Severe 21st-century ocean acidification in Antarctic Marine Protected Areas

Ocean acidification (OA) driven by anthropogenic carbon uptake lowers seawater pH and carbonate mineral saturation states, threatening marine biodiversity. The Southern Ocean is especially susceptible due to cold, CO2-rich waters, upwelling of low-pH deep waters, and a high Revelle factor. Coarse-resolution Earth system models have projected emergence of seasonal aragonite undersaturation in open Southern Ocean surface waters by mid-century, but have struggled to resolve Antarctic continental shelf processes and ice-shelf cavity dynamics, leaving OA projections for MPA-relevant shelf waters uncertain. Antarctic MPAs (adopted: South Orkney Islands Southern Shelf, Ross Sea region; proposed: Weddell Sea, East Antarctic, Western Antarctic Peninsula) protect diverse ecosystems spanning primary producers, krill and other mid-trophic species, fishes, marine mammals, seabirds, and rich benthic communities. Multiple climate stressors (warming, deoxygenation, sea-ice loss, circulation change) may act additively with OA. This study uses a high-resolution ocean–sea-ice–biogeochemistry model that resolves shelf processes and ice-shelf cavities to project 21st-century OA within the five Antarctic MPAs under four SSP emission scenarios, assessing shelf versus open-ocean contrasts, vertical structure, timing of undersaturation onset, and climate-change feedback contributions.

Prior work has established OA as a global threat to marine calcifiers and ecosystems, with the Southern Ocean identified as a hotspot due to physical-chemical sensitivity. Earth system models indicate that open Southern Ocean surface waters may become seasonally undersaturated with respect to aragonite by the 2030s–2040s, but their coarse resolution limits fidelity on continental shelves and within ice-shelf cavities. Observations and regional models show Antarctic shelves are present-day sinks of anthropogenic carbon due to strong biological uptake in summer and winter sea-ice suppression of outgassing; dense-water formation can propagate shelf OA signals to the abyss. Biological response studies indicate OA can reduce phytoplankton productivity at high pCO2, impair krill recruitment (though adults may be resilient), induce shell dissolution and developmental delays in pteropods, affect fish metabolism and mortality (especially under combined warming), and deform benthic bivalve and sea urchin larvae, with significant life-stage and species-specific variability and knowledge gaps. These findings motivate high-resolution projections in ecologically critical, management-relevant MPA regions.

The study uses the global Finite Element Sea ice Ocean Model (FESOM v1.4) coupled to the Regulated Ecosystem Model (REcoM2), which resolves C, N, Si, Fe, and O2 cycles with two phytoplankton and two zooplankton groups and variable stoichiometry. The mesh is eddy-permitting on Antarctic shelves (<5 km in the south; coarser in the open ocean up to ~150 km) with 99 vertical z-levels and includes a sea-ice component and ice-shelf component with fixed geometry (RTopo-2). Surface atmospheric forcing (3-hourly momentum, heat, freshwater fluxes and sea level pressure) is taken from the AWI Climate Model (AWI-CM) CMIP6 simulations. The historical period (1950–2014) is followed by four SSP scenarios (2015–2100): SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5, with atmospheric CO2 rising to 446, 603, 867, and 1135 ppm by 2100, respectively. Simulations: (i) simA (net fluxes, variable climate and variable CO2) for historical and each SSP; (ii) simB control (constant climate and 1950 CO2) to assess drift; (iii) simC (variable CO2, constant climate) for historical, SSP2-4.5, SSP5-8.5 to isolate climate-change effects (simA - simC); (iv) simD (variable climate, constant 1950 CO2) for historical and SSP5-8.5 to estimate anthropogenic carbon, C_anth = DIC(simA) - DIC(simD). Carbonate chemistry (monthly pH, Ω_arag, Ω_calc) is computed offline using mocsy v2.0 from model fields (DIC, alkalinity, temperature, salinity, nutrients) with surface-referenced potential density for unit conversion. OA is assessed within five MPAs (masks from published coordinates), considering only areas outside ice-shelf cavities and distinguishing continental shelf vs open ocean at the 2000 m isobath. Metrics include top-200 m mean and bottom pH, Ω_arag, Ω_calc, vertical profiles to 2000 m, temporal evolution (1990s vs 2090s decadal means), seasonal onset of aragonite undersaturation, and the fraction of water volume in saturation-state classes. Climate-change feedback contributions to pH changes are quantified as simA - simC. Model drift is characterized (minimal at surface; some subsurface drift) and found small relative to OA signals.

• Anthropogenic carbon accumulates much more strongly on continental shelves than in the open ocean by 2100. Under SSP5-8.5, C_anth in the top 200 m and at the bottom is at least twice as high on shelves; shelf bottom C_anth exceeds 120 μmol kg−1 in many areas vs up to ~70 μmol kg−1 in parts of the open ocean. Enhanced vertical mixing and dense-water formation on shelves transport C_anth rapidly to depth, with shelf waters exceeding 50 μmol kg−1 down to ~750 m by 2100, while such values are largely confined to the top ~200 m offshore.
• pH declines markedly across all MPAs, with projected decreases up to 0.36 (total scale) in the top 200 m by 2100. The vertical pH gradient reverses under higher-emission scenarios: surface waters become more acidic than deeper waters by the 2090s (e.g., gradient reversal up to −0.07 in the East Antarctic MPA under SSP5-8.5). On shelves, downward transfer of the OA signal reduces the contrast between surface and depth compared to open ocean.
• Expressed as hydrogen ion concentration increase (2090s vs 1990s), acidity on continental shelves rises by up to 144% (top 200 m; Weddell Sea) and 130% (bottom; Weddell Sea), exceeding the whole-MPA averages (up to 131% and 80%, respectively).
• The aragonite saturation horizon (Ω_arag = 1) shoals from 351–678 m (1990s) to the surface in all MPAs by 2100 under SSP3-7.0 and SSP5-8.5. Under SSP2-4.5 and SSP1-2.6 it rises to ~40–70 m and 80–253 m, respectively.
• End-of-century undersaturation prevalence: For the three highest-emission scenarios, virtually all (>95%) waters above 2000 m in the MPAs are undersaturated with respect to aragonite. Calcite undersaturation emerges widely under SSP5-8.5: 17–63% of waters across MPAs and 60–75% of shelf waters become Ω_calc < 1. Under SSP1-2.6, the upper ocean generally remains aragonite-supersaturated, though extensive undersaturation still occurs in deeper/shelf waters in some MPAs.
• Timing of surface aragonite undersaturation (Ω_arag < 1) is scenario-driven and relatively consistent across MPAs. Under SSP5-8.5, onset at the surface occurs around 2052–2060 in winter (JJA), 2066–2069 annually, and 2076–2085 in summer (DJF). Under SSP3-7.0, annual mean surface undersaturation begins ~2074–2083. Under SSP2-4.5, wintertime surface undersaturation occurs before 2100 in several MPAs; SSP1-2.6 maintains surface supersaturation through century’s end.
• Seasonal progression and adaptation windows: In the upper 50 m, several decades separate 50% and 90% undersaturated volume thresholds, with 20–28 years under SSP5-8.5 and 28–39 years under SSP3-7.0, implying longer potential acclimation windows under stronger mitigation.
• Climate-change feedbacks (beyond atmospheric CO2 rise) amplify shelf OA, especially under SSP5-8.5. In the Weddell and Ross MPAs, climate effects account for ~19% and 16% of total top-200 m pH decline and ~20% and 19% at the bottom, respectively. Mechanisms include sea-ice retreat and increased freshwater from ice-shelf basal melt enhancing CO2 uptake and altering circulation. Open-ocean bottom pH often shows smaller declines or slight increases due to reduced AABW formation and water-mass shifts.
• Regional contrasts: The Weddell Sea shows slightly delayed OA onset and retains the largest volume of supersaturated water early on, but by 2100 exhibits OA severity comparable to other MPAs under all but SSP1-2.6. Under SSP5-8.5, all regions are nearly ice-free in summer by 2100, facilitating enhanced air-sea CO2 exchange.
• Compared to CMIP6 Earth system models, upper-ocean OA under SSP5-8.5 is similar, but bottom acidification rates on Antarctic shelves are about twice as large in this high-resolution model (pH declines >0.35 vs <0.2).

The projections directly address the question of how OA will evolve in Antarctic MPA waters when shelf processes are explicitly resolved. Enhanced vertical mixing and dense-water formation on continental shelves drive stronger, more pervasive OA than in adjacent open ocean, reversing traditional surface–deep pH gradients under intermediate-to-high emissions and removing refugia for aragonite-califying organisms. Climate-change feedbacks—particularly sea-ice loss and increased ice-shelf meltwater—substantially intensify shelf OA in high-emission scenarios, with notable contributions to the total pH decline in the Weddell and Ross MPAs. While the Weddell Sea has been considered a potential refuge due to delayed onset and higher initial supersaturation, end-of-century conditions converge toward severe undersaturation similar to other MPAs unless emissions are minimized. The biological implications span many trophic levels: primary producers, krill recruitment, pteropods and other calcifiers, fishes, and especially benthic calcifiers, which may be highly vulnerable due to exposure to low variability environments at depth. OA interacts with concurrent stressors—warming and deoxygenation—likely compounding impacts and disrupting food webs. Management implications are significant: even with MPAs in place, OA pressures will be similar inside and outside protected areas, implying the need for both emission mitigation and broader ecosystem-based management (e.g., reduced fishing pressure) to lessen cumulative stressors and enhance resilience. The results underscore the value of high-resolution modeling of shelf seas and ice-shelf interactions for risk assessment and conservation planning.

High-resolution projections show that Antarctic continental shelf waters within both adopted and proposed MPAs will experience severe OA through the 21st century under intermediate-to-high emission scenarios. By 2100, aragonite undersaturation becomes ubiquitous across MPAs for SSP2-4.5, SSP3-7.0, and SSP5-8.5, with widespread calcite undersaturation under SSP5-8.5; only the lowest-emission pathway (SSP1-2.6) avoids pervasive surface undersaturation. Enhanced vertical mixing and dense-water formation intensify shelf OA relative to the open ocean, while climate-change feedbacks further accelerate pH declines, especially in the Weddell and Ross seas. These findings call for strong greenhouse gas mitigation and parallel management actions—such as expanding and effectively managing MPAs and reducing fishing pressure—to reduce cumulative ecosystem stress. Future research should integrate multi-stressor projections (OA, warming, oxygen, sea-ice change, circulation) to identify spatial-temporal refugia and high-risk areas, quantify biological vulnerabilities and adaptation capacities across life stages, and refine fine-scale shelf and cavity processes to improve risk assessments for Antarctic biodiversity and fisheries management.

• Anthropogenic carbon (C_anth) is diagnosed relative to 1950 (simD uses constant 1950 CO2), not preindustrial; reported C_anth likely underestimates total anthropogenic accumulation relative to observation-based estimates that span the full industrial era.
• Climate-effect attribution (simA − simC) is available only for SSP2-4.5 and SSP5-8.5; simD (for C_anth) only for SSP5-8.5, limiting scenario coverage for some diagnostics.
• Model drift is minimal at the surface but present at depth in the control; although small relative to OA signals, it can affect low-emission, low-signal cases (e.g., Antarctic Peninsula under SSP1-2.6).
• In the East Antarctic open ocean, simulated bottom C_anth in the 1990s is lower than observations, likely due to missing sufficiently dense shelf waters to efficiently ventilate the abyss; observational estimates vary by method, complicating direct comparison.
• Spatial resolution is coarse in parts of the open ocean (≤150 km) and very limited cell count in the South Orkney Islands MPA (22 cells); the SOISS MPA is entirely “open ocean” in model topography, which may limit representativeness.
• Biological impact inferences are constrained by limited experimental studies, species-specific variability, and uncertainty in long-term acclimation/adaptation responses; the study does not explicitly couple ecological dynamics beyond the biogeochemical model components.