{"id":"dc36e86a-2e64-4b2a-9a7f-3064de3fdc26","arxiv_id":"2506.13994","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"high","formal_verification":"none","parameter_count":5,"one_line_summary":"A review paper argues GCM-based climate attribution overstates anthropogenic warming, understates solar/astronomical influences, and overestimates equilibrium climate sensitivity (ECS), with empirical models projecting moderate 21st-century warming.","lead":"This paper argues that global climate models (GCMs) have major unsolved problems: they cannot reproduce natural climate swings like the Medieval Warm Period and the Holocene Thermal Maximum, and they may exaggerate human-caused warming. The author claims that solar and astronomical influences are underestimated and that real climate sensitivity to CO2 is lower than the IPCC's best estimate, so future warming may be moderate and Net-Zero policies may not be necessary.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 20/50/30 attribution rests on an unconstrained solar-sensitivity ratio (αS/αA≈5) estimated from collinear low-frequency forcings; if that ratio fails, the low-ECS and moderate-warming conclusions collapse.","rationale":"Rejecting the paper because its conclusions differ from the IPCC would be circular; the relevant question is whether the paper's own evidence supports its central claim. It does not, for a concrete technical reason. The claim is not merely that GCMs have a 'hot model' problem or that TSI reconstructions are uncertain — both are legitimate scientific positions and the paper cites real literature for them. The load-bearing step is the conversion of those uncertainties into a specific 20/50/30 attribution and ECS ≈ 1.1 °C. That conversion runs through Eq. 9, where the solar and anthropogenic forcings are separated by regression under an assumed TSI shape. Since the two forcings are collinear at low frequencies and since the paper itself identifies the high-variability TSI as the contested input, α_S/α_A ≈ 5 cannot be distinguished from the prior choice of F_S. The paper concedes the out-of-sample fragility of its harmonic models (Section 5.5), but the problem is already present in the calibration period. My proposed test settles whether the attribution ratio is robust or an artifact. The reader's weakest-assumption analysis pointed to the same region (the empirical low-ECS models, the 4–6× solar sensitivity), though it framed the issue mainly as extrapolation to 2100; I would frame it as identifiability of α_S in the calibration period. Heavy reliance on the author's own prior papers is a selection-bias concern, but the decisive defect is technical. I therefore recommend no change to the reader's REJECT verdict.","tokens_in":50379,"tokens_out":7029,"duration_ms":57501,"concrete_test":"Re-fit Eq. 9 of Section 5.4 to the same temperature and volcanic data, replacing the high-variability TSI with the Matthes et al. (2017) TSI forcing used by CMIP6. If α_S/α_A drops from ≈5 to ≈1 and the implied ECS rises above 2 °C, the 50/30 attribution is not robust to the contested forcing input. Stronger control: generate a synthetic temperature series from a known model with ECS = 3 °C, α_S = α_A, and low-variability TSI, then apply the paper's regression with high-variability TSI; if it returns α_S/α_A ≈ 5 and ECS ≈ 1.1 °C, the estimator is biased and the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.4's Eq. 9 is a linearized 0-D energy balance model: ΔT(t) − ΔT(t−1) = α_A F_A(t) + α_V F_V(t) + α_S F_S(t) − β ΔT(t−1). The paper's headline attribution (Section 6: ~50% solar, ~30% anthropogenic, ECS ≈ 1.1 ± 0.4 °C) is not a measurement; it is obtained by allowing α_S ≠ α_A and reporting α_S/α_A ≈ 5.1 (Section 5.6). The decisive problem is identifiability. Over 1850–2020, F_A (greenhouse gases plus aerosols) and the high-variability F_S are both slowly varying, low-frequency series with strong overlap. A multiple regression can separate their contributions only if the shape and amplitude of F_S are known a priori. That shape is exactly the contested 'high-secular-variability TSI' reconstruction (Section 4.2.2), which the paper itself contrasts with the CMIP6 low-variability Matthes et al. (2017) forcing. The paper offers no independent physical constraint on α_S; the 4–6× 'hypersensitivity' is an output of the fitted model, not a measured climate property. It also concedes (Section 5.5) that such empirical/harmonic fits 'may perform well within the calibration interval, and yet they could diverge substantially from reality outside the regression period.' If α_S/α_A is an artifact of the TSI choice, the 50% solar share, the residual 30% anthropogenic share, and the ECS ≈ 1 °C estimate all collapse, taking the moderate-warming/no-Net-Zero policy conclusion with them.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript argues that the CMIP/IPCC assessment that nearly all post-1850 warming is anthropogenic is undermined by data biases, underestimated solar and astronomical variability, and model deficiencies. It reviews evidence on urban heat island effects, homogenization, satellite–surface discrepancies, total solar irradiance reconstruction, the Medieval Warm Period and Holocene conundrums, and then presents empirical models: a zero-dimensional energy-balance regression (Eq. 9), an astronomically based harmonic model (Eq. 10), and a synthetic solar-based climate model (Fig. 22). These yield an equilibrium climate sensitivity of about 1.1 ± 0.4 °C, a decomposition of observed warming into roughly 20% record bias, up to 50% solar contribution, and about 30% anthropogenic contribution, and a projection that 21st-century warming under SSP2-4.5 remains below 2 °C. The conclusion is that Net-Zero policies are not necessary to meet the Paris targets.","tokens_in":50693,"tokens_out":11005,"duration_ms":90341,"significance":"If the central claims were established, the manuscript would overturn the current detection and attribution consensus and would have immediate policy consequences, so the stakes are high. The paper has useful features: it assembles a broad and current literature, reproduces many source figures, provides data links, and explicitly concedes in Section 5.5 that empirical harmonic fits can diverge outside the calibration interval. However, the quantitative claims are conditional on unvalidated choices—high-variability TSI, αS ≠ αA, fixed harmonic periods, and assumed observational uncertainties—and, as documented below, the main results are not reproducible from the equations as written. The paper therefore does not provide a sound basis for its headline attribution or policy conclusion.","major_comments":[{"comment":"The paper's central attribution—roughly 50% solar, 30% anthropogenic, 20% record bias, and ECS ≈ 1.1 ± 0.4 °C—is a regression output, not a measurement. In Eq. (9), the coefficients αA, αV, αS, and β are fitted to observed temperature, and the headline scenario requires both a high-variability TSI forcing and the assumption αA = αV ≠ αS. Over 1850–2020, F_A and F_S are both smooth, slowly varying series, so the ratio αS/αA ≈ 5.1 is identified only by the shape and amplitude of the chosen F_S—the very quantity contested in Section 4.2.2. No independent physical constraint or out-of-sample test is offered for this ratio; the same model with equal sensitivities and the CMIP6 low-variability TSI gives ECS ≈ 2.1 ± 0.7 °C (Section 5.4). The 20/50/30 decomposition in the Conclusion therefore collapses if the contested TSI reconstruction is replaced.","section":"§5.4, §5.6, §6"},{"comment":"The derivation of ECS is internally inconsistent as written. Immediately before Eq. (4), the text defines αA = kA/τ and β = 1/τ, but Eq. (4) asserts the equality kA F_A(t) = αA F_A(t), which requires αA = kA. Eq. (7) then defines ECS = 3.7 kA. Using the reported fitted values in Section 5.6 (αA = 0.083 °C m²/(W y), β = 0.303 y⁻¹), the two interpretations give different results: αA = kA/τ with τ = 1/β implies kA ≈ 0.274 °C/(W/m²) and ECS ≈ 1.0 °C, whereas αA = kA would give ECS ≈ 0.31 °C. A reader cannot reproduce the claimed ECS ≈ 1.1 ± 0.4 °C from the equations and definitions as stated.","section":"§5.4, Eqs. (4)–(7)"},{"comment":"The 21st-century projection that warming stays below 2 °C under SSP2-4.5 rests on the harmonic model in Eq. (10), with periods 60.95, 114.78, 129.95, 983.40, and 2318 years and amplitudes estimated from the calibration period. The manuscript itself concedes that such fits 'may perform well within the calibration interval, and yet they could diverge substantially from reality outside the regression period.' No out-of-sample forecast validation is presented, and the astronomical justification for the periods does not rule out the possibility that the detected peaks are spectral artifacts of red noise. In the absence of a genuine forecast test, the policy conclusion drawn from Fig. 21 is not supported.","section":"§5.5, Eq. (10), Fig. 21"},{"comment":"The argument for a dominant solar contribution is circular. Section 4.2.2 uses the paleoclimate correlation to argue for high-variability TSI models; that TSI series is then used in Eq. (2)/(9) to recover αS/αA ≈ 5.1 and ECS ≈ 1.1 °C (§5.6); Section 6 then presents this low ECS as evidence for 'hypersensitivity' and for speculative corpuscular and cloud mechanisms. The paper acknowledges in §4.2.3 that the CLOUD experiment found cosmic rays insufficient for nucleation, yet the conclusion does not resolve this conflict. The low-ECS result is therefore not an independent line of evidence; it is a restatement of the prior choice of TSI and sensitivity assumptions.","section":"§4.2.2–§4.2.3 and §6"}],"minor_comments":[{"comment":"The uncertainty ranges for the UAH-MSU and NOAA-STAR satellite records are 'assumed here' rather than estimated from the data; this assumption and its influence on the 20–30% excess-warming percentages should be stated explicitly in the text.","section":"§4.1.2, Table 1"},{"comment":"'Unlikely Eq. 1' should read 'Unlike Eq. 1.'","section":"§5.4, paragraph following Eq. (1)"},{"comment":"The 13 harmonic amplitudes, frequencies, and phases are not listed in the paper but are only referenced to previous work, which makes the central projection impossible to reproduce from the manuscript alone.","section":"§5.5, Eq. (10)"},{"comment":"There are several typographical errors: 'Inded' in §3.2, 'exgerate' in the Highlights, 'JJenkins' in the reference list, and a cross-reference to 'Section 3.3.1' in §4.1.2 where no such section exists (the hot-model discussion appears in §3.4.1).","section":"General"}],"recommendation":"reject","confidential_remarks":"The manuscript is largely a synthesis of the author's prior publications (e.g., Scafetta 2013a, 2023a, 2024), and the quantitative results are not independently verifiable from the material presented here. Given the policy stakes attached to the claims, the manuscript would need substantially stronger identification, out-of-sample validation, and internal consistency before it could be considered as a basis for revising the IPCC attribution assessment. Recommendation: reject."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper before citing it: it is a review/position piece, not new research. The headline numbers — 20% warm bias, 50% solar, 30% anthropogenic, ECS 1.1°C — reproduce the author's prior work and are not derived here. What is new is the packaging: a single document that assembles the main skeptical arguments against the IPCC consensus and connects them to a Net-Zero policy conclusion.\n\nCredit where it is due. The paper does some things honestly. It explicitly rejects the CO2 saturation fallacy, which is a common bad argument in this space. It concedes that GCMs remain essential tools. It also gathers a set of genuine open problems: the CMIP6 'hot model' bias, the Medieval Warm Period and Holocene conundrums, the TSI reconstruction controversy, and the urban heat island/satellite discrepancy. A reader wanting a catalog of these critiques will find a useful, well-referenced summary. The section on homogenization uncertainty and the Tmin/Tmax divergence is a legitimate research topic.\n\nNow the soft spots, and they are load-bearing. The central attribution claim in Section 6 is not a measurement. It rests on the zero-dimensional energy balance model in Eq. 9, where αS and αA are fitted to the same temperature record. The paper reports αS/αA ≈ 5.1, but that ratio is not identifiable with the low-frequency, collinear forcings used over 1850–2020. The shape of the solar forcing is itself the contested 'high-variability TSI' reconstruction. If you pick a different TSI series, the ratio and the ECS move a lot. The paper's own Section 5.5 admits that such harmonic models can diverge substantially outside the calibration period. So the 50% solar share, the 30% anthropogenic share, and the ECS ≈ 1°C are outputs of a fit, not independent constraints. The policy conclusion that Net-Zero is unnecessary inherits all of that fragility.\n\nCitation pattern is worth noting: heavy reliance on the author's own prior papers for the key numbers. That is not automatically a flaw, but here those papers are not formally verified or independently reproduced, and the new paper does not add independent evidence.\n\nWho should read it: people working on climate attribution who want a survey of the skeptical corpus and a reminder that some model-data discrepancies are real. It is not a reliable source for the 20/50/30 numbers. If I were advising a student, I would say: read Section 3.1 and 3.4 for the valid critiques, then treat Section 5 and the Conclusion as a worked example of overreach.\n\nRecommendation: this deserves peer review — the claims are important enough that a serious journal should get expert referees to evaluate the identifiability problem and the policy implications. But I would expect the reviewers to reject the central quantitative claims or demand major revision with a clear separation of the legitimate critiques from the unsupported attribution breakdown. As a preprint, it should not be cited as evidence for low ECS.","headline":"A well-written review of real gaps in climate modeling that overreaches badly: the 20/50/30 attribution is not derived, it is an unidentifiable fit dressed as a result.","tokens_in":51350,"tokens_out":1954,"would_cite":false,"duration_ms":22096,"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":"The paper argues that IPCC climate assessments overstate human-caused warming because the CMIP models underpinning them underrepresent natural solar and multidecadal variability, and that empirical models yield a lower climate sensitivity…","keywords":["climate change attribution","equilibrium climate sensitivity","solar forcing","total solar irradiance","empirical climate models","urban heat island bias","natural climate variability","climate policy"],"falsifier":"Track the predicted multidecadal phase: the empirical framework implies that the 1970–2000 warming was largely the warm phase of a quasi-60-year cycle, so global temperature should flatten or cool from the early 2000s through roughly 2030, so a sustained increase through 2035 matching the 1980–2000 trend would refute the natural-variability attribution. A second check is whether lower-troposphere satellite records continue to warm about 20–33% less than the homogenized surface records, since the paper's warm-bias hypothesis predicts that gap will persist or widen as urban-heat-island blending grows.","tokens_in":1990,"feed_emoji":"☀️","tokens_out":2295,"duration_ms":75684,"temperature":0.7,"pith_summary":"The paper sets out to show that the central IPCC attribution result—that essentially all observed global surface warming since 1850–1900 is human-caused—is not established, because the CMIP GCMs used to produce it underrepresent natural variability at decadal, millennial, and orbital timescales and are validated against surface temperature records with uncorrected warm biases. It argues that empirical and semi-empirical models, which directly fit the observed temperature record to a small number of forcings and astronomical harmonics, offer a more holistic alternative. If the paper is right, the equilibrium climate sensitivity is lower than the IPCC's central range, the reported warming is partly a measurement artifact, and the human share of the observed warming is roughly 30% rather than approximately 100%. That would matter because it would imply moderate 21st-century warming under realistic scenarios and would undercut the scientific case for costly Net-Zero policies.","feed_headline":"Solar and data biases may cut human warming share to 30%","feed_subtitle":"The paper argues CMIP models understate natural variability; climate sensitivity would be low and Net-Zero unnecessary.","key_machinery":"The load-bearing object is an empirical/semi-empirical family of models built on a zero-dimensional energy-balance equation with a single response time-constant, whose discretized form regresses temperature against anthropogenic, volcanic, and solar forcing plus a fast noise term. In one variant the solar component is replaced by thirteen astronomical harmonics, notably with periods of 60.95, 114.78, 129.95, 983.40, and 2318 years, fixed from astronomical considerations rather than freely fitted. What these models do is convert the same observed temperature record into an attribution split and an ECS estimate without depending on the cloud and water-vapor feedback parameterizations that drive the GCMs' high sensitivities. The mechanism that produces the low-ECS, high-solar result is a fitted sensitivity of climate to total solar irradiance that is four to six times larger than the sensitivity to greenhouse or volcanic forcing, which the paper attributes physically to solar-corpuscular and cosmic-ray–cloud modulation that the GCMs omit.","core_discovery":"The paper's central claim is that the near-100% anthropogenic attribution in the latest IPCC assessment is not empirically settled: the CMIP GCMs on which it rests systematically underproduce natural multidecadal and millennial variability, such as the quasi-60-year Atlantic oscillation, the Medieval Warm Period, and the Holocene Thermal Maximum; they use solar forcing functions with artificially low secular variability; and they are compared against surface temperature records that carry uncorrected warm biases. Reworking the same temperature history with empirical and semi-empirical models—a zero-dimensional energy-balance response to anthropogenic, volcanic, and solar forcings, plus an astronomically based harmonic component—yields a lower equilibrium climate sensitivity, as low as approximately \\(1.1 \\pm 0.4\\,^{\\circ}\\text{C}\\), a solar contribution up to about 50% of the reported warming since 1850, a non-climatic warm bias of roughly 20%, and a residual human contribution of roughly 30%. The paper concludes that 21st-century warming under the moderate SSP2-4.5 scenario would stay below about \\(2\\,^{\\circ}\\text{C}\\), so Net-Zero policies would not be necessary to meet the Paris Agreement targets.","pith_inferences":["The paper's attribution split of roughly 20% record bias, up to 50% solar, and 30% anthropogenic is a composite of several model variants rather than a single mechanistic estimate; a natural extension would be formal uncertainty propagation and out-of-sample hindcasts from 1850–1950 to 1950–2020.","The quasi-60-year and quasi-millennial harmonics imply a testable near-term prediction: after the current warm phase, a relative flattening or cooling should appear within the next decade or two, which existing observing systems could verify directly.","If the solar-corpuscular/cloud amplifier is real, a targeted causal test would compare cloud-cover anomalies from satellite data with cosmic-ray Forbush decreases and the 11-year solar cycle, moving beyond the correlations the paper presents.","The claim that Net-Zero policies are unnecessary also depends on SSP2-4.5 being a realistic emissions path; the paper cites work labeling higher SSPs as unlikely, but readers should note that even moderate-emission scenarios could become unrepresentative if carbon-cycle feedbacks strengthen."],"forward_implications":["If the 20% warm-bias estimate holds, reported historical warming and land-temperature trends would need to be revised downward, shifting both detection baselines and adaptation planning.","If solar and astronomical contributions really account for up to half the observed warming, GCM attribution studies that assume a near-zero natural trend would need to be recalibrated, and solar forcing sets with larger secular variability would need to be adopted.","If equilibrium climate sensitivity is near \\(1.1\\)–\\(2.1\\,^{\\circ}\\text{C}\\) rather than the IPCC's \\(2.5\\)–\\(4\\,^{\\circ}\\text{C}\\) central range, projected warming under SSP2-4.5 would stay moderate and the \\(2\\,^{\\circ}\\text{C}\\) threshold would not be crossed this century.","If the GCMs' failure to reproduce the Medieval Warm Period and the Holocene Thermal Maximum is real, paleoclimate validation would become a standard prerequisite for trusting GCM-based attribution claims.","If the empirical projections are correct, the scientific justification for Net-Zero-by-2050 policies as a Paris-compliant strategy would be seriously weakened, and adaptive or lower-cost mitigation approaches would appear sufficient."],"supporting_citations":[{"why":"Supplies the zero-dimensional energy-balance regression model that yields the empirical ECS estimate of \\(1.1 \\pm 0.4\\,^{\\circ}\\text{C}\\) and the four-to-six-fold solar hypersensitivity.","marker":"(Scafetta, 2023a)"},{"why":"Provides the empirical-model projections under SSP2-4.5 showing 21st-century warming staying below \\(2\\,^{\\circ}\\text{C}\\).","marker":"(Scafetta, 2024)"},{"why":"Gives an independent empirical-model result of moderate warming under business-as-usual emissions, corroborating the paper's projection.","marker":"(Connolly et al., 2020)"},{"why":"Defines the CMIP6 solar forcing that the paper criticizes for having minimal secular variability, which is central to the underestimation argument.","marker":"(Matthes et al., 2017)"},{"why":"Is the AR6 attribution claim of near-100% anthropogenic warming that the paper sets out to challenge.","marker":"(IPCC, 2021)"},{"why":"Shows that urban-inclusive station networks warm at 0.89°C per century versus 0.55°C for rural-only stations, supporting the warm-bias estimate.","marker":"(Soon et al., 2023)"},{"why":"Provides a reanalysis of climate-sensitivity evidence with a lower median ECS of about \\(2.16\\,^{\\circ}\\text{C}\\), buttressing the paper's low-sensitivity position.","marker":"(Lewis, 2023)"},{"why":"Criticizes the IPCC's reliance on a single Common Era temperature reconstruction, supporting the paper's claim that the Medieval Warm Period has been understated.","marker":"(Esper et al., 2024)"}],"fun_headline_variants":["CMIP models undercount natural variability, inflating human warming share","New analysis: human share of warming may be just 30%, climate sensitivity low","Data biases and solar shifts may slash human warming role to 30%","Low climate sensitivity, high solar influence: Net-Zero questioned","Earth's natural variability may explain half of observed warming"],"cache_read_input_tokens":53248,"weakest_assumption_plain":"The load-bearing premise is that the empirical models' fitted coefficients—the single response time of the climate system, the amplitudes and phases of the astronomical harmonics, and the regression parameters linking forcing to temperature—remain valid when extrapolated to 2100; the paper itself concedes that the harmonic model may perform well inside the calibration interval yet diverge substantially from reality outside it.","fun_headline_variants_meta":{"raw":{"variants":["CMIP models undercount natural variability, inflating human warming share","New analysis: human share of warming may be just 30%, climate sensitivity low","Data biases and solar shifts may slash human warming role to 30%","Low climate sensitivity, high solar influence: Net-Zero questioned","Earth's natural variability may explain half of observed warming"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000246,"raw_usage":{"total_tokens":1617,"prompt_tokens":1097,"completion_tokens":520,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":713,"completion_tokens_details":{"reasoning_tokens":429}},"tokens_in":713,"tokens_out":520,"duration_ms":4646,"temperature":1.0,"reasoning_tokens":429,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:09:02.908328+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Track the predicted multidecadal phase: the empirical framework implies that the 1970–2000 warming was largely the warm phase of a quasi-60-year cycle, so global temperature should flatten or cool from the early 2000s through roughly 2030, so a sustained increase through 2035 matching the 1980–2000 trend would refute the natural-variability attribution. A second check is whether lower-troposphere satellite records continue to warm about 20–33% less than the homogenized surface records, since the paper's warm-bias hypothesis predicts that gap will persist or widen as urban-heat-island blending grows.","supporting_citations":[],"review_version":1}