feat: add Pan-STARRS (PS1) survey

[aboucaud]

Summary

Adds Pan-STARRS1 (PS1) as a new survey: surveycodex/data/PanSTARRS.yaml (filename/survey name confirmed with the maintainer — no hyphen, unlike the mission's "Pan-STARRS" branding). Telescope: the 1.8m PS1 telescope at Haleakala, Hawaii. Instrument: Gigapixel Camera #1 (GPC1). Filters: g, r, i, z, y, plus the wide w filter. zeropoint/effective_wavelength are computed with speclite's panstarrs filter group, which was confirmed to resolve for all 6 bands.

Also adds "PanSTARRS": "panstarrs" to SPECLITE_SURVEY_PREFIXES in scripts/check_zeropoints.py and scripts/check_effective_wavelengths.py.

Sourced values and citations

Field	Value(s)	Source
pixel_scale	0.258 arcsec/px	Chambers et al. 2016 (arXiv:1612.05560), Section 2.4 / Table 3
mirror_diameter	1.8 m	Chambers et al. 2016, Table 2 ("Primary mirror 1800 mm diameter")
obscuration	0.32078172	Derived from Chambers et al. 2016 Table 2, which quotes the effective aperture directly as 0.65 × π × (92.2 cm)² = 17284 cm² "including diffraction and obscuration"; obscuration = 1 − 17284 cm² / (π × 90 cm²) relative to the 1.8 m primary
gain	0.97 e⁻/ADU	Tonry et al. 2012 (arXiv:1203.0297), Section 2.5 (p. 15): measured gain of the OTA34/cell33 amplifier used for standard-star calibration
zeropoint_airmass	1.2	Tonry et al. 2012, text preceding Table 4: PS1 natural-system zeropoints are explicitly defined as "the AB magnitude of a neutral color... star that would produce 1 e⁻/sec in the detector with 1.2 airmasses of extinction"
sky_brightness (g/r/i/z/y/w)	21.92 / 20.83 / 19.79 / 19.24 / 18.24 / 20.62	Tonry et al. 2012, Table 4, column µ_obs (observed sky brightness) — one consistent source covering all 6 filters, including w, which has no entry in Chambers et al. 2016's 3π-survey-only tables
full_exposure_time (g/r/i/z/y)	516 / 480 / 540 / 360 / 360 s	Chambers et al. 2016 Table 11 nominal single-exposure time (43/40/45/30/30 s) × 12 dithered visits over the survey lifetime (Section 3.2.3: "Over 3.5 years this would give ... 12 exposures in each band")
full_exposure_time (w)	45 s	Denneau et al. 2013 (arXiv:1302.7281), Table 1 ("Exposure times as of October 2012"). Chambers et al. 2016 Section 3.4 (Solar System Survey) is unpopulated in that paper and no stacked-visit design is documented for wP1, so the single nominal exposure is used rather than a guessed total
psf_fwhm (g/r/i/z/y)	1.47 / 1.31 / 1.19 / 1.14 / 1.09 arcsec	Chambers et al. 2016 Table 11, median of the PSF FWHM distribution
psf_fwhm (w)	1.31 arcsec	No wP1 row exists in Table 11; approximated with rP1's median FWHM since wP1's effective wavelength (608 nm, Tonry et al. 2012 Table 4) is closest to rP1 (617 nm) among the tabulated 3π filters, and no direct wP1 seeing measurement was found in the literature
zeropoint, effective_wavelength	computed	speclite's panstarrs-* filters

On zeropoint_airmass specifically

Unlike decam2014/sdss2010, speclite's panstarrs filters carry no airmass metadata (load_filter('panstarrs-r').meta.get('airmass') returns None). Rather than reuse another survey's airmass or guess, I went to the primary source defining the PS1 photometric system (Tonry et al. 2012) and found the reference airmass explicitly stated as 1.2 in the text describing Table 4's zeropoint definition. This is a real, citable value, not an assumption.

Methodology validation

Before trusting the compute_zeropoint_mag / speclite code path for PS1, I reproduced DES's already-shipped decam2014-g/r/i/z zeropoints and effective wavelengths using DES.yaml's own mirror_diameter/obscuration, and got an exact match (e.g. g: 26.72 mag / 489.004 nm) before applying the same method to panstarrs-*.

Test plan

• pytest --cov — 36 passed
• pre-commit run --all-files — all hooks pass (isort, black, flake8, pyupgrade, etc.)
• python -m surveycodex — PanSTARRS prints correctly with all 6 filters
• surveycodex --refs — references block prints correctly
• python scripts/check_zeropoints.py — speclite vs surveycodex match exactly for all 6 filters (g/r/i/z/y/w)
• python scripts/check_effective_wavelengths.py — speclite vs surveycodex match exactly for all 6 filters

🤖 Generated with Claude Code

[aboucaud]

Addressed code review findings:

• Fixed a worked-example arithmetic error in the obscuration comment: it quoted the radius as "92.2 cm" but Chambers et al. 2016 Table 2 actually states "92 cm" (0.65×π×92² = 17284 cm², matching the paper's own stated area). The shipped obscuration value was already numerically correct since it used the paper's final area directly — only the comment's worked derivation was wrong.
• Fixed zeropoint_airmass's section citation: "Section 2.6" → "Section 3" (Tonry et al. 2012) — Section 2.6 is an unrelated "Photometry Refinement" section; the quoted PS1 natural-system definition is in Section 3.
• Narrowed the try/except ValueError added to scripts/check_zeropoints.py/check_effective_wavelengths.py so it only swallows speclite's "No such group" error (meaning the installed speclite version lacks the filter group entirely), re-raising anything else. Previously it would have also silently swallowed a genuine "No such filter in this group" error for any survey using this shared code — e.g. masking a future filter-name mapping bug instead of failing loudly.

All validation re-run and clean (pytest, pre-commit, all 6 filters match speclite exactly).