# GNILC dust templates with the CIB monopole fix (temporary evaluation copy)

Regenerated GNILC dust templates (inputs of the PySM 3 `d9`/`d10` models, and of the
`d11` large scales) after fixing the NumPy indexing bug that prevented the monopole
subtraction in the preprocessing notebook
(https://github.com/galsci/pysm/issues/281).

What changed relative to the released `gnilc_dust_template_nside2048_2023.02.10.fits`:

- the zero-level offset of the GNILC F353 input map, **0.1254 MJy/sr**, measured with
  the HI4PI HI column-density correlation
  (https://www.zonca.dev/posts/2026-11-03-gnilc-dust-zero-level-offset), is now
  actually subtracted from intensity outside the GAL080 mask (the 0.13 MJy/sr from
  Planck 2018 XII that the pipeline intended to subtract was never applied due to the
  bug);
- all downstream artifacts are recomputed from the corrected intensity: small-scale
  logpoltens spectra, temperature and polarization modulation maps, large-scale
  logpoltens ALMs, Galactic plane fix file;
- the small scales use the same random seed as the 2023 run.

Main effects: the intensity mean outside the Galactic plane drops by 0.114 MJy/sr
(offset times the 0.911 color correction), the polarization fraction at high latitude
roughly doubles (median p outside GAL080: 4.1% -> 8.0%), and the TT/EE/BB spectra are
preserved at large scales while the small scales are moderately enhanced.

Contents: `gnilc_dust_template_nside2048_fix281_2026.11.03.fits` (uK_RJ, 353 GHz,
same format as the released templates) and `raw/` with the regenerated artifacts.

**This is a temporary location for evaluation, not a replacement of the released
templates.** Generated 2026-11-03 on Popeye (SDSC) by Andrea Zonca.