BOSZ spectra are staged as local source data under data/atmospheres/. The STScI
HLSP page labels the archive path as bosz2024, but its 2025 Sept. 24 update
states that the models were recomputed and earlier 2024 files should be
replaced.
Local layout¶
data/atmospheres/bosz/2025-recomputed/
provenance/
wavelength/
bosz2024_wave_r10000.txt
scripts/
raw/
r10000/
m+0.00/
processed/
catalog.parquet
catalog_fragments/
validation/
bosz_2025_recomputed.zarr/Everything under data/atmospheres/ is gitignored. Raw gzip spectra and
processed artifacts are local cache products, not package data.
Precision policy¶
The converter parses the two released resampled columns as float64 for
validation, then stores both arrays as float32 in the processed Zarr artifact.
The shared wavelength grid is stored as float64.
This mirrors the NewEra policy: dense spectra are the storage bottleneck, while
the wavelength coordinate is small and should remain precise. Downstream
interpolation or integration may cast spectra back to float64 for accumulation.
The converter preserves the released BOSZ resampled columns as flux and
continuum arrays. Unit-normalization and surface-flux interpretation belong at
the downstream physical interface, not in the raw cache converter.
Conversion¶
Install the optional data tooling and run a dry-run first:
uv run --extra data python scripts/convert_bosz_resampled.py --dry-runConvert the staged bridge subset:
uv run --extra data python scripts/convert_bosz_resampled.py --overwriteFor large BOSZ expansion, convert in batches by resolution, metallicity, and composition. Give each batch a stable group name:
uv run --extra data python scripts/convert_bosz_resampled.py \
--raw-dir data/atmospheres/bosz/2025-recomputed/raw/r10000/m-0.50 \
--wavelength data/atmospheres/bosz/2025-recomputed/wavelength/bosz2024_wave_r10000.txt \
--group-name r10000_mminus0_50_v2 \
--delete-raw-after-validateRaw files are deleted only after the batch readback validation succeeds.