// the find
corteva/rioxarray
geospatial xarray extension powered by rasterio
rioxarray is an xarray accessor that brings rasterio's geospatial operations (reprojection, clipping, CRS management, nodata handling) into the xarray data model. It's for geoscientists and remote sensing engineers who want to work with raster data in NumPy/Dask arrays without dropping into GDAL directly. Backed by Corteva Agriscience, so it has real production usage.
The accessor pattern is the right call — `ds.rio.reproject('EPSG:4326')` is genuinely nicer than juggling rasterio contexts manually. Dask write support means you can process rasters larger than RAM without rewriting your pipeline. The convention system (CF, Zarr) handles CRS metadata portability across formats, which is the part that usually breaks silently. Test suite is integration-heavy with real raster files, not mocked fixtures — you actually trust what passes.
618 stars for a library that solves a legitimately hard problem suggests limited discovery outside the AgriScience/earth-observation niche — don't expect a large community if you get stuck. The README is nearly empty; all docs live on the GitHub Pages site, which means a broken deployment leaves new users with nothing. CRS inference from non-standard NetCDF files still requires hand-holding — the `_convention` module has edge cases that bite anyone not using CF-compliant data. No support for vector-raster operations beyond clipping; you still need to reach for rasterio or geopandas for anything fancier.