The Polar Data Cube

As part of the initiative, and to facilitate easier analysis of and access to scientific Antarctica data, some of the presented datasets are combined into zarr stores, reprojected to a common 100x100m grid in the EPSG:3031.

Putting datasets on a common projection, resolution, and coordinate system reduces the amount of preprocessing needed before analysis. Instead of repeatedly handling reprojection, spatial joins, resampling, and file-format differences, you can focus on comparing variables across the same x and y grid.

Content¶

The notebooks in this folder introduce the Antarctica data cube workflow from two angles: using prebuilt remote cubes, and understanding how multiple datasets can be aligned into a shared analysis grid to produce ARD.

1. 1_remote_cube_access.ipynb opens the remote Zarr stores lazily with xarray, selects a small region of interest, and plots example layers without downloading the full cube.

2. 2_build_single_cube_demo.ipynb demonstrates the construction pattern on a compact ROI by combining gridded, vector, and raster inputs onto one EPSG:3031 grid.

3. 3_large_area_processing.ipynb shows how to work with larger cube regions using Dask.

Additionally there are notebooks that show how one can access external datasets and combine it with the cube data under ./accessing_external_datasets

Visualization¶

You can see a visualisation of the data in eodash:

https://esa-earthcode.github.io/polar-science-cluster-dashboard/

Data Collections Overview¶

• All spatial layers are aligned to one projected epsg:3031 x/y grid. The working grid uses 100 m cells, with x from -2867900 to 2867900 and y from -2457900 to 2457900, giving 57,358 columns by 49,158 rows.

• The same x, y position refers to the same place across every variable.

• Some datasets already closely match the target grid. Others had to to be reprojected, shifted, resampled and interpolated. The default interpolation method is nearest neighbour.

• Vector datasets are converted into raster variables by burning their geometries onto the target grid.

• If a dataset contains extra dimensions, such as time, depth, height, band, or category, the spatial transformation was applied across the horizontal grid while preserving those dimensions. All datasets are then combine in several stores according tot heir common dimensions:

cube_paths = [ “https://s3.waw4-1.cloudferro.com/EarthCODE/OSCAssets/antarctica_cube/icetemp.zarr”, “https://s3.waw4-1.cloudferro.com/EarthCODE/OSCAssets/antarctica_cube/sec.zarr”, “https://s3.waw4-1.cloudferro.com/EarthCODE/OSCAssets/antarctica_cube/antarctica-combined.zarr”, “https://s3.waw4-1.cloudferro.com/EarthCODE/OSCAssets/antarctica_cube/icemask_composite.zarr/”, “https://s3.waw4-1.cloudferro.com/EarthCODE/OSCAssets/antarctica_cube/ice_velocity.zarr”, ]

Current Assumptions and Future Refinements¶

The current approach and any other approach to upsampling or downsampling the original data will invite distortions. In some cases the data then becomes unusable for some scientific analysis. This is why we also transform and keep the original data in a cloud-friendly format (see 1_Datasets) to enable other scientific workflows that do not require a whole integrated data layer.