Comment on “2023 summer warmth unparalleled over the past 2,000 years” by Esper et al. (Nature, 2024)

Abstract

Esper et al. [1] assert that June–August 2023 temperatures over Northern Hemisphere extra-tropical land (30–90°N) were the warmest in 2,000 years, exceeding the upper 95% confidence bound of natural variability by >0.5 °C and producing a pre-Anthropocene-to-2023 temperature range of 3.93 °C relative to the volcanically cooled summer of CE 536. These claims rest on a narrow observational domain, circular baseline manipulations, a documented data error in the identification of the warmest pre-instrumental summer, and a 15-member tree-ring ensemble whose enormous subjectivity and low-frequency deficiencies have been thoroughly exposed. As McIntyre [2,3] demonstrates through detailed reverse-engineering, the paper deploys an updated “Mike’s Nature trick” via manipulative re-scaling and re-centering to manufacture a hockey-stick shape while concealing divergence. The underlying proxies, by the authors’ own prior admission [11], fail to record major millennial-scale temperature changes driven by orbital insolation. Independent peer-reviewed analyses further demonstrate that tree-ring chronologies are biased by CO₂ fertilization [10], suffer from the divergence problem [8], lose centennial-to-millennial variance due to the segment-length curse [7], and produce unreliable reconstructions when subjected to proper statistical scrutiny [16,12,13,14,15]. The central conclusions are therefore scientifically untenable and represent a continuation of long-discredited practices in tree-ring paleoclimatology.

1. Documented Data Error and Collapse of the Exceedance Claim

Table 1 and the core text erroneously identified the warmest pre-instrumental summer as CE 246 (+0.88 °C; 95% range −0.03 to +1.50 °C). This was a simple mislabeling of CE 242 data. The subsequent correction revises CE 246 to +0.89 °C (range +0.01 to +1.68 °C). The original assertion that 2023 (+2.07 °C observed anomaly) exceeded natural variability by >0.5 °C is thereby falsified: the corrected upper bound yields only a 0.39 °C exceedance, which vanishes within the ensemble’s methodological scatter [4]. This basic labeling error in a flagship journal already erodes confidence in the analysis of extremes.

2. Cherry-Picked Domain, Seasonal Target, and Spatial Bias

The study deliberately restricts analysis to 30–90°N land JJA temperatures from Berkeley Earth. This subdomain maximizes recent continental and summer-season warming signals while excluding oceans, tropics, and the Southern Hemisphere, where trends and variability are substantially smaller. Tree-ring proxies are overwhelmingly concentrated in NH mid-latitude boreal forests, inheriting the same geographic bias. Extrapolation to hemispheric or global policy contexts (e.g., Paris Agreement baselines) is invalid [5].

3. Circular and Manipulative Baseline Adjustments

Esper et al. [1] invoke a 0.24 °C “warm bias” in 1850–1900 instrumental data, derived by direct comparison with their own tree-ring ensemble (Extended Data Fig. 3), to inflate the 2023 anomaly from +2.07 °C to +2.20 °C relative to a longer pre-instrumental mean. This adjustment is circular and ad hoc. Schneider et al. [6] show the offset shrinks substantially when spatial coverage is properly matched, exposing sampling artifacts rather than systematic instrumental bias. The further incorporation of cold Common Era periods to depress the baseline lacks physical justification for evaluating a single-year extreme.

McIntyre [3] reveals additional manipulative re-centering and re-scaling operations applied to the Büntgen et al. [4] ensemble mean (Rmean). The authors re-center Rmean to the 1901–2010 period, scale its standard deviation to match Berkeley Earth observations, then re-center again to 1851–1900, operations with no statistical necessity whose sole apparent purpose is to reduce visible divergence between proxies and instrumental temperatures in the calibration era and to force a sharper hockey-stick appearance. When re-centered consistently to 1961–1990, the “adjusted” instrumental series aligns closely with the unmanipulated Büntgen target, exposing the artifice. This echoes the pattern of flawed centering and scaling first exposed in the original hockey-stick reconstructions [12,13,14].

4. Fundamental Flaws in the Proxy Ensemble and Confidence Intervals

The 15-member ensemble derives from only nine long tree-ring chronologies [4]. The authors themselves previously documented that “subjectivity in site and series selection, correction for biological age trends (detrending), and the climate calibration procedure” produces reconstructions that “differ in their mean, variance, amplitude, sensitivity, and persistence” [4]. McIntyre [3] demonstrates this subjectivity in practice: individual ensemble members range from near-perfect overfitting of recent instrumental temperatures (via inverse regression, including some that improperly incorporate instrumental data) to flat or declining series that show no 20th-century warming. Averaging these wildly inconsistent outputs into a mean (and deriving 95% intervals from their spread) is methodologically indefensible.

Tree-ring width chronologies suffer from the well-known “segment length curse” [7], which systematically suppresses centennial-to-millennial variability, and from the divergence problem, the post-1960s/1970s decoupling of tree growth from rising temperatures that further invalidates recent calibration and artificially inflates modern exceedance [8,9,5]. Graybill and Idso [10] additionally demonstrate that recent tree-ring growth is biased upward by aerial CO₂ fertilization, contaminating the temperature signal. Esper et al. [1] exacerbate these issues by applying post-hoc scaling that “airbrushes” divergence, exactly as McIntyre [3] documents.

Most damningly, Esper et al. [11], cited favorably by McIntyre [2], explicitly demonstrate in their Supplementary Information that tree-ring width records from high-latitude Northern Hemisphere sites fail to capture enormous multi-millennial summer temperature changes driven by orbital insolation (a 35 W/m² decline since the early Holocene; still ~6 W/m² since Roman times). Proxies such as treeline position and glacier equilibrium lines register these changes, but the tree-ring chronologies used in 2024 do not. Broader statistical evaluations confirm that such noisy, biased proxies cannot reliably reconstruct surface temperatures over the past 1,000–2,000 years [16,17,15]. The pseudo-confidence intervals attached to such insensitive proxies are therefore meaningless for bounding natural variability over 2,000 years.

5. Misrepresentation of Forcing, Single-Year Extremes, and Policy Overreach

The paper attributes the 2023 anomaly primarily to greenhouse gases amplified by El Niño while ignoring internal variability and the fact that the current El Niño was still unfolding. The 3.93 °C range to CE 536 is presented as proof of Anthropocene uniqueness, yet volcanic events dominate short-term variability while the proxies systematically underestimate pre-industrial warm periods. Broader multiproxy syntheses that incorporate non-tree-ring archives show far smaller Common Era variability [18], directly contradicting the “unparalleled” claim.

The title, abstract, and conclusion leap from this regionally and seasonally restricted, proxy-manipulated analysis to urgent calls for net-zero emissions. No sensitivity tests are performed for alternative domains, calibration periods, or detrending methods. Data archiving is minimal: Esper et al. [1] released only a single table of the final reconstruction; the underlying ensemble and measurement data remain incomplete or inaccessible [3].

Conclusion

Esper et al. [1] employ selective domain choice, circular adjustments, post-hoc scaling tricks to conceal divergence, and proxies demonstrably insensitive to long-term variability, in direct contradiction of peer-reviewed evidence on CO₂ fertilization bias, the divergence problem, the segment-length curse, and statistical unreliability of tree-ring reconstructions, in order to manufacture an appearance of unprecedented 2023 warmth. The published data error, combined with the authors’ own evidence of tree-ring limitations and McIntyre’s [2,3] forensic exposure of methodological manipulation, renders the central claims scientifically untenable. These results should not inform policy or public understanding of climate change.

References
[1] Esper, J., Torbenson, M. & Büntgen, U. 2023 summer warmth unparalleled over the past 2,000 years. Nature 631, 94–97 (2024) [and Author Correction: Nature 641, E11 (2025)].

[2] McIntyre, S. Jan and Ulf’s Nature Trick: The Hottest Summer in 2000 Years. Climate Audit, 24 May 2024. https://climateaudit.org/2024/05/24/jan-and-ulfs-nature-trick-the-hottest-summer-in-2000-years/

[3] McIntyre, S. Reconstructing the Esper Reconstruction. Climate Audit, 2 June 2024. https://climateaudit.org/2024/06/02/tracing-the-esper-confidence-intervals/

[4] Büntgen, U. et al. The influence of decision-making in tree ring-based climate reconstructions. Nat. Commun. 12, 3411 (2021).

[5] Anchukaitis, K. J. & Smerdon, J. E. Progress and uncertainties in global and hemispheric temperature reconstructions of the Common Era. Quat. Sci. Rev. 286, 107537 (2022).

[6] Schneider, L. et al. Constraining the nineteenth-century temperature baseline for global warming. J. Climate 36, 6261–6272 (2023).

[7] Cook, E. R. et al. The segment length curse in long tree-ring chronology development for paleoclimatic studies. Tree-Ring Bull. 55, 1–24 (1995).

[8] Loehle, C. A mathematical analysis of the divergence problem in dendroclimatology. Clim. Change 94, 347–353 (2009).

[9] D’Arrigo, R. et al. On the ‘Divergence Problem’ in northern forests: A review of the tree-ring evidence and possible causes. Glob. Planet. Change 60, 289–305 (2008).

[10] Graybill, D. A. & Idso, S. B. Detecting the aerial fertilization effect of atmospheric CO₂ enrichment in tree-ring chronologies. Glob. Biogeochem. Cycles 7, 81–96 (1993).

[11] Esper, J. et al. Orbital forcing of tree-ring data. Nat. Clim. Change 2, 862–866 (2012).

[12] McIntyre, S. & McKitrick, R. Corrections to the Mann et al. (1998) proxy data base and Northern Hemisphere average temperature series. Energy Environ. 14, 751–771 (2003).

[13] McIntyre, S. & McKitrick, R. Hockey sticks, principal components, and spurious significance. Geophys. Res. Lett. 32, L03710 (2005).

[14] McIntyre, S. & McKitrick, R. The M&M critique of the MBH98 Northern Hemisphere climate index: Update and implications. Energy Environ. 16, 69–100 (2005).

[15] McIntyre, S. & McKitrick, R. Proxy inconsistency and other problems in millennial paleoclimate reconstructions. Proc. Natl Acad. Sci. USA 106, E10 (2009).

[16] McShane, B. B. & Wyner, A. J. A statistical analysis of multiple temperature proxies: Are reconstructions of surface temperatures over the last 1000 years reliable? Ann. Appl. Stat. 5, 5–44 (2011).

[17] von Storch, H. et al. Reconstructing past climate from noisy data. Science 306, 679–682 (2004).

[18] PAGES 2k Consortium. Consistent multidecadal variability in global temperature reconstructions and simulations over the Common Era. Nat. Geosci. 12, 643–649 (2019).