Fast, practical static plots of HEALPix data with matplotlib and cartopy — with ellipsoidal (WGS84) support for geoscience workflows.

healpix-plotting prioritises getting a usable figure quickly over perfectly accurate cell-geometry rendering. It rasterises HEALPix data via nearest-neighbour resampling onto a regular lon/lat grid and renders the result with Cartopy's imshow.

Unlike astronomy-focused HEALPix tools, this library is built with Earth observation and geoscience in mind: the underlying coordinate operations are provided by healpix-geo, which supports geodetically correct reference ellipsoids such as WGS84.

The library is well suited for:

• Exploratory analysis and quality control of EO / climate data
• Debugging and sanity checks
• Quick snapshots for reports and discussions

It is not intended as a true boundary-polygon renderer.

pip install healpix-plotting

1. Build a target sampling grid — a regular lon/lat grid inferred from the data, defined by a bounding box, or specified via an affine transform.
2. Resample — each sampling point is mapped to a HEALPix cell id and filled by nearest-neighbour lookup (with optional aggregation for duplicate ids).
3. Render — the resulting raster is drawn on a Cartopy axis with imshow(..., transform=PlateCarree()).

Because the library rasterises via nearest-neighbour resampling, it does not attempt exact polygon boundary filling.

Standard HEALPix was originally defined on the unit sphere. For geoscience applications — such as Earth observation, numerical weather prediction, or climate modelling — coordinates are commonly expressed in geodetic lon/lat on a reference ellipsoid (most often WGS84).

healpix-plotting delegates coordinate conversions to healpix-geo, which implements ellipsoidal HEALPix conversions. Passing ellipsoid="WGS84" to HealpixGrid propagates this choice through all internal operations.

For a full explanation of how ellipsoidal HEALPix works, see the healpix-geo ellipsoids tutorial.

HEALPix is a spherical tessellation. Depending on your use case you may need to consider:

• True cell boundaries — cells are not lon/lat rectangles; boundaries are curved on the sphere.
• Projection effects — any map projection changes apparent cell geometry.
• Resolution and aliasing — the target sampling-grid resolution determines sharpness and artifact level.
• Dateline and polar behaviour — extent wrapping and polar distortion can introduce discontinuities.
• Sphere vs. ellipsoid — for precise geoscience work, prefer ellipsoid="WGS84" over the default "sphere".

• Rendering is raster-based; exact HEALPix cell boundaries are not drawn.
• Output quality depends on the sampling-grid resolution (speed vs. aliasing trade-off).
• "bilinear" interpolation is currently not implemented (raises NotImplementedError).
• The helper pattern healpix_grid.operations.healpix_to_lonlat(cell_ids, **healpix_grid.as_keyword_params()) does not work for indexing_scheme="zuniq" — see the caveat in HealpixGrid.

Need a quick geoscience plot (EO, NWP, climate)?  → healpix-plotting  ✓  (WGS84 support)
Working in the ECMWF/earthkit ecosystem?           → earthkit-plots    ✓
Need exact cell boundaries / polygon operations?   → xdggs             ✓
Standard healpy full-sky (astronomy) workflow?     → healpy            ✓

For each sampling-grid point, the corresponding HEALPix cell id is computed. Ids not present in the source data are masked; the remaining values are placed into the raster using searchsorted-based indexing.

Before resampling, duplicate cell_ids are collapsed with numpy-groupies using the function specified by agg (default: "mean"). The sorted unique ids are then used as the lookup table for the raster fill.

Plotting uses imshow with transform=ccrs.PlateCarree() and interpolation="nearest". For non-global subsets, the extent is set before plotting to obtain a smoother result with Cartopy.