Vertical structures of marine heatwaves

Marine heatwaves (MHWs)—periods of anomalously high ocean temperatures—have become more frequent and longer in duration and can cause widespread ecological and socio-economic impacts. Prior research has primarily focused on surface expressions because of satellite SST availability, yet observations show MHW signals can extend into deeper layers. The vertical structure of MHWs at the global scale remains poorly characterized. Existing regional studies (e.g., East Australian Current) suggest events with differing penetration depths and vertical shapes (shallow, intermediate, deep), with varying dominance of air-sea fluxes versus ocean advection. Notably, some events exhibit subsurface cooling beneath surface warming, while others have maximum anomalies below the surface, indicating diverse underlying processes. The purpose of this study is to identify and classify typical vertical structures of global MHWs, quantify their penetration depths, map their spatial distributions, examine regional drivers (e.g., atmospheric forcing, planetary waves, currents, eddies, mixing), and assess temporal trends over the Argo era. Understanding vertical structures is crucial for evaluating physical drivers, heat content changes, and ecosystem impacts that extend beyond the surface layer.

The paper synthesizes findings that MHW frequency and duration have increased and are projected to rise with continued warming. Documented impacts include coral bleaching, seagrass and kelp loss, disruptions to biogeochemical cycles and habitats, mass mortalities across trophic levels, fisheries impacts, and cultural service losses. MHWs are linked to large-scale climate modes (ENSO, IOD, NAO, PDO, SAM) via atmospheric/oceanic teleconnections, including Kelvin and Rossby waves. Mechanistic drivers include air-sea heat flux anomalies (shortwave, longwave, latent, sensible), horizontal and vertical advection (geostrophic, Ekman, eddies), and lateral/vertical mixing. Regional studies revealed subsurface extensions of MHWs in western boundary current systems and distinguished shallow, intermediate, and deep events, with deeper events often advection-driven and shallower events flux-driven. However, some shallow/intermediate events show subsurface cooling and not all deeper events have subsurface temperature maxima, motivating a global assessment of vertical structures.

Detection of MHWs: A fixed-baseline definition following Hobday et al. is applied to NOAA OISST V2.1 daily SST (1982–2021), where MHWs are periods of at least 5 consecutive days with SST above the 90th percentile of the 1982–2011 seasonal climatology. High-resolution SST (0.25°) is upscaled to 2° to match subsurface data sparsity. Subsurface data and anomalies: Argo T/S profiles (CARDC) from 2001–2020, sampling from ~−5 to 2000 dbar at ~10-day intervals with ~70 levels, are mapped to a 2° grid and interpolated onto 58 depth levels. Temperature anomalies Ta(z) are computed relative to climatological seasonal profiles from CARS2009. For each MHW, vertical structure is the average Ta(z) across Argo profiles occurring during the MHW intervals. Mixed-layer depth (MLD) is the depth where potential density differs from the 10-dbar value by 0.03 kg m−3. Thermocline depth is the depth of maximum vertical temperature gradient. Mixed-layer heat budget (1993–2020): The mixed-layer temperature tendency is decomposed into surface flux forcing, horizontal advection, and a residual (including vertical entrainment and mixing): ∂T/∂t = (Qs−Qb)/(ρwhCp) − ∇·(uT) + ∇²T + Res. Tendency is from OISST; surface fluxes from NCEP-DOE Reanalysis 2 and MLD from GLORYS12V1; horizontal advection from OSCAR currents combined with OISST (0.25°), then interpolated to 2°; composites are made during MHW development (start to day of maximum intensity) to assess dominant drivers regionally. Vertically cumulative temperature anomaly: CTa(p) = Σ Ta(p)Δp from surface to pressure p; the maximum MCTa is integral down to the level where Ta returns to zero. Impact depth definition and classification: Impact depth (IDMHW) is the depth where CTa reaches 0.85 × MCTa. Classification rules:

• Shallow MHW: IDMHW shallower than min(MLD, 100 dbar), with anomalies decaying with depth and without strong subsurface warming or cooling (subsurface anomalies below 50th percentile thresholds).
• Deep MHW: IDMHW deeper than min(MLD, 100 dbar), with decaying anomalies and without strong subsurface warming/cooling (as above).
• Subsurface-reversed MHW: surface warming with subsurface cooling beneath; subsurface minimum Tamin below its 50th percentile.
• Subsurface-intensified MHW: maximum warming in the subsurface; Tamax − SSTa exceeds its 50th percentile. Eddy influence: Mesoscale eddies (AVISO eddy trajectories) are used to estimate eddy-induced temperature anomalies as differences between anomalies within anticyclonic/cyclonic eddies and the background without eddies, distinguishing surface versus subsurface eddy cores. Auxiliary datasets include CMEMS/DUACS geostrophic velocity and eddy kinetic energy for context.

• Four vertical types identified globally: shallow, subsurface-reversed, subsurface-intensified, and deep MHWs, each with distinct vertical profiles and spatial patterns.
• Spatial distributions: shallow and deep MHWs are more common in mid- to high latitudes; subsurface-reversed and subsurface-intensified types are widespread with hotspots in the tropics. In the tropical Pacific, subsurface-reversed hotspots occur in the central basin, and subsurface-intensified hotspots in the eastern and western basins; in the tropical Indian Ocean, subsurface-intensified dominates (especially west); in the tropical Atlantic, subsurface-intensified is higher in the east while subsurface-reversed is higher in the west.
• Impact depths (global means and regional correlations): • Shallow: 32 ± 16 dbar; depth pattern correlates with MLD (r = 0.56). • Subsurface-reversed: 57 ± 38 dbar; depth pattern correlates with thermocline depth (r = 0.39); deeper in western Pacific/Atlantic, shallower in the east, consistent with tilted thermocline; shallow in the Indian Ocean thermocline dome. • Subsurface-intensified: 348 ± 190 dbar; deeper in subtropical gyres, shallower in tropics; depth correlates with MLD (r = 0.43) and thermocline depth (r = 0.35). • Deep: 316 ± 202 dbar; distribution resembles subsurface-intensified with similarity to MLD (r = 0.54).
• Tropical mechanisms: Thermocline modulation by wind-forced planetary waves (downwelling Kelvin, upwelling Rossby) shapes subsurface-intensified and subsurface-reversed structures. Correlations between depths of maximum subsurface cooling/warming and thermocline depth are stronger in tropics (r = 0.59 and 0.78) than globally (r = 0.46 and 0.55).
• Subtropics: Air-sea heat flux anomalies drive many events (shallower, shorter), but subduction and advection allow penetration to depth, explaining prevalence of deep MHWs (e.g., NE Pacific Blob events, NW Atlantic 2011/12).
• High latitudes: ACC advection and eddies foster deep MHWs; south of ACC, vertical processes (upwelling/mixing) dominate, with subsurface signals and mode-water processes contributing to subsurface-reversed/intensified types.
• Western boundary currents: Enhanced horizontal advection and eddy activity favor deep MHWs (e.g., Tasman Sea 2015/16). Anticyclonic eddies more associated with shallow/deep warming; cyclonic eddies with subsurface-reversed; subsurface anticyclones reinforce subsurface warming, cyclones reinforce subsurface cooling.
• Eastern boundary currents: Coastal Kelvin waves and wind anomalies modulate upwelling and alongshore heat advection; all four types occur with comparable proportions due to complex air-sea coupling. In SE Indian Ocean, Leeuwin Current advection and high EKE favor deep MHWs.
• Trends in area (2001–2020, Argo era): Statistically significant increases in the total ocean area experiencing each type: subsurface-intensified 1.52 × 10^4 m^2 yr^−1 (p < 0.01), shallow 2.13 × 10^4 m^2 yr^−1 (p < 0.01), subsurface-reversed 1.27 × 10^4 m^2 yr^−1, deep 7.38 × 10^3 m^2 yr^−1 (p < 0.01).
• Interannual variability: Global area peaks during strong El Niño events (2009/10, 2015/16); 2019/20 peak linked to extreme positive IOD and subsequent basin-wide Indian Ocean warming; NE Pacific peaks also tied to PDO/NPGO.
• Depth trends: Impact depths of all types show increasing trends over 2001–2020, with mid-high latitude subsurface-intensified and deep MHWs contributing strongly due to increased upper-ocean heat content; in some tropical regions, increased variance of upper-ocean heat content deepens subsurface MHWs.
• Overall, results highlight ocean stratification, large-scale circulation, waves, and eddies as key modulators of vertical MHW structure and penetration.

The study addresses the global knowledge gap on vertical MHW structures by defining four robust types based on Argo-derived subsurface temperature anomalies and a cumulative anomaly-based impact depth. The distinct geographic patterns and correlations with MLD and thermocline depth show that ocean stratification and dynamics (planetary waves, advection by currents, eddies, and mixing) govern how surface heat anomalies penetrate and transform subsurface structures regionally. Tropical teleconnections via Kelvin and Rossby waves are pivotal in setting subsurface maxima or reversals; in subtropics and boundary currents, advection and subduction underpin deep penetration; at high latitudes, ACC and ventilation processes propagate warming to depth. The observed increases in both occurrence area and impact depth across all types over the Argo era are consistent with continued ocean heat uptake under climate change, with superimposed interannual modulation by ENSO, IOD, and PDO. Ecologically, subsurface-intensified events—often with stronger subsurface anomalies than at the surface—imply that surface-only metrics underestimate impacts, particularly on mesopelagic and deeper biota. The findings underscore the need for vertically resolved monitoring and process-based forecasting to anticipate ecosystem risks and to inform fisheries, aquaculture, and conservation management.

This work establishes a global framework for categorizing MHW vertical structures into four types—shallow, subsurface-reversed, subsurface-intensified, and deep—quantifies their impact depths, maps their spatial distributions, and links them to regional physical drivers. It demonstrates significant increases since 2001 in both the area affected and the penetration depths for all types, consistent with ongoing ocean heat uptake, and highlights strong regional modulation by climate modes and ocean dynamics. The results emphasize that subsurface signatures can exceed surface anomalies, necessitating subsurface observations for accurate ecological impact assessments and improved prediction. Future research should: enhance sustained, vertically resolved observations (e.g., expanded Argo, including BGC-Argo and deep Argo); improve representation of boundary currents, mesoscale/submesoscale processes, and air-sea exchanges in high-resolution models; and develop forecasting systems that integrate vertical structure to better predict MHW onset, depth, and duration under continued warming.

• Subsurface observations from Argo have relatively low temporal resolution (~10-day sampling) and uneven spatial coverage, necessitating SST upscaling to 2° and leading to uncertainties in capturing event evolution and regional details.
• The Argo era (2001–2020) is too short to clearly separate long-term trends from decadal-to-multidecadal climate variability (e.g., PDO, AMO).
• Mixed-layer heat budget includes a residual term that aggregates processes (e.g., entrainment, vertical mixing) not explicitly resolved, potentially conflating mechanisms.
• Challenges remain for models to accurately simulate boundary currents, mesoscale/submesoscale dynamics, coastal processes, and air-sea exchanges that influence MHW structure and variability.
• Classification thresholds (e.g., 50th percentiles, 0.85 MCTa depth) and gridding choices may influence typology and derived depths, and dependence on climatologies (e.g., CARS2009) introduces baseline uncertainties.