Sir Bu Nair Island (SBN) located ∼70km offshore from the southeast coast of the UAE lies within a well-enforced protected area with limited human disturbances (Hadj-Hammou et al., 2025), which extends over 49.6 km2, including ∼36.3km2 of marine habitat surrounding the island (Fig. 1) (Bento et al., 2021). The island is surrounded by a relatively steep bathymetry gradient reaching up to ∼30m depth just ∼3-5km from shore (Fig. 1) (Cavalcante et al., 2024). Geologically, SBN Island's near-circular shape and the surrounding steep bathymetric gradient result from the surface expression of a Hormuz salt dome rising from the stable shelf of the Arabian basement, influencing local seafloor topography (Ellison et al., 2021; Thomas et al., 2015).

The dominant astronomical constituents responsible for the water level fluctuations in the Gulf are the principal semi-diurnal (M2, S2) and diurnal (K1, O1) tides (Fig. 2 – shows M2 and K1). While the Gulf as a whole is often described as mixed, mainly semi-diurnal, the location of SBN is near the central area where the relationship between the semi-diurnal and diurnal tides is crucial (Burt and Paparella, 2024). Previous studies of current meter data near SBN have shown that the local hydrodynamic conditions are influenced by both the diurnal (∼24h) and semi-diurnal (12.5 h) tidal periods (Fig. 2), leading to a complex daily pattern of water flow (Cavalcante et al., 2024).

Time-series measurements of vertical currents, temperature, and absolute water depth were collected and analyzed around Sir Bu Nair Island (SBN). Two ADCP moorings were deployed (Fig. 1). On the northwest flank (SBN-NW; 25.251° N, 54.194° E), an upward-looking Nortek Signature 1000 kHz ADCP was bottom-mounted at ∼17m water depth. The instrument sampled with 0.5 m vertical bins, averaging 120 pings at 1 Hz into 2 min ensembles, saved every 20 min, from 18 May 2021–31 October 2022. On the south side (SBN-S; 25.208° N, 54.221° E), a bottom-mounted Teledyne RDI Workhorse Sentinel 600 kHz ADCP at ∼18m depth used a comparable setup, but due to an equipment malfunction produced two shorter records: 18 May 2021–21 July 2021 and 21 August 2021–19 September 2021.

To characterize the vertical thermal structure at SBN, six moorings equipped with HOBO (TidbiT) temperature loggers were deployed across near-surface to near-bottom depths (refer to Table 1). However, only two arrays were successfully recovered. The SBN-S temperature mooring was co-located with the ADCP frame at ∼18m depth, while the SBN-NE mooring was located ∼750m offshore at ∼16m depth (Fig. 1). Both arrays recorded hourly temperature data from 1 August 2021–30 October 2022. Although the SBN-NW temperature array was not recovered, thermal conditions on the northwest flank were partially reconstructed using an earlier deployment of array of temperature loggers attached to the ADCP frame at nominal depths of 4 and 9 m, which provided hourly data from 19 May–19 September 2021. Earlier temperature records at SBN-S were also incorporated to extend temporal coverage. All records were quality-controlled and harmonized to a common temporal resolution prior to analysis. Details of data availability and integration are provided in Table 1.

In the absence of in-situ wind observations at Sir Bu Nair (SBN), we used 10 m winds (u10, v10) from the ERA5 atmospheric reanalysis for the grid cell enclosing SBN (https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=download (last access: 15 September 2025) (Hersbach et al., 2020). The hourly ERA5 wind fields were extracted for the full mooring period (18 May 2021–31 October 2022), using the nearest 0.25° grid cell without spatial interpolation. All time stamps were converted to a common time base and expressed in local time. Positive (u10) points east and positive (v10) north.

To evaluate ERA5 skill for local winds, we compared against the closest available in-situ record from an oceanographic buoy located ∼24 km southeast of SBN (24.992° N, 54.350° E) (Fig. 1) over 24 April 2025–1 September 2025. Despite the validation period falling outside the mooring window, the analysis indicates strong agreement in vector components (ru=0.84 (along-axis) and rv=0.92 (cross-axis)), supporting the use of ERA5 as a proxy for local wind forcing at SBN. We therefore use ERA5 winds (and derived wind stress where specified) as the atmospheric driver in the analysis.

To assess the spatial and temporal representativeness of the observed temperature conditions, we used daily NOAA Coral Reef Watch (CRW) CoralTemp v3.1 SST for 1985–2025 at 5 km spatial resolution (Skirving et al., 2020). CRW SST was first compared with daily mean temperatures from the shallowest in situ sensors at the three SBN sites (Fig. 1). Agreement was strong across the available records (r=0.964-0.996; RMSE=0.44-0.57°C), supporting its use to provide longer-term and regional thermal context.

We used an analogous approach to assess the spatial and temporal context of the observed circulation using Copernicus Marine Service (CMEMS) current fields for 1993–2025 (Jean-Michel et al., 2021). CMEMS currents have daily temporal and horizontal resolution of 1/12° (∼8km around SBN). The uppermost model layer (0.494 m depth) was used to characterize regional surface circulation.

Hourly ERA5 10 m winds (u10, v10) were converted to eastward and northward wind-stress components (τx, τy) components using the Eqs.(1 and 2): 1τx=ρaCDUu102τy=ρaCDUv10 where ρa=1.225kgm-3 is air density, U=u102+v102 is wind speed, and CD is the neutral drag coefficient following Large–Pond style (Large and Pond, 1981): 1.3×10-3 for U≤11ms-1, CD=0.49×10-3+0.065×10-3U. We then projected (τx, τy) onto the principal axis obtained from principal component analysis (PCA) to obtain along (τa) and across (τc) components and finally applied the same low-pass filter as for currents. To assess sensitivity to drag-coefficient parameterization and to separate dynamic forcing from simple wind-speed covariance, all analyses were repeated using rotated and filtered wind velocity components (u10,v10) in place of wind stress.

Raw 20 min Acoustic Doppler Current Profiler (ADCP) ensembles were averaged to hourly eastward (u) and northward (v) velocity components (ms-1) prior to analysis. Water level, η (m), was derived from the ADCP absolute depth record and averaged to the same resolution. ERA5 winds (u10, v10) and temperature observations were interpolated onto a common hourly time base. Short gaps (≤3h) were linearly interpolated, whereas longer gaps were retained as missing values and excluded pairwise from subsequent calculations. For characterization of the wind regime, wind-speed percentiles were calculated using a 3 d sliding window, and the median was smoothed using a 9 d moving mean. Linear trends were removed prior to filtering to minimize low-frequency leakage (Thomson and Emery, 2024; Bendat and Piersol, 2011).

Subtidal variability was isolated using a 48 h Lanczos digital filter. To account for topographic steering around the island, currents were rotated into along-axis and cross-axis coordinates using the principal axis of variance derived from the low-pass currents. This orientation therefore reflects the dominant subtidal flow relevant to wind-current coupling (Kanarik et al., 2025; Van Der Molen et al., 2022).

The barotropic velocity was defined as the depth-mean flow, and the baroclinic component as the deviation from that mean. Kinetic energy was calculated for each component and expressed relative to total kinetic energy (Simpson et al., 1990). Tidal constituents in currents and sea level were estimated using the MATLAB package t_tide (Pawlowicz et al., 2002). In addition, along-axis and cross-axis currents were regressed against sea level (η) and its Hilbert transform to estimate amplitude and phase relationships with sea level variability (Piecuch et al., 2022).

Linear relationships between wind forcing and currents were quantified using Pearson correlation at zero lag and cross-correlation over lags of ±72h. Analyses were performed between along-axis forcing (τa or u10,a) and along-axis currents (ua), and likewise between cross-axis forcing (τc or u10,c) and cross-axis currents (vc). Lags are reported in hours, with positive values indicating current variations lagging the wind forcing. Pairwise deletion was used where missing values were present. As a sensitivity test, analyses were repeated using three current metrics: the uppermost ADCP bin, the mean of the upper three bins, and the depth-averaged current, to identify the layer exhibiting the strongest wind-driven response.

Frequency-domain relationships were examined with Welch spectra and cross-spectra using a 50 % overlap and applied a Hanning filter (Welch, 1967). Magnitude-squared coherence (MSC) was used to quantify frequency-dependent coupling between forcing and response series (Bendat and Piersol, 2011). The complex transfer function between wind stress (τa, τc) and current response (ua, vc) was used to estimate gain H and phase ϕ. Gain and phase were interpreted only where MSC exceeded the approximate 95 % confidence level based on the number of independent Welch segments (Elipot and Gille, 2009). Spectral filtering and frequency-domain procedures followed standard methods described by Emery and Thomson (2024).

To assess the relative importance of dynamical drivers, band-pass filtered ΔT (temperature difference between 4 and 18 m depth) was regressed against similarly filtered τa, τc, ua, uc, and sea level η. Frequency-band decomposition was performed using fourth-order Butterworth filters applied with zero-phase filtering. Predictor importance was quantified as the reduction in full-model R2 when each predictor was removed in turn (semi-partial variance contribution). These values therefore represent the unique fraction of ΔT variance associated with each predictor after accounting for covariance among forcings. Because predictors are partially correlated, contributions are interpreted as variance attribution rather than strict causality (Grömping, 2007; Peres-Neto et al., 2006).

The wind regime at Sir Bu Nair (SBN) was quantified from the distribution of 10 m wind speed magnitudes and directional frequencies. Figure 3 shows the annual cycle based on ERA5 data, while Fig. 4 provides the corresponding directional distribution.

The 9 d moving mean of the median (red curve, Fig. 3) showed wind speeds fluctuating between 4 and 7 ms-1 throughout the observation period, while the 95th percentile identified discrete peaks in wind magnitude. During winter, these values consistently exceed 10 ms-1. A maximum 95th percentile speed of approximately 15 ms-1 was recorded on 21 January 2022, followed by a secondary peak near 13 ms-1 on 14 May 2022. The 5th–95th percentile range narrowed notably from June through August, indicating reduced variability in daily wind magnitude.

The directional distributions (Fig. 4) showed a bimodal wind field at SBN. The primary mode originated from the Northwest (NW), while a secondary, less frequent mode originated from the Southeast (SE). The highest velocity classes (>10ms-1) were concentrated within the NW quadrant and associated with Shamal winds (“shamal” means north in Arabic). Winds from the SE were predominantly weaker, with most observations within the 3–6 ms-1 range. The NW mode showed the greatest directional stability, particularly during the months when the 95th percentile wind speeds were highest.

Over the study period, the annual temperature range exceeded 15 °C (Fig. 5), with summer temperatures reaching ∼35.5°C and winter temperatures declining to ∼21.2°C. Vertical temperature differences were most pronounced in late spring and summer, reaching maxima of 4.8 °C at SBN-NW, 4.5 °C at SBN-S, and 3.8 °C at SBN-NE. These events were transient and occurred across multiple depth sensors, indicating short-lived periods of pronounced thermal stratification superimposed on the seasonal cycle.

Across all sites, the highest daily mean temperatures occurred in mid-August 2021, when temperatures exceeded 35.0 °C across multiple depths. At SBN-NW, temperatures ≥35.0°C persisted for up to 20 consecutive days, compared with 18 d at SBN-NE and 24 d at SBN-S. Exceedance of 35.5 °C was restricted to shallow depths and lasted up to three consecutive days at SBN-S and SBN-NW.

The regional analysis showed that the warm conditions observed in 2021 were not restricted to SBN. The 1985–2025 record showed that JAS 2021 was anomalously warm at SBN (mean SST=34.02°C; +0.79°C relative to the long-term JAS mean; 95th percentile), whereas JAS 2022 was close to climatological conditions (33.14 °C; -0.09°C; 32nd percentile) (Fig. S1 in the Supplement). Spatial SST anomalies showed that the 2021 warming extended across much of the southeastern Gulf, whereas anomalies in 2022 were substantially weaker (Fig. S1).

The harmonic analysis results for SBN-NW and SBN-S are summarized in Table 2. The tidal form number, F=(K1+O1)/(M2+S2) was ∼0.95 at both sites, indicating mixed tides with semidiurnal dominance. M2 was the largest constituent (36–37 cm), followed by K1 and O2 with mean semidiurnal range of ∼0.72-0.74m. Under constructive interference, the astronomical tidal range could exceed ∼1.5m.

Phase differences for the dominant constituents were small between stations, indicating a coherent regional tidal signal. Tidal current ellipses were predominantly rectilinear, with strong currents at SBN-NW than SBN-S Maximum semimajor axes occurred for M2 and K1 at SBN-NW (15–16 cms-1), compared with ∼12cms-1 for M2 and ≤3cms-1 for K1 at SBN-S. Diurnal constituents were oriented approximately north–south at SBN-NW but were more variable at SBN-S. Shallow-water constituents were small (∼1-2 cm), indicating weak tidal distortion at both sites.

Currents at SBN-NW and SBN-S exhibited distinct directional structures (Fig. 6). At SBN-NW, flow was strongly aligned along a principal axis oriented ∼29°, accounting for >94% of the variance and indicating weak cross-axis motion and strong vertical coherence (Fig. 6A and B). At SBN-S, currents were oriented predominantly east–west (∼91°), with directional dominance decreasing from ∼85% near the bottom to ∼67% in the upper layer and greater cross-axis variability (Fig. 6C and D). Upper-layer along-axis currents at SBN-NW reached ∼0.6ms-1 (95th percentile), compared with ∼0.3ms-1 at SBN-S (Figs. 7 and 8), while cross-axis velocities generally remained within ±0.1-0.2ms-1 at both sites.

The tidal harmonics explained 80 % of the along-axis current variance at SBN-NW but only 35 % of the cross-axis variance. Tidal influence was weaker at SBN-S, explaining 39 % and 12 % of the along and cross-axis variance, respectively. Sea level relationships showed a similar spatial pattern, with stronger coupling at SBN-NW.

At the regional scale, the 1993–2025 CMEMS record showed that surface-current speeds during the 2021 observational period were moderately below the long-term mean but remained within historical variability (0.096 versus 0.102 ms-1; 24th percentile) (Fig. S2a). The mean current at SBN-S (0.089 ms-1, 78.2°) was broadly consistent with the regional CMEMS flow (0.081 ms-1, 98.7°), whereas the mean current at SBN-NW (0.1 ms-1, 26.6°) was strongly rotated relative to the regional flow (0.083 ms-1, 103.6°) (Fig. S2b–e).

Low-pass filtered velocities showed a persistent residual flow of ∼0.08-0.1ms-1 along the principal axis at both sites (Figs. 7 and 8). Subtidal flow was predominantly barotropic, with 94 % and 89 % of the kinetic energy associated with the depth-mean component at SBN-NW and SBN-S, respectively, and strongly anisotropic, with 97 % and 92 % aligned with the principal axis at SBN-NW and SBN-S, respectively. Subtidal kinetic energy was substantially higher at SBN-NW (0.09 m2s-2) than at SBN-S (0.01 m2s-2).

Wind–current relationships were evaluated for along- and cross-axis currents using wind stress and velocity, considering depth-averaged currents and the average of the top three ADCP bins, with representative cases shown in Fig. 9. Results were insensitive to 40–48 h filter cutoffs and consistent across seasonal windows (not shown).

Wind forcing showed a moderate relationship with the residual currents. At SBN-NW, the strongest relationship occurred in the cross-axis component, with correlated reaching r=0.62 for wind velocity and r=0.50 for wind stress at ∼1h lag (Fig. 9a). At SBN-S, coupling was weaker and occurred primarily along-axis (r=0.35 for wind velocity; Fig. 9b), while cross-axis correlations were negligible.

Wind–current coherence increased at subtidal periods, particularly between 3 and 8 d. At SBN-NW, cross-axis coherence reached 0.4–0.5 (Fig. 9c), whereas at SBN-S, along-axis coherence reached ∼0.4 (Fig. 9d). Transfer functions showed a corresponding increase in gain at 2–8 d periods for these components, with phase estimates corresponding to hour-scale lags (Fig. 9e and f). Overall, wind coupling was strongest in the cross-axis currents at SBN-NW, with a weaker but detectable along-axis response at SBN-S.

At SBN-S, the surface-to-bottom temperature difference (ΔT) increased during spring and summer and repeatedly decreased towards zero, whereas autumn and winter were characterized by small and occasionally negative differences (Fig. 10). These transitions occurred during both warm seasons (2021 and 2022), indicating recurrent thermal variability. A period of pronounced vertical thermal variability from 20 May–25 June 2021 was selected for detailed analysis because concurrent observations were available at both SBN-NW and SBN-S (Fig. 11).

Both sites were strongly thermally stratified before ∼10 June, interrupted by a brief weakening around 28–29 May followed by rapid re-establishment of vertical temperature differences. Temperatures then converged across depths over ∼2-3 days around 10–12 June, producing near-thermal homogeneity at both sites (Fig. 11), although higher-frequency variability was more pronounced at SBN-NW. Before this transition, vertical temperature differences (ΔT=4-18m) were generally ∼2-4°C and occasionally exceeded 4 °C, with surface temperatures of ∼32-33°C overlying bottom temperatures of ∼28-30°C.

During this period, along-axis currents at SBN-NW frequently exceeded 0.3 ms-1 and were vertically coherence, with episodic intensification throughout the water column (Fig. 12a and c). At SBN-S, currents were weaker (∼0.2ms-1) and less vertically coherence (Fig. 12b and d). Progressive vector diagrams showed stronger, more directional northeastward displacement at SBN-NW, whereas trajectories at SBN-S were weaker, more dispersed, and predominantly eastward (Fig. 13b).

Spectral analysis of (ΔT) showed similar frequency-dependent variability at both sites (Fig. 14a), with most energy at lower frequencies (2–4 d), a pronounced diurnal peak (∼0.04cph), and a smaller semidiurnal peak. The diurnal and lower-frequency signals were also evident in the full-record SBN-S spectrum (Fig. S3). Sea level (η) and currents showed pronounced variability at tidal frequencies, with site and component-specific differences, whereas wind stress was dominated by lower-frequency variability (Fig. 14b and c).

At SBN-NW, coherence with ΔT was strongest in the diurnal band, reaching ∼0.55 for cross-axis wind stress (τc) and along-axis currents (ua), with lower coherence for sea level (η), and along-axis wind stress (τa) (Fig. 15a). Cross-spectral density was also largest in this band, particularly for ua (Fig. 15c). At SBN-S, diurnal coherence was weaker (∼0.35), with correspondingly lower cross-spectral density (Fig. 15b and d).

Secondary coherence occurred near semidiurnal frequencies, although ΔT spectral energy was small. A quarter-diurnal (M4) signal was evident in the hydrodynamic field, particularly between sea level and along-axis currents at SBN-NW (coherence ∼0.50), but its coherence with ΔT remained <0.25. Coherence at the 2–4 d periods was generally modest. Significant diurnal relationships corresponded to phase-derived lags of a few hours.

Band-limited variance partitioning showed that ΔT variability was greatest in the diurnal band, followed by the 2–4 d and semidiurnal bands (Fig. 16). At SBN-NW, the diurnal band explained 67 % of ΔT variance, with the largest unique contributions from cross-axis wind stress (τc; 9.8 %) and along-axis currents (ua; 8.8 %). The 2–4 d band explained 39 %, with sea level providing the largest unique contribution (12 %), whereas the semidiurnal band explained 31 % and was primarily associated with currents (Fig. 16a and c).

At SBN-S, the diurnal band explained 54 % of ΔT variance, with the largest unique contribution from along-axis wind stress (τa; 14.5 %), sea level (η; 11.1 %). The 2–4 d band explained 36 %, again with the sea level providing the largest unique contribution (12 %), while the semidiurnal band explained 16 % (Fig. 16b and c).

The observations provide the first integrated description of circulation around Sir Bu Nair Island (SBN) and show that local hydrodynamics reflect the interaction between tidal forcing, regional circulation, and the island's steep surrounding bathymetry. At both mooring sites, currents were highly anisotropic and aligned with a dominant principal axis, indicating rectilinear flow strongly influenced by local bathymetry rather than freely rotating shelf circulation.

At SBN-NW, upper-layer along-axis currents commonly reached ∼0.6ms-1, with episodic peaks approaching 0.8–1.0 ms-1, consistent with values reported from a shorter deployment (Cavalcante et al., 2024). Currents at SBN-S were weaker, commonly reaching ∼0.3ms-1. This directional asymmetry is typical of island and reef environments where tidal and residual flows can be steered by bathymetry gradients and coastline geometry into preferred pathways (Lowe and Falter, 2015; Wolanski and Hamner, 1988). The regional CMEMS comparison further showed that the mean flow at SBN-S was broadly aligned with the regional circulation, whereas the flow at SBN-NW was strongly rotated relative to the regional field. This contrast was consistent with substantial local modification of the regional flow around the island. Circulation around SBN therefore cannot be represented by a single island-scale state but varied substantially between adjacent flanks.

Subtidal currents were also predominantly barotropic, with most kinetic energy contained in the depth-mean flow and concentrated along the principal axis, indicating that the rectilinear circulation pattern persisted at lower frequencies. This vertically coherent flow is consistent with previous Gulf studies showing that tidal shear, bottom friction, and shallow bathymetry promote strong mixing and depth-uniform currents over broad areas of the basin (Alosairi and Pokavanich, 2017; Kämpf and Sadrinasab, 2006; Salim et al., 2024; Cavalcante et al., 2012). Periods of thermal stratification in late spring and summer, however, were accompanied by increased vertical differentiation in the flow, particularly from late May to mid-June 2021 (Fig. 11). During these periods, surface waters were 2–4 °C warmer than bottom waters and vertical shear increased.

Tidal forcing was the primary driver of current variability at SBN, particularly along the principal flow axis. Harmonic analysis showed that tidal constituents explained ∼80% of the along-axis variance at SBN-NW, but only ∼35% of the cross-axis variance. At SBN-S, tidal control was weaker, accounting for ∼39% of the along-axis and ∼12% of the cross-axis variance. Spectral analyses further showed that the principal tidal energy was concentrated in the diurnal and semidiurnal bands, consistent with previous Gulf-wide studies (Reynolds, 1993; Cavalcante et al., 2011), although both peaks were less pronounced at SBN-S.

This interpretation was reinforced by the coherence and phase relationships between sea level and currents. At SBN-NW, sea level was strongly coherent with along-axis currents in both the diurnal and semidiurnal bands, while cross-axis currents also exhibited substantial coherence, consistent with local modification of the tidal flow around the island. At SBN-S, the strongest coupling occurred in the semidiurnal along-axis band, whereas other bands were weaker and more variable. Together with the stronger current amplitudes and directional organization at SBN-NW, these results indicated more energetic and organized tidal exchange at SBN-NW than at SBN-S.

A secondary quarter-diurnal peak (M4, ∼0.161cph) was also present in the spectra. Previous work interpreted this feature as possible seiche activity around SBN (Cavalcante et al., 2024). However, its coincidence with the expected M4 overtide of the dominant semidiurnal tide, together with its weak association with ΔT in the present analysis, was more consistent with nonlinear tidal harmonic generation around the island and adjacent shoals. By contrast, energy near the 25–26 h period identified from multi-year Gulf sea level records is more consistent with previously proposed Gulf-scale seiche dynamics (Afshar-Kaveh et al., 2020). Thus, although tides dominated currents and sea level variability, they did not fully explain the observed current variability, particularly in the cross-axis component and at subtidal timescales.

Wind forcing emerged as an important secondary influence within this tide-dominated regime. The clearest responses occurred in the cross-axis flow at SBN-NW and the along-axis flow at SBN-S, where correlations peaked at short lags (0–2 h) and coherence increased at subtidal periods of 3–8 d. These responses were consistent with synoptic timescales of Shamal variability (Al Senafi and Anis, 2015; Makarem et al., 2026), which influences Gulf circulation and sea level setup (Thoppil and Hogan, 2010; Kämpf and Sadrinasab, 2006; Pous et al., 2004; Reynolds, 1993). Although secondary to tides, wind forcing could therefore have modified residual transport pathways around SBN (Cavalcante et al., 2020). The short response times were consistent with rapid local barotropic adjustment through wind-driven setup/set-down and associated pressure gradients, rather than a classical offshore Ekman response.

The repeated development and rapid reduction, typically over 2–3 d, of vertical temperature differences at SBN characterized a highly dynamic thermal regime. During the May–June 2021 event (Fig. 11), ΔT exceeded 4 °C before temperatures converged rapidly across the observed depths, indicating an abrupt thermal homogenization rather than gradual erosion of the vertical temperature gradient. Comparable episodes occurred across all three mooring locations during 2021–2022, with vertical temperature differences commonly reaching 2–4 °C, indicating that these transitions were not restricted to individual sites. Their seasonal timing was consistent with strong late spring and summer surface heating in the Gulf (Burt and Paparella, 2024; Paparella et al., 2019), which promotes the development of warm surface waters over cooler deeper layers, as previously observed around SBN (Cavalcante et al., 2024).

The persistence of large vertical temperature differences despite energetic tides indicated that tidal stirring alone was insufficient to suppress summer thermal stratification. Within the diurnal band, ΔT variability was associated with both current and atmospheric variability. Coherence and variance partitioning (Figs. 15 and 16) showed particularly strong associations at SBN-NW between ΔT, cross-axis wind stress (τc) and along-axis currents (ua). The diurnal band explained 67 % of ΔT variance at SBN-NW and 54 % at SBN-S. These relationships indicated that diurnal variability in wind forcing and currents was closely associated with changes in vertical thermal structure, superimposed on the tidally energetic background circulation.

At lower frequencies, the 2–4 d band accounted for 36 %–39 % of ΔT variance at both sites, with sea level (η) providing the largest unique contribution (∼12%; Fig. 16). This timescale overlapped with synoptic wind variability in the Gulf, including Shamal events, suggesting that atmospheric forcing and associated sea-level and current adjustments may have contributed to the development and subsequent reduction of vertical temperature gradients. The rapid, near-synchronous convergence of temperatures across observed depths during the June 2021 event (Fig. 11) indicated a rapid transition toward thermal homogeneity, but the absence of concurrent salinity observations prevented assessment of density stratification and limited our ability to distinguish among vertical mixing, horizontal advection, or their combined effects. The observed variability therefore suggested an interaction between lower-frequency preconditioning and shorter-timescale tidal and atmospheric forcing, rather than control by a single mechanism.

Over the past decade, coral reefs across much of the southern Gulf have declined following recurrent marine heatwaves, with losses of more than 80 % of live coral documented in some areas (Burt, 2024). SBN stands out in this context, with only one bleaching event recorded over the same period, comparatively modest coral losses, and the only known large stands of heat-sensitive Acropora table corals remaining in the southern Gulf after their extirpation elsewhere in the region (Hadj-Hammou et al., 2025). The hydrodynamic setting identified here may contribute to this relative resilience. thermal exposure across reef depths was strongly modulated by vertical temperature variability

SBN's offshore location (>70km) reduces its exposure to coastal anthropogenic pressures (Howells et al., 2020; Hadj-Hammou et al., 2025), while our results showed that thermal exposure across reef depths was strongly modulated by vertical temperature variability. During summer 2021, temperatures>35°C persisted for up to 24 consecutive days (Fig. 5). Under these conditions, energetic tidal currents were insufficient to prevent persistent vertical temperature gradients and prolonged exposure of shallow reef habitats to high temperatures. This period coincided with the first recorded bleaching event at SBN and losses of heat-sensitive Acropora species (Hadj-Hammou et al., 2025).

These observations suggest an important relationship between physical forcing and reef thermal exposure at SBN. Under typical summer conditions, episodic wind forcing may contribute to the disruption of vertical temperature gradients and exchange between cooler subsurface waters and shallow reef habitats, potentially reducing cumulative thermal stress (Paparella et al., 2019; Yao and Johns, 2010b). Conversely, prolonged weak-wind conditions may favour the persistence of strong vertical temperature gradientsand heat retention in the shallow reef zone, increasing exposure to temperatures associated with bleaching. The potential role of wind forcing in modulating reef thermal exposure may become increasingly relevant under climate change, given the reported decline in summer Shamal frequency since the mid-1990s (Al Senafi and Anis, 2015), and increasingly recurrent bleaching events across Gulf reefs (Riegl et al., 2018).

Spatial variability in circulation may also influences SBN's role in regional connectivity. Previous modelling studies identified SBN as a potential larval source for degraded reefs along the southern Gulf coast (Riegl et al., 2017; Cavalcante et al., 2016). The present observations provide physical context for this interpretation, showing substantial differences in current strength and direction between adjacent island flanks and in their relationship with the regional circulation. Such spatial heterogeneity is likely to influence pathways of larval export and retention around the island. SBN's capacity to function as both a relative thermal refuge and a source of larvae therefore likely depends on the interaction between regional and locally modified circulation and episodic thermal stress.

Overall, circulation and thermal variability at SBN reflected a hierarchy of interacting processes, with tides establishing the dominant current variability and flow structure, and wind and lower-frequency sea level and current variability associated with changes in residual circulation and vertical thermal structure. Together with strong topographic steering, these interacting processes produced spatially and temporally heterogeneous conditions around the island, with implications for physical exchange, thermal exposure, and reef connectivity.