STRRINGS: Stellar Streams in Nearby Galaxies

• STRRINGS is a residual-image based survey that catalogues tidal features and extracts 35 modelling-ready stellar stream systems from over 19,000 nearby galaxies.
• The methodology leverages DESI Legacy Imaging Surveys with original, model, and residual images to enhance detection and segmentation of low-surface-brightness streams.
• Key results include a tidal feature incidence of 11.9%, detailed photometric and geometric measurements, and stream-to-host mass ratios indicative of minor merger events.

STRRINGS, short for STReams in Residual Images of Nearby GalaxieS, is a residual-image-based survey and stream catalogue built to identify nearby extragalactic stellar streams that are suitable for dynamical modelling. It combines a large visual census of tidal features with a modelling-oriented first release of 35 stream systems selected from 19,387 nearby galaxies in the Siena Galaxy Atlas 2020 with $z \leq 0.02$, using original, model, and residual images from the DESI Legacy Imaging Surveys. Its stated purpose is not only to inventory tidal debris, but to assemble a sample of long, narrow streams that can be compared with simulations to obtain population-level constraints on halo properties, especially dark-halo flattening (Sola et al., 4 Aug 2025).

1. Scope, scientific context, and definition

STRRINGS is explicitly presented as both a large visual tidal-feature census of nearby galaxies and the first released subsample of 35 stellar-stream systems chosen to be suitable for modelling (Sola et al., 4 Aug 2025). The project is framed within hierarchical galaxy formation in $\Lambda$CDM, where mergers and accretion generate distinct tidal structures including tails, streams, shells, plumes, and disturbed stellar haloes.

Within that taxonomy, stellar streams are singled out because they arise from tidally disrupted satellites in minor mergers and can trace the progenitor orbit within the host potential. The paper emphasizes that long, curved streams can constrain the host potential geometrically, and that populations of many streams can constrain halo properties statistically. It further states that future work aims to infer dark-halo flattening from a sample of order $>50$ streams (Sola et al., 4 Aug 2025).

A central novelty is the use of residual images from DESI-LS. These are generated by subtracting source models from original survey images, leaving low-surface-brightness asymmetries such as streams more visible. The project therefore targets not merely the detection of tidal features in general, but the construction of a modelling-ready stream sample (Sola et al., 4 Aug 2025).

2. Data basis and parent-galaxy sample

The underlying imaging comes from the DESI Legacy Imaging Surveys, specifically DECaLS, BASS, and MzLS, covering about $14{,}000\ \mathrm{deg}^2$ in $g,r,z$ with approximate $5\sigma$ depths for faint galaxies of $g = 24.0$, $r = 23.4$, and $z = 22.5$ AB magnitudes. DESI-LS source modelling uses The Tractor through legacypipe, and for every image provides the original image, the model image, and the residual image (Sola et al., 4 Aug 2025).

Residuals are central because standard source-modelling algorithms are optimized for compact or high-surface-brightness components, whereas diffuse tidal features are often not well modelled and remain in the subtraction products. The paper also notes that segmentation is easier on residual images because the stream is less contaminated by bright host-galaxy light and nearby sources (Sola et al., 4 Aug 2025).

Residual-image surface-brightness limits were estimated using $3\sigma$, $\Lambda$0 boxes placed randomly on masked images, yielding $\Lambda$1 in $\Lambda$2, $\Lambda$3 in $\Lambda$4, and $\Lambda$5 in $\Lambda$6 (Sola et al., 4 Aug 2025).

The parent catalogue is the Siena Galaxy Atlas 2020 (SGA-2020). After applying the redshift cut $\Lambda$7, the sample contains 20,077 individual galaxies. Combining galaxies in close SGA-defined groups into single objects yields 19,387 objects, which form the visually inspected parent sample (Sola et al., 4 Aug 2025). The paper states that the sample is 23.4% early-type galaxies, 64.9% late-type galaxies, and 11.7% of unknown morphology; 92.3% are isolated (Sola et al., 4 Aug 2025).

Host stellar masses are estimated from Tractor $\Lambda$8 and $\Lambda$9 fluxes using the Bell et al. prescription

$>50$0

with $>50$1 and $>50$2. The host-galaxy stellar-mass distribution has median $>50$3 and 16–84% quantiles $>50$4–$>50$5 (Sola et al., 4 Aug 2025).

3. Visual classification and catalogue construction

The first-stage inspection used pre-made PNG montages from the SGA site containing the original $>50$6 colour image, model image, and residual image. The field of view was 3 times the galaxy size, based on D25. For inspection, the residuals were converted to black-and-white and clipped to the 35th–90th percentiles of the pixel distribution to enhance faint structure (Sola et al., 4 Aug 2025).

The visual classification used mutually exclusive categories: No feature, Streams, Tails/Arms, Plumes, Shells, and Disturbed/asymmetric stellar halo, with optional Grand-design spiral and Curiosity flags (Sola et al., 4 Aug 2025). Definitions were morphological in projection.

A total of nine classifiers contributed. In the first pass, two classifiers inspected all 19,387 objects and flagged 5,120 as showing a feature according to at least one of the two. Those candidates were then classified by the other seven classifiers as well (Sola et al., 4 Aug 2025).

The voting system first decided whether a galaxy had any feature or no feature; in ties, the galaxy was counted as having a feature. For feature-class ties, the priority order was

$>50$7

which the paper explicitly notes biases results toward streams (Sola et al., 4 Aug 2025). Confidence was defined as

$>50$8

No hard score cut was applied in the main catalogue. For stream classifications, 30.9% have score $>50$9 and 6.9% have score $14{,}000\ \mathrm{deg}^2$0, indicating substantial classifier disagreement (Sola et al., 4 Aug 2025).

4. Tidal-feature census results

The main census result is that $14{,}000\ \mathrm{deg}^2$1 of the 19,387 inspected objects host a tidal feature, where tidal feature means streams, tails, plumes, or shells (Sola et al., 4 Aug 2025). The majority, $14{,}000\ \mathrm{deg}^2$2, show no feature.

Category	Fraction
Streams	$14{,}000\ \mathrm{deg}^2$3
Tails	$14{,}000\ \mathrm{deg}^2$4
Shells	$14{,}000\ \mathrm{deg}^2$5
Plumes	$14{,}000\ \mathrm{deg}^2$6
Disturbed halo	$14{,}000\ \mathrm{deg}^2$7
Grand-design spiral	$14{,}000\ \mathrm{deg}^2$8
Curiosities	$14{,}000\ \mathrm{deg}^2$9

Morphology dependence is strong. The fraction hosting any tidal feature is $g,r,z$0 for early-type galaxies, $g,r,z$1 for late-type galaxies, and $g,r,z$2 for unknown morphology (Sola et al., 4 Aug 2025). Streams occur in $g,r,z$3 of early types and $g,r,z$4 of late types, while shells are especially concentrated in early types at $g,r,z$5 versus $g,r,z$6 in late types (Sola et al., 4 Aug 2025).

The incidence also rises strongly with stellar mass, from $g,r,z$7 around $g,r,z$8 to $g,r,z$9 around $5\sigma$0, with a sharper rise above roughly $5\sigma$1 (Sola et al., 4 Aug 2025). Environment matters as well: the tidal-feature fraction is $5\sigma$2 in groups and $5\sigma$3 in isolated galaxies; the stream fractions are $5\sigma$4 and $5\sigma$5 respectively (Sola et al., 4 Aug 2025).

These results are accompanied by explicit caveats. The catalogue uses mutually exclusive labels even though real galaxies may host multiple feature types; the voting scheme favors streams; DESI-LS is shallower than dedicated low-surface-brightness surveys; residuals may contain artefacts and oversubtraction; and the 3$5\sigma$6 galaxy-size inspection field can miss large-radius streams (Sola et al., 4 Aug 2025).

5. The modelling-ready sample of 35 streams

Among the full census, 861 galaxies had streams as their dominant feature. These were re-examined by four classifiers using a focused three-way scheme: Not streams, Stream not selected for modelling, and Stream selected for modelling. The final release includes only objects for which all four classifiers agreed on “stream selected for modelling,” producing 35 galaxies/stream systems (Sola et al., 4 Aug 2025).

The selection criteria were tailored to the intended dynamical framework. Desired systems were relatively thin, noticeably curved, sufficiently faint, and around hosts whose central regions appear undisturbed (Sola et al., 4 Aug 2025). The stated aim was to isolate streams most consistent with kinematically simple minor mergers and a relatively stable host potential.

Segmentation was performed manually with Jafar, using DESI-LS residual cutouts in $5\sigma$7 combined by the median of the three bands, clipped to 35–90% of the pixel distribution, rescaled to $5\sigma$8 pixels, and converted to HiPS format (Sola et al., 4 Aug 2025). When a stream overlapped or wrapped around itself, it was split into multiple masks assigned to distinct angular regions.

Track reconstruction followed a procedure in which the stream is segmented into non-overlapping masks, the image is divided into angular bins, radial residual-light profiles are measured in each bin, and one or more Gaussians are fitted to those profiles to recover the ordered stream track (Sola et al., 4 Aug 2025). This provides the basis for geometric measurements such as length, width, angular extent, and galactocentric radius.

6. Measured stream properties and candidate progenitors

For the 35 selected systems, photometry used DESI-LS viewer residual cutouts in $5\sigma$9 at native pixel scale $g = 24.0$0. Remaining sources were masked with SEP, and flux, surface brightness, and colour were measured inside the stream mask. Because residual images can contain negative values from oversubtraction, the flux distribution was truncated to keep only positive values when measuring stream photometry (Sola et al., 4 Aug 2025).

The reported median stream surface brightnesses are $g = 24.0$1 in $g = 24.0$2, $g = 24.0$3 in $g = 24.0$4, and $g = 24.0$5 in $g = 24.0$6. The faintest detected stream reaches $g = 24.0$7 in $g = 24.0$8, $g = 24.0$9 in $r = 23.4$0, and $r = 23.4$1 in $r = 23.4$2 (Sola et al., 4 Aug 2025). The typical median $r = 23.4$3 colour is 0.5.

The sample is deliberately long by construction. The median length is 124 kpc, the mean length is 224 kpc, and the 5–95% range is 24–1151 kpc. Widths are more conventional: median and mean 5.6 kpc, with a 5–95% range of 2.3–9.9 kpc (Sola et al., 4 Aug 2025). Most streams lie within projected radius $r = 23.4$4 kpc, with a few extending to 40–60 kpc, so the sample predominantly probes the inner dark-matter distribution (Sola et al., 4 Aug 2025).

The median stream stellar mass is $r = 23.4$5. Stream-to-host stellar mass ratios have median 0.007, mean 0.02, and only 4 streams exceed 0.05, which the paper states is consistent with minor mergers, roughly in the range $r = 23.4$6 to $r = 23.4$7 (Sola et al., 4 Aug 2025).

Several reported trends organize the sample. Stream stellar mass correlates positively with stream length; there is a hint that more massive streams are redder; more massive streams are somewhat wider; and physically wider streams tend to lie at larger galactocentric radius (Sola et al., 4 Aug 2025). For the angular-width ratio

$r = 23.4$8

the paper gives linear fits

$r = 23.4$9

for median values and

$z = 22.5$0

for all individual measurements, with the former having p-value = 0.03 from Wald’s test (Sola et al., 4 Aug 2025).

The segmentation stage identified six streams containing a compact candidate dwarf progenitor: ESO356-012, PGC1092512, NGC0804, PGC430221, NGC5513_GROUP, and UGC09239. The paper concludes that five are likely genuine progenitors, while the NGC5513_GROUP case is more doubtful (Sola et al., 4 Aug 2025).

7. Interpretation, limitations, and projected development

STRRINGS is presented as a proof-of-concept sample for extragalactic stream modelling at scale. Its intended scientific roles are to trace recent minor accretion, to characterize merger demographics as a function of stellar mass, morphology, and environment, and to constrain the host gravitational potential and dark-matter halo structure. The paper explicitly states that the catalogue will be used in future work to constrain the flattening of dark matter haloes at the population level (Sola et al., 4 Aug 2025).

The project is also explicit about its limitations. DESI-LS is relatively shallow compared with dedicated low-surface-brightness surveys; residual images can contain artefacts and oversubtraction; stream photometry on residuals likely underestimates total stream light; stream-to-host mass ratios are therefore lower bounds; and manual segmentation introduces subjectivity (Sola et al., 4 Aug 2025). Most systems probe only inner halo radii, not extreme outskirts.

A plausible implication is that the catalogue’s strongest present value lies less in completeness than in curation: the sample is intentionally restricted to streams that satisfy the assumptions of a particular modelling framework. This is consistent with its finding that only 35 of 861 stream-dominant systems were retained for first-release modelling (Sola et al., 4 Aug 2025).

Planned extensions include expanding beyond $z = 22.5$1, using residual images for automated detection and segmentation, fine-tuning tools such as Zoobot, and scaling to future surveys such as Rubin/LSST and Euclid (Sola et al., 4 Aug 2025). In that sense, STRRINGS is both a catalogue and a methodological template for turning residual-image tidal-feature searches into a modelling-oriented program for constraining halo structure in external galaxies (Sola et al., 4 Aug 2025).