Satellite-observed strong subtropical ocean warming as an early signature of global warming

The study addresses how much warming to anticipate from rising greenhouse gases and the current status of global warming, focusing on the ocean as the primary heat sink. While long historical SST datasets (e.g., ICOADS) exist, they carry uncertainties due to inconsistent sampling; satellite observations now provide about 40 years of improved, homogeneous coverage. Satellite-derived warming patterns show enhanced warming along subtropical western boundary current extensions and a cooling patch in the subpolar North Atlantic, as well as stronger Northern Hemisphere warming and slight Southern Ocean cooling. These features have been attributed to internal variability (e.g., PDO) and circulation changes (e.g., AMOC). The authors hypothesize instead that the dominant subtropical warming with weak subpolar warming is a dynamically constrained early, transient response to increased CO2, linked to surface water convergence/divergence and upper-ocean circulation, and that on longer timescales subpolar warming will dominate as deep ocean warming progresses.

Prior work identified accelerated warming over subtropical western boundary current extensions (explained by intensification and poleward shift of these currents) and a subpolar North Atlantic cold patch linked to AMOC weakening. Hemispheric asymmetry shows stronger Northern Hemisphere warming; slight Southern Ocean cooling has been connected to sea-ice expansion, deep upwelling, and Antarctic ozone depletion. Observed patterns resemble the negative PDO phase, often interpreted as internal variability. Paleoclimate evidence from the mid-Pliocene, Last Glacial Maximum, Miocene, Eocene, and Cretaceous consistently indicates greater subpolar sensitivity to GHG forcing than subtropics, due to sea-ice constraints and nonlinear radiative and evaporative feedbacks. Previous studies also emphasized the role of mean ocean circulation in shaping surface warming patterns and the transient nature of Southern Ocean and tropical Pacific responses under different forcings.

Data and models used: (1) Satellite SST: NOAA OI SST V2 (1982–2022) to derive observed trends. (2) CMIP6 simulations: abrupt4xCO2, 1pctCO2, historical, and piControl from up to 47 models (22-model ensemble for main abrupt4xCO2 analysis). SST anomalies in abrupt4xCO2 are computed relative to a three-year piControl baseline around each model’s initial condition. All fields are bilinearly interpolated to 1°×1°. Pattern correlations between observed trends (70°S–70°N) and model anomalies are calculated over time. (3) PlioMIP2 mid-Pliocene simulations: ensemble mean from seven models (including COSMOS), with CO2 set to 400 ppm for Pliocene and 280/284 ppm for preindustrial, compared with PRISM proxy SSTs (95 sites using faunal assemblages, Mg/Ca, alkenones). (4) Long integration: a 3000-year abrupt4xCO2 experiment with AWI-ESM (ECHAM6 atmosphere; FESOM2 ocean-ice, variable resolution ~25–110 km; 47/46 vertical levels), initialized from a preindustrial control, reaching near-equilibrium with TOA imbalance <0.15 W m−2. (5) Ocean-only sensitivity: FESOM1.4 simulations under CORE2 forcing, with a simplified bulk heat flux scheme where net surface heat flux is linear in SST–air temperature difference. A Uniform4K experiment applies a uniform +4 K surface air temperature anomaly over ice-free oceans (60°S–60°N); five 40-year ensemble members are compared to the last 100 years of a 1000-year control. (6) Ocean age tracer: AWI-ESM preindustrial run with an age tracer reset to 0 at the surface each time step and incremented below; integrated 3000 years to diagnose mean age of the upper 300 m, indicating source regions and ventilation pathways. (7) Removal of internal variability: PDO and AMO signals are identified via linear regression of HadISST (1900–2021) onto PDO/AMO indices; reconstructed variability Y(x,y,t)=R(x,y)·V(t) is subtracted from observed SST to reassess trends. (8) Diagnostics: barotropic streamfunction computed from depth-integrated transport to delineate gyres; analyses include spatial overlays of gyre boundaries and trend maps, and zonal-mean vertical sections of temperature anomalies. Physical framework references include Clausius–Clapeyron and Stefan–Boltzmann laws to interpret latitude-dependent sensitivity.

- Satellite observations (1982–2022) show pronounced subtropical ocean warming and cooling in the eastern equatorial Pacific; the subtropical enhancement occurs across Pacific, Atlantic, and Indian Oceans and persists even after removing PDO and AMO signals, indicating it is not dominated by internal variability. - Enhanced subtropical warming aligns with subtropical gyre structures and regions of surface convergence/downwelling (western intensification). FESOM1.4 Uniform4K experiments, despite uniform atmospheric warming, produce amplified subtropical SST warming co-located with gyre maxima, demonstrating ocean dynamics focus surface heat into subtropics. - CMIP6 abrupt4xCO2 experiments reproduce the observed early-stage spatial warming pattern: pattern correlation 0.53 in year 1 and 0.67 after a decade. The resemblance exists for only 1–2 decades; thereafter patterns diverge as longer-term responses develop. Similar evolution occurs in 1pctCO2 runs. - Ocean age tracer simulations show young upper-ocean water (0–5 years) in subtropics and older water in subpolar and eastern equatorial Pacific upwelling zones, supporting rapid subtropical surface warming versus delayed subpolar response. - Long-term AWI-ESM 3000-year abrupt4xCO2 integration shows ongoing global mean ocean warming (0.04 °C per century at simulation end) with largest surface warming emerging in subpolar regions and deep ocean as the system approaches quasi-equilibrium, consistent with mid-Pliocene reconstructions and PlioMIP2 models. - Paleoclimate evidence indicates stronger subpolar sensitivity: mid-Pliocene subpolar SST anomalies ~2–5 °C; Eocene deepwater around 10 °C when CO2 ~4× preindustrial, implying high polar ocean temperatures. - The observed global mean surface temperature increase of 1.31 °C to date occurred while many upwelling zones exhibited little warming or cooling; accounting for the pattern effect implies committed warming at current GHG levels near 2.31 °C, exceeding the Paris Agreement threshold. - Discrepancies: CMIP6 fails to reproduce observed southern Indian Ocean cooling (likely missing iceberg drift/melt processes) and initial eastern equatorial Pacific cooling (linked to mean-state cold tongue biases and possibly aerosols or ozone). Some models with improved mixing (e.g., FIO-ESM2.0) or idealized CESM experiments can capture initial EEP cooling for a limited period. - Timescales: Reversal from subtropical-dominant to subpolar-dominant surface warming expected within a century in the North Pacific; requires millennial timescales in the North Atlantic and Southern Hemisphere due to deep-ocean warming dependence.

The findings support that the current satellite-observed pattern of strong subtropical warming with weak or negative trends in upwelling regions is an early, dynamically constrained response to increased CO2, governed by surface water convergence (downwelling in subtropical gyres) and divergence (upwelling in subpolar gyres and equatorial regions). This challenges the interpretation of the observed PDO-like pattern as primarily internal variability. As heat penetrates the deep ocean and sea ice retreats, subpolar amplification becomes dominant, aligning with paleoclimate reconstructions and long integrations. The transient subtropical-enhanced warming pattern can temporarily intensify meridional temperature gradients, shift oceanic and atmospheric circulation poleward, strengthen the Walker circulation, and enhance Southern Hemisphere westerlies and storm tracks. Over centennial to millennial timescales, as upwelling regions warm more strongly, these circulation changes can reverse. The recognition of this temporal evolution helps reconcile modern observations with paleo evidence and implies higher committed warming due to pattern effects, with profound implications for ice-sheet stability and sea-level rise.

This study demonstrates that enhanced subtropical ocean warming observed by satellites is a short-term, forced signature of anthropogenic CO2 increase, arising from upper-ocean circulation that concentrates surface heat into subtropical gyres. Climate model experiments and ocean age diagnostics corroborate the dynamics, while long-term simulations and mid-Pliocene reconstructions indicate that subpolar warming will ultimately exceed subtropical warming as the climate approaches quasi-equilibrium. The results imply substantial committed warming beyond current observations and potential risks to marine-terminating ice sheets as high-latitude oceans warm. Future research should: incorporate iceberg drift and meltwater processes into models to improve Southern Ocean and Indian Ocean regional responses; reduce equatorial Pacific mean-state and mixing biases to capture initial EEP cooling; better quantify regional timescales of pattern reversal and deep-ocean heat uptake; enhance observations of upper-ocean convergence/divergence and ventilation; and assess impacts of evolving warming patterns on atmospheric circulation and ice-sheet–ocean interactions.

- CMIP6 models generally lack explicit iceberg drift and melt processes, likely contributing to mismatches with observed southern Indian Ocean cooling. - Many models exhibit a cold tongue mean-state bias and related deficiencies that prevent reproducing the observed initial cooling in the eastern equatorial Pacific; aerosol forcing and ozone representation may also contribute to discrepancies. - The abrupt4xCO2 experiment is idealized; while useful for signal-to-noise, it does not represent historical forcing pathways. The FESOM Uniform4K experiment holds winds fixed, omitting circulation shifts that occur under warming. - Satellite SST trends span only four decades, limiting detection of slower deep-ocean and subpolar responses. Proxy reconstructions carry uncertainties in magnitude and spatial coverage. - Model-dependent results and ensemble averaging may smooth important regional features; ocean mixing and ventilation biases affect timing and magnitude of pattern evolution.