{"id":"7dcf9d64-ffe8-4190-93f6-c91f6a5ee646","arxiv_id":"2607.23541","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Area-weighted global-mean SAT change exceeds a histogram-shift “global-scale” warming metric by 0.19–0.23°C (2000–2019 vs 1979) and 0.27–0.54°C by 2080–2099, mainly from Arctic amplification.","lead":"The paper finds that the usual global-mean temperature rise exceeds a histogram-shift measure of warming by about 0.2°C recently and up to ~0.5°C by 2100, mostly from Arctic amplification. That gap is used to argue the mean overstates “global-scale” warming and is a risky sole policy metric.","discovery_kind":"new_application","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The residual ΔTmean − ΔTgs is labeled \"net regional-scale warming\" and quantitatively decomposed (71.8% / 28.2%), but histogram location statistics perform no spatial-scale separation — the residual can be produced entirely by smooth, hemispheric-scale patterns such as Arctic amplification itself.","rationale":"The reader identified the right neighborhood — the unjustified elevation of ΔTgs to \"the\" global warming rate — but located the weakness primarily in the normative framing (\"if the energy-relevant rate remains the area-weighted mean, the overestimate framing fails\"). I agree with that, but I think the sharper, more technically attackable point is one step further into the argument: the interpretive decomposition itself. The paper does not merely claim ΔTmean > ΔTgs; it assigns the residual a physical identity (\"net regional-scale warming effect\") and quantitative percentages (71.8%/28.2%, 78.9%/21.1%), and these percentages carry the policy-relevant weight of §4 (the 27%-of-2°C comparison). That decomposition is the load-bearing structure connecting the measurement to the title, and it rests on a category error: histogram central-tendency statistics cannot partition warming by spatial scale, and the paper's own evidence (Arctic amplification dominance) points to a large-scale driver.\n\nThis is a correctness concern about interpretation, not about the reported numbers, which are internally consistent across four reanalyses and 20 CMIP6 models and robustly signed. The empirical quantification (mean exceeds median/mode/shift by 0.19–0.23°C observed, 0.27–0.54°C projected; Arctic exclusion collapses the gap) would survive my proposed test regardless of outcome. What the test settles is whether the residual deserves the label \"regional-scale\" and hence whether ΔTgs deserves the label \"global-scale.\" If the low-pass-filtered field reproduces much of the residual, the honest framing is \"mean exceeds typical-area/modal warming due to positively skewed change, dominated by hemispheric polar amplification\" — close to what the reader's CONDITIONAL already demands. Hence UNCHANGED: my concern reinforces and sharpens the reader's condition rather than overturning the verdict; the fix is reframing and dropping the scale-decomposition language, not new data.","tokens_in":15655,"tokens_out":2792,"duration_ms":48029,"concrete_test":"Take the ERA5 2000–2019 minus 1980–1999 SAT-change field and apply a spherical-harmonic low-pass filter (e.g., truncate at total wavenumber n ≤ 5, keeping only hemispheric/planetary scales; discard everything smaller). Compute ΔTmean and the histogram shift ΔTgs on this purely large-scale field using the paper's exact §2.2 procedure (0.01°C bins, 1979-base cross-correlation). If mean − shift remains a substantial fraction of the original 0.11°C residual for a field containing no regional-scale structure by construction, the \"net regional-scale warming\" attribution and the 71.8%/28.2% decomposition are artifacts of skewness arithmetic, not scale physics. As a control, verify that a rigid uniform shift of the same filtered field yields mean = shift, confirming the method itself is not the source.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central interpretive move goes beyond the (sound) numerical finding that ΔTmean > ΔTgs > median/mode. In §3.1 the authors state the residual \"implies that there is a net positive warming effect of regional-scale temperature change,\" and decompose ΔTmean into 71.8% global-scale and 28.2% regional-scale contributions (repeated as 78.9%/21.1% for trends in §3.2, and invoked for the Paris-target framing in §4, where 0.54°C is \"as high as 27%\" of the 2°C target).\n\nThe problem: the histogram shift (ΔTgs) and the arithmetic mean are both scale-blind statistics of the same field. Their difference measures skewness of the SAT-change distribution — mean minus a robust location estimate — not warming at any particular spatial scale. Arctic amplification, which the paper itself identifies as the dominant source of the residual (Fig. 2c–d: excluding 65°N–90°N collapses the gap to 0.01–0.02°C/decade), is not a \"regional-scale\" phenomenon in the decomposition sense; it is a smooth, hemispheric-scale, physically coupled pattern (sea-ice albedo feedback, poleward heat transport) that is part of the large-scale forced response. If a purely large-scale, spatially smooth warming pattern can generate mean − shift ≠ 0, then attributing the entire residual to \"regional-scale warming\" and assigning percentage contributions has no physical basis — the same arithmetic would follow. This does not make the measured numbers wrong, but it undercuts the load-bearing interpretive claim (and the title's \"overestimates global warming rate\") that ΔTgs isolates global-scale warming and the remainder is a distinct regional component. A genuine scale separation requires spatial filtering of the field, not central-tendency statistics of its histogram.","agreement_with_reader":"partial"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The manuscript argues that the conventional global-mean surface air temperature change (ΔTmean) systematically exceeds a proposed \"global-scale\" warming measure (ΔTgs), defined as the horizontal shift of the global SAT image histogram that maximizes cross-correlation with the 1979 histogram. Using four reanalyses (ERA5, JRA55, NCEP1, NCEP2) and 20 CMIP6 models under three SSPs, the authors show that the SAT-change distribution is positively skewed, so the arithmetic mean exceeds the median, mode, and histogram shift: by 0.19–0.23°C for 2000–2019 relative to 1979 in reanalyses, and by 0.27–0.54°C for 2080–2099 in projections. Excluding the Arctic (65°N–90°N) collapses the trend differences to 0.01–0.02°C/decade, identifying Arctic amplification as the dominant source of the skew. The residual ΔTmean − ΔTgs is interpreted as a \"net regional-scale warming effect\" and decomposed into percentage contributions (71.8%/28.2% for the two-decade change; 78.9%/21.1% for trends), with the SSP5-8.5 residual (0.54°C) described as \"as high as 27%\" of the 2°C Paris target.","tokens_in":16095,"tokens_out":3434,"duration_ms":153677,"significance":"If the numerical results hold — and they appear to, given four independent reanalyses and a 20-model CMIP6 ensemble producing consistent mean-vs-shift gaps (0.19–0.23°C for 2000–2019; 0.27–0.54°C by 2080–2099), plus a clean Arctic-exclusion test (Fig. 2c–d) — the paper documents a robust and underappreciated property of spatial warming distributions: the area-weighted mean SAT change systematically exceeds the warming experienced over the majority of the Earth's surface, by an amount that grows with emissions. The methods are transparent and the numbers are in principle reproducible from public data. However, the paper's headline significance as framed (\"ΔTmean overestimates the global warming rate,\" with implications for the Paris targets) rests on an interpretive claim — that mean minus histogram-shift isolates \"regional-scale warming\" — that the diagnostics cannot support, and this currently overstates what the analysis delivers.","major_comments":[{"comment":"The central interpretive move — that ΔTmean − ΔTgs represents \"net positive regional-scale warming\" — is not supported by the statistics used. Both ΔTmean (area-weighted mean) and ΔTgs (maximum-cross-correlation histogram shift) are scale-blind location statistics of the same global field; their difference measures the skewness of the SAT-change distribution, not warming at any particular spatial scale. A perfectly smooth, hemispheric-scale pattern such as Arctic amplification (which the paper itself identifies as the dominant contributor, Fig. 2c–d, where excluding 65°N–90°N collapses the trend gap to 0.01–0.02°C/decade) will produce mean − shift ≠ 0 exactly as observed. The decomposition of ΔTmean into 71.8% 'global-scale' and 28.2% 'regional-scale' contributions (repeated as 78.9%/21.1% for trends in §3.2) therefore has no physical basis as stated. The numerical finding (mean > ΔTgs >","section":"§3.1 (and repeated in §3.2, §4)"},{"comment":"The definition of ΔTgs as 'global-scale SAT change' — the horizontal shift of the SAT histogram maximizing cross-correlation with the 1979 histogram — is presented as the physically appropriate measure of 'the global warming rate,' against which ΔTmean is an 'overestimate.' This is asserted, not justified. The area-weighted arithmetic mean is not an estimator that 'assumes symmetry' (as the abstract and §1 claim); it is the area integral of the temperature anomaly and is the quantity that enters radiative feedbacks, ocean heat uptake, and the GMST metric used in assessments. The histogram-shift statistic is instead a robust, mode-like location measure: it tracks the warming at the most common SAT values, which is a defensible diagnostic but not obviously 'the' global warming rate. The title and abstract claim ('overestimates global warming rate') is therefore definitional rather than dem","section":"§2.2, Abstract, and Title"},{"comment":"The statement that the SSP5-8.5 ΔTmean − ΔTgs difference (0.54°C) is 'as high as 27% compared to the 2°C target of the Paris Agreement' is a category error with a baseline mismatch. The 2°C target is defined on ΔTmean relative to preindustrial; the 0.54°C is the difference between two location statistics of the same field relative to a 1979 base year. The ratio of a within-field statistical difference to an absolute policy threshold has no decision-relevant meaning as presented, and the 1979 baseline makes the numerical comparison additionally incommensurate. This passage should be removed or recast (e.g., as the residual's size relative to projected twenty-first-century warming within the same baseline).","section":"§4 (Discussion) and §3.3, referencing Table 1"},{"comment":"ΔTgs is the paper's central constructed quantity, yet no sensitivity analysis is provided for its free choices: the 0.01°C bin width (with the acknowledged empty-bin problem at fine resolution, §2.2 step 3), the 1979 base year, the smoothing applied to histograms, and the effect of genuine shape change over time (Fig. S3–S4 show skewness and variance trends, so the near-rigid-shift assumption is only approximate — under SSP5-8.5 the shape change is large, precisely where the largest residuals are reported). The single sentence that RMS-difference matching 'gives the same result' is reassuring but insufficient. Table 1's headline numbers (0.19–0.23°C; 0.27–0.54°C) should be shown to be stable under reasonable perturbations of bin width, base period, and shift-estimation method.","section":"§2.2 (Method) and Table 1"}],"minor_comments":[{"comment":"The Author Contributions section lists Y.G., J.X.L.W., W.Z., and W.Q., but the byline contains only three authors (Leung, Gan, Zhang). Please reconcile.","section":"Author contributions"},{"comment":"Typos: 'reigonal-scale' should be 'regional-scale'; 'original draf' should be 'draft'.","section":"§3.2, penultimate paragraph; Author contributions"},{"comment":"The percentages in parentheses in Table 1 use ΔTgs as the denominator (e.g., 0.19/0.33 = 57.6%), whereas the same residual is expressed in §3.1 as 28.2% of ΔTmean. The two conventions should be unified or explicitly labeled to avoid confusion.","section":"Table 1"},{"comment":"Caption of Fig. S4 reads 'Same as Fig. S4, except for…' — a self-reference; presumably Fig. S3 is intended.","section":"Figure S4 caption"},{"comment":"The choice of the single year 1979 as the anomaly base is unconventional (a climatological reference period is standard) and is never justified; please explain and note any sensitivity, particularly for the reanalysis intercomparison in Table 1.","section":"§2.2 and §3.2"},{"comment":"Fig. 2(b) caption ends mid-word ('...and ΔT75th include.'); also 'include' should be 'included'.","section":"Figure 2 caption"},{"comment":"The skewness formula in §3.1 is typeset ambiguously (the exponents and the summation placement are unclear in the rendered equation); please reformat.","section":"§3.1, skewness definition"}],"recommendation":"major_revision","confidential_remarks":"The manuscript sits within a cluster of papers by the same group applying the image-histogram approach to climate fields (Leung et al. 2022; Gan et al. 2023, 2026a, 2026b; Liu et al. 2025; Qian et al. 2016, 2019). In particular, Gan et al. (2026a, npj Climate and Atmospheric Science, \"Beyond global mean temperature: increasing asymmetry of global warming in past and future climate change\") appears, from its title and citation placement, to address closely related ground; the editor may wish to verify the degree of overlap and whether the present manuscript's incremental contribution over that paper is clearly delineated. I do not raise this with the authors as a criticism, but a short statement of how this work differs from Gan et al. (2026a) would strengthen the submission."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful core here is straightforward and well executed: SAT-change fields are positively skewed, so area-weighted ΔTmean sits above median, mode, and a maximum-cross-correlation histogram shift (their ΔTgs) by ~0.19–0.23 °C (2000–2019 vs 1979) across four reanalyses and by 0.27–0.54 °C late-century in a 20-model CMIP6 ensemble, with Arctic latitudes explaining most of the gap. Methods are transparent (0.01 °C bins, correlation shift, t-tested trends), multi-source consistency is real, and the tables/figures make the numbers easy to check. That packaging of their prior histogram work into a systematic mean-vs-shift comparison is the actual increment; inhomogeneous warming and mean≠median under skew are not new.\n\nThe soft spot is interpretive, not numerical. ΔTmean − ΔTgs is just mean minus a robust location estimate of the same field—i.e., a skewness residual. It does not perform spatial-scale separation. Arctic amplification, which they correctly flag as the main driver, is a large-scale coupled pattern, not a leftover “regional” term you can percentage-decompose (71.8/28.2, 78.9/21.1) and then hold up against the Paris 2 °C target. The title/abstract claim that ΔTmean “overestimates global warming rate” therefore rests on elevating ΔTgs to the true global-scale rate by definition. If the energy-budget and policy quantity remains the area-weighted mean, the overestimate framing fails even while the inequality holds. Base year, bin width, and Arctic cutoff are free parameters but not load-bearing; the definitional move is.\n\nWho it is for: people who already track multi-metric SAT reporting or communication of polar amplification. A serious referee should see it—after the overestimate language is narrowed to “mean exceeds typical-area/histogram-shift warming.” I would engage the numbers, not the residual story as written.","headline":"Solid multi-dataset quantification that mean exceeds a histogram-shift location statistic under skew, but the title and “regional-scale” residual framing overreach what the statistics can support.","tokens_in":17219,"tokens_out":514,"would_cite":false,"duration_ms":10204,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Global-mean surface temperature change systematically exceeds the warming experienced over most of Earth's surface because SAT change is positively skewed.","keywords":["global mean temperature","global warming rate","global-scale warming","regional-scale warming","warming asymmetry","image histogram","Arctic amplification"],"falsifier":"Recompute ΔTgs with an alternative similarity measure or after masking the Arctic; if the mean–shift gap collapses or the histogram shape change itself accounts for most of the discrepancy, the overestimate interpretation fails.","tokens_in":16871,"feed_emoji":"🌡️","tokens_out":943,"duration_ms":25704,"temperature":0.7,"pith_summary":"Climate policy treats the area-weighted global-mean surface air temperature change, ΔTmean, as the single measure of the global warming rate. This paper shows that surface temperature change is positively skewed, so the arithmetic mean is pulled upward by extreme regional warming and therefore exceeds the temperature shift that actually occurs over the largest share of the planet. Using image histograms of global SAT, the authors define a global-scale change ΔTgs as the horizontal shift that best matches the base-year histogram; in reanalyses ΔTmean already exceeds ΔTgs by roughly 0.2 °C for 2000–2019 relative to 1979, and CMIP6 projections widen the gap to 0.27–0.54 °C by 2080–2099. The excess is interpreted as net regional-scale warming, dominated by Arctic amplification. The result is offered as a warning that relying on ΔTmean alone can misrepresent how much of the planet is warming at the headline rate and can therefore mislead both impact assessments and mitigation targets.","feed_headline":"Global mean temperature overstates warming over most of Earth","feed_subtitle":"Skewed SAT change, led by the Arctic, pushes the mean 0.2–0.5 °C above the shift felt by most of the surface","key_machinery":"The image-histogram representation of global SAT and the derived metric ΔTgs: the horizontal temperature shift that maximizes cross-correlation with the 1979 histogram, taken as the warming experienced over the majority of Earth's surface.","core_discovery":"Because the spatial distribution of surface-air-temperature change is positively skewed, the conventional global-mean change ΔTmean is systematically larger than the global-scale change ΔTgs obtained from the rigid horizontal shift of the SAT image histogram. The difference reaches 0.19–0.23 °C in recent decades and 0.27–0.54 °C by the end of the century under rising emissions, and is attributed mainly to Arctic amplification acting as a net positive regional warming contribution.","pith_inferences":["If negotiators treated ΔTgs rather than ΔTmean as the controlled variable, the allowable emissions path consistent with a 1.5 °C or 2 °C “global-scale” limit would be less stringent than present pathways framed on the mean.","The same histogram-shift logic could be applied to other spatially skewed fields (precipitation intensity, extreme heat days) to separate planetary-scale from regional-scale contributions.","Stabilizing or reducing Arctic amplification would shrink the mean–ΔTgs gap even if global-mean warming continued, offering a distinct regional lever on the headline indicator."],"forward_implications":["Headline global-warming figures used in policy already contain a several-tenths-of-a-degree regional contribution that is not experienced over most of the surface.","Under higher-emission pathways the mean–global-scale gap widens, so the same ΔTmean target corresponds to less uniform planetary warming.","Impact metrics tied to polar or regional extremes (ice loss, sea-level rise, local heat) will diverge further from ΔTmean-based projections.","Reporting median, mode or histogram-shift measures alongside the mean would give a more complete picture of how much of the planet is warming at the stated rate."],"fun_headline_variants":["Global-mean SAT change runs 0.2°C hotter than most of Earth's surface","Arctic skew makes ΔTmean overstate the warming most places feel","Histogram shift shows mean warming exceeds global-scale change","Positive skew from Arctic amplification inflates the global mean","ΔTmean exceeds ΔTgs by 0.19–0.54°C as warming grows uneven"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The claim that the maximum-cross-correlation horizontal shift of the global SAT histogram is the physically correct definition of “global-scale warming,” so that any excess of the arithmetic mean over that shift counts as an overestimate.","fun_headline_variants_meta":{"raw":{"variants":["Global-mean SAT change runs 0.2°C hotter than most of Earth's surface","Arctic skew makes ΔTmean overstate the warming most places feel","Histogram shift shows mean warming exceeds global-scale change","Positive skew from Arctic amplification inflates the global mean","ΔTmean exceeds ΔTgs by 0.19–0.54°C as warming grows uneven"]},"model":"grok-4.5","effort":"low","cost_usd":0.00374,"raw_usage":{"total_tokens":1263,"prompt_tokens":905,"num_sources_used":0,"completion_tokens":101,"cost_in_usd_ticks":37404000,"prompt_tokens_details":{"text_tokens":905,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":257,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":905,"tokens_out":101,"duration_ms":5703,"temperature":1.0,"reasoning_tokens":257,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-30T19:34:37.414751+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Recompute ΔTgs with an alternative similarity measure or after masking the Arctic; if the mean–shift gap collapses or the histogram shape change itself accounts for most of the discrepancy, the overestimate interpretation fails.","supporting_citations":[],"review_version":1}