PANORAMIC JWST Imaging Survey
- PANORAMIC is a wide-area, pure parallel extragalactic imaging survey using JWST/NIRCam across 1–5 μm to capture early galaxy evolution.
- The survey employs a 'wedding cake' design with multiple tiers of shallow wide fields and select ultra-deep pointings to reduce cosmic variance.
- Covering ~530 sq. arcmin and utilizing 6–7 filters in 192 hours of observation, PANORAMIC sets a new benchmark for Cycle 1 extragalactic studies.
PANORAMIC is a pure parallel extragalactic imaging program with JWST/NIRCam observed during JWST Cycle 1. Conceived to address early galaxy evolution with both depth at $2$– and statistically large, widely separated sightlines, it exploits NIRCam’s simultaneous short- and long-wavelength imaging via a dichroic to build a multi-band “wedding cake” survey comprising wide, shallower pointings and a smaller number of ultra-deep pointings. The survey obtained square arcmin of NIRCam imaging from $1$–, totaling hours of science integration time, including square arcmin of novel sky area with at least six broad-band filters and square arcmin with a seventh filter, while delivering photometric redshift performance comparable to CANDELS without ancillary data (Williams et al., 2024).
1. Program definition and scientific rationale
PANORAMIC, JWST GO-02514 under PIs Williams and Oesch, was designed around a specific observational problem: many of the most consequential questions in early galaxy evolution require both deep $2$– imaging and many independent sightlines. The survey therefore prioritizes wide-area sampling rather than a single contiguous field. This design minimizes cosmic variance, especially at the bright end of luminosity functions where clustering dominates uncertainties, while preserving sufficient depth for robust photometric redshifts and source characterization (Williams et al., 2024).
A defining feature is its use of pure parallel observing. In the PANORAMIC design, pure parallel observations naturally produce a “wedding cake” structure, in which many fields are relatively wide and shallow while a subset become substantially deeper. This tiered geometry is not incidental but central to the survey’s scientific utility: it supports simultaneous study of rare bright galaxies and fainter, more abundant systems across a broad redshift range.
The survey was also framed relative to the broader Cycle 1 extragalactic landscape. It complements narrower, deeper programs such as JADES, CEERS, and PRIMER, and broader but fewer-band imaging such as COSMOS-Web with four bands. Its distinguishing contribution is wide-area NIRCam imaging with six or more filters at 0–1, sufficient to obtain robust photometric redshifts without followup or ancillary data.
2. Observational design, filters, and footprint
The observational configuration combines six broad NIRCam bands—F115W, F150W, F200W, F277W, F356W, and F444W—with additional medium-band coverage where available. F410M was observed over 2 square arcmin, and one special GOODS-N association replaced broad bands with six medium bands: F162M, F182M, F210M, F300M, F430M, and F460M. Field selection prioritized low 3 zodiacal background at the 10th percentile, high Galactic latitude 4, at least two mechanism moves to enable three filter-pairs, and prime-program exposures allowing at least four groups in ramps for robust cosmic ray rejection. Mid-Cycle, the minimum NIRCam parallel exposure per pointing was relaxed to 42 minutes from a 1-hour goal in order to maximize area, while long-parallel slots were conservatively truncated to 5 hours per pointing to respect spacecraft data volume limits (Williams et al., 2024).
The realized survey geometry comprises 55 pointings with at least two filters and 45 pointings or associations with at least six broad bands, spread across roughly 40 associations. Approximately 200 square arcmin falls in or adjacent to major extragalactic deep fields, including GOODS-South, GOODS-North, COSMOS, COSMOS-Web, UDS, UDS-PRIMER, XMM/UDS, Abell 370, MACS0416, SSA22, and EGS. This arrangement both enlarges the novel-sky area and increases the legacy value of established deep-field footprints.
| Quantity | Value | Note |
|---|---|---|
| Total NIRCam imaging | 6 sq arcmin | Cycle 1 pure parallel program |
| Novel area with 7 broad bands | 8 sq arcmin | 45 pointings across 9 associations |
| F410M coverage | 0 sq arcmin | Seventh filter |
| Science integration time | 1 hours | Largest Cycle 1 GO extragalactic NIRCam time by nearly a factor of 2 |
| F444W 2 depth | 3–4 ABmag | 5-radius apertures |
Depth varies systematically across the wedding-cake tiers. Across the full survey, 6 point-source depths in F444W range from 7 to 8 ABmag in 9-radius circular apertures. The wide tier typically spans 42–100+ minutes total and yields F444W $1$0 depths of approximately $1$1–$1$2 ABmag, while the top 20% deepest pointings reach approximately $1$3–$1$4 ABmag and approach the depth–area combination of CEERS. For performance simulations, typical average depths were taken as $1$5, $1$6, $1$7, $1$8, $1$9, and 0 ABmag for the six-filter baseline, and 1, 2, 3, 4, 5, 6, and 7 ABmag for the seven-filter configuration.
3. Data reduction, calibration, mosaics, and catalogs
The reduction framework is built around grizli version v1.9.13.dev26. STScI stage-1 products are ingested, after which the pipeline performs artifact masking for snowballs using snowblind, 8 noise suppression, and treatment of scattered-light features such as claws and wisps. Astrometry is anchored by aligning F444W to Gaia DR3, or by bootstrapping through DESI Legacy Imaging Surveys catalogs when required; the remaining filters are then aligned to the F444W source catalog on a field-by-field basis. Mosaics are generated with astrodrizzle, with short-wavelength images sampled at 9 per pixel and long-wavelength images at 0 per pixel, and with additional modeling and subtraction of diffraction spikes at 1 and 2 relative to the detector axes (Williams et al., 2024).
Source detection uses SourceExtractor in dual mode, with the detection image defined as the inverse-variance weighted long-wavelength stack 3. Photometry is measured in circular apertures, with 4 radius adopted as fiducial, after PSF-matching all images to F444W. PSFs are derived from webbpsf on a per-field basis with jitter_sigma=0.02 and rotated to match the pointing position angle; ancillary HST PSFs are constructed empirically with photutils EPSFBuilder, and matching kernels are computed with pypher. Curve-of-growth tests show agreement between webbpsf and empirical stellar PSFs at the adopted apertures to within 5 at 6 diameter, but webbpsf underestimates F444W PSF cores by approximately 7–8 inside 9 diameter. As a result, those PSFs are not recommended for structural measurements, since sizes would be biased larger.
Total-flux estimation follows a two-step procedure. Aperture fluxes are first scaled to a Kron-like aperture on the PSF-matched detection image, then divided by the encircled energy appropriate to the F444W PSF’s Kron ellipse. Depths are estimated from the flux distributions in 5000 random, source-free 0-radius apertures per field and per filter.
Catalog quality control is explicit and field-specific. Bright point sources with F444W 1 are flagged as stars when 2, and Gaia DR3 proper-motion matches within 3 are further used to mask stars and diffraction spikes. Data-quality flags propagate zero-weight pixels into isophotal and Kron footprints, with binary dilation for conservative edge masking. Spurious detections are removed by extreme sizes, specifically 4 pixels or Kron radius 5 pixels, and by hot-pixel-like compactness with 6 pixels. Catalog inclusion requires 7 in at least one NIRCam wide filter and valid photo-8 fits. The consolidated flag use_phot=1 identifies sources with no flagged pixels in any of the long-wavelength wide filters contributing to the detection stack.
4. Photometric-redshift formalism and achieved performance
Photometric redshifts are derived with EAZY using the blue_sfhz template set, which contains 14 templates including one based on a JWST/NIRSpec spectrum of an extreme emission-line galaxy at 9. The fitting configuration adopts an error floor of 0, a redshift range of 1–2, and three iterations of internal zeropoint optimization. Typical zeropoint corrections are 3–4, with specific larger corrections in F115W for j043844m6849 at approximately 5, and in F150W and F200W for j010500p0217 at approximately 6 and 7, respectively (Williams et al., 2024).
The survey defines its photometric-redshift statistics explicitly. Let 8. Then
9
and, for outliers with threshold $2$0,
$2$1
Performance forecasts use JAGUAR mock catalogs joined across 10 realizations to improve high-$2$2 statistics, with noise assigned to match the average field depths. In those simulations, the six-filter baseline achieves $2$3 and $2$4. Adding F410M improves the results to $2$5 and $2$6. The paper also reports that at $2$7, the six-filter baseline gives photometric-redshift performance comparable to CANDELS, with $2$8 and $2$9, despite relying on NIRCam-only photometry.
The same section formalizes the photometric system and depth conversion. AB magnitudes are written as
0
with 1 in 2, while the 3 flux limit is
4
Under background-limited conditions, the signal-to-noise ratio scales approximately as 5 with integration time 6, which is the formal basis for the observed depth range across PANORAMIC’s tiers.
5. Survey geometry, cosmic variance, and scientific reach
The spatial logic of PANORAMIC is as important as its depth. Because the survey samples approximately 40 independent sightlines, cosmic variance decreases approximately as 7. For the bright end of the ultraviolet luminosity function at 8, the survey reports a factor of approximately 3 improvement in cosmic-variance uncertainty relative to an equivalent contiguous-area survey (Williams et al., 2024).
This geometry broadens the survey’s science domain across redshift. At 9–00, the six-plus-band design enables robust three-filter dropout selection and high-fidelity photometric redshifts for bright-end ultraviolet luminosity-function work, while the widely separated sightlines sharpen constraints on the star-formation-rate density at 01. At 02, the 03–04 coverage reveals red, massive, and dusty populations that are poorly captured by HST+Spitzer combinations, including dust-obscured starbursts, quiescent candidates, and “Little Red Dots,” described as abundant faint AGN candidates at 05, extending down to surprisingly low masses of 06–07.
The survey also opens significant discovery space outside traditional deep fields. The random wide-area sightlines yield strong galaxy–galaxy lenses and candidate clusters with arcs, extremely red bright sources with 08 ABmag, and 09 dropouts that are likely brown dwarfs. A plausible implication is that the survey’s scientific value is not limited to the canonical high-10 luminosity-function problem; its field distribution also enhances sensitivity to rare classes that would be undersampled in more concentrated deep programs.
In comparative terms, PANORAMIC expands the existing Cycle 1 area covered by similar six-band NIRCam data by approximately 11. Its deep-tier area and depth resemble CEERS, while its full coverage at the shallow-tier depth of approximately 12 mag in the weighted long-wavelength stack establishes a wide-tier baseline for NIRCam imaging at 13–14. This combination of area, tiering, and filter multiplicity gives the program its specific legacy role within the first JWST imaging cycle.
6. Limitations, data release, and operational use
The survey is affected by several operational and calibration constraints. Science pure-parallels began only in December 2022, compressing the schedule for Cycle 1. A minimum exposure time per filter required at least four groups for ramp fitting, equivalent to 15 s, and data-volume limits capped long NIRCam pure-parallel integrations at approximately 16 hours per pointing. Some fields were observed with only four filters because of software issues in March–April 2023, and those fields were excluded from DR1 catalogs. Orientation changes caused by occasional guide-star failures reduced common filter overlap in some associations, while NIRSpec MSA shorts corrupted some parallel imaging, particularly in GOODS-N. Additional systematics include wisps, claws, and persistence from Saturn, with masking or omission from DR1 where deeper uncorrupted data existed (Williams et al., 2024).
DR1 releases science-ready mosaics in each NIRCam filter for novel-area associations, photometric catalogs per association, weight and flag maps from grizli, and fitsmap interactive viewers. Associations are named by coordinates in the form j[RA]p/m[Dec]. For associations entirely overlapping deeper data, such as CEERS/EGS, COSMOS-PRIMER, and UDS-PRIMER, PANORAMIC imaging is merged into DAWN JWST Archive releases rather than duplicated, although COSMOS-Web footprints in which PANORAMIC provides the longest on-sky integration are co-added and released.
The recommended usage protocol is correspondingly explicit. Source finding should be performed on the long-wavelength detection stack; photometry should use PSF-matched 16-radius aperture fluxes with the catalog total-flux corrections; use_phot=1 should be applied for robust subsets; Gaia DR3 proper-motion matches within 17 should be removed to exclude stars and spikes; and the 18 and Kron-radius flags should be respected when rejecting spurious detections. For structural analyses, empirical PSFs from similar datasets are preferred over the PANORAMIC webbpsf products because of the F444W core discrepancy.
Taken as a survey system rather than a collection of fields, PANORAMIC demonstrates that pure parallel NIRCam imaging can deliver wide-area legacy coverage, an ultra-deep sub-tier, and CANDELS-like photometric-redshift performance using NIRCam alone. Its principal significance lies in combining large-area sampling, six-plus-band 19–20 photometry, and minimized cosmic variance into a single Cycle 1 framework that extends the statistical basis for studies of rare bright galaxies, dusty and quiescent populations, and unexpected luminous outliers across cosmic time.