific ROI.

Note that the all domains currently implemented are using a RegularLatLonGrid (a simple regular latitude longitude grid as used by open-meteo), but some domains are defined on other grids such as RotatedLatLon or Stereographic. Implementing these different projection grids will be a bit of work.

The alternative approach would be to have the latitude and longitude arrays exposed in a static-folder, such as dwd_icon_d2/static. This would involve the overhead of always accessing the files exposed there and would require some nearest neighbor search algorithm to efficiently find the points of interest.

Thanks for the example and the work in PR #32, clear and useful.

[terraputix]

I added another small example script in https://github.com/open-meteo/python-omfiles/blob/main/examples/spatial_xarray.py showing how to use xarray to read data directly from AWS S3.

In the future we'll see how to stabilize metadata access, but this discussion will be continued in #22 or in the main open-meteo repository.