The shelf break region in the southern Weddell Sea (Fig. ) is an area of great climatic interest. This is where cold and dense water masses, formed on the continental shelf and underneath the Filchner–Ronne Ice Shelf (FRIS), cross the shelf break and descend the continental slope , ultimately contributing to the formation of Antarctic Bottom Water which spreads out into the major oceans at abyssal depths .

Furthermore, warm off-shelf water, referred to as Warm Deep Water (WDW), crosses the shelf break during austral summer in the form of modified Warm Deep Water MWDW, . The MWDW flows southward towards the Filchner Ice Shelf along the eastern flank of the Filchner Depression (see map in Fig.  for location; ) and reaches, at least occasionally, the Filchner Ice Shelf front during austral autumn . Some climate models suggest a larger inflow and a dramatic increase in basal melt rates below the FRIS within the next century .

Physical processes at the shelf break and on the continental slope influence both the cold outflow and the warm inflow in terms of their hydrographic properties and strengths. The variable depth of the thermocline, for example, which is controlled mainly by wind forcing and eddy overturning will determine if and when warm water can access the continental shelf .

Meanwhile, variability in off-shelf water properties will alter the density contrast between the cold outflow and the ambient water, and thus the strength of the geostrophically balanced outflow . It will also influence the properties of the descending dense plume, since it is a mixture of outflow water and ambient water . The co-location of the critical latitude for the tidal component M2 and a critical slope leads to enhanced turbulence levels in the region . Mixing can be expected to be further enhanced at the shelf break by the strong diurnal tidal currents . The strong diurnal tidal currents in the study region have been linked to the presence of continental shelf waves , a class of coastal trapped waves (CTWs).

CTWs can be generated by, e.g. tides or wind . Additionally, a connection between the generation of the waves and the outflow of dense shelf water through troughs has been suggested . CTWs with sub-inertial frequencies propagate along a trapping boundary, e.g. a coastal wall or a sloping bottom . The waves require the support of such a boundary to exist, and their energy decays exponentially with increasing distance from it . While the direction of phase propagation is (with shallow water) to the left (in the Southern Hemisphere), the group velocity cg of CTWs, and thus the energy associated with the waves, can propagate in either direction (Fig. ). If the group velocity is zero, i.e. for a maximum in the dispersion curve of a wave, energy cannot propagate. When the frequency of this maximum (hereafter called resonant frequency, RF) in the dispersion relation coincides with the frequency of tidal forcing, resonance may occur and tidal currents will be amplified. In practice, energy likely escapes in one or the other direction along the slope. Leakage of energy occurs, for example, because of irregularities in the bathymetry and because the bottom slope changes i.e. isobaths converge or diverge;. Therefore, we use the term near-resonance rather than resonance.

Such near-resonant diurnal CTWs were first recorded on the shelf of the Outer Hebrides of Scotland by and have been observed and modelled at numerous locations on numerous occasions since then e.g. . In our study region, and first suggested that CTWs caused the observed strong diurnal tidal currents, and they attributed a weakening of currents in austral winter to a seasonally varying stratification. Later, and found a particularly strong enhancement of the K1 tidal constituent during austral summer. The summer maximum was hypothesised to be due to the interaction of barotropic CTWs with topography in the presence of a seasonally variable mean current . The authors showed how changes in the background current will affect the phase of diurnal CTWs, which are assumed to be generated upstream, as they arrive in the study region. These studies were based on a small number of moorings and a barotropic shelf wave model neglecting the effects of stratification. Studies by, e.g. , , and indicate that, in addition to the currents, seasonally varying hydrography alters the properties of CTWs.

Another seasonal phenomenon which can potentially influence the CTW generation is sea ice. Frictional damping of tidal CTWs due to sea ice is suggested to be the cause of the observed reduction of tidal currents over the shelf in the Sea of Okhotsk during winter when the sea ice cover exceeds 80 % . Strong tidal currents in turn are important for local sea ice deformation and hence sea ice concentration . A study by showed that CTWs can locally enhance sea ice drift.

To the east of our study region, the continental slope steepens considerably (Fig. ), and isobaths hence diverge in the direction of CTW propagation. Numerical simulations from the Barents Sea region showed that tidally generated CTWs were confined to a region of divergent bathymetry .

Our study expands on previous investigations of tidally generated CTWs at diurnal frequencies in the shelf break region of the southern Weddell Sea. Records from 29 moorings collected over more than 4 decades are used to quantify the strength of the diurnal tidal currents in the area and to describe their temporal and spatial variability. The extended data set confirms the existence of an enhancement of the diurnal tidal currents during austral summer and shows that it is a persistent phenomenon. Previous studies – based on a small sub-set of our data records – have suggested that the summertime enhancement is due to changes in the oceanographic ”background”, as it determines the dispersion relation for the CTWs which are responsible for the tidal amplification in the area. We investigate this further by using a numerical code to study the sensitivities of the CTW properties and the RF to seasonal changes in hydrography and background current. With this aim, we synthesise observational data from the continental slope providing a novel description of the seasonal changes in shelf break hydrography. The results from the numerical code suggest that the effect of a changing stratification dominates the effect of the background current. Finally, we investigate the role of divergent bathymetry and the seasonal variability of sea ice cover in the southern Weddell Sea, and we discuss the potential relevance for tidally generated CTWs.

Current metre data from 29 moorings located on the continental slope and shelf in the area surrounding the Filchner Depression have been analysed. The records span the years 1968 to 2014 and are each of 1–2 years duration. The locations of the moorings are shown in Fig. , and deployment details are listed in Table . The mean currents included in Fig.  are vertical means for moorings M1 to M5, while for moorings C, W2, W3, F3 and F4 only the uppermost instrument has been considered to minimise the effects of the Ice Shelf Water plume. Temperature records suggest that plume water is rarely present at these levels.

The coordinate system is rotated clockwise to align the y axis with the isobaths, agreeing with the set-up of the numerical code in the Southern Hemisphere. u is thus directed on the shelf and v along the continental slope (Fig. ). The rotation angle φ, positive for clockwise rotation, is listed in Table ; it is inferred for each mooring from the local bathymetry based on the GEBCO_2014 bathymetry grid (the GEBCO_2014 grid, version 20150318; http://www.gebco.net) and using an average length scale of the order of 10 km. The estimated accuracy of φ is approximately ±10∘.

Time series of kinetic energy (KE) associated with the diurnal tidal currents are constructed as follows: the hourly-averaged current metre data are divided into intervals with a length of 1.5 months beginning every 14th day. For each interval, the power spectral densities are estimated using Welch's method and 14-day, 50 % overlapping Hanning windows.

The diurnal tidal KE is obtained by integrating the velocity spectra, KE=∫ω1ω2(Su+Sv)dω, where, following , ω1 and ω2 correspond to periods of 26.9 and 21.3 h, respectively.

Diurnal tidal KE has also been inferred using tidal predictions from the Circum-Antarctic Tidal Simulation version 2008b (CATS2008b), an updated version of the linear tidal inverse model described by . The barotropic currents at the specific tidal frequencies are predicted for the respective time and location of every mooring deployment, and tidal KE is directly calculated as a monthly running mean from the amplitudes of the predicted tidal currents.

Tidal ellipses, i.e. major and minor axes, inclinations and Greenwich phases, have been obtained from the mooring records using harmonic analysis T_TIDE; , a Matlab version of the Fortran code developed by .

Records of temperature and salinity from mooring M3, located at the 725 m isobath just east of the Filchner Depression sill (Fig. ), are used to describe the seasonal changes in hydrography at the shelf break and upper continental slope. The mooring records are complemented by a conductivity–temperature–depth (CTD) profile obtained during the deployment cruise in 2009 and by hydrographic measurements obtained in the vicinity of the M3 location (within 10 km; Fig. ) provided by seals tagged with small CTD sensors described inhereafter referred to as “seal data”. The accuracies of the seals' temperature and salinity measurements are stated to be 0.005 ∘C and 0.02, respectively .

In addition, we use wind observations from the Halley Research Station, located at 75∘35′ S, 26∘39′ W (Fig. ), from 1957 to 2014 and satellite-derived records of sea ice concentration , available for the period 1978 to 2014. The sea ice concentration is averaged over the study area (inset in Fig. ).

Observations from the continental slope in the southern Weddell Sea show anomalously strong tidal currents at diurnal frequencies . Our extended analysis – including all current metre records (1968–2014) from the region – confirms previous findings suggesting that the strong currents are the result of tidally forced CTWs .

The observations agree qualitatively with the mode 1 CTW generated in the numerical code provided by . As expected, the rotational direction of the observed and simulated currents changes from anticlockwise on the upper part of the continental slope to clockwise on the deeper part of the slope, and the strength of the diurnal tidal currents increases towards the shelf break (Fig. ). The wavelengths inferred from the observations are generally consistent with those obtained from the numerical code for cases where CTWs of diurnal frequencies are allowed.

Time series of the KE associated with the diurnal tides show a persistent pattern with two annual peaks (one in austral summer and one in austral winter) and minima in spring and autumn. The austral summer peak is enhanced for all moorings by 30 to 180 % compared to the austral winter peak.

The semi-annual signal, i.e. the two peaks, is at least partly the result of astronomical forcing: around the equinoxes, the sun is nearly above the Equator (and also the moon within about a fortnight so that both their declinations are small); hence, the diurnal forcing is at its minimum. In June and December, the solar declination is at its maximum and the solar contribution to the diurnal tide is greatest.

This results in two semi-annual peaks of diurnal tidal KE with similar amplitude (Fig. b). At the mooring locations M3 and M5, the diurnal KE of the tides directly calculated from the prediction of the CATS tidal model is reduced by about 20–25 % near the equinoxes, while the observations typically show a reduction of more than 50 % compared to the winter values.

It is possible that other factors contribute to the observed semi-annual variability. For example, there may be a semi-annual cycle of mixing (and hence stratification) caused by the semi-annual tidal signal. Alternatively, the observed variability of diurnal KE may be the sum of the effect of an annual cycle in stratification that is phase shifted relative to the effect of the changing background current (L. Padman, personal communication, 2016).

We hypothesise that the enhancement of tidal KE during austral summer is caused by near-resonance of diurnal CTWs. The obtained dispersion relations suggest that diurnal CTWs may be near-resonant, i.e. that the group velocity is zero or close to zero at diurnal frequencies so that energy cannot propagate out of the area, which results in amplified diurnal tidal currents. Near-resonance of diurnal CTWs in the study region was also suggested by using a barotropic shelf wave model . The dispersion relation is shown to be relatively sensitive to changes in the upper ocean stratification, and the observed seasonal changes in upper ocean hydrography discussed below cause the RF to move through the diurnal tidal band, so that diurnal CTWs are “nearer” to resonance during austral summer than during austral winter, when the RF falls above the diurnal band. Tidal currents may be enhanced for a range of frequencies surrounding the RF, and thus the RF does not need to coincide with one exact tidal frequency for amplification to occur .

The largest seasonal changes in the shelf break hydrography in the region occur above the pycnocline, similar to regions farther east in the Weddell Sea . Cooling and a gradual increase in salinity (due to ice freezing and brine rejection) during austral autumn and winter lead to a gradual deepening of the surface layer. Towards the end of the winter (August–September), the upper 400 m are relatively homogeneous. During austral summer, the winter layer is capped by a fresh and relatively warm surface layer which likely is the result of local sea ice melt and solar heating. The layer of summer surface water is thin (10–100 m; see CTD profile in Fig. ) and greatly increases the stratification by creating a seasonal, shallow pycnocline. The sensitivity test (Fig. ) shows that the value of the RF is sensitive to the stratification in the upper layer (SM) and that it increases for decreasing stratification. The response in RF to realistic changes in SM is of sufficient magnitude to cause the RF to move through the diurnal tidal band.

While the RF is influenced by changes in the strength of the permanent (deeper) pycnocline (SSM), which is the transition from WW above to MWDW and WDW below, there is no observational evidence suggesting that it would change in magnitude. The depth of the permanent pycnocline (SSD), however, increases from about 400 m in austral summer to about 600 m, but changes in SSD have little or no influence on the RF.

Despite the fact that the stratification affects the dispersion curve of CTWs considerably (Fig. ), CTWs are relatively barotropic in the region Figs.  and ; . This is expected since the Burger number is Bu=NHfL2≈O(10-4)≪1, where N2 is the stratification (N= (0.97–1.7) × 10-3 s-1), f the Coriolis factor (-1.4 × 10-4 s-1) and H (400–600 m) and L (300–500 km) are representative depth and length scales, respectively.

suggested that the observed seasonality was linked to the variability of the slope current. The available observations of the slope current are limited and do not allow a detailed description, but the data from moorings M1 to M5 suggest, in agreement with observations upstream , that the westward flowing slope current is intensified during austral winter. When a barotropic, westward background current is included in our numerical set-up, the dispersion curve (and thus the RF) is shifted toward higher frequencies Fig.  and, but the effect is small compared to the effect of stratification changes. The combined effect of stratification and current, however, is considerable, as the stronger current observed during austral autumn and winter will add to the effect of the low wintertime stratification and move the RF to higher frequencies.

The low values of tidal KE during austral winter can potentially be caused by frictional damping of CTWs by sea ice. When sea ice is in free drift , no tidal energy is dissipated at the ocean–ice interface. However, as ice concentration increases and internal ice stresses prevent the ice from responding to local tidal currents, the stress at the ocean–ice interface may be significant compared with friction at the seabed, thus removing tidal energy and reducing tidal currents .

This mechanism was suggested by to explain the reduction in tidal CTW energy observed at one of their mooring sites during winter. The reduction was only observed at the site away from the region of CTW generation, indicating the cumulative effect of frictional dampening over distance. However, the authors fail to explain why the period with reduced tides is much shorter (and misaligned) compared to the period with dense (> 80 %) sea ice cover.

The ice cover in the study region normally exceeds 90 % during austral winter (mid-April to mid-November) and decreases to a minimum of on average 50 % in February (Fig. ), and high (low) sea ice concentration hence coincides with low (high) diurnal tidal energy levels, as expected if frictional damping by sea ice is important. The semi-diurnal tidal currents, however, are observed to be larger during austral winter than during summer , which is inconsistent with the response if damping by ocean–ice interactions was a significant factor. We can unfortunately neither quantify nor rule out the relevance of sea ice concentration and sea ice damping to the observed seasonality of the diurnal tidal currents.

While the tidal force is the main generation mechanism for CTWs in the diurnal tidal band , CTWs can also be generated by wind . Short duration storms have been observed to excite near-resonant CTWs of mode 1 , i.e. the response to storms would, in our case, resemble the tidally forced waves. Time series of wind from the nearby Halley Research Station (see Fig.  for location), however, show that storms (wind speed > 20 m s-1) are rare during austral summer and, as expected, more frequent during winter and early spring. CTWs induced by storms can hence not explain the summer enhancement of the diurnal tidal currents. Fourier analysis of the time series reveals a daily cycle in wind strength with an increase of magnitude of up to 1.4 m s-1 around noon, which likely results from local boundary layer effects: the stable boundary layer which develops during the night is destroyed during the day by mixing due to solar insolation see, e.g. . Since the signal is weak, we conclude that these oscillations are not responsible for the observed summer amplification.

The CTWs owe their existence to topography, and the dispersion curve is sensitive to changes in bathymetry e.g.and Fig. . Just east of our study region, the continental slope is much steeper (Fig. ), i.e. the isobaths diverge towards the west. Divergent bathymetry has been shown to favour the generation of diurnal CTWs . On a divergent slope, tidal energy travelling along the slope may slow down and converge as the group velocity decreases due to the changing topography, while energy travelling in the other direction will speed up and diverge. There are no direct observations of tidal currents from the steep eastern region, but the tidal motion of sea ice above the shelf break there suggests weaker currents compared to the study region . The wavelengths of the diurnal CTWs are typically large compared to the length scale over which the changes in bathymetry discussed above occur and the scale of other topographic features in the area. While the implications of this are beyond the scope of the current study, we note that the CTWs modelled by similarly have wavelengths which are considerably larger than the topographic scales. A full 3-D analysis, similar to the one by , would be needed to fully explore the effect of bathymetry on CTWs in the study region.

The anomalously large diurnal tidal currents and the CTWs will potentially influence the exchange of MWDW across the shelf break. In an idealised model study, CTWs were shown to enhance the inflow of warm water through a trough cross cutting the continental shelf , similar to the Filchner Depression. We note that the depth of the WW–WDW transition (here identified by the -1 ∘C isotherm) varies on diurnal timescales (Fig. a), and, e.g. in December 2009, the vertical excursion of the isotherm associated with the diurnal tides is > 100 m (Fig. b). The depth of the transition is likewise affected by CTWs with a 35 h period Fig. ;. The existence and strength of diurnal and longer; CTWs in the region must hence be expected to directly influence the availability of warm water above the shelf depth, i.e. at depths where it can potentially access the continental shelf through the influence of other processes, e.g. a background mean flow, rectified tidal flows, friction or eddy exchanges. In addition, the tides in the area will greatly influence mixing and hence modify the stratification in the shelf break area. Modelling efforts aiming to describe and predict the oceanic heat transport towards the FRIS cavity thus ought to include tidal forcing to correctly capture the dynamics at the shelf break.

Finally, we mention that the observed diurnal tidal currents are up to 1 order of magnitude larger than those predicted by the tidal model CATS2008b Fig. ,, which does not include stratification or the variability of mean circulation required to predict seasonal modulation of tidal currents. Moreover, due to discrepancies in the model bathymetry, the predicted peak tidal currents are not consistently aligned with the shelf break when running CATS along a cross-shelf section through the locations of moorings M1 and M2 (see Figs.  and  for location of the section). Hence, care must be taken when using CATS to de-tide velocity observations from the study region.

Velocity measurements at 29 moorings located on the continental slope and shelf in the southern Weddell Sea during the period 1968 to 2014 show pronounced diurnal tidal variability. Diurnal tidal currents are strongest at the shelf break and substantially enhanced during austral summer. The summer enhancement is not predicted by the tidal model CATS2008b , as the model does not include stratification or the variability of mean circulation. We investigated the possibility for near-resonant CTWs causing the enhanced diurnal tidal currents by using a 2-D numerical code to obtain CTW properties . Dispersion curves of mode 1 CTWs have a maximum in frequency (the resonant frequency, RF), which results in zero group velocity, i.e. trapped energy. The RF moves in and out of the diurnal tidal frequency band depending on the stratification and the slope current which both vary seasonally as hydrographic and current observations at the shelf break reveal. For the weakly stratified water column and strong slope current during austral winter, the RF is found above the diurnal band, suggesting the generation of weak, non-resonant tidal CTWs which quickly propagate out of the generation area. For austral summer conditions, i.e. a more stratified upper water column combined with a weaker slope current, the RF can fall into the diurnal band, thus leading to near-resonant diurnal CTWs enhancing the tidal currents.

While no direct influence of wind on the diurnal tidal currents and no evidence of sea ice affecting the diurnal CTWs have been found, the varying bathymetry east of the study area likely affects the propagation of the CTWs. Studies with realistic 3-D ocean models are needed to quantify these influences as well as to detect the generation site of the CTWs.

The shelf break region in the southern Weddell Sea is an area of great climatic interest. Cold, dense water descends the continental slope and contributes eventually to the formation of Antarctic Bottom Water, while warm WDW and MWDW flowing onto the shelf prospectively may reach the cavity below FRIS, thus enhancing basal melt rates. The strong diurnal tidal currents at the shelf break facilitate the cross-shelf exchange of water masses and contribute to mixing, hence influencing the hydrographic properties of both the cold outflow and warm inflow.

The mooring data used in this study are available under 10.1594/PANGAEA.870436, 10.1594/PANGAEA.792882, 10.1594/PANGAEA.792883, 10.1594/PANGAEA.792884, 10.1594/PANGAEA.792885, 10.1594/PANGAEA.869799, 10.1594/PANGAEA.869820, 10.1594/PANGAEA.869813, 10.1594/PANGAEA.869804, 10.1594/PANGAEA.870449, 10.1594/PANGAEA.870518, 10.1594/PANGAEA.869905 and 10.1594/PANGAEA.871146. The CTD profile at mooring M3 is available under PANGAEA.854148. The numerical model described in is available at http://www.whoi.edu/cms/files/Matlab_Code_30467.htm. The seal data were collected and made freely available by the International MEOP Consortium and the national programs that contribute to it (http://www.meop.net).