{"id":"1a172372-7de1-4c2f-902e-f1c9cc679391","arxiv_id":"1908.05493","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Applying the same processing to both Kodaikanal Ca II K digitizations shows the newer 16-bit series is higher quality and should be preferred, while most quality degradation is intrinsic to the original plates.","lead":"This paper compares two digital scans of the same century-long archive of calcium-K solar images from Kodaikanal Observatory and identifies which scan is more reliable. The newer 16-bit scan has better image quality and far more images, and most data problems come from the original photographs, not the scanning.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Intrinsic-origin conclusion rests on the untested absence of a shared processing-pipeline bias; applying two independent analysis chains to both digitizations would settle it.","rationale":"The paper is a careful, well-executed intercomparison: it applies a consistent pipeline, documents limitations, and provides the DS2 series. The main recommendation (prefer DS2) is robust to the concern, because DS2's larger size, 16-bit depth, and better post-1990 resolution stand on their own. The load-bearing assumption is not the recommendation but the strong interpretive claim that common quality trends are intrinsic to the original observations. The independent scanner hardware makes shared digitisation artefacts unlikely, but the shared analysis pipeline is a genuine common-cause channel that has not been falsified. The proposed test is feasible with existing data and would discriminate between pipeline bias and image-intrinsic trends. Until such a test is run, the intrinsic-origin conclusion should be read as a well-supported inference rather than a proven fact; this does not change the acceptance of the paper.","tokens_in":18855,"tokens_out":11556,"duration_ms":117716,"concrete_test":"Run an independent processing chain (different limb fit, different resolution metric, different background/CLV estimator) on a stratified sample of ~100 DS1 and DS2 images spanning 1904–2007, and compare the derived secular trends with those reported in Section 3.2. If the trends persist under both chains and both digitizations, the intrinsic-origin claim is confirmed; if they weaken or change sign under the second chain, the original pipeline is the common cause and the conclusion in Section 5 needs qualification.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.2 and Section 5 conclude that the time-dependent trends common to DS1 and DS2 (disc eccentricity, spatial resolution, large-scale inhomogeneities, CLV fit chi-squared) are intrinsic to the original photographic observations and not to the digitisation. The two scanning campaigns are indeed independent in hardware, bit depth, and timing, which makes a shared scanner artefact unlikely. However, the two datasets were processed in this study with one and the same pipeline: the same limb-detection routine feeds the eccentricity and radius estimates, the same 98%-power-spectral-density estimator yields the resolution, and the same iterative background/CLV fitting produces the inhomogeneity and chi-squared metrics. A systematic bias in any of these routines that scales with plate scale, solar radius, dynamic range, or noise level would be impressed on both series and could mimic an intrinsic secular degradation. The Rome/PSPT control covers only 1996–2007, so it cannot validate the 1904–1995 portion of the trends. Because the 'intrinsic to the original observations' interpretation underlies the paper's claim that the archive's quality problems are inherent, this shared-pipeline channel is the single most load-bearing unexcluded alternative.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper compares two digitizations of the Kodaikanal Ca II K spectroheliogram archive: DS1 (22,158 mostly 8-bit JPG images, 1907--1999) and DS2 (48,928 16-bit FITS images, 1904--2007). The authors apply the same photometric calibration, limb-darkening compensation, and plage-segmentation pipeline to both series, and also to Rome/PSPT data as a modern reference. They quantify disc eccentricity, spatial resolution, large-scale inhomogeneities, CLV-fit chi-squared, and density/contrast statistics, and derive plage areas. The main findings are: (i) both series show similar time-dependent degradation (increasing eccentricity, worsening resolution, growing inhomogeneities, changing CLV), which the authors attribute to intrinsic properties of the original photographic material; (ii) DS2 has better and more consistent image quality and roughly twice as many images; (iii) plage areas from DS1 and DS2 are highly correlated (Pearson 0.95 daily) but differ by RMS ~0.006 of disc area, with cycle-dependent offsets; and (iv) differences among published plage-area series are largely traceable to different processing choices and solar-disc definitions. The authors conclude that DS2 should be preferred.","tokens_in":19056,"tokens_out":10935,"duration_ms":104995,"significance":"If the conclusions hold, this is a valuable reference for users of the Kodaikanal archive and helps explain the discrepancies among previously published plage-area series. The consistent application of one pipeline to both digitizations is a real strength, as it removes processing differences as an explanation for DS1--DS2 differences, and the image-quality metrics are quantified with uncertainties. The explicit statement that the plage-area errors are tentative, not formal, is appropriately cautious. The external Rome/PSPT comparison provides a useful modern anchor. The practical recommendation that DS2 be preferred is well supported by its larger sample, better post-1990 resolution, and more stable density statistics. The main weakness is that the causal attribution of the common time-dependent trends to the original plates, rather than to the shared processing pipeline, is asserted more strongly than the evidence supports.","major_comments":[{"comment":"The conclusion that the time-dependent trends common to DS1 and DS2 (disc eccentricity, spatial resolution, large-scale inhomogeneities, CLV fit chi-squared) are intrinsic to the original observations is not fully supported, because both series are processed with the same analysis pipeline. The same limb-detection routine feeds the eccentricity and radius estimates, the same 98%-power-spectral-density estimator yields the resolution, and the same iterative background/CLV fitting produces the inhomogeneity and chi-squared metrics. A systematic bias in any of these routines that scales with plate scale, solar radius, dynamic range, or noise level could be impressed on both series and could mimic intrinsic secular degradation. The Rome/PSPT control covers only 1996--2007, so it cannot validate the 1904--1995 portion of the trends. Please either process both series with a second, independent analysis chain and check that the common trends persist, or add sensitivity tests of the metrics to pipeline choices (e.g., limb-detection threshold, PSD window/cutoff, background polynomial order) and soften the causal wording in Sections 3.2 and 5 from \"should be ascribed to\" to \"are consistent with\" the original photographic observations.","section":"Section 3.2 / Section 5"},{"comment":"The DS2--Rome/PSPT overlap comparison, which is the main external validation of the DS2 plage areas, shows an RMS difference of 0.01 in disc fraction and a pronounced seasonal pattern (Rome/PSPT higher in winter, DS2 higher in summer), with a maximum absolute difference of 0.08. The paper reports this pattern but does not analyze its origin. Since a seasonal systematic in DS2 would affect the reliability of the recommended series, please investigate whether this modulation reflects a calibration effect (e.g., CLV or flat-field residuals) and explicitly discuss its impact on the claim of a \"good match\" between DS2 and Rome/PSPT.","section":"Section 4.1 / Figure 13"}],"minor_comments":[{"comment":"The sentence \"We find only 13,835 images of each set referring to the same solar observations\" is ambiguous; please rephrase to clarify that these are paired scans of the same plates, one image from each digitization.","section":"Section 2.1"},{"comment":"The phrase \"normalised to the maximum value from the respective digitisation\" is ambiguous; state whether the normalisation is to the maximum over all images of each series or to a per-image maximum.","section":"Section 3.2 / Figure 9"},{"comment":"The sentence \"The maximum (RMS) error in the plage areas for the average disc eccentricity found for Kodaikanal data is 0.013 (0.0005)\" is hard to parse; please spell out the maximum error and the RMS error separately.","section":"Section 3.2"},{"comment":"The sentence \"The multiplicative factor for identifying plage was chosen to be 8.5\" should explicitly state that this factor multiplies the standard deviation of the quiet-Sun intensity values, as implied later in the text.","section":"Section 2.2"},{"comment":"There is a typo in \"Priyalet al. (2017)\"; the word \"al.\" should be separated from \"Priyal\".","section":"Section 4.2"}],"recommendation":"major_revision","confidential_remarks":"This is a solid and useful data-characterization paper with no fatal flaw. The main technical gap is the shared-pipeline alternative to the \"intrinsic origin\" conclusion; if the authors add a sensitivity analysis or a second independent processing chain, or soften the claim, the paper should be acceptable. The DS2-versus-PSPT seasonal pattern also deserves a more careful discussion. Both points are within the scope of a revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a solid, practical paper. The genuinely new thing is that both digitizations of the Kodaikanal Ca II K archive are run through the same photometric calibration and segmentation pipeline, and compared quantitatively with modern Rome/PSPT data. That allows the authors to separate digitization effects from processing effects, and it explains a good deal of the scatter among published plage-area series. The recommendation to prefer the 16-bit DS2 series is well supported: it has more images, more stable transmittance, better resolution after 1990, and the plage areas agree with Rome/PSPT at least as well as DS1.\n\nThe paper is also honest about its limits. The plage-area error bars are explicitly labeled as tentative, the fixed segmentation threshold is argued to be immaterial for the relative comparison, and the metadata timing problems are documented as affecting both series.\n\nThe main soft spot is the conclusion in Section 5 that common time-dependent trends — worsening resolution, increasing eccentricity, growing inhomogeneities, changing CLV — are intrinsic to the original plates. The two digitizations were made with different hardware and a decade apart, so a shared scanner artifact is unlikely. But both series were processed with the same software: same limb detection, same PSD estimator, same background/CLV fitting. If any of those routines has a systematic bias that scales with plate scale, dynamic range, or noise, it would be impressed on both series and could mimic intrinsic degradation. The Rome/PSPT control only covers 1996–2007, so it cannot validate the earlier secular trends. This is not a fatal flaw — the main recommendation to prefer DS2 stands regardless — but the intrinsic-origin statement is an inference from two realizations of the same archive through one pipeline, not a proven fact. It would be more accurate to say 'consistent with an intrinsic origin' or to test with an independent processing chain on one series.\n\nFor the target audience — people building plage-area time series and irradiance reconstructions from historical Ca II K data — this is an important reference. It deserves a serious referee. I would accept it after minor revision, asking for a softer or better-guarded wording of the intrinsic-origin conclusion, and ideally a sensitivity test of the pipeline on one series. Otherwise, the analysis is careful and the data handling is transparent, so I expect it will be widely used.","headline":"A careful and useful comparison of the two Kodaikanal digitizations; the recommendation to prefer the 16-bit series is well supported, but the 'intrinsic artifact' conclusion is somewhat stronger than the evidence.","tokens_in":19672,"tokens_out":2791,"would_cite":true,"duration_ms":24958,"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":"The paper establishes that the newer 16-bit digitisation of the Kodaikanal Ca II K archive is superior, while the shared quality degradation in both scans is intrinsic to the original plates.","keywords":["Ca II K observations","Kodaikanal archive","plage areas","archive digitisation","image quality","solar cycle","instrumental effects"],"falsifier":"Scan a stratified sample of the same Kodaikanal plates (for example, 20 plates per decade from 1904–2007) with a modern calibrated scanner and recompute the same metrics—disc eccentricity, spatial resolution, large-scale inhomogeneities, centre-to-limb variation residuals, and plage areas. If the secular trends shrink or vanish, the shared trends in DS1 and DS2 were partly scanner artefacts; if they reproduce at the same slope, the archive itself is degrading.","tokens_in":18675,"feed_emoji":"🌞","tokens_out":6433,"duration_ms":56930,"temperature":0.7,"pith_summary":"This paper asks whether the two recent digitisations of the century-long Kodaikanal Ca II K photographic archive differ because of the scanning or because of the original plates. The authors process both datasets with identical photometric calibration and plage-segmentation software, then compare image-quality metrics and derived plage areas. They find that the slow degradation seen in the older 8-bit scan—worsening spatial resolution, increasing disc eccentricity, growing large-scale inhomogeneities, and changing centre-to-limb variation—also appears in the newer 16-bit scan, so it is intrinsic to the archive. The 16-bit series has more consistent photometry, roughly twice as many images, and higher resolution after 1990, making it the dataset to prefer.","feed_headline":"Kodaikanal's new scan wins: double the data, same plate flaws","feed_subtitle":"The 16-bit series has more images and steadier photometry; its quality drifts are inherited from the original plates.","key_machinery":"The load-bearing instrument is a single processing pipeline applied unchanged to both datasets. Images are converted to density, photometrically calibrated by fitting the quiet-Sun centre-to-limb variation to a modern reference, corrected for limb darkening, and segmented into plage using a threshold of 8.5 times the quiet-Sun contrast standard deviation. Because the pipeline is held fixed, any systematic difference between DS1 and DS2 must be attributed either to the digitisation or to the photographic material, and the matching secular drifts in eccentricity, resolution, inhomogeneities, and centre-to-limb variation residuals point to the latter.","core_discovery":"The central claim is that the two digitisations of the same physical archive carry the same time-dependent quality trends, and therefore those trends belong to the original photographic observations rather than to either scanner. Plage areas derived from DS1 and DS2 with identical techniques agree to an RMS difference of about 0.006 in disc fraction on daily values (0.005 for observations taken on the same day and time), with Pearson coefficients near 0.95–0.98, and both sets track the modern Rome/PSPT coverage in the overlap period, with Kodaikanal slightly lower. On this evidence DS2 supersedes DS1: it provides more than double the images (48,928 versus 22,158), more stable transmittance, and better spatial resolution after 1990, while its remaining imperfections are properties of the plates themselves.","pith_inferences":["A third digitisation of the same plates with modern equipment may not add much scientific value beyond DS2 for most purposes, because the quality bottleneck appears to be the photographic material itself; a testable extension would be scanning a stratified sample of plates to check whether any remaining artefacts disappear.","The common time-dependent trends could in principle be modelled as a quality-correction function and applied to plage areas, but that is an extension beyond what this paper attempts.","If plate degradation is intrinsic, then other photographic solar archives of similar age may hide comparable latent trends, so cross-archive composites may need to include plate-age or observational-epoch covariates.","The suggestion that Kodaikanal's effective bandwidth may be broader than the nominal 0.5 Å, or that the line was observed off-centre, is a testable inference that could be checked by examining the shape of the centre-to-limb variation over time."],"forward_implications":["The 16-bit DS2 series should be the default dataset for future Kodaikanal plage-area studies, with DS1 retained mainly for years where it has better coverage, such as 1907, 1909, 1911, 1913, 1955, 1973, and 1993.","Long-term activity reconstructions from Kodaikanal data must account for time-dependent image quality, since disc eccentricity, spatial resolution, and large-scale inhomogeneities drift secularly in both digitisations.","Errors in the date and time metadata affect both series, so daily-mean analyses or careful metadata cleaning are needed before comparing Kodaikanal data to other archives.","The scatter among published Kodaikanal plage-area series stems mainly from different calibration methods and different definitions of the solar radius and normalisation area, not from the digitisation itself.","The archive's intrinsic degradation limits the precision of century-long plage-area trends and makes cross-calibration with other Ca II K archives necessary."],"supporting_citations":[{"why":"Supplies the DS1 dataset, the older 8-bit scan of 22,158 plates from 1907–1999 whose quality is compared against.","marker":"Makarov et al. (2004)"},{"why":"Supplies the DS2 dataset, the newer 16-bit CCD scan of 48,928 plates from 1904–2007 that this paper recommends.","marker":"Priyal et al. (2014)"},{"why":"Provides the photometric calibration and limb-darkening-compensation method applied identically to DS1 and DS2.","marker":"Chatzistergos et al. (2018a)"},{"why":"Gives the plage-area processing, error estimates, and the earlier DS1-based series that anchor the comparisons.","marker":"Chatzistergos et al. (2019a)"},{"why":"Supplies prior quality metrics and a DS1-based plage series whose values are recomputed and compared here.","marker":"Ermolli et al. (2009a)"},{"why":"One of the published DS2-based plage series used to test how much of the scatter comes from processing choices rather than the data.","marker":"Chatterjee, Banerjee, and Ravindra (2016)"},{"why":"Provides the homogeneous sunspot-area series used as an independent reference for judging the relative heights of solar cycles.","marker":"Balmaceda et al. (2009)"}],"fun_headline_variants":["New Kodaikanal scan doubles data, inherits old plate flaws","16-bit Kodaikanal archive outshines 8-bit, but flaws persist","Plate flaws, not scanners, explain Kodaikanal archive gaps","Kodaikanal's best digitization yet: more images, same plate issues","Superior 16-bit scan: more Kodaikanal images, same inherited drifts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion depends on the assumption that the two digitisation campaigns, carried out more than a decade apart with different hardware, do not share some systematic artefact that happens to produce the same slow drift in image quality; only then does the matching degradation point to the original plates.","fun_headline_variants_meta":{"raw":{"variants":["New Kodaikanal scan doubles data, inherits old plate flaws","16-bit Kodaikanal archive outshines 8-bit, but flaws persist","Plate flaws, not scanners, explain Kodaikanal archive gaps","Kodaikanal's best digitization yet: more images, same plate issues","Superior 16-bit scan: more Kodaikanal images, same inherited drifts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000446,"raw_usage":{"total_tokens":2295,"prompt_tokens":1029,"completion_tokens":1266,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":645,"completion_tokens_details":{"reasoning_tokens":1162}},"tokens_in":645,"tokens_out":1266,"duration_ms":9678,"temperature":1.0,"reasoning_tokens":1162,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:11:59.349105+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Scan a stratified sample of the same Kodaikanal plates (for example, 20 plates per decade from 1904–2007) with a modern calibrated scanner and recompute the same metrics—disc eccentricity, spatial resolution, large-scale inhomogeneities, centre-to-limb variation residuals, and plage areas. If the secular trends shrink or vanish, the shared trends in DS1 and DS2 were partly scanner artefacts; if they reproduce at the same slope, the archive itself is degrading.","supporting_citations":[{"cited_title":"Journal of Geophysical Research: Space Physics 114(A7)","cited_arxiv_id":null,"evidence_quote":"Provides the homogeneous sunspot-area series used as an independent reference for judging the relative heights of solar cycles."}],"review_version":1}