ztf-dust

Scripts to reproduce the results and figures of:

Giunta, Karchev & Trotta (2026), The colour variability of low-z SNe Ia is entirely explained by dust

The analysis uses the simplebayesn package, which must be installed first.

---

Dependencies

Install simplebayesn and its dependencies:

pip install git+https://github.com/marco-giunta/simplebayesn.git

Additional packages needed for the Foundation SALT2 fits:

pip install sncosmo tqdm

---

Data

• ZTF DR2: download from https://ztfcosmo.in2p3.fr/
• Foundation DR1: clone from https://github.com/djones1040/Foundation_DR1
• SALT2.JLA-B14 model files: available from the SNANA distribution under snana/models/SALT2/SALT2.JLA-B14

---

Workflow

The scripts are listed below in the order they should be run. Each script accepts --input / --output arguments; run any script with --help for details.

1. Foundation SALT2 fits

The official Foundation DR1 release only provides SALT2 fits for the cosmological subsample. This script recomputes fits for both the cosmo and no-cosmo samples using sncosmo with the SALT2.JLA-B14 model, matching the original SNANA setup:

python foundation_salt_fit.py \
    --input /path/to/Foundation_DR1/ \
    --output data/foundation_salt.csv \
    --salt /path/to/SALT2.JLA-B14/

2. Gibbs sampling

ZTF HQ VL (full sample, no colour cut):

python ztf_gibbs.py \
    --input data/ztf_dr2.csv \
    --output chains/ztf_gibbs.h5

ZTF HQ VL with $|c| \leq 0.3$ colour cut:

python ztf_03_color_cuts_gibbs.py \
    --input data/ztf_dr2.csv \
    --output chains/ztf_03_gibbs.h5

Foundation (full and cosmo-only):

python foundation_gibbs.py \
    --input data/foundation_salt.csv \
    --output chains/foundation_gibbs.h5

python foundation_cosmo_gibbs.py \
    --input data/foundation_salt.csv \
    --output chains/foundation_cosmo_gibbs.h5

ZTF subsamples split by host-galaxy environment (stellar mass and DLR):

python ztf_gibbs_mass_splits.py \
    --input_data data/ztf_dr2.csv \
    --input_global data/ztf_globalhost.csv \
    --input_local data/ztf_localhost.csv \
    --output chains/splits/

3. Selection bias simulation

These scripts demonstrate the bias induced by unmodelled colour cuts, using simulated ZTF-like data:

# Gibbs sampler on simulated data with and without cuts
python ztf_sim_sel_gibbs.py \
    --input data/ztf_dr2.csv \
    --output chains/ztf_sim_gibbs.h5

# emcee sampler with likelihood renormalisation on the cut simulated data
python ztf_sim_sel_emcee.py \
    --input data/ztf_dr2.csv \
    --output chains/ztf_sim_emcee.h5

4. Figures and tables

Each script reads one or more chain .h5 files and saves its output. Run with --help for argument details.

Script	Output
ztf_foundation_mb_c.py	Fig. 1: observed magnitude-colour distributions of ZTF and Foundation cosmo/no-cosmo
ztf_sim_sel_cornerplot.py	Fig. 2: selection bias simulation cornerplot
ztf_03_color_cuts_gibbs_posterior_cornerplot.py	Fig. 3: ZTF colour parameter posteriors with/without cut
ztf_gibbs_extrinsic_latent_population_plot.py	Fig. 4: ZTF latent colour-magnitude distribution
ztf_foundation_gibbs_posterior_cornerplot.py	Fig. 5 and Fig. A2: ZTF vs Foundation posteriors
ztf_gibbs_latents_plot.py	Figs. 6-7: ZTF per-SN latents vs host environment
ztf_gibbs_mass_splits_posterior_cornerplots.py	Figs. 8-9: ZTF split-sample posteriors
ztf_gibbs_posterior_cornerplot.py	Fig. A1: full ZTF parameter cornerplot
ztf_fnd_gibbs_table.py	Tab. 3: parameter table

---

Citation

If you use these scripts, please cite:

Giunta, Karchev & Trotta (2026), The colour variability of low-z SNe Ia is entirely explained by dust