{"id":"a0414e31-62e9-4e8f-aaa9-04b9e76c5b94","arxiv_id":"2411.12257","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Using morphological matching between 3D dust maps and HI/CO emission, the authors reconstruct line-of-sight velocities of the local ISM and find streaming motions of about 10.8 km/s for HI and 6.6 km/s for CO relative to Galactic rotation.","lead":"This paper builds a three-dimensional map of how fast the gas in our cosmic neighborhood is moving toward or away from the Sun, by matching the shapes of dust clouds seen at different distances with the shapes of atomic and molecular gas seen at different velocities. A smart generalist might read it because it offers a new way to measure the local interstellar medium's kinetic energy and dynamics, which underpin star formation and the structure of the Milky Way.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Morphological correlation may trace large-scale disk structure rather than co-located gas; a Galactic-scale simulation test is needed before the kinetic tomography velocities can be trusted.","rationale":"The paper is a careful application of the HOG method, with extensive parameter exploration, jackknife tests, and a single-cloud simulation test. The reconstructed quadrupolar rotation pattern is a strong sanity check. However, the central claim that the assigned velocities represent physical streaming motions depends on the assumption that morphological similarity between a 3D dust distance slice and a line-emission velocity channel indicates co-location. This assumption is load-bearing because both datasets trace the same large-scale Galactic disk; the morphological match could be driven by shared large-scale structure rather than by the same gas at the same distance. The existing validation is too sparse: Appendix D uses one isolated cloud without a background disk or multi-component sightlines, and Appendix E has only five masers, which show large velocity differences from the HOG reconstruction. These differences exceed the claimed streaming-motion dispersions, leaving the absolute velocity scale uncertain. The reader's weakest assumption is the same co-location premise, so I agree with the reader's identification. Since the proposed Galactic-scale simulation test would provide ground-truth distances and velocities, including large-scale structure and realistic line-of-sight confusion, it would settle whether the concern actually lands. If the method fails that test, the quantitative streaming motions and energy densities are unsupported; if it passes with accuracy better than the claimed dispersions, the co-location assumption is credible. This does not change the reader's conditional verdict but clarifies the specific validation needed for acceptance.","tokens_in":45165,"tokens_out":6350,"duration_ms":74853,"concrete_test":"Construct a synthetic Milky Way-like disk from a high-resolution MHD simulation (e.g., a SILCC-type stratified box embedded in a rotating potential or a TIGRESS galactic patch) with known 3D density, HI/CO emissivity, and velocity fields. Generate mock 3D dust extinction and HI/CO PPV cubes with the same angular resolution, 10°x10° tiling, and vLOS range (-25 to 25 km/s) as in the paper. Run the full HOG pipeline (Eq. 2, max-Vd assignment, Sec. 3.3 thresholds) and compare the reconstructed vHOG_LOS per distance channel against the density-weighted true vLOS of the simulation. The co-location assumption survives only if the rms reconstruction error is substantially smaller than the claimed 6.6 km/s (CO) and 10.8 km/s (HI) streaming-motion dispersions, and if the derived streaming-motion map recovers the true departures from rotation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that HOG-based kinetic tomography assigns physically co-located distance-velocity pairs rests on the unstated premise in Sec. 3.3 that the maximum Vd (Eq. 2) identifies the same gas in a 3D dust slice and a line-emission velocity channel. High Vd could instead arise because both the dust and the HI/CO emission independently trace the same large-scale Galactic disk structure (warp, spiral-arm ridges, vertical distribution), producing correlated morphologies across broad distance and velocity ranges even when the dust at distance d is not the gas at velocity v. The paper's validation is insufficient to exclude this: Appendix D tests one isolated simulated cloud without a background disk or multiple LOS velocity components, and Appendix E has only five VLBI masers whose velocities differ from the HOG velocities with standard deviations of 13.5 and 11.3 km/s, larger than the reported streaming-motion dispersions (10.8 and 6.6 km/s). The jackknife tests in App. A.2.3 demonstrate morphological uniqueness, not physical association. If this co-location premise fails, every assigned vLOS and all derived energy and momentum maps are invalid. The input-vLOS truncation is acknowledged as a lower-limit effect, but the co-location assumption is not explicitly tested at Galactic scales.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a reconstruction of line-of-sight (LOS) velocities of the local interstellar medium within 1.25 kpc of the Sun and |b|<5 degrees, using the histogram of oriented gradients (HOG) method to correlate 3D dust density slices from Edenhofer et al. (2024) with HI and CO velocity channels. From the resulting distance-velocity association, the authors build face-on maps of LOS velocity, subtract a Galactic rotation model to define streaming motions, and derive effective densities, kinetic energy densities, momentum densities, and mass flow rates. The main claims are that most dust-associated gas follows the large-scale rotation pattern, that streaming-motion dispersions are about 10.8 km/s (HI) and 6.6 km/s (CO), and that the associated kinetic energy densities, about 0.11 and 0.04 eV/cm3, are comparable to other local ISM energy densities, with factor-of-ten overdensities toward the Radcliffe Wave, the Split, Vela C, and other structures.","tokens_in":45398,"tokens_out":5165,"duration_ms":58280,"significance":"If the central distance-velocity assignment is valid, this is a genuinely novel application of HOG-based kinetic tomography: it is the first use of high-quality 3D dust reconstructions to assign LOS velocities to distance channels over a large Galactic-plane volume, and it produces physically interpretable maps of streaming motions and kinetic energy density that can be compared with stellar kinematics and simulations. The paper is also strong in its extensive parameter testing (kernel size, tile segmentation, jackknife chance-correlation tests, expanded velocity windows in Appendices A and C), in its use of the public astroHOG code, and in its candid acknowledgment of several limitations, including the restricted input velocity range and the small VLBI maser sample. The astrophysical conclusions about near-equipartition and energy overdensities are interesting and potentially important, but they inherit the uncertainty of the underlying co-location assumption, which the current validation does not yet secure.","major_comments":[{"comment":"The central identification of a distance-velocity pair with physical co-location rests on an unstated premise: that the maximum Vd between a 3D dust distance slice and a line-emission velocity channel implies the same gas is seen in both. Vd measures morphological similarity, and that similarity can also arise when dust and HI/CO independently trace the same large-scale Galactic disk structure (warp, spiral-arm ridges, vertical stratification) over broad distance and velocity ranges. The current validation is too weak to exclude this: Appendix D tests one isolated simulated cloud without a background disk or multiple velocity components, and Appendix E contains only five VLBI masers, with residuals whose standard deviations (13.5 km/s for HI, 11.3 km/s for CO) exceed the reported streaming dispersions. A Galactic-scale synthetic test with a known distance-velocity relation, a background disk, and multiple LOS components is needed to establish that the assigned vLOS is the co-located velocity rather than a morphological coincidence; without such a test, the velocity field and all derived energy and momentum maps in Sec. 5 rest on an untested assumption.","section":"Sec. 3.3 and Eq. (2)"},{"comment":"The input line-emission window |vLOS|<25 km/s is chosen in Sec. 2 from the Reid et al. (2019) rotation model for d<1.25 kpc, and the streaming motions are then defined in Sec. 4.2 as vR19_LOS - vHOG_LOS. Consequently, the recovery of the rotation pattern in Fig. 11 and the near-zero mean streaming motions in Fig. 12 are partly built into the analysis rather than being independent discoveries. The paper acknowledges in Sec. 2 that the restricted window makes the reconstructed streaming amplitudes lower limits, but it does not explicitly state that the same window also guarantees that the recovered large-scale velocity field resembles rotation. This should be stated, and the central claim in the abstract and Sec. 7 should be tempered accordingly.","section":"Sec. 2 and Sec. 4.2"},{"comment":"The VLBI maser comparison provides only weak validation of the reconstructed velocities. The five masers within the volume have residuals with standard deviations of 13.5 km/s (HI) and 11.3 km/s (CO), both larger than the reported streaming-motion dispersions of 10.8 and 6.6 km/s. This means the maser sample cannot confirm the small-scale streaming pattern; it only places a coarse constraint on the method. The text should explicitly state that the maser comparison does not discriminate between the co-location interpretation and a morphological-coincidence interpretation, and it should describe what additional independent distance-resolved velocity constraints (for example, HI absorption toward continuum sources, or stellar absorption-line kinematics) would be needed to validate the reconstruction.","section":"Appendix E and Sec. 6.3"},{"comment":"The effective density used for all energy, momentum, and mass-flow quantities depends on an ad hoc block-level threshold Vd>1.0, and the paper does not report how neff, Ek, p, and Mdot vary when this threshold is changed over a plausible range. Since the factor-of-ten energy overdensities and the total kinetic energy estimates in Secs. 5.2 and 6.4 are derived from this quantity, a robustness test against this threshold is needed before the near-equipartition and overdensity claims can be considered secure.","section":"Sec. 5.1 and Eq. (3)"}],"minor_comments":[{"comment":"The caption states \"Values of Vd<2.87 correspond to mostly antiparallel gradients,\" which is a sign error; it should read \"Values of Vd<-2.87\" (or equivalently |Vd|>2.87 with negative sign).","section":"Fig. 4 caption"},{"comment":"In the sentence \"the most prominent of which is found around the location of the North America MC\" the text later reads \"700◦< d< 1100 pc toward l≈ 270◦,\" where the first quantity should be \"700 pc < d < 1100 pc\" rather than \"700◦.\"","section":"Sec. 6.4.2"},{"comment":"The comparison values for starlight and far-infrared radiation are given as \"0.54 and 0.31\" without units; they should be written as 0.54 eV/cm3 and 0.31 eV/cm3 to match the preceding quantities.","section":"Sec. 6.4.1"},{"comment":"The statement that Vd≈2.87 is \"roughly equivalent to a 3σ confidence interval\" is imprecise because the projected Rayleigh statistic's distribution and effective number of independent samples matter; it would be clearer to state the exact test or the empirical threshold used.","section":"Sec. 3.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is potentially important and the authors are clearly aware of many limitations, but the central co-location assumption is load-bearing and, in my reading, is not yet validated at the scales required by the claims. The single-cloud synthetic test and the five-maser comparison are insufficient, and the input velocity window makes part of the main rotational-pattern result constructional. I would encourage the editor to request a Galactic-scale synthetic test with a known distance-velocity field and realistic background disk, as well as a sensitivity analysis for the Vd>1.0 effective-density threshold, before considering the paper for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this is a real step forward in combining 3D dust maps with gas kinematics, and the authors are careful about their limitations, but the central trick — treating a high morphological correlation between a dust distance slice and a velocity channel as evidence that the same gas is at that distance and velocity — is load-bearing and only sparsely validated. I'd send it to review, but the referee should push for a Galactic-scale simulation test and for error bars on the derived maps.\n\nWhat's genuinely new: they take the HOG method, previously applied to Taurus, and run it across the full |b|<5 plane out to 1.25 kpc using the Edenhofer 3D dust map. That produces the first large-area kinetic tomography of the local ISM: face-on maps of LOS velocity, streaming motions, kinetic energy density, momentum density, and mass flow. The recovered large-scale quadrupolar rotation pattern is a nontrivial sanity check. They also test parameter choices (kernel size, tile segmentation, velocity range, jackknife chance-correlation limits, the 12 dust posterior samples, and Monte Carlo noise). The code and data are public. They explicitly state that the restricted input velocity window truncates streaming motions, so their quantitative values are lower limits. That is honest and correct.\n\nThe soft spots are just as the stress-test note says. The co-location premise — that maximum Vd means the dust and gas are physically associated — is never stated as an assumption and is not tested at Galactic scale. The synthetic test in App. D is one isolated simulated cloud, no background disk and no multiple velocity components. The five VLBI masers are too few and their scatter (13.5 and 11.3 km/s) is larger than the reported streaming dispersions (10.8 and 6.6 km/s). The input velocity window is chosen from the same rotation model used to define streaming motions, so the recovery of rotation is partly circular. And the energy/momentum maps come without propagated uncertainties. None of these is a fatal flaw, but together they mean the headline values should be treated as provisional.\n\nThe fix is concrete: run the HOG analysis on a Galactic-scale simulation with known distance-velocity structure and foreground/background confusion, compare the recovered velocities to the true ones, and publish uncertainty maps. I'd also ask the authors to state the co-location assumption explicitly and to discuss the circularity in Sec. 3.2 more directly.\n\nWho should read it: anyone working on the local ISM, 3D dust, or gas kinematics. It deserves a serious referee and, with the additions above, a place in A&A. I would not desk-reject it.","headline":"A serious and useful first large-area kinetic tomography of the local ISM, but the distance-velocity association rests on an under-validated co-location premise; publish after a Galactic simulation test and uncertainty maps.","tokens_in":46057,"tokens_out":3251,"would_cite":true,"duration_ms":36293,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper reconstructs line-of-sight motions of the local interstellar medium by matching 3D dust structures with HI and CO emission, and finds most gas follows Galactic rotation with local departures of 10.8 and 6.6 km/s.","keywords":["interstellar medium","kinetic tomography","3D dust maps","HI line emission","CO line emission","histogram of oriented gradients","Galactic rotation","streaming motions"],"falsifier":"A larger sample of independent distance-velocity anchors inside 1.25 kpc, such as parallax-measured masers or stars with known three-dimensional velocities, would settle whether the assigned velocities are real: unbiased HOG velocities should scatter randomly around the anchors, while systematic residuals that grow with distance or longitude would expose morphological misassociation.","tokens_in":44925,"feed_emoji":"🌌","tokens_out":9368,"duration_ms":90606,"temperature":0.7,"pith_summary":"The paper tries to turn the Solar neighborhood's interstellar medium into a four-dimensional picture: where each parcel of dust sits along the line of sight, and how fast it is moving toward or away from us. It does so by matching the shapes seen in a three-dimensional dust map with the shapes seen in hydrogen (HI) and carbon monoxide (CO) emission at different velocities, using the histogram of oriented gradients (HOG), a computer-vision measure of morphological similarity. The authors argue that a high match between a dust slice and a gas velocity channel means the same structure is being seen, so the gas velocity can be assigned to the dust distance. On this basis they reconstruct the line-of-sight velocity field within 1.25 kpc of the Sun and report that most gas follows Galactic rotation, with local streaming departures of roughly 10.8 km/s for HI and 6.6 km/s for CO. This matters because it yields a new, independent map of kinetic energy and momentum in the local interstellar medium, with values comparable to other interstellar energy densities and overdensities near structures such as the Radcliffe Wave.","feed_headline":"Nearby gas departs from Milky Way rotation by 10.8 and 6.6 km/s","feed_subtitle":"Computer-vision matching of 3D dust with HI and CO emission maps builds a 4D view of gas flows within 1.25 kiloparsecs.","key_machinery":"The load-bearing tool is the histogram of oriented gradients (HOG) method: a computer-vision comparison that computes the angle between spatial gradients of a 3D dust distance slice and a line-emission velocity channel, then aggregates those angles into a direction-sensitive projected Rayleigh statistic ($V_d$, Eq. 2). High positive $V_d$ means the two images' structures are morphologically similar; the paper uses this to assign the velocity of the best-matching line channel to each dust distance slice, and later to define effective densities that isolate the portions of a dust slice associated with each tracer.","core_discovery":"The paper's central claim is that morphological correlation between 3D dust density slices and line emission channels is strong enough to assign a line-of-sight velocity to each dust distance slice over most of the |b|<5° Galactic plane within 1.25 kpc. The reconstructed velocity field reproduces the quadrupolar pattern expected from Galactic rotation, and subtracting a standard rotation model leaves streaming departures with standard deviations of about 10.8 km/s for HI and 6.6 km/s for CO. The kinetic energy densities tied to these motions average about 0.11 and 0.04 eV/cm³, comparable to other local interstellar energy densities, while energy and momentum overdensities of about a factor of ten concentrate toward the Radcliffe Wave, the Split, and the Vela and Ara regions.","pith_inferences":["The ±25 km/s input clip means the quoted 10.8 and 6.6 km/s dispersions are probably floors; applying the same pipeline to future higher-resolution 3D dust reconstructions could recover larger streaming excursions that current angular resolution would misattribute to chance correlation.","A decisive check that would separate physical co-location from chance alignment is to compare HOG-assigned velocities with parallax-based three-dimensional velocities of many individual stars inside 1.25 kpc; the current five-maser comparison is too sparse to settle the question.","The roughly 15 km/s velocity dispersion between HI and CO at the same distance slices suggests the method could be repurposed as a probe of momentum coupling between warm atomic and cold molecular gas in feedback regions, a use the paper only begins to explore."],"forward_implications":["The HOG method yields the first quantitative global map of line-of-sight velocities for the interstellar medium within 1.25 kpc, combining HI and CO tracers into a distance-velocity picture rather than relying on kinematic-distance assumptions.","The bulk of the local gas moves with Galactic rotation, so the large-scale quadrupolar velocity pattern is present even in this small volume; departures from it are modest in the mean but reach standard deviations of 10.8 km/s for HI and 6.6 km/s for CO.","The kinetic energy densities from streaming motions are about 0.11 eV/cm³ (HI) and 0.04 eV/cm³ (CO), comparable to thermal, magnetic, cosmic-ray, and radiation energy densities in the local ISM, supporting near-equipartition.","Energy and momentum overdensities by roughly a factor of ten concentrate toward the Radcliffe Wave, the Split, Vela, and Ara, while no clear imprint of the local spiral arm is found; the distribution points to a combination of large-scale forcing and supernova feedback.","Because the input velocity range is clipped to ±25 km/s, the reconstructed streaming amplitudes and derived energies are likely lower limits rather than full values."],"supporting_citations":[{"why":"Supplies the 3D dust density reconstruction between 69 and 1250 pc that fixes the distance channels used in the morphological matching.","marker":"Edenhofer et al. (2024)"},{"why":"Introduces the histogram of oriented gradients method for quantifying morphological correlation between images.","marker":"Soler et al. (2019)"},{"why":"Provides the projected Rayleigh statistic that the paper adapts into its direction-sensitive correlation metric $V_d$.","marker":"Jow et al. (2018)"},{"why":"Delivers the all-sky HI 21-cm line emission observations used as one of the two velocity tracers.","marker":"HI4PI Collaboration et al. (2016)"},{"why":"Supplies the CO (J=1–0) survey that defines the |b|<5° coverage and the molecular-gas velocity tracer.","marker":"Dame et al. (2001)"},{"why":"Defines the Galactic rotation model from which expected line-of-sight velocities and streaming motions are computed.","marker":"Reid et al. (2019)"},{"why":"The earlier kinetic-tomography reconstruction that provides the baseline for comparing the HOG-based velocity field.","marker":"Tchernyshyov & Peek (2017)"},{"why":"Identifies the local structures (Radcliffe Wave, Split, Local Arm) used to contextualize the energy overdensities.","marker":"Zucker et al. (2023)"},{"why":"Provides the cluster families and supernova counts used to argue that stellar feedback contributes to the reconstructed energy input.","marker":"Swiggum et al. (2024)"}],"fun_headline_variants":["Gas flows near Sun mapped in 4D, depart from Galactic rotation","Local ISM kinetic energy near equipartition, overdensities x10","Computer vision links dust and gas to reveal streaming motions","4D map of local ISM: rotation plus streaming, energy balanced"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method assumes that when a dust slice and a gas velocity channel look alike on the sky, they are physically the same gas; if the similarity is a coincidence or comes from large-scale structure, the assigned velocities would not be real.","fun_headline_variants_meta":{"raw":{"variants":["Gas flows near Sun mapped in 4D, depart from Galactic rotation","Local ISM kinetic energy near equipartition, overdensities x10","Computer vision links dust and gas to reveal streaming motions","4D map of local ISM: rotation plus streaming, energy balanced"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00021,"raw_usage":{"total_tokens":1491,"prompt_tokens":1103,"completion_tokens":388,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":719,"completion_tokens_details":{"reasoning_tokens":312}},"tokens_in":719,"tokens_out":388,"duration_ms":4971,"temperature":1.0,"reasoning_tokens":312,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:45:09.599071+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A larger sample of independent distance-velocity anchors inside 1.25 kpc, such as parallax-measured masers or stars with known three-dimensional velocities, would settle whether the assigned velocities are real: unbiased HOG velocities should scatter randomly around the anchors, while systematic residuals that grow with distance or longitude would expose morphological misassociation.","supporting_citations":[{"cited_title":"D., Beuther , H., Rugel , M., et al","cited_arxiv_id":null,"evidence_quote":"Introduces the histogram of oriented gradients method for quantifying morphological correlation between images."},{"cited_title":"& Peek , J","cited_arxiv_id":null,"evidence_quote":"The earlier kinetic-tomography reconstruction that provides the baseline for comparing the HOG-based velocity field."}],"review_version":1}