The ocean losing its breath under the heatwaves

Ocean warming and deoxygenation, largely driven by anthropogenic climate change, are accelerating. Beyond long-term trends, extreme manifestations—marine heatwaves (MHWs) and low-oxygen extreme (LOE) events—pose acute risks to marine ecosystems and dependent human communities. While deoxygenation processes and MHWs have been studied separately, the global-scale coupling between MHWs and LOEs, especially their co-occurrence in the surface layer where marine life and fisheries are most sensitive, is not well quantified. This study asks: How frequently and where do MHWs and LOEs co-occur, how has this changed in recent decades, and what roles do anthropogenic forcing and internal climate variability play in driving these compound events? The work aims to establish a global assessment framework and clarify mechanisms and drivers behind the evolving compound low-oxygen-and-heatwave (CLH) risks.

Prior work documents long-term ocean warming and oxygen loss, with mechanisms including reduced solubility, enhanced biological consumption, stronger stratification, and weakened ventilation. Marine heatwaves have intensified and lengthened under climate change, causing ecological and socio-economic impacts. Literature on LOE events, though more recent, shows abrupt deoxygenation can push organisms to critical oxygen limits sooner than expected from gradual trends, with drivers including anomalous water mass distributions and mesoscale eddies. Case studies indicate MHWs can catalyze LOEs (e.g., the North Pacific “Blob,” Baltic Sea seabed heatwaves, SW Atlantic events), and multi-stressor exposures (high temperature and low oxygen) can synergistically impact marine organisms and fisheries. However, a comprehensive global assessment of MHW–LOE coupling, its evolution, and its attribution to anthropogenic forcing versus internal variability has been lacking.

• Data and study domain: Focus on the ocean surface layer (0–10 m). Combined in situ observations and climate model simulations.
• Observations: World Ocean Database 2018 (WOD18; ~1.3 million profiles since 1960) for temperature and dissolved oxygen; gridded at 2.5°×2.5°, aggregated into nine basins with monthly means. World Ocean Atlas 2018 (WOA18) used for climatological evaluation. Additional datasets: OISST (0.25°, daily since 1981) and GOBAI-O2 (1°, monthly, 2004–2022). Observations were harmonized to consistent resolutions and overlapping periods for comparisons.
• Climate model: CESM2.1.3 (POP2 ocean, MARBL biogeochemistry), nominal 1° horizontal grid, 60 levels. Historical forcing per CMIP6 (1850–2014); analysis limited to 1960–2014.
• Event definitions: Seasonally varying thresholds per grid or basin computed over 1960–2014. MHW: temperature > 90th percentile; LOE: dissolved oxygen < 10th percentile. For daily model outputs, minimum duration 5 days; events separated by ≤2 days merged. Compound low-oxygen extreme and heatwave (CLH): simultaneous exceedance of both thresholds with minimum duration 2 days (daily) or corresponding monthly identification for observations. Annual statistics include frequency, mean duration, and total event days.
• Joint distribution classification: Annual event days for each extreme binned into Rare, Occasional, Frequent, Common using the 40th, 70th, and 90th percentiles across oceans to visualize joint occurrence changes.
• Likelihood Multiplication Factor (LMF): LMF = P(CL H) / [P(MHW)×P(LOE)], computed per grid (model) and per basin (model and observations) to quantify dependence relative to independence.
• Separation of drivers: Ensemble Empirical Mode Decomposition (EEMD) applied to temperature and oxygen time series in each grid (noise amplitude 0.2σ, ensemble=400, 6 IMFs). IMFs 1–5 represent internal variability; IMF 6 represents long-term anthropogenic signal. Extreme detection applied separately to reconstructed anthropogenic and variability components to quantify their contributions to CLH changes.
• Attribution to climate modes: Following Holbrook et al., assessed associations between CLH occurrence and large-scale climate indices (Niño-3.4, EMI, DMI, NAO, PDO, TPI, Atlantic Niño, SAM, NPGO). Monte Carlo AR(4)-based surrogate testing (10,000 realizations) determined significance and dominant phases.
• Model evaluation: Compared simulated climatologies, trends, and temporal evolution of MHW, LOE, and CLH against WOD/WOA, OISST, and GOBAI-O2. Reported biases, RMSE, pattern correlations, and trend consistency.

• Strong global coupling between MHWs and LOEs:
  • Observations show globally averaged LMF ≈ 4.0; model shows ≈ 5.1, indicating 4–5× higher co-occurrence than independence.
  • LMF > 1 over >90% of the ocean; strongest (>5) in subtropical mid-latitude Pacific and Atlantic; LMF < 1 in equatorial East Pacific and parts of the Indian Ocean.
  • Temperature–oxygen anomalies are significantly negatively correlated in most basins; detrended correlations remain negative.
• Coherent spatial patterns of extremes:
  • LOE frequency and duration patterns resemble those of MHWs (pattern correlation > 0.8, p < 0.01).
• Rapid intensification and expansion of compound events:
  • Global annual CLH days increased with a trend of +0.96 days per year (p < 0.01), rising by ~25 additional days per year in the 2010s relative to a baseline of ~14 days in the 1980s.
  • Increases in annual CLH days occur across >70% of the global ocean; maximal regional increases up to ~100 days, with decreases in parts of the eastern tropical Pacific and high-latitude Southern Ocean.
  • Extreme event time under climate change increases by ~0.07–0.09 months per year since the 1960s (observations and model).
• Shift in co-occurrence categories:
  • Regions with ‘occasional’ co-occurrence of both MHW and LOE expanded from ~10% (late 20th century) to ~24.5% (early 21st century).
  • Areas with ‘rare’ co-occurrence declined from ~69.3% to ~25% between the two periods (1985–1999 vs. 2000–2014).
• Rising fraction of LOE days accompanied by MHW:
  • 41.0% (1970–1984), 45.5% (1985–1999), and >50% (2000–2014) of LOE days co-occurred with MHWs globally; similar strengthening across most basins (except East Pacific).
• Elevated risk in high-biomass and key fishing regions:
  • CLH increases are larger where fish biomass is high. Where biomass >50 g m−2, CLH days trend ≈ +1.47 days per year (~50% above global average).
  • Four fishing zones show strong increases in annual CLH days: North Sea (+3.64 d/yr), Hokkaido (+2.95 d/yr), East China Sea (+2.66 d/yr), Newfoundland (+3.34 d/yr).
  • Only 8% of regions with biomass <5 g m−2 exceed the global-average CLH increase, versus 66% where biomass >200 g m−2.
  • Post-2000 CLH events exhibit stronger anomalies: temperature +13%, oxygen −11% compared to pre-2000 events.
• Drivers and mechanisms:
  • Anthropogenic forcing (long-term warming and oxygen decline) is the primary contributor to global increases in CLH days; internal variability shapes spatial patterns.
  • ENSO modulates CLH in the East Equatorial Pacific and NW Indian Ocean; SAM south of 40°S; IOD around Indonesia; PDO contributes in mid-latitudes; NAO influence in the North Atlantic (30–50°N) is comparatively minor.
  • Mechanistically, heatwaves reduce oxygen solubility (enhancing ocean-to-atmosphere O2 flux), and enhanced stratification (mixed-layer shallowing) inhibits ventilation, compounding LOEs during MHWs.

The study demonstrates that marine heatwaves and low-oxygen extremes are not independent hazards; rather, they frequently co-occur, with dependence strong enough to multiply joint risk several-fold relative to chance. This coupling has intensified in recent decades, particularly in mid-latitude basins and high-biomass regions, elevating compound stress on ecosystems and fisheries. The findings directly address the research questions by quantifying the magnitude, spatial distribution, and temporal evolution of CLH events, and by attributing their changes to anthropogenic forcing versus internal variability. Anthropogenic warming is found to dominate the global increase in compound-event days, while internal climate variability (e.g., ENSO, IOD, SAM, PDO) organizes regional patterns and episodic modulation. Mechanistic analyses substantiate that reduced oxygen solubility during heatwaves and enhanced stratification limiting ventilation jointly intensify LOEs when temperatures are anomalously high. These results underscore escalating compound extreme risks for marine ecosystems and dependent communities, informing adaptation strategies, risk assessments, and fisheries management under continued warming.

This work provides the first global, observation- and model-based assessment of the coupling between marine heatwaves and low-oxygen extremes in the surface ocean, defining and quantifying compound CLH events. It establishes that co-occurrence has become more frequent and persistent, with hotspots in mid-latitudes and high-biomass regions, and that anthropogenic forcing is the primary driver of the global increase, while internal variability shapes regional expressions. The study advances understanding of compound ocean extremes and highlights their growing importance for ecosystems and fisheries. Future research should employ higher-resolution, daily-resolved models to better resolve coastal and eddy processes, conduct detailed attribution to disentangle contributions from specific forcings (e.g., aerosols, volcanic eruptions), and integrate ecological response models to assess biological impacts and inform adaptation and management in vulnerable regions.

• Model resolution and coastal biases: The CESM (∼1°) under-resolves boundary currents, eddies, and coastal dynamics, likely underestimating coastal MHWs and thus CLH severity; actual coastal impacts may be stronger.
• Observational constraints: Surface-focused (0–10 m) analysis with WOD18 monthly aggregation and incomplete spatiotemporal coverage; mapping/averaging affected by missing values. GOBAI-O2 is monthly and limited to 2004–2022; conversion from months to days assumes 30 days per month.
• Attribution method scope: EEMD separates long-term and oscillatory components but does not isolate specific forcings (e.g., aerosols, volcanism); further attribution is needed.
• Model biases: Documented temperature and oxygen mean-state and trend biases regionally (e.g., temperature biases in North/South Pacific; O2 overestimation extra-equatorially), potentially affecting regional CLH metrics, though large-scale patterns and trends are robust.
• Basin-level observational LMF and CLH metrics are at coarser temporal resolution than model (monthly vs daily), which may smooth event characteristics.