Observability of Finite-Depth Double-Diffusive Exchange from Sparse Temperature-Salinity Measurements

Sriram P. Kalathoor

REVIEW 3 major objections 5 minor 16 references

Observability of Finite-Depth Double-Diffusive Exchange from Sparse Temperature-Salinity Measurements

T0 review · 3 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The useful observing unit for hidden finite-depth double-diffusive exchange is a small ensemble or section, not an isolated temperature-salinity cast.

desk verdict The qualitative hierarchy (ensembles beat isolated casts) is credible and worth publishing; the quantitative accuracies need a leave-one-simulation-out robustness pass first. read the letter →

arxiv 2607.16272 v1 pith:K6VGW2QJ submitted 2026-07-08 physics.geo-ph cond-mat.mtrl-sciphysics.chem-phphysics.flu-dyn

Sriram P. Kalathoor This is my paper · ORCID

classification physics.geo-phcond-mat.mtrl-sciphysics.chem-phphysics.flu-dyn

keywords double-diffusiveexchangesaltfingersoceanobservingsystemshydrographicprofilesobservabilityprofilebundlesverticalsamplingresolutioninterfacetracking

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks how much of a spatially organized double-diffusive exchange history survives sparse ocean measurements. Using four controlled simulations as known truth, it shows that a single vertical profile is a weak route identifier: late-time strict-case accuracy stays below 0.48 and route-family accuracy is around 0.53–0.56, with most profiles closer to a different-route profile than to a same-route one. Small bundles of profiles recover route family much better: two-profile bundles reach 0.917 route-family accuracy, three reach 0.961, and four reach 0.994. Coarsening vertical sampling preserves broad route separation while distorting interface-width estimates, and sections only recover horizontal modes when station spacing resolves them. The paper concludes that route-relevant information is a property of the observing ensemble, not the single cast.

What carries the argument

The analytical engine is a set of five profile observables computed from vertical temperature-salinity casts—salinity-gradient interquartile width, central and outer gradient fractions, gradient entropy, and profile T-S correlation—projected into a standardized feature space. Route-family recovery is measured by leave-one-out nearest-centroid and k-nearest-neighbor classifiers on single profiles, and by exhaustive enumeration of all profile bundles of a given size from nine locations per simulation. Section observables (dominant horizontal mode and interface roughness) extend the same bookkeeping to horizontal sampling. The key mechanism is that aggregation converts a locally ambiguous cast

What would settle it

Run the same profile and bundle classification on a new set of independently generated finite-depth exchange simulations with different horizontal spectra, background parameters, or random seeds; if isolated casts reach high route-family accuracy, or if two-profile bundles do not reliably beat single profiles, the central ensemble claim fails. A quick numeric check: train on many random phase realizations and test whether single-profile leave-one-out accuracy exceeds the roughly 0.5 mark reported here.

Watch

Extended reading notes

Core claim

The central claim is that finite-depth double-diffusive exchange routes are observable in principle, but not through isolated casts. In a truth set of four simulated exchange histories with identical background parameters and different initial horizontal structure, profile-derived metrics—gradient width, central and outer gradient fractions, gradient entropy, and temperature-salinity correlation—collapse distinct routes: a solitary cast cannot reliably say whether the interface is compact, broad, low-mode, or mixed. Adding a second profile raises route-family identification from a 0.694 one-profile baseline to 0.917, and four-profile bundles essentially saturate the controlled set at 0.994.

Load-bearing premise

The result rests on the four simulations and the five scalar-profile metrics being an adequate stand-in for the full space of finite-depth double-diffusive exchange; if the real ocean's route diversity is richer, or a different feature set reshuffles the distances, the accuracy thresholds and bundle-size conclusions are tied to that controlled truth set.

Editorial extensions

If this is right

Ocean observing strategies for double-diffusive regimes should treat a cluster of nearby profiles, rather than a single cast, as the basic unit for inferring exchange history.
Two to four properly spaced profiles can separate broad route families in favorable finite-depth settings, with accuracy rising from about 0.69 for one profile to 0.917, 0.961, and 0.994 for two, three, and four profiles.
Vertical resolution should be chosen according to the target: coarse spacing can support broad route classification, but quantitative interface-width estimates need fine sampling.
Section designs need station spacing fine enough to resolve the expected horizontal mode; otherwise dominant modes alias even when roughness information survives.
The same scalar metrics can be computed on real high-resolution profiles and section products, allowing practical observing formats to be compared with route-known synthetic truth.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The specific bundle-size thresholds (0.917, 0.961, 0.994) are conditioned on a four-member taxonomy; a real ocean with a continuum of routes likely needs larger or differently spaced bundles for the same confidence.
If the goal is field diagnostics, combining small profile bundles with a coarse section could recover both vertical route structure and horizontal scale more efficiently than either alone.
The single-profile failure may partly reflect the restricted feature set; adding velocity shear, microstructure, or tracer information could shift some route information back into individual casts, so the ensemble result is a bound on current scalar-profile observables.
The separation between broad route-state observability and quantitative metric fidelity suggests that field surveys could be tiered: coarse networks first for route classification, then targeted dense sampling where route boundaries or sharp interfaces matter.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Read this one with the stress-test note in hand. The bundle accuracies in §3.2 are the load-bearing numbers, and the authors never state whether the bundle classifier was validated by leaving out whole simulations or just individual profiles. The nine profiles per route come from one evolving field and are spatially correlated; leave-one-profile-out can leak near-duplicate information into training. That means the 0.917/0.961/0.994 ladder may be inflated. The single-profile LOO numbers (0.528–0.556) suffer the same bias, so the gap that drives the ensemble conclusion could be smaller than advertised. The qualitative conclusion—route-family information lives in spatial ensembles, not single casts—is plausible and probably survives a stricter validation, but the specific thresholds are not yet trustworthy.

What is genuinely good: the analysis is honest about its scope. Table 2 explicitly disclaims universal classification; the real ITP/Argo/GO-SHIP products are used only as format context, not validation; features are computed without route labels; single-profile classification is leave-one-out. The coarsening result—broad route-state separation survives coarse vertical sampling while interface-width metrics distort—is a clean and useful separation. The section-mode aliasing discussion is also informative, though the non-monotonic behavior in Table 5 (7→16→8→16 for high-annulus) needs diagnostic comment: it looks like interpolation or detection artifacts, and the paper just reports it.

Minor issues: the two single-profile baselines (0.53 LOO vs 0.694 bundle-enumeration baseline) are reported as different tests, which is fine, but a reader has to dig to see why they aren't inconsistent. The '22 of 36 profiles closer to a different-route profile' statistic is near chance for a three-family grouping (~23/36) and is reported without that baseline. No uncertainty intervals anywhere—point estimates on a four-simulation truth set. No code or processed data released yet, though the companion Zenodo record exists.

Who this is for: oceanographers designing mixed-layer observing campaigns, especially anyone arguing about whether single floats or clustered arrays are needed for double-diffusive structure. It is a legitimate, useful contribution even if the headline numbers soften after a stricter validation.

Recommendation: send to peer review. An editor should not desk-reject this. But the reviewer should insist on leave-one-simulation-out (or leave-one-route-out) classification, a reconciliation of the two single-profile metrics, and uncertainty estimates before publication.

Referee Report

3 major / 5 minor

Summary. The paper asks which route-relevant features of finite-depth double-diffusive exchange remain observable after reduction to sparse temperature-salinity measurements. Four controlled finite-depth simulations (high-annulus, low-mode, mixed baseline, mixed seed) are treated as known truth fields and sampled by vertical profiles, profile bundles, coarsened profiles, Argo-style records, and hydrographic sections. The authors compute five profile observables (salinity-gradient IQR width, central/outer gradient fractions, gradient entropy, T–S correlation) and section metrics, then test whether route labels can be recovered from these observables. The main claim is that isolated profiles are weak route identifiers, that two–four profile bundles recover route-family information with high accuracy (0.917–0.994), that coarse vertical sampling can preserve broad route separation while distorting interface-width estimates, and that sections add horizontal-mode information only when station spacing resolves the relevant mode. The paper concludes that the useful observing unit is an ensemble or section, not an isolated cast.

Significance. If the quantitative results hold, the paper provides a useful measurement-design hierarchy for sparse ocean observations of double-diffusive exchange. The controlled-truth methodology is a strength: observables are computed without route labels, the paper explicitly disclaims universal classification in Table 2, and it carefully separates route-family recovery from quantitative metric fidelity. However, the central quantitative claims rest on a classifier-validation scheme that is not fully specified and appears to use dependent profiles from the same simulation, and on a truth set of only four simulations with two realizations collapsed into one route family. The contribution is potentially significant, but the accuracy ladder and the resulting thresholds must be revalidated before the paper's conclusions can be accepted.

major comments (3)

[§3.1, §2.2, Abstract] The 'leave-one-out profile classification' appears to be leave-one-profile-out. Nine profiles per simulation are drawn from the same evolving mid-plane field (§2.2), so profiles within a route history are spatially correlated. With leave-one-profile-out, training includes eight correlated profiles from the same simulation, leaking information and inflating accuracy. This directly affects the reported single-profile accuracies (0.528/0.556) and the bundle accuracy ladder (0.917/0.961/0.994) that supports the central claim. The correct validation for generalization to an unseen route or unseen spatial region is leave-one-simulation-out (or at least leave-one-route-family-out). Please redo the analysis with that scheme and report the resulting differences.
[§3.2, Table 4] The bundle-enumeration accuracy is not defined. The manuscript does not state what classifier is used for bundles, how a bundle is represented in feature space, how 'all possible bundles' are enumerated, or how accuracy is computed. The one-profile bundle-enumeration baseline (0.694) is inconsistent with the single-profile LOO route-family accuracy (0.53) in Table 4; the explanation that 'the tests are different' is insufficient. Without a precise definition of the classifier and the train/test protocol, the reader cannot audit the central quantitative result.
[§2.1, §3.1] The truth set contains only four simulations, and the 'route family' definition collapses the two mixed realizations into one class, leaving three effective classes with one class represented twice. Because the mixed seed is intentionally a different phase realization of the same mixed baseline, its profiles may be nearly redundant with the mixed baseline, inflating family accuracy. This is acknowledged as a controlled truth set, but the abstract and §8 present the bundle thresholds ('two profiles reach 0.917') without sufficiently emphasizing that these numbers are conditioned on a single, small, and possibly redundant truth set. Please add more independent family realizations or systematically vary the family grouping and report the sensitivity.

minor comments (5)

[§5, Table 3, Fig. 4, Table 6] The number of retained ITP profiles is inconsistent: §5 says 166, while Table 3, Fig. 4, and Table 6 use 'ITP 84'. Please reconcile and ensure the reported medians correspond to the correct sample size.
[§4.2] The 'route separation' ratio is not defined. Please give the exact formula (e.g., distance between route centroids in standardized feature space, or a resubstitution distance) and specify how it is normalized by the native value.
[§2.3, Table 1] The interface-centered inner and outer band widths used for the central/outer gradient fractions are not specified. Please state the band widths or the rule used to determine them, as they are free parameters of the observable definition.
[§6.3] The 'minimal interface-tracking rule' used for the CCHDO/GO-SHIP section is not defined. Please specify the rule, including how candidate interface pressures are selected and when profiles are rejected as unusable.
[Fig. 2b] The bundle-accuracy curves show point values without error bars or confidence intervals. Since the enumeration is finite, please report the number of bundles at each size and, if possible, a bootstrap interval.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the observability numbers are computed from fixed external truth fields and explicitly scoped, not reduced to fitted inputs.

full rationale

The derivation chain is internally non-circular. The truth fields are fixed inputs from the author's companion simulations, but the present observability results are computed after the fact from those fields: §2.1 states 'The fields are fixed inputs: a controlled set of known exchange histories' and 'the observable metrics are computed without using that label.' The profile observables (§2.3, Table 1) are defined on sampled temperature-salinity profiles, and the single-profile classification uses 'two leave-one-out classifiers ... applied in standardized observable space' (§3.1), so the route label is not used to construct the features. The bundle-accuracy ladder (§3.2, Table 4) is a measurement-design statistic evaluated on those fixed truth sets; it is neither a parameter fitted to the labels nor a prediction renamed from a fit. The self-citations (Kalathoor 2026a,b) supply the simulations and route-selection physics, not the bundle accuracies or the ensemble conclusion, so the load-bearing input is independent of the target result. The paper also repeatedly scopes its claims: 'The two- or three-profile threshold is specific to this controlled truth set' (§3.2), the accuracy statements are 'under the current feature set' (Abstract), and the real ocean products are explicitly 'not used as direct validation targets' (§1). A genuine statistical concern remains: the paper does not state whether bundle classification uses leave-one-simulation-out rather than leave-one-profile-out, and within-simulation profile correlation could inflate the reported accuracies; however, that is an experimental-validity issue, not a circular reduction of the kind where an output is equivalent to an input by construction.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests entirely on the author's own simulation pipeline. Free parameters are mostly analysis choices (band widths, thresholds, tolerances, locations, times) that are disclosed in spirit but not numerically specified; they do not appear to be fitted to match the target result, but they do control the reported accuracies. The axioms that matter most are domain assumptions: the companion-paper truth set is assumed to span finite-depth double-diffusive exchange, and the five profile observables are assumed sufficient for route observability. No external benchmark constrains any of these choices — the real ocean products are deliberately used only as format context. Invented entities: none; the route families are categories of the author's simulations, not new physical entities.

free parameters (5)

Interface-centered inner/outer band widths for central and outer gradient fractions
Table 1 defines C_S and O_S relative to 'an interface-centered inner band'/'outer band'; the band widths are never numerically specified, yet central/outer fractions feed the classification feature space and the resolution-degradation tables (Table 6).
Route-family accuracy threshold of 0.95 = 0.95
Table 4 and §3.2 define the bundle-size requirement as the minimum number of profiles for ≥0.95 route-family accuracy; this hand-chosen threshold produces the headline 'three profiles suffice' claim.
Strict tolerances for 'all-metric stable fraction'
§3.3 reports stable fractions of 0.000/0.089/0.889/1.000 for 1/3/8/9 profiles, but the tolerances defining stability are not given; the progression is conditional on them.
Evaluation times (t = 15, 30, 45, 60)
All headline numbers are 'at the final comparison time' (t=60); Table 4 shows single-profile accuracy varies 0.39–0.53 across times, so the choice of late time matters.
Nine profile locations per route history and their coordinates
§2.2 samples nine hand-selected mid-plane locations per route; all bundle accuracies are computed over these specific locations, with no sensitivity analysis to location choice.

assumptions (5)

domain assumption The four finite-depth simulations (high-annulus, low-mode, mixed baseline, mixed seed) from the companion study (Kalathoor 2026b) define the complete relevant space of finite-depth double-diffusive exchange histories.
§2.1: truth set = four simulations with identical background parameters differing in initial spectral content/realization. The route-family taxonomy and all classification targets derive from this unvalidated choice.
ad hoc to paper The chosen profile observables (gradient IQR width, central/outer fractions, entropy, T–S correlation) are sufficient features for route observability.
§2.3 defines five profile metrics as the feature space; classification accuracy is conditional on this specific feature set ('under the current feature set,' §3.1), with no justification that they capture the discriminating information.
domain assumption Classification accuracy in standardized Euclidean feature space with nearest-centroid/3-NN measures physical observability.
§3.1 operationalizes observability as leave-one-out classifier accuracy; a different classifier or feature normalization would change the numbers (as the paper's own 0.53 LOO vs 0.694 bundle-baseline discrepancy shows).
standard math Station-spacing/Nyquist reasoning governs section mode recovery.
§6.2: 'the high-annulus mode aliases to 7 because the Nyquist mode is 8'; standard Fourier sampling theory.
domain assumption ITP, Argo, and CCHDO/GO-SHIP products are representative of real observing formats and their sampling structure.
§2.4: real products anchor observing formats (spacing, QC structure) but are not validation targets; the transfer of synthetic conclusions to real double-diffusive exchange assumes these formats are the relevant ones.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Observability of Finite-Depth Double-Diffusive Exchange from Sparse Temperature-Salinity Measurements." pith.science (2026). https://pith.science/paper/K6VGW2QJ

@misc{pith2026260716272,
  author       = {Pith},
  title        = {Pith review of: Observability of Finite-Depth Double-Diffusive Exchange from Sparse Temperature-Salinity Measurements},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K6VGW2QJ}},
  note         = {Machine review of arXiv:2607.16272}
}

read the original abstract

Double-diffusive interfaces can support exchange histories that are spatially organized but only sparsely observed. A resolved three-dimensional calculation contains the route by which scalar gradients broaden, remain compact, connect with remote parts of a finite-depth layer, or organize horizontally, whereas field products usually provide profiles, repeated casts, autonomous-float records, or hydrographic sections. We ask which route-relevant features of finite-depth double-diffusive exchange remain observable after that measurement reduction. Four controlled finite-depth exchange histories are treated as known truth fields and sampled with vertical-profile, profile-bundle, vertical-coarsening, Argo-style, and section-like observing formats. The resulting observables are compared with selected Ice-Tethered Profiler, Argo, and CCHDO/GO-SHIP products to place the synthetic measurements in realistic observing contexts. Isolated profiles are weak route identifiers: at the final comparison time, strict case accuracy remains below 0.48, route-family accuracy is about 0.53--0.56 in leave-one-out profile classification, and 22 of 36 profiles are closer to a different-route profile than to a same-route profile under the current feature set. Profile bundles are substantially more informative. In the bundle-enumeration framework, the one-profile late-time route-family baseline is 0.694, while two-, three-, and four-profile bundles reach route-family accuracies of 0.917, 0.961, and 0.994. Coarse vertical sampling can preserve broad route separation while distorting local interface-width estimates, and section-like sampling adds horizontal-scale information only when station spacing resolves the relevant mode. The useful observing unit for hidden finite-depth double-diffusive exchange is therefore an ensemble or section, not an isolated cast.

Figures

Figures reproduced from arXiv: 2607.16272 by the authors.

**Figure 1.** Measurement reduction from a finite-depth scalar field to observable profile and section products. The full field contains the hidden exchange route, while profiles and section stations retain only restricted parts of that route history. 2.3 Observable Metrics The profile observables are deliberately narrower than the full simulation state. They are quantities that can be computed from sampled temperature-salinity p… view at source ↗

**Figure 2.** Profile observability of hidden route family. Isolated profiles remain ambiguous, while small profile bundles recover route-family structure much more reliably in the controlled truth set. 4 Vertical Resolution and the Fidelity of Profile Observables 4.1 Coarsening Degrades Interface-Width Estimates Vertical coarsening was applied to synthetic and real profile products to determine how much metric fidelity is lost w… view at source ↗

**Figure 3.** b shows this retained route separation. A retained separation ratio near or above one indicates that the particular reduced feature set preserves broad state differences. The same coarse profiles can still distort the physical width or outer-gradient estimate that an interface interpretation would need. Broad state observability therefore survives more easily than quantitative metric fidelity. 1 2 5 10 20 Vertical s… view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Real high-resolution profile context in the same observable coordinates as the synthetic profile reductions. The comparison is branch-aware: it tests shared scalar bookkeeping without requiring a one-to-one dynamical match between Arctic staircases and salt-finger rout…

**Figure 5.** Figure 5: Section observability at the final comparison time. Station spacing controls whether horizontal modes can be recovered, while roughness can retain partial interfaceamplitude information even when the dominant mode aliases. 6.3 CCHDO/GO-SHIP Provides Real Section Bookk…

**Figure 6.** Figure 6: Supporting vertical-resolution quantities for profile observability [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]

**Figure 7.** Figure 7: Argo-style support material for autonomous-profile sampling. C CCHDO/GO-SHIP Section Support The CCHDO/GO-SHIP appendix records the section geometry that connects hydrographic-station spacing to the synthetic section sampling [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]

**Figure 8.** Figure 8: CCHDO/GO-SHIP I06S section support material. Statements and Declarations Funding. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Competing interests. The author declares no competing interests. 17 …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 1 canonical work pages

[1]

, title =

Schmitt, Raymond W. , title =. Annual Review of Fluid Mechanics , volume =. 1994 , doi =

1994
[2]

, title =

Ruddick, Barry R. , title =. Deep Sea Research Part A. Oceanographic Research Papers , volume =. 1983 , doi =

1983
[3]

Deep Sea Research Part I: Oceanographic Research Papers , volume =

You, Yuzhu , title =. Deep Sea Research Part I: Oceanographic Research Papers , volume =. 2002 , doi =

2002
[4]

Schmitt, R. W. and Ledwell, J. R. and Montgomery, E. T. and Polzin, K. L. and Toole, J. M. , title =. Science , volume =. 2005 , doi =

2005
[5]

Kelley, D. E. and Fernando, H. J. S. and Gargett, A. E. and Tanny, J. and Ozsoy, E. , title =. Progress in Oceanography , volume =. 2003 , doi =

2003
[6]

Radko, Timour , title =
[7]

and Toole, J

Timmermans, M.-L. and Toole, J. and Krishfield, R. and Winsor, P. , title =. Journal of Geophysical Research: Oceans , volume =. 2008 , doi =

2008
[8]

Shibley, N. C. and Timmermans, M.-L. and Carpenter, J. R. and Toole, J. M. , title =. Journal of Geophysical Research: Oceans , volume =. 2017 , doi =

2017

Show all 16 references

[9]

and Koetsier, J

van der Boog, Carine G. and Koetsier, J. Otto and Dijkstra, Henk A. and Pietrzak, Julie D. , title =. Earth System Science Data , volume =. 2021 , doi =

2021
[10]

Journal of Atmospheric and Oceanic Technology , volume =

Krishfield, Richard and Toole, John and Proshutinsky, Andrey and Timmermans, Mary-Louise , title =. Journal of Atmospheric and Oceanic Technology , volume =. 2008 , doi =

2008
[11]

Oceanography , volume =

Toole, John and Krishfield, Richard and Timmermans, Mary-Louise and Proshutinsky, Andrey , title =. Oceanography , volume =. 2011 , doi =

2011
[12]

Oceanography , volume =

Roemmich, Dean and Johnson, Gregory and Riser, Stephen and Davis, Russ , title =. Oceanography , volume =. 2009 , doi =

2009
[13]

and Claustre, Herve and Johnson, Kenneth and others , title =

Roemmich, Dean and Alford, Matthew H. and Claustre, Herve and Johnson, Kenneth and others , title =. Frontiers in Marine Science , volume =. 2019 , doi =

2019
[14]

and Wanninkhof, Rik and Kramp, Martin and Johnson, Gregory C

Sloyan, Bernadette M. and Wanninkhof, Rik and Kramp, Martin and Johnson, Gregory C. and others , title =. Frontiers in Marine Science , volume =. 2019 , doi =

2019
[15]

, title =

Kalathoor, Sriram P. , title =. 2026 , eprint =. doi:10.48550/arXiv.2606.21131 , url =

2026 doi arXiv
[16]

, title =

Kalathoor, Sriram P. , title =. 2026 , publisher =. doi:10.5281/zenodo.20806266 , url =

2026 doi

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [1]

, title =

Schmitt, Raymond W. , title =. Annual Review of Fluid Mechanics , volume =. 1994 , doi =

1994

[2] [2]

, title =

Ruddick, Barry R. , title =. Deep Sea Research Part A. Oceanographic Research Papers , volume =. 1983 , doi =

1983

[3] [3]

Deep Sea Research Part I: Oceanographic Research Papers , volume =

You, Yuzhu , title =. Deep Sea Research Part I: Oceanographic Research Papers , volume =. 2002 , doi =

2002

[4] [4]

Schmitt, R. W. and Ledwell, J. R. and Montgomery, E. T. and Polzin, K. L. and Toole, J. M. , title =. Science , volume =. 2005 , doi =

2005

[5] [5]

Kelley, D. E. and Fernando, H. J. S. and Gargett, A. E. and Tanny, J. and Ozsoy, E. , title =. Progress in Oceanography , volume =. 2003 , doi =

2003

[6] [6]

Radko, Timour , title =

[7] [7]

and Toole, J

Timmermans, M.-L. and Toole, J. and Krishfield, R. and Winsor, P. , title =. Journal of Geophysical Research: Oceans , volume =. 2008 , doi =

2008

[8] [8]

Shibley, N. C. and Timmermans, M.-L. and Carpenter, J. R. and Toole, J. M. , title =. Journal of Geophysical Research: Oceans , volume =. 2017 , doi =

2017