This space is open for questions and discussion around the Polar Hackathon. Please feel free to post any questions here, whether they are about datasets, notebooks, workflows, technical issues, or ideas you would like to explore during the hackathon.
You can also create a new topic in the Polar Science Cluster category if your question or discussion deserves its own thread:
Looking forward to seeing your questions, ideas, and hackathon outcomes!
Resources are available below in the link below:
Cloning the repository seems to work well. To create the environment, I used the python package manage pixi, with the following commands:
pixi init
pixi add xarray dask geopandas rasterio rioxarray pyproj shapely pandas numpy matplotlib pystac ipykernel zarr aiohttp
These packages were listed in the readme, except for ipykernel zarr aiohttp.
Nice! Thanks @mdtanker, I’ve added these to the repo setup as an alternative pixi install and added the libraries
Just to add that cartopy is also missing from the requirements, and +1 on using pixi @mdtanker
s3fs is also required for the calving fronts notebook (not sure if for any others require this also)
and hvplot for ice velocity
yea, I think it will be required for remote geopandas access for all vectors. thanks
Oh yeah I just saw that too, add it to the list in the GitHub issue: Use environment.yml instead of conda-linux-64.lock for building local environment · Issue #2 · ESA-EarthCODE/polar_hackathon · GitHub
thanks - added a environment yaml that’s platform agnostic
Does the EOxHub service hosting the Jupyter platform run in the same Cloudferro data centre hosting the data? Just curious
the workspace runs on OTC, now T Cloud Open, I believe; i.e. the compute for this particular workspace is not collocated with the data for the data cube you’re seeing in the notebooks
That’s good to know, thanks. Are there any plans for the compute and data to be colocated in the future?
We are are cloud agnostic when choosing where to run the compute and we have Polar TEP workspaces running with about four providers at the moment. There are many factors to consider in the decision. Currently, geopolitical considerations seem to be outweighing technical performance.
Thanks David. I just happened to start reading the new Scientific Data paper on CDSE (Kovács et al. 2026; Copernicus Data Space Ecosystem establishes public cloud processing for earth observation data | Scientific Data) and it mentions that there is a direct link between CloudFerro and OTK, which is interesting.
The CDSE platform runs on top of two Public Clouds, CloudFerro Cloud (CFC) and Open Telekom Cloud (OTC). The clouds are directly connected to each other, the GÉANT scientific network, and the internet, using redundant broadband, low latency multiple 100 G links to allow smooth operation across the clouds.
So there might not be as big a performance hit with compute and data being physically separated in this instance as you’d expect if the connection was just over the open internet?
I’m having trouble figuring out the CS+AO dataset CRS, as I don’t think it is in EPSG:3031. The attributes in the ZARR say the projection is LAEA, but I am also struggling to get rioxarray to read anything meaningful from the attributes in order to reproject into EPSG:3031.
When plotting the data, it looks like we need some kind of x/y offset, but I can’t see these anywhere in the attributes. If I subtract half the max-min for the x/y coordinates, I get something that seems to align with EPSG:3031. Do you know what the correct CRS should be, or am I missing something in the datacube?
You can see the dataset attributes listed here: CS+AO Sea Ice Thickness - Polar Hackathon
Based on the below attributes I think the the dataset is in ccrs.LambertAzimuthalEqualArea(
central_latitude=-90.0,
central_longitude=0.0
)
grid_type : sp2ease
projection :laea
lon_0 : 0
lat_0 : -90