{"id":"77fedf73-51c1-45cc-a294-759cf3f08dce","arxiv_id":"2507.17670","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"APRIL, an automatic registration method, aligns ground-based coronagraph green-line images with SDO/AIA 211 Å images to map them into Helioprojective Cartesian Coordinates with claimed precision better than 0.1 arcsecond.","lead":"This paper introduces APRIL, an automatic algorithm that aligns ground-based coronagraph images of the Sun's green corona with space-based EUV images, then maps them to a standard solar coordinate frame. It is validated on 100 days of data from two Chinese coronagraphs and claims sub-0.1 arcsecond registration precision.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 0.1 arcsec accuracy claim is a convergence tolerance and a repeatability statistic, not a measured absolute error; without injected-shift or ephemeris validation the headline is unsupported.","rationale":"The paper presents a plausible and well-documented registration pipeline, and the release of catalogs and registered FITS files is a concrete contribution. The central quantitative claim, however, is not supported by the measurements shown. The 0.1 arcsec value is exactly the convergence threshold of the iterative solver, and the Table 1 Xstd/Ystd/Rstd values are scatter over four RANSAC realizations, which measures repeatability rather than absolute accuracy. A successful convergence to zero in the internal median-distance metric would occur even if both images shared a common systematic offset relative to the true helioprojective coordinates. The absolute accuracy is entirely inherited from the SDO/AIA 211 reference frame and the observer geometry, and no independent check is provided. Section 4's explicit limitation that APRIL requires clear coronal structures and is not applicable during solar minimum is honestly stated and narrows the generality claim, but it is not the controlling issue. The reader's conditional verdict is appropriate: the registration method is likely sound as a relative alignment tool, but the absolute accuracy statement should be reworded or externally validated.","tokens_in":13533,"tokens_out":20899,"duration_ms":236779,"concrete_test":"Take at least 50 YOGIS images from 2024, apply known similarity transformations (translations 0.1-20 pixels, rotations +/-5 degrees, scale 0.97-1.03) to each, register the modified image against the same AIA 211 reference with APRIL, and compare recovered versus injected parameters. Report the RMS and median residual in translation (arcsec) and the equivalent displacement at 1.25 solar radii for rotation and scale errors. If the translation residual RMS exceeds 0.1 arcsec, or the rotation/scale residual exceeds 0.1 arcsec at 1.25 solar radii, then the Section 3.1 convergence cutoff does not support the accuracy claim; if it does, the claim still needs rewording as registration precision unless an independent absolute reference is added.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central accuracy claim is not established by the evidence. In Section 3.1 the iteration terminates when |dx| and |dy| are below 0.05 pixel, 'corresponding to approximately 0.1 arcsec'; this is a stopping tolerance, not a measured error. In Section 3.2 the reported Xstd, Ystd and Rstd are standard deviations across four RANSAC repetitions, i.e. repeatability, and the rank in Equation (10) is built from those same internal statistics. The only external check mentioned is that the recovered daily solar radius 'closely matches' the expected value, with no quantitative residual given. The abstract and conclusion therefore conflate convergence and precision with absolute accuracy. The absolute HCC mapping is inherited from the AIA 211 WCS and the assumed observer geometry; no star crossings, synthetic shifts, or ephemeris comparison are used to validate the final coordinates. The method may be a good relative registration tool, and the released catalogs and FITS files are useful, but the headline accuracy is under-supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents APRIL, an automatic registration algorithm that aligns ground-based coronal green-line images (from YOGIS and SICG) with SDO/AIA 211 Å images using local cross-correlation in polar coordinates, edge enhancement, and RANSAC, and then derives Helioprojective Cartesian Coordinates (HCC) from the inferred similarity transformation. The method is tested on 100 days of YOGIS data spanning 2013–2024, plus one day of SICG data, and the authors release catalogs, FITS files, and movies. The headline claim is a registration accuracy better than 0.1 arcsecond under optimal data quality and a precision no worse than 0.4 arcsecond in most cases.","tokens_in":13689,"tokens_out":3471,"duration_ms":35347,"significance":"If properly validated, the method addresses a real need: ground-based coronagraphs often lack reliable absolute pointing, and a robust registration against AIA 211 Å images provides a practical path to HCC coordinates for multi-instrument studies. The use of an external reference frame (the AIA WCS) is a strength, and the release of registration parameters and FITS files is a useful community resource. The explicit limitation that the method requires coronal structures and is not applicable near solar minimum is honest and appropriate. However, the central accuracy claim currently conflates an internal convergence threshold and RANSAC repeatability with absolute accuracy; the scientific significance depends on adding an independent absolute validation or clearly reframing the claim as internal precision.","major_comments":[{"comment":"The claim of 'accuracy no less than 0.1 arcsecond' is not supported by the reported evidence. Section 3.1 states that the iteration terminates when |Δx| < 0.05 pixel and |Δy| < 0.05 pixel, which is a stopping criterion, not a measured registration error against ground truth. The only external comparison mentioned is that the recovered daily solar radius 'closely matches' the expected value, but no quantitative residual is given. Please either provide an absolute validation (e.g., injected synthetic shifts, comparison with star positions, or an independent ephemeris/pointing solution) or revise the abstract and conclusion to state that 0.1 arcsecond is the convergence tolerance / internal precision rather than absolute accuracy.","section":"Section 3.1 and Abstract/Conclusion"},{"comment":"The metrics Xstd, Ystd, Rstd, and the rank in Eq. (10) are computed from standard deviations over only four RANSAC repetitions of the same image. These quantify repeatability of the stochastic solver, not the accuracy of the mapping relative to the true helioprojective coordinates. The statement that 'precision remained robust ... within 0.2 pixels (~0.4 arcsecond) and could achieve an accuracy better than 0.05 pixel (~0.1 arcsecond) under optimal conditions' therefore overstates external performance. Please relabel these as repeatability/precision metrics and add a separate accuracy assessment against known coordinates.","section":"Section 3.2, Table 1, Eq. (10)"},{"comment":"The absolute HCC mapping inherits the SDO/AIA 211 Å WCS and the assumption that the AIA pointing is an accurate absolute reference. Because no independent check (star crossings, simulated shifts, or comparison with solar ephemeris residuals) is performed, the absolute accuracy of the final HCC coordinates is not established. This is not a flaw in the registration concept, but it must be stated explicitly as an assumption and either tested or removed from the headline accuracy claims.","section":"Section 2.2 / Section 3.1, Eq. (9)"}],"minor_comments":[{"comment":"There is a typo in the sentence defining the aligned solar center: '(xg⊙, yg⊙) = (xe′⊙, xe′⊙)' should read '(xg⊙, yg⊙) = (xe′⊙, ye′⊙)'.","section":"Section 3.1, after Eq. (9)"},{"comment":"The table reports Xstd, Ystd, Rstd, and R⊙ without stating their units; the text later refers to pixels, but this should be stated in the table header or caption for clarity.","section":"Table 1"},{"comment":"The phrase 'with an accuracy of no less than 0.1 arcsecond' is ambiguous; 'accuracy of 0.1 arcsecond or better' would be clearer, and the claim should explicitly distinguish convergence tolerance from measured accuracy.","section":"Abstract"},{"comment":"The azimuthal width w is stated to have an optimal value of 150 degrees, but no sensitivity analysis or justification is provided; a brief plot or discussion of how the results vary with w would strengthen the method description.","section":"Section 2.2, Step 2"},{"comment":"The RANSAC-based standard deviations are computed from only four repetitions, which is a small sample for estimating stochastic variability; this should be noted as a limitation or the number of repetitions should be increased.","section":"Section 3.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is likely publishable after revision. The main risk is the unsupported absolute-accuracy claim; the authors should be asked to either validate against an independent reference or reframe the headline as internal precision. The data release and the breadth of testing are valuable, and the stress-test concern about the 0.1 arcsecond claim is fully borne out by Section 3.1."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is a genuinely useful contribution. It presents the first automatic registration of ground-based coronagraph green-line images to SDO/AIA 211 Å images, and the authors have put real work into validating it on 100 days of YOGIS data plus one day of SICG data spanning 11 years. The method itself is a sensible combination of known pieces — local cross-correlation, polar transforms, RANSAC, edge enhancement — and the edge-enhancement step is shown to matter for convergence. The release of catalogs and Level 0.5 FITS files is a plus; people studying coronal transients with these ground-based instruments will find it useful. The limitations (solar minimum, waveband restriction) are stated plainly, and the citation pattern looks appropriate.\n\nThe main soft spot is the accuracy claim. The abstract and conclusion say mapping accuracy is better than 0.1'', but that number is actually the iteration stopping tolerance — |Δx|, |Δy| < 0.05 pixel — and the reported Xstd/Ystd/Rstd statistics in Table 1 are repeatability across four RANSAC runs, not errors against an absolute reference. The rank metric in Equation (10) is built from those same internal statistics. The only external check mentioned is that the recovered daily solar radius 'closely matches' the expected value, but no quantitative residual is given. So the headline claim conflates convergence, precision, and accuracy. This is fixable: inject known shifts into real images, compare against star crossings if any exist in the FOV, or at least reword the claim to say 'repeatability better than 0.1'' under optimal conditions' and leave absolute accuracy as inherited from AIA pointing. The method is likely sound as a relative registration tool; the absolute HCC mapping depends on AIA's WCS being correct, which is a reasonable assumption but should be stated as such.\n\nMinor point: the convergence rate column in Table 1 includes cases below 100%, and the paper doesn't discuss what happens on those failures in the released data — are those images dropped from the catalogs? That should be clarified.\n\nOverall, this is solid work worth refereeing, but the accuracy language needs a serious revision pass. A referee should request either injected-shift validation or a careful rewording. I'd bring it to a reading group, and I'd cite it if I worked on ground-based coronagraph data.\n\nRecommendation: send to peer review, with the accuracy claim as the main point to fix.","headline":"Useful enabling tool with honest limitations, but the headline 0.1'' accuracy claim outruns what is actually measured.","tokens_in":14324,"tokens_out":1302,"would_cite":true,"duration_ms":16024,"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 algorithm registers ground-based coronal images to Helioprojective Cartesian coordinates using 211 Å EUV reference images, reaching sub-0.1-arcsecond convergence on good data.","keywords":["solar corona","coronagraph","image registration","helioprojective coordinates","green-line emission","SDO/AIA 211 Å","RANSAC","cross-correlation"],"falsifier":"Take a space-based 211 Å image, apply a known similarity transformation (rotation, scale, translation), hide the disk behind an occulter-sized mask, add noise at the level of a typical ground-based coronagraph, and run APRIL; if the recovered parameters differ from the injected ones by more than about 0.4 arcseconds on good-quality frames, the precision claim is falsified.","tokens_in":13306,"feed_emoji":"☀️","tokens_out":6726,"duration_ms":65593,"temperature":0.7,"pith_summary":"The paper presents APRIL, an automatic registration method that aligns ground-based coronal images taken in the Fe XIV 5303 Å green line to space-based EUV images at 211 Å, then uses the known pointing of the EUV images to assign Helioprojective Cartesian coordinates to every green-line pixel. The method matters because ground-based coronagraphs hide the solar disk behind an occulter, so neither the solar center nor the radius is directly visible and small pointing drifts are otherwise hard to correct. By matching coronal structures in overlapping annular subregions, the method recovers translation, rotation, and scale, and the paper reports that converged registrations reach roughly 0.1 arcseconds under good conditions and stay within about 0.4 arcseconds in most of the 100 test days spanning an 11-year period. The result would let decades of existing green-line coronagraph data be co-analyzed with EUV observations and used for studies of transient coronal activity.","feed_headline":"Ground-based coronal images mapped to solar coordinates to 0.1 arcsec","feed_subtitle":"New algorithm aligns green-line and EUV images so ground-based coronagraphs yield high-precision solar positions.","key_machinery":"The load-bearing mechanism is an iterated similarity-transform registration built from local phase information. Each iteration transforms the EUV and green-line images to polar coordinates around the current estimate of the solar center, keeps only the annulus from about 1.04 to 1.25 solar radii, and applies the Scharr edge filter to turn diffuse coronal structure into sharper features. Thirty-six overlapping azimuthal windows, each 150 degrees wide, are cross-correlated in the Fourier domain; a localized upsampled discrete Fourier transform locates each correlation peak to subpixel accuracy. The resulting point pairs feed a least-squares RANSAC solver for translation, rotation, and scale, and the EUV image is resampled and the process repeated until the parameter updates fall below 0.05 pixel. The edge-enhancement step is what keeps the correlation peaks from drifting, and the iteration is what absorbs the initial error in assuming the occulter center coincides with the solar center.","core_discovery":"The central claim is that the low-contrast, occulter-obscured green-line corona can be registered to 211 Å EUV images accurately enough to recover the green-line image's solar center, solar radius, and polar-axis angle, and therefore to map it into Helioprojective Cartesian coordinates. The registrations are obtained automatically: after a coarse alignment using the occulter geometry, the images are converted to polar coordinates over the radial band from 1.04 to 1.25 solar radii, edge-enhanced with the Scharr operator, divided into 36 overlapping azimuthal subregions, and matched by Fourier cross-correlation with subpixel refinement. A RANSAC fit of a similarity transform rejects bad matches, and the whole procedure is iterated until the translation and rotation updates converge to zero. The paper reports that on good data the iteration converges to sub-0.1-arcsecond alignment and that repeated runs on most days show scatter below 0.2 pixels, about 0.4 arcseconds, with the recovered solar radius following its expected annual variation. On this basis the method is offered as a general tool for coronagraphs that observe the green line.","pith_inferences":["Because the absolute reference is the EUV image pointing, the true absolute accuracy is capped by the accuracy of that instrument's pointing solution; the 0.1 arcsecond figure measures convergence, not an independent absolute check.","The method should be testable on synthetic data: warp a 211 Å image by known rotation, scale, and translation, inject realistic noise and an occulter mask, and see whether APRIL recovers the injected parameters to the claimed tolerance.","The same subregion-cross-correlation machinery could plausibly register other coronal emission lines to one another once a stable structural correspondence is established, extending the approach beyond green-line data.","The requirement of visible coronal structure means the method is season-limited; a practical pipeline would need to flag low-contrast epochs rather than silently failing."],"forward_implications":["Existing multi-year green-line coronagraph archives can be re-reduced into Helioprojective Cartesian coordinates without needing a visible solar limb.","Base-difference movies built from APRIL-registered frames show only real coronal changes, not apparent shifts from pointing drift.","The drift of the solar center relative to the occulter, tracked over a day, gives a quantitative record of guiding-system error that can be used to correct pointing.","Coronagraphs without any guiding system, including small or balloon-borne instruments, can still produce absolutely positioned data if their images contain green-line structure.","Registrations are computed per frame, so fast, short-timescale transients can be studied without first averaging frames."],"supporting_citations":[{"why":"Establishes the 0.89–0.99 correlation between green-line and 211 Å intensities that justifies using EUV images as the registration reference.","marker":"X. Zhang et al. (2022)"},{"why":"Supplies the local-statistical-correlation-plus-feature-matching strategy that APRIL adapts from solar-disk images to coronal images.","marker":"T. Feng et al. (2018)"},{"why":"Provides the upsampled discrete Fourier transform used to locate cross-correlation peaks to subpixel precision.","marker":"M. Guizar-Sicairos et al. (2008)"},{"why":"Describes the SDO/AIA instrument and its 211 Å channel, the source of the reference images.","marker":"J. R. Lemen et al. (2012)"},{"why":"Describes the 10 cm green-line coronagraph design on which the YOGIS observations are based.","marker":"K. Ichimoto et al. (1999)"},{"why":"Earlier phase-correlation registration of eclipse coronal images that required manual parameter adjustment, the limitation APRIL removes.","marker":"M. Druckmüller (2009)"},{"why":"Demonstrates cross-correlation registration of eclipse corona without rotation and scaling, the gap APRIL fills.","marker":"Y. Liang et al. (2021)"}],"fun_headline_variants":["Auto-align ground corona to EUV at 0.1 arcsec","Solar coronal mapping hits sub-arcsecond precision","Green-line corona registered to EUV with 0.1'' accuracy","Coronal image registration yields 0.1 arcsec positions","Precise helioprojective mapping from ground-based corona"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method inherits its absolute pointing from the EUV reference images: if that instrument's pointing is off, every derived solar-center coordinate is off by the same amount, and the paper does not check the derived coordinates against an independent reference such as star positions or known simulated shifts.","fun_headline_variants_meta":{"raw":{"variants":["Auto-align ground corona to EUV at 0.1 arcsec","Solar coronal mapping hits sub-arcsecond precision","Green-line corona registered to EUV with 0.1'' accuracy","Coronal image registration yields 0.1 arcsec positions","Precise helioprojective mapping from ground-based corona"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000188,"raw_usage":{"total_tokens":1358,"prompt_tokens":1000,"completion_tokens":358,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":616,"completion_tokens_details":{"reasoning_tokens":270}},"tokens_in":616,"tokens_out":358,"duration_ms":4315,"temperature":1.0,"reasoning_tokens":270,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:18:50.370049+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a space-based 211 Å image, apply a known similarity transformation (rotation, scale, translation), hide the disk behind an occulter-sized mask, add noise at the level of a typical ground-based coronagraph, and run APRIL; if the recovered parameters differ from the injected ones by more than about 0.4 arcseconds on good-quality frames, the precision claim is falsified.","supporting_citations":[{"cited_title":"2018, title High-accuracy Registration Method of Solar High-resolution Observation Images and Full-disk Solar Images, AR&T, 15, 69","cited_arxiv_id":null,"evidence_quote":"Supplies the local-statistical-correlation-plus-feature-matching strategy that APRIL adapts from solar-disk images to coronal images."}],"review_version":1}