feat: Add function sample_points to the raster accessor as a performant alternative to zonal_stats.

[LuukBlom]

Issue addressed

Fixes #521

Explanation

After discussing with @dalmijn, came up with this solution where we provide a quicker alternative to raster.zonal_stats.
Essentially, the function captures this behaviour:

• find sample points: one per geometry. (determined by sampling_strategy) & determine index (row, col) for each sample point
• directly sample the xr.Dataset. (this keeps xarray from doing massive amounts of overhead when the dataset is large)
• spatial datasets with no time dim, skip any stat calculations and return the value immediately. (makes no sense to take the mean/min/max of a single value)
• spatial datasets with time dim: sample each point at each time step, reduce by applying the stats

When I was testing this, I wrote a little notebook to help visualize bugs:
showcase-sample.ipynb

TODO:

• coverage
• changelog
• performance test compared to zonal_stats?

General Checklist

• Updated tests or added new tests
• Branch is up to date with main
• Tests & pre-commit hooks pass
• Updated documentation
• Updated changelog.rst

Data/Catalog checklist

• data/catalogs/predefined_catalogs.yml has not been modified.
• None of the old data_catalog.yml files have been changed
• data/changelog.rst has been updated
• new file uses LF line endings (done automatically if you used update_versions.py)
• New file has been tested locally
• Tests have been added using the new file in the test suite

Additional Notes (optional)

Add any additional notes or information that may be helpful.

[hboisgon]

Good idea. Wonder why you did not use raster.sample in the background? Good it the method is very simialr in terms of arguments and return object to raster.zonal_stats.

[LuukBlom]

Thanks for the review @hboisgon ! it made me reconsider my assumptions.

I have refactored the function and implemented your comments:

• Now uses raster.sample
• I also renamed it the function from sample_stats to sample_points (can still be changed, but already better imo)
• no longer accesses the raster point by point, but selects all indices at once

[sonarqubecloud]

Quality Gate failed

Failed conditions
78.0% Coverage on New Code (required ≥ 80%)

See analysis details on SonarQube Cloud