Cracking the Pacific Puzzle: Why Part of the Ocean is Cooling While the World Warms

Introduction to the Pacific Puzzle

For more than a decade, a profound contradiction between observational climate data and global climate simulations has perplexed researchers, representing one of the most significant unresolved issues in modern climate dynamics. While global mean temperatures have unequivocally risen in response to anthropogenic greenhouse gas emissions, satellite-era observations have revealed a persistent multidecadal cooling trend in the eastern tropical Pacific Ocean and the Pacific sector of the Southern Ocean. This trend has been particularly pronounced over the past forty-five years, with critical observational periods frequently analyzed between the years 1979 and 2014.¹ This localized cooling contradicts the output of nearly all standard low-resolution climate models, such as those comprising the Coupled Model Intercomparison Project Phase 6, which predominantly simulate widespread, uniform warming across these ocean basins.¹

This glaring divergence between observed reality and model projections has been colloquially termed the Pacific puzzle.¹ The inability of traditional models to capture the observed cooling pattern is not merely an academic curiosity; it has profound implications for global climate projections. The spatial pattern of sea surface temperature change in the tropical Pacific acts as a global pacemaker, dictating atmospheric circulation, global rainfall distribution, and the frequency of extreme weather events from the Americas to Australasia.² Furthermore, because models simulate warming where cooling is actually occurring, standard multimodel ensembles systematically overestimate historical global surface warming.²

Recently, a significant advancement achieved by researchers at the Max Planck Institute for Meteorology has offered the first comprehensive physical explanation for this discrepancy.¹ Utilizing a new generation of kilometer-scale, highly resolved coupled climate models, specifically the ICOsahedral Non-hydrostatic model, researchers successfully reproduced the observed multidecadal cooling trends in both the Southern Ocean and the southeastern tropical Pacific.² Driven by massive advancements in exascale supercomputing and international software refactoring initiatives, this high-resolution modeling framework resolves oceanic mesoscale eddies and regional topography that standard models can only estimate through mathematical parameterization.¹

The resulting historical simulations reveal a complex, interhemispheric teleconnection pathway. The findings demonstrate that explicitly resolved ocean eddies enhance heat export in the Southern Ocean, initiating a high-latitude cooling signal. This signal is transported equatorward by atmospheric circulation, amplified by coastal stratocumulus cloud feedbacks constrained by the Andes Mountains, and ultimately propagated westward across the equatorial Pacific via the coupled ocean-atmosphere dynamics known as the Bjerknes feedback.² This exhaustive report examines the historical context of the Pacific puzzle, the diverse theoretical frameworks previously proposed, the computational architecture that enabled its resolution, and the intricate physical mechanisms that drive these planetary-scale teleconnections.

The Historical Context of Tropical Pacific Sea Surface Temperatures

To understand the magnitude of the recent modeling breakthrough, it is necessary to thoroughly examine the historical trajectory of the Pacific puzzle and the specific observational metrics that define it. The discrepancy first gained widespread attention during the early 2010s, a period often associated with the global warming hiatus.⁵ During this time, the rate of global mean surface temperature increase appeared to slow down despite steadily rising atmospheric carbon dioxide concentrations, leading to intense scientific scrutiny and public debate regarding the efficacy of anthropogenic climate forcing.⁵ While later analyses with updated datasets confirmed that long-term global warming had not ceased and that the hiatus was largely an artifact of internal variability and observational coverage gaps, the spatial pattern of temperature change remained highly anomalous.⁵

Specifically, between 1979 and 2014, robust observations from datasets such as the Extended Reconstructed Sea Surface Temperature version 5 and the European Centre for Medium-Range Weather Forecasts Reanalysis version 5 showed a pronounced cooling trend in the eastern Pacific cold tongue and the Pacific sector of the Southern Ocean.² This cooling trend strongly resembles a prolonged La Nina-like state, characterized by a strengthening of the east-west sea surface temperature gradient across the equatorial Pacific Ocean.⁸ In a La Nina-like mean state, the western Pacific warm pool becomes anomalously warm, while the eastern Pacific cold tongue experiences enhanced upwelling and anomalous cooling. In stark contrast, historical simulations from standard atmosphere-ocean general circulation models almost universally predicted an El Nino-like warming pattern, where the eastern Pacific warms faster than the western Pacific, thereby weakening the zonal temperature gradient and dampening the prevailing trade winds.⁸

The persistence of the observed trend over several decades strongly suggested a systematic forced response rather than a random internal fluctuation.⁷ An extensive analysis comparing multiple sea surface temperature metrics, including east-west gradients, north-south gradients, and thermocline depth changes, revealed that the observed trends were at the extreme edge or entirely beyond the range of modeled internal variability in large multimodel ensembles.¹¹ The combination of thermocline shoaling and a lack of warming in the equatorial upwelling region simply could not be replicated by models relying on coarse spatial grids.¹¹ The scientific community was thus forced to confront the reality that the failure lay not in the observational record, but in the structural limitations of the climate models themselves, particularly their inability to simulate specific, small-scale dynamic processes that govern regional heat uptake, ocean circulation, and cloud formation.⁷

Evolution of Theoretical Frameworks and Competing Hypotheses

Prior to the deployment of kilometer-scale coupled simulations, climate dynamicists proposed numerous competing theories to explain the model-observation discrepancy. These theoretical frameworks attempted to bridge the gap using available coarse-resolution data and idealized modeling experiments. Exploring these hypotheses provides crucial context for understanding why the explicit resolution of physical processes eventually proved necessary.

The Transient Ocean Thermostat Mechanism

One of the most prominent theories advanced to explain the eastern Pacific cooling was the ocean thermostat mechanism. This theory posits that the fundamental dynamics of equatorial upwelling buffer the eastern Pacific against the immediate thermodynamic effects of greenhouse gas forcing.¹⁰ In the tropical Pacific, the prevailing easterly trade winds continuously push warm surface water toward the west, causing cold, deep water to upwell in the east. Under global warming, the western Pacific warms rapidly because its deep thermocline prevents cold water from reaching the surface. The eastern Pacific, however, warms much more slowly because the upwelled water originates from deep layers that have not yet been exposed to anthropogenic warming.¹⁰

According to this mechanism, the differential warming between the west and the east increases the zonal sea surface temperature gradient. This enhanced gradient strengthens the Walker circulation and the trade winds, which in turn drives even more upwelling of cold water in the east, effectively creating a self-sustaining thermostat that opposes the global warming trend.¹⁰ While observational data, including causal relationship analyses between sea surface temperatures, surface zonal winds, and subsurface temperature changes, supported the existence of this transient ocean thermostat, standard climate models consistently failed to reproduce it. Models, regardless of their specific coarse resolution configurations, exhibited a weak ocean thermostat effect due to asynchronous trends among the coupled variables and a systematic underestimation of the atmospheric heat flux damping required to sustain the feedback.¹⁰

The Role of Internal Multidecadal Variability

Another prevailing hypothesis argued that the observed cooling was primarily a manifestation of internal multidecadal climate variability, such as the Interdecadal Pacific Oscillation or the Pacific Decadal Oscillation, which happened to phase out of sync with external anthropogenic forcing.¹¹ The Interdecadal Pacific Oscillation naturally oscillates between warm and cool phases every twenty to thirty years. Proponents of this theory suggested that the climate system entered a strong negative phase in the late 1990s, temporarily masking the underlying forced warming signal in the eastern Pacific.¹¹

However, subsequent research cast significant doubt on this hypothesis as a complete explanation. While internal variability undoubtedly plays a role in decadal climate fluctuations, the magnitude and duration of the observed cooling—spanning more than forty years—pushed the boundaries of what could be statistically attributed to natural cycles alone. Furthermore, comparisons using the Multimodel Large Ensembles Archive demonstrated that the observed combination of enhanced east-west gradients and thermocline shoaling was extremely unlikely to be consistent with modeled internal variability, reinforcing the argument that the trends represent a distinct response to radiative forcing that models simply fail to capture.¹¹

Antarctic Meltwater and Aerosol Forcing

A growing body of literature also explored the potential influence of high-latitude forcing mechanisms, specifically the input of fresh meltwater from the Antarctic ice sheet. Studies suggested that Antarctic meltwater forcing could be partially responsible for the observed surface cooling and sea ice expansion over the Southern Ocean. The influx of fresh, buoyant water stratifies the ocean surface, limiting the upwelling of warmer deep waters and trapping the cooling signal at the surface.¹⁵ While meltwater experiments demonstrated a cooling effect in the Southern Ocean, the simulated responses often failed to replicate the precise magnitude of the observed cooling, and the strength of the subsequent teleconnection to the tropical Pacific was highly model-dependent and often too weak to fully explain the Pacific puzzle.¹⁵

Similarly, the effects of anthropogenic aerosols were investigated as a potential driver. Aerosols reflect incoming solar radiation and modify cloud properties, generally exerting a cooling effect on the climate. Some researchers hypothesized that asymmetrical aerosol forcing between the Northern and Southern Hemispheres could shift the Intertropical Convergence Zone and alter Pacific trade winds.¹² However, as aerosol emissions stabilized or declined in many regions due to clean air regulations while the Pacific cooling trend persisted, the aerosol hypothesis lost traction as the primary explanatory mechanism for the specific spatial pattern observed in the eastern Pacific.¹²

The Land-Sea Heating Contrast

More recently, dynamicists proposed the land-sea heating contrast as a potential driver of the multidecadal cooling. Because land surfaces have a lower heat capacity than the oceans, continents warm much faster than the surrounding seas under greenhouse gas forcing. In targeted experiments where carbon dioxide concentrations were artificially increased only over land masses, global coupled climate models exhibited a pronounced transient cooling of the eastern and equatorial Pacific.⁷ This land-sea thermal contrast drives a northward shift of the Intertropical Convergence Zone, a westward shift of convection over the Western Pacific, and a strengthening of the subtropical high-pressure systems via Rossby wave propagation.⁷ While this mechanism provides valuable insights into the atmospheric dynamics of transient cooling, it relies on the atmospheric warming effects overwhelming the oceanic cooling effects, which again is highly dependent on the accurate representation of cloud radiative feedbacks and regional oceanography.⁷

The following table summarizes the various theoretical frameworks proposed to explain the Pacific puzzle and their inherent limitations within the context of standard CMIP6 climate modeling.

Theoretical Framework	Core Mechanism	Limitations in Standard CMIP6 Models
Transient Ocean Thermostat	Upwelling of deep, un-warmed water maintains cold eastern Pacific; enhances zonal gradient and trade winds.	Models exhibit asynchronous coupling; atmospheric heat flux damping is too weak to sustain the feedback.
Internal Variability (IPO)	Natural multidecadal oscillations temporarily mask forced anthropogenic warming.	Observed cooling magnitude and duration fall outside the extreme bounds of modeled internal variability.
Antarctic Meltwater	Freshwater from ice melt stratifies Southern Ocean, trapping cooling at the surface.	Forced responses fail to replicate exact magnitude; teleconnections to tropics are overly weak.
Anthropogenic Aerosols	Hemispheric asymmetry in aerosol cooling shifts atmospheric circulation.	Timing of aerosol emission reductions does not align perfectly with the persistent cooling trend.
Land-Sea Heating Contrast	Rapid continental warming alters global atmospheric waves and strengthens subtropical highs.	Highly dependent on accurate representation of cloud radiative feedbacks, which are poorly parameterized.

Ultimately, none of these hypotheses could be fully validated using standard, coarse-resolution climate models because the models themselves were structurally incapable of resolving the fine-scale ocean eddies, precise coastal topographies, and marine stratocumulus cloud formations required to accurately simulate the coupled ocean-atmosphere response to external forcing.¹

The Paradigm Shift: The ICON Earth System Model

The resolution of the Pacific puzzle required a fundamental paradigm shift in computational climate science. Standard CMIP6 models typically employ horizontal grid spacings of roughly one hundred kilometers. At this coarse resolution, critical physical processes such as atmospheric deep convection, mesoscale ocean eddies, and highly localized coastal wind systems cannot be explicitly simulated. Instead, they must be represented through parameterizations—mathematical approximations that estimate the bulk statistical effects of small-scale processes on the larger grid.² Parameterizations are a well-documented source of uncertainty and systematic bias, particularly regarding the simulation of cloud radiative feedbacks, oceanic vertical mixing, and horizontal heat transport.⁷

To overcome these profound structural limitations, researchers turned to the ICOsahedral Non-hydrostatic Earth System Model. Jointly developed by a consortium of leading European institutions, including the Max Planck Institute for Meteorology, the German Weather Service, the German Climate Computing Center, and the Karlsruhe Institute of Technology, the model features a unified, highly scalable architecture designed for both operational numerical weather prediction and long-term, multidecadal climate projections.¹⁷ The transition to kilometer-scale resolutions—often referred to as storm-resolving in the atmosphere and eddy-rich in the ocean—represents a monumental technical achievement, requiring massive computational resources and extensive software engineering to run efficiently on heterogeneous supercomputer architectures.¹⁸

Configuration and Physical Resolution

For the specific investigation into the Pacific sea surface temperature discrepancy, researchers executed a coupled historical simulation utilizing the model with an unprecedented nominal horizontal grid spacing of ten kilometers in the atmosphere and five kilometers in the ocean.² This configuration is frequently denoted as R2B8 for the atmospheric grid and R2B9 for the oceanic grid.²⁰ The vertical resolution was equally ambitious, featuring ninety vertical levels in the atmosphere extending up to seventy-five kilometers, and one hundred and twenty-eight vertical levels in the ocean utilizing a z-star coordinate system to accurately capture both the planetary boundary layer and the deep ocean stratification.²¹ The dynamic time step for the simulation ranged between forty and one hundred and twenty seconds, requiring immense processing power to integrate over multidecadal timescales.²¹

Because the grid spacing was sufficiently fine, researchers were able to deactivate the parameterizations for numerous gray-zone processes. Deep convective cloud formation, orographic and non-orographic gravity waves in the atmosphere, and mesoscale turbulent eddies in the ocean were explicitly resolved according to the fundamental laws of fluid dynamics rather than approximated via statistical assumptions.² The coupled configurations utilize sophisticated sub-components, including the JSBACH or TERRA modules for land surface processes, the HAMOCC module for ocean biogeochemistry, and the YAC coupler to seamlessly transfer fluxes between the ocean, land, and atmosphere.¹⁸

Following standardized HighResMIP protocols, the historical simulation was branched from a rigorous control spin-up, allowing the model to generate its own internal climate dynamics free from the artificial constraints of forced parameterizations.² The spin-up process involved initializing the ocean with climatological distributions for temperature and salinity, followed by decades of integration under 1950s greenhouse gas boundary conditions.² The resulting historical simulation produced a global-mean surface temperature increase of 0.56 Kelvin over a thirty-five-year study period, closely matching the amplitude observed in the European Centre for Medium-Range Weather Forecasts Reanalysis version 5 dataset, while simultaneously capturing the elusive regional cooling trends in the Pacific and Southern Oceans.²

The following table contrasts the structural capabilities of the explicitly resolved kilometer-scale model against the standard parameterization-heavy models utilized in typical historical climate ensembles.

Model Characteristic	Standard CMIP6 Ensembles	Explicitly Resolved ICON Model
Atmospheric Grid Spacing	~100 kilometers	10 kilometers (R2B8)
Oceanic Grid Spacing	~100 kilometers	5 kilometers (R2B9)
Vertical Resolution	Typically 30 to 50 levels	90 Atmosphere / 128 Ocean levels
Mesoscale Ocean Eddies	Parameterized via diffusion equations	Explicitly resolved fluid dynamics
Deep Atmospheric Convection	Parameterized bulk mass flux	Explicitly resolved updrafts
Coastal Topography	Highly smoothed, broad representation	High-resolution, steep terrain mapping
Cloud Radiative Feedback	Structurally weak, highly uncertain	Physically realistic, emergent property

The Exascale Computing Frontier: The WarmWorld Initiative

The successful execution of a coupled Earth system model at kilometer-scale resolution is not merely a scientific achievement; it is a triumph of advanced software engineering and high-performance computing. The computational cost of running a five-kilometer ocean and ten-kilometer atmosphere model over multidecadal timescales is astronomical. Such simulations can consume hundreds of thousands of central processing unit node-hours on world-class supercomputers such as the Levante system at the German Climate Computing Center or the JUWELS Booster system at the Julich Supercomputing Centre.²³ To realize this capability, the European scientific community established massive funding initiatives, most notably the WarmWorld project, alongside interconnected programs such as Next Generation Earth Modeling Systems, Destination Earth, and European Eddy-Rich Earth System Models.¹⁹

The German federally funded WarmWorld project is specifically designed to develop the scalability of the climate model for exascale supercomputing, enabling the operational resolution of global oceanic and atmospheric coupled circulation systems.²⁵ The initiative brings together more than thirty scientists and software engineers from a dozen research institutes and universities, meticulously dividing the labor into four highly specialized operational modules: Better, Faster, Easier, and Smarter.²⁷

Module Better: Physical Fidelity and Configuration

The Better module focuses on defining, testing, and optimizing the physical configurations of the highly resolved model systems.²⁷ As state-of-the-art climate models push toward ever-higher resolutions, fine-grained physical processes that were previously ignored or parameterized find explicit representation within the model.²⁸ The Better module ensures that these newly resolved processes, such as turbulent mixing schemes based on turbulent total energy, function correctly without destabilizing the model's energy conservation.²⁰ Researchers within this module rigorously evaluate the impact of mixed-phase and ice-cloud processes, identify software bugs in turbulent mixing packages that cause runaway surface cooling over sea ice, and validate the model output against high-resolution observational data campaigns, such as the MOSAiC expedition in the Arctic.²⁹

Module Faster: Code Refactoring and Portability

The Faster module, coordinated by the German Climate Computing Center, takes on the critical software engineering challenge of transforming the legacy codebase into an open, scalable, and modularized framework known as ICON-consolidated.³⁰ Traditional climate models written in monolithic Fortran code struggle to utilize the heterogeneous architectures of modern supercomputers, which increasingly rely on highly parallel graphics processing units.³² The Faster module refactors the core dynamical solver and physics routines using directive-based programming models like OpenACC and advanced stencil loop languages, enabling portable performance improvements across diverse hardware environments.²³

This refactoring effort is essential for achieving the simulation throughput required for climate research, targeting metrics such as simulated years per day.³² The success of the Faster module's refactoring efforts culminated in prestigious international recognition when the model development team was awarded the Gordon Bell Prize for Climate Modeling, reflecting a major milestone in pushing the computational frontiers of high-resolution Earth system modeling.³⁴

Module Easier: Data Accessibility and Workflows

Running a multidecadal simulation at kilometer-scale resolution produces a staggering volume of data. A single model run can generate petabytes of raw output, creating a massive bottleneck for post-processing and scientific analysis.³⁶ The Easier module is explicitly tasked with making this exascale climate information visible, accessible, and interoperable for downstream users and the broader academic community.²⁵ The module develops novel data-centric workflows, utilizing modern, cloud-optimized data formats like Zarr, hierarchical equal area iso-latitude pixelation grids, and SpatioTemporal Asset Catalogs to abstract the data complexity away from the end-user.²⁹ This infrastructure ensures that researchers can rapidly query and visualize specific regional phenomena without needing to download massive global datasets.²⁵

Module Smarter: Applied Mathematics and Informatics

The fourth module, Smarter, seeks to integrate novel techniques from the applied mathematics and informatics communities to further enhance model performance.²⁵ This module funds projects addressing advanced data compression algorithms, simulation acceleration techniques, and in-situ analysis workflows where data is analyzed directly in the supercomputer's memory while the simulation is running, bypassing the slow process of writing petabytes of data to disk.²⁵ By soliciting proposals from external computer science domains, the Smarter module ensures that the climate modeling infrastructure remains at the cutting edge of technological innovation.²⁷

The coordinated efforts of the WarmWorld project demonstrate that solving long-standing climatological puzzles requires not just deep physical insight, but also world-class software engineering and data management capabilities.

WarmWorld Module	Primary Objective	Key Technical Focus Areas
Better	Improve physical fidelity at high resolutions.	Energy conservation, turbulent mixing schemes, validation against observational campaigns.
Faster	Achieve exascale computational performance.	Code refactoring, GPU porting, OpenACC directives, modularization into ICON-consolidated.
Easier	Manage exascale data output.	Zarr data formats, STAC catalogs, HEALPix grids, streamlined data access workflows.
Smarter	Integrate advanced computer science techniques.	Data compression, simulation acceleration, in-situ memory analysis.

Mechanism Part I: Southern Ocean Heat Uptake and Eddy Dynamics

With the computational architecture established, the high-resolution simulation was able to uncover the precise sequence of physical events that constitute the Pacific puzzle. The first critical mechanism originates not in the tropics, but in the hostile, turbulent waters surrounding Antarctica. The Southern Ocean is a region of immense global importance, serving as a primary sink for both planetary heat and anthropogenic carbon dioxide. However, traditional climate models notoriously fail to capture recent sea surface temperature trends in this region, generally simulating widespread artificial warming.¹⁴

In the observational record, such as the European Centre for Medium-Range Weather Forecasts Reanalysis version 5, the pronounced cooling of Southern Ocean sea surface temperatures occurs primarily in the Pacific sector. Paradoxically, this surface cooling occurs despite an increase in the net downward atmospheric heat flux entering the ocean from the atmosphere.² Basic thermodynamic principles dictate that if an area is absorbing more heat from the atmosphere but its surface is actively cooling, the thermal energy must be transported elsewhere by complex fluid dynamics. The observations indicate that this cooling extends deep throughout the water column, confirming that lateral heat export by ocean currents, rather than passive advection or localized surface cooling, is the primary driver of the temperature anomaly.²

The capacity to explicitly resolve ocean mesoscale eddies—swirling, turbulent fluid structures roughly ten to one hundred kilometers in diameter—is the primary reason the highly resolved model succeeds where standard CMIP6 models fail.² In coarse-grid models, the complex interactions between atmospheric wind stress, ocean boundary currents, and these turbulent eddies are parameterized using diffusion equations. Consequently, low-resolution models often simulate an excessively sluggish ocean response and fail to accurately transport heat across the massive, powerful fronts of the Antarctic Circumpolar Current.²

In the five-kilometer ocean resolution of the kilometer-scale model, the direct representation of vertical eddy heat transport along the Antarctic Circumpolar Current fronts enables critical dynamical adjustments.² Explicitly resolved eddies facilitate the enhanced uptake of heat from the surface into the deeper ocean interior. Furthermore, these fine-scale dynamical adjustments cause a local, northward displacement of the Antarctic Circumpolar Current fronts.² This meridional displacement reflects an expansion of the outcropping portion of deep, cold polar water masses, which are continually brought to the ocean surface by strong wind-driven upwelling.²

Because the turbulent eddies efficiently transport the excess absorbed heat northward and export it into other ocean basins, the localized surface waters in the Pacific sector of the Southern Ocean experience a net loss of heat, resulting in the observed, persistent cooling trend.² This high-latitude cooling mechanism operates as the initial thermodynamic trigger, setting off a complex chain reaction that ultimately reaches the equator. Without the kilometer-scale spatial resolution required to explicitly model these eddy dynamics, the initial cooling signal is never generated, causing the entire subsequent teleconnection pathway to fail entirely in standard models.²

Mechanism Part II: Equatorward Teleconnections and Topographic Blocking

The cooling signal generated in the Pacific sector of the Southern Ocean does not remain isolated in the high latitudes. It exerts a powerful, governing influence on the tropical eastern Pacific through a highly efficient atmospheric teleconnection pathway.² This teleconnection bridges the geographical gap between the high-latitude oceanic heat sink and the equatorial climate pacemaker, demonstrating the deeply interconnected nature of the global Earth system.

The transmission of the cooling signal begins via equatorward advection. The anomalously cold sea surface temperatures in the Southern Ocean induce a corresponding high-pressure anomaly in the overlying atmosphere.² This atmospheric high-pressure anomaly intensifies the climatological southeasterly trade winds that blow consistently toward the equator along the western coast of South America.² As these intensified winds sweep over the ocean surface, they dramatically enhance evaporation rates. Because evaporation is a latent heat transfer process that consumes thermal energy, the wind-evaporation-sea surface temperature feedback further depresses local ocean temperatures, efficiently carrying the cooling signal northward toward the tropics.²

As the cooling signal approaches the coasts of Chile and Peru, a second critical advantage of the high-resolution modeling framework becomes immediately apparent: the accurate representation of steep continental topography. The Andes Mountains form a massive, continuous wall extending along the entire western edge of the South American continent. In coarse climate models, topographic features must be artificially smoothed out over hundreds of kilometers to maintain numerical stability, which inadvertently allows warm, dry continental air from the South American interior to artificially spill westward over the coastal ocean. In the kilometer-scale model, the steep, jagged terrain of the Andes is explicitly represented, effectively blocking warm air advection from the interior toward the coast.¹

This precise topographic blocking is essential for maintaining a strong, low-level atmospheric temperature inversion—a distinct meteorological condition where cool, moist marine air is trapped near the ocean surface beneath a layer of warmer, drier subsiding air aloft.² Furthermore, the high-resolution mapping of coastal geography vastly improves the simulation of highly localized coastal wind systems. These parallel winds drive the mechanical upwelling of cold, nutrient-rich deep water directly along the shoreline through Ekman transport, further reinforcing the cooling trend near the continent.²

Mechanism Part III: Stratocumulus Cloud Radiative Feedbacks

The combination of cold upwelled coastal water, trapped marine boundary layer moisture, and a strong temperature inversion creates the perfect thermodynamic environment for the formation of marine stratocumulus clouds.² The Peruvian stratocumulus deck is one of the most extensive, persistent, and climatically important cloud systems on Earth. Unlike other subtropical cloud regions that are highly influenced by variable conditions in the free troposphere, the Peruvian stratocumulus region is highly sensitive to underlying sea surface temperature variations.²

As the ocean surface cools due to equatorward advection and enhanced coastal upwelling, the low-level atmospheric inversion strengthens, promoting the development of thicker, more extensive stratocumulus cloud decks.² These low, highly reflective clouds act as giant solar mirrors, exceptionally effective at reflecting incoming shortwave solar radiation back into space. By reflecting sunlight away before it has the opportunity to penetrate and heat the ocean surface, the stratocumulus clouds generate a massive positive feedback loop, dramatically amplifying the initial Southern Ocean cooling signal to a realistic magnitude.¹

In standard CMIP6 models, subtropical stratocumulus cloud feedback is notoriously weak or entirely misrepresented due to the reliance on bulk parameterization schemes that fail to capture the delicate physics of cloud-top entrainment and boundary layer turbulence.⁷ Regional cloud-locking experiments have definitively confirmed that the efficiency of the Southern Ocean teleconnection to the tropical Pacific is heavily dependent on the strength and accuracy of this subtropical cloud feedback.⁷ Because the finer grid of the explicitly resolved model allows for greater amplitude variations in individual grid cells and accurately captures both the steep topography and the coastal upwelling physics, the model naturally generates a sufficiently strong shortwave cloud radiative feedback to cool the eastern tropical Pacific to the magnitude observed in satellite data.¹

Mechanism Part IV: The Bjerknes Feedback and Equatorial Propagation

Once the amplified cooling signal reaches the southeastern tropical Pacific, it interacts with the fundamental coupled dynamics of the equatorial ocean and atmosphere. The physical mechanism responsible for dragging the coastal cooling westward along the equator is known as the Bjerknes feedback, named after the pioneering meteorologist Jacob Bjerknes, who first described the intricately coupled nature of the El Nino-Southern Oscillation.³⁹

The equatorial Pacific is naturally characterized by a stark zonal temperature asymmetry. The western Pacific warm pool hosts some of the warmest ocean waters on Earth, driving intense atmospheric convection. In contrast, the eastern equatorial Pacific features a pronounced cold tongue driven by continuous upwelling.⁴¹ This distinct temperature contrast drives the massive atmospheric Walker circulation: warm air rises over the western Pacific, travels eastward in the upper troposphere, sinks over the cooler eastern Pacific, and returns westward near the ocean surface as the prevailing easterly trade winds.¹⁰

The surface trade winds physically push warm surface waters toward the west, piling them up against the maritime continent and causing the oceanic thermocline—the steep thermal boundary separating warm surface water from cold deep water—to tilt. The thermocline becomes deep in the west and highly shallow in the east.⁴³ This shallow thermocline in the eastern basin allows the trade winds to easily upwell cold deep water to the surface, maintaining the temperature contrast and perfectly closing the coupled feedback loop.⁴³ The conceptual mathematical model underlying this system, often referred to as the recharge oscillator, dictates that the rate of change of sea surface temperatures depends on the delicate balance between local atmospheric damping, air-sea interaction loops, and the delayed response of equatorial Rossby waves reflecting off the western boundary.⁴⁵

When the Southern Ocean cooling effect, highly amplified by the stratocumulus clouds, arrives at the equatorial Pacific, it immediately merges into this existing coupled system.⁴⁷ The introduction of anomalously cold water into the eastern Pacific artificially increases the zonal temperature gradient between the east and the west. According to the mechanics of the Bjerknes feedback, this heightened temperature contrast forces a strengthening of the surface easterly trade winds.⁴⁷

The invigorated trade winds exert significantly greater mechanical stress on the ocean surface. This enhanced wind stress further lifts the thermocline in the eastern basin, increasing the mechanical upwelling of cold subsurface water.⁴⁷ Concurrently, the stronger winds increase evaporative latent heat flux along the equator, providing an additional cooling source.⁴⁷ The climatological Intertropical Convergence Zone acts as an atmospheric guide, steering the propagation of this cooling signal steadily westward across the basin.² The continuous, positive reinforcement between the cooling ocean, the strengthening winds, and the shoaling thermocline solidifies a distinct, triangular cold patch extending from the southeastern Pacific coast into a wide zonal band across the equator.⁴⁷

Standard models frequently suffer from a fundamentally weak Bjerknes feedback due to compensatory structural errors, where overly weak atmospheric heat flux damping offsets underlying dynamical biases.¹⁴ By accurately simulating the initial coastal cooling via the Southern Ocean teleconnection, and possessing sufficient spatial resolution to capture the tight coupling between wind stress and coastal upwelling, the high-resolution model correctly triggers the Bjerknes feedback, faithfully reproducing the observed multidecadal cooling pattern across the entire Pacific basin.²

The chronological sequence of physical mechanisms required to reproduce the Pacific puzzle is summarized in the following table.

Sequential Step	Geographic Region	Primary Physical Mechanism	Effect on Local Climate
1. Heat Export	Southern Ocean (Pacific Sector)	Explicit mesoscale eddies transport heat laterally across ACC fronts.	Pronounced surface cooling despite increased atmospheric heat flux.
2. Advection	Southeastern Pacific	High-pressure anomalies strengthen southeasterly winds.	Cooling signal is carried equatorward via wind-evaporation-SST feedback.
3. Amplification	Peruvian Coast	Andes block warm air; low inversion triggers stratocumulus clouds.	Massive reflection of shortwave radiation severely drops ocean temperatures.
4. Propagation	Equatorial Pacific	Enhanced zonal gradient triggers the coupled Bjerknes feedback.	Trade winds strengthen, thermocline shoals, creating a westward triangular cold patch.

Evaluating Model Performance Against Global Observations

The success of the kilometer-scale modeling framework is best understood when directly contrasted against observational data and the widespread failures of legacy models. While single-member ensemble runs cannot fully isolate internal variability from forced response using the statistical averaging techniques employed by large multimodel ensembles, the explicit, first-principles resolution of physics provides profound analytical deductions that coarse-grid statistical averaging simply cannot achieve.²

The inability of standard CMIP-class models to reproduce the observed cooling trends in the Southern Ocean and the tropical Pacific leads directly to an overestimation of historical global warming.² Because coarse models simulate warming in regions that are actively acting as massive heat sinks, their global energy budgets become systematically skewed. In contrast, global land surface temperature trends, which are less dependent on complex fluid eddy dynamics, are generally well-captured across most model types, including both standard and high-resolution configurations.²

When evaluating the explicitly resolved historical simulation, the model yields a global-mean surface temperature increase of 0.56 Kelvin over the specific thirty-five-year study period.² This value closely matches the exact amplitude observed in the European Centre for Medium-Range Weather Forecasts Reanalysis version 5 dataset.² The close agreement between the single-member simulation and the observational record is particularly notable given the exceedingly low likelihood of standard CMIP models reproducing the observed magnitude of regional sea surface temperature trends.² This alignment verifies that correctly simulating regional oceanic biases through high-resolution physics is absolutely crucial for stabilizing global energy budget calculations and producing accurate historical climate baselines.²

Implications for Future Climate Projections and Adaptation

Solving the historical Pacific puzzle is a monumental scientific achievement, yet it immediately raises critical, pressing questions regarding the future trajectory of the Earth's climate system. The primary unknown facing climate dynamicists is whether this Southern Ocean-driven, La Nina-like cooling trend is a permanent feature of the modern anthropogenic climate state, or a temporary, transient response that will eventually reverse as the system equilibrates.⁴

If the physical mechanisms driving the cooling—specifically the eddy-driven high-latitude heat export and the robust stratocumulus cloud feedback—are fundamental, long-term thermodynamic adjustments to greenhouse gas forcing, the eastern Pacific may remain anomalously cool relative to the global average for decades to come. This persistent La Nina-like state would continually alter global weather patterns, potentially increasing the probability of severe, multidecadal droughts in regions such as the western United States, fundamentally altering the global distribution of agricultural rainfall, and shifting the frequency of extreme weather events across the tropics.⁹

Conversely, some leading researchers anticipate a future scenario where these temporary buffering mechanisms are eventually overwhelmed by the continuous, unrelenting accumulation of greenhouse gases in the atmosphere. If the transient ocean thermostat breaks down, or if shifting large-scale wind patterns eventually disrupt the delicate stratocumulus cloud decks off the coast of Peru, the eastern equatorial Pacific could experience an accelerated, intensified warming phase as the system seeks to re-establish a new equilibrium.⁴ A sudden reversal of the current cooling trend in the East Pacific over the next few decades could lead to a rapid spike in global-mean temperatures, unleashing a period of accelerated catch-up warming with profound, potentially devastating socioeconomic and environmental consequences.⁴⁹

The persistent unreliability of standard, coarse-resolution models in simulating the historical period has cast a long shadow over near-term regional climate projections, eroding confidence in local adaptation strategies.² By definitively establishing that kilometer-scale resolution is necessary to accurately capture the complex extratropical-to-tropical teleconnections that regulate this vast oceanic system, the high-resolution simulation provides a clear scientific pathway to restore confidence in these critical projections.²

The next logical step for the international climate research community is to deploy these highly resolved, exascale-ready models to forecast future decades. Researchers must rigorously analyze whether the physical drivers of the historical cooling—the ocean eddies and the cloud feedbacks—are sustained under higher emission scenarios, or if they degrade.¹ The assessment and reduction of the uncertainties surrounding forced precipitation and temperature changes in the tropical Pacific remain paramount for designing effective global climate adaptation and mitigation strategies.⁷

Conclusion

The Pacific puzzle has long served as a glaring, uncomfortable reminder of the structural limitations inherent in global climate modeling. For decades, the inability of standard, coarse-resolution models to replicate the distinct multidecadal cooling of the eastern tropical Pacific and the Southern Ocean sowed deep uncertainty regarding the accuracy of regional and near-term climate predictions. The recent success of the kilometer-scale Earth System Model provides a comprehensive, physics-based resolution to this enigma, fundamentally shifting the paradigm of computational climate science.

The research unequivocally demonstrates that the global climate system is highly sensitive to localized, small-scale physical processes that cannot be adequately parameterized using bulk statistical approximations. By explicitly resolving ocean mesoscale eddies across the massive fronts of the Antarctic Circumpolar Current, the high-resolution model accurately captures the dynamic, lateral export of heat that creates a high-latitude oceanic cooling sink. Furthermore, the precise preservation of steep coastal topography along the Andes Mountains facilitates the accurate simulation of strong marine temperature inversions, which in turn generate robust stratocumulus cloud feedbacks. These highly reflective coastal cloud decks massively amplify the advected cooling signal, which is then broadcast across the entire equatorial Pacific basin via the intricate, coupled ocean-atmosphere dynamics of the Bjerknes feedback loop.

This scientific breakthrough underscores the absolute, indispensable value of massive exascale supercomputing initiatives and collaborative, international software refactoring projects. Moving forward, the climate science community must increasingly rely on these storm-resolving and eddy-rich models to untangle the profound complexities of interhemispheric teleconnections. While it remains to be seen whether the eastern Pacific will continue to act as a mitigating heat sink or eventually succumb to accelerated warming, the continued development of kilometer-scale coupled models ensures that future climate projections will be firmly grounded in explicit physical reality rather than statistical approximation.

Works cited

1. New Generation of Climate Models Sheds First Light on Long-Standing Pacific Puzzle, accessed February 21, 2026, https://mpimet.mpg.de/en/communication/detail-view-news-homepage/new-generation-of-climate-models-sheds-first-light-on-long-standing-pacific-puzzle

2. Km-scale coupled simulation and model–observation SST trend discrepancy - PNAS, accessed February 21, 2026, https://www.pnas.org/doi/10.1073/pnas.2522161123

3. accessed December 31, 1969, https://phys.org/news/2026-02-generation-climate-pacific-puzzle.html

4. Deciphering the Mechanisms Controlling the Pacemaker of Global Warming, accessed February 21, 2026, https://www.asiaresearchnews.com/content/deciphering-mechanisms-controlling-pacemaker-global-warming

5. Climate change trends in Vietnam: Evidence from 14 provinces and cities - ResearchGate, accessed February 21, 2026, https://www.researchgate.net/publication/394038114_Climate_change_trends_in_Vietnam_Evidence_from_14_provinces_and_cities

6. (PDF) Heat Hide and Seek - ResearchGate, accessed February 21, 2026, https://www.researchgate.net/publication/261180826_Heat_Hide_and_Seek

7. Subtropical clouds key to Southern Ocean teleconnections to the tropical Pacific | Request PDF - ResearchGate, accessed February 21, 2026, https://www.researchgate.net/publication/362708385_Subtropical_clouds_key_to_Southern_Ocean_teleconnections_to_the_tropical_Pacific

8. Emergence of the enhanced equatorial Atlantic warming as a fingerprint of global warming - PMC, accessed February 21, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC12865183/

9. The Antarctic ozone hole and the pattern effect on climate sensitivity - ResearchGate, accessed February 21, 2026, https://www.researchgate.net/publication/362855078_The_Antarctic_ozone_hole_and_the_pattern_effect_on_climate_sensitivity

10. Understanding CMIP6 Multi‐Model Ensemble Projected Pacific Warming Pattern Variability, accessed February 21, 2026, https://www.researchgate.net/publication/400460292_Understanding_CMIP6_Multi-Model_Ensemble_Projected_Pacific_Warming_Pattern_Variability

11. (PDF) Two Competing Drivers of the Recent Walker Circulation Trend - ResearchGate, accessed February 21, 2026, https://www.researchgate.net/publication/376045141_Two_Competing_Drivers_of_the_Recent_Walker_Circulation_Trend

12. Global impacts of recent Southern Ocean cooling - PNAS, accessed February 21, 2026, https://www.pnas.org/doi/10.1073/pnas.2300881120

13. Fast and Slow Responses of the Tropical Pacific to Radiative Forcing in Northern High Latitudes in - AMS Journals, accessed February 21, 2026, https://journals.ametsoc.org/view/journals/clim/36/16/JCLI-D-22-0622.1.xml

14. The ICON-based Earth System Model for climate predictions and projections (ICON XPP v1.0) - GMD, accessed February 21, 2026, https://gmd.copernicus.org/articles/18/9385/2025/

15. Two-Way Teleconnections between the Southern Ocean and the Tropical Pacific via a Dynamic Feedback in - AMS Journals, accessed February 21, 2026, https://journals.ametsoc.org/view/journals/clim/35/19/JCLI-D-22-0080.1.xml

16. Changes in climate extremes in small Pacific Islands from km-scale model simulations - WebThesis, accessed February 21, 2026, https://webthesis.biblio.polito.it/37103/1/tesi.pdf

17. Computing the Full Earth System at 1km Resolution - MPG.PuRe, accessed February 21, 2026, https://pure.mpg.de/rest/items/item_3686525_2/component/file_3686533/content

18. ICON: Towards vertically integrated model configurations for numerical weather prediction, climate predictions and projections - AMS Journals, accessed February 21, 2026, https://journals.ametsoc.org/view/journals/bams/aop/BAMS-D-24-0042.1/BAMS-D-24-0042.1.pdf

19. The fast development of DestinE's Climate Change Adaptation Digital Twin, accessed February 21, 2026, https://destine.ecmwf.int/news/the-fast-development-of-destines-climate-change-adaptation-digital-twin/

20. 1 nextGEMS: 30-year scenario simulations with IFS-FESOM and ICON at km-scale - DKRZ, accessed February 21, 2026, https://luv.dkrz.de/media/2025/report_1235.pdf

21. nextGEMS simulations overview — easy.gems documentation, accessed February 21, 2026, https://easy.gems.dkrz.de/DYAMOND/NextGEMS/simulation_overview.html

22. The Ocean Component of the ICON Earth System Model—Global Simulation Characteristics and Local Telescoping Capabilit - MPG.PuRe, accessed February 21, 2026, https://pure.mpg.de/rest/items/item_3407144_13/component/file_3432686/content?download=true

23. Report on the joint project ”WarmWorld” bk1341 - DKRZ, accessed February 21, 2026, https://luv.dkrz.de/media/2025/report_1341.pdf

24. ICON-Sapphire: simulating the components of the Earth system and their interactions at kilometer and subkilometer scales - GMD, accessed February 21, 2026, https://gmd.copernicus.org/articles/16/779/2023/

25. Max Planck Institute for Meteorology: WarmWorld, accessed February 21, 2026, https://mpimet.mpg.de/en/research/modeling/warmworld

26. The Destination Earth digital twin for climate change adaptation - EGUsphere, accessed February 21, 2026, https://egusphere.copernicus.org/preprints/2025/egusphere-2025-2198/egusphere-2025-2198.pdf

27. WarmWorld: Exascale EarthSystem Models to anticipate changes in a Warmer World, accessed February 21, 2026, https://www.dkrz.de/en/communication/news-archive/warmworld_start

28. Module Overview - WarmWorld, accessed February 21, 2026, https://warmworld.de/modules/

29. Report on the joint project ”WarmWorld” bk1341 - DKRZ, accessed February 21, 2026, https://luv.dkrz.de/media/2026/report_1341.pdf

30. Faster - WarmWorld, accessed February 21, 2026, https://www.warmworld.de/faster/

31. Earth System Data Exploration - Forschungszentrum Jülich, accessed February 21, 2026, https://www.fz-juelich.de/en/ias/jsc/about-us/structure/research-groups/esde

32. A highly-efficient automated optimization approach for kilometer- level resolution Earth system models on heterogeneous many-cor - EGUsphere, accessed February 21, 2026, https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5297/egusphere-2025-5297.pdf

33. Kilometer-Level Coupled Modeling Using 40 Million Cores: An Eight-Year Journey of Model Development - arXiv, accessed February 21, 2026, https://arxiv.org/html/2404.10253v1

34. Faster Module Meeting: 3 years of progress and the road ahead, accessed February 21, 2026, https://www.dkrz.de/en/communication/news-archive/faster-module-meeting2025

35. Faster Module Meeting 2025: Taking stock of achievements and paving plans - WarmWorld, accessed February 21, 2026, https://warmworld.de/news/faster_2025_dwd/

36. Achievements in the previous allocation period - DKRZ, accessed February 21, 2026, https://luv.dkrz.de/media/2026/report_1344.pdf

37. WarmWorld project at full speed - Max-Planck-Institut für Meteorologie, accessed February 21, 2026, https://mpimet.mpg.de/en/communication/news/das-warmworld-projekt-laeuft-auf-hochtouren

38. Impact of host climate model on contrail cirrus effective radiative forcing estimates - ACP, accessed February 21, 2026, https://acp.copernicus.org/articles/25/473/2025/acp-25-473-2025-relations.html

39. ENSO as an Integrating Concept in Earth Science, accessed February 21, 2026, https://www.whoi.edu/cms/files/mcphaden06sci_285704.pdf

40. Simulated and projected SST of Asian marginal seas based on CMIP6 models - Frontiers, accessed February 21, 2026, https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2023.1178974/full

41. Over-projected Pacific warming and extreme El Niño frequency due to CMIP5 common biases - Oxford Academic, accessed February 21, 2026, https://academic.oup.com/nsr/article/8/10/nwab056/6212231

42. Time Scales and Mechanisms for the Tropical Pacific Response to Global Warming: A Tug of War between the Ocean Thermostat and Weaker Walker in - AMS Journals, accessed February 21, 2026, https://journals.ametsoc.org/view/journals/clim/33/14/JCLI-D-19-0690.1.xml

43. A climate change signal in the tropical Pacific emerges from decadal variability - PMC - NIH, accessed February 21, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC11437070/

44. Influence of Ocean Background State in Atlantic Niño Diversity in - AMS Journals, accessed February 21, 2026, https://journals.ametsoc.org/view/journals/clim/39/1/JCLI-D-25-0170.1.xml

45. Climate Change, Variability and Prediction: Recent Publications - Geophysical Fluid Dynamics Laboratory, accessed February 21, 2026, https://www.gfdl.noaa.gov/climate-dynamics-and-prediction-publications/

46. Wang, C., and J. Picaut, Understanding ENSO physics - A review, in Earth's Climate: The Ocean, accessed February 21, 2026, https://www.aoml.noaa.gov/phod/docs/2004_Wang_Picaut.pdf

47. Subtropical clouds key to Southern Ocean teleconnections to the tropical Pacific - PMC - NIH, accessed February 21, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC9407304/

48. Subtropical clouds key to Southern Ocean teleconnections to the tropical Pacific | PNAS, accessed February 21, 2026, https://www.pnas.org/doi/10.1073/pnas.2200514119

49. A fluctuation-dissipation theorem perspective on radiative responses to temperature perturbations - ResearchGate, accessed February 21, 2026, https://www.researchgate.net/publication/383307186_A_fluctuation-dissipation_theorem_perspective_on_radiative_responses_to_temperature_perturbations

50. Quantifying the time of emergence of the anthropogenic signal in the global land carbon sink, accessed February 21, 2026, https://bg.copernicus.org/articles/23/767/2026/bg-23-767-2026-relations.html