DECADE Shear Catalogs Overview

• The DECADE shear catalogs are calibrated, weak-lensing shape datasets derived from heterogeneous public DECam imaging for robust cosmic shear inference.
• They employ DES-style pipelines with stringent quality cuts, Metacalibration, and NGMix to ensure accurate shear measurements and reliable redshift calibration.
• The methodology rigorously validates inhomogeneity effects through end-to-end simulations, subset stress tests, and analytic covariance checks to mitigate systematic biases.

The DECADE shear catalogs are the calibrated weak-lensing shape catalogs of the Dark Energy Camera All Data Everywhere program, built from public DECam imaging in regions outside the Dark Energy Survey footprint and designed for cosmology-ready cosmic shear analyses. In the northern Galactic cap, the foundational catalog contains 107,371,478 galaxies over $5{,}412\,\mathrm{deg}^2$, with an effective number density of $4.59\,\mathrm{arcmin}^{-2}$, and it is processed with the Dark Energy Survey Data Management pipeline as part of the DECADE campaign that underlies DELVE Early Data Release 3 (Anbajagane et al., 24 Feb 2025). Subsequent DECADE analyses treat these catalogs as the core observational input for cosmic shear inference, photometric-redshift calibration, inhomogeneity stress testing, and likelihood-level combinations with DES Year 3 and, later, with a newly added southern Galactic cap sample (Anbajagane et al., 24 Feb 2025, Anbajagane et al., 24 Feb 2025, Anbajagane et al., 24 Feb 2025, Anbajagane et al., 3 Sep 2025).

1. Definition, scope, and survey context

DECADE stands for Dark Energy Camera All Data Everywhere. Its central idea is to exploit the full public DECam archive rather than only a dedicated survey footprint, and to process that archive consistently with the DES Year 6 image-processing pipeline or, in the initial catalog paper, with the DES Data Management infrastructure used in DES (Anbajagane et al., 24 Feb 2025, Anbajagane et al., 24 Feb 2025). The DECADE shear catalogs therefore differ from canonical Stage-III lensing products in origin: they are assembled from public DECam data including survey and standard observing programs, or, equivalently, from community archival DECam imaging collected up to December 2022 (Anbajagane et al., 24 Feb 2025, Anbajagane et al., 24 Feb 2025).

This archival origin determines the catalogs’ defining observational property: substantial spatial inhomogeneity. The imaging combines many observing programs with different depths, seeing, airmass, extinction, sky background, exposure time, and related conditions (Anbajagane et al., 24 Feb 2025). The DECADE cosmic shear analysis explicitly characterizes the dataset as an amalgamation of community-led DECam imaging campaigns in the northern Galactic cap rather than a purpose-built weak-lensing survey, and treats the resulting heterogeneity as a stress test for Stage-III shear-calibration and cosmology pipelines (Anbajagane et al., 24 Feb 2025).

The relationship between DECADE and DELVE is operational rather than merely nominal. The DECADE campaign is the consistent processing effort, DELVE EDR3 is the public imaging/catalog product generated from that processing, and the DECADE shear catalog is the weak-lensing shape catalog built on top of those products (Anbajagane et al., 24 Feb 2025). The full DELVE EDR3 footprint comprises about $8{,}500\,\mathrm{deg}^2$ in the north Galactic cap, but after lensing-specific cuts the shear sample is reduced to $5{,}412\,\mathrm{deg}^2$ (Anbajagane et al., 24 Feb 2025).

A recurrent implication in the DECADE series is that image-depth variation and source-density variation are not identical. The papers emphasize that, although the underlying imaging is highly heterogeneous, the final source sample is selected so that its number-density pattern is substantially more controlled. In particular, a cut of $m_i<23.5$ is stated to reduce the strongest depth-driven nonuniformity in the source catalog (Anbajagane et al., 24 Feb 2025). This suggests that DECADE’s main methodological challenge is not only raw observing-condition variability, but the need to propagate its effect consistently through selection, shear calibration, redshift calibration, and covariance estimation.

2. Catalog construction and shape-measurement formalism

The shear catalogs are built from a base object catalog derived from SourceExtractor detections on coadds, followed by cuts analogous to DES Gold selection and then Metacalibration-based lensing-quality cuts (Anbajagane et al., 24 Feb 2025). The base selection includes SourceExtractor flag cuts, removal of failed Fitvd measurements, masking around 2MASS stars, Gaia eDR3 stars, Yale bright star catalog objects, BSCP5 very bright stars, HyperLEDA galaxies, and Harris globular clusters, plus additional masks for high stellar density and high extinction (Anbajagane et al., 24 Feb 2025). After these cuts, the base catalog contains 470,812,637 objects over about $5{,}750\,\mathrm{deg}^2$ (Anbajagane et al., 24 Feb 2025).

The weak-lensing shapes are measured with Metacalibration, using only the $r,i,z$ bands; the $g$ band is dropped because of PSF modeling concerns (Anbajagane et al., 24 Feb 2025). Shapes are measured with NGMix by fitting an elliptical Gaussian convolved with the PSF model, and Metacalibration estimates the shear response by artificially shearing each object’s image at the pixel level and remeasuring the shapes on five versions of the image: no shear, $1+$, $1-$, $4.59\,\mathrm{arcmin}^{-2}$0, and $4.59\,\mathrm{arcmin}^{-2}$1, with $4.59\,\mathrm{arcmin}^{-2}$2 in each component (Anbajagane et al., 24 Feb 2025).

The response formalism is given explicitly as

$4.59\,\mathrm{arcmin}^{-2}$3

with a corresponding selection response

$4.59\,\mathrm{arcmin}^{-2}$4

and total response

$4.59\,\mathrm{arcmin}^{-2}$5

The catalog follows the DES Y3 convention of using the diagonal response and averaging the diagonal terms for the scalar shear calibration used in cosmology (Anbajagane et al., 24 Feb 2025).

The main lensing-quality cuts are

$4.59\,\mathrm{arcmin}^{-2}$6

$4.59\,\mathrm{arcmin}^{-2}$7

$4.59\,\mathrm{arcmin}^{-2}$8

and

$4.59\,\mathrm{arcmin}^{-2}$9

followed by photometric and color cuts

$8{,}500\,\mathrm{deg}^2$0

$8{,}500\,\mathrm{deg}^2$1

Obvious stars and binary systems are rejected using

$8{,}500\,\mathrm{deg}^2$2

together with the size–magnitude boundary

$8{,}500\,\mathrm{deg}^2$3

These cuts are described as necessary both to remove contaminants and to make the sample and its redshift calibration more stable (Anbajagane et al., 24 Feb 2025).

For the full sample, Table 2 of the catalog paper reports raw number density $8{,}500\,\mathrm{deg}^2$4, $8{,}500\,\mathrm{deg}^2$5, $8{,}500\,\mathrm{deg}^2$6, and shape noise $8{,}500\,\mathrm{deg}^2$7 (Anbajagane et al., 24 Feb 2025). The paper often summarizes this as an effective number density of about $8{,}500\,\mathrm{deg}^2$8–$8{,}500\,\mathrm{deg}^2$9, depending on definition (Anbajagane et al., 24 Feb 2025). Shear weights are assigned as

$5{,}412\,\mathrm{deg}^2$0

with the catalog binned in 20 logarithmic bins of size ratio $5{,}412\,\mathrm{deg}^2$1 and 20 logarithmic bins in SNR (Anbajagane et al., 24 Feb 2025).

3. Photometric-redshift calibration and tomographic structure

The DECADE shear catalogs are cosmologically usable only in conjunction with a redshift-calibration pipeline that matches the weighted source sample entering the shear correlations. The DECADE photometric-redshift paper therefore treats the shape catalog as fixed and calibrates the redshift side of the weak-lensing analysis (Anbajagane et al., 24 Feb 2025). It states that the analysis uses the galaxy fluxes and flux errors in the $5{,}412\,\mathrm{deg}^2$2 bands from Metacalibration, and propagates the same shear weights into redshift calibration through

$5{,}412\,\mathrm{deg}^2$3

where $5{,}412\,\mathrm{deg}^2$4 is the shear weight and $5{,}412\,\mathrm{deg}^2$5 is the Metacalibration shear response (Anbajagane et al., 24 Feb 2025).

The redshift calibration uses three linked ingredients: the wide-field DECADE source catalog; a deeper, lower-noise multi-band sample from the DES Y3 deep fields using $5{,}412\,\mathrm{deg}^2$6; and a redshift sample combining COSMOS, PAUS+COSMOS, C3R2, VVDS, and zCOSMOS (Anbajagane et al., 24 Feb 2025). After quality cuts and requiring detection in the wide data at least once, the fiducial deep-field sample has 178,301 galaxies (Anbajagane et al., 24 Feb 2025).

The transfer function between deep and wide photometry is estimated with a Balrog-style synthetic source injection catalog built specifically for DECADE. Deep-field galaxies are injected into real DECADE imaging, the images are reprocessed through the full pipeline, and the result estimates the probability of how a galaxy observed in the deep data would appear in the noisier wide field (Anbajagane et al., 24 Feb 2025). In the paper’s summary, this is the bridge that allows deep-field color–redshift information to be transferred probabilistically to the actual DECADE source sample (Anbajagane et al., 24 Feb 2025).

The adopted methodology is SOMPZ, Self-Organizing Maps Photometric Redshifts. The paper uses a wide SOM trained on DECADE wide $5{,}412\,\mathrm{deg}^2$7 fluxes and flux errors with size $5{,}412\,\mathrm{deg}^2$8, and a deep SOM trained on deep $5{,}412\,\mathrm{deg}^2$9 photometry with size $m_i<23.5$0 (Anbajagane et al., 24 Feb 2025). Wide cells are ranked by their median redshift, estimated from Balrog galaxies whose deep counterpart has redshift information, and then grouped into four bins of roughly equal galaxy counts (Anbajagane et al., 24 Feb 2025). The tomographic bin edges are

• Bin 1: $m_i<23.5$1–0.381
• Bin 2: 0.381–0.619
• Bin 3: 0.619–0.803
• Bin 4: 0.803–2.0 (Anbajagane et al., 24 Feb 2025)

The mean redshifts of the SOMPZ $m_i<23.5$2 distributions are reported as

$m_i<23.5$3

with fiducial SOMPZ-only mean redshifts 0.3332, 0.5013, 0.7064, and 0.9056 for bins 1 through 4 (Anbajagane et al., 24 Feb 2025). The total uncertainties on the mean redshifts are stated as $m_i<23.5$4, with dominant contributions from sample variance, zeropoint calibration uncertainty, and redshift bias (Anbajagane et al., 24 Feb 2025).

An independent validation is provided by clustering redshifts obtained by cross-correlating the source galaxies with BOSS/eBOSS reference samples: CMASS, LOWZ, LRG, ELG, and QSO (Anbajagane et al., 24 Feb 2025). The paper concludes that the SOMPZ-based mean $m_i<23.5$5 realizations are consistent with the WZ measurements; all SOMPZ realizations pass the $m_i<23.5$6 consistency threshold; and the agreement holds at about $m_i<23.5$7 robustness level, which is much larger than the quoted mean-redshift uncertainty and therefore rules out catastrophic redshift failures (Anbajagane et al., 24 Feb 2025).

4. Systematics characterization and shear calibration

The DECADE shear-catalog program combines null tests on the data with end-to-end image simulations. The main catalog paper states that the coadd images used to derive the data have median limiting magnitudes of $m_i<23.5$8, $m_i<23.5$9, and $5{,}750\,\mathrm{deg}^2$0, estimated at $5{,}750\,\mathrm{deg}^2$1 in a 2 arcsecond aperture (Anbajagane et al., 24 Feb 2025). It also emphasizes that the depth distribution is much broader than DES Y3: about 10% of the DECADE area is deeper than the 99th-percentile depth of DES Y3 (Anbajagane et al., 24 Feb 2025). These facts motivate extensive validation against PSF, DCR, survey-property, and parity tests.

PSF modeling is mapped across the DECam focal plane. The paper reports large-scale interpolation residuals in ellipticity and CCD-scale “tree ring” features in the size residuals, consistent with instrumental effects and deemed small (Anbajagane et al., 24 Feb 2025). Brightness-dependent PSF tests find residual size errors at the $5{,}750\,\mathrm{deg}^2$2–$5{,}750\,\mathrm{deg}^2$3 level, corresponding to additive shear contamination at only $5{,}750\,\mathrm{deg}^2$4–$5{,}750\,\mathrm{deg}^2$5 level (Anbajagane et al., 24 Feb 2025). Color-dependence tests show that PSF size varies by less than about 1% across the relevant color range and the PSF ellipticity error remains below $5{,}750\,\mathrm{deg}^2$6, which the authors treat as negligible for cosmology (Anbajagane et al., 24 Feb 2025).

The catalog paper also measures mean shear as a function of SNR, size ratio, PSF properties, and color using 20 percentile bins and linear regressions. Most slopes are consistent with zero within $5{,}750\,\mathrm{deg}^2$7, with the strongest trend being shear versus color, described as consistent with DCR expectations (Anbajagane et al., 24 Feb 2025). Tangential shear around bright and faint stars, and around field centers, is consistent with zero, with quoted $5{,}750\,\mathrm{deg}^2$8-values $5{,}750\,\mathrm{deg}^2$9, $r,i,z$0, and $r,i,z$1, respectively (Anbajagane et al., 24 Feb 2025). $r,i,z$2-modes measured using the HybridEB estimator over $r,i,z$3 to $r,i,z$4 are consistent with null with $r,i,z$5 (Anbajagane et al., 24 Feb 2025).

PSF two-point diagnostics are summarized through Rowe statistics $r,i,z$6 and Tau statistics $r,i,z$7. The PSF-related contamination model is written as

$r,i,z$8

with fitted coefficients

$r,i,z$9

and $g$0, which the paper interprets as consistent with no significant extra PSF contamination (Anbajagane et al., 24 Feb 2025).

The multiplicative shear bias is calibrated with an image-simulation pipeline that selects 1600 coadd tiles, injects synthetic galaxies and stars, renders images with GalSim, convolves with the PSFEx model, adds noise, background, and masks from the real data, processes the simulations through the full DESDM pipeline, and runs SourceExtractor and Metacalibration on the simulated outputs (Anbajagane et al., 24 Feb 2025). Synthetic galaxies come from DES deep fields overlapping COSMOS, with HST-based morphology models and DES photometry; stars are drawn from a TRILEGAL-based LSST stellar catalog converted to DES bands (Anbajagane et al., 24 Feb 2025).

The bias model is

$g$1

and the final calibration uses a variance-cancellation estimator with simulations at $g$2 (Anbajagane et al., 24 Feb 2025). The tomographic multiplicative biases reported in Table 3 are

• Bin 1: $g$3
• Bin 2: $g$4
• Bin 3: $g$5
• Bin 4: $g$6
• Full sample: $g$7 (Anbajagane et al., 24 Feb 2025)

The full-sample additive bias is reported as

$g$8

A subtle residual PSF-size mismatch is also explicitly tested: the model PSF is on average about 0.2% larger than the stellar PSF, and the paper finds it would induce $g$9, which is within the final calibration uncertainty (Anbajagane et al., 24 Feb 2025).

A later cross-survey reanalysis uses TXPipe to apply a uniform suite of diagnostics to KiDS-1000, DES-Y3, and HSC-Y3, rather than to DECADE directly, but its conclusions provide context for interpreting shear-catalog validation practice more broadly (Jefferson et al., 6 May 2025). That paper emphasizes that null tests should not be treated as ad hoc or purely binary pass/fail checks and recommends a standardized, metadata-rich, and cosmology-aware diagnostic framework for future surveys (Jefferson et al., 6 May 2025). This suggests a methodological backdrop against which the DECADE validation program can be understood: not as the elimination of all residuals, but as the demonstration that residuals are subdominant for the intended cosmological analyses.

5. Inhomogeneity stress tests and end-to-end pipeline robustness

A distinctive feature of the DECADE catalog series is that inhomogeneity is treated as a first-order scientific issue rather than only as background survey characterization. The pipeline-validation paper states that the DECADE dataset is highly heterogeneous in observing conditions, but that the final source sample has a more controlled number-density pattern after the $1+$0 cut (Anbajagane et al., 24 Feb 2025). The central question is therefore whether standard cosmic shear inference remains reliable once all selection-dependent ingredients are recomputed for the actual subset of data under study (Anbajagane et al., 24 Feb 2025).

The robustness program is fully end-to-end. For each subset, the analysis re-derives shear weights and response grids, recomputes multiplicative shear bias using the existing image simulations but with the subset-specific selection function, reselects galaxies into tomographic bins, recomputes the subset’s $1+$1 with SOMPZ, re-estimates the Balrog transfer function, recomputes $1+$2 and $1+$3, updates the survey geometry and mask, generates a new analytic covariance with CosmoCov, remeasures the full shear two-point data vector, and reruns the cosmological likelihood analysis (Anbajagane et al., 24 Feb 2025). This is one of the most specific methodological claims in the DECADE series: the pipeline is not merely applied to fixed subcatalogs, but rebuilt consistently for each split sample (Anbajagane et al., 24 Feb 2025).

The paper defines 23 split criteria, producing 46 subsets total because each split is analyzed both for the selected half and its complement (Anbajagane et al., 24 Feb 2025). The survey-property splits include airmass, seeing or PSF width, DCR-related quantities, east versus west sky, $1+$4, sky background brightness, sky background variation, exposure time, magnitude limit, number of exposures, extinction coefficients, and Gaia stellar density (Anbajagane et al., 24 Feb 2025). Additional object-property splits include color $1+$5, magnitude in $1+$6, signal-to-noise, object size, PSF size, and object-to-PSF size ratio (Anbajagane et al., 24 Feb 2025).

Consistency in $1+$7 is quantified using

$1+$8

where the two measurements are from either a subset and its complement, or a subset and the fiducial full catalog (Anbajagane et al., 24 Feb 2025). The principal result is that all 46 subset analyses are consistent with the fiducial cosmology within $1+$9, most are within $1-$0, no split shows a shift larger than expected from shape noise and cosmic variance, and subset-versus-complement comparisons also remain within $1-$1 in all cases (Anbajagane et al., 24 Feb 2025).

The same paper validates covariance modeling for inhomogeneous data. It reports that the analytic shape-noise covariance matches simulation-based estimates to within 5%, both for the actual inhomogeneous galaxy positions and for a version with randomized homogeneous positions (Anbajagane et al., 24 Feb 2025). It also states that, without scale cuts, baryonic contamination shifts the constraints by $1-$2 in the $1-$3–$1-$4 plane, whereas after applying fiducial scale cuts the shift drops to $1-$5, below the target threshold of $1-$6 (Anbajagane et al., 24 Feb 2025). The paper further notes that switching from HMCode to HaloFit causes only a small shift, using NLA instead of TATT slightly tightens constraints by about 8%, fixing IA parameters improves the $1-$7 uncertainty by about 33%, and widening the redshift-prior uncertainty increases the $1-$8 uncertainty by about 20% (Anbajagane et al., 24 Feb 2025).

These results support the series’ broader claim that existing weak-lensing analysis methods can be fairly resilient toward inhomogeneous datasets (Anbajagane et al., 24 Feb 2025). A plausible implication is that DECADE’s significance is methodological as well as cosmological: it converts a heterogeneous archive into a controlled benchmark for testing which components of the Stage-III cosmic shear pipeline are genuinely fragile and which are robust after consistent recalibration.

6. Cosmological deployment, extensions, and public release

The northern DECADE shear catalog is used as the observational basis for the DECADE cosmic shear analysis over $1-$9, with four tomographic bins and redshift distributions from SOMPZ, cross-checked with a WZ method (Anbajagane et al., 24 Feb 2025). The catalog-level observable entering the cosmological analysis is the two-point correlation function

$4.59\,\mathrm{arcmin}^{-2}$00

measured in tomographic and angular bins after scale cuts (Anbajagane et al., 24 Feb 2025). The likelihood is Gaussian,

$4.59\,\mathrm{arcmin}^{-2}$01

with covariances generated using CosmoCov and inference performed with CosmoSIS, using Nautilus as the fiducial sampler and Polychord as a cross-check (Anbajagane et al., 24 Feb 2025).

For $4.59\,\mathrm{arcmin}^{-2}$02CDM, the DECADE-only result is

$4.59\,\mathrm{arcmin}^{-2}$03

with $4.59\,\mathrm{arcmin}^{-2}$04 (Anbajagane et al., 24 Feb 2025). In $4.59\,\mathrm{arcmin}^{-2}$05CDM, the DECADE-only result is

$4.59\,\mathrm{arcmin}^{-2}$06

which the paper states is consistent with $4.59\,\mathrm{arcmin}^{-2}$07 and provides only a weak upper-bound constraint on $4.59\,\mathrm{arcmin}^{-2}$08 (Anbajagane et al., 24 Feb 2025). The DECADE-only and reanalyzed DES Y3 constraints agree at the $4.59\,\mathrm{arcmin}^{-2}$09 level, and DECADE is statistically consistent with Planck at about $4.59\,\mathrm{arcmin}^{-2}$10 in $4.59\,\mathrm{arcmin}^{-2}$11CDM and $4.59\,\mathrm{arcmin}^{-2}$12 in the combined analysis (Anbajagane et al., 24 Feb 2025).

A later DECADE analysis adds a southern Galactic cap sample. In that work, the DECADE shear catalogs are explicitly described as the core shape-measurement data product from the Dark Energy Camera All Data Everywhere cosmic shear project (Anbajagane et al., 3 Sep 2025). The analysis combines three cosmic shear data vectors: DECADE NGC, DECADE SGC, and DES Year 3 (Anbajagane et al., 3 Sep 2025). The quoted survey sizes are $4.59\,\mathrm{arcmin}^{-2}$13 for DECADE NGC, $4.59\,\mathrm{arcmin}^{-2}$14 for DECADE SGC, and $4.59\,\mathrm{arcmin}^{-2}$15 for DES Y3, for a total of about $4.59\,\mathrm{arcmin}^{-2}$16 and 270 million galaxies (Anbajagane et al., 3 Sep 2025). Because the footprints are independent sky patches, the combination is performed at the likelihood level without cross-covariance terms between surveys (Anbajagane et al., 3 Sep 2025).

The fiducial $4.59\,\mathrm{arcmin}^{-2}$17CDM result from the combined DECam 13k dataset is

$4.59\,\mathrm{arcmin}^{-2}$18

which the paper compares to Planck 2018 and describes as consistent within $4.59\,\mathrm{arcmin}^{-2}$19 in the $4.59\,\mathrm{arcmin}^{-2}$20–$4.59\,\mathrm{arcmin}^{-2}$21 plane (Anbajagane et al., 3 Sep 2025). The paper also studies six different models for baryonic suppression of the matter power spectrum—BCEmu, Bacco, HMx, and BaryonForge implementations of these models—and concludes that current scale-cut approaches leave a residual bias of about $4.59\,\mathrm{arcmin}^{-2}$22 in the $4.59\,\mathrm{arcmin}^{-2}$23 posterior, while using all scales with baryon modeling yields only a $4.59\,\mathrm{arcmin}^{-2}$24 improvement in the figure of merit (Anbajagane et al., 3 Sep 2025). The stated interpretation is that additional non-lensing datasets, and/or calibrations of the baryon model, will be required to access the full statistical power of the lensing measurements (Anbajagane et al., 3 Sep 2025).

The public-release status of the catalogs is also explicit. The combined analysis states that all DECADE cosmic shear data products are publicly accessible; the data vectors and likelihoods are available in the cosmosis-standard-library; the data vectors are also hosted on the project’s public website; the shear catalogs are hosted by NOIRLab; and the DECADE shear database is slated for public online access at

$4.59\,\mathrm{arcmin}^{-2}$25

with other DELVE/DECADE products available through DELVE DR3 (Anbajagane et al., 3 Sep 2025).

An objective summary of the series therefore has two parts. First, DECADE shear catalogs are DES-style, Metacalibration-based weak-lensing catalogs built from public DECam imaging and validated for cosmological use despite substantial survey inhomogeneity (Anbajagane et al., 24 Feb 2025, Anbajagane et al., 24 Feb 2025). Second, the later DECADE program uses these catalogs not only to obtain competitive $4.59\,\mathrm{arcmin}^{-2}$26 constraints, but also to argue that data previously categorized as “unusable” for cosmic shear analyses can be incorporated if calibration, redshift characterization, covariance validation, and systematic stress testing are sufficiently rigorous (Anbajagane et al., 24 Feb 2025).