HEALPix-geo benchmark replication

A replication study of the discrete-global-grid benchmarks from Law & Ardo (2024), “Using a discrete global grid system for a scalable, interoperable, and reproducible system of land-use mapping” (Big Earth Data, Law & Ardo (2024)), re-run on HEALPix via healpix-geo — with a sphere vs WGS84 ellipsoid comparison the original H3-based study could not make.

It is the HEALPix companion to the H3 replication: same paper, same benchmarks, but indexed on the GRID4EARTH ellipsoidal-HEALPix stack so the effect of the reference surface becomes measurable.

What the replication tests¶

The paper’s central claim is that indexing vector geometries onto a discrete global grid makes raster–vector land-use operations dramatically faster while staying reproducible. We reproduce that benchmark with HEALPix cells (addressed by depth, the GRID4EARTH term for the refinement level), on both a bare sphere and the WGS84 ellipsoid.

Key findings¶

HEALPix-geo matches H3 performance¶

All three DGGS implementations validate the paper’s headline result with consistent numbers (all at HEALPix depth 9, matching H3 resolution 9 for a like-for-like comparison):

The WGS84 ellipsoid matters for European EO data¶

Because HEALPix on the sphere and on the WGS84 ellipsoid place a latitude-dependent share of points in a different cell, the reference surface is not a free choice for Copernicus-resolution analysis:

The notebook Why the reference surface matters reproduces this pattern from scratch, in seconds, with healpix-geo.

Reproduce it¶

• Source & Docker: annefou/healpix_geo_replication_2026

• Archived release: Fouilloux (2026)

• Core library: healpix-geo (GRID4EARTH) — HEALPix on the WGS84 ellipsoid