Drivers of Antarctic sea ice advance

The study addresses why and how the timing of Antarctic sea-ice advance (the onset of the ice season) varies in space and time. Antarctic sea ice undergoes pronounced seasonal cycles with large climatic and ecological impacts. Although satellite records document substantial regional changes in the timing of advance and retreat over the past four decades, interpretation is complicated by strong interannual variability and multiple contributing ocean–atmosphere processes. The authors focus on the fundamental drivers of the date of sea-ice advance and its linkage to the prior date of retreat. Two candidate mechanisms can set advance: local seawater freezing after the mixed layer cools to the freezing point, and import of drifting ice from already frozen regions into open water. Oceanic heat entrainment or advection near the winter ice edge can inhibit freezing. The paper hypothesizes that upper-ocean heat content, surface heat fluxes, sea-ice thermodynamics, and ice drift jointly control advance, and seeks to quantify their relative roles and the mechanistic link between retreat, mixed-layer heat content, and subsequent advance.

Prior work using passive microwave observations identified substantial regional variability and trends in the timing of Antarctic sea-ice advance and retreat, often attributed to wind-driven ice transport and thermodynamic ice–ocean feedbacks. Studies have shown that the winter ice edge is maintained largely by dynamic import rather than local freezing, and that changes in winds and ocean heat supply can modulate the ice edge. Interannual links between retreat and subsequent advance have been reported in both Antarctic and Arctic, suggesting a thermodynamic feedback wherein earlier retreat allows greater summer ocean heating and later freeze-up. However, the complexity of drivers (winds, advection, entrainment, air–sea fluxes) and strong interannual variability have limited process-based understanding, and models exhibit biases in Southern Ocean processes that affect projections of sea-ice seasonality.

Data and diagnostics: The study uses daily passive microwave sea-ice concentration (OSI SAF; OSI-450/430-b) over 1982–2018 to derive climatological dates of retreat (dr) and advance (da), defined as the first day smoothed (15-day filter) SIC falls below or exceeds 15%, respectively. To characterize ocean surface conditions, a daily satellite SST product (ESA CCI L4, 0.05°) from 1982 is used, along with analysis uncertainty. Mixed layer depth (MLD) and stratification come from a gap-filled monthly 1979–2018 in situ climatology combining CTD, Argo, and marine mammal-borne sensors (MEOP). Surface radiative fluxes during the open-water season are from ISCCP FH-series (1982–2016), and robustness is checked with NOAA AVHRR OI SST. All datasets are mapped to the OSI-SAF EASE2 25 km grid. Sea-ice concentration budget decomposition: To separate dynamic (Dy) and thermodynamic (Th) drivers around advance, the authors use a daily sea-ice concentration budget (2003–2010) derived from AMSR-E ice drift and NASA Team SIC. The SIC tendency is decomposed into advective/divergent ice flux divergence (dynamic term) and a residual (treated as thermodynamic; mechanical redistribution expected negligible at low SIC). Because drift and SIC uncertainties prevent evaluating the budget at SIC <15% (prior to da), Dy and Th contributions are integrated over a window Δt after da (tested 15–60 days; chosen 30 days) to compute ΔSIC, Dy = (∫∇·(u SIC) dt)/ΔSIC, Th = (∫residual dt)/ΔSIC, and the ratio |Dy/Th|. Regions with |Dy/Th|<1 are thermodynamics-dominated; |Dy/Th|>1 are dynamics-dominated. The |Dy/Th|=1 contour delineates an inner zone (freezing-dominated) and an outer zone (transport/melt-dominated near the ice edge). Ocean thermal metrics and mixed-layer heat content: Seasonal diagnostics include SSTmax and its date dSST during the ice-free season (between dr and da), and SST on the dates of advance and retreat (SSTda, SSTdr). Because SST alone cannot capture variable MLD, an observational mixed-layer heat content (MLH) proxy is defined using monthly MLD and SST: MLH(t) = ρ c_p MLD_t × SST(t). From this, MLHmax = ρ c_p MLD_dSST × SSTmax, MLHda = ρ c_p MLD_da × SSTda, etc. Spatial relationships are analyzed via least-squares regressions and 2D histograms between anomalies of da, SSTmax, MLHmax, and dr, considering only inner-zone points where appropriate. Heat budget framework: A simple mixed-layer heat budget links da to MLHmax and the average net heat loss during the cooling period, <Q̄>, integrating from dSST to da: da − dSST = (MLHmax − MLHda)/<Q̄>. This yields an expected linear MLHmax–da relation with slope inversely proportional to <Q̄>. Similarly, MLHmax relates to dr through the heating period budget. The slope-derived <Q̄> is interpreted as the net mixed-layer cooling rate (dominated by air–sea fluxes). Interannual analysis: To assess whether mean-state mechanisms apply interannually, detrended time series (1982–2018) are used to correlate anomalies of dr with subsequent SSTmax, SSTmax with subsequent da, and dr with subsequent da. SST serves as a proxy for MLH at interannual scales due to limited time-varying MLD coverage. Spatial patterns of correlation and the standard deviation of da are evaluated, with significance at the 95% level.

- Two regimes govern sea-ice advance: In the inner seasonal ice zone (south of the |Dy/Th|=1 contour), local seawater freezing dominates sea-ice concentration increases after da. In a circumpolar outer band a few degrees wide near the ice edge (32% of the seasonal ice zone area), dynamic import of sea ice into warmer waters dominates, and net melting can occur on the day of advance. - Sea surface temperature at the date of advance indicates warmer-than-freezing mixed layers in the outer zone: median SSTda − Tf ≈ 0.6 ± 0.3 °C vs 0.4 ± 0.2 °C in the inner zone; the top 5% of SSTda − Tf (>1 °C) cluster near the outer-zone boundary. Hydrography shows thermal instability at the base of the mixed layer in the outer zone during the first three months after advance, consistent with entrainment opposing ice growth. - Robustness of zone delineation: The spatial patterns of Th>0, Dy>0, |Dy/Th|>1, and regions of net melt (Th<0) are insensitive to the chosen 15–60 day window following da. - Thermodynamic control of da via mixed-layer heat content: Spatially, da correlates strongly with SSTmax (R² = 0.81) but even more with MLHmax (R² = 0.89) when accounting for MLD variability, removing nonlinearity observed where mixed layers are deep (>80 m). The inferred mean net cooling rate from the MLHmax–da slope is <Q̄> ≈ 80 W m⁻², consistent with independent estimates of fall air–sea heat loss. - Retreat sets summer heat content: MLHmax is tightly linked to dr (R² = 0.80), indicating that earlier/later retreat regulates the amount of heat stored in the summer mixed layer through modulation of surface radiative input. Consequently, dr and da are strongly and negatively linked spatially (later dr → earlier da; R² = 0.78) in the inner zone. - Outer-zone relationships are weaker but present: MLHmax–da R² = 0.83; dr–MLHmax R² = 0.72; dr–da R² = 0.61, with larger residuals likely due to spatial variability in net heat fluxes and entrainment/advection of warm waters. - Interannual variability: Detrended correlations show significant negative dr–subsequent SSTmax (r ≈ −0.6 ± 0.2, p<0.05) and positive SSTmax–subsequent da (r ≈ 0.5 ± 0.2, p<0.05), yielding significant negative dr–subsequent da correlations (r ≈ −0.5 ± 0.2, p<0.05) over large areas. However, correlations are weak/insignificant near the ice-edge and in East Antarctic/Maud Rise sectors, indicating stronger roles for drift variability and ocean heat input there. The standard deviation of da peaks within/near the outer zone, consistent with enhanced dynamical and ocean-heat influences on timing interannually. - Overall, climatological da in the inner zone is set by the onset of freezing controlled by mixed-layer heat content, whereas in the outer zone it reflects a balance between imported ice and ocean heat-driven basal melt. Across the seasonal ice zone, MLHmax is a primary determinant of climatological da.

The findings demonstrate that the spatial climatology of sea-ice advance is fundamentally constrained by upper-ocean thermodynamics: the amount of heat accumulated in the mixed layer during the ice-free season and the subsequent rate of cooling to the freezing point. Retreat timing modulates summer radiative heat uptake, setting MLHmax, which in turn sets da. Near the ice edge, dynamic import and ocean heat supply complicate the balance, but MLHmax still explains a large portion of da variability. Interannually, the same thermodynamic linkage operates widely but is less dominant than in the mean state, especially where winds, ice drift, and oceanic heat entrainment/advection are highly variable. These results provide a process-based framework (dr–MLHmax–da) to interpret observed variability and to evaluate climate models’ representation of Southern Ocean sea-ice seasonality and the ice concentration budget during advance. They also imply that long-term changes featuring earlier retreat and warmer surface waters would tend to delay advance via increased MLH, while concurrent stratification changes that shoal the mixed layer could offset this by reducing heat content. This balance will be critical for projecting future Antarctic sea-ice seasonality.

The study identifies two distinct regimes controlling Antarctic sea-ice advance: inner-zone advance governed by local freezing and outer-zone advance governed by import of drifting ice into warm waters, often with concurrent basal melt. It establishes that the summer mixed-layer heat content (MLHmax) is a strong predictor of the climatological advance date across the seasonal ice zone and that MLHmax is itself largely set by the timing of sea-ice retreat. A simple mixed-layer heat budget quantifies the mean net cooling rate (~80 W m⁻²) linking MLHmax to advance. Interannually, the dr–SSTmax–da chain is evident but weaker, with dynamical and ocean-heat processes contributing more strongly than in the mean state. The inner–outer zone framework and the dr–MLHmax–da relationships offer practical constraints for evaluating and improving climate models and for anticipating long-term changes. Future research should resolve how evolving stratification and mixed-layer depth interact with surface warming to shape MLH and the seasonal timing of Antarctic sea ice, and clarify regional roles of entrainment, advection, and drift in interannual variability.

- The sea-ice concentration budget cannot be evaluated prior to advance because of missing ice-drift data and large SIC errors at low concentration; Dy/Th is therefore inferred from integration over 15–60 days after da (30 days used), assuming this approximates processes near da. - The residual term of the SIC budget is treated as purely thermodynamic; although mechanical redistribution (ridging/rafting) should be negligible at low SIC, it is not directly separated. - SST-based diagnostics have uncertainties; SSTda − Tf differences and the inner–outer boundary location carry uncertainty. Alternative SST products and longer periods were tested for robustness, but product limitations remain. - MLH is estimated using monthly climatological MLD and satellite SST; interannual analyses rely on SST as a proxy for MLH due to limited time-resolved MLD, potentially omitting variability from mixed-layer depth changes. - Linear regression frameworks assume relatively uniform average net heat fluxes during cooling/heating; regional departures (e.g., Maud Rise, East Antarctica) indicate important roles for entrainment/advection and winds that are not fully quantified. - The outer-zone definition based on |Dy/Th|=1 and the estimated area fraction (32%) depend on the 2003–2010 AMSR-E period and may vary with different datasets or years. - Spatial and temporal coverage gaps (e.g., drift fields, hydrography) and gridding/interpolation may introduce biases in regional diagnostics and correlations.