JADES DR5: JWST Extragalactic Data Release

• JADES DR5 is an extragalactic data release providing fully calibrated, ultra-deep infrared imaging from JWST and HST across 35 space-based filters.
• Advanced processing pipelines, including non-negative matrix factorization for wisp subtraction, ensure robust artifact correction and photometric accuracy.
• The release underpins studies on high-redshift galaxy evolution, structural analysis, and SED modeling from Cosmic Dawn to the epoch of reionization.

The JWST Advanced Deep Extragalactic Survey (JADES) Data Release 5 (DR5) represents a landmark public data release in extragalactic observational astronomy, delivering ultra-deep, wide-field infrared imaging and derived catalogs for the GOODS-South and GOODS-North legacy fields. This release provides fully calibrated mosaics and catalogs in 35 space-based filters spanning 0.43–25 μm, sourced from JWST/NIRCam, JWST/MIRI, HST/ACS, and HST/WFC3, constructed from ≳1250 hours of JWST imaging and including reductions from 19 external JWST programs. Comprehensive pipeline innovations, advanced artifact removal (notably wisp subtraction via non-negative matrix factorization, NMF), and robust catalog methodologies underpin its utility for studying high-redshift galaxy populations, structural evolution, and cosmological processes from Cosmic Dawn to the epoch of reionization (Alberts et al., 22 Jan 2026, Johnson et al., 22 Jan 2026, Robertson et al., 22 Jan 2026, Carreira et al., 22 Jan 2026, Wu et al., 22 Jan 2026, Hainline et al., 22 Jan 2026).

1. Imaging Footprint and Filter Coverage

JADES DR5 covers 469 arcmin² with at least one NIRCam filter (245 arcmin² in GOODS-S, 224 arcmin² in GOODS-N), and 250 arcmin² with full coverage in the eight “core JADES” NIRCam filters (F090W–F444W). A total of up to 18 NIRCam wide and medium bands (SW: F070W–F210M, LW: F250M–F480M) are included, with MIRI imaging obtained in F770W, F1280W, and F1500W. The ultra-deep MIRI “GS-Deep” field achieves 155 ks exposure (F770W, 7.7 μm) over 10.4 arcmin², reaching a 5σ depth of 19 nJy (AB 28.2). Medium-depth MIRI parallels in GOODS-S/N cover 36 arcmin² (F770W), 25 arcmin² (F1280W), and 22 arcmin² (F1500W). Imaging depths vary by band and region, with NIRCam wide filters reaching AB ∼29.6–30.1 mag (for 0.1″ radius, 5σ) in the deepest regions (Johnson et al., 22 Jan 2026, Alberts et al., 22 Jan 2026, Robertson et al., 22 Jan 2026).

2. Data Processing and Calibration Pipelines

Reduction pipelines employ the official JWST Calibration Pipeline (v1.14.0–1.16.1), augmented by bespoke JADES procedures. For NIRCam, steps include ramp fitting, crosstalk subtraction using amplifier coefficients, augmented bad-pixel and hot-pixel masking from on-orbit darks, persistence and saturation masking, and “sky-flat” creation for LW channels via deep source-masked median stacks. 1/f noise removal employs amplifier row/column/pedestal fitting. In MIRI, custom warm/hot pixel detection (>5σ from the local median or above dark threshold) and advanced background subtraction (super-background median construction, row/column filtering, low-order polynomial removal) improve uniformity. Persistence is modeled by propagating saturated pixels (Ngood < 3) forward in time across all filters and flagging as DO_NOT_USE in all subsequent exposures. Astrometric alignment is anchored to Gaia DR3 with sub-10 mas accuracy (Johnson et al., 22 Jan 2026, Alberts et al., 22 Jan 2026).

3. Artifact Correction: Wisp Subtraction via NMF

“Wisps”—scattered-light artifacts intrinsic to JWST/NIRCam—are modeled and subtracted via a non-negative matrix factorization (NMF) algorithm, replacing template-based approaches. Each exposure (after flat-fielding, background subtraction, 1/f correction, and source masking) is arranged in a data matrix $X\in \mathbb{R}^{m\times n}$; NMF seeks $W\ge0$, $H\ge0$ minimizing $\tfrac12 \|\Omega \odot (X-WH)\|_F^2$ with $\Omega$ the inverse-variance mask. Multi-component, filter- and detector-specific templates (up to $k=3$) capture both dominant and variable wisp structures. Iterative amplitude fitting, 1/f correction, and bias correction converge within 500 iterations (downsampled and full-resolution). Compared to previous STScI median-template subtraction, NMF reduces excess RMS in wisp-affected regions by a factor of 2–3 and suppresses photometric bias to ≲1%. All wisp templates, subtraction code, and documentation are public and intended for use in Stage 2 pipelines prior to coaddition or photometry (Wu et al., 22 Jan 2026).

4. Photometric and Morphological Catalog Construction

Source detection and deblending use signal-to-noise (S/N) stacks for both long- and short-wavelength NIRCam bands, refined via photutils segmentation, iso-significance contour deblending, and 2D Gaussian regression for centroids and shape moments. Forced circular-aperture photometry (several radii plus an 80% encircled energy PSF aperture, “CIRC0”) and ellipsoidal Kron photometry ($k=2.5$ and “short” $k=1.4$) are performed for every source in every band. Empirical curve-of-growth metrics are provided. Photometric uncertainties employ a pixel-level regression, combining per-pixel RMS “uncertainty mosaics” (to capture correlated noise scaling $\sigma(N)=\sigma_1 N^\beta$, $1\le\beta\le2$) with regression for aperture size, shape, and local background.

Morphological parameters are derived via Bayesian inference of single-component Sérsic profiles per filter and field using the No-U-Turn Sampler (NUTS) in NumPyro, median posterior as the best-fit, and full posterior samples for uncertainty estimation. Subsets with improved fits in F444W undergo bulge–disk decomposition, with bulge ($n_b$ free) and disk ($n_d=1$ fixed) components. Catalogs record positions, Sérsic index $n$, effective radius $R_\mathrm{eff}$, axis ratio, position angle, fluxes, and fit quality (Robertson et al., 22 Jan 2026, Carreira et al., 22 Jan 2026).

5. Catalog Contents and Data Access

JADES DR5 releases all products as High-Level Science Products (HLSPs) via MAST (https://archive.stsci.edu/hlsp/jades, DOI: 10.17909/8tdj-8n28). Imaging mosaics are distributed per filter, subregion, and full field, with multi-extension FITS arrays (SCI, ERR, WHT, EXP, NIM). The main photometric catalog supplies ∼500,000 sources with circular and Kron photometry, curve-of-growth data, morphological fits, photometric redshifts (EAZY, both small-aperture and Kron), source positions and shapes, program IDs and image provenance (PID_HASH), and per-filter exposure/quality flags. Morphological catalogs supply posterior samples, best-fits, and per-field/filter multi-extension tables. Separate catalogs are provided for photometrically selected $z_\mathrm{phot}>8$ galaxy candidates, with 2081 objects (GOODS-S: 1597, GOODS-N: 484), including UV slopes, effective radii, and spectroscopic cross-matches. Wisp templates, masks, and subtraction software are separately released for reproducibility (Robertson et al., 22 Jan 2026, Alberts et al., 22 Jan 2026, Hainline et al., 22 Jan 2026, Wu et al., 22 Jan 2026, Carreira et al., 22 Jan 2026).

6. Scientific Results and Applications

DR5 enables rest-frame near-IR and optical studies from $z\sim0$ to $z\gtrsim15$. MIRI imaging supplies the critical $\lambda_\mathrm{obs}>5\,\mu\mathrm{m}$ regime: e.g., GS-Deep F770W secured the first photometric detection of Hβ+[O III] boosting in a $z=14.18$ galaxy and upper limits at $z=13.86$, essential for nebular EW and stellar population modeling. For $z\sim3$–6, F770W provides rest-frame J-band, empirically breaking age–dust degeneracies in massive early galaxy samples.

Morphological analysis across 24,692 galaxies with $1$r_\mathrm{eff}\propto(1+z)^{-0.635\pm0.013}$ kpc (rest-frame optical, $\lambda\approx5000$ Å), with surface luminosity densities within 1 kpc ($\Sigma_{1\,\mathrm{kpc}}$) remaining constant with redshift—implying size–core luminosity scaling offsets the $(1+z)^4$ dimming. Bulge–disk decompositions (8390 HUDF galaxies) demonstrate $$r_{\mathrm{eff,bulge}} \propto (1+z)^{-0.234\pm0.030}$$ kpc and $$r_{\mathrm{eff,disk}} \propto (1+z)^{-1.091\pm0.043}$$ kpc, with a bulk population of $B/T > 0.6$ at all redshifts. Among $z_\mathrm{phot}>8$ candidates, 27% are extended (≳1 kpc), with many showing multi-knot or clumpy morphologies out to $z\sim13$–14, indicating early assembly processes.

SED modeling confirms that deep NIRCam can recover stellar masses/SFRs to within factors of ∼2 for typical $z\sim8$ galaxies, but MIRI coverage is necessary for robust fitting of dusty galaxies and accurate emission-line constraints in the epoch of reionization (Alberts et al., 22 Jan 2026, Carreira et al., 22 Jan 2026, Hainline et al., 22 Jan 2026).

7. Limitations and Considerations

Fit failures affect ∼10% of sources, principally near detector edges or for highly extended, low-SNR profiles. Single-component Sérsic models inadequately capture morphologically complex or clumpy systems. Users are advised to impose rest-frame selection and mass completeness for structural/demographic studies and to check morphological fit diagnostics (e.g., χ², p-value, total flux recovery). Wisp subtraction via NMF is robust, but rare “bright wisp” events and “claws” (star-induced stray light) require additional masking or custom modeling. Systematic uncertainties in MIRI PSF matching and local backgrounds are controlled to ≲5%.

Catalogs omit stellar-mass and color/SF selection; future multivariate analyses will require external inputs for full population studies. Detailed documentation is provided in Alberts et al. (2025) and Robertson et al. (2026) (Alberts et al., 22 Jan 2026, Carreira et al., 22 Jan 2026).

---

JADES DR5 represents a foundational resource, delivering deep, uniform, artifact-corrected imaging and catalogs for extragalactic science at the highest current sensitivity and uniformity, underpinned by extensive public access and documentation (Johnson et al., 22 Jan 2026, Robertson et al., 22 Jan 2026, Alberts et al., 22 Jan 2026, Carreira et al., 22 Jan 2026, Wu et al., 22 Jan 2026, Hainline et al., 22 Jan 2026).