Marine heatwaves in a shifting Southern Ocean induce dynamical changes in primary production

Marine heatwaves (MHWs)—periods when sea surface temperature exceeds a seasonally varying high-percentile threshold for multiple consecutive days—have increased in frequency and intensity alongside long-term ocean warming. Their ecological impacts depend on both the magnitude of SST anomalies and event characteristics (frequency, duration). While MHWs have been widely studied at low and mid-latitudes, higher-latitude Southern Ocean (SO) responses remain less explored despite this region’s critical role in global climate and ocean ventilation. The SO is undergoing complex changes driven by atmospheric and ocean warming, freshwater inputs from increased precipitation and ice-shelf melt, and changing winds, all of which influence sea ice extent, stratification, and the Antarctic Circumpolar Current. This study aims to (1) characterize SO MHW spatiotemporal patterns since 1982, (2) identify their physical drivers, and (3) quantify linkages between MHWs and primary production, including potential causal pathways and regional asymmetries.

Prior work shows globally rising MHW frequency and duration, with drivers including atmospheric blocking, surface heat fluxes, advection, and climate modes such as ENSO. Fixed climatological baselines are commonly used to evaluate ecological impacts of MHWs. In the SO, anthropogenic forcing (greenhouse gases, ozone changes) intensifies the polar vortex and increases subsurface heat content; freshwater inputs from melt alter circulation and biogeochemistry. Stratification linked to warming typically depresses chlorophyll at low-to-mid latitudes, but high-latitude phytoplankton responses to MHWs are underexplored. Recent studies highlight subantarctic phytoplankton changes due to meltwater-driven stratification and suggest zonally asymmetric biological responses tied to regional limiting factors (light, iron) and climate modes (ENSO, SAM, TSA).

Study domain: Southern Ocean and adjacent basins south of 40°S. Time periods: primarily 1982–2021 (SST/SIC, drivers), with biogeochemistry and MLD analyses from 1998–2021 where applicable. Sea ice mask: grid cells with mean sea ice concentration (SIC) > 20% masked for MHW detection. Data sources: (1) ESA CCI & C3S daily Level-4 SST and SIC at 0.05° (regridded to 0.5° for comparison), 1981–2021; (2) ECMWF ERA5 near-surface air temperature (2 m), hourly data aggregated to daily/annual means at 0.5°; (3) GLORYS12V1 reanalysis mixed layer depth (MLD) at 1/12°, regridded to 0.5°, daily/monthly, 1998–2021; (4) Mercator-Ocean global biogeochemistry hindcast (NEMO-PISCES) at 0.25°, monthly surface (upper 30 m) chlorophyll and nutrients (Fe, NO3, Si, PO4), regridded; (5) Carbon-based Production Model (CbPM) satellite NPP (SeaWiFS/MODIS/VIIRS), monthly 1998–2021, regridded to 0.5°. Marine heatwave detection: Applied 95th percentile threshold relative to a fixed climatology (1982–2012), minimum duration ≥5 days, allowing gaps <3 days, and an additional criterion requiring SST above local long-term mean summer temperature (LMST) to avoid wintertime false positives. Daily thresholds smoothed via 31-day moving average with an 11-day centered window. Annual MHW metrics: frequency, mean and maximum SSTA, duration (mean and total), cumulative intensity (°C-days), and areal coverage. Climate indices: ONI (ENSO), TSA, and SAM obtained from standard sources to assess teleconnections. Local driver analysis: Used GFDL ESM2M (MOM4p1) heat budget terms over the upper 10 m from a 500-yr preindustrial control and an eight-member 1982–2021 historical/RCP8.5 ensemble. Computed anomalies of heat budget components and averaged during MHW onset (build-up) and decay (dissipation): advection (horizontal/vertical resolved and parameterized), air–sea heat exchange (shortwave, longwave, latent, sensible), vertical diffusion (local KPP and tidal mixing), non-local convective vertical mixing (KPP), and residual processes. Converted heat flux to temperature tendency using standard constants and dz=10 m. Trend analysis: Linear trends spatially via OLS (with significance), circumpolar averages via Theil–Sen robust estimator; biogeochemical and NPP trends via modified Mann–Kendall accounting for autocorrelation (α=0.05). Causal inference: Empirical Dynamic Modelling with Convergent Cross Mapping (CCM) on monthly gridded time series (1998–2021) for Max SSTA, NPP, SIC, and MLD. Selected delay τ=3 months via mutual information and embedding dimension E=6 via manifold unfolding tests. Evaluated cross-map skill (Pearson’s ρ) and directionality among variables. Case studies: Time series analyses (2015–2018) in Davis Sea and Amundsen–Bellingshausen Sea to examine compound MHW and high NPP (HNPP, NPP > 95th percentile) events and co-evolution with N-SAT, SST, SIC, and MLD.

- Spatial patterns: Maximum MHW SST anomalies (SSTA) ranged from ~0.6–11 °C south of 40°S, with elevated values at lower latitudes of the Pacific, Atlantic, and Indian sectors, plus Drake Passage, Amundsen–Ross, Davis Sea, and south Indian sector (often 4–6 °C). Frequency hotspots (2–3 events yr−1) include the Davis Sea, Tasman Sea surroundings, New Zealand region, and the southwestern Atlantic. The Pacific sector shows longer events (average ~12 days, with local means 20–30 days). - Trends and temporal evolution: Circumpolarly averaged metrics decreased until ~1994/1995, then rose to a peak around 2015 and remained elevated thereafter. Maximum SSTA averaged 3.19 °C (1982–2015) vs 3.7 °C (2015–2021). Frequency increased from ~1 event yr−1 (lasting ~11 days) in 1982–2015 to ~2 events yr−1 (lasting ~15 days) in 2015–2021, raising total annual MHW days from ~16 to ~34. Cumulative intensity increased from ~16 to ~28 °C-days (average SO value ~17.73 °C-days yr−1). Ocean area south of 40°S experiencing at least one MHW per year increased by ~20% in recent years. Frequency and duration are negatively spatially correlated (r ≈ −0.37). Regional heterogeneity: Atlantic sector exhibits larger magnitudes; Pacific sector hosts longer events; Indian sector more recently frequented by extremes. - Climate modes and asymmetries: In the Pacific, interannual variability and peaks in MHW metrics align with ENSO (notable El Niño and La Niña years). Decoupling between Pacific and Atlantic metrics appears during strong La Niña years, indicating zonally asymmetric relationships with climate modes (ENSO, TSA, SAM). - Physical drivers: Heat budget analysis shows MHW onset primarily driven by air–sea heat fluxes and vertical diffusion, with vertical diffusion emerging as the dominant contributor during 1982–2021 onset and also an important cooling process during decay. Circumpolar heat gain during onset in the ensemble (1982–2021) is ~39.8 W m−2, largely explained by vertical diffusion (~26 W m−2; ~65.2%). Advection follows ACC pathways but plays a more localized/secondary role (potentially underestimated due to model resolution). Non-local convective mixing counteracts warming during onset. Strongest and weakest events share similar dominant drivers. - Link to atmosphere/sea ice: Nov–Mar SSTAs are significantly positively correlated with near-surface air temperature. Sea ice increased until 2015, reflecting more solar radiation and cooler SST; after 2015, rapid SIC loss correlates with warmer SST, enhanced MHW activity, and regional wind-driven redistribution of ice. - Biogeochemical responses: Following the post-2015 SIC drop and peak MHW activity in subantarctic regions, CbPM NPP increased. Across Nov–Mar months (1998–2021), mean NPP is higher during MHWs by ~100 mg C m−2 d−1; the minimum NPP increases from ~1 to ~74 mg C m−2 d−1 under MHW conditions. MHW periods show lower SIC and shallower MLD, retaining meltwater-derived iron and nutrients in the euphotic layer. Modeled Fe and macronutrients show post-2015 trend inversions indicating enhanced biological consumption. - Causality: CCM reveals nonlinear causal couplings between physical variables and NPP. Cross-map skill for NPP from Max SSTA is moderate and spatially heterogeneous (stronger in southern subantarctic band), while MLD shows the strongest and most homogeneous predictive skill for NPP, underscoring the central role of mixed-layer shoaling. SIC also exhibits causal influence in the southernmost subantarctic regions. The variables are interlinked, with bidirectional coupling between temperature extremes and MLD, and SSTA strongly influencing SIC. - Case studies and compound events: In the Davis Sea (2016/17) and Amundsen–Bellingshausen Sea (late 2016–mid 2017), sustained warm air and ocean conditions, low SIC, and shallower MLD coincide with compound MHW and high NPP (HNPP) events, especially in late summer after seasonal minima of MLD and SIC. - Regional asymmetry: North of the Antarctic Polar Front, primary producer responses are zonally asymmetric, reflecting regional differences in limiting factors and dynamics (e.g., ACC pathways).

The study demonstrates that Southern Ocean MHWs have intensified in frequency, duration, and cumulative intensity, particularly after the abrupt sea ice decline around 2015. Heat budget analysis identifies air–sea fluxes and especially vertical diffusion as dominant local drivers of onset-phase warming, while convective mixing damps extremes. These physical changes are tightly linked to upper-ocean stratification and sea-ice dynamics, producing shallower mixed layers that, in nutrient-replete high-latitude environments and with enhanced iron supply from meltwater, favor higher net primary production. Empirical dynamic modeling confirms nonlinear causal connections from temperature extremes, sea-ice changes, and mixed-layer depth to NPP, with MLD emerging as the strongest proximate control. Regional and zonal asymmetries, influenced by climate modes such as ENSO, SAM, and TSA, modulate MHW characteristics and biological responses, with hotspots like the Davis Sea and longer events in the Pacific sector. Together, these findings address the core research question by showing that increasing MHW activity can promote biological carbon assimilation via stratification-driven enhancement of phytoplankton productivity in parts of the SO, thereby potentially influencing the biological carbon pump. However, responses are not uniform; in highly dynamic ACC regions and depending on local limitation (light, iron, grazing), the magnitude and even direction of biological change can differ, and phenological shifts may create trophic mismatches.

This work provides a circumpolar assessment of Southern Ocean MHWs (1982–2021), identifies their principal local drivers (air–sea heat fluxes and vertical diffusion), and empirically links them to enhanced primary production through mixed-layer shoaling and sea-ice-mediated nutrient supply, particularly in the southern subantarctic band. MHW metrics have increased markedly since 2015, with broader spatial extent and stronger cumulative intensity. Causal analysis highlights MLD as the dominant physical control of NPP variability, with additional contributions from temperature extremes and SIC. These insights imply that short-term thermal extremes can modulate SO carbon uptake via impacts on the biological carbon pump. Future research should (1) extend records to better separate multi-decadal internal variability from anthropogenic trends, (2) quantify long-term interactions between MHWs, ecosystem structure (including phytoplankton functional types and grazers), and CO2 fluxes, (3) improve resolution of ocean models to capture mesoscale advection and vertical exchanges, and (4) refine satellite-derived productivity estimates and reconcile model–observation differences in the SO.

- Attribution limits: The 1982–2021 satellite era may be too short to unambiguously separate multi-decadal internal variability from anthropogenic trends in SO MHWs. - Model resolution: MOM4p1’s relatively coarse horizontal grid likely underestimates mesoscale advection impacts on local MHW development and decay. - Data masks and sea-ice influence: MHW detection is less reliable where SIC is high; areas with mean SIC > 20% were masked, potentially omitting relevant processes. - Productivity estimates: CbPM-derived NPP carries model uncertainties specific to the SO; different satellite NPP models yield varying magnitudes. - Biogeochemical forcing: The biogeochemistry hindcast lacks data assimilation and may not capture all nutrient/iron dynamics. - Ecological complexity: Phytoplankton responses are species- and region-specific, influenced by grazing and competition; results cannot be generalized as a single uniform response. - Causality inference: While CCM reveals nonlinear couplings, bidirectionality and confounding co-drivers (e.g., winds) limit straightforward mechanistic attribution.