The North Atlantic Ocean and northwest European shelf
experience intense low-pressure systems during the winter months. The effect
of strong winds on shelf circulation and water properties is poorly
understood as observations during these episodes are rare, and key flow
pathways have been poorly resolved by models up to now. We compare the
behaviour of a cross-shelf current in a quiescent period in late summer,
with the same current sampled during a stormy period in midwinter, using
drogued drifters. Concurrently, high-resolution time series of current speed
and salinity from a coastal mooring are analysed. A Lagrangian analysis of
modelled particle tracks is used to supplement the observations. Current
speeds at 70 m during the summer transit are 10–20 cm s
The shelf seas and abyssal ocean are often treated as separate systems as they exhibit very different behaviours despite the lack of a physical barrier between the two. The shallow continental shelves are responsible for roughly 25 % of global primary production (Simpson and Sharples, 2012) and play host to the vast majority of human–ocean interactions. The exchange of water between the ocean and shelf seas is still an evolving field of study, as is our understanding of how oceanic and shelf-edge processes play out at the coast (Brooks and Townsend, 1989; Holt et al., 2009; Huthnance et al., 2009; Münchow and Garvine, 1993). In this paper we investigate the behaviour of a newly characterised across-shelf current, the Atlantic Inflow Current (AIC; Porter et al., 2018), and the role it plays in transporting oceanic water across the northwest European shelf.
The northwest European shelf is bounded to the west by the northeastern
Atlantic and Rockall Trough basins (Fig. 1a). A topographically steered
slope current flows along the shelf edge between Biscay and Norway, becoming
increasingly consistent in flow speed and direction north of the Celtic Sea.
The stability and persistence of the slope current, particularly north of
55
While schematics of shelf circulation typically infer a permanent residency of oceanic water on regions of the shelf (Ellett and Edwards, 1983; Ellett and MacDougal, 1985; Inall et al., 2009; Simpson et al., 1979), in reality these intrusions appear to be sporadic both in duration and geographic extent. Evidence for the spatial mobility of fronts and water masses on the Malin Shelf is provided by satellite observations, as well as in situ cruise and mooring data (Ellett and Edwards, 1983; Inall et al., 2009; Jones et al., 2018; Jones, 2016; Porter et al., 2018). The variable occupation of the Malin Shelf by oceanic water means that coastal water properties in some regions exhibit high temporal variability (Jones et al., 2018). We seek to characterise the causes and nature of these oceanic intrusions.
Our investigation builds on findings from a drifter release on the Malin
Shelf during the FASTNEt shelf-edge observation campaign (Porter et al.,
2018, Fig. 1b). Drogued drifters released into the slope current in July
2013 moved on-shelf downstream of a canyon system at 55.5
Salinity at the Tiree Passage Mooring at 20 m. Grey bars denote
winter months (DJFM). Green lines show instances of daily mean westerly
winds on Malin Shelf exceeding 18 m s
We also utilise salinity and current observations from a fixed mooring in
Tiree Passage off the west coast of Scotland. The Tiree Passage Mooring
(TPM) tracks the highly variable mix of coastal water, freshwater runoff, and
oceanic water that flows through the Inner Hebrides, collectively referred
to as the Scottish Coastal Current (SCC). The SCC originates in the
baroclinically driven outflow from the Irish Sea (Hill, 1987; Hill and
Simpson, 1988; Jones, 2016) and receives contributions from rivers and
sea lochs, causing it to become less saline and increase in volume as it
progresses northward. Residual currents at the TPM are typically poleward
with an average speed of 10 cm s
Once or twice during most winters a brief pulse of very high-salinity water is observed at the TPM, and these high-salinity pulses (HSPs) are often associated with storm events (Jones et al., 2018). An HSP constitutes a brief but significant change from usual circulation patterns because nearly undiluted oceanic water is observed at a coastline normally buffered from the Atlantic by the SCC. HSPs may be an important mediator of winter coastal water properties as oceanic water is a source of both heat and nutrients to the shelf seas (Painter et al., 2016; Porter et al., 2018; Siemering et al., 2016). They may have other impacts at the coast, for instance the import of organisms typically restricted to waters further offshore.
Understanding of the drivers of these events is scant: while HSPs are associated with storm activity, salinity at the TPM does not correlate simply with wind forcing on the shelf (Jones et al., 2018). Also, it is not known whether accepted transport pathways from the outer shelf hold true when wind and wave action mask the weaker baroclinic flows which prevail during quiescent periods. Observations of HSPs are few as they typically occur in midwinter when most satellite sensors are obscured by cloud. In addition, they are associated with stormy periods in a region notorious for rough seas, so they are generally not captured by oceanographic cruises which cannot sample in such conditions.
This study capitalises on the fortuitous recirculation of two drifters out of the 30 released in July 2013 during the FASTNEt programme (Porter et al., 2018). While most drifters had exited the Malin Shelf by October 2013, two drifters drogued at 70 m were captured by an eddy in the Rockall Trough shortly after release and only crossed onto the Malin Shelf in December 2013. This transit coincided with an HSP being measured by the TPM, so it has the potential to explain the origin and nature of these phenomena. We therefore examine the Lagrangian properties of the December drifter tracks in comparison with the fixed time series at the TPM. In addition, we compare the shelf conditions during this winter shelf transit with the late summer conditions sampled by the first cluster of drifters. This suite of observations is complemented by a modelled particle-tracking experiment.
On 17 July 2013, 30 satellite-tracked GPS drifters were released from the
RRS
The drifters were configured to report GPS positions every 3 h. To exclude high-frequency motions such as tides and inertial oscillations from the drifter tracks, the data were filtered with a 10th-order zero-phase Butterworth low-pass filter with a cut-off at 2 cpd (cpd: cycles per day). There were numerous instances of a single missing GPS position, but longer gaps were rare, with a maximum gap size of 20 h. In all cases, positional data gaps were linearly interpolated. Drifter velocities were calculated using displacements from the filtered location data.
The TPM is situated in northern Tiree Passage at 56.6
Daily 10 m wind and sea level pressure data from the ERA-Interim
This study uses model output from AMM15 (Atlantic Margin Model, 1.5 km resolution), which was developed and validated by Graham et al. (2018a, b) through the UK Joint Weather and Climate Research Programme. This incarnation of the model builds on AMM7 (7 km resolution), which has been utilised and validated by numerous studies (O'Dea et al., 2017, 2012). Both models are based on NEMO (Nucleus for European Modelling of the Ocean) architecture.
AMM15 bathymetry is derived from EMODnet (EMODnet Portal, September 2015
release), and it uses a
To test the implication of the observed drifter trajectories, we conducted a series of offline particle-tracking experiments using AMM15-modelled velocities, with daily mean full-depth U/V velocities obtained from a hindcast for the period of interest. From this dataset, horizontal velocities at 20 and 70 m depths were extracted to coincide with the TPM and drifter observation depths. We contrasted two periods sampled by the drifters: 1 to 11 August and 15 to 25 December 2013. In each case, the 10 d interval was deemed an “observation period”. Particles were released at daily intervals across the local model domain for the 40 d preceding the observation period and during the 10 d observation period (Fig. 3). Particles which were advected into the observation polygons during the observation period were identified and their release location noted. The experiment was repeated five times to sample a range of diffusive random walks, resulting in a total of 250 unique particles being released from each location in Fig. 3a. Our analyses focus on the origin of the particles which reached the observation polygons on the inner shelf.
The particles were tracked using a 2D Lagrangian scheme. The 2D location of
a particle
In this study only the on-shelf portions of drifter tracks (depth < 200 m) are investigated; for a detailed analysis of shelf-edge dynamics from
the drifter observations, see Porter et al. (2018). In addition, we use only
tracks from drifters drogued at 70 m as those drogued at 15 m all moved
on-shelf in August 2013, thus precluding an autumn–winter comparison. The
drifters were initially released into the slope current but diverged at the
canyon system at 55.5
Trajectories of deep (drogued at 70 m) drifters, coloured by
velocity and separated into drifters which travelled on-shelf in
Cross-shelf drifter progress in the complex Malin Shelf region is dependent
on local bathymetry and the location at which each drifter crossed the shelf edge.
However, drifters from both groups were advected eastwards in the AIC and
passed through the same region north of Ireland, so the time taken to reach
this point from the shelf edge provides a metric of cross-shelf progress
(Fig. 4). Beyond this point the tracks turn northward and scatter, and
inter-comparison of drifters is once again problematic. Consequently, we
chose a meridional line at 8
Seven drifters travelled between the shelf edge and 8
Flow at 70 m was generally on-shelf but with numerous sub-tidal meanders and
reversals. Drifters crossing the shelf edge between 55.4 and
56.1
To appraise the meteorological and oceanographic conditions on the shelf
during the experiment, we compared the periods of drifter shelf transit with
time series of reanalysis wind data on the Malin Shelf, the 20 m current
speed at the TPM, and the 20 m salinity at the TPM (Fig. 5). Note that the
TPM was undergoing maintenance during the early drifter experiment so TPM
current speed and salinity data commence in late August 2013. The
cross-shelf transit time of drifters in group 1 (August 2013) is between 18
and 42 d. Wind forcing during this period is typical for the time of year
at this location, with numerous short episodes of SW–NW winds of 10–15 m s
The combination of drifter observations and the TPM salinity observations provides contrasting snapshots of cross-shelf flow during intense storms and a more quiescent period. However, the observations are in some ways not comparable: for example, when analysing the storm event, we compare the salinity at the TPM at 20 m of depth and the behaviour of drifters drogued at 70 m of depth which could not pass through Tiree Passage due to its shallow bathymetry. To present a more compelling picture of the periods under investigation, we supplement these observations with a series of particle-tracking experiments. Specifically, we seek to answer the following question: did the episode of stormy weather in December 2013 significantly alter the origins of water reaching the Scottish west coast throughout the water column? To address this question, we performed a series of particle-tracking experiments, focussing on particles terminating at the inner Malin Shelf. The particles were released at two depths: 20 m (the nominal depth of TPM observations) and 70 m (the drogue depth of the deep drifters) at the locations shown in Fig. 3a.
Particle distribution maps showing the origin of particles
advected through Tiree Passage and the inner Malin Shelf. The green polygon
shows the observation region for the 20 m particles in
Figure 6 compares the tracks of particles terminating at the inner Malin
Shelf during August 2013 (the first episode of on-shelf drifter advection)
and during December 2013 when there was rapid on-shelf advection of two
further drifters. Each cell is coloured by the percentage of particles
released at that location which were subsequently advected through the
observation region. In August, most particles at both 20 and 70 m
originated within 50 km of the observation polygon, with a few cells west of
8
Particle distribution maps showing the origin of particles
advected through Tiree Passage and the inner Malin Shelf. The green polygon
shows the observation region for the 20 m particles in
A measure of typical particle advection times can be obtained by colouring
particle origin cells by the average time their particles took to arrive at
the observation polygons (Fig. 7). Note that Figs. 6 and 7 were produced
using separate (but identical) batches of particle releases for
computational reasons, with each batch consisting of five repeat experiments.
Thus, the figures exhibit small differences in particle distribution due to
the diffusive component of the particle motion. Again, the contrast between
the 20 m particles in August and those in December highlights differences
in the origins of the water. The source region of the December particles
includes a broad section of the shelf edge, whereas in August the waters are
sourced exclusively from the shelf. In addition, the December distributions
indicate a relatively rapid pathway between the shelf edge at 55
In this study we examined a period of intense shelf sea flows driven by a cluster of winter storms. The salinity and current speeds measured at the TPM during this event were amongst the highest measured by the mooring during its multi-year occupation. The spike in salinity at the TPM, coupled with the trajectories of two drogued drifters, confirmed that the origin of the water passing through Tiree Passage switched to the outer shelf during this period. In Sect. 4.1 we discuss the rapid (1–2 d) increase in currents due to geostrophic flows driven by wind-induced pressure gradients on the northwest European shelf. This precedes the increase in salinity at the TPM by 8 d, which we interpret as the time taken to advect high-salinity water from a remote location in the enhanced shelf currents. The nature of the high-salinity intrusion is discussed in Sect. 4.2. We then consider the additional insight provided by the particle tracking in Sect. 4.3 and the weather conditions associated with HSPs in Sect. 4.4. In Sect. 4.5 we investigate the likelihood of an HSP occurring in each winter.
There is strong evidence in the literature that wind-induced pressure
gradients can quickly set up or enhance currents on the Malin Shelf and that
their pathways are influenced by the wind direction (Davies and Xing, 2003;
Inall et al., 2009; Jones, 2016; Xing and Davies, 2001). Similarly, the flow
through the North Channel of the Irish Sea onto the Malin Shelf is highly
correlated with wind aligned with the channel with a lag time of a few hours
(Bowden and Hughes, 1961; Brown and Gmitrowicz, 1995). We may expect a rapid
setup of inner shelf currents in response to a wind-induced surface pressure
gradient as the barotropic adjustment time will be limited only by the speed
of a long wave in shallow water. This speed is set by
In addition to shelf-scale pressure gradients, winds with a westerly component will set up a local Ekman drift towards Northern Ireland. We would expect this flow to result in an elevated sea surface height and depressed isopycnals towards the northern Irish coast and in turn develop a geostrophic jet flowing eastwards along the northern Irish coast. Porter et al. (2018) noted depressed isopycnals associated with the core of the AIC during July 2013, so it is not unreasonable to surmise that Ekman effects act to enhance or complement the AIC.
If we consider the AIC and Tiree Passage to be subject to the same forcing influences, one might expect a link to exist between the speed of the drifters tracking the AIC and the poleward current speed concurrently measured at the TPM. However, we find little coherence between these measures (not shown), though both the drifters and the TPM do show increased current speed in December compared with August. We surmise that the instantaneous sub-tidal speed of the drifters is a complex aggregate of wind-driven currents and local bathymetric flow intensification, so their speed does not correlate simply with that measured at the fixed mooring.
As a tracer of oceanic-origin water, the transport of salt is limited by the current speed on the shelf, and the 8 d lag between the increase in current speed and the onset of high salinity at the TPM points to the advection of the high-salinity water from a remote source. This lag period is in accord with the time taken by the December drifters to reach the inner shelf from the shelf edge (between 6 and 10 d).
Integrating the TPM poleward current flow over the interval between the
onset of enhanced flow (6 December 2013, event B) and the arrival of
high-salinity water at the TPM (event C, 14 December 2013) gives a
total displacement of 145 km. If we assume this water arrived in Tiree
Passage via the AIC, the minimum distance from the shelf edge to the TPM via
this route is 210 km, and the December drifters in fact crossed onto the
shelf further south (54.75
Unlike salinity, the temperature measured at the TPM (not shown) does not
exhibit any notable deviation from the expected seasonal cooling at the time
of the HSP investigated here. This is because there is very little
difference between oceanic (near-surface) and coastal water temperatures on
the Malin Shelf in December (Ellett and Edwards, 1983; Inall et al., 2009;
Jones et al., 2018). A change in water origins from coastal to oceanic,
however abrupt, would therefore have little impact on water temperatures at
this time. However, between January and March coastal waters are cooler (6–8
Downstream of the TPM, coastal waters typically flow towards northwards
towards the North Sea (McKay et al., 1986; McKinley et al., 1981). The Minch
region between the islands of Skye and the Outer Hebrides represents a
partial barrier to flow, so during the high transports associated with an HSP
much of the imported water is likely to pass around the outside of the Outer
Hebrides on the outer continental shelf. North of 58
Average horizontal current velocity in the AMM15 model for
The model particle releases demonstrate a striking contrast between shelf
behaviour in August 2013 and that during the storm event in December 2013.
As might be expected from the evidence thus far, current speeds were slower
in August and particles in the observation polygons originated more locally
over the 50 d tracking period. This was true for both particles in the
seasonal thermocline (20 m) and below it (70 m). By contrast, most particles
tracked during the storm event originated between the mid-shelf and the
shelf edge. The long transit time of the August particles would be
associated with greater mixing between oceanic and shelf waters before
reaching the coast. We would therefore expect that the high salinity
associated with the inflow would be more dilute than during December. The
preferential across-shelf pathway taken by most particles in Fig. 6d
closely matches the route taken by the two drifters in Fig. 4b. This
supports the notion of the AIC as a narrow, jet-like current, which was
proposed by Porter et al. (2018). Further clarity on the modelled flow is
obtained by comparing the average modelled current speed during the two
10 d observation periods (Fig. 8). The key difference between the August
and December observation periods is the rapid across-shelf flow in December,
which is not present during August. This disparity illustrates why most
particle recruitment in August was restricted to the inner shelf: there was
no pathway to enable transport from the shelf edge. In December, on-shelf
flow exceeded 0.3 m s
Due to the design of the release and observation periods we might expect about one in five local releases to result in an observation in the polygons given steady flow. This is because particles were released over a period of 50 d but were only tallied in the observation polygons for the final 10 d of the experiment (Fig. 3b), so the majority of local particles would pass through the observation polygon without being counted. However, in the December experiments we see high concentrations of particles originating remotely. This is a result of flow speeds increasing during the simulation, resulting in an effective convergence of particles during the observation period (days 40–50).
Sea level pressure maps for the three storm events associated with the
winter 2013–2014 high-salinity pulse (HSP). These storms were named Xaver,
Bernd, and Dirk, respectively. Sea level pressure data derived from the ECMWF
ERA-Interim
Mode 1 EOF winter (DJFM) gale days for North Atlantic. Based on
ERA-Interim daily averaged 10 m wind speed and the WMO gale definition of wind
speed exceeding 17.2 ms
At the shelf edge, the insulating effect of the steep bathymetry maintains a
strong control on ocean–shelf interaction for the model particles. There are
almost no examples of 70 m particles crossing the shelf edge, and only a
small proportion of 20 m particles crossed from ocean to shelf. Graham et
al. (2018b) reported an on-shelf volume flux of roughly
Graham et al. (2018b) found that cross-shelf fluxes in the surface layer (0–20 m) were larger during winter (January) than in the summer (July). The enhanced cross-slope flow during winter is likely to be due to increased wind forcing driving a downwelling circulation on the shelf, as described by Holt et al. (2009). The storm event depicted in the present study can be regarded as a snapshot of the processes contributing to the long-term winter average.
Time series (eigenvalues) of EOF mode 1 eigenvector pattern
illustrated in Fig. 10 (blue line), defined as the North Atlantic gale
index. NAO index after Hurrell (1995). The winter is defined as the year
containing JFM; i.e. for a winter December 2013–March 2014, the year is
2014. The two indices correlate with
Given the regular passage of low-pressure systems across the UK during most
winters, it is perhaps surprising that only one to two HSPs per year are recorded
by the TPM. This finding indicates that further prerequisites are required
to transport oceanic water to this inner shelf location. There is much
evidence that the Malin Shelf is subject to an underlying baroclinic flow
originating in the Irish Sea from tracers, oceanographic observations, and
rotating tank experiments (Ellett, 1979; Ellett and Edwards, 1983; Hill,
1987; Jones et al., 2018; McKay et al., 1986; McKay and Baxter, 1985). Jones
et al. (2018) conjectured that the relative infrequency of HSPs was because
the inner shelf had a tendency to return to this baroclinic state in the
absence of wind forcing. They argued that the displacement of water driven
by a single low-pressure system may not be sufficient to advect oceanic
water across the “buffer” of coastal waters which typically occupy the inner
shelf. If we use, for example, the approximate linear relationship between
wind and current speed in shallow seas suggested by Whitney and Garvine (2005),
Strong on-shelf flow towards the Scottish west coast only occurs during southerly to westerly winds, as other wind directions either drive Irish Sea water onto the shelf or retard shelf flow more generally (Davies and Xing, 2003; Jones et al., 2018). To illustrate the relevance of this note to the present study, sea level pressure maps of the major wind events associated with the December 2013 HSP are shown in Fig. 9. In each case an area of intense low pressure passed to the north of the UK, resulting in a westerly airflow over the Malin Shelf. The path of low-pressure systems over the northeastern Atlantic is captured by the North Atlantic Oscillation (NAO) index (Hurrell, 1995), which compares the atmospheric pressure over Iceland and the Azores. Positive NAO winters such as that of 2013–14 feature a northerly storm track, whereas during negative NAO years low-pressure systems tend to pass south of the UK. Consequently, conditions likely to result in an HSP episode are more likely to occur during positive NAO years. In the following section we explore the relationship between the NAO and storminess over the northeastern Atlantic.
We now analyse wintertime storminess over the North Atlantic by using daily
10 m wind speed data from the ERA-Interim
Following Qian and Saunders (2003) we define a set of wind speed indices as
the numbers of days in each winter (DJFM) on which wind speeds exceed different
force levels on the Beaufort wind scale and World Meteorological Organization
(WMO) wind speed classification. These levels are near gale force (Beaufort
scale 7, 13.9–17.1 ms
Analysing the period 1979 to 2015, we perform an EOF decomposition of “gale
force” winds over the whole North Atlantic (i.e. 17.2 ms
The mode 1 EOF pattern (Fig. 10) shows the storm track with a WSW to ENE
orientation, as is typically associated with NAO-high winters. The most
energetic landfall of this pattern is between 56 and
58
Between 1989 and 1994, Fig. 14 of Inall et al. (2009) shows a prolonged
period of more saline shelf waters at 10 m of depth from Ellett Line CTD
stations. This period corresponds to a high gale index period in the EOF
mode 1 eigenvalue time series (Fig. 11). Further, the relative dip in the
gale index in winter 1991 apparently coincides with a modest westward
relaxation of isohaline contours, with fresher waters reappearing east of
6.5
In this paper we have characterised an oceanic inflow onto the northwest European shelf during intense storm activity using drifters, a moored time series, and modelled particle tracking.
We found that during the storm event, across-shelf flow increased from 10–20
to 60 cm s
The spike in coastal salinity associated with storm activity was one of
several in the 13-year mooring time series. To assess the likelihood of
these sporadic storm episodes occurring during a given winter, we
constructed a “gale index”: a measure of the number of days wind speeds
exceeded predefined thresholds across the North Atlantic. Despite high
correlation between the gale index and the NAO index (
AMM15 model data are archived on the Met Office mass storage system and can
be accessed through the STFC-CEDA platform JASMIN (Lawrence et al., 2013).
The NEMO code and the parameterisations used are outlined in Graham et
al. (2018a). The drogued drifter data and Tiree Passage Mooring data are banked at the British Oceanographic Data Centre. Dataset information is available at
SJ prepared the paper with contributions from all co-authors. MI secured the funding for the work, developed the WMO gale index (Sec. 4.5), and made significant contributions to the particle-tracking experiment design. MP designed and executed the drogued drifter release experiment and helped to shape the present study. JG was a lead contributor to the design and testing of the AMM15 model and was instrumental in ensuring that the particle tracker was appropriately tuned for model diffusivity. FC helped to secure the funding for the work, provided mentorship in the form of PhD supervision to SJ, and identified the relation between drifter behaviour and HSPs in the mooring salinity record.
The authors declare that they have no conflict of interest.
This study was funded by the UK Natural Environment Research Council (NERC) projects Fluxes Across Sloping Topography of the North East Atlantic (FASTNEt) (NE/I030151/1) and Overturning in the Subpolar North Atlantic Program (OSNAP) (NE/K010700/1). Marie Porter was supported by the AtlantOS project (European Union Horizon 2020 research and innovation programme, grant 633211). The authors would like to thank colleagues at the Met Office for their assistance in providing access to archived model output.
This research has been supported by the UK Natural Environment Research Council (NERC) (grant nos. NE/I030151/1 and NE/K010700/1).
This paper was edited by Erik van Sebille and reviewed by Robert Marsh and one anonymous referee.