Projected amplification of summer marine heatwaves in a warming Northeast Pacific Ocean

Marine heatwaves are increasing in frequency and severity, with the 2019 Northeast Pacific event setting all-season records. Summer marine heatwaves pose particular risks to marine species and ecosystems, with documented impacts on fisheries and regional weather. The study addresses two questions: (i) how anthropogenic climate change influenced the 2019 event (attribution), and (ii) how a similar event would evolve in a warmer climate (projection). The total climate-change signal is decomposed into regional mean warming versus event-specific changes driven by local feedbacks, analogous to frameworks for terrestrial heatwaves where soil-moisture feedbacks can amplify extremes. Traditional probabilistic approaches require large ensembles and face challenges defining event boundaries and subtype clustering. Here, a storyline approach is used by nudging free-tropospheric winds to observations while allowing other components to evolve freely, isolating thermodynamic effects and enabling process-based assessment of feedbacks governing marine heatwaves.

Prior work shows marine heatwaves have become longer, more frequent, and more intense due to anthropogenic warming, with significant ecological and socioeconomic impacts. The 2013–2016 Northeast Pacific heatwaves caused widespread biological mortality and contributed to droughts. Mechanisms implicated in the 2019 event include weakened North Pacific high, reduced winds and mixing, reduced mixed-layer depth (MLD), and decreased low-cloud cover increasing shortwave radiation. CMIP6 models often underestimate MHW intensity, potentially due to coarse ocean resolution. Probabilistic attribution studies link increased occurrence to rising global and regional SSTs but are challenged by event definition and dynamics variability. Storyline methods have been applied to land heatwaves to separate dynamics from thermodynamics; this work extends that approach to ocean extremes to quantify event-specific feedbacks (e.g., MLD shoaling, cloud changes, air-mass advection) relative to regional mean warming.

The study uses the coupled climate model AWI-CM-1-1-MR, comprising ECHAM6.3.04p1 atmosphere (T127L95, ~100 km, 95 levels to ~0.01 hPa) and the FESOM v1.4 ocean (unstructured mesh, ~80 km subtropical Pacific, 30–60 km Northeast Pacific, ~8 km Gulf Stream). Storyline simulations nudge free-tropospheric winds (vorticity and divergence, 700–100 hPa) toward ERA5 with a 24 h relaxation and spectral truncation at zonal wavenumbers ≤20, constraining synoptic to planetary scales while leaving boundary layer and small scales free. Five-member nudged ensembles are run for 2017–2020 using initial states from free-running ensembles to create three background climates: preindustrial (branched 1851), present-day (2017), and +4 °C world (2093; ≈ +4 °C global mean surface air temperature vs PI), each integrated four years with transient forcings. Free-running ensembles (1850–2100; historical+ssp370) provide 10-year climatologies for PI (1850–1859), present-day (2015–2024), and +4 °C (2091–2100). SST anomalies are computed relative to the simulated 1984–2014 climatology, with ERA5 anomalies relative to ERA5 1984–2014. The storyline warming is the difference between summer ensemble means of two climates; regional mean warming is from free-running climatological summer means; event-specific change is storyline minus regional mean warming. Statistical robustness is assessed via non-overlapping 5-member ensembles (equivalent to p ≈ 0.008 by Mann–Whitney). A mixed-layer heat-budget framework quantifies contributions to temperature changes between climates: ΔSSTtotal ≈ (Qnet)/(ρ Cp MLD) × Δt + ε, where Qnet = QSW + QLW + Qlat + Qsens absorbed within MLD and ε is residual ocean dynamics (advection/entrainment). Event-specific budget-driven warming is decomposed into Flux-varying (changes in surface heat fluxes at mean MLD) and MLD-varying (changes due to MLD anomalies at mean fluxes) contributions using a linearized Taylor expansion. Regional analyses focus on the marine heatwave core, a coastal region, and the central North Pacific, as defined in the figures.

- Validation of storyline approach: Present-day storylines reproduce the 2019 event’s spatial horseshoe SST anomaly pattern and temporal evolution with high fidelity (Pearson r = 0.93; p < 0.0001 against ERA5 after seasonal-cycle removal). The model exhibits a warm bias north of the core and a cool bias in the core due to climatology biases. - Attribution (Present-day vs PI): Maximum SSTs in the heatwave core reach 20.4 °C (present) vs 19.0 °C (PI). Storyline summer warming averages 1.4 ± 0.2 °C in the core and coastal region, and 1.7 ± 0.2 °C in the central North Pacific. This exceeds the simulated summer GMSST increase of 1.0 °C (i.e., +40% to +70%). Regional mean warming explains most of the core/coastal increase (≈1.6 ± 0.2 °C, ~+60% over GMSST), while event-specific processes slightly dampen warming in these regions by up to −0.2 ± 0.25 °C. In the central North Pacific, event-specific amplification adds +0.45 ± 0.35 °C beyond a regional mean of 1.25 ± 0.3 °C. - Projection (+4 °C vs present): Summer SSTs during a 2019-like circulation are 2–3 °C warmer regionally, with maximum SST in the core projected at 23.7 °C in late August (+3.3 ± 0.15 °C vs present; +4.7 ± 0.15 °C vs PI). Peak intensity occurs more than a week later, prolonging risk. Area-mean summer warming is 2.9 ± 0.15 °C in the core, coastal region, and central North Pacific—about +52% above the simulated summer GMSST increase of 1.9 °C. Event-specific amplification is modest in the core (+0.2 ± 0.2 °C) but substantial at the periphery: coastal +0.6 ± 0.2 °C; central North Pacific +0.7 ± 0.3 °C—leading to lateral expansion of heatwave conditions. Signals are highly robust (non-overlapping ensembles). - Physical drivers of amplification (+4 °C vs present): Enhanced net surface heat flux into the mixed layer dominates event-specific amplification, primarily via increased shortwave radiation from reduced low-cloud cover, and to a lesser extent reduced evaporative cooling efficiency (latent heat). In the core, Flux-varying contribution ≈ +0.5 °C is partly offset by MLD-varying ≈ −0.3 °C because already shallow MLD limits further shoaling. At the periphery, deeper mean MLD allows larger MLD shoaling, yielding MLD-driven amplification ≈ +0.4 °C in both coastal and central North Pacific. Anomalous northerly winds advect warmer subpolar air in a +4 °C climate, reducing air–sea temperature contrast and surface cooling in the central North Pacific, further enhancing flux-driven warming; anomalous southerlies over the core may limit amplification there. - Process contrast across periods: From PI to present, residual ocean dynamics (advection/entrainment) likely dampen warming in the core/coast and contribute to peripheral amplification, whereas from present to +4 °C, surface forcing (Qnet and MLD) changes dominate the event-specific amplification.

By constraining atmospheric dynamics and isolating thermodynamic changes, the storyline approach demonstrates that anthropogenic warming has already substantially increased SSTs during 2019-like summer marine heatwaves, beyond global-mean ocean warming. In a +4 °C climate, event-specific processes—reduced low-cloud cover increasing shortwave input, enhanced MLD shoaling where mean MLDs are deeper, and warmer air advection from subpolar regions—amplify and expand marine heatwave conditions beyond regional mean warming. The amplification is strongest at the heatwave periphery, implying larger spatial footprints and prolonged seasonal impacts, while core warming shows limited additional amplification due to MLD shoaling saturation and southerly air advection. The differing roles of ocean dynamics (more important PI→present) versus surface forcing (dominant present→+4 °C) clarify how feedbacks will shift with continued warming. These results underscore heightened risks for ecosystems and fisheries and support the utility of storylines to inform adaptation by depicting plausible evolutions of specific extreme events in different climate states.

The study extends nudged, coupled storyline simulations to ocean extremes and shows that 2019-like Northeast Pacific summer marine heatwaves are already intensified by anthropogenic warming and will be further amplified and laterally expanded in a +4 °C climate. Regional mean warming accounts for much of the increase, but event-specific feedbacks—low-cloud reductions, MLD shoaling, and altered air-mass advection—add substantial warming beyond regional means, especially at the periphery. The approach provides a high signal-to-noise, process-based complement to probabilistic attribution. Future work should: (i) apply the method to other basins and seasons; (ii) quantify ocean dynamical contributions with eddy-resolving models; (iii) assess sensitivities to cloud feedback uncertainties; (iv) evaluate changes in the likelihood of relevant circulation regimes; and (v) link physical amplification to ecological and biogeochemical impacts.

- Model resolution: Medium-resolution ocean likely under-represents mesoscale eddies and small-scale processes, leading to underestimated marine heatwave intensities and uncertain ocean dynamical contributions. - Biases: Present-day climatology biases (cool in core, warm north) affect anomaly magnitudes. - Dynamics constrained: Storylines hold large-scale dynamics fixed, not addressing changes in frequency/likelihood of circulation regimes under warming. - Heat budget residual: Ocean dynamics (advection/entrainment) are treated as residual, particularly affecting PI→present interpretation; exact contributions remain uncertain without eddy-resolving analyses. - Cloud feedback uncertainty: Low-cloud feedbacks in some regions (e.g., subtropical trades) may be overestimated in some models, affecting amplification projections. - Saturation effects: Very shallow MLDs can limit further shoaling-driven amplification in cores; nonlinearities may not be fully captured by linearized budget decomposition.