the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
SWOT fast-sampling observations of topographically modulated coastal-eddy propagation in the Algerian Basin
Abstract. The detailed evolution of coastal mesoscale eddies propagating along continental margins is difficult to observe with conventional nadir altimetry due to the limited spatio-temporal sampling. Here we use observations from the one-day repeat fast-sampling phase of the high-spatial-resolution Surface Water and Ocean Topography (SWOT) mission to examine the propagation of a small anticyclonic coastal eddy along the Algerian Coast in the Western Mediterranean Sea in May–June 2023. SWOT sea level anomaly fields are compared with conventional altimetry, sea surface temperature, and chlorophyll-a observations to assess the eddy position, structure, and evolution. SWOT resolves a more realistic nearshore eddy structure, leading to an improved characterization of its eastward displacement compared to conventional gridded altimetry, particularly during the eddy interaction with capes and steep bathymetric transitions. Ellipse-based centroid tracking indicates localized apparent propagation-speed maxima of about 20 cm s-1, whereas conventional gridded altimetry produces smoother and weaker velocities, generally below 10 cm s-1. These speed maxima coincide with a narrowing continental shelf and sharper shelf-break slopes, suggesting kinematic variability organized by the local coastal topography. This case study demonstrates the value of SWOT fast-sampling observations for resolving coastal eddy deformation and event-scale kinematic variability, highlighting new opportunities for understanding the interaction between small mesoscale eddies and coastal topography.
- Preprint
(2227 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3276', Anonymous Referee #1, 13 Jul 2026
-
RC2: 'Comment on egusphere-2026-3276', Anonymous Referee #2, 10 Nov 2026
The manuscript uses SWOT KaRIn data from the 2023 one-day repeat (fast-sampling) phase to track a small anticyclonic eddy along the Algerian coast from early May to mid-June. The SWOT sea level anomaly (SLA) fields are compared against the gridded L4 altimetry product, sea surface temperature (SST) and chlorophyll-a. To track the eddy in SWOT, the authors fit an ellipse to a selected SLA contour on each available day and follow the centre of the ellipse. They find two brief peaks in apparent alongshore speed of about 20 cm s⁻¹ near Capes Caxine and Matifou, where the shelf narrows and the slope steepens, whereas the speed derived from the gridded product remains below about 10 cm s⁻¹. They conclude that coastal topography modulates the propagation of the eddy and that SWOT captures rapid kinematic adjustments that gridded altimetry smooths out.
General assessment
The topic is of clear interest. Coastal eddies are hard to observe with nadir altimetry, and the one-day repeat phase of SWOT offers a rare chance to follow an individual eddy from day to day. The case is well chosen, and the comparison with the gridded L4 product is useful for anyone working with both datasets. A solid study of this event would be a welcome contribution.
In its current form, however, the analysis does not support the main conclusions of the paper. I have four main concerns.
1. The method cannot be reproduced or properly evaluated from the text alone. The spatial filtering, the choice of the SLA contour, the boundary reconstruction, the ellipse fit and the speed calculation are all compressed into one short paragraph (L88–104). The Code and data availability statement mentions Zenodo without giving a link, while the DOI appears only in the reference list. I looked up the repository (10.5281/zenodo.20445357) and went through the notebooks. They clarify how the analysis was done, but they also show that several key settings were tuned by hand for each day and are not mentioned in the manuscript (details below).
2. The claimed scope is broader than the actual analysis. The eddy is framed as a test case for a broader observational problem (L71–74), yet the analysis covers a single eddy over five weeks. The paper does not discuss what carries over to other regions, to other latitudes and eddy scales, or to the nominal 21-day repeat orbit of SWOT.
3. The central finding in the title is not convincingly demonstrated. The link between the speed peaks and the shelf geometry rests on just two speed values and a visual comparison. No statistical test or uncertainty estimate is provided. In addition, the code shows that both speed peaks fall on intervals where the manually chosen tracking settings change abruptly from one day to the next.
4. Structure and clarity. The Results section mixes data description with physical interpretation, and several claims in the Abstract and Discussion go beyond what the results actually show. In particular, the key paragraph of Sect. 3.3 (L155–171) is very difficult to follow.
For these reasons, I recommend a major revision.
Specific comments
Introduction and aims
1. Aims (L66–74). The stated goal shifts from tracking one eddy to addressing "a broader observational problem", which makes it hard to tell what specific questions the paper sets out to answer. Please state the aims as concrete questions, for instance whether SWOT captures the position and shape of the eddy more accurately than the gridded product, and whether its propagation speed changes near the capes. Then, in the Discussion, please explain what can be generalised to other settings, including other latitudes (where eddy scales and the km size of degree grids differ) and the nominal 21-day orbit.
Data and methods
2. Data (L77–87). Please give exact product names, versions and DOIs in Sect. 2, and specify which SWOT variable is used (the code reads ssha_filtered). Please also define how SLA is referenced, including the reference surface and the daily spatial mean subtracted in the code (in the ellipse-fitting notebook, this mean is removed from the May fields but not from the June ones). The bathymetry dataset is not named anywhere in the text (the code uses GEBCO at 30 arc-seconds). In addition, please describe how SWOT samples the eddy: which passes cross it (016 and 003 in the code), which dates were usable, and how long the intervals between consecutive observations are (1 to 4 days). Finally, horizontal resolutions are given partly in degrees and partly in km (1/16°, 2 km, 7 km). Please give all of them in km at this latitude so that the grids can be compared directly.
3. Eddy detection on SWOT (L88–104). This method is at the heart of the paper. If it has been published before, please cite the reference. If it is new, it requires a complete description and a sensitivity test.
- Filter. Why was a second-order Butterworth filter chosen, and why a cutoff of 33 km? How is the filter applied across the nadir gap and next to the coast? In the code, the cutoff is set to 0.03 cycles per grid point, which on a 2 km grid corresponds to a wavelength of about 67 km, not 33 km. Please check this calculation. A filter of this size removes much of the fine detail that SWOT provides for an eddy roughly 50 km across.
- Contour. How is "a chosen SLA level representative of the eddy boundary" selected, and why use a tolerance band of ±5 mm? In the code, the contour level is chosen by hand for each day, ranging from 0.005 to 0.15 m, along with day-specific spatial cuts and alpha-shape parameters. Tracking requires an objective criterion (for example, the outermost closed contour or the contour of maximum geostrophic speed) together with a sensitivity test (see also comment 10).
- Boundary and ellipse. Please explain how the incomplete boundary is "reconstructed" (the code uses an alpha shape). Also note that the ellipse is fitted in degrees of longitude and latitude rather than in metric coordinates, which distorts the geometry at 37° N.
- Methodological figure. A figure illustrating the processing steps for one or two dates (filtered field, selected points, reconstructed boundary and fitted ellipse) would make the procedure much easier to follow. Note that Fig. 3G–L shows SLA contours, not the fitted ellipses.
4. Two methods for two datasets (L88–92). For readers unfamiliar with PyEddyTracker, please add a sentence or two explaining how it detects and tracks eddies. More importantly, because SWOT and the gridded L4 product are analysed with different tracking methods, differences in estimated speed may come from the method rather than the data. Applying the same method to both datasets, or running the ellipse fit on SWOT fields smoothed to the L4 resolution, would separate the two effects.
5. Speed (L99–104). "Apparent alongshore speed" needs a precise definition. In the code, it is computed as the total distance between consecutive centres divided by the time step, with two values made negative by hand. Because this includes cross-shore displacement, it is a total speed rather than an alongshore component. Please project the motion into alongshore and cross-shore components, state the time interval of each estimate and provide an uncertainty.
Results
6. Sect. 3.1, Fig. 2. The comparison with SST and chlorophyll-a relies on visual inspection of two dates, and the text itself calls it qualitative (L107–108). Terms such as "validation" (L108), "very good correspondence" (L109) and "close agreement" (L176–177) are therefore too strong. Please either quantify the agreement across more cloud-free dates (for example, from the offset between SLA contours and tracer fronts) or tone down the wording. For the small-scale features near the coast, please show that they match coherent structures in the tracers and are not noise, given that SWOT errors are larger near the coast (L61–62). A direct comparison of the SWOT and gridded SLA fields on those dates would also help.
7. Sect. 3.2, Fig. 3. Claims that the eddy "loses intensity", "regains intensity" (L123–130) or shows "a more detailed representation" (L130–133) rest solely on the colours of the maps. The plotted contour ("maximum SLA minus 3.5 cm") is an indirect proxy for gradient strength, and it is not the contour used to fit the ellipses. To support claims about eddy intensity, please compute geostrophic velocities and show time series of amplitude, radius and maximum rotational speed for both products. It would also help to open Sect. 3 with a brief overview of the eddy (first detection, lifetime and overall path).
8. Fig. 4. The caption does not say what the figure is meant to show. The colour bar is labelled with calendar dates, whereas the caption states that it shows the absolute difference in days from 26 May. In addition, the source of the black contour is not stated, and the open circles are not explained. While the figure shows why the gridded map of 26 May is degraded in position and shape, it does not explain the drop in amplitude, because SWOT also shows a weaker eddy on that day (L127–128). How the gridded L4 product is mapped from along-track data should be described in Sect. 2.
9. Speeds, Fig. 6 (L148–154). First, the text refers to "Figure 5A" and "Figure 5B", which are actually Figs. 5 and 6. Second, each speed estimate spans between 1 and 4 days, yet the points are plotted at interval midpoints and connected by straight lines, which gives the one-day peaks more visual weight than the multi-day slow periods. Since both series track the same eddy over the same period, they should yield a similar net displacement. Please show this, for instance by plotting cumulative distance or centre longitude against time for both products. If the total distances are similar, SWOT shows short jumps followed by near-stagnation rather than faster propagation overall, which makes "weaker velocities" in the Abstract misleading. Likewise, the claim that SWOT "more effectively captures" the kinematics (L153–154) is not demonstrated. Higher spatial resolution in the raw fields does not by itself prove that the differences between ellipse centres give a more accurate estimate of eddy translation.
10. The two speed peaks coincide with manual changes in the method. In the code, the contour level, the spatial cropping bounds and the alpha parameter are adjusted by hand for each day. Both speed peaks fall on intervals where these settings change markedly from one day to the next. On 24–25 May (21.7 cm s⁻¹), the chosen contour level drops from 0.035 to 0.005 m and a new longitude cut is applied. On 31 May–1 June (about 19 cm s⁻¹), the analysis switches from pass 016 to pass 003, the contour level jumps from 0.016 to 0.12 m, and the June fields are filtered without removing the daily spatial mean. A different contour yields a different ellipse, and a shift of its centre by 10 km within one day already appears as a speed of about 12 cm s⁻¹. While part of the signal may be physical, the paper does not mention these manual adjustments, and its main conclusion rests on these two values. Please document all parameter choices, explain why they were needed, and show that the speed peaks remain when a single, consistent rule is applied to all days.
11. Speed and bathymetry (L155–171). This paragraph carries the main result of the paper, yet it is hard to follow and the link is not convincingly shown. Because shelf width and slope steepness appear only as circle colour and size along an irregular time series, the reader can neither read off the values nor judge how well they relate to speed. Please provide the values in a table, or plot speed directly against each bathymetric parameter with a correlation coefficient and its p-value. As it stands, the conclusion rests on two peaks out of 21 estimates (see comment 10), while several other intervals with a narrow shelf or steep slope show low speeds (for example 25–29 May, right after the first peak). The physical mechanism also needs to be set up more clearly. The Introduction lists general factors of eddy–slope interaction (L36–43) and cites Sutyrin et al. (2003), but the paper never specifies which mechanism is expected here, in which direction it should alter the propagation speed, or how the ratio of eddy radius (about 25 km) to shelf width (2–14 km) comes into play. In addition, the two bathymetric proxies are not defined in Sect. 2 (in the code, the coast is taken as the 20 m isobath, each coast point is paired with the 200 m isobath at the nearest longitude, and the value is read at the longitude of the eddy centre). Finally, L142–143 place the eddy near Cape Caxine around 29 May, whereas L159–162 place it there around 24 May.
Discussion
12. L186–188. The fact that daily SWOT observations exist only for the fast-sampling phase is an important limitation that should be stated in Sect. 2 and linked to the aims. In the same paragraph, the claim that uncertainties in coastal geophysical corrections were "probably weak due to the low tidal signal and relatively dry weather" is not supported by anything in the paper. Please either provide evidence (such as tidal amplitudes and wet-troposphere correction values) or remove the statement.
13. L195–203. The comparison with gap-filling approaches (de Marez et al., 2026) is not backed by any test on these data. Moreover, the code fills missing values by nearest-neighbour interpolation before applying the spatial filter, so the claim that the present method "preserves the native structure of the observed anomaly" needs to be qualified.
Code availability
14. Please give the Zenodo DOI directly in the Code and data availability statement. Please also make sure that the code and the text describe the same procedure. At present, several settings in the code, including the filter cutoff, the day-by-day contour levels and the speed calculation, either differ from the manuscript or are not described in it (comments 3, 5 and 10).
Citation: https://doi.org/10.5194/egusphere-2026-3276-RC2
Interactive computing environment
Codes used to produce the results of the article "SWOT fast-sampling observations of topographically modulated coastal-eddy propagation in the Algerian Basin" Martina Auditore https://doi.org/10.5281/zenodo.20445357
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 180 | 67 | 35 | 282 | 39 | 36 |
- HTML: 180
- PDF: 67
- XML: 35
- Total: 282
- BibTeX: 39
- EndNote: 36
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript presents a consistent description of a coastal eddy observed during the SWOT fast-sampling phase in the Algerian Basin. The physical interpretation is generally plausible, and the comparison between SWOT, conventional gridded altimetry, SST, and chlorophyll-a is visually convincing. However, in my opinion the manuscript does not provide sufficient scientific scope for publication in Ocean Science in its present form.
The main result of the paper is that SWOT resolves the structure and short-term evolution of a coastal eddy better than standard interpolated L4 altimetry maps. While this is an interesting illustration, it is also largely expected. The manuscript repeatedly compares SWOT observations with daily 1/16° gridded products derived from conventional altimetry. It is already well established within the altimetry community that such mapped products smooth mesoscale and submesoscale variability and have difficulty representing features evolving on time scales shorter than several weeks. Consequently, the fact that SWOT performs better in this comparison does not, by itself, constitute a sufficiently novel scientific result.
Furthermore, the capability of SWOT to observe mesoscale and even submesoscale eddies has already been demonstrated by a rapidly growing body of literature. In this context, the present study remains primarily descriptive. The analysis does not introduce a clearly innovative eddy-detection methodology, nor does it provide a quantitative assessment of the improvements brought by SWOT relative to conventional products. The comparison with SST and chlorophyll-a is useful as qualitative support, but it is not developed into an independent validation of eddy position, propagation, or deformation.
The manuscript does not appear to provide substantial new insight into the oceanography of the Algerian Basin. The study documents an interesting case, but it does not convincingly demonstrate a new dynamical mechanism nor quantify the role of topography in a way that significantly advances our understanding of coastal eddy propagation. In particular, the interpretation of the apparent velocity maxima as evidence of topographic modulation remains suggestive rather than demonstrated, especially since the estimated velocities are derived from ellipse-centroid displacement during phases of strong eddy deformation.
Finally, the manuscript reads as a useful and pedagogically effective case study. It clearly illustrates that SST and chlorophyll-a patterns can be consistent with sea-level contours and that SWOT provides a much sharper two-dimensional representation of the eddy than conventional mapped altimetry. However, I do not think this is sufficient for the standards of Ocean Science, which states that it publishes studies with "important implications for our understanding of the state and behaviour of the ocean". In its present form, the manuscript demonstrates an expected observational advantage of SWOT but does not yet translate this advantage into a sufficiently significant scientific contribution.
To strengthen a potential resubmission, the authors should move beyond a qualitative description and provide a more quantitative analysis. This could include objective metrics of the improvement obtained with SWOT, independent estimates of eddy displacement based on SST or chlorophyll-a observations, a sensitivity analysis of the ellipse-fitting methodology, or a clearer separation between eddy translation and deformation. A more rigorous assessment of the relationship between the observed propagation-speed maxima and bathymetric parameters would also be necessary if the claim of topographically modulated propagation is to be retained.
Another potentially valuable direction would be the inclusion of the SWOT-MIOST product, in which SWOT observations are incorporated into mapped altimetric fields. Such a comparison would be more informative than a direct SWOT-versus-L4 comparison and would allow the authors to evaluate whether SWOT-enhanced mapped products improve the representation of high-frequency coastal variability. Moreover, SWOT-MIOST products are available beyond the CAL/VAL phase and extend into the science phase of the mission, increasing both the relevance and the broader applicability of the study.
For these reasons, I do not recommend publication of the manuscript in its current form. However, because I still see potential in the study, I recommend that the manuscript be considered for major revision rather than rejection. The requested revisions are substantial and will require significant additions to the analysis. A substantially revised manuscript could become suitable for publication if the authors either provide a considerably stronger quantitative and dynamical analysis or reformulate the work more explicitly as a methodological or product-evaluation study rather than as a scientific advance in coastal eddy dynamics.
Below I add some line-by-line additional comments:
Lines 10–12 — The final sentence of the abstract reads rather generic and shows signs of AI-generated phrasing. This is true for many other sentences in the manuscript. If AI tools were used in the preparation of the manuscript, this should be explicitly stated according to the journal policy, for example in the acknowledgements or in the dedicated AI-use declaration if required by the journal.
Lines 39–42 — The sentence referring to the “β-effect, lower-layer circulation, shelf width and depth, and slope characteristics” is technically correct but not very accessible. Please explain this point in simpler physical terms, so that the reader can understand which mechanisms are expected to influence eddy propagation near a continental slope.
Lines 47–48 — The statement that SST and ocean-colour observations are sensitive to cloud presence is generally true, but it is not clear whether this is a major limitation for the specific region and period analysed in this study. Please clarify whether cloud contamination was actually a significant issue for the Algerian Basin case considered here.
Lines 52–55 — The limitations of conventional altimetry in coastal regions are cited using Vignudelli et al. (2011), but this reference is quite old and does not reflect the progress made in coastal altimetry over the last 15 years, including improvements in sensor technology, retracking, data processing, and geophysical corrections. Since the study analyses recent data, please either update this statement with more recent references or remove the sentence if it is not directly relevant to the present analysis.
Lines 59–61 — The claim that SWOT provides observations “much closer to the coast compared to conventional along-track altimetry” is misleading. Reprocessed high-rate conventional altimetry can provide valid measurements up to a few hundred metres from the coast in favourable cases. The real added value of SWOT in the coastal zone should instead be framed as its two-dimensional mapping capability with low noise over the swath. Please revise this statement accordingly.
Lines 60–62 — The phrase “potential persistent small-scale uncertainties” is unclear. Please specify what type of uncertainty is meant, whether it refers to geophysical corrections, wet-tropospheric effects, residual errors, or another source of error, and explain why it is expected to be persistent.
Lines 86–87 — The manuscript states that SST and ocean-colour data were downloaded from the Copernicus Marine Service. Please provide the corresponding product DOIs in this section, or explicitly indicate where the reader can find the DOI information later in the paper.
Lines 88–90 — The manuscript states that PyEddyTracker was used for conventional altimetry. Please clarify how this was implemented. Was the code used as distributed publicly? Was it modified by the authors? If a specific version or customised implementation was used, please indicate how the reader can access it.
Lines 94–103 — The ellipse-fitting procedure appears to be a central methodological component of the study, but it is not described in sufficient detail to allow reproducibility. Please provide a clearer step-by-step description of the method, including the filtering, contour selection, boundary reconstruction, ellipse fitting, and centroid-speed calculation. Ideally, this should include the mathematical formulation and/or a worked example showing how the procedure is applied to one SWOT scene.
Lines 106–114 — The comparison with SST and chlorophyll-a is useful as qualitative support, but it should not be described as a “validation” of the SLA-derived eddy position and shape. In particular, the fact that SWOT gives higher velocities or sharper structures does not automatically imply that they are more accurate. If the authors want to use SST as an independent reference for the eddy displacement, the eddy centroid or another comparable positional metric should also be estimated from SST. Otherwise, the wording should be softened to “qualitative consistency” or “supporting evidence” rather than validation.
Lines 148–154 — The interpretation of the SWOT-derived speed peaks needs to be treated with more caution. The higher propagation velocities obtained from the ellipse-centroid displacement are not necessarily more accurate than those from conventional altimetry, especially because the eddy is deforming during the same period. To support the statement that SWOT better captures propagation variability, the authors should compare the centroid displacement with an independent estimate, for example from SST, or explicitly state that these are apparent kinematic velocities affected by deformation.