Elephant seal dives reveal deep submesoscales year-round but seasonal in the Southern Ocean

The mixed-layer paradigm and the first hints of an interior regime

Two decades of submesoscale research established the surface mixed layer as the primary theater of 1–20 km fronts, where mesoscale strain sharpens buoyancy gradients and mixed-layer instability converts stored potential energy into eddy kinetic energy [5]. That framework was extended only tentatively to the first ~100 m beneath the mixed layer, and observations of deeper submesoscale structures date back to the 1980s but remained geographically sparse [1]. A 1/48° realistic simulation of the Antarctic Circumpolar Current with tidal forcing showed the interior departing from quasigeostrophic equilibrium down to 900 m, with order-one Rossby and Richardson numbers, cyclone–anticyclone asymmetry, and vigorous ageostrophic frontogenesis driving intense upward heat fluxes [7]. That result reframed the interior as potentially ageostrophic, but it was a model result in one region, not a basin-scale observational census.

The anchor study converts that hypothesis into an observational claim. Using 25 tags deployed 2014–2020, it assembled 133,598 dives with a median inter-dive spacing of 781 m and median depth of 301 m, yielding 118,121 profiles at 1 km horizontal and 1 m vertical resolution across roughly 2,000 km × 9,000 km of the Indian sector [1]. The key measurement is the filtered lateral buoyancy gradient bx: 60% of values below the mixed layer exceed 0.25 × 10⁻⁷ s⁻² versus 40% within it, and 35% versus 20% exceed 0.5 × 10⁻⁷ s⁻² [1]. In other words, the strongest fronts are not at the surface but beneath it.

Spring intensification tracks surface eddy energy, not mixed-layer instability

Deep submesoscale activity in the East region peaks in austral spring (Sep–Nov) and weakens in fall (Mar–May), and this cycle closely mirrors the seasonality of surface eddy kinetic energy from the NeurOST altimetry product [1]. Between 170 and 500 m, the mean correlation coefficient between EKE and bx is 0.86, and the depth-invariant behavior of submesoscale intensity below 170 m contrasts with the depth-dependent structure expected from mixed-layer instability [1]. The authors invoke Burger-number scaling: with N ≈ 5 × 10⁻³ s⁻¹ and |f| ≈ 10⁻⁴ s⁻¹, a 50 km mesoscale eddy has a vertical scale near 1,000 m, so its strain field can reach through much of the upper ocean and sharpen fronts well below the mixed layer [1].

This mechanism is plausible but not proven by correlation alone, and the authors say so. The altimetry product resolves scales ≳100 km, so it cannot detect the smaller eddies whose spice lenses appear in the seal transects [1]. An alternative or complementary explanation is that subsurface coherent eddies smaller than 50 km stir temperature and salinity around their edges, producing slanted lateral buoyancy gradients [1]. The competing literature offers a different generative route entirely: in the East China Sea, submesoscale eddies of 5–20 km radius are distributed antisymmetrically across the Kuroshio, with horizontal shear and vertical buoyancy flux both feeding eddy kinetic energy, and vertical buoyancy flux dominating at the smallest scales [2]. That is a western-boundary-current regime, not the ACC, but it shows that mesoscale strain is not the only pathway to submesoscale fronts.

Why seals could see what ships and gliders cannot

The observational advance rests on a biologging program that has deployed over 1,000 tags since 2004, including 746 on southern elephant seals, producing more than 420,000 temperature–salinity–depth profiles from Antarctic to tropical waters [4]. Elephant seals dive up to 80 times per day to depths exceeding 1,000 m, and four tags in this study recorded high-resolution data through austral winter for up to ~5 months, filling a seasonal gap that conventional platforms cannot cover [1]. The CTD data themselves required a dedicated postprocessing chain: thermal-mass correction for temperature and conductivity, salinity-spike and density-inversion removal, yielding accuracies of ±0.02°C and ±0.03 g kg⁻¹ for high-resolution profiles [6].

Those accuracies matter because the signal is small. The study filters out internal gravity waves using along-isopycnal spice gradients, which are not generated by IGWs, and shows that seasonality and EKE–bx correlation survive without the filter [1]. But the authors are explicit that the filter reduces rather than eliminates IGW contamination, and that the balanced inverse Richardson number is derived assuming geostrophic balance, which is questionable in the strongly ageostrophic frontal regions being studied [1]. Independent work using the same class of seal data found that interior-plus-surface quasigeostrophic reconstruction captures mesoscale structures but misses the seal-observed smaller-scale signals, and that 2D and 3D omega-equation diagnoses of vertical velocity diverge substantially [9]. That is a direct caution against over-reading the vertical-exchange implications from the present dataset alone.

What deep fronts mean for heat, nutrients, and the energy cascade

Submesoscale currents do not usually mix directly, but they cascade energy and tracer variance from the adiabatic mesoscale down to scales where diapycnal mixing can act, and they redistribute buoyancy, heat, freshwater, and biogeochemical tracers [5]. The anchor paper argues that because deep buoyancy gradients and temperature anomalies are comparable in magnitude to those in a prior localized study, the vertical velocities up to ~100 m day⁻¹ and vertical heat transports of order 100 W m⁻² inferred there may apply across a much broader region [1]. If so, nutrient transport accompanying these motions could be substantially more widespread than recognized, with implications for biogeochemistry and climate [1].

The energy-cascade context is more nuanced than a simple local story. Satellite altimetry analyzed with a coarse-graining framework shows that the surface kinetic energy cascade is predominantly spatially nonlocal: energy lost or gained at larger scales can be redistributed to other regions rather than feeding local smaller scales, and this nonlocality is pronounced in the Southern Ocean and Kuroshio Extension [3]. That means deep submesoscale activity need not be energetically closed locally, and the seasonal EKE–bx correlation could reflect regional energy transport as much as local frontogenesis. The anchor paper's own discussion acknowledges that the seasonality of mesoscale EKE itself has multiple drivers, including mixed-layer instabilities feeding an inverse cascade with a lag of a few months, and wind stress and wind stress curl that strengthen winter baroclinic instability [1].

Where the conclusion stops: one sector, one species, one season of coverage

The results are based on the Indian sector of the Southern Ocean, and the authors state directly that regions with different EKE and background stratification may host a different deep submesoscale regime [1]. The East region provides continuous 2014–2020 coverage and the full seasonal cycle, but the South region has data only for summer, fall, and winter of four years, the Northwest only for summer and spring of five years, and the Southwest for summer, fall, and spring [1]. The smallest seasonal sample is 1,541 profiles (summer in the South) and the largest is 25,438 (summer in the East), so regional seasonality claims rest on uneven sampling [1].

The observing system that made this possible is also a constraint. Animal-borne tags are deployed on a limited set of species and routes, and the 20-year IMOS record shows that sustained coverage depends on continued deployments and international collaboration [4]. Sensor limitations persist: dissolved oxygen sensors on these tags show offsets and temporal drifts that require cautious interpretation [10], and glider sampling of submesoscale variability carries orientation-dependent biases with a mean error of 52% that are insensitive to deployment length or additional gliders [8]. The anchor paper points to the SWOT altimeter and near-surface drifters as the path to quantifying three-dimensional deep submesoscale dynamics and vertical velocities [1], but until seal and SWOT observations are concomitant, the vertical-exchange magnitudes remain an inference from a single sector rather than a measured global budget.

About These Sources

This research page is built on 10 peer-reviewed studies — published from 2015 to 2026, 4 from 2024 or later, 1 in Q1–Q2 journals, collectively cited 140 times — selected as the most relevant from 13 studies that passed quality screening, drawn from 95 papers retrieved from a database of over 500 million.

Sources used in this answer