Physical connectivity and Lagrangian transport patterns in the Bay of Biscay across 30 years

Steiger, Nadine; Manso-Narvarte, Ivan; Caballero, Ainhoa; Rubio, Anna

doi:10.5194/egusphere-2026-4302

Preprints

https://doi.org/10.5194/egusphere-2026-4302

Preprints

24 Jul 2026

| 24 Jul 2026

Physical connectivity and Lagrangian transport patterns in the Bay of Biscay across 30 years

Nadine Steiger, Ivan Manso-Narvarte, Ainhoa Caballero, and Anna Rubio

Abstract. Understanding ocean physical connectivity is essential for characterizing and predicting transport patterns, which in turn play a fundamental role in several marine processes. Here, we present an analysis of 30 years of surface water connectivity for the Bay of Biscay, derived from backward Lagrangian particle simulations driven by hourly high-resolution surface velocity reanalysis fields. The Lagrangian simulations are used to characterize transport pathways within and between key subregions across integration times of 7, 30, and 90 days, thereby capturing connectivity from weekly to seasonal scales.

Results show that the obtained seasonal cycle of connectivity reflects the seasonal variability of regional ocean circulation: i) The northward transport along the shelf by the Iberian–Poleward Current, ii) the presence of transport barriers along the continental slope, and iii) the enhanced cross-shelf transport from the Spanish shelf during spring and from the French shelf during summer. Regional maps of particle origins and transit times reveal areas of strong isolation along the French coast and zones of intensive mixing at the French Spanish border. While interannual variability is evident, the simulations also indicate a slight but significant long-term decrease in transport from the Spanish to the French shelf.

Overall, these results provide an overview of the main pathways of transport and retention within the Bay of Biscay at different time scales, offering insights into its potential role in different key ocean processes such as marine litter or plankton dispersal, genetic exchange, and ecosystem functioning.

Received: 17 Jul 2026 – Discussion started: 24 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Nadine Steiger, Ivan Manso-Narvarte, Ainhoa Caballero, and Anna Rubio

Status: final response (author comments only)

RC1: 'Comment on egusphere-2026-4302', Anonymous Referee #1, 11 Aug 2026

This study is thorough and well designed, and the manuscript is generally very well written.

I recommend publication after considering the comments and suggestions below.

The first sentence in Introduction reads:

“Ocean connectivity is a fundamental process in the marine system.”

However, connectivity is not a fundamental process in itself, but rather the property or outcome of fundamental processes such as transport/advection. Thus this sentence should be reformulated.
Figure 1:
- For readers without connection to the Bay of Biscay area, it might be a little hard to locate this area. I would suggest to include a little more land to the east and west of the given area, and perhaps also a little to the south and north.
- The legend uses the short names “fr_shelf”, “es_slope” etc, which are also defined in the text. These shortnames make sense in the programming/analysis phase, as names of files and figures etc. However, I am not sure if these shortnames are needed in the manuscript, as it might be clearer to simply write explicitly (also in this legend) as “Frensh shelf” and “Spanish slope” etc. This goes throughout the paper. The exception might be Figures 7-9 where compact representation is needed. The shortnames could be defined and used for these figure labels. I will not insist on this change, but it could be considered.
Section 2.2, line 125:

Ocean model resolution is specified as 1/36 degree of longitude and latitude. It would be useful to also say what this refers to in km, approximately since one degree of longitude depends on the latitude.
It is also a general thing throughout the paper that distances are given in degrees of longitude and latitude. Most places (like here) it would be good to also give the corresponding distance in km. E.g. line 128 says that particles are seeded on 0.08 x 0.08 degree grid - however, this gives different spacing northwards and eastwards. This is not a problem in itself, but should be made more clear by specifying also distances in km.
Section 2.2, line 137. I would suggest to rewrite these to sentences to something like:

“The model configuration included no wind drift or diffusion, and a time step of 15 minutes. Particles hitting the coast were not stranded/removed, but left in place to move ashore once the current has an offshore component.”
Section 2.3, line 156: “preceding two months”

Figure 3: Again it might be a little hard to locate the area, as both land and ocean are white. It could be considered to put a color (black?) on land. But this is hard to know without testing, and thus authors can decide what they think is best.
Figure 4: Lines for 200 and 1000 m isobats are very close, indicating that the shelf is quite steep. This could be commented on somewhere.
Section 3.2, line 258: Although ODI is defined earlier, I would suggest to remind the reader that it means Origin diversity index.
Section 3.2, line 268: within “2-5 degrees” is again a little vague, since a degree of longitude is different from a degree of latitude. This is another example where distances in km would be better.
Section 3.4, line 361: a trend of -0.0006 degrees N is given, corresponding to 67 meters per month. This sounds like a very small trend although stated as significant,, and also I would think that Figure 10a shows rather a decline until ~2012 and thereafter an increase, i.e. a long term oscillation if anything.

However, more prominent on Fig 10a is the apparent oscillation on ~5 yers scale. This is quite interesting, and should be commented in the text. Could it be related to El Nino or other known indices/oscillations?
Section 4, line 383: The paper/simulations uses model surface currents without additional wind drift, which is fine. However, “floating marine litter” is here mentioned, and which is known to need additional wind drift which can be quite significant (typically ~3% of wind). Thus, I would agree that results are valid for microplastics (and plankton etc), but I would comment that connectivity including wind drift could be quite different. E.g. floating plastics is also subject to stranding whenever onshore winds/waves.

Citation: https://doi.org/10.5194/egusphere-2026-4302-RC1
RC2:
'Comment on egusphere-2026-4302', Rodrigo Duran, 29 Oct 2026
The authors run backward Lagrangian particle simulations with OpenDrift, driven by hourly surface currents from the 1/36° IBI reanalysis (1993–2024). They divide the Bay of Biscay into five seeding subareas and three open boundaries, and track particles back 7, 30 and 90 days. From these runs they compute particle origins, transit times, an origin diversity index (ODI), connectivity matrices, seasonal cycles and 30-year trends. Their main findings are:
The shelves are strongly retentive, especially the central French shelf.

The slopes behave as dispersive transition zones.

Connectivity has a clear seasonal cycle linked to the Iberian Poleward Current (IPC).

Northward transport of Spanish shelf water onto the French shelf has declined slightly, while transport from the basin toward the Spanish shelf and slope has increased.

A basin-wide connectivity reference for the Bay of Biscay would be valuable for fisheries, marine litter and eDNA work. The writing is mostly clear and the figures are informative.
Several problems need to be addressed before the conclusions can be trusted:
The model currents, and the persistent transport patterns built on them, are not compared with observations.

Backward-in-time results are often described as if they were forward transport. In addition, the number of particles traced back to the Spanish shelf, as implied by Fig. 7 and Table 1, is puzzling and needs explanation.

The p-value reported for a key trend contradicts the figure.

The significance tests for the trends are not described, and the monthly values used are not independent.

Comments 1 and 3–5 address these points and are essential. The other comments would strengthen the paper. There are also several figure-reference errors and leftover editing marks that should be corrected.
Major comments
Validation of the velocity fields and transport patterns

The whole study depends on the IBI surface currents being realistic, but no comparison with observations is shown. The Discussion states that the patterns are consistent with earlier studies, but only qualitatively. The authors themselves acknowledge (lines 410–413) that closer validation against local observations remains essential for smaller-scale applications. This matters because the 1/36° product is very new (released November 2025, line 407).
At minimum, please summarize what the product’s validation documentation (e.g., the CMEMS Quality Information Document) reports for surface currents in the Bay of Biscay. Please also add at least one quantitative comparison with observations. Suitable observations exist:
HF radar. Two stations have provided hourly surface currents in the southeastern Bay since 2009 (Solabarrieta et al., 2014). They measure near-surface currents, closely comparable to the hourly surface fields used here, and they resolve the surface signature of the IPC. A comparison of mean currents and their seasonal cycle in this area would be a direct test.

Drifters. Charria et al. (2013) built a seasonal climatology of Bay of Biscay circulation from drifter trajectories collected between 1992 and 2009. These drifters were drogued below the surface, so a fair comparison needs the (daily) model velocities at drogue depth.

A simple quantitative test is the percentage of model particles and of observed drifters that cross the slope over time, as done by Gough et al. (2019). For a more complete, optional comparison, Maslo et al. (2020) provide a template for connectivity studies:
They validated their model against floats and moorings by comparing mean velocities and velocity variance ellipses.

They then compared connectivity matrices built from real floats with those built from model particles.

As they note, uneven deployment biases observation-based matrices, so subsampling the model particles to match drifter coverage gives a better comparison.

Comment 9 describes a further, efficient way to check the persistent barriers and isolated regions against satellite and drifter observations.
Please also state whether the hourly surface fields include tidal currents and how tides affect the results, and which model depth the “surface” velocity represents.
The physical setup does not match some of the claimed applications

The simulations use no windage, no Stokes drift and zero diffusion (line 137). Omitting windage and Stokes drift neglects components of near-surface transport: Stokes drift affects both near-surface water and floating material, and windage affects floating material.
The paper repeatedly presents its results as relevant to floating marine litter (e.g., lines 20, 383 and 451). The manuscript itself notes (lines 68–71) that fronts modulate the convergence, retention and offshore export of floating material. It also notes that the position and persistence of these fronts are strongly influenced by wind forcing and windage effects (Bertin et al., 2024, 2026). You can either restrict the claims to surface water parcels moving with currents, or add a sensitivity test with windage and Stokes drift.
One efficient way to test this is to compute climatological Lagrangian coherent structures (cLCS; Comment 9) from currents plus windage. Allende-Arandía et al. (2023) did this for the Caribbean:
A coastal barrier present in the currents-only fields was weakened with 1% windage and erased with 2%.

Independent trajectories computed with windage followed the windage-modified structures toward the coast.

Because Stokes drift can be parameterized as a percentage of the wind (Clarke and Van Gorder, 2018), windage-adjusted cLCS can also represent the combined effect of windage and Stokes drift. A similar analysis would show which of the barriers reported here also hold for floating litter, and how they change with wind. How barriers change with wind is a common operational question.

Visualizing transport barriers as a function of wind allows for a transparent discussion of your results. More on this below.
Please also state the time integration scheme used in OpenDrift (e.g., Euler or fourth-order Runge–Kutta), as this can affect comparisons between forward and backward trajectories (Reijnders et al., 2026).
Backward results are interpreted as forward transport, and the particle counts need explanation

This is an important conceptual problem.
(a) Different questions, different denominators. In a backward simulation, C_ij answers: “Of the water now in area i, what fraction came from area j?” It does not answer: “Of the water that started in j, what fraction ended up in i?” Even if both directions share the same set of trajectories, the percentages divide by different totals: particles arriving in i versus particles starting in j. In an area-preserving flow sampled uniformly by particles:
forward fraction ≈ backward fraction × (particles in destination ÷ particles in source)
As an analogy, “63% of passengers landing at a small regional airport came from the hub” does not mean “63% of passengers leaving the hub fly to that airport.”
Several passages mix the two. Line 331 reports “63% connectivity from Spanish shelf to slope in June” as off-shelf transport. The off-shelf transport itself is real. The issue is that 63% is the share of Spanish slope water that came from the shelf, not the share of shelf water that was exported.
Using the 90-day matrix (Fig. 7c) as an example: 49% of Spanish slope water originated on the Spanish shelf. With 241 versus 460 particles per seeding (Table 1), roughly 49% × 241/460 ≈ 26% of Spanish shelf water would reach the slope.
Similar wording appears elsewhere. For example, lines 460–461 use Fig. 4 to infer the offshore dispersion of shelf water; for larvae, the forward question (where do they go?) is the relevant one.
Two cases are less affected:
Retention. For the diagonal of the matrix, the backward and forward fractions coincide in an area-preserving flow.

Comparisons of the same entry. Comparisons of the same matrix entry across seasons or years are less affected, because the denominator is fixed.

The distinction is recognized in connectivity studies. Connectivity is usually defined in the forward sense, as the probability that water from one location reaches another (Cowen et al., 2006). Maslo et al. (2020) explicitly attribute the asymmetry in their east–west percentages to the different sizes of the regions involved.
(b) A puzzling particle count. Multiply each row of Fig. 7 by the number of particles seeded in that subarea (Table 1) and sum the es_shelf column. Compared with the 460 particles seeded on the Spanish shelf, this gives about the following number of particles traced back to the Spanish shelf per seeding:
565 at 7 days;

910 at 30 days;

1,440 at 90 days.

The excess grows with integration time, and no other subarea exceeds its seeded number. If particles sampled the surface uniformly and each were assigned a single origin, one would not expect many more particles to trace back to a subarea than are seeded in it, unless the surface flow there is strongly non-area-preserving.
Please clarify how the matrix is computed. Is the origin the position at the end of the backward integration, or the boundary crossed? Is each row divided by the number of particles seeded in that subarea (see the minor comment on Eq. 1)?
Please explain these counts. Two possibilities worth checking are the coastline setting (coastline_action = “previous”) and the tendency of backtracking to overrepresent divergent zones as sources (Reijnders et al., 2026). If the excess is confirmed, it could also explain why the Spanish shelf appears as such a widespread origin in Fig. 4
If a problem is confirmed, this could also explain why the Spanish shelf appears as such a widespread origin in Fig. 4.
I recommend that the authors:
reword the affected passages in strictly backward terms;

clarify how the connectivity matrix is computed;

report the number of particles found in each subarea at the end of the backward integration relative to the number seeded there, and show where any excess occurs;

report what fraction of particles are next to the coast (e.g., within one grid cell of land) at the end of the backward integration, investigate whether coastline effects are a problem;

if an excess is concentrated at the coast, discuss how it affects the inferred origins (Figs. 4 and 7) and the northward-extent metric (Comment 8);

compare forward and backward trajectories for a subset of seedings, as recommended by Reijnders et al. (2026);

add forward connectivity estimates for questions that are inherently forward, such as larval export and litter fate. These could come from forward runs, for example for a subset of years, or from reweighting the backward results, with the caveat that reweighting assumes area preservation and uniform sampling.

If you suspect area is not preserved, the Jacobian determinant of the flow map, as used by Duran et al. (2018, Supporting Information) and Gouveia et al. (2021, 2025), would quantify how much the velocity field itself expands or compresses surface water along the trajectories.
Contradictory p-value

Line 361 reports a significant trend in the northward extent of Spanish shelf water (p = 0.015). Figure 10 prints p = 0.0876 for the same trend, which is not significant at the 95% level. This metric is presented as supporting the long-term decline reported in the abstract and conclusions, so the discrepancy must be resolved. Please also state which series each p-value refers to (e.g., monthly values or seasonal-year means).
The panel labels in Figure 10 also do not match the caption or the text. In the figure, (a) is the yearly anomaly time series, (b) the seasonal cycle and (c) the Hovmöller diagram. The caption and text use a different order.
Trend significance may be overstated

Several issues affect the trend analyses (Figs. 9 and 10).
Test not described; values not independent.

The Methods do not state which significance test was used.

Consecutive monthly values are built from overlapping forcing periods: for the 90-day runs, consecutive months share about 90 of roughly 120 days of forcing; for the 30-day runs, about 30 of 60.

If an ordinary regression t-test on the several hundred monthly values was used, significance is overstated.

Please state the test, and either account for autocorrelation or test annual means.

Multiple testing. Fig. 9 shows about 120 trend tests. At the 95% level, about six would appear significant by chance even if the tests were correct, so isolated significant cells should be interpreted with caution.

Reanalysis artifacts. Reanalyses can show spurious trends when the observing system changes, such as the arrival of Argo floats in the early 2000s or changes in altimetry missions. Please discuss this, and ideally check whether the trends hold over sub-periods or appear in the Eulerian velocity fields.

A dip, not a steady decline. The authors note a strong minimum in 2012/13 followed by partial recovery. A linear trend over the whole period may be driven mainly by that event. Does reporting the trend over sub-periods bring clarity?

The trends are small compared with the variability. For example, −0.0006° per month amounts to about −0.2° over 30 years, against monthly anomalies of about +-1° (line 360). Are such changes are physically or ecologically meaningful?
The proposed IPC mechanism needs reconciling and testing

The Discussion (lines 415–418) explains shelf retention by the IPC acting as a barrier along the slope. Fig. 8 does show the lowest contributions from the shelves to the basin in winter, consistent with a slope barrier. However, shelf retention peaks in May–June (lines 324–325), when the IPC is weak, presumably because in winter the IPC also supplies the shelves with water from the south. Please reconcile these two aspects in the Discussion.
The authors note that linking the trends to a weakening IPC (lines 438–441) requires further analysis. This could be tested directly by computing an Eulerian IPC index (e.g., mean along-slope velocity off the Cantabrian coast) from the same reanalysis and correlating it with the Lagrangian metrics. Alternatively, cLCS from two sub-periods (e.g., 1993–2012 and 2013–2024) could also be compared (see comments on cLCS below).
Results depend on subarea definitions

The authors acknowledge this (lines 426–436), but the effects deserve more attention.
Size and shape. The deep basin holds about 20 times more particles than each slope area (Table 1). High basin retention and low slope retention partly reflect size and shape, not only dynamics. A simple null model, such as the retention expected from random motion with matched dispersion, would help separate geometry from dynamics. Maslo et al. (2020) make the same point about their own connectivity analysis.

The “Spanish shelf” mixes two regimes. It extends south to 41° N and includes the Galician and northern Portuguese coasts (lines 116–117), so it combines the western Iberian upwelling coast with the Cantabrian shelf. This affects the interpretation of the “northward extent of Spanish shelf water” metric. Splitting it, or renaming it “Iberian shelf,” would help.

Boundary crossing. Lines 107–108 state that particles crossing the 48.5° N, 41° N or 11° W limits are assigned to the respective boundary. Please clarify whether this uses the first crossing, and whether particles can re-enter the domain. If the origin is instead taken as the position at the end of the backward integration, particles that leave and re-enter would be labeled as interior water. Figure 3 shows trajectories extending into the Channel and west of 11° W.

Comment 9 describes a method that identifies barriers and isolated regions without predefined subareas.
Robustness of the northward-extent metric

The 99.9th percentile of latitude is described as robust to outliers (line 191). However, it is set by the top 0.1% of particles, i.e., between a few and about ten particles per monthly simulation, depending on how particles are counted (Comment 3b).
In addition, if the metric uses seeding positions, it cannot exceed 48.5° N, the northern limit of the seeding area. Around November, when the climatology peaks at about 47.8° N (line 359), positive anomalies would then be capped.
Please state which positions are used, test the sensitivity to the percentile chosen (e.g., 95th or 99th), and discuss any cap.
Suggestion: compare the persistent patterns with climatological Lagrangian coherent structures and observations

Disclosure: I am a cp-author of several of the cLCS studies cited here. These comments are suggestions, not a condition for acceptance.
The authors infer barriers, isolated regions and accumulation zones indirectly, from subarea-based percentages and the ODI. Climatological Lagrangian coherent structures (cLCS; Duran et al., 2018) identify these features directly from the flow, without predefined subareas. The method has three steps:
Average the velocity into a daily climatology.

Compute 7-day backward strain (Cauchy–Green) tensors.

Average those tensors monthly and extract the most attracting lines, together with a map of attraction strength (cρ).

The results do not depend on the 7-day integration time: the barriers and pathways stay the same, and only the attraction strength increases with longer integrations.
Low-pass filtering (the climatological velocity) is part of the design, because the method targets the persistent component of transport. Duran et al. (2018) showed that the ensemble-mean Lagrangian transport patterns of the unfiltered velocity (i.e., without climatological averaging) are well explained by the cLCS computed from the climatological velocity. The climatological velocity therefore preserves meaningful Lagrangian information when it is analyzed with suitable tools.
For this study, the relevant point is that low cρ marks weak attraction. Together with the particle results, this can help identify stagnant or isolated water (regions of low cρ).
The software is freely available in Python (https://github.com/MireyaMMO/cLCS) and Matlab (https://bitbucket.org/rodu/clcss/src/master/). It is automated, and the computation typically takes a few hours on a laptop, even for larger domains. All the methods need is a multi-year horizontal velocity time series,
The method and a close variant have been applied with HYCOM, NEMO and ROMS simulations, both free-running and data-assimilative. Applications cover boundary currents, shelves, monsoon systems, semi-enclosed seas and the deep ocean. Each study compared the results with tracers, drifters, floats or other independent observations:
Gulf of Mexico (Duran et al., 2018; 12-year data-assimilative HYCOM).

Ensemble-mean tracer advection with the unfiltered velocity over all years was organized by the cLCS.

The cLCS agreed with Deepwater Horizon and Ixtoc oil-spill observations, including beaching records, and with GLAD drifters.

Northwestern Gulf of Mexico (Gough et al., 2019; 18-year free-running NEMO at 1/36°).

The cLCS were compared with ensemble tracer densities from daily releases over all years.

They were also compared with historical drifters from several programs.

The percentage of drifters crossing the shelf-break barrier was quantified after 3 to 28 days.

Deep Gulf of Mexico (Maslo et al., 2020; free-running ROMS at about 5 km).

The model was first validated against deep floats and moorings, and connectivity matrices from floats and model particles were compared directly.

A variant of the method used strain tensors from 20-day backward integrations of the unfiltered velocity, averaged over five years.

This variant explained the low east–west connectivity seen in both data sets: a deep gyre forms an enclosed region, with exchange limited to the boundary currents. It also shows that the integration window can be lengthened for slower flows.

Brazil Current (Gouveia et al., 2021; 13-year ROMS at 1/36°). The cLCS were compared with:

satellite SST;

drogued drifters (with probability density estimates), synthetic drifters and iSphere drifters;

the satellite outline of the 2011 Frade oil spill and 2019 oil-beaching records.

Yucatan shelf (Kurczyn et al., 2021).

The model was first checked against moored ADCPs, then combined with CTD surveys and back-trajectories.

The cLCS supported recurrent westward transport also seen in a 39-year satellite SST record.

Equatorial Atlantic to Caribbean (Allende-Arandía et al., 2023; 25-year data-assimilative HYCOM climatology).

A particle model calibrated with Global Drifter Program drifters was used, and the densities of about 1.9 million particles followed the cLCS pathways and barriers.

The cLCS were recomputed with 1% and 2% windage.

Independent windage-forced trajectories from the unfiltered HYCOM velocity moved along the windage-adjusted cLCS toward the coast.

Bay of Bengal (Trinadha Rao et al., 2024; 24-year global HYCOM, an independent group).

Of the drifters, two followed the cLCS closely and the rest partly; mismatches were attributed to submesoscale eddies.

For a satellite (SAR) oil slick, the cLCS captured the destination though not the exact path.

The cLCS were also compared with a freshwater plume in satellite salinity and with oil-model particle trajectories.

Black Sea (Gouveia et al., 2025; Copernicus NEMO reanalysis at about 1/27° × 1/36°).

The boundary-current position was checked against satellite altimetry.

The cLCS were compared with satellite SST and ocean-color tracers (coccolithophore blooms, particulate inorganic carbon, chlorophyll), which outlined the identified fronts, barriers and recurrent eddies.

The Jacobian diagnostic mentioned in Comment 3 was applied.

Across these studies, agreement is strongest for persistent transport. Discrepancies appear mainly in submesoscale-dominated areas, but note that wind can change the barriers (Allende-Arandía et al., 2023). The Black Sea application used a Copernicus NEMO reanalysis comparable to IBI, in a basin with strong shelf–slope interaction and coastal accumulation questions.
Computing cLCS from the IBI velocities would have different uses:
Independent check against observations. Comparing cLCS with satellite SST fronts, ocean-color features, HF radar currents and the drifter data would be an efficient test of the persistent barriers, isolated regions and accumulation zones reported here. Comparing cLCS from IBI with the authors’ own particle results would only test consistency, since both use the same velocities, although cLCS would provide a different means of visualization and generalization of trajectory results. Comparisons with observations would be an independent evaluation.

Testable expectation. As noted above, ensemble-mean transport under the unfiltered velocity is organized by cLCS. Where the boundaries between origins in Fig. 4 are sharp and persistent, neighboring water on either side traces back to different subareas, i.e., it separates in backward time. Such boundaries are expected along attracting structures, toward which water from different origins converges. They should therefore coincide with strongly attracting cLCS, and a systematic mismatch along them would indicate a problem worth investigating.

Coastal accumulation. Gouveia et al. (2021) found that the 2019 oil was more likely to re-beach where coastal attraction was highest. Similar results were reported in Duran et al. (2018). Both studies identified coastal regions of maximum attraction as likely accumulation zones. A coastal attraction map for the Bay of Biscay could similarly be compared with the litter accumulation data the authors already cite. Their backward connectivity framework cannot easily answer this question.

Windage and Stokes drift. Computing cLCS from currents plus windage, as in Allende-Arandía et al. (2023), would show which barriers persist for floating material and how they shift with wind (Comment 2).

Interpretation requires care. As Kurczyn et al. (2021) explain, averaging the strain tensors formally removes the strict barrier property, and chevron-shaped cLCS mark jets rather than barriers. Strongly attracting cLCS can mark barriers or recurrent pathways. The comparison should also be read within its scope: cLCS address the seasonal claims about barriers, isolation and accumulation.
The ODI is a weak measure of mixing

The ODI counts how many different origins appear in a grid cell. A single particle from a rare origin raises it as much as a large share would, so it depends on particle numbers and subarea geometry. Each 0.12° cell contains only about two seeding positions on the 0.08° grid, i.e., about 15 particles per month. Whether a rare origin is counted therefore depends heavily on sampling.
At 7 days the ODI mostly reflects distance to subarea borders, as the authors note. A weighted measure, such as the Shannon entropy of the origin fractions, would represent mixing better.
The ODI is averaged over years, which should normally give non-integer values, yet Fig. 6 shows discrete classes. Please state whether the averaged values were binned or rounded. The explanation for a maximum of 6 rather than 8 (lines 295–297) also needs clarifying: Fig. 4 shows the most frequent origins, which does not explain a count of unique origins.
Errors in describing the time windows

Line 154 states that 7- and 30-day trajectories are influenced only by circulation in the seeding month. This is incorrect. For example, a particle seeded on the 1st and tracked back 7 days lies entirely in the previous month.
The 90-day case is described correctly (lines 155–156), but the seasonal labels should reflect it: the 90-day “February (winter)” maps largely reflect November–January circulation.
Please state these lags clearly, and consider three-month seasonal composites instead of single representative months.
Transit time definition is unclear

In a backward run, “first entry into each subarea” (line 162) is ambiguous. It could mean the first time the particle is found in the subarea as it is tracked backward from its seeding position, which in real time is the last time the particle was in that subarea. It could also mean the earliest time the particle entered the subarea in real time. The two differ for particles that visit a subarea more than once. Please state which is used.

The mean is also computed only over particles that reach the subarea within 90 days, so it is a conditional mean. For distant sources it reflects only the fastest particles. Please define the metric precisely and state that it is conditional.
Inconsistent retention numbers

Lines 458–459 state that about 50% of shelf particles remain on the shelf after one month and cite Fig. 5. However, Fig. 5 shows 90-day percentages, and Fig. 7b gives 72% shelf retention at 30 days. Please clarify where the 50% comes from.
Minor comments
Eq. 1 sums over i (seeding subareas) in the denominator, which would normalize by origin. The text (lines 177–179) says the normalization is by the number of particles seeded in subarea i, so the sum should run over j. The rows of Fig. 7 sum to about 100%, so this is probably a typo, but please confirm the denominator (see Comment 3b).

Leftover editing marks: “[NS5.1]” (line 31), “[NS6.1]” (line 41), “[NS7.1][NS7.2]” (line 54).

Lines 245–246: “At 30 years… at 7 years” should be “days.”

Lines 242–243: Fig. 4i–l shows the third-most frequent origin, not the second.

Lines 288 and 293: “Fig. 5b, e, h, k” and “Fig. 5c, f, i, l” should refer to Fig. 6.

Line 290: “southern French shelf in winter” should probably be “autumn,” since winter already opens the sentence.

Line 438: the increase in basin-to-shelf transport is a trend result shown in Fig. 9, not Fig. 8.

Line 346: the text gives 0.008% per month for basin to Spanish shelf, but Fig. 9c shows 9 (×0.001). This may be a rounding difference.

Lines 348–349: negative trends need a minus sign (−0.006, −0.007, etc.).

Line 347 is garbled (“they to total increases… accumulate”), as is the Fig. 9 caption (“between each for”).

Line 133: “7.958” and “55.706” should be “7,958” and “55,706.” Table 1 sums to 7,957, not 7,958.

Line 162: “seeing time” should be “seeding time.”

Line 273: I could not find “Supplementary Material, Fig. 16”.

Line 45: “At the open ocean” should be “In the open ocean.”

Lines 49–50: the narrow-shelf and limited-river-discharge rationale applies to the Spanish shelf. The French shelf is wide and receives the Loire and Gironde (see also line 105).

Abstract (line 15): the abstract says French shelf cross-shelf export is enhanced in summer, while the results give September as the peak (line 332). It also says Spanish shelf export is enhanced in spring, while line 331 gives early summer (June).

Line 132: monthly simulations “for every month” from 1993 cannot include the first months, given the 120-day backward runs. Please state the exact period used for climatologies and trends.

References: there are many errors with your references please review carefully spelling, matching citations and references, etc, and consider:

Amaro et al. (2016) concerns the Whittard Canyon on the Celtic margin, a questionable source for Bay of Biscay bathymetry.

Lett et al. (2010) describes a Lagrangian modeling tool but is cited at lines 28–29 and 77 for statements about fronts and the biogeography of the northern boundary. Please check these citations.

Figs. 4, 6, A4 and A5: purple (French slope) and yellow (Spanish slope) are hard to read in small clusters. Perhaps a colorblind-friendly palette would help?

Fig. 7: add a measure of variability (e.g., standard deviation across years).

A quantitative comparison with existing connectivity estimates for the region would help place the results, e.g., the global connectivity dataset of Assis et al. (2025) or larval dispersal estimates such as Ayata et al. (2010).

References
Allende-Arandía, M. E., Duran, R., Sanvicente-Añorve, L., and Appendini, C. M. (2023). Lagrangian characterization of surface transport from the Equatorial Atlantic to the Caribbean Sea using climatological Lagrangian coherent structures and self-organizing maps. J. Geophys. Res. Oceans, 128, e2023JC019894. https://doi.org/10.1029/2023JC019894

Bertin, S., Rubio, A., Hernández-Carrasco, I., Orfila, A., and Sentchev, A. (2026). The impact of windage on the transport and accumulation of marine litter: A Lagrangian analysis in the Bay of Biscay. J. Mar. Syst., 255, 104209. https://doi.org/10.1016/j.jmarsys.2026.104209

Charria, G., et al. (2013). Surface layer circulation derived from Lagrangian drifters in the Bay of Biscay. J. Mar. Syst., 109–110, S60–S76. https://doi.org/10.1016/j.jmarsys.2011.09.015

Clarke, A. J., and Van Gorder, S. (2018). The relationship of near-surface flow, Stokes drift and the wind stress. J. Geophys. Res. Oceans. https://doi.org/10.1029/2018JC014102

Cowen, R. K., Paris, C. B., and Srinivasan, A. (2006). Scaling of connectivity in marine populations. Science, 311, 522–527. https://doi.org/10.1126/science.1122039

Duran, R., Beron-Vera, F. J., and Olascoaga, M. J. (2018). Extracting quasi-steady Lagrangian transport patterns from the ocean circulation: An application to the Gulf of Mexico. Sci. Rep., 8, 5218. https://doi.org/10.1038/s41598-018-23121-y

Gough, M. K., Beron-Vera, F. J., Olascoaga, M. J., Sheinbaum, J., Jouanno, J., and Duran, R. (2019). Persistent Lagrangian transport patterns in the northwestern Gulf of Mexico. J. Phys. Oceanogr., 49, 353–367. https://doi.org/10.1175/JPO-D-17-0207.1

Gouveia, M. B., Duran, R., Lorenzzetti, J. A., Assireu, A. T., Toste, R., Assad, L. P. de F., and Gherardi, D. F. M. (2021). Persistent meanders and eddies lead to quasi-steady Lagrangian transport patterns in a weak western boundary current. Sci. Rep., 11, 497. https://doi.org/10.1038/s41598-020-79386-9

Gouveia, M. B., Peneva, E., Gancheva, I., and Stanev, E. V. (2025). Persistent Lagrangian transport patterns in the Black Sea identified from climatological Lagrangian coherent structures. Sci. Rep., 15, 20006. https://doi.org/10.1038/s41598-025-00432-5

Kurczyn, J. A., Duran, R., Beier, E., and Souza, A. J. (2021). On the advection of upwelled water on the western Yucatan Shelf. Front. Mar. Sci., 8, 723452. https://doi.org/10.3389/fmars.2021.723452

Maslo, A., Azevedo Correia de Souza, J. M., Andrade-Canto, F., and Rodríguez Outerelo, J. (2020). Connectivity of deep waters in the Gulf of Mexico. J. Mar. Syst., 203, 103267. https://doi.org/10.1016/j.jmarsys.2019.103267

Reijnders, D., Denes, M. C., Rühs, S., Breivik, Ø., Nordam, T., and van Sebille, E. (2026). Stability bias in Lagrangian (back)tracking in divergent flows. J. Adv. Model. Earth Syst., 18, e2025MS005470. https://doi.org/10.1029/2025MS005470

Rubio, A., Caballero, A., Orfila, A., Hernández-Carrasco, I., Ferrer, L., González, M., Solabarrieta, L., and Mader, J. (2018). Eddy-induced cross-shelf export of high Chl-a coastal waters in the SE Bay of Biscay. Remote Sens. Environ., 205, 290–304. https://doi.org/10.1016/j.rse.2017.10.037

Solabarrieta, L., Rubio, A., Castanedo, S., Medina, R., Charria, G., and Hernández, C. (2014). Surface water circulation patterns in the southeastern Bay of Biscay: New evidences from HF radar data. Cont. Shelf Res., 74, 60–76. https://doi.org/10.1016/j.csr.2013.11.022

Trinadha Rao, V., Suneel, V., Gulakaram, V. S., and Sravani, C. L. (2024). Extraction of persistent Lagrangian coherent structures for the pollutant transport prediction in the Bay of Bengal. Sci. Rep., 14, 8761. https://doi.org/10.1038/s41598-024-58783-4

Rodrigo Duran

Planetary Science Institute
Citation: https://doi.org/10.5194/egusphere-2026-4302-RC2
EC1:
'Comment on egusphere-2026-4302', Erik van Sebille, 02 Nov 2026
Publisher’s note: this comment was edited on 4 November 2026. The following text is not identical to the original comment, but the adjustments were minor without effect on the scientific meaning.
Review of “Physical connectivity and Lagrangian transport patterns in the Bay of Biscay across 30 years: by Steiger et al
This is a very interesting and carefully-written manuscript that I think could fit well within Ocean Science.
My major points are
While there is some reflection about more physics-informed clustering of the different subregions in the Discussion section, the definitions of these subregions in section 2.1 seem rather ad-hoc and top-down. It would be good to consider or flag more carefully how sensitive the results are to the definitions of the subregions. And perhaps also mention alternative ways to define the subregions earlier on. The work by Vincent Rossi and Enrico Ser-Giacomi in the Mediterranean comes to mind, for physics-based clustering.

I find the definition of the ODI (lines 169-174) very difficult to follow. Would it help to add an equation here too, like in the Connectivity matrix? Or otherwise add more words to explain it more carefully?

I find the argument to use the shelf instead of the slope “due to the stronger signal in connectivity and its trend” rather weak. This suggests you look for maximal result (cherrypicking?), instead of an oceanographically-relevant result. Could you phrase this motivation it terms of oceanographic requirements, instead of the outcomes from the data?

The figures could all use a legend for the colours, so that readers don’t have to flip scroll back to Figure 1

Other minor points
Line 8-11: Also briefly mention the relevance of the Bay of Biscay in the abstract?

Line 31 and other places: what do [NS5.1] and similar words refer to?

In the caption of figure, perhaps make clearer that these are not three sets of simulations for the different integration lengths – but rather that the analysis uses different subsets of the full simulation data.

Line 183: ij should be subscript here?

Line 191: what does “remaining robust” mean here? Be more specific?

In Figure 3, randomize the lines tso that some regions are not completely on top of others?

Line 281: should this be “dashed” instead of “ligh grey”?

Line 383: I don’t think I fully follow the argument here. Why would the surface flow be better constrained? Since waves and wind complicate surface flow, I would argue this is even poorer constrained than the sub-surface (geostrophic) flow

Line 521: I strongly suggest that you upload the code now – so that this can also be checked for the next iteration of reviews
Citation: https://doi.org/10.5194/egusphere-2026-4302-EC1

Nadine Steiger, Ivan Manso-Narvarte, Ainhoa Caballero, and Anna Rubio

Viewed

Total article views: 481 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
277	164	40	481	32	31

HTML: 277
PDF: 164
XML: 40
Total: 481
BibTeX: 32
EndNote: 31

Views and downloads (calculated since 24 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	68	14	10	92
Aug 2026	25	21	6	52
Sep 2026	58	43	9	110
Oct 2026	78	69	9	156
Nov 2026	48	17	6	71

Cumulative views and downloads (calculated since 24 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	68	14	10	92
Aug 2026	25	21	6	52
Sep 2026	58	43	9	110
Oct 2026	78	69	9	156
Nov 2026	48	17	6	71

Viewed (geographical distribution)

Total article views: 456 (including HTML, PDF, and XML) Thereof 456 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 06 Nov 2026

Short summary

Ocean currents transport water, marine life, nutrients, and pollution. Using computer simulations, we studied the movement and origin of surface waters in the Bay of Biscay. We found seasonal offshore transport of coastal waters, relatively isolated waters on the French shelf, and strong exchange from the Spanish to the French shelf that has gradually weakened over time. These results improve our understanding of the transport of organisms, marine litter, and other materials in the region.


Total:	0
HTML:	0
PDF:	0
XML:	0