{"id":"90ef1a62-faba-4522-b717-3509b67ca22b","arxiv_id":"2508.02780","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A reproducible VISTA/VIRCAM reduction pipeline, NASIM, recovers low surface brightness features down to ~27.7 mag/arcsec^2 in the Ks band, ~67 times deeper than 2MASS.","lead":"NASIM is a new, fully automated pipeline that reduces VISTA/VIRCAM near-infrared images while preserving faint, diffuse structures that standard pipelines miss. Applied to the deep KEDFS survey, it reaches a surface brightness depth of about 27.7 mag/arcsec^2, enabling studies of galaxy outskirts, ultra-diffuse galaxies, and intracluster light.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central depth and preservation claim lacks an end-to-end test with known diffuse flux; background corrections (adaptive flat, per-frame sky, final polynomial sky model) could remove or create the 27.7 mag/arcsec^2 signal.","rationale":"The reader's conditional verdict is well placed. I do not see internal logical inconsistency in the reduction steps; the pipeline is described in unusual detail and the Maneage framework provides strong reproducibility. The main gap is quantitative validation of the losslessness of the background corrections. The reader singles out the adaptive flat's 11-frame window; I agree that is a risk, but the post-stack polynomial sky model in Section 3.10 is at least as dangerous because it explicitly subtracts a smooth large-scale function from the coadded science image, and the adopted order is chosen without quantitative comparison. The depth number 27.7 is quoted without error bars and appears derived from one galaxy's radial profile rather than from a dedicated aperture-noise measurement in the final stack. The commit hash discrepancy (4d32667 in the abstract versus 9d9968e in code/data availability) is real but is a reproducibility issue, not the central scientific claim. The proposed injection test would settle the matter by measuring the pipeline's transfer function for diffuse signal; it is feasible because the workflow is fully reproducible. I therefore keep the conditional verdict unchanged rather than escalating to rejection.","tokens_in":20019,"tokens_out":6155,"duration_ms":73628,"concrete_test":"Run the Maneage workflow (commit 9d9968e) end-to-end on one KEDFS tile in two branches: (a) unmodified, and (b) with synthetic diffuse exponential discs (central surface brightness 24-28 mag arcsec^-2, half-light radii 10-60 arcsec, known total flux) inserted into the dark-subtracted, pre-flat images and propagated through flat-fielding, sky subtraction, polynomial sky modelling, and stacking using identical NoiseChisel masks and parameters. Measure recovered versus input surface brightness in the injected regions at radii corresponding to 26-28 mag arcsec^-2. If recovered flux is within about 10% and no systematic gradient is introduced, the preservation claim is supported; if the pipeline suppresses or biases the injected signal by more than the reported 3-sigma depth, the headline depth and LSB recovery claims are pipeline-dependent and need revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's headline result is a Ks surface brightness limit of about 27.7 mag arcsec^-2 (3 sigma over 100 arcsec^2) with diffuse emission preserved. The load-bearing assumption is that NASIM's large-scale background corrections are lossless at precisely the surface brightness levels it is designed to recover. This is not yet established. The pipeline applies three successive large-scale corrections: an adaptive flat built from 11 consecutive exposures including the target image itself (Section 3.3.2, Method 4), a single-value sky subtraction per frame estimated from NoiseChisel tiles (Section 3.5), and a second-order polynomial sky model subtracted from the final stack before re-stacking (Section 3.10). The choice of second-order polynomial is justified only by qualitative evaluation, and the depth claim rests on one radial profile of NGC 1494 (Section 4.1, Figure 7) with no uncertainty budget for the 27.7 value. The paper itself notes residual 1/f patterns remain after correction (Section 3.3.3) and that the VIDEO comparison field has non-uniform background (Section 5.2). The stated 0.05% 1/f amplitude is about 70 times larger than the sky-subtracted flux of a 27.7 mag arcsec^-2 feature, and the adaptive flat can in principle self-subtract large-scale emission that persists across the 11-frame window. Without injecting diffuse sources of known surface brightness into the raw frames and recovering them through the full pipeline, the central claim that NASIM reveals rather than manufactures LSB features is not quantitatively supported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents NASIM, a fully automated and Maneage-reproducible reduction pipeline for VISTA/VIRCAM observations, applied to the KEDFS Ks-band survey. The pipeline introduces an adaptive flat-fielding scheme in which each science frame is divided by a master flat built from 11 consecutive exposures, followed by conservative single-value sky subtraction per frame, 1/f noise removal, and a second-order polynomial sky model subtracted from the final stack before re-stacking. The authors claim a surface brightness depth of approximately 27.7 mag arcsec^-2 (3 sigma over 100 arcsec^2), about 67 times deeper than 2MASS and 11 times deeper than VHS, and demonstrate the output on galaxy outskirts, a UDG, and intracluster light, with a comparison against the standard VDFS/VIDEO reduction.","tokens_in":20367,"tokens_out":5665,"duration_ms":60565,"significance":"If validated, this is a valuable contribution: it opens a largely unexplored ground-based NIR low-surface-brightness regime, it is built on an exemplary reproducible framework (Maneage/Gnuastro) with version-controlled software and data lineage, and it releases useful KEDFS tiles as legacy data. The adaptive-flat concept and the explicit non-aggressive sky-subtraction strategy are well motivated and directly address a real limitation of existing VISTA pipelines. However, the central quantitative claims currently rest on visual comparisons and a single radial profile, and the paper lacks end-to-end recovery tests for diffuse flux. The pipeline parameters are configuration choices made by inspecting outputs rather than fitted to the showcased galaxies, so I see no circularity problem; the issue is instead that the validation is not yet quantitative enough to support the headline depth and preservation claims.","major_comments":[{"comment":"The headline depth claim ('approximately 27.7 mag arcsec^-2 (3 sigma over 100 arcsec^2)') is stated without a derivation or uncertainty budget. The text should specify how this value is measured from the radial profile: the aperture used for the noise estimate, the number of independent 100 arcsec^2 apertures, the sigma-clipping recipe, and the photometric calibration error. It should also report the uncertainty on the 27.7 value and show the noise floor on the same figure. As written, the abstract's central quantitative result cannot be reproduced or compared against other surveys.","section":"Section 4.1, Figure 7"},{"comment":"The two dominant large-scale corrections — the adaptive flat built from 11 consecutive exposures including the target image itself, and the second-order polynomial sky model subtracted before the final re-stack — act on exactly the spatial scales and brightness levels of the diffuse emission that NASIM claims to preserve. The paper's evidence that these steps do not self-subtract astrophysical signal is visual (Figures 3 and 5). I request an end-to-end injection-recovery test: add diffuse sources with known surface brightness (spanning roughly 26-28 mag arcsec^-2 and angular scales from tens of arcseconds to arcminutes) into the raw dark-subtracted frames, run the full NASIM chain, and report recovered versus input flux as a function of scale and brightness. This test is particularly important because Section 3.3.3 states that residual 1/f patterns of about 0.05% of the sky remain after correction; that amplitude is about 70 times the sky-subtracted flux of a 27.7 mag arcsec^-2 feature, so the tolerance for background systematics must be demonstrated quantitatively rather than assumed.","section":"Section 3.3.2, Method 4; Section 3.10; Section 3.3.3"},{"comment":"The choice of the adopted sky model is not quantitatively justified. The text reports that Chebyshev, first-order, and second-order polynomial models were tested and that the second-order polynomial was adopted, but it gives no selection criterion and no residual statistics. Similarly, the NoiseChisel settings minskyfrac=0.9 and the increased interpnumngb are asserted to give a clean sky estimate without a comparison of alternative settings. At minimum, the authors should report the residual large-scale power (for example, the RMS in empty regions before and after subtraction as a function of model order) and demonstrate that the adopted settings do not remove signal at the claimed depth around bright galaxies.","section":"Section 3.10; Section 3.5"},{"comment":"The claim that NASIM does not compromise compact source detection rests on a single number: 5-sigma limiting magnitudes of 23.67 mag (NASIM) versus 23.72 mag (VDFS) in 2 arcsec diameter apertures. A limiting magnitude alone does not constrain the full detection function. I ask for source injection-recovery completeness curves, or at least number counts relative to the VDFS catalogue, over a range of magnitudes and source sizes, plus an uncertainty estimate for the two limiting magnitudes, before concluding that point-source sensitivity is preserved.","section":"Section 5.2"}],"minor_comments":[{"comment":"The sentence containing 'the software environment, analysis pipeline, and and text of this paper' has a duplicated 'and'; the paper also mixes British and American spellings (for example 'focusses' and 'focuses').","section":"Section 3, first paragraph"},{"comment":"The abstract cites Maneage commit 4d32667, while the 'Code and data availability' section cites Git commit 9d9968e and a different latest Maneage commit (8161194); these version identifiers should be reconciled or explained.","section":"Abstract versus Code and data availability"},{"comment":"The Reference column entry 'KEDFS; Abstract' is not a standard citation; a proper reference for the KEDFS survey, or an explicit note that it is described in this paper, is needed.","section":"Table 1"},{"comment":"The 1/f noise correction is described only qualitatively; stating the measured amplitude before and after correction, and the residual pattern in the final stack, would make the claimed improvement testable.","section":"Section 3.3.3, Figure 4"}],"recommendation":"major_revision","confidential_remarks":"For the editor: this is a methods/software paper well within the journal's scope, and I saw no attribution or novelty concerns. The reproducibility infrastructure is a genuine strength. The main gate for publication should be the requested injection-recovery validation and the explicit derivation of the depth limit; without those, the abstract's quantitative claims are stronger than the evidence presented in the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The adaptive flat-fielding is the real contribution. Constructing a unique master flat for each exposure from a sliding window of 11 images is a sensible response to VIRCAM's sky variability and persistence, and the paper's side-by-side comparison of Methods 1-4 is convincing. The Maneage-based reproducibility and the release of KEDFS tiles are also genuine assets, as is the VIDEO comparison showing that NASIM preserves compact-source depth (23.67 vs 23.72) while recovering a dwarf galaxy's outer envelope.\n\nThe soft spot is the load-bearing claim. The pipeline applies three successive large-scale corrections - the adaptive flat, a per-frame scalar sky subtraction, and a second-order polynomial sky model - and the paper argues that these corrections preserve diffuse flux at the very levels they are meant to recover. That is not established. There is no end-to-end test with injected diffuse sources of known surface brightness, and the depth value of ~27.7 mag/arcsec^2 is presented without an error budget or a clear description of how it was measured. The paper's own admission that residual 1/f patterns remain in the final stack makes the number hard to interpret, and the 0.05% amplitude of those patterns is large compared to the signal of a 27.7 mag/arcsec^2 feature. The choice of a second-order polynomial for the sky model is also justified only by qualitative evaluation.\n\nThese concerns are addressable, not fatal. The qualitative evidence is strong - galaxy outskirts, a catalogued UDG, and ICL are all plausibly recovered - and the VIDEO comparison suggests the pipeline is not over-subtracting compact sources. What is missing is a standard validation: inject mock diffuse sources into raw frames, run them through the full pipeline, and show that they are recovered at the claimed depth. Also worth fixing: the commit hash is inconsistent between the abstract (4d32667) and the code/data section (9d9968e).\n\nI would send this to a serious referee. The paper deserves review, and the referee should require the injection tests and a defensible depth measurement before publication. After that it becomes a valuable methods paper and a useful data release.","headline":"A genuinely useful and reproducible LSB pipeline for VISTA, but the headline depth claim needs stronger validation before it can be taken at face value.","tokens_in":20904,"tokens_out":2787,"would_cite":true,"duration_ms":34061,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A new pipeline reaches 27.7 mag per square arcsecond in Ks-band imaging while preserving faint diffuse emission.","keywords":["low surface brightness","near-infrared imaging","image reduction pipeline","adaptive flat-fielding","VISTA","VIRCAM","KEDFS","reproducible workflow"],"falsifier":"Re-reduce a single KEDFS tile with adaptive flats built from 3-, 5-, 11-, and 21-exposure windows and compare the radial profile of NGC 1494 and the measured 3$\\sigma$ depth; if the profiles and depth are indistinguishable, the window length is not what creates the claimed gain. A complementary test is to add synthetic exponential-disc sources of known surface brightness around 26\\,--\\,28 mag arcsec$^{-2}$ to the raw frames and check what fraction NASIM recovers in the final stack.","tokens_in":19869,"feed_emoji":"🌌","tokens_out":9552,"duration_ms":98212,"temperature":0.7,"pith_summary":"This paper argues that the bright, time-varying near-infrared sky is not an insurmountable barrier to low surface brightness science from the ground if the flat-field correction is built on the right timescale. It presents NASIM, a fully automated and reproducible reduction pipeline for VISTA/VIRCAM, whose central move is an adaptive flat: one master flat per science exposure, made from that exposure plus the five before and five after it. Applied to the deep Ks-band survey KEDFS, the pipeline reaches a surface brightness limit of $27.7\\,\\mathrm{mag\\,arcsec^{-2}}$ (3$\\sigma$ over 100 arcsec$^2$), roughly 67 times deeper than 2MASS and 11 times deeper than VHS. A reader should care because, if the claim holds, retired VISTA data become a mineable archive for galaxy outskirts, ultra-diffuse galaxies, and intracluster light, with point-source sensitivity essentially unchanged.","feed_headline":"Reveals diffuse light to 27.7 mag arcsec-2 in near-infrared","feed_subtitle":"Adaptive flat-fielding recovers galaxy outskirts and intracluster light that standard VISTA reductions erase.","key_machinery":"The load-bearing object is the adaptive flat: for each science frame $N$, NASIM stacks the 11 consecutive exposures centred on it (5 before, 5 after, plus $N$ itself) with $\\sigma$-clipped median combination to form a dedicated master flat, after dark subtraction and NoiseChisel-based masking of astrophysical sources. This confines the flat to the time window in which the sky is actually coherent, removing large-scale gradients and detector patterns without imposing a night-long average. The pipeline then corrects the image with that flat, subtracts a single sky value per frame using only clean tiles, removes low-level 1/f readout stripes by collapsing and re-expanding the image along the readout axis, resamples to a common 0.2 arcsec grid, and coadds with weights set by sky standard deviation; a final loop models residual sky in the stack with a second-order polynomial under a strong mask and restacks.","core_discovery":"The core discovery is that standard VISTA reductions lose diffuse emission not because the data are too shallow but because their flat-fielding smears sky variations over too long a time. NASIM instead builds a unique master flat for each exposure from 11 consecutive dark-subtracted and object-masked frames, tracks the sky on the roughly one-hour timescale on which it actually varies, and pairs this with non-aggressive single-value sky subtraction, removal of 1/f readout patterns, and weighted $\\sigma$-clipped stacking. On KEDFS this yields Ks-band images that trace the outskirts of NGC 1494 to $\\sim27.7\\,\\mathrm{mag\\,arcsec^{-2}}$, recover an ultra-diffuse galaxy and intracluster light that are invisible in VHS, and extend the NGC 895 profile far beyond the point where the standard VIDEO reduction truncates it. The 5$\\sigma$ point-source limit in 2 arcsec apertures is 23.67 mag for NASIM versus 23.72 for the comparison pipeline, so the LSB gain does not come at the cost of compact-source depth.","pith_inferences":["An extension the paper leaves implicit: the adaptive-flat timescale argument should transfer to any ground-based NIR camera with a quickly varying sky, so archival data from other instruments could be re-reduced for LSB science without new observations.","A testable extension: inject synthetic diffuse sources of known surface brightness into raw KEDFS frames, run NASIM, and measure recovered flux near $27.7\\,\\mathrm{mag\\,arcsec^{-2}}$; the recovery fraction would separate genuine signal from flat-field artifacts.","If the claim is right, re-reducing wide but shallow surveys like VHS would not reproduce the 67x gain over 2MASS, since that depth is specific to KEDFS, but it would still open the brightest LSB regimes over tens of thousands of square degrees."],"forward_implications":["Selected KEDFS tiles are released with the paper, giving immediate public access to Ks-band imaging deep enough to study galaxy outskirts, ultra-diffuse galaxies, and intracluster light.","The same pipeline can re-reduce any VISTA/VIRCAM survey, so legacy archives such as VHS could be reprocessed for large-area low-surface-brightness studies.","Because NASIM keeps the 5$\\sigma$ compact-source limit within 0.05 mag of the point-source-optimised VIDEO reduction, a single reduction can serve both LSB and traditional source catalogues.","With VIRCAM decommissioned, KEDFS is positioned as the ground-based Ks anchor for multi-wavelength follow-up with Euclid, JWST, Roman, LSST, Spitzer, and ALMA in the Euclid Deep Field South.","The full workflow is reproducible from a version-controlled project, allowing independent verification and re-use on the entire VISTA archive."],"supporting_citations":[{"why":"Supplies the NoiseChisel segmentation algorithm used to mask astrophysical sources before building the adaptive flats and to estimate sky values.","marker":"Akhlaghi & Ichikawa (2015)"},{"why":"Provides the image-processing functions that carry out flat correction, sky subtraction, resampling, and stacking.","marker":"Akhlaghi (2019a)"},{"why":"Supplies the reproducible workflow layer that automates the pipeline and records complete data lineage.","marker":"Akhlaghi et al. (2021)"},{"why":"Documents the VIRCAM detector layout and pawprint structure that the flat-fielding and mosaicking steps must handle.","marker":"Sutherland et al. (2015)"},{"why":"Demonstrates the use of science exposures to construct flats for low-surface-brightness work, the basis of the adaptive-flat strategy.","marker":"Trujillo & Fliri (2016)"},{"why":"Defines the VIDEO survey reduction used as the point-source-optimised baseline in the NGC 895 comparison.","marker":"Jarvis et al. (2013)"}],"fun_headline_variants":["NASIM rewrites VISTA data to reveal faint galaxy halos","Adaptive flat-fielding recovers diffuse light VISTA misses","Faint universe emerges from legacy VISTA near-infrared data","NASIM pipeline exposes galaxy outskirts down to 27.7 mag"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the sky's large-scale structure stays stable over the roughly one-hour window spanned by the 11 exposures that make each adaptive flat; if it varies faster, the flat imprints gradients that can mimic or erase the very features NASIM aims to recover.","fun_headline_variants_meta":{"raw":{"variants":["NASIM rewrites VISTA data to reveal faint galaxy halos","Adaptive flat-fielding recovers diffuse light VISTA misses","Faint universe emerges from legacy VISTA near-infrared data","NASIM pipeline exposes galaxy outskirts down to 27.7 mag"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000534,"raw_usage":{"total_tokens":2613,"prompt_tokens":1036,"completion_tokens":1577,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":652,"completion_tokens_details":{"reasoning_tokens":1502}},"tokens_in":652,"tokens_out":1577,"duration_ms":12222,"temperature":1.0,"reasoning_tokens":1502,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:36:50.070817+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-reduce a single KEDFS tile with adaptive flats built from 3-, 5-, 11-, and 21-exposure windows and compare the radial profile of NGC 1494 and the measured 3$\\sigma$ depth; if the profiles and depth are indistinguishable, the window length is not what creates the claimed gain. A complementary test is to add synthetic exponential-disc sources of known surface brightness around 26\\,--\\,28 mag arcsec$^{-2}$ to the raw frames and check what fraction NASIM recovers in the final stack.","supporting_citations":[{"cited_title":"F., et al","cited_arxiv_id":null,"evidence_quote":"Supplies the reproducible workflow layer that automates the pipeline and records complete data lineage."},{"cited_title":"2015, A&A, 575, A25","cited_arxiv_id":null,"evidence_quote":"Documents the VIRCAM detector layout and pawprint structure that the flat-fielding and mosaicking steps must handle."},{"cited_title":"J., Bonfield, D","cited_arxiv_id":null,"evidence_quote":"Defines the VIDEO survey reduction used as the point-source-optimised baseline in the NGC 895 comparison."}],"review_version":2}