The study, led by Daria Paul of the Alfred Wegener Institute in Bremerhaven together with colleagues at the Helmholtz-Zentrum Hereon, the German Climate Computing Center, and the University of Hamburg, draws on field campaigns aboard the research icebreaker RV Polarstern during three austral summers between 2018 and 2021. Across 28 ice stations in the Weddell Sea, the team sampled 19 first-year ice floes and 9 multi-year ice floes, excavating 50 snow pits and collecting 921 profiles with a SnowMicroPen, a motorized penetrometer that drives a sensor tip into the snow at a constant speed of 20 millimeters per second. The instrument records penetration resistance at roughly 4-micrometer vertical resolution, capturing about 250 force measurements per millimeter of snow, a level of detail no manual snow pit can match.

Raw penetration force signals, however, do not directly reveal snow types. To convert thousands of force curves into stratigraphic information, the researchers trained a one-dimensional convolutional neural network on 155 manually classified profiles, teaching it to recognize the characteristic signatures of four major snow classes: wind slab, rounded and faceted grains, depth hoar, and melt-freeze forms. The network, built from four convolutional layers with a kernel size of five, processed the profiles in 1-millimeter depth bins, generating more than 7.7 million training data points from the labeled subset. Validation showed that wind slab and rounded-faceted snow were classified correctly more than 80 percent of the time, while depth hoar and melt-freeze forms reached around 70 percent, with most errors confined to transitions between structurally similar classes. The trained model then classified the remaining 766 unlabeled profiles.

With the full dataset classified, a clear pattern emerged. The intrinsic properties of each snow type, including density and specific surface area, the measure of ice crystal surface area per unit mass that governs how snow interacts with light and microwaves, were broadly similar on first-year and multi-year ice. Wind slabs and melt-freeze layers were dense, averaging around 355 to 360 kilograms per cubic meter, while rounded-faceted snow and depth hoar were looser, at roughly 252 to 290 kilograms per cubic meter. What separated the two ice regimes was composition. On first-year ice, rounded and faceted crystals dominated, making up 57 percent of the snowpack, while melt-freeze forms accounted for only 13 percent. On multi-year ice, the proportions nearly inverted: melt-freeze forms rose to 35 percent and rounded-faceted snow fell to 30 percent.

This compositional shift has a physical explanation rooted in the contrasting histories of the two ice types. First-year ice is thin, allowing strong conductive heat fluxes from the relatively warm ocean below to sustain steep temperature gradients within the snow, which drive dry metamorphism and the growth of faceted grains and depth hoar. Multi-year ice, by contrast, is thicker and its snow survives multiple summers, experiencing repeated episodes of surface melt. Meltwater percolates into the snowpack and refreezes, producing hard, dense melt-freeze layers near the base that are often interleaved with depth hoar. The multi-year snowpacks averaged six distinct layers compared with five on first-year ice, and their bulk density was significantly higher, at 321.7 versus 301.7 kilograms per cubic meter, not because individual layers were denser but because dense melt-freeze layers were roughly three times more prevalent and twice as thick.

Perhaps the most consequential result concerns spatial scales. Using Moran’s I spatial autocorrelation analysis along transects up to 64 meters long, the team quantified how far snow properties remain correlated horizontally. On first-year ice, correlations collapsed rapidly, falling from values of 0.4 to 0.6 at short separations to below 0.2 within about 15 meters. On multi-year ice, snowpack composition stayed far more coherent, with autocorrelation values of 0.6 to 0.8 for most snow types and sustained values around 0.4 out to roughly 20 meters. The contrast reflects the greater surface stability of multi-year floes, whose long-lived snow covers smooth out small-scale roughness, whereas first-year ice surfaces are continually reshaped by deformation, wind redistribution, and episodic flooding.

The implications for measurement strategy are sobering. When the researchers compared the variance of bulk density within individual floes against the variance across all floes of the same ice class, they found that a single floe captures only about half, roughly 50 percent for first-year ice and 49 percent for multi-year ice, of the density variability characteristic of its regime. In other words, even a well-sampled floe is only partially representative of its ice type, and point measurements on first-year ice become statistically meaningless within meters. Robust characterization of Antarctic snow demands sampling across multiple floes, a requirement that clashes with the logistical realities of polar fieldwork.

The study also revealed a hierarchy of variability that matters for how models and satellites represent snow. Relative variability, expressed as the ratio of standard deviation to mean, was highest for snow-type fractions and layer thicknesses, nearly double that of specific surface area and snow depth. Bulk density, by contrast, was the most spatially homogeneous property, meaning snow load estimates can be upscaled more reliably than stratigraphic detail. But the authors caution that this does not make bulk representations sufficient. Microwave emission and scattering, the physical basis of satellite retrievals of sea ice concentration and thickness, are sensitive to the vertical distribution of density, grain size, and refrozen melt layers. Radar altimeters such as those on CryoSat-2, operating in the Ku-band, frequently scatter within the snowpack rather than at the true snow-ice interface, introducing substantial uncertainty into derived ice thicknesses.

There is also a hemispheric twist. Arctic studies have reported the opposite spatial pattern, with higher correlations on first-year ice than on multi-year ice, linked to the heterogeneous summer surfaces of Arctic multi-year floes, pockmarked by melt ponds and bare ice. The Antarctic result, obtained in late summer when depth hoar is comparatively minor and dense basal melt-freeze layers dominate multi-year snowpacks, underscores that snow stratigraphy assumptions cannot simply be transferred between hemispheres or seasons. The consistency of specific surface area within snow types across ice regimes, however, offers something valuable: snow-type-specific parameters that may be transferable for microwave emission and radiative-transfer modeling, even as whole-snowpack properties diverge.

The team acknowledges limitations, including the penetrometer’s inability to penetrate the hardest basal ice layers, which likely means multi-year ice stratigraphic complexity and bulk density are underestimated, and the coupling of ice type to sampling region in the Weddell Sea. Yet the core conclusion stands robust: Antarctic snow on sea ice is a stratigraphically structured, ice-regime-dependent medium whose heterogeneity is governed by layer composition and geometry rather than bulk properties. As the authors note, capturing this variability is essential for improving sea ice and climate models and the satellite products that validate them. In a warming Antarctic where the extent and character of sea ice are shifting, knowing exactly how snow is stacked, and how quickly that stacking changes over a few steps across the ice, may prove fundamental to reading the polar climate record correctly.

Subject of Research: Spatial variability of snow stratigraphy and properties on Antarctic late-summer sea ice

Article Title: Variability of internal snow properties over Antarctic late summer sea ice on different spatial scales

Article References: Paul, D., Caus, D., Keil, P., Kadow, C., & Arndt, S. (2026). Variability of internal snow properties over Antarctic late summer sea ice on different spatial scales. The Cryosphere, 20(9), 5491-5508. https://doi.org/10.5194/tc-20-5491-2026

Image Credits: AI Generated

DOI: 10.5194/tc-20-5491-2026

Keywords: Antarctic sea ice, snow stratigraphy, SnowMicroPen, machine learning, convolutional neural network, Weddell Sea, first-year ice, multi-year ice, snow density, specific surface area, remote sensing, climate models

The stakes are considerable. Antarctic Bottom Water forms when dense shelf water, chilled by contact with Antarctic air and ice, spills off the continental shelf around Antarctica, plunging down the slope and mixing with overlying deep water. In the Weddell Sea, this process produces Weddell Sea Deep Water and the even colder Weddell Sea Bottom Water. The deep water overflows the South Scotia Ridge into the Scotia Sea, threads through passages around South Georgia, and pools in the Argentine Basin, where its core properties hover near a potential temperature of −0.10 °C and a salinity of 34.67. From there, the only significant exit toward the equator is the Vema Channel, a narrow, deep trench near 31° S that funnels the coldest bottom water into the Brazil Basin and, eventually, toward the North Atlantic. Every link in this chain matters for the Atlantic Meridional Overturning Circulation, yet direct observations of the abyssal limb remain sparse, limited to a handful of trans-basin moored arrays and repeated shipboard sections.

That observational scarcity is precisely why ocean models and reanalyses carry so much weight. Reanalyses such as ECCO, SODA, and GLORYS blend observations into a numerical simulation to produce a continuous picture of the ocean, while forward models like OFES evolve freely under prescribed atmospheric forcing. Santos and his colleagues began by auditing four such products against the World Ocean Atlas 2018 climatology and against in situ measurements, including thirteen years of conductivity–temperature–depth casts along the SAMBA-West line at 34.5° S and temperature records from three moorings deployed within the Vema Channel. The verdict was sobering. All four simulations broadly captured the large-scale distribution of abyssal water masses, but each displayed persistent warm biases in the deep ocean, and none clearly reproduced the cascading of dense shelf water down the Antarctic continental slope. Most tellingly, all of them showed a discontinuity in abyssal properties between the Argentine and Brazil Basins, as if the cold water simply failed to make it through the Vema Channel.

The temperature–salinity diagrams told the story in stark terms. In the Argentine Basin, the reference climatology shows a well-defined curve extending to waters colder than −0.5 °C. In the Brazil Basin, that curve shifts toward warmer, lighter values, because sills along the Vema Channel block the coldest water and mixing erodes what passes through. The models, however, exaggerated this transition artificially. ECCO, SODA, and OFES shifted their entire deep temperature–salinity curves toward lighter densities north of the channel, while GLORYS lost a striking fraction of its densest water classes. In the channel itself, the picture was even worse: ECCO represented the passage with a single column of grid points, SODA contained no points there at all, and even the higher-resolution products misaligned, widened, or shallowed the trench, leaving no water colder than 0.0 °C inside it.

Because the four simulations differed in so many respects—atmospheric forcing, data assimilation schemes, bathymetry, and sea-ice representation—the intercomparison alone could not pin the blame on any single culprit. So the team turned to the Finite-volumE Sea ice–Ocean Model, FESOM 2, an unstructured-mesh model developed at the Alfred Wegener Institute that is at least nine times faster than its finite-element predecessor. They built a reference configuration, FESOM R, designed to mimic the vertical grid of OFES, with 105 vertical levels and horizontal spacing of about 30 kilometers over most of the globe, refined to 10 kilometers in the South Atlantic. Then they ran three targeted variants, changing nothing else: FESOM V added twenty vertical levels concentrated between 3800 and 5350 meters, bringing the spacing near the seafloor down to roughly 50 meters; FESOM H refined the horizontal grid to 7.5 kilometers in the South Atlantic and 4 kilometers around the Vema Channel; and FESOM VH combined both refinements.

The results were unambiguous. FESOM V, the vertically refined experiment, delivered the largest basin-scale improvement. Cold waters below −0.7 °C spread across a broader swath of the central Weddell Sea, waters at or below 0.0 °C filled the entire Argentine Basin, and a colder class of bottom water finally flowed through the Vema Channel into the Brazil Basin. The warm bias that plagued every configuration shrank and became spatially uniform, without the abrupt basin-to-basin discontinuities seen in the reference run. The density evolution of the deep basins also became more coherent: in FESOM R and FESOM H, entire density classes of abyssal water vanished from the Argentine and Brazil Basins within decades of the simulation, whereas the vertically refined runs retained them. The mechanism is physical rather than numerical sleight of hand. Colder abyssal layers extend upward through the lower water column, allowing the coldest water to reach the depth of the Vema Channel sill and spill northward, instead of being trapped below the deepest model level connected through the passage.

FESOM H, by contrast, was the weakest performer of the four. Refining the horizontal grid alone produced warmer abyssal layers, erased the coldest waters from the central Weddell Sea, and left the Argentine Basin without any water below 0.0 °C. The likely explanation lies in the delicate balance between resolved and parameterized turbulence. The Gent–McWilliams eddy parameterization, which stands in for mesoscale eddies the model cannot explicitly resolve, scales its diffusivity with grid spacing. When the grid becomes fine enough to partially resolve eddies but the parameterization is not retuned, the model can end up double-counting eddy effects, distorting the very mixing that shapes abyssal water masses. The authors caution that this interpretation remains speculative, since they deliberately kept all parameterizations unchanged to isolate the effect of resolution, but the lesson is clear: horizontal refinement without recalibration of mixing schemes can be actively harmful in the deep ocean.

The combined experiment, FESOM VH, offered a more nuanced picture. At the basin scale it did not outperform vertical refinement alone, and its abyssal waters ran slightly warmer than FESOM V’s. But in the Vema Channel it achieved something no other configuration could: a realistic representation of the channel’s geometry, complete with a cold bottom-intensified core deflected toward the eastern wall. That eastward deflection is a genuine dynamical feature, attributed to Ekman transport driven by bottom friction, and it is consistently observed in moored measurements and regional simulations of Antarctic Bottom Water flow. Capturing it required sufficient resolution in both dimensions—vertical refinement to cool the water column and horizontal refinement to resolve the narrow trench—demonstrating that the two grid directions play complementary roles.

Comparisons with the moorings and CTD stations reinforced the hierarchy. Across three historical mooring periods in the Vema Channel, the vertically refined configurations FESOM V and FESOM VH systematically reduced warm biases and reproduced the observed variability more faithfully, while FESOM H consistently produced the largest warm biases. Along the SAMBA-West line, FESOM V again provided the closest match to observed mean temperatures at Sites C and D, though all configurations overestimated the observed warming trend. Correlations with individual mooring time series remained modest across the board, a reminder that even a well-resolved global model cannot be expected to nail phase agreement at single points in the abyss. What vertical resolution reliably controls is the mean state and variance of deep temperature—the very quantities that determine whether cold bottom water survives its journey across bathymetric barriers.

The implications reach well beyond the South Atlantic. Global ocean models feed climate projections, sea-level estimates, and carbon budgets, and the abyssal ocean absorbs a substantial share of the excess heat that the climate system accumulates. If the deep thermal structure is systematically too warm because the vertical grid is too coarse, those biases propagate into every downstream product. Previous studies had hinted that enhanced vertical resolution improves the representation of dense water masses and baroclinic modes, but this study provides an unusually clean demonstration, holding everything constant except the grid. Its message to model developers is direct: before spending scarce computing resources on ever-finer horizontal meshes, invest in the vertical grid near the seafloor, and when horizontal resolution does increase, retune the mixing parameterizations that quietly govern the deep ocean. The coldest waters on the planet, it turns out, live or die in the last fifty meters of a model’s water column.

Subject of Research: The effect of horizontal and vertical grid resolution on the representation of Antarctic Bottom Water and abyssal ocean properties in global numerical ocean models

Article Title: Impact of grid resolution on the abyssal ocean representation in numerical models

Article References: Santos, D. M. C., Van Caspel, M., Timmermann, R., & Sato, O. T. (2026). Impact of grid resolution on the abyssal ocean representation in numerical models. Ocean Science, 22(5), 3079-3104. https://doi.org/10.5194/os-22-3079-2026

Image Credits: AI Generated

DOI: 10.5194/os-22-3079-2026

Keywords: Antarctic Bottom Water, ocean modeling, grid resolution, FESOM 2, Vema Channel, Weddell Sea, Atlantic Meridional Overturning Circulation, abyssal ocean, ocean reanalysis, vertical resolution, Argentine Basin, deep ocean warming

The study focuses on HadGEM3-GC3.1-LL, the low-resolution configuration of the UK’s CMIP6 climate model, which simulates the atmosphere, ocean, land, and sea ice as a fully coupled system. The atmosphere component, the Met Office Unified Model, runs on a grid of roughly 135 kilometers, which means that much of Antarctica’s rugged terrain is too small to be resolved directly. Instead, the model relies on a parameterization of orographic gravity wave drag, or OGWD, to represent the forces that sub-grid-scale mountains impose on the flow. This parameterization splits the total drag into two components: flow-blocking drag, which arises when air is too stable to climb over an obstacle and is deflected around it, and gravity wave drag, which arises when air does surmount the terrain and generates waves that break aloft and deposit momentum.

Which of these two regimes dominates depends on a delicate balance captured by the non-dimensional parameter Nh/U, where U is wind speed, N is atmospheric stability, and h is obstacle height. When Nh/U greatly exceeds one, blocking prevails; when it is near or below one, the flow passes over the mountain. In the Unified Model, the effective sub-grid mountain height is set by a tuning constant called n-sigma, with a default value of 2.5 that is among the least constrained parameters in the entire model. The researchers exploited this lever to run two sets of sensitivity experiments over Antarctica: one, called F_BLOCK, in which n-sigma was doubled to 5, making virtual mountains effectively taller and forcing the flow into a blocking regime, and another, called F_OVER, in which n-sigma was reduced to 1.5, nearly halving the barrier height and letting winds sweep across the terrain while generating strong gravity waves. Crucially, the modifications were applied only over Antarctica, so any circulation change in the simulations could be attributed unambiguously to the southernmost continent.

The results were striking. In the baseline historical simulation, pan-Antarctic sea ice area begins a sustained decline around the early 1980s, driven largely by the Weddell Sea, one of the continent’s main sea ice factories. In the flow-blocking experiments, that decline is delayed by roughly ten to twenty years, with the ensemble mean holding steady until just before 1995. In the flow-over experiment, the decline begins about ten years earlier still, around 1972. Because the delayed decline appears consistently across all four members of the flow-blocking ensemble, the team could rule out natural variability as the explanation: the differences are a systematic response of the coupled climate system to how orographic drag is represented.

The immediate physical mechanism operates through the atmosphere’s vertical structure. In the flow-over configuration, larger-amplitude gravity waves break in the lower and mid-troposphere, decelerating the background westerly flow and weakening the tropospheric polar vortex. In the flow-blocking configuration, weaker wave amplitudes mean less drag and stronger zonal winds aloft. A stronger polar vortex insulates the Antarctic continent, exchanging less heat with the mid-latitudes, so surface air temperatures drop in F_BLOCK and rise in F_OVER. Colder conditions strengthen the katabatic winds, the rivers of cold, dense air that drain from the high interior toward the coast, which push sea ice offshore in the Weddell Sea and promote the formation of fresh ice. Within the first one to two years of simulation, sea ice area rises by about one million square kilometers in F_BLOCK and falls by a similar amount in F_OVER relative to the historical run.

The long-term story, however, unfolds in the Weddell Sea, and it hinges on the Antarctic Peninsula acting as a formidable barrier to the prevailing westerlies. In the flow-blocking experiment, the incoming flow cannot clear the ridge, so westerly winds both upstream over the Amundsen and Bellingshausen Seas and downstream over the Weddell Sea are weakened. Mean sea level pressure anomalies generated by the blocking propagate vast distances around the continent, echoing earlier work by Sandu and colleagues showing that orographic blocking leaves fingerprints far beyond the mountains that create it. June-to-November westerly winds at 850 hectopascals across the Weddell Sea are reduced by about half a meter per second, roughly sixteen percent, in F_BLOCK, and strengthened by a comparable margin in F_OVER.

Those wind changes translate directly into the ocean. The total surface stress exerted on the Weddell Sea, the combined push of atmosphere and sea ice on the water, is much weaker under flow-blocking conditions and much stronger under flow-over conditions, partly because stress scales with the square of wind speed and partly because thinner or reduced ice allows more efficient momentum transfer. Stronger winds spin up the cyclonic Weddell gyre, whose depth-integrated circulation, measured by the barotropic streamfunction, strengthens steadily through the historical run. A stronger gyre makes the surface flow more divergent, and by continuity, relatively warm, salty water is pumped upward from depth, with Ekman pumping in the historical simulation rising from roughly 55 meters per year in 1980 to about 85 meters per year by the end of the run.

Near Maud Rise, at about 66 degrees south and 3 degrees east, weak stratification and favorable topography set the stage for open-ocean deep convection, the vertical mixing that draws heat from the deep ocean to the surface. In the model, these events are diagnosed when the ocean mixed layer exceeds 2000 meters in depth, and polynya-scale events are flagged when the convecting area surpasses 80,000 square kilometers, the size of the real Weddell Sea polynya of 2016 and 2017. The correspondence with sea ice is dramatic: every cluster of deep convection years coincides with a sharp drop in sea ice area. In the historical simulation, convection recurs almost every year from 1989 onward, and in F_OVER it appears even earlier and more often. In F_BLOCK, by contrast, deep convection is nearly absent, with only a handful of small events across sixty-five years, and the ice survives largely intact.

The team then turned the correlation into a diagnostic tool. Plotting annual sea ice area against ocean surface stress reveals a strong negative relationship, with a correlation coefficient of minus 0.75, and the historical simulation migrates over time from a low-stress, high-ice state toward a high-stress, low-ice state that mirrors the flow-over experiment. Using the observational range of sea ice area from the HadISST1 dataset as an emergent constraint, the researchers estimate that plausible surface stress values lie between 0.042 and 0.066 newtons per square meter, a range that the flow-blocking simulation satisfies best. That result suggests the default model may underestimate atmospheric blocking by the Peninsula, and that a calibrated tuning of the sub-grid orography there would be advisable.

The broader implications reach beyond model tuning. The study shows that a more positive Southern Annular Mode does not necessarily mean stronger winds over the Weddell Sea, because the Peninsula’s barrier effect can override the large-scale annular signal. It also suggests that the springtime strengthening of westerlies over the Weddell Sea seen in the historical run, which is consistent with ERA5 reanalysis, may partly reflect a long-term shift in the Peninsula’s flow regime toward weaker blocking, driven by declining static stability, rather than greenhouse and ozone forcing alone. And because the spurious deep convection is fueled by wind-driven Ekman pumping, a large-scale process the model resolves explicitly, simply tweaking vertical mixing parameterizations will not fix it. The authors point instead to revising how ocean surface stress is represented, potentially through variable drag coefficients that respond to the evolving state of the ice and atmosphere, an approach already under development for the next version of HadGEM3. In the end, the message is humbling and exhilarating in equal measure: a single tuning constant, buried in the treatment of invisible mountains, can decide whether a flagship climate model loses its Antarctic sea ice a decade early, a decade late, or not at all.

Subject of Research: The influence of orographic gravity wave drag and flow regimes over the Antarctic Peninsula on Weddell Sea sea ice and deep convection in a coupled climate model

Article Title: Southern Hemisphere sea ice response to different flow regimes over the Antarctic Peninsula

Article References: Guarino, M.-V., Ridley, J. K., Farneti, R., Kucharski, F., & Tompkins, A. M. (2026). Southern Hemisphere sea ice response to different flow regimes over the Antarctic Peninsula. Climate Dynamics, 64(9), Article 410. https://doi.org/10.1007/s00382-026-08350-6

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08350-6

Keywords: Antarctic sea ice, Antarctic Peninsula, orographic gravity wave drag, Weddell Sea, deep convection, polynyas, HadGEM3, CMIP6, wind stress, Ekman pumping, Southern Annular Mode, climate modeling