Cosmicflows-4 Catalog Overview

• Cosmicflows-4 Catalog is a heterogeneous extragalactic distance compilation that unifies eight methods via Bayesian overlap-calibration anchored by Cepheid, TRGB, and maser distances.
• It organizes 55,877 galaxies into 38,065 groups, using group averaging to reduce random errors and improve cross-method consistency in peculiar-velocity studies.
• The catalog facilitates precise velocity-field analyses by integrating diverse techniques to refine the cosmic distance ladder and yield improved Hubble parameter estimates.

The Cosmicflows-4 Catalog (CF4) is a heterogeneous extragalactic distance compilation designed to place redshift-independent distances, galaxy grouping, and peculiar-velocity inference on a common scale for the nearby Universe. In its core release, CF4 compiles distances for 55,877 galaxies gathered into 38,065 groups, reaches to $z=0.1$ in the CMB frame, and combines eight distance methodologies within a Bayesian overlap-calibration framework anchored by Cepheid, tip of the red giant branch, and maser distances (Tully et al., 2022).

1. Definition, scope, and catalog architecture

CF4 is organized as a catalog for velocity-field work rather than as a single-method survey. Its central purpose is to map deviations from pure Hubble expansion, so the catalog stores both distance information and the metadata required to move consistently between heliocentric, Local Sheet, and CMB frames. The release tabulates method-specific distance moduli and uncertainties, a combined MCMC-merged distance modulus and uncertainty, celestial, Galactic, and supergalactic coordinates, and group assignments. The grouping scheme is not auxiliary bookkeeping: it is a structural element of the catalog, because group averages reduce random distance errors and increase method overlap for cross-calibration (Tully et al., 2022).

The public distribution is centered on three master tables at the Extragalactic Distance Database (EDD): All CF4 Individual Distances, CF4 All Groups, and CF4 All Group Velocities. The individual table carries per-galaxy identifiers, coordinates, velocities, and per-method distances; the group table carries group-averaged moduli and per-method counts; the group-velocity table adds distances, velocities in different frames, curvature-adjusted $fV$, peculiar velocities by multiple estimators, Hubble parameters, and supergalactic Cartesian positions. Method-specific catalogs remain available on their native zero-points, whereas the merged CF4 products apply the global calibration (Tully et al., 2022).

The redshift reach is method-dependent. Coverage is dense to $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$, extends to $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$ in the celestial north through SDSS Fundamental Plane data, and reaches $z_{\rm CMB}=0.1$ through the supernova component. The catalog therefore spans the Local Volume, the intermediate-distance regime dominated by Tully–Fisher and Fundamental Plane distances, and the low-redshift supernova regime within one internally linked structure (Tully et al., 2022).

2. Distance methodologies and the absolute scale

CF4 merges eight distance methodologies. The catalog is dominated numerically by the spiral-galaxy Tully–Fisher family and the early-type-galaxy Fundamental Plane, but its absolute scale is fixed by local ladder methods and by the geometric maser distance to NGC 4258.

Method	Scale in CF4	Technical role
Tully–Fisher (TF)	12,412 galaxies	Spiral-galaxy luminosity–rotation distances
Baryonic TFR (BTFR)	9,967 distances	Gas+stellar-mass form of TF
Fundamental Plane (FP)	34,059 SDSS; 7,099 6dFGSv; $\approx 1{,}508$ older surveys	Main high-$z$ early-type distance set
Type Ia SNe	1,008 hosts	Overlap scaffold to $z_{\rm CMB}=0.1$
Surface brightness fluctuations	multiple optical and IR subsamples	Precision early-type distances
SN II	96 events in 94 hosts	Standardized-candle distances
Cepheid PL (CPLR)	76 hosts	Population I absolute anchor
TRGB and masers	489 TRGB galaxies; 6 maser galaxies	Population II and geometric anchors

The TF relation is written as

$M = a\log_{10}W + b,$

while the FP is expressed as

$$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$

CF4 also adopts the standard SNe Ia form

$fV$0

and the TRGB calibration

$fV$1

These relations are not merely listed side by side: they are stitched together through galaxy and group overlaps, so that subsample zero-points are optimized before cross-method integration (Tully et al., 2022).

The absolute scale is anchored by TRGB, Cepheid, and maser distances. CF4 adopts a Gaia-informed TRGB zero-point shift of $fV$2 mag relative to the Hipparcos-based Rizzi calibration to enforce agreement with the geometric maser distance to NGC 4258. The comparison between TRGB and Cepheid distances gives $fV$3–$fV$4 mag with $fV$5 mag rms across 16 shared galaxies. After internal method consolidation, the five-method ensemble scale is shifted onto the anchor scale by a zero-point offset determined from 121 calibrator overlaps, $fV$6 mag. A SNe Ia-only reference would imply a smaller offset of $fV$7 mag, corresponding to a $fV$8 difference in $fV$9, and this explicitly identifies residual method-dependent systematics within the global calibration problem (Tully et al., 2022).

3. HI linewidth infrastructure, Tully–Fisher inputs, and CF4 distance subsets

A large fraction of CF4 distances depends on homogeneous H I linewidths. The underlying reason is that TF distances are sensitive to linewidth definitions, spectral resolution, and profile processing, so heterogeneous archival measurements can introduce zero-point offsets and inflate peculiar-velocity errors. The Cosmicflows H I program therefore standardized linewidth extraction through the All-Digital H I catalog and a common “Wmean50” pipeline (Courtois et al., 2014).

In the 2014 archival consolidation, the All-Digital H I Catalog reported 20,343 H I spectra for 17,738 galaxies, with 14,802 galaxies having accurate linewidth measurements useful for TF work; that release added 4,117 new measurements, a 34% augmentation over the previous catalog. The catalog stores the PGC identifier, telescope/source code, systemic heliocentric velocity $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$0, linewidths $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$1, $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$2, and $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$3, uncertainties, integrated H I fluxes, channel resolution, and flux/channel at the 50% level. “Accurate” entries require $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$4 and $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$5 at the flux level used for width extraction; contaminated or low-S/N profiles remain archived but are excluded from TF use (Courtois et al., 2014).

The linewidth definition used for CF4 is

$V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$6

the width at 50% of the mean flux computed over the channels enclosing 90% of the total integrated H I flux. The correction for redshift and instrumental broadening is

$V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$7

with $V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$8. The conversion toward twice the maximum rotation speed uses

$V_{\rm cmb}\lesssim 16{,}000\ {\rm km\ s^{-1}}$9

with $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$0 and $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$1. The deprojection is then

$V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$2

These definitions were carried through later H I expansions intended explicitly to feed CF4 (Dupuy et al., 2021).

The 2021 ADHI update added 1,274 new good-quality measurements to the database—385 from the Green Bank Telescope and 889 from reprocessed Nançay spectra—and identified 1,515 additional ALFALFA galaxies whose linewidths could be mapped into the ADHI/Cosmicflows standard. After that release, ADHI contained 18,874 galaxies, of which 15,433 had good-quality data for TF use. The same paper gives the ALFALFA mapping

$V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$3

and, for $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$4,

$V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$5

Those transformations were introduced to enlarge the TF-ready input set while maintaining the Cosmicflows linewidth standard (Dupuy et al., 2021).

Two major CF4 distance subsets are built directly on this H I foundation. The TF catalog provides distances for 9,792 spiral galaxies within $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$6, using calibrated relations in SDSS and WISE bands, inclination corrections, bias corrections, and a random-forest estimate of dust attenuation for spirals lacking infrared photometry; it reports a combined value $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$7 with potential systematics up to $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$8 (Kourkchi et al., 2020). The baryonic Tully–Fisher subset contributes 9,984 accepted spiral-galaxy distances extending to $V_{\rm cmb}\approx 30{,}000\ {\rm km\ s^{-1}}$9, defines

$z_{\rm CMB}=0.1$0

and derives $z_{\rm CMB}=0.1$1 as a statistical result from that subset alone (Kourkchi et al., 2022).

4. Group construction, overlap geometry, and Bayesian merging

Grouping is central to CF4’s statistical design. The catalog assembles galaxies into physically motivated systems using scaling relations tied to the second turnaround or splashback radius, with $z_{\rm CMB}=0.1$2 and $z_{\rm CMB}=0.1$3. Within $z_{\rm CMB}=0.1$4 it uses the nearby catalog of Kourkchi & Tully (2017); for $z_{\rm CMB}=0.1$5 it favors the 2MASS-based groups of Tully (2015); and beyond the useful 2MASS range it adopts the SDSS groups of Tempel et al. (2017). The immediate effect is a reduction in distance and velocity errors by averaging over members; the broader effect is the creation of a dense overlap network across methods that would otherwise have few direct galaxy-by-galaxy intersections (Tully et al., 2022).

Merging is performed in two stages. Within each methodology, CF4 optimizes subsample zero-point offsets $z_{\rm CMB}=0.1$6 by minimizing differences over group-averaged moduli, using independent Gaussian uncertainties and inverse-variance weights. Posterior sampling is carried out with emcee using 128 chains of length 10,000 and a burn-in of 1,000 steps. Outliers are removed with a $z_{\rm CMB}=0.1$7 criterion; the TF example quoted for this stage is 275 rejected cases out of 22,233 measurements, or about 1.2%. After those internal consolidations, TF, FP, SBF, SNe Ia, and SN II are connected on a common arbitrary scale, chosen as the grouped SNe Ia scale, and only then shifted onto the absolute TRGB/Cepheid/maser scale (Tully et al., 2022).

The geometry of the overlap network is a defining property of the catalog. CF4 highlights, for example, Coma with 209 FP and 50 TF distances, Leo with 66 FP and 49 TF, and Virgo with 32 FP, 49 TF, 132 SBF, and 4 SNe Ia. Rich clusters can thus drive group-averaged moduli down from the $z_{\rm CMB}=0.1$8–25% single-galaxy scatter typical of TF and FP to few-percent moduli for the group. This is the operational reason CF4 is better described as a grouped, overlap-calibrated distance network than as a simple concatenation of method-specific source lists (Tully et al., 2022).

5. Peculiar velocities, field reconstructions, and released derived products

Distances and CMB-frame velocities in CF4 are converted into group Hubble parameters and peculiar velocities. For a group,

$z_{\rm CMB}=0.1$9

with

$\approx 1{,}508$0

and

$\approx 1{,}508$1

CF4 quotes a mean value from groups with $\approx 1{,}508$2 of $\approx 1{,}508$3, corresponding to $\approx 1{,}508$4, with a formal statistical uncertainty of $\approx 1{,}508$5 and a potential systematic floor of $\approx 1{,}508$6. Peculiar velocities are provided through a low-$\approx 1{,}508$7/high-$\approx 1{,}508$8 transition between the Davis–Scrimgeour and Watkins–Feldman estimators, ramped by $\approx 1{,}508$9; for groups with $z$0 mag, the peculiar-velocity distribution is nearly Gaussian with $z$1 (Tully et al., 2022).

CF4 has also been released as a field-reconstruction resource. A 2022 study publicly distributed three-dimensional reconstructions of the local gravitational field computed from about 56,000 galaxy distances and 1,008 Type Ia supernova distance moduli. Those reconstructions use a Bayesian forward model with Hamiltonian Monte Carlo, linear-theory relations between density and velocity, and gridded outputs at $z$2 resolution up to $z$3, with velocity-field standard-deviation grids for uncertainty quantification. That work reported $z$4 for ungrouped CF4, $z$5 for grouped CF4, and $z$6 for the CF4 SNe Ia subsample, together with a grouped-CF4 bulk flow of $z$7 at 300 Mpc (Courtois et al., 2022).

A 2023 reconstruction of the large-scale velocity field from grouped CF4 applied the Bias Gaussianization correction to remove lognormal distance and velocity biases, then used Wiener filtering and constrained realizations to recover the linear density and velocity fields out to $z$8. In that analysis, the $z$9 and $z_{\rm CMB}=0.1$0 profiles of CF4 without the 6dFGS component were consistent with cosmic variance within $z_{\rm CMB}=0.1$1, whereas the 6dFGS sample dominated the bulk-flow profile beyond $z_{\rm CMB}=0.1$2 and drove it to roughly a $z_{\rm CMB}=0.1$3 excess at $z_{\rm CMB}=0.1$4. The inferred flow was described as somewhat atypical but not compellingly in tension with $z_{\rm CMB}=0.1$5CDM (Hoffman et al., 2023).

6. Systematics, tensions, and methodological disputes

CF4 explicitly identifies several systematic issues. The main catalog paper discusses Malmquist and selection biases, including the 6dFGSv magnitude-limit truncation and TF/BTFR H I flux-limit effects; morphology and environment systematics, including a morphology-dependent FP correction and host dependencies in SNe Ia standardization; and sky-coverage problems, especially the Galactic plane gap and the relative sparsity outside SDSS beyond $z_{\rm CMB}=0.1$6. The practical recommendations are correspondingly conservative: prefer group-averaged distances and velocities, combine multiple methods where possible, and use distances beyond $z_{\rm CMB}=0.1$7 for $z_{\rm CMB}=0.1$8 and large-scale flow studies (Tully et al., 2022).

Subsequent CF4 analyses have nevertheless produced sharply different assessments of the same local velocity field. A minimum-variance bulk-flow study using the grouped CF4 catalog found $z_{\rm CMB}=0.1$9 at $M = a\log_{10}W + b,$0 toward $M = a\log_{10}W + b,$1, with $M = a\log_{10}W + b,$2 under $M = a\log_{10}W + b,$3CDM, and $M = a\log_{10}W + b,$4 at $M = a\log_{10}W + b,$5 toward $M = a\log_{10}W + b,$6, with $M = a\log_{10}W + b,$7 (Watkins et al., 2023). By contrast, a CF4 group-based parallel peculiar-velocity correlation analysis reported

$M = a\log_{10}W + b,$8

using total-uncertainty covariance, and a “local” value

$M = a\log_{10}W + b,$9

when only statistical uncertainty covariance was retained; that study emphasized the effects of north–south anisotropy, large distance errors, and estimator choice (Wang et al., 9 Apr 2026).

A prior-free forward-modeling analysis of grouped CF4 reconstructed radial and bulk flows in concentric shells out to 300 Mpc/$$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$0 under only a flat FLRW background assumption and found a preferred $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$1 from radial-flow suppression, together with a localized $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$2 tension in the supergalactic $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$3 component and an associated $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$4 bulk-flow tension around $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$5 (Duangchan et al., 29 Jul 2025). A separate study of dipolar anisotropy in local $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$6 measurements mapped angular–radial variations in shells of distance modulus $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$7 and found a clear, distance-decreasing dipole when uncorrected CMB-frame velocities were used, but a much weaker and often statistically insignificant dipole once peculiar velocities from a forward model were subtracted (Salzano et al., 2 Dec 2025). Taken together, these results show that CF4’s scientific interpretation depends strongly on the chosen peculiar-velocity estimator, grouping, weighting, and covariance model; that dependence is an observed feature of the literature rather than an external criticism.

7. Extensions beyond CF4 and successor products

CF4 has already been extended into field-level and cosmic-web products. Cosmicflows-4++ provides 65,331 galaxy distances out to $$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$8 and, uniquely, supplies both observed redshift-space coordinates and reconstructed peculiar-velocity-corrected real-space positions for the same objects. Using identical two-point-correlation pipelines in real and redshift space, a joint analysis found

$$\log R_e = a\log \sigma + b\log \langle I\rangle_e + c.$$9

at effective redshift $fV$00, and a growth-rate constraint

$fV$01

This turns the Cosmicflows framework from a pure peculiar-velocity resource into a catalog for joint real-/redshift-space clustering analysis (Franco et al., 1 May 2026).

A further extension, CF4++ZOA, is tailored to the Galactic Zone of Avoidance. Using an explicit ZOA mask, a linear iterative HMC reconstruction on a $fV$02 grid spanning $fV$03, and a V-web classification with $fV$04, that program identifies 37 voids and 42 knots within the reconstructed survey volume. The reported void effective radii span $fV$05–$fV$06, while knot volumes range from $fV$07 to $fV$08 (Hollinger et al., 11 Jun 2026).

Comparative studies of density-field products have also sharpened the catalog’s practical use cases. One 2026 comparison between the CF4 mean posterior field and a cloned galaxy catalog concluded that CF4 is preferred in the Zone of Avoidance, where cloning generates artificial reflection-symmetric structures, whereas the external catalog may better capture the direction and integrated mass of structures outside the ZOA. In that comparison, the CF4 field was represented on a Cartesian $fV$09 grid with spacing of about $fV$10 Mpc (Li et al., 28 Jan 2026). A plausible implication is that “the CF4 catalog” now denotes not only the 2022 distance tables but also a family of linked products—group catalogs, H I standards, field reconstructions, real-/redshift-space extensions, and cosmic-web classifications—that retain a common peculiar-velocity and calibration ancestry while serving different technical applications.