Sub-kilometer current divergence and vertical velocities in the macrotidal Iroise Sea from airborne radar interferometry and models

Fine-scale ocean dynamics strongly influence coastal mixing, sediment transport, and vertical exchange, yet remain difficult to observe at the spatial scales at which they occur. Here, we use total surface current vectors measured by the airborne Ocean Surface Current Airborne Radar (OSCAR) to investigate sub-kilometer divergence, vorticity, shear rate, and vertical velocities in the tidally dominated Iroise Sea around Ushant Island. Observations acquired during ebb and flood tides in May 2022 reveal two distinct mechanisms generating intense fine-scale dynamics. North and west of Ushant Island, interactions between strong tidal currents and complex bathymetry produce alternating zones of divergence and convergence associated with flow acceleration and deceleration over bathymetric gradients. South of the island, a tidal jet flowing alongside comparatively calm waters generates strong lateral shear, producing divergence of up to O(20 f) and vertical velocities approaching 0.2 m.s −1. Comparison with the high-resolution MARS2D model shows that currents and bathymetry-driven divergence are reproduced reasonably well, whereas the shear-driven divergence associated with the southern tidal jet is underestimated by approximately an order of magnitude. Coarsening the OSCAR observations from 200 m to 1 km reduces divergence, vorticity, and shear rate from O(10 f) to O(1 f), demonstrating the strong sensitivity of derivative quantities to spatial resolution. These results highlight the importance of sub-kilometer observations for resolving fine-scale coastal dynamics, provide a unique observational benchmark for evaluating high-resolution numerical models, and illustrate the complementary roles of airborne, in situ and satellite measurements in characterizing coastal and submesoscale processes across a wide range of spatial and temporal scales.

Fine-scale dynamic processes (defined here as occurring at spatial scales between 100 meters to 10 kilometers) play a fundamental role in the global ocean, including polar, shelf, and coastal environments, where they regulate horizontal transport, lateral mixing, vertical exchanges, and air–sea interactions (see Gommenginger et al. (2019) and references therein; Zhu et al., 2024; Viitak et al., 2025). These predominantly ageostrophic dynamics strongly influence marine ecosystems by enhancing vertical nutrient fluxes (Lévy et al., 2012). In coastal seas, where they are particularly energetic, they govern sediment transport, erosion, and freshwater entrainment, thereby modulating water quality, pollutant dispersal, and biogeochemical cycling, with significant implications for ecosystem health, aquaculture, and fisheries.

The need to understand fine-scale dynamics extends beyond oceanography. The rapidly expanding marine renewables sector requires improved characterization of ocean dynamics in coastal and shelf seas to better assess resources, risks and environmental impacts, and to optimize infrastructure siting and energy production. Existing current charts and survey datasets are often outdated and insufficient for these purposes (Thiébaut and Sentchev, 2017), and the acquisition of new surveys remains costly and time-consuming. Beyond their economic importance, fine-scale coastal dynamics affect communities and ecosystems worldwide: approximately one billion people live within 10 km of the coast (Cosby et al., 2024) and 3.3 billion people rely on fish and seafood as their primary source of animal protein (Viana et al., 2023). Understanding how these fine-scale dynamics influence local circulation, ecosystems and the transport of sediments, carbon, nutrients and pollutants is therefore increasingly important, particularly in the face of a changing climate. Better observations and understanding are essential to improve the representation of fine-scale dynamics in coastal and regional models, and to develop more realistic parameterizations in global ocean and climate models.

Despite their significance, fine-scale ocean dynamics remain poorly quantified due to the inherent difficulty of accurately measuring two-dimensional vector fields of surface currents at high resolution. Today’s understanding stems primarily from numerical simulations, theoretical models and a relatively small number of experimental field campaigns (e.g. Farrar et al., 2025). The scarcity of high-resolution synoptic observations of current vector fields limits model validation. This is particularly significant in shallow regions, especially in highly energetic tidally dominated seas with strong currents (Hessner et al., 2014), where complex bathymetry can force both vertical and horizontal movement of water, creating eddies, shear zones and strong divergence over short horizontal distances.

High spatial resolution is essential because many signatures of ageostrophic dynamics are strongly scale dependent. Recent studies have demonstrated that divergence can range from O(1 f) at 1-km resolution to below 0.15 f at scales of 6–12 km (Esposito et al., 2021), with similar effects found for vorticity. Largescale measurements average out spatial variability, underestimating divergence, and failing to resolve ageostrophic dynamics. Consequently, observations capable of resolving scales of O(< 1 km) across spatial extents of several kilometers are required to investigate these processes.

Yet, present-day observing systems struggle to meet these requirements. Nadir satellite altimeters offer quasi-global spatial coverage, but are limited to geostrophic currents and mesoscales larger than 70 km, with data quality and geophysical corrections becoming less reliable within 10 km from the coast (Gommenginger et al., 2011). SAR altimeters (e.g. Cryosat, Sentinel-6) and the SWOT mission measure sea surface height with much finer spatial resolution (down to 500 meters); however, like all altimeters, these can only infer geostrophic currents, and therefore miss important ageostrophic dynamics. Some in situ and ground-based systems can resolve the required time and spatial scales, but each has limitations. Surface drifters provide good temporal resolution but have insufficient spatial resolution and coverage, particularly in coastal regions. Acoustic Doppler Current Profilers (ADCPs) provide point measurements (when moored) or transects (from ships), missing the two-dimensional synoptic coverage required to investigate fine-scale dynamics. Ground-based High-Frequency (HF) radars and X-band marine radars can provide synoptic maps of currents at high spatial and temporal resolutions and have been used successfully to investigate fine-scale processes in specific regions (Hessner et al., 2014). However, their spatial coverage is limited by the range of the ground stations and by line-of-sight obstruction from land. Furthermore, the global network is unevenly distributed, concentrated mainly in industrialized nations, resulting in substantial observational gaps around the global coastal ocean (Roarty et al., 2019; Hauser et al., 2023).

Recent advances in remote sensing have renewed interest in direct measurements of total surface currents. The airborne Doppler scatterometer, DopplerScatt, developed by Rodríguez et al. (2018), has been deployed as part of the Sub-Mesoscale Ocean Dynamics Experiment (S-MODE; Farrar et al., 2025) to assess the contributions of submesoscales to ocean circulation and vertical transport in the California Current system. Using 2-km resolution DopplerScatt surface current observations, Torres et al. (2024) reported divergence values of O(1 f). The spatial coverage provided by remote sensing systems like DopplerScatt offers valuable broader context for interpreting localized observations from ships, autonomous vehicles, and fixed instruments such as ADCPs or X-band radars.

The primary source of data in this paper is the Ocean Surface Current Airborne Radar (OSCAR), a novel airborne remote sensing instrument that provides quasi-instantaneous high-resolution images of Total Surface Current Vectors (TSCV) and Ocean Surface Vector Winds (OSVW) at spatial resolutions down to 200 meters (McCann et al., 2024). The instrument is a Synthetic Aperture Radar Along-Track Interferometer (SAR-ATI), developed as the airborne demonstrator for the SeaSTAR satellite mission concept submitted to the European Space Agency (ESA) Earth Explorer 11 programme (Gommenginger et al., 2019; Martin et al., 2022; Gommenginger et al., 2023). OSCAR provides observations at a spatial resolution approximately one order of magnitude finer than DopplerScatt, offering the opportunity to investigate the structure and magnitude of coastal fine-scale dynamics at different scales.

In this paper, we investigate how bathymetry and tidal jets generate sub-kilometer divergence, vorticity, shear and vertical motions around Ushant Island in the Iroise Sea, using 200 m OSCAR observations and high-resolution numerical models. From the observed current fields, we derive diagnostic quantities including divergence, vorticity, shear rate, and estimates of vertical velocities, and assess the impact of spatial resolution on their magnitude and distribution. These observation-based diagnostics are subsequently compared with output from high-resolution numerical models to evaluate the representation of fine-scale dynamics in models in relation to bathymetry.

OSCAR field campaign in Iroise Sea

The study focuses on the waters surrounding Ushant Island in the Iroise Sea, west of Brittany, France, a highly energetic environment characterised by strong tidal currents, complex coastlines, and pronounced bathymetric variability that generate large current gradients and fine-scale circulation features (Sentchev et al., 2013). OSCAR observations were acquired during an airborne campaign conducted in May 2022 as part of the ESA SEASTARex project (McCann et al., 2024).

The strong tidal currents around Ushant Island are clearly visible in optical satellite imagery through plumes of suspended sediment (Figure 1), which reveal the presence of energetic tidal jets north and south of the island. While the seabed immediately surrounding Ushant consists primarily of exposed rock, the slightly deeper areas are covered by gravel and sand (Guillou and Thiébot, 2016). These sediments are readily resuspended by strong currents and contribute to the turbidity patterns visible in optical images.

During the campaign, HF radar, X-band radar and ADCP measurements were also collected for validation of OSCAR total surface current vectors (TSCV). Comprehensive validation of OSCAR TSCV has been presented elsewhere (McCann et al., 2024; Gommenginger et al., 2023) and is therefore not repeated here. HF radar data have previously been used in this region to investigate tidal currents (Thiébaut and Sentchev, 2017), the Ushant front (Le Boyer et al., 2009) and Lagrangian residual circulation (Muller et al., 2009). However, HF radar data could not be used in this study because observations were unavailable north of Ushant Island owing to radar shadowing by the island. Similarly, the spatial coverage of TSCV data from an X-band radar system on the La Jument lighthouse (McCann et al., 2024) was too limited for the present analyses. The moored ADCP was located too far south of the study area and provided only point measurements.

The study focuses on TSCV images obtained from six OSCAR flights over Ushant Island on 17 and 22 May 2022. All flights were timed to coincide with periods of maximum tidal flow. The observations on 17 May were acquired during spring tide, with two overpasses during maximum ebb conditions at 09:32 and 09:38 UTC. Winds were from the south-east with speed around 7.9 m.s −1. The observations on 22 May were acquired one day before neap tide, with four overpasses around maximum flood conditions at 05:39, 05:47, 05:55 and 07:23 UTC. Winds were from the north-east with speeds around 5.5 m.s −1. Wind conditions on both days were favourable for airborne operations and provided sufficient sea surface roughness and SAR backscatter for high-quality SAR Doppler measurements and surface-current retrievals.

OSCAR instrument and geophysical inversion

The Ocean Surface Current Airborne Radar (OSCAR) is a multi-azimuth SAR Along-Track Interferometer operating at Ku-band (13.5 GHz). The instrument consists of one antenna pointing in the broadside direction (across-track) and two pairs of ATI antennas squinted ±45 ° fore and aft of broadside. The foreand aft-pointing ATI pairs measure radial surface velocity (RSV) and normalized radar cross-section (NRCS) in two orthogonal lines-of-sight, while the broadside antenna measures NRCS only.

This unique three-azimuth configuration enables the simultaneous retrieval of total surface current vectors (TSCV) and ocean surface vector winds (OSVW) from a single overpass at high spatial resolution and precision (0.1 m.s −1 at 1x1 km 2) (Gommenginger et al., 2019; Martin et al., 2022; Gommenginger et al., 2023). The simultaneous retrieval of TSCV and OSVW is performed using the inversion described by Martin et al. (2024) and McCann et al. (2024) using the SeaSTAR project codebase (Martin et al., 2023).

In this study, the inversion uses the KuDop geophysical model function (GMF) by Mouche et al. (2012) to model RSV and the NSCAT4DS KuMod GMF by Wang et al. (2017) to model NRCS. The resulting TSCV and OSVW fields are delivered at 200 m spatial resolution. For each overpass, OSCAR produces quasi-instantaneous images of TSCV and OSVW over a trapezoidal area with a swath width up to 5 km, and in this campaign, a track length of approximately 20 km in the along-track direction. The present study focuses exclusively on OSCAR TSCV observations.

Ambiguity removal and filtering

As in conventional scatterometry (Stoffelen and Anderson, 1997), the OSCAR geophysical inversion yields four ambiguous solutions for each wind-current vector pair. Ambiguity removal is required before TSCV fields and derivative quantities can be obtained. An ambiguity removal algorithm needs to be applied to select one of the four solutions before TSCV fields and derivative quantities can be obtained.

The ambiguity removal approach used by McCann et al. (2024) selects the wind-current solution whose wind component is closest to the ERA5 background wind field (Hersbach et al., 2023), based on the Euclidean distance between each retrieved wind solution and the ERA5 wind vector. While this approach is well suited to mapping TSCV fields, the calculation of derivative products places much stricter requirements on spatial consistency, because even isolated retrieval errors can produce large spurious gradients. To minimise such artefacts, a more advanced spatial ambiguity-removal approach was adopted in this study. A novel Spatial Simultaneous Ambiguity Removal (SSAR) method was developed and implemented within the OSCARplus Python package (Michalski, 2026).

Like standard scatterometer ambiguity removal algorithms (e.g. Stoffelen and Anderson, 1997), the SSAR method assumes that current and wind vectors vary gradually between neighboring grid cells, without abrupt changes in direction or magnitude. Hence, the optimal current-wind solution for a given cell is the one that minimizes the difference with its surrounding cells for both winds and currents simultaneously. The algorithm proceeds in two stages.

First, the ERA5-based ambiguity removal provides an initial solution for all cells. In the second stage, for each cell, the SSAR algorithm calculates the sum of Euclidean distances, D n, between each candidate ambiguity, v n, for the central cell and the eight previously selected solutions, v i, of the neighboring cells. Separate distance sums are calculated for currents and winds (Equation 1):

The current and wind distances are then combined into a total cost function, τ n, with the currents term weighted by an adjustable scaling factor, m (Equation 2):

For this study, a scaling factor of m = 1.5 was used. Tests showed that the algorithm is only weakly sensitive to the precise value of this scaling parameter. The effect of different values of m on the retrieved TSCV fields is illustrated in Supplementary Figure S1.

The candidate ambiguity with the lowest total cost is selected to replace the previous solution, after which the algorithm advances to the next grid cell. As SSAR progresses through the field, each newly selected solution is immediately used during ambiguity selection for subsequent cells. In this study, the procedure was repeated four times, alternating between along-track and across-track passes to minimize directional bias. Despite these iterations, a small number of isolated cells remained inconsistent with their surroundings. These residual outliers were removed by applying a 3×3 median filter to the final ambiguity field. The effect is confined primarily to single-grid-cell (200 × 200 m) features and removes the most extreme derivative values.

OSCAR TSCV derivative products

Central to the interpretation of fine-scale ocean dynamics are surface-current derivative quantities such as the normalized divergence δ / f = ∂ u ∂ x + ∂ v ∂ y, vorticity ζ / f = ∂ v ∂ x − ∂ u ∂ y, and shear rate σ / f = ∂ v ∂ x + ∂ u ∂ y, where f is the planetary vorticity, f = 2Ωsin ϕ, with Ω the Earth’s angular velocity and ϕ the latitude (e.g. Rudnick, 2001; Shcherbina et al., 2013; D’Asaro et al., 2018). Divergence is commonly used as a proxy for vertical transport and mixing, while vorticity characterises rotational motions and departures from geostrophic balance, which become increasingly important at fine scales as frictional and inertial effects compete with the Coriolis force (Ohlmann et al., 2017). Shear rate quantifies lateral velocity gradients and highlights regions of strong flow deformation. Together, these quantities provide a concise description of the fine-scale dynamics associated with bathymetric forcing, tidal-jet shear, and the resulting vertical exchanges.

Because these quantities depend on spatial derivatives of the two-dimensional TSCV field, they are particularly sensitive to measurement resolution and noise. Derivative products like divergence, vorticity and shear rate require accurate measurements of total surface current vectors. Most studies to date have relied on data from moored or shipborne ADCP, drifters, High-Frequency (HF) radars or numerical models.

OSCAR inferred vertical velocity

An estimate of vertical velocity (w) was obtained using shallow water theory (Equation 3), where h is the water depth and u is the horizontal current vector (TSCV).

This formulation neglects nonhydrostatic and topographic-slope effects and assumes that horizontal scales are much larger than vertical scales, and that horizontal currents u are approximately uniform throughout the water column. Around Ushant Island, the characteristic horizontal scales of the observed convergence/divergence structures are approximately 400–600 m, while water depth ranges from 40 to 100 m. The assumption that horizontal scales are much larger than vertical scales is therefore not equally well satisfied throughout the domain, and the resulting vertical velocities need to be interpreted accordingly. These limitations are discussed further in Section 4.4. A full derivation can be found in Appendix A.

Numerical model data

This study uses TSCV output from two operational numerical models, MARS3D and MARS2D (Lazure and Dumas, 2008), both operated by the Laboratoire d’Océanographie Physique et Spatiale (LOPS).

MARS3D provides total horizontal currents every hour at 2.5 km spatial resolution. No vertical currents are provided in the products. The model resolves both barotropic and baroclinic dynamics using terrainfollowing coordinates distributed over 40 vertical layers. Open-boundary forcing includes the principal tidal constituents (M2, S2, K2, N2, O1, K1, P1 and Q1).

MARS2D provides depth-averaged currents every 15 min at 250 m spatial resolution. The model uses the same forcing as MARS3D, but does not account for density-driven circulation and vertical structure.

Both models use bathymetry derived from EMODnet (EMODnet Bathymetry Consortium, 2022) and SHOM-Ifremer (SHOM-IFREMER, 2014). In the vicinity of Ushant Island, the bathymetry is derived exclusively from the higher-resolution SHOM–Ifremer dataset.

Bathymetry

The bathymetry used in the study is provided by EMODnet (EMODnet Bathymetry Consortium, 2022). Water depth is relative to the Lowest Astronomical Tide (LAT) datum, gridded at 1/16 arc minute (approx. 115 m in the study area). The EMODnet bathymetry is broadly consistent with the bathymetry used in the numerical models.

OSCAR observations and resolution dependence

Ebb-tide observations (17 May 2022)

OSCAR observations acquired at 09:32 UTC on 17 May 2022 (Figure 2) reveal strong surface currents up to 2.8 m.s −1, consistent with previous observations of this energetic tidal environment (Sentchev et al., 2013). Two distinct tidal jets are observed: a northern jet passing north of Ushant Island, and a southern jet generated by flow through the strait between Ushant and Bannec Islands.

The northern jet flows primarily through waters deeper than 60 m, approximately 2 km north of Ushant Island. Only a fraction of the flow crosses the adjacent coastal region, where water depth decreases to less than 40 m and the bathymetry becomes complex. In Figure 2, the solid blue arrow along the northern jet indicates the position of the transect shown in Figure 3A. Along this transect, the bathymetry exhibits rapid depth variations, including a narrow canyon a few hundred meters wide where the water depth doubles to over 50 m. These rapid depth changes are associated with strong oscillations in divergence, reaching values close to 10 f on the upslope before decreasing sharply to −10 f on the downslope. In contrast, the estimated vertical velocity remains small (± 0.03 m.s −1) and varies gradually, increasing slightly as water depth decreases but showing only a weak response to the canyon itself.

The southern jet exhibits a markedly different setting and behavior. The small arrows in Figure 2A show that the flow emerges strongly from the gap between Ushant and Bannec Islands, but some northward recirculation is evident from longitudes around 5.13 ° W. The dashed blue arrow across the southern jet indicates the position of the transect shown in Figure 3B. Unlike the northern transect, this section crosses the southern jet perpendicularly across relatively uniform bathymetry (≈ 80 m deep), revealing an exceptionally sharp lateral velocity gradient, with velocities increasing from 0.2 m.s −1 to 2 m.s −1 over approximately 750 m. This sharp edge coincides with very large values of divergence, reaching up to 20 f. Because the water depth is nearly constant, the estimated vertical velocity closely tracks the divergence, reaching magnitude of up to 0.2 m.s −1 on the edge of the jet.

Flood-tide observations (22 May 2022)

OSCAR observations acquired during flood tide at 05:47 UTC on 22 May 2022 (Figure 4) show a similar bathymetric influence on the flow, although currents are generally weaker than on 17 May 2022. Maximum northward currents of up to 2.2 m.s −1 occur in the deeper part of the OSCAR domain, while strong current gradients are observed west of Ushant Island where the flow encounters rapidly decreasing water depths.

The solid arrow in Figure 4 indicates the location of the transect shown in Figure 5. Along this transect, water depth decreases from approximately 80 m to less than 40 m over a distance of a few kilometers. As the flow crosses the shallowing bathymetry, current speed increases from approximately 0.9 m.s −1 to 1.5 m.s −1. This acceleration is accompanied by enhanced divergence and estimated vertical velocity, with the largest positive values occurring over the steepest bathymetry. Similar to the northern transect during ebb tide, divergence responds rapidly to changes in bottom slope, while the estimated vertical velocity varies more smoothly owing to its dependence on both divergence and water depth.

Influence of spatial resolution

Figure 6 compares the distributions of divergence, vorticity, and shear rate derived from the OSCAR observations at their native 200 m resolution (red) and after coarsening to 1 km resolution (blue). The coarsened current fields were obtained by block-averaging the original 200 m TSCV data within each 1 km grid cell (i.e. each 1 km cell represents the mean of 25 original 200 m cells). The derivative quantities were then recalculated from the coarsened current field.

All three quantities show a substantial reduction in magnitude at the coarser resolution. The long tails of the 200 m distributions, associated with large values at the edges of the jets, are absent at 1 km resolution. The reduction is particularly pronounced for divergence, whose range decreases by more than a factor of two after coarsening (from approximately 30 f at 200 m resolution to less than 10 f at 1 km resolution). Likewise, the standard deviation of all three quantities is almost halved, reflecting the inability of coarser resolution data to resolve the strongest gradients and most energetic fine-scale structures.

The shape of the distributions is also affected by spatial resolution. Divergence and shear rate show marked changes in skewness, although interestingly, the vorticity distributions show only minor changes in skewness. These results indicate that a substantial fraction of the divergence, vorticity, and shear signal resides at scales smaller than 1 km. Consequently, observations at kilometer-scale resolution significantly underestimate the intensity and intermittency of fine-scale ageostrophic dynamics in this energetic tidal environment.

Comparison with numerical models

MARS2D representation of tidal jets

MARS2D output for 09:30 UTC on 17 May 2022 reproduces the two principal tidal jets observed by OSCAR north and south of Ushant Island (Figure 7A). Current magnitudes are broadly consistent between the two datasets, with a maximum speed in MARS2D of 3.3 m.s −1, slightly larger than the maximum 2.8 m.s −1 observed by OSCAR.

Important differences exist however in the spatial structure of the flow. In the northern sector, MARS2D places the main jet within 1 km of Ushant Island, considerably closer to the coast than OSCAR observations (Figure 2A). The model therefore predicts a larger fraction of the flow crossing the shallow bathymetry immediately north of the island. In the southern sector, MARS2D predicts a pronounced southward deflection of the jet upon reaching a water depth around 80 m, a behavior that is not evident in the OSCAR observations.

It is important to note that previous validation against X-band radar and NovaSAR satellite radar imagery demonstrated that the OSCAR locations of the northern and southern jets are robust (McCann et al., 2024).

MARS2D representation of divergence and estimated vertical velocity

In the northern jet, both datasets present a similar spatial pattern, with strong divergence when the flow approaches shallow water, and strong convergence when water depth increases. However, the divergences derived from MARS2D (Figure 7B) are generally a factor of two smaller than those derived from OSCAR observations (Figure 2B). In the southern jet, the differences are considerably larger: divergence predicted by MARS2D is approximately one order of magnitude weaker than that observed by OSCAR, and the intense shear-induced divergence observed along the edge of the southern jet is entirely absent in the model.

Differences are also evident in the estimated vertical velocity fields. The spatial patterns obtained with MARS2D differ substantially from those observed by OSCAR in both the northern and southern jets. In the northern jet, MARS2D predicts peak vertical velocity magnitude of up to 0.18 m.s −1, although values remain below 0.06 m.s −1 across most of the domain (Figure 2C). By comparison, the vertical velocities estimated from OSCAR observations (Figure 7C) are significantly larger with more pronounced spatial signatures.

MARS3D TSCV, divergence and vertical structure

Figures 7D, E show MARS3D surface currents and surface-current divergence for 17 May 2022 at 10:00 UTC at 2.5 km resolution. Figure 7F shows vertical profiles of current velocity at two grid points, A and B, laying within the northern and southern jets predicted by MARS3D.

Although its coarser resolution provides less detail than MARS2D and OSCAR, MARS3D reproduces tidal jets of similar intensity and overall structure (Figure 7D) to those predicted by MARS2D (Figure 7A) and observed by OSCAR (Figure 2A). In contrast, the divergence derived for MARS3D is about an order of magnitude smaller than for MARS2D, consistent with the reduction observed between the 200 m and 1 km OSCAR data.

The vertical profiles (Figure 7E) show a logarithmic decrease in horizontal velocity with depth, indicative of a mixed water column influenced by bottom friction. At 80% of the total water depth, the horizontal velocity remains approximately 87% of the surface value at Point A and 79% at Point B (see white stars in subplot D). The horizontal velocity remains relatively uniform throughout approximately the top 60% of the water column at all three locations.

MARS2D and OSCAR comparisons in deep and shallow water

As discussed in Section 3.2.1, the dominant current features are not located in the same geographical positions in MARS2D and OSCAR. This is evident from comparing the spatial distributions of TSCV, divergence and estimated vertical velocities in Figures 2A–C, 7A–C, as well as the side-by-side maps of vorticity (Supplementary Figure S2) and shear rate (Supplementary Figure S3). Consequently, a pointby-point comparison between the two datasets is not meaningful. Instead, the observations and model output are compared as a function of water depth, to assess the dynamical behavior in the two datasets independently of their spatial misalignment.

In this region, water depth provides a useful proxy for the dominant dynamical regimes. Shallow regions (< 60 m) exhibit large relative variations in depth and strong bathymetric gradients, where bathymetry-driven dynamics dominate. In contrast, deeper regions have relatively flat bathymetry, where the influence of bathymetric variations is reduced and other dynamical processes become more prominent. The shallow/deep classification is therefore used here as a simple means of distinguishing between bathymetry-dominated and other dynamical regimes.

Figure 8 presents TSCV magnitude, normalized divergence, normalized vorticity, and normalized shear rate from OSCAR and MARS2D on 17 May 2022, plotted against water depth. The MARS2D domain used for this comparison extends 0.01 ° beyond the OSCAR domain to include the southern jet predicted by MARS2D, which lies partially outside the OSCAR coverage. The data are divided into two depth classes: shallow water (0 to 60 m; light blue) and deep water (over 60 m; dark blue).

TSCV magnitudes (Figures 8A, B) show broadly similar distributions between MARS2D and OSCAR. Both datasets show a wide range of velocities between 0 and 3 m.s −1 in both depth classes, with weaker currents on average in shallow waters.

In contrast, divergence distributions differ significantly, particularly in deep water. For depths greater than 60 m, OSCAR-derived divergence presents a large spread of values between −20 f and +30 f, with a median of −1.25 f and a standard deviation of 8.77 f. MARS2D predicts a much narrower distribution centered near zero (median −0.02 f; standard deviation 1.76 f). In shallow water, both distributions display a broad spread of divergence values, although OSCAR shows more negative divergence on average (median −4.72 f), whereas the MARS2D distribution remains centered close to zero.

Vorticity distributions also differ between the two datasets. In shallow waters, vorticity is predominantly positive for both MARS2D (median 7.62 f) and OSCAR (median 3.14 f) but MARS2D shows much greater spread (standard deviation 14.61 f against 8.27 f for OSCAR). The vorticity skewness is similar in the two datasets, with positive values of 1.06 for MARS2D and 1.1 for OSCAR. At depths greater than 60 m, however, the model shows a concentrated cluster of negative vorticity values, not seen with OSCAR, resulting in a median of −0.3 f and a near-zero skewness of 0.06, compared with a positive skewness of 0.75 for OSCAR.

The shear-rate distributions for MARS2D and OSCAR are comparable in shallow water. Although MARS2D shows a slightly negative median (−0.25 f) compared with OSCAR’s positive median (0.96 f), the overall range and standard deviations are similar (MARS2D 8.06 f, OSCAR 9.4 f). In deeper waters, the distributions differ markedly, with OSCAR dominated by strongly negative values (to −40 f), whereas MARS2D predicts strongly positive values (to +30 f).

Bathymetric control of fine-scale dynamics

OSCAR observations on 17 May reveal a strong influence of bathymetry on the northern tidal jet. The sharp decrease in water depth north of Ushant Island acts as an obstacle, deflecting the bulk of the current northward (Figure 2A). Contrary to expectations based on horizontally constrained flows, this deflection leads to a slowdown over the nearshore shallows, as water is redirected away from the island rather than accelerated over the shallow region. In contrast, MARS2D predicts the jet to cross the shallows directly (Figure 7A), indicating that the model does not fully reproduce the observed bathymetric response.

Along the northern transect (Figure 3A), OSCAR observations show highly variable currents across the shallow region, with local accelerations over bathymetric highs where water depth decreases rapidly to approximately 20 m. These variations generate strong oscillations in divergence. Both OSCAR and MARS2D (Figures 2B, 7B) show a consistent pattern of positive divergence over upslopes followed by negative divergence (convergence) over downslopes (Figure 3A). The close correspondence between divergence and bathymetric gradients demonstrates the strong dependence of fine-scale divergence on water depth, current direction and current magnitude.

The estimated vertical velocity responds differently. Although divergence varies rapidly across the bathymetric features, vertical velocity remains comparatively weak because the effects of divergence and depth partly compensate for each other in the shallow-water formulation (Equation 3).

Similar bathymetric controls are also observed west of Ushant Island during flood tide on 22 May (Figures 4, 5), although the smaller divergence and estimated vertical velocities during flood tide may reflect weaker mixing under neap-tide conditions. Here, the flow acceleration over shallowing bathymetry again produces strong divergence, together with enhanced vertical velocity. The overall agreement between bathymetric gradients and OSCAR-derived current patterns provides additional confidence in both the bathymetric dataset and the retrieved current fields. The displacement of bathymetrically constrained flow features in MARS2D could therefore partly reflect differences in the model bathymetry, which incorporates higher-resolution SHOM–Ifremer data around Ushant Island. A detailed comparison of the bathymetric products would be required to determine the extent of this contribution.

Shear-driven divergence at the edge of tidal jets

The southern jet presents a markedly different dynamical regime. Here, the flow travels over relatively uniform bathymetry and follows a largely straight trajectory after exiting the channel between Ushant and Bannec Islands (Figure 2A). In contrast to the northern jet, the dominant current gradients arise from strong lateral velocity contrasts between the energetic tidal jet and the sheltered waters south of Ushant Island. The strongest divergence is co-located with these strong lateral velocity gradients and associated shear, rather than with marked bathymetric variations. We therefore refer to this regime as “shear-driven divergence”, to distinguish it from the bathymetry-driven divergence observed in the northern jet.

The cross-jet transect (Figure 3B) reveals extremely large divergence and estimated vertical velocities at the edge of the jet. Peak divergence reaches approximately +20 f and estimated vertical velocity 0.2 m.s −1, roughly twice the values observed along the northern transect (Figure 3A). Because bathymetric variations are small in this region, changes in estimated vertical velocity closely follow changes in divergence (Equation 3).

The OSCAR observations indicate that two distinct mechanisms generate intense divergence around Ushant Island. In the northern jet, divergence is primarily controlled by bathymetric forcing, and this mechanism is partially reproduced by MARS2D. In the southern jet, divergence is generated by strong lateral shear associated with the boundary between the fast-moving tidal jet and the sheltered waters south of Ushant Island. The shear-driven divergence is particularly notable because it is absent from the MARS2D simulations despite the model reproducing similar current magnitudes.

Model representation of fine-scale dynamics

The comparisons of OSCAR and MARS2D currents and derived quantities paint a contrasting picture. Although the two datasets produce broadly similar current velocity distributions (Figures 8A, B), the derived quantities show much larger differences. This indicates that reproducing the currents alone is insufficient to accurately represent the associated fine-scale dynamics.

The model performance varies with water depth. Agreement is generally better in shallow waters (0–60 m) than in deeper waters (60 m and over), a finding that is supported by the spatial comparisons in Figures 2, 7. In the northern jet, where divergence is closely linked to bathymetric gradients, both OSCAR and MARS2D reproduce similar patterns of divergence and convergence (Figures 2B, 7B), although the model-derived values are typically around a factor of two smaller. In contrast, the shear-driven divergence observed along the southern jet is almost entirely absent from the model, with MARS2D-derived divergence approximately an order of magnitude smaller than that observed by OSCAR.

These differences are reflected in the divergence statistics (Figures 8C, D). At depths greater than 60 m, OSCAR-derived divergence exhibits a wide distribution (median 1.25 f, standard deviation 8.77 f) against a much narrower distribution for MARS2D centered near zero (median −0.02 f; standard deviation 1.76 f). In shallow waters, the two distributions are more similar, both displaying a broad spread of divergence, although OSCAR divergence is negative on average. The stronger agreement in shallow water reflects the fact that both datasets capture the bathymetrically controlled divergence observed north of Ushant Island, whereas the largest discrepancies occur in the deeper waters associated with the southern tidal jet.

A similar pattern emerges for vorticity (Figures 8E, F). In shallow waters, both datasets exhibit predominantly positive vorticity and similar skewness values (1.06 for MARS2D and 1.1 for OSCAR). At depths greater than 60 m, however, MARS2D develops a concentration of negative vorticity values, resulting in a median of −0.3 f and a skewness of only 0.06, compared with 0.75 for OSCAR. The reduced skewness suggests that the model does not reproduce the asymmetric distribution of intense rotational structures observed by OSCAR in deeper water.

The shear-rate distributions exhibit the largest discrepancies (Figures 8G, H). In shallow waters, MARS2D and OSCAR show similar statistics (standard deviations of 8.06 and 9.4 respectively), although the mean shear rate is considerably smaller in MARS2D (0.19 compared with 3.22 for OSCAR). In deeper waters, however, OSCAR observations are dominated by negative shear rates, whereas MARS2D predicts predominantly positive values. This reversal is consistent with the contrasting trajectories of the southern tidal jet. In OSCAR, the jet turns northwards and interacts with the sheltered waters south of Ushant Island, whereas MARS2D shows a straighter jet path.

The OSCAR observations demonstrate that regions of high shear can host some of the largest divergence and estimated vertical velocities in the study area. These mechanisms are particularly challenging to observe with traditional systems, as instruments such as ADCPs can miss the important two-dimensional structures, while HF radars may not have the spatial resolution or specialized processing required to operate reliably in energetic regions (Dumas et al., 2025).

Estimated vertical velocities and influence of spatial resolution

The estimation of vertical velocities from surface-current divergence (Section 2.5) assumes that horizontal currents are approximately uniform throughout the water column. In this tidally dominated area, the water column is likely well mixed, although horizontal velocities are expected to decrease towards the seabed due to bottom friction. Using the OSCAR surface velocity as representative of the water column may therefore introduce a positive bias in the magnitude of the estimated vertical velocities. The MARS3D simulations provide a useful consistency check on this assumption for the strong tidal jets observed in this study. The MARS3D vertical profiles in Figure 7F predict a logarithmic decrease in velocity with depth, but horizontal velocity remains relatively uniform over the top 80 − 85% of the water column within the jets, with departures from surface velocities becoming noticeable only within the bottom 15 − 20% of the water column. The simulated vertical structure is therefore broadly consistent with the assumption of depth-uniform horizontal velocity, although this does not independently validate the assumption, particularly given MARS3D’s underestimation of shear-driven dynamics of the southern jet (Section 3.2.3). In the absence of independent subsurface current measurements in this region, such as ADCP profiles, that could be used in this study, the resulting uncertainty in the vertical-velocity estimates cannot be directly quantified. Nevertheless, the weak vertical variation simulated through most of the water column supports the use of the shallow-water approximation to estimate vertical velocities in this study.

The analysis of vertical velocities is additionally constrained by the limitations of shallow-water theory. The theory assumes that horizontal scales are much larger than vertical scales. In the study area, the characteristic horizontal scales of the observed dynamics are approximately 400600m. Hence, the assumption is reasonably well satisfied in nearshore regions, where water depths are typically 40-60m, but becomes less well satisfied towards the edges of the OSCAR domain, where water depths reach 90–100 m, increasing the uncertainty in the vertical velocity estimates. In addition, nonhydrostatic and topographic-slope effects may be significant in this energetic macrotidal environment with complex, strongly varying bathymetry. A more comprehensive observational campaign including ADCP current profiles would be required to validate the OSCAR vertical velocity estimates and assess the applicability of the shallow-water approximation across this region.

The OSCAR observations reveal estimated vertical velocities reaching up to ±0.2 m.s −1 along the edge of the southern jet. These values are substantially larger than those typically reported in the scientific literature. Surface or near-surface vertical velocity is most commonly measured in situ using ADCPs or derived from surface currents and divergence estimated from Lagrangian drifter data (Lodise et al., 2020). In the open ocean, observed vertical velocities at spatial resolutions of 1–10 km are generally O(10 −3 m.s −1) (Johnson et al., 2001; Tarry et al., 2022). Direct measurements of vertical velocities in tidally-dominated conditions are rare due to the experimental challenges associated with the strong currents, but vertical velocities O(0.1 − 0.5 m.s −1) have been observed with ADCPs in constrained tidal flow (Baschek et al., 2006; Li and Zheng, 2016). The large estimated vertical velocities observed with OSCAR around Ushant Island are therefore consistent with the dynamics observed in these exceptionally energetic tidal currents.

The importance of spatial resolution is demonstrated directly by the OSCAR observations. Coarsening the OSCAR data from 200 m to 1 km substantially reduces divergence, vorticity, and shear rate (Figure 6). The long tails associated with extreme values largely disappear, and the standard deviation of all three quantities is almost halved. These results indicate that a large fraction of the observed dynamics is concentrated at scales below 1 km. This is confirmed by the spatial structure of the tidal jets, where the transition from weak to strong flow occurs over approximately 750 m at the edge of the southern jet, and approximately 1 km along the northern jet.

The sensitivity to spatial resolution is consistent with previous observations of divergence and vorticity near submesoscale fronts. Esposito et al. (2021) compiled measurements of maximum divergence and vorticity reported by multiple research teams across a range of spatial scales, showing maxima of O(1 f) at resolutions around 1 km, increasing to O(10 f) and above at scales below 300 m. OSCAR shows a similar scale dependence but with consistently larger magnitudes: maximum divergence and vorticity increase from O(5-10f) at 1 km to almost 40f at 200 m, compared with approximately 1–5f and 10–20f, respectively, at similar scales in Figure 13 of Esposito et al. (2021). The larger OSCAR values likely reflect the unusually energetic tidal environment around Ushant Island and intense gradients produced by the strong interactions with bathymetry and coastlines.

The marked decrease in divergence, vorticity, shear rate, and estimated vertical velocity with coarsening indicates that a large fraction of the dynamics and mixing around Ushant Island occurs at sub-kilometer scales. In the northern jet (ebb) and west of Ushant Island (flood), these processes are driven primarily by interactions between strong tidal currents and complex bathymetry. In the southern jet, they result from intense lateral shear between the fast moving tidal jet and the adjacent calm waters sheltered by the island. Both mechanisms generate strong vertical motions that enhance mixing and transport at the edges of the jets. This interpretation is supported by the satellite optical imagery (Figure 1), where the tidal jets appear lighter than surrounding waters, indicating enhanced sediment resuspension and transport.

This study used high-resolution total surface current vectors (TSCV) measured by the airborne radar interferometer OSCAR to investigate fine-scale dynamics in the tidally dominated Iroise Sea around Ushant Island. Observations acquired during ebb tide on 17 May 2022 and flood tide on 22 May 2022 reveal that strong divergence, shear, and estimated vertical velocities arise from two distinct mechanisms. Since observations were acquired on only two days, the study is intended to investigate the fine-scale dynamic structures observed under the sampled tidal conditions rather than to characterise their full range or frequency around Ushant Island.

The first mechanism occurs north and west of Ushant Island, where tidal currents interact with complex shallow bathymetry. Abrupt depth changes deflect the tidal jets away from the island and generate alternating zones of divergence and convergence as the flow crosses bathymetric hills and troughs. Both OSCAR observations and numerical models show positive divergence over upslopes and convergence over downslopes. Although divergence can locally reach large values, the corresponding estimated vertical velocities remain comparatively moderate because the effects of divergence and depth partly compensate each other in the shallow-water formulation.

The second mechanism occurs south of Ushant Island during ebb tide, where a strong tidal jet emerging from the channel between Ushant and Bannec Islands flows alongside comparatively calm waters south of Ushant Island sheltered by the island. The resulting lateral velocity gradients generate intense shear-driven divergence and estimated vertical velocities. In this region, divergence reaches values of O(20 f) and estimated vertical velocity magnitudes approach 0.2 m.s −1. These values exceed those observed in the northern jet and demonstrate the importance of tidal-jet shear as a source of fine-scale vertical transport and mixing.

Comparison with MARS2D shows that the model reproduces the overall magnitude of the tidal currents and captures the bathymetry-driven divergence patterns observed in shallow waters. However, important discrepancies remain. MARS2D places the northern jet closer to the island than observed and predicts the southern jet to turn southward earlier than indicated by OSCAR. As a result, the intense shear-driven divergence observed along the southern jet is underestimated by approximately an order of magnitude, while the predicted shear-rate shows opposite sign from the observations. Statistical comparisons of surface currents and derived quantities indicate that agreement between model and observations is considerably better in shallow waters (0 − 60 m) than in deeper waters (> 60 m), suggesting that the model represents bathymetry-driven fine-scale dynamics more successfully than shear-driven processes associated with tidal-jet boundaries. Further research is required to identify the causes of these discrepancies and determine the relative contributions of model resolution, model configuration and observational factors.

The OSCAR observations further demonstrate the strong sensitivity of derivative quantities to spatial resolution. Coarsening the observations from 200 m to 1 km reduces the normalized divergence, vorticity, and shear rate from O(10 f) to O(1 f), as the narrow transition zones along the jet edges, typically 750 − 1000 m wide, become unresolved. These results show that even kilometer-scale observations may substantially underestimate the intensity of fine-scale coastal dynamics and the associated vertical transport. They also highlight the limitations of point-measurements, such as ADCPs, which cannot capture the two-dimensional structure of tidal jets or their interactions with bathymetry.

The large spatial extent and high spatial resolution provided by OSCAR make it uniquely suited to observing fine-scale coastal dynamics. Future missions such as the Earth Explorer 10 Harmony mission (European Space Agency, 2020) will provide new measurements of surface currents from space from which divergence, vorticity and shear can be estimated. However, their planned spatial resolution (approximately 1 km) and TSCV accuracy (0.2 m.s −1) may still be insufficient to fully resolve the strongest divergence, shear and estimated vertical velocities observed in this study. Airborne systems such as OSCAR therefore remain an important complement to future satellite missions. Although manned airborne deployments are relatively expensive, they offer considerable flexibility in targeting specific locations, tidal phases and oceanographic phenomena. Continued development of airborne TSCV mapping capability could provide versatile temporal sampling of rapidly evolving coastal and submesoscale processes, while also supplying the essential high-resolution observations needed to characterize fine-scale ocean dynamics, validate high-resolution models, and complement satellite, ground-based radar and in situ measurements from instruments such as drifters, ADCPs and autonomous platforms.