{"id":"711bbc3c-bc40-4687-b5a4-69643c45d42a","arxiv_id":"2411.14956","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A VGOS-only celestial reference frame from 155 sessions, 418 sources, with median formal errors of 30 to 47 microarcseconds, improves geodetic VLBI baseline and Earth orientation estimates.","lead":"Astronomers built a celestial reference frame from five years of new broadband VGOS radio telescope observations, cataloging 418 radio sources with median positional precision near 30 to 47 microarcseconds. The frame improves baseline length and Earth orientation estimates by up to a few millimeters, a step toward the 1mm global reference frame goal.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Self-consistency of using the same 155 sessions to build and test the CRF is the key unresolved threat; a holdout test would settle whether the 1-3 mm geodetic gains are real.","rationale":"The reader's weakest_assumption correctly identifies the same central issue: the improvement from fixing to VIE2023-VG in Sect. 6.3 risks being a self-consistency artifact because the CRF is estimated and tested on the same 155 sessions. I agree that this is the load-bearing concern and that it is specific and addressable through holdout validation, external comparison, or error calibration. I add two supporting details from the manuscript: (1) the datum constraints in Sect. 4 (NNR on 71 sources plus a forced zero-slope declination constraint) mean that part of the frame is intentionally tied to ICRF3-SX, so the VSH comparison is not an independent accuracy test; (2) the paper itself states in Sect. 5 that accuracy follow-up is needed and that raw formal errors are not realistic uncertainties, which weakens the precision half of the central claim. However, these points do not make the paper rejectable: the product is new, the pipeline is documented, the calibration caveat is explicit, and the proposed tests are feasible. The appropriate outcome remains conditional acceptance, with the revision expected to include a holdout or external validation and calibrated uncertainties. I would recommend UNCHANGED relative to the reader's CONDITIONAL verdict rather than a stricter one, because the concern, while real, is an evaluation-design issue that the authors can address without invalidating the underlying CRF construction.","tokens_in":24886,"tokens_out":1801,"duration_ms":15626,"concrete_test":"Perform a holdout validation: split the 155 VGOS sessions into two disjoint halves (e.g., by time or by session index), build a CRF from half A, and evaluate single-session baseline/EOP wrms on half B by fixing to that held-out CRF versus ICRF3-SX. If the wrms improvement persists at the 1-3 mm level, the artifact explanation fails; if it shrinks substantially or changes sign, the reported improvement is partly self-consistency and the headline should be reworded. An additional cross-check is to repeat the S1/S2 comparison fixing to VIE2023sx (a CRF built from S/X data through 2024.0) to separate frequency-consistency from session-consistency effects.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline geodetic claim (Sect. 6.3) is that fixing VIE2023-VG instead of ICRF3-SX lowers baseline-length wrms by up to 3 mm and EOP wrms by 10-15%. But the compared catalogs are not symmetric: ICRF3-SX was built from S/X data through 2018 and is not adjusted to VGOS, whereas VIE2023-VG was estimated from the very same 155 VGOS sessions (plus the same NNT/NNR/zero-slope datum constraints) that are then analyzed in S2. Any session-specific systematic error absorbed into the VGOS source positions will be re-absorbed consistently when those positions are fixed in single-session solutions, artificially reducing the scatter. The paper explicitly flags network imbalance and a hemisphere-dependent zero-slope constraint (Sect. 4) and shows large VSH systematic terms (Sect. 6.1, Table 5), so the possibility that the improvement is partly a self-consistency artifact is concrete, not hypothetical. The comparison to ICRF3-SX is also confounded by the frequency difference (S/X vs broadband), and the paper does not compare against an independently estimated VGOS CRF or an S/X CRF with VGOS data included (VIE2023sx is used only for source-position comparisons, not for the geodetic test). The formal-error precision claims are explicitly acknowledged as uncalibrated accuracy (Sect. 5), so the central quantitative evidence for the new frame's value rests on this circularity-sensitive internal test.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs VIE2023-VG, a celestial reference frame estimated from 155 multi-baseline 24-hour VGOS sessions up to 2024.0, using 1.39 million group delays and yielding 418 source positions. It documents VGOS source selection and scheduling, compares the new frame with ICRF3-SX and with the S/X-based VIE2023sx frame using normalized angular separations and vector spherical harmonics, tests alternative tropospheric delay estimation intervals, and evaluates the frame geodetically by fixing source coordinates in single-session analyses to either VIE2023-VG or ICRF3-SX. The headline claims are median formal errors of 30 microarcseconds in right ascension and 47 microarcseconds in declination, VSH distortions up to about 60 microarcseconds relative to ICRF3-SX, and improvements in baseline-length weighted root mean square of up to 3 mm and in EOP wrms of approximately 10-15% when the VGOS-based frame is fixed instead of ICRF3-SX.","tokens_in":25130,"tokens_out":4390,"duration_ms":45558,"significance":"If the frame and its evaluation are accepted, the paper would be a timely and useful contribution to the VLBI transition to VGOS: it provides a publicly available VGOS-only CRF, carefully documents scheduling strategies, and quantifies the impact of source-position priors on geodetic products. The detailed treatment of the VGOS-R&D sessions and the source-centric scheduling approach is a genuine strength, as is the provision of the catalog through a DOI. However, the geodetic validation is not independent of the frame construction, because the same 155 sessions are used to estimate VIE2023-VG and then to test it, and the VSH comparisons are partially constrained by the datum definition. Those limitations do not invalidate the catalog itself, but they mean the headline accuracy improvements should be viewed as provisional until an independent validation is performed.","major_comments":[{"comment":"The central geodetic evaluation is self-referential: the VIE2023-VG CRF used in solutions S2 and S4 is estimated from exactly the same 155 VGOS sessions that are then analyzed session-wise with that frame fixed. Any session-specific systematic errors absorbed into the global source coordinates, for example from the sparse southern network or from tropospheric mismodeling, will be consistently re-absorbed when those coordinates are fixed in the single-session solutions, artificially reducing the baseline-length and EOP scatter relative to a comparison with ICRF3-SX, which shares no data with the VGOS sessions. The reported improvements of up to 3 mm in baseline length and 10-15% in EOP wrms are therefore not an independent measure of the frame's accuracy. I recommend either a holdout test (estimating the CRF from one subset of sessions and evaluating it on the remaining sessions) or a comparison against an independently estimated VGOS CRF from another analysis center, and the manuscript should state clearly that such validation has not yet been performed.","section":"Sect. 6.3, Table 8, Fig. 15"},{"comment":"The VSH comparison with ICRF3-SX is partly forced by the datum definition. As described in Sect. 4, the global solution applies an NNR condition on 71 ICRF3 defining sources plus a zero-slope constraint on the declination differences with respect to ICRF3-SX. Consequently, the rotation and declination-slope components of the VSH fit in Table 5 are not free parameters, and their small or moderate values do not by themselves demonstrate that the frame has small systematic distortions relative to ICRF3-SX. The ambiguity is visible in the text: the 16-parameter VSH solution gives R1 = +56 +/- 16 microas, while the three-rotation-only solution gives R1 = -8 +/- 10 microas, confirming strong correlations among the VSH components. The abstract and conclusion statements that the largest systematic distortions 'reach about 60 microas' should be reworded to state which components are affected by the datum constraints and to avoid implying that the distortions are independently measured.","section":"Sect. 4 and Sect. 6.1, Table 5"},{"comment":"The claim of 'excellent source position precision' rests on raw formal errors from the least-squares adjustment. The authors themselves note in Sect. 5 that a careful accuracy investigation is needed before reporting realistic uncertainties, and the ICRF3 analysis applies a 1.5 scaling factor and a 30 microas noise floor to formal errors. Given that the comparison with ICRF3-SX in Sect. 6.1 yields wrms differences of 80-100 microas for well-behaved sources, the median formal errors of 30/47 microas should not be presented without an explicit caveat that they are uncalibrated precision estimates. At minimum, the abstract should be changed from 'excellent source position precision' to a statement that the formal precision is excellent but unvalidated, or the paper should provide an independent accuracy assessment such as comparison with an external catalog that accounts for the datum and noise-floor issues.","section":"Sect. 5, Table 2, Abstract"}],"minor_comments":[{"comment":"The paper refers to ten VGOS-R&D sessions, but Sect. 2 names VR2101, VR2201-VR2206, VR2301, VR2302, and VR2304 and excludes VR2303 because its database was not released; please state explicitly that the ten analyzed R&D sessions exclude VR2303.","section":"Sects. 1-2"},{"comment":"The figure legend uses 'gr180' while the text and Table 7 use 'zwd30gr180' or 'grd180'; unify the notation to avoid confusion.","section":"Fig. 11"},{"comment":"The table note 'The second and third columns show comparisons...' is slightly ambiguous because the table has five numeric columns; rephrase to identify columns 2-3 and 4-5 by their headings (ICRF3-SX and VIE2023sx, for Dn<15 and Dn<5).","section":"Table 5 notes"}],"recommendation":"major_revision","confidential_remarks":"The manuscript appears to be the accepted A&A version, and my concerns are interpretive rather than about data accumulation. The catalog and scheduling documentation are valuable community products, but the paper's strongest quantitative claims rely on a self-consistency test and on uncalibrated formal errors. A major revision that adds a holdout or independent validation, or that explicitly re-scopes the claims, would make the contribution reliable for the readership."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is the first VGOS-only celestial reference frame, built from 155 sessions and 1.39 million delays, and the authors make the catalog publicly available. That alone makes it worth a serious referee. The paper also does the community a service by documenting source selection and scheduling strategies in detail, and by testing shorter tropospheric estimation intervals. The scheduling analysis is genuinely informative: the source-centric approach in the R&D sessions produces much more balanced scan distributions, and the quantitative comparison is convincing.\n\nThe main soft spot is the evaluation in Sect. 6.3. The authors compare single-session solutions where source positions are fixed to either ICRF3-SX or their new VIE2023-VG CRF. But the VGOS CRF was estimated from exactly the same 155 sessions being tested. Any session-specific error absorbed into the global source positions will be re-absorbed consistently when those positions are fixed, artificially lowering the baseline-length wrms. The up-to-3 mm improvement on long baselines, and the 10-15% EOP wrms gains, could be partly a self-consistency artifact. The paper does not address this, and it is the central quantitative evidence for the frame's value.\n\nThe comparison to ICRF3-SX is also confounded by frequency (broadband vs S/X) and by time span, though the authors do compare to VIE2023sx for source positions. The VSH analysis shows that the large systematic terms (around 60 microas) shrink to below 30 microas when the outlier threshold is tightened from Dn<15 to Dn<5, so those distortions are not robust. The precision claims rest on raw formal errors, which the authors explicitly flag as uncalibrated. These are not fatal flaws, but they are real limits on what can be concluded right now.\n\nOn the positive side, the paper is transparent about the southern hemisphere problem, the zero-slope constraint, and the Hb down-weighting. The outlier source investigation, with structure indices and time series from both S/X and VGOS, is careful and informative. The data product itself is reproducible.\n\nMy recommendation: send it to peer review, but ask the authors to validate the geodetic improvement with a holdout test (e.g., build the CRF from a subset of sessions and test on the rest) or against an independent VGOS CRF. Even a direct discussion of the circularity and its likely magnitude would strengthen the paper. The frame itself is a useful product, and the scheduling and tropospheric results stand on their own; the geodetic claim just needs a cleaner experiment.","headline":"First VGOS-only CRF with a public catalog and careful documentation, but the headline geodetic gain needs a holdout test to rule out self-consistency.","tokens_in":25791,"tokens_out":2929,"would_cite":true,"duration_ms":26598,"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":"Five years of VGOS broadband VLBI data define a 418-source celestial frame with sub-50-microarcsecond formal errors that improves geodetic baseline analysis.","keywords":["celestial reference frame","VGOS","VLBI","ICRF3-SX","source position","baseline length repeatability","Earth orientation parameters","vector spherical harmonics"],"falsifier":"Split the 155 VGOS sessions into two disjoint halves, estimate VIE2023-VG-style CRFs from each half, fix each frame in the single-session analysis of the other half, and compare baseline-length and EOP wrms against ICRF3-SX; if the up-to-3 mm improvement disappears, the current comparison is largely a self-consistency effect. A second check is to remove the zero-slope declination constraint and see whether the VSH distortion terms grow, which would show that the constraint, rather than the data, is carrying the frame's declination alignment.","tokens_in":24585,"feed_emoji":"📡","tokens_out":5700,"duration_ms":47115,"temperature":0.7,"pith_summary":"The paper reports the first celestial reference frame built from five years of VGOS broadband VLBI observations alone, and argues that this frame is both precise enough and better suited to geodetic use than the adopted S/X-based ICRF3-SX. The frame, called VIE2023-VG, contains 418 radio sources from 1.39 million group delays across 155 sessions, with median formal errors of 30 $\\mu$as in $\\alpha^* = \\alpha\\cos\\delta$ and 47 $\\mu$as in declination. Fixing this VGOS frame instead of ICRF3-SX in the analysis of individual sessions lowers the weighted rms of baseline lengths by up to roughly 3 mm on the longest baselines, and lowers the scatter of all five Earth-orientation parameters by about 10–15%. The paper also shows that four specialized research-and-development sessions introduced 172 new sources, and that denser tropospheric estimation improves declination formal errors for mid-declination sources. This matters because a VGOS-only frame is needed to approach the GGOS goal of a terrestrial reference frame accurate to 1 mm.","feed_headline":"VGOS-only radio frame cuts long-baseline errors by 3 mm","feed_subtitle":"A VGOS-only frame fixes 418 radio sources at ~30 microarcseconds and reduces baseline and Earth-orientation scatter.","key_machinery":"The load-bearing objects are the group delays from VGOS broadband sessions, estimated in a global adjustment with piece-wise linear offsets for clocks, zenith wet delay, tropospheric gradients, Earth-rotation parameters, and celestial pole offsets. The frame is aligned to ICRF3-SX through a no-net-rotation condition on 71 ICRF3 defining sources with more than 1000 observations, plus a zero-slope constraint in the declination difference. Comparisons are made with vector spherical harmonics (VSH) up to degree and order two, and with the normalized angular separation $D_n$ built from coordinate differences scaled by the combined formal errors; the VSH analysis is what turns raw position differences into statements about systematic rotation, dipole, and quadrupole distortions. In the geodetic evaluation, baseline-length wrms and the five Earth-orientation parameters computed with the frame fixed to VIE2023-VG versus ICRF3-SX form the quantitative bridge from astrometry to geodesy.","core_discovery":"The central claim is that VIE2023-VG, a global least-squares CRF estimated from VGOS group delays only, achieves source-position formal errors of 30 $\\mu$as in $\\alpha^*$ and 47 $\\mu$as in $\\delta$ (median over all 418 sources), and that using this VGOS-based frame as the fixed a priori in single-session geodetic analysis improves baseline-length repeatability by up to 3 mm on baselines longer than about 12,000 km, with weighted-mean baseline-length offsets up to 2 mm. In systematic terms, the largest vector-spherical-harmonic distortions relative to ICRF3-SX reach about 60 $\\mu$as for degree and order up to two, and these reduce by roughly a factor of two when the 85 sources with normalized angular separation $D_n>5$ are removed, implicating source structure rather than a global frame defect. Because southern-hemisphere observations are scarce, the solution applies a zero-slope constraint on the declination difference versus ICRF3-SX on top of the no-net-rotation condition; the authors take this as evidence that adding southern VGOS stations is a prerequisite for a fully self-consistent VGOS frame.","pith_inferences":["If the baseline-length improvement is genuine and not a self-consistency artifact, then every geodetic analysis of VGOS data should adopt a VGOS-only CRF, and the 1 mm GGOS target depends less on the frame's formal precision than on southern network geometry.","The roughly linear growth of the wrms improvement with baseline length (about 0.14 mm per megameter) suggests the gain is dominated by declination errors mapping into low-elevation observations and into the north component of station positions, which is testable by comparing north versus east station-position scatter.","A stronger test would split the 155 sessions into two disjoint halves, estimate a CRF from each half, fix each frame in the single-session analysis of the other half, and re-measure the baseline-length and EOP gains; if the up-to-3 mm improvement survives disjoint estimation, it is a real frame-quality gain.","The same VSH machinery applied to future VGOS frames with southern stations would show whether the roughly 60 $\\mu$as distortion terms shrink or rotate, giving a quantitative check on how much the network geometry contributes to the frame."],"forward_implications":["VGOS-only session analysis can replace ICRF3-SX as the operational a priori, improving the height and north components of station positions at essentially all stations.","The source-centric scheduling strategy used in VR2301 and VR2302 should become the default for VGOS, since it roughly doubles source counts per session with a much more even scan distribution.","The next bottleneck is southern sky coverage: frame distortions and the need for the zero-slope declination constraint will persist until more southern VGOS stations join the network.","Denser tropospheric estimation (5–10 minute zenith wet delay) benefits mid-declination sources but hurts far-southern sources until southern stations gain better sky coverage.","The outlier analysis implies that a combined S/X and VGOS frame would degrade the best VGOS-only positions, so frequency-separated frames, not one mixed frame, are the path to highest accuracy."],"supporting_citations":[{"why":"Supplies ICRF3-SX, the adopted reference frame that all VGOS comparisons and the fixed a priori baseline solution S1 rest on.","marker":"Charlot et al. 2020"},{"why":"Provides the vector spherical harmonics formalism used to quantify systematic rotations, dipoles, and quadrupoles between VIE2023-VG and ICRF3-SX.","marker":"Mignard & Klioner 2012"},{"why":"Defines the normalized angular separation statistic $D_n$ used to identify outlier sources.","marker":"Mignard et al. 2016"},{"why":"Describes the source-centric scheduling method fully utilized in VR2301 and VR2302 that extended the VGOS source list.","marker":"Schartner et al. 2023"},{"why":"Sets the analysis parameterization and supplies the VIE2023sx S/X frame used as a comparison solution.","marker":"Krásná et al. 2023"},{"why":"Defines the source structure index used to explain why the outlier sources show large position differences.","marker":"Fey & Charlot 1997"},{"why":"Provides the earlier study of CRF effects on EOP repeatabilities against which the improvement claim is framed.","marker":"Diamantidis & Haas 2023"}],"fun_headline_variants":["VGOS frame pins 418 sources to 30 microarcseconds","New VGOS frame cuts baseline errors by 3 mm","418-source VGOS frame hits 30 microarcseconds","VGOS frame improves source precision, cuts errors by 3 mm"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The evaluation of the frame's geodetic benefit assumes that fixing a frame derived from the same 155 VGOS sessions is a fair comparison; if the global solution absorbs session-specific errors, the reported baseline-length and EOP improvements would shrink or vanish when the frame is tested on data that did not contribute to it.","fun_headline_variants_meta":{"raw":{"variants":["VGOS frame pins 418 sources to 30 microarcseconds","New VGOS frame cuts baseline errors by 3 mm","418-source VGOS frame hits 30 microarcseconds","VGOS frame improves source precision, cuts errors by 3 mm"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001096,"raw_usage":{"total_tokens":4711,"prompt_tokens":1215,"completion_tokens":3496,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":831,"completion_tokens_details":{"reasoning_tokens":3423}},"tokens_in":831,"tokens_out":3496,"duration_ms":23808,"temperature":1.0,"reasoning_tokens":3423,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:40:38.514821+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Split the 155 VGOS sessions into two disjoint halves, estimate VIE2023-VG-style CRFs from each half, fix each frame in the single-session analysis of the other half, and compare baseline-length and EOP wrms against ICRF3-SX; if the up-to-3 mm improvement disappears, the current comparison is largely a self-consistency effect. A second check is to remove the zero-slope declination constraint and see whether the VSH distortion terms grow, which would show that the constraint, rather than the data, is carrying the frame's declination alignment.","supporting_citations":[{"cited_title":"S., Gordon, D., et al","cited_arxiv_id":null,"evidence_quote":"Supplies ICRF3-SX, the adopted reference frame that all VGOS comparisons and the fixed a priori baseline solution S1 rest on."},{"cited_title":"& Klioner, S","cited_arxiv_id":null,"evidence_quote":"Provides the vector spherical harmonics formalism used to quantify systematic rotations, dipoles, and quadrupoles between VIE2023-VG and ICRF3-SX."},{"cited_title":"H., & Soja, B","cited_arxiv_id":null,"evidence_quote":"Describes the source-centric scheduling method fully utilized in VR2301 and VR2302 that extended the VGOS source list."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the source structure index used to explain why the outlier sources show large position differences."},{"cited_title":"& Haas, R","cited_arxiv_id":null,"evidence_quote":"Provides the earlier study of CRF effects on EOP repeatabilities against which the improvement claim is framed."}],"review_version":1}