2023 Temperatures Reflect Steady Global Warming and Internal Sea Surface Temperature Variability

The study addresses whether the exceptional global mean surface temperature anomaly (GSTA) in 2023 indicates an acceleration in surface warming or can be explained by internal variability and known regional forcings. While anthropogenic warming has proceeded at a broadly steady rate (~0.2 °C/decade since ~1970), recent literature notes a modest step-up in warming rates around 1990, a continued rise in Earth’s energy imbalance, and accelerated ocean heat content accumulation. Potential drivers include ongoing greenhouse gas increases and reduced aerosol cooling due to sulfur emission reductions (e.g., in China and international shipping). Despite the shift to El Niño conditions and unprecedented SSTs in multiple basins (Equatorial and North Pacific, North Atlantic, Southern Ocean), the magnitude of the 2023 record prompted questions about rapid system changes or underestimated sensitivities. The purpose is to quantify the contribution of sea surface temperature (SST) spatial patterns to the 2023 anomaly and compare with historical strong-anomaly years to assess consistency with internal variability under steady anthropogenic warming.

The paper builds on research documenting steady anthropogenic warming rates since the 1970s, with suggestions of a minor acceleration around 1990, increases in Earth’s energy imbalance, and faster ocean heat accumulation. Proposed causes include sustained greenhouse gas forcing and diminished aerosol cooling following emissions controls (notably in China and the shipping sector). Studies have explored potential roles of the 2021 Hunga Tonga eruption, shipping aerosol changes, and implications for effective climate sensitivity. The SST “pattern effect” literature shows that the spatial distribution of SST anomalies modulates global temperature via atmospheric dynamics and feedbacks. Green’s function approaches, developed and tested across multiple Earth system models, quantify how localized SST perturbations influence radiation, clouds, water vapor, and ultimately global temperatures. Prior work indicates broadly consistent spatial response patterns across models, and that such functions can capture key features of interannual variability and teleconnections.

• Approach: Use a Green’s function (GF) based on the NCAR CESM1.2.1-CAM5.3 Earth System Model to isolate the portion of interannual (monthly/annual) GSTA variability attributable to observed SST spatial patterns after removing the global warming trend at each grid point.
• Observational datasets: Primary data are HadCRUT5 gridded monthly surface temperature anomalies, version 5.0.2.0, spanning January 1850–December 2023. For consistency checks, three additional monthly gridded products are used through Dec 2023: GISTEMP v4, NOAAGlobalTemp v5.1, and Berkeley Earth.
• Trend removal: To isolate internal variability, a 10-year boxcar smoothing is applied at each grid point relative to an 1850–1900 baseline, removing long-term global warming, spatial patterns of warming, and seasonal differences. Endpoint mirroring ensures consistent weighting near series ends. This method does not remove all decadal-scale regional variability.
• Green’s function construction and application: The CESM1-derived GF links an idealized local SST increase to atmospheric radiative/dynamical responses and global-mean surface air temperature (2 m). It is based on 40-year fixed-SST simulations with SST perturbations over 74 partially overlapping ocean patches (80° longitude × 40° latitude). Monthly observational fields are regridded to the GF resolution (2.5° latitude, 1.9° longitude). For each month, the detrended SST pattern is multiplied by the corresponding monthly GF and summed over ocean-dominated grid points to yield the SST pattern-induced modulation (correction) of GSTA. The method makes no assumptions about the causes (forced vs natural) of SST trends/variability.
• Ocean basin analyses: Global corrections are decomposed into contributions from major ocean regions defined by longitude/latitude boxes, including Equatorial West/East Pacific, North Pacific, Tropical and Subtropical North Atlantic, Indian Ocean, and Southern Ocean (coordinate ranges provided in Table 1 of the paper).
• Model comparison: CMIP6 historical (1850–2014) and SSP extensions (2015–2100; primarily SSP1-2.6, SSP5-8.5; results insensitive to SSP choice) provide an ensemble-based distribution of annual SST pattern-induced corrections. Multiple ensemble members for some models enhance sampling of internal variability. Only monthly mean global surface air temperature at 2 m is used. The GF response patterns have been shown comparable across models in prior studies, and the applicability to monthly/interannual variability is supported by rapid atmospheric adjustment timescales (e.g., Rossby wave propagation).

• Record margin context: Across four temperature reconstructions (HadCRUT5, NOAA, Berkeley Earth, GISTEMP), most post-1970 records were set by ~0.05 °C margins; El Niño years reached up to ~0.15 °C. In 2023, record margins were ~0.17 °C in HadCRUT5 and Berkeley Earth, and comparable to other strong El Niño years in the other series. There is no clear trend of increasing record margins over time, consistent with steady surface warming modulated by internal variability.
• SST pattern contributions: After removing SST pattern influences via the GF-based correction, 2023 remains the warmest year on record but aligns with recent long-term warming trends rather than exceeding them anomalously.
• Historical consistency: Monthly SST-induced corrections in 2023 are within the historical spread (1950–2023). Years with similarly strong corrections include 1952, 1969, 1998, and 2016. Annual/global corrections for these years are comparable to 2023 across HadCRUT5 and corroborated by the other observational series.
• CMIP6 comparison: The 2023 annual SST pattern-induced correction lies in the upper 5th percentile of model-simulated historical variability but is not exceptional.
• Warming rates after filtering: GF-filtered warming rates (HadCRUT5) are 0.29 °C/decade (2014–2023), 0.27 °C/decade (2004–2023), and 0.19 °C/decade (1974–2023).
• Geographic contributions: 2023 saw simultaneous warm anomalies across several basins—El Niño in the Equatorial Pacific and anomalously warm Tropical North Atlantic. The Tropical North Atlantic (0–30° N) made its strongest contribution on record; the Subtropical/North Atlantic also contributed strongly but remained within previous observations. Southern Ocean results are more uncertain due to incomplete observational coverage in some years.
• Robustness across datasets: All four observational series identify the same strongest-correction years and support that 2023’s SST pattern influence was strong but not unprecedented.

The analysis indicates that the exceptional 2023 global temperature anomaly can be largely explained by the combined warm state of multiple ocean basins superimposed on steady anthropogenic warming, rather than a sudden acceleration in global surface warming. While individual basins were not uniquely anomalous relative to historical variability, their concurrent warmth amplified the global anomaly. The results address the core question by showing that SST pattern effects comparable to 2023 have occurred in prior strong El Niño years (e.g., 1998, 2016), and that record margins fit expectations under steady warming modulated by internal variability. However, the method does not diagnose the causes of the SST patterns themselves. Potential contributors—such as changes in aerosol emissions (including shipping), effects from the Hunga Tonga eruption, shifts in Earth’s energy imbalance, cloud changes, or unusual evolution of ENSO following a multi-year La Niña—could underlie the observed SST configurations and thus indirectly influence the corrections. Therefore, while 2023 is consistent with internal variability plus steady forcing, multiple mechanisms may have contributed to the SST patterns that produced the observed anomaly.

Using a Green’s function-based approach to quantify SST pattern effects, the study finds that 2023’s record-high global temperature is consistent with steady anthropogenic warming combined with internal SST variability and regional forcing, rather than signaling a step change in warming rates. The SST pattern influence in 2023 was strong but within the range of past observations and model-simulated variability, and after correction 2023 aligns with the ongoing warming trend. Future work should investigate the causes of recent SST patterns, including assessments of upper-ocean heat uptake across basins, the climatic impacts of recent aerosol emission changes (e.g., shipping regulations), potential effects of volcanic injections, and the dynamics of ENSO transitions following prolonged La Niña conditions. Enhanced observations in data-sparse regions (e.g., the Southern Ocean) and cross-model GF evaluations will further constrain SST pattern effects.

• Causality: The GF method attributes temperature modulation to observed SST patterns but does not identify the underlying drivers (forced vs natural), so external factors (aerosols, volcanic effects, energy imbalance changes, cloudiness) could still be influential.
• Model dependence: Results depend on the CESM1-derived Green’s function; although inter-model similarities exist, some sensitivity to model response patterns is possible.
• Temporal response: Applying equilibrium-derived GFs to monthly variability assumes rapid atmospheric adjustments; slower components may be imperfectly captured.
• Detrending choice: The 10-year boxcar smoothing does not remove decadal-scale regional variability, potentially leaving some low-frequency signals in the residuals.
• Observational coverage: Incomplete historical observations in certain regions (notably the Southern Ocean) introduce uncertainty in basin-specific corrections.
• Dataset differences: Minor differences among observational temperature reconstructions can affect rankings and magnitudes, though overall conclusions are robust.