{"id":"a2e3c2ac-e8f8-4323-b021-a1aa1aa7f401","arxiv_id":"2506.14894","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"QRaFT traces quasi-radial structures in coronagraph images and aligns with the model magnetic field to within about 4 to 7 degrees on synthetic tests.","lead":"This paper describes QRaFT, an image-processing method that traces quasi-radial coronal structures in coronagraph photos to estimate the direction of the Sun's open magnetic field. Tests on synthetic MHD images show alignment to the model field within a few degrees, and the authors argue the outputs can help validate and improve solar wind and space weather models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 4–7° alignment is an in-sample, post-filter statistic computed on the same MAS snapshot used for parameter tuning; out-of-sample accuracy is undemonstrated.","rationale":"The reader's weakest assumption is that validation relies on the MAS thermodynamic simulation rather than on real coronal observations. My stress-test sharpens this into a more specific and immediately checkable issue: even within the model, the 4–7° figure is not an out-of-sample measurement. The processing parameters in Table 2 are different for MAS, COR1, and TSE images, and Section 2.2 explicitly says the parameters are optimized for each image type; yet the same MAS snapshot is used both to select parameters and to evaluate accuracy. Furthermore, Fig. 7 statistics appear to cover only validated features, and the validation filters (Eqs. 25–31) preferentially remove short, curved, or faint features—exactly the detections one would expect to be poorly aligned. The paper is transparent about many of these points, including the closed-loop limitation and the 14-day-old boundary data for COR1, which counts in its favor. But transparency does not remove the overfitting risk. A second-simulation, no-retuning test would settle whether the 4–7° alignment is a property of QRaFT or of the particular MAS run and tuning. I do not think this warrants rejection: the method is clearly described, the code is public, and the in-model demonstration is a reasonable first validation. Conditional acceptance is the right level, consistent with the reader's verdict, so I recommend UNCHANGED.","tokens_in":21391,"tokens_out":4026,"duration_ms":44886,"concrete_test":"Generate synthetic pB images from a second, independent thermodynamic MHD corona (e.g., a different MAS eclipse run or an AWSoM solution) at an unseen vantage point. Apply QRaFT with the exact Table 2 'PSI MAS' keyword set, without any retuning, and compute Eq. 24 misalignment angles against that run's magnetic field. Report mean/median |θ| and P(|θ|<10°) for both validated and unfiltered features. If these numbers do not reproduce the 4–7° / ~80% benchmarks, the reported accuracy is an artifact of in-sample parameter selection rather than a method-level guarantee.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim is computed entirely from a single MAS solution (Section 3.1): both the synthetic density/pB images and the ground-truth magnetic field come from the same run, and the QRaFT processing keywords (Table 2) are customized for that data source. The paper states in Section 2.2 that these parameters must be optimized for image resolution, noise, and radial range, and it provides no train/test separation: the same snapshot that was used to choose the parameters is also used to measure the 4–7° misalignment. In addition, the θ statistics of Fig. 7 are computed for features that survive the validation filters (Eqs. 25–31), which discard short, curved, or faint detections. Because those filters preferentially remove the most misaligned and closed-loop features, the quoted alignment is a post-selection property of the pipeline, not an unconditional accuracy of tracing. If these settings and filters are what make QRaFT accurate, then the claim 'extracted optical features are aligned within ~4–7 degrees' may not transfer to real coronagraph images, where the density-to-field correspondence is different and no model field is available to guide filtering. This is the load-bearing assumption for the paper's usefulness as an empirical open-flux proxy.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper introduces QRaFT, an image-processing pipeline that detects quasi-radial, field-aligned structures in white-light coronagraph images and uses their local orientation as a proxy for the plane-of-sky direction of the open coronal magnetic field. The method is described in detail: radial detrending, transformation to polar coordinates, anisotropic smoothing, unsigned second-order azimuthal differencing, adaptive percentile thresholding with a sliding inner boundary, cluster labeling, polynomial interpolation of feature nodes, computation of local orientation angles, and automatic feature validation filters. Validation is performed on synthetic images from a single thermodynamic MAS MHD simulation—both a central-plane density array and a FORWARD-computed pB image—where the ground-truth magnetic field is known, and the authors report a characteristic misalignment of about 4–7 degrees for surviving features. The paper also demonstrates the pipeline on a STEREO/COR1 image and on a 2017 total-solar-eclipse image, and it argues that QRaFT provides an empirical constraint for coronal and solar-wind models.","tokens_in":21745,"tokens_out":3564,"duration_ms":54130,"significance":"If the stated accuracy holds in an out-of-sample sense, QRaFT would fill a genuine gap: it is, to my knowledge, the first published method aimed specifically at tracing open-flux coronal structures in routinely collected non-eclipse coronagraph images. The strengths of the paper are substantial: the algorithm is specified in enough mathematical detail to be reimplemented; the source code is publicly available on GitHub; the synthetic test is a real consistency check, because the algorithm does not take the model magnetic field as input and no equation reduces to fitted values; and the authors are unusually candid about the limitations, including closed-loop contamination, the model-data mismatch in the COR1 comparison, and the deferral of systematic validation to a companion paper. The principal weakness is that the headline 4–7 degree claim is measured on the same model snapshot used to tune the processing parameters and only on features that survive the validation filters, so its transferability to real observations is not established.","major_comments":[{"comment":"The central quantitative claim—that QRaFT features are aligned within ~4–7 degrees of the model magnetic field—is computed from a single MAS snapshot, and the QRaFT processing keywords in Table 2 (PSI MAS column) are customized for that data source, as Section 2.2 states that these parameters must be optimized for image resolution, noise, and radial range. Moreover, the statistics in Fig. 7 are computed only for features that survive the validation filters of Eqs. (25)–(31), which reject short, curved, and faint features. Because those filters preferentially remove the least field-aligned and most loop-like detections, the quoted accuracy is a post-selection property of the pipeline, not an unconditional property of the tracing. The paper provides no train/test separation, no second model snapshot, and no independent model check. To support the abstract's claim, please provide out-of-sample validation (e.g., a different MAS time step or a different MHD model) and, ideally, report the misalignment statistics for all features before filtering as well as for the validated subset.","section":"§3.1, Fig. 7, Table 2"},{"comment":"The COR1 comparison is presented as a performance demonstration, but the ground truth is the same MAS model whose boundary magnetograms predate the COR1 observation by up to 14 days (Section 3.1). The measured degradation in Fig. 9 (P(10°) = 51.8%) is therefore an inseparable mixture of QRaFT error and model-data mismatch, as the authors themselves acknowledge. This means the manuscript contains no quantitative validation of QRaFT against an independent source of real-image ground truth. If the paper is to support the claim that QRaFT transfers to real coronagraph images, the authors need either to compare against a more contemporaneous model run or to state explicitly, in the abstract and conclusions, that real-image validation is currently qualitative only.","section":"§3.2, Fig. 9"},{"comment":"The paper defers the 'more systematic performance and error analysis' to Rura et al. (2025), which is cited as under review. The present manuscript thus does not, on its own, contain the evidence needed to support the headline 4–7 degree accuracy claim beyond a single, in-sample, post-filter measurement. I recommend that the authors make the current manuscript self-contained for its main claim—for example, by including a multi-snapshot or multi-viewing-angle error analysis, or by explicitly rephrasing the claim as a preliminary finding pending the companion paper. This is a load-bearing issue because the abstract's numerical claim is the paper's primary takeaway.","section":"§4, §3.1"}],"minor_comments":[{"comment":"In the paragraph following Eq. (4), the sentence 'the detrended image Idetr(x, y) undergoes is transformed into plane polar coordinates' contains a grammatical error ('undergoes is transformed'); please remove 'is'.","section":"§2.2"},{"comment":"There is a typo in the text after Eq. (25): 'local magnetic filed orientation' should read 'local magnetic field orientation'.","section":"§2.5, Eq. (25) vicinity"},{"comment":"The phrase 'The first raw of the panels' in the caption of Fig. 7 should be 'The first row of the panels'.","section":"Fig. 7 caption"},{"comment":"The section mentions that some visually identifiable quasi-radial structures are missed by QRaFT, but no quantitative detection-completeness measure is provided; a simple statement of the fraction of visual structures recovered would strengthen the demonstration.","section":"§3.3, Fig. 11"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is clearly one of a two-paper set, with the systematic validation deliberately placed in a companion paper under review. The editor may wish to consider whether the main quantitative claim should be peer-reviewed in this journal before publication, or whether the current paper should be reframed as a methods description with the accuracy claim explicitly marked as preliminary. There is also a novelty claim in §2.1 ('no method other than QRaFT is currently available') that, as written, is stronger than the cited literature supports; eclipse-image tracing methods exist, and the novelty lies specifically in routine non-TSE open-field tracing, which is a narrower and defensible claim. These are scope and framing concerns rather than technical errors."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe useful take: this is a real methodological advance. QRaFT is the first pipeline I know of that traces quasi-radial open-flux structures in ordinary coronagraph images, not just eclipse shots, and the code is public. The core idea—enhance azimuthal gradients via second-order differencing in polar coordinates, then adaptively threshold and trace—is clearly explained, the math is simple and reproducible, and the paper is refreshingly honest about its limitations. It even warns that the COR1 comparison says more about model error than tracer error.\n\nWhat's new: the combination of polar transform, second-order azimuthal differencing, adaptive thresholding, polynomial feature tracking, and validation filters, packaged as an open-source IDL tool. That is a legitimate contribution with practical value for model validation and space weather.\n\nThe soft spot is precisely the validation. The headline 4–7° alignment comes from a single MAS snapshot, with the processing parameters in Table 2 tuned on that same snapshot, and the statistics are computed only for features that survived the validation filters (Eqs. 25–31). Those filters delete short, curved, faint, and highly misaligned detections, so the quoted accuracy is a post-selection property. The paper does not report all-feature statistics or how many features were discarded. The real-image demonstrations are qualitative: COR1 is cross-calibrated against an outdated model snapshot, and the TSE example has no ground-truth comparison. So the claim that QRaFT is accurate on real data is asserted rather than demonstrated.\n\nI don't think this is a fatal flaw. The paper never claims out-of-sample accuracy, and it points to a companion paper for more systematic validation. The method does not use the model field as input, so circularity is not an issue. Still, before this is published, a serious referee should ask for: train/test separation (tune on one snapshot, test on another), statistics with and without filtering, and ideally an independent observational benchmark. Those are fixable.\n\nVerdict: send to peer review. The method is solid, the code is available, and the limitations are acknowledged. It just needs a stricter validation to back up the accuracy headline.\n\nRecommendation: engage with it, but expect revision.","headline":"A useful, clearly-described open-source tracing method whose headline accuracy number is in-sample and post-filter; worth peer review but needs stricter validation.","tokens_in":22239,"tokens_out":2960,"would_cite":true,"duration_ms":34233,"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":"QRaFT recovers field-aligned structures in coronagraph images, with numerical tests showing 4–7 degree alignment to the underlying MHD magnetic field.","keywords":["solar corona","coronagraph image segmentation","open magnetic flux","quasi-radial field-line tracing","MHD simulation","white-light imaging","space weather","plane-of-sky magnetic field"],"falsifier":"Apply QRaFT to synthetic white-light images from a structurally different, independently built coronal MHD model; if the median misalignment against that model's magnetic field clearly exceeds 7 degrees, the claimed accuracy is a property of the test model rather than a general property of the tracing method.","tokens_in":21232,"feed_emoji":"☀️","tokens_out":9513,"duration_ms":83086,"temperature":0.7,"pith_summary":"This paper presents QRaFT, an image-processing method for tracing quasi-radial, field-aligned structures in white-light images of the faint open solar corona. Its central claim is that the traced features approximate the plane-of-sky orientation of the steady-state open magnetic field: in synthetic images built from a thermodynamic MHD simulation, the misalignment between extracted features and the model field is about 4–7 degrees, with roughly 80% of traced locations within 10 degrees. If that accuracy carries over to real observations, QRaFT gives modelers a usable empirical handle on the open-flux geometry that shapes the solar wind and the propagation of coronal mass ejections. The paper demonstrates the method on MHD-synthesized density and polarized-brightness images, a STEREO COR1 coronagraph image, and a 2017 total solar eclipse image.","feed_headline":"Open-flux coronal field traced to within 4–7 degrees","feed_subtitle":"Segmentation method QRaFT turns faint white-light corona images into magnetic field-line maps.","key_machinery":"The load-bearing object is the enhanced polar image $I_{\\rm enh}(\\varphi,\\rho) = I''/I''_{\\rm tr}$, the unsigned second-order azimuthal derivative of the smoothed, radially detrended image normalized by its large-scale azimuthal trend. In plane-polar coordinates a quasi-radial structure appears as an azimuthal intensity crest, and the unsigned second derivative produces sharp local maxima at the crest and its wings that adaptive thresholding can detect. The detector loops over percentile thresholds and inner radial boundaries, labels contiguous pixel clusters, and represents each cluster by a polynomial fit to its radial sequence of azimuthal centroids, yielding continuous traced features with local orientation angles $\\xi$ and a misalignment angle $\\theta$ relative to the outward model field.","core_discovery":"The central discovery is that the azimuthal structure of a radially detrended coronagraph image encodes the local orientation of the open coronal magnetic field. QRaFT deliberately sacrifices radial intensity gradients, which carry little field-geometry information, in favor of azimuthal gradients that mark narrow quasi-radial density structures. After remapping to polar coordinates, anisotropic smoothing, unsigned second-order azimuthal differencing, adaptive percentile thresholding, and a sliding inner radial boundary, the detected pixel clusters are interpolated into continuous features whose segment angles match the outward plane-of-sky magnetic field of the underlying thermodynamic MHD solution to within roughly 4–7 degrees. To the authors' knowledge, no other method traces open-field structures in routinely collected non-eclipse coronagraph images, and QRaFT is offered as a fill for that gap.","pith_inferences":["If the 4–7 degree alignment is confirmed with independent field measurements, misalignment maps could be used as a data-driven spatial score for coronal model skill, highlighting specific longitudes and heights where a model's field geometry is wrong.","Comparing QRaFT features against an emissivity-weighted, line-of-sight-integrated model field, rather than the plane-of-sky field at each node, would probably shrink the apparent misalignment in real coronagraph data and could be tested immediately on the paper's synthetic pB images.","Running QRaFT on a sequence of coronagraph images over a solar cycle could track the evolution of open-flux boundaries in white light and connect them to coronal-hole observations and solar wind stream structure.","Because QRaFT intentionally discards closed-loop features, pairing it with an active-region loop tracer would produce a combined observational connectivity map of the whole corona."],"forward_implications":["QRaFT output can be overlaid on model field-line maps as a direct visual and quantitative check of where a simulated corona's open-field geometry matches the observed white-light structure.","Misalignment statistics such as the COR1 example can be interpreted as data-derived measures of model boundary quality, since a boundary magnetogram 14 days old produces larger apparent errors.","The method's accuracy is highest in open-flux regions and degrades in closed-flux streamers, so practical use should focus on open-corona segmentation and treat closed-loop encounters as flagged outliers.","As space-borne coronagraphs approach eclipse-image quality, QRaFT-style tracing could be run systematically for operational solar wind and CME forecasting."],"supporting_citations":[{"why":"Supplies the thermodynamic MAS simulation whose density and magnetic field produce the synthetic images and the ground-truth geometry for the alignment test.","marker":"Mikić et al. 2018"},{"why":"The FORWARD code that converts the simulated three-dimensional density into a synthetic polarized-brightness image including line-of-sight Thomson scattering.","marker":"Gibson et al. 2016"},{"why":"The companion paper carrying the more systematic validation and error analysis of QRaFT, to which this paper defers for full statistics.","marker":"Rura et al. 2025"},{"why":"Earlier demonstration that coronagraph segmentation can constrain synoptic magnetograms, serving as the direct predecessor and motivation for QRaFT.","marker":"Jones et al. 2020"},{"why":"The publicly released QRaFT package whose algorithms and processing keywords this paper documents.","marker":"Uritsky 2024"},{"why":"Supplies the coronal flux-tube anisotropy premise that long-lived optical gradients are field-aligned.","marker":"Klimchuk 2006"}],"fun_headline_variants":["QRaFT maps coronal magnetic field from faint images","New method extracts magnetic field lines from coronagraph images","Coronal field orientation from azimuthal gradients alone","QRaFT traces open-flux field to 4–7 degrees","Faint coronagraph images reveal open magnetic field lines"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method's claimed accuracy rests on the premise that the simulated corona used to build the test images is geometrically similar enough to the real corona that the 4–7 degree alignment measured against the simulation's own magnetic field transfers to real coronagraph observations.","fun_headline_variants_meta":{"raw":{"variants":["QRaFT maps coronal magnetic field from faint images","New method extracts magnetic field lines from coronagraph images","Coronal field orientation from azimuthal gradients alone","QRaFT traces open-flux field to 4–7 degrees","Faint coronagraph images reveal open magnetic field lines"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000567,"raw_usage":{"total_tokens":2688,"prompt_tokens":947,"completion_tokens":1741,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":563,"completion_tokens_details":{"reasoning_tokens":1675}},"tokens_in":563,"tokens_out":1741,"duration_ms":12033,"temperature":1.0,"reasoning_tokens":1675,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:10:26.900413+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Apply QRaFT to synthetic white-light images from a structurally different, independently built coronal MHD model; if the median misalignment against that model's magnetic field clearly exceeds 7 degrees, the claimed accuracy is a property of the test model rather than a general property of the tracing method.","supporting_citations":[{"cited_title":"E., Uritsky, V","cited_arxiv_id":null,"evidence_quote":"The companion paper carrying the more systematic validation and error analysis of QRaFT, to which this paper defers for full statistics."},{"cited_title":"I., Uritsky, V","cited_arxiv_id":null,"evidence_quote":"Earlier demonstration that coronagraph segmentation can constrain synoptic magnetograms, serving as the direct predecessor and motivation for QRaFT."},{"cited_title":"M., & Klimchuk, J","cited_arxiv_id":null,"evidence_quote":"The publicly released QRaFT package whose algorithms and processing keywords this paper documents."}],"review_version":1}