Frozen input data for the healpix-resample paper experiments

Résumé fourni par la source

Description This dataset provides the frozen input data used to reproduce the experiments, figures, and tables of the healpix-resample paper. It is designed to let the publication notebooks run with OFFLINE=True, without querying mutable remote catalogues, object stores, or imagery services. The record contains two archives: healpix-resample-paper-core-data-v3.zip — four Sentinel-2 Level-2A B04 extracts representing urban, water, forest, and agricultural scenes (exact product identifiers recorded in data_manifest.csv and in the Zarr metadata); an ERA5 ensemble surface-flux extract used for the conservative-resampling experiment; and the provenance set: data_manifest.csv, git_commit.txt, ATTRIBUTION.md, DATA_LICENSE.md, DATA_DEPOSIT.md. healpix-resample-paper-sentinel2-regions-v1.zip — 40 Sentinel-2 L2A B04 region patches (one per validation region) used by the multi-region reduced-resolution experiment and the throughput benchmark. This record contains no Esri-derived imagery. The high-resolution scene textures and the 360 multi-region patches used by the synthetic experiments are derived from Esri World Imagery, whose terms do not cover redistribution of machine-readable derived datasets. They are instead regenerated locally and deterministically by the acquisition cells of the publication notebooks, from the pinned, immutable World Imagery Wayback release 26334 (2026-08-05): every user fetches identical tiles regardless of when the step is run, and data_manifest.csv pins the expected SHA-256 of every regenerated store, so the regeneration is verifiable, not merely repeatable. All archived and regenerable assets are listed in data_manifest.csv with their repository paths, source identifiers, sizes, and SHA-256 checksums (directory checksums are computed deterministically over every Zarr metadata file and chunk). Both archives are built by a deterministic packing script (sorted members, fixed timestamps), so their MD5 checksums can be re-derived by anyone with the inputs. The archives preserve paths beginning with notebooks/. After cloning the associated software repository at the commit recorded in git_commit.txt, extract both archives at the repository root and verify: python notebooks/build_data_manifest.py --check --doi 10.5281/zenodo.22210945 A successful fresh installation reports 45/45 archived assets available; the 364 Esri-derived assets are reported as regenerable and absent, which is expected. After running the notebooks' acquisition cells once (see REPRODUCING.md, step 2b), the check reports all 364 regenerated assets available with matching SHA-256. The notebooks then run with their default OFFLINE=True configuration and raise an explicit error if an immutable input is missing. The associated software is available from: Repository: https://github.com/GRID4EARTH/healpix-resample Software concept DOI: https://doi.org/10.5281/zenodo.21723671 Software version DOI: https://doi.org/10.5281/zenodo.21723672 This archive contains primary experimental inputs and provenance information. Generated figures, result tables, temporary caches, and notebook outputs are intentionally excluded because they can be regenerated from the archived inputs. Data attribution Contains modified Copernicus Sentinel data 2026. ERA5 data were obtained from the Copernicus Climate Change Service operated by ECMWF. The source dataset is Complete ERA5 global atmospheric reanalysis, DOI: https://doi.org/10.24381/cds.143582cf. No Esri or third-party imagery-provider pixels are included in this record. The notebooks that regenerate the Esri-derived textures apply the attribution "Sources: Esri, Maxar, Earthstar Geographics, and the GIS User Community"; applicable terms are documented in the accompanying DATA_LICENSE.md and ATTRIBUTION.md files.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.