Decadal variability on the Northwest European continental shelf

10 Decadal scale time series of the shelf seas are important for understanding both climate and 11 process studies. Despite numerous investigations of long-term temperature variability in the shelf 12 seas, studies of salinity variability are few. Salt is a more conservative tracer than temperature in 13 shallow seas, and it can reveal changes in local hydrographic conditions as well as transmitted 14 basin-scale changes. Here, new inter-annual salinity time series on the northwest European shelf 15 are developed and a 13 year high resolution salinity record from a coastal mooring in western 16 Scotland is presented and analysed. We find strong temporal variability in coastal salinity on 17 timescales ranging from tidal to inter-annual, with the magnitude of variability greatest during 18 winter months. There is little seasonality and no significant decadal trend in the coastal time 19 series of salinity. We propose 4 hydrographic states to explain salinity variance in the shelf area 20 west of Scotland based on the interaction between a baroclinic coastal current and wind-forced 21 barotropic flow: while wind forcing is important, we find that changes in the buoyancy-driven 22 flow are more likely to influence long-term salinity observations. We calculate that during 23 prevailing westerly wind conditions, surface waters in the Sea of the Hebrides receive a mix of 62 24 % Atlantic origin water to 38 % coastal sources. This contrasts with easterly wind conditions, 25 during which the mix is 6 % Atlantic to 94 % coastal sources on average. This ‘switching’ between 26 hydrographic states is expected to impact nutrient transport and therefore modify the level of 27 primary productivity on the shelf. This strong local variability in salinity is roughly an order of 28 magnitude greater than changes in the adjacent ocean basin, and we infer from this that Scottish 29 coastal waters are likely to be resilient to decadal changes in ocean climate. 30

4 tracer of shelf behaviour because it is more conservative than temperature, it is easy to distinguish 105 coastal from oceanic origins and its use in observational campaigns has been routine for decades, 106 enabling long time series to be constructed. The sensitivity of salt to advection in particular can 107 make it a more critical measure of model skill in semi-confined basins than temperature, e.g. Young 108 and Holt (2007). This work complements the analyses of coastal current physics by investigating 109 their implications over inter-annual time scales. 110 111 1.1 Regional setting 112 The shallow waters west of Scotland are subject to both coastal and oceanic influences (Burrows et 113 al., 1999;Hill et al., 1997;Inall et al., 2009;McKay et al., 1986). The region is of a greater complexity 114 than many shelf sea areas, as it features numerous influences on the physical oceanography 115 combined with a convoluted coastline. The offshore Atlantic water displays well documented inter-116 annual variability in temperature, salinity and nutrient content (Holliday, 2003a Simpson and Hill, 1986). Further offshore, the island chain of the 128 Outer Hebrides forms a partial barrier between the north-west Scottish coastline and the North 129 Atlantic. The resultant tapering south-north oriented channel is known as the Minch.   Southampton. Of primary interest to this study are the 17 CTD stations regularly occupied on or 262 near the Malin Shelf. Significant temporal gaps exist in these records, so to achieve a seasonal data 263 resolution across the shelf, historical data from the International Council for the Exploration of the 264 Sea (ICES) were used to supplement the core Ellett Line dataset (see Appendix C for more details). 265 2.3 ICES CTD and surface data 266 ICES maintains a large database of historical CTD and bottle data which has been assembled from 267 numerous sources and subjected to basic quality control (Berx and Hughes, 2009 Daily 10 m wind data were acquired from two reanalysis models maintained by ECMWF 281 (www.ecmwf.int). The ERA-interim model covers the period between 1979 and present, however as 282 this study required data from 1975 onwards, the output of the ERA-20C model was used to provide 283 data for this interval. A comparison of the two models over a common time-frame showed virtually 284 no discrepancy between the two, so the transition between models was deemed to be valid. Data 285 were downloaded on a 1 ° x 1 ° grid, from which a time series on the central Malin Shelf (56 ° N, 7 ° 286 W) was extracted. Wind stress was calculated from the velocity data using: 287 Where is the density of air in kg m -3 and the 10 m wind speed in m s -1 at a height of 10 m. 289 is the coefficient of drag, given by the approximation suggested by Large and Pond (1981): 290 = (0.63 + 0.066 ) × 10 −3 (2) 291 11 The use of ECMWF reanalysis wind data on the European Shelf is commonplace (Holt and Proctor,292 2003; Huthnance et al., 2009;Pingree et al., 1999) and its accuracy on inter-annual time-scales is 293 considered to be sufficient for the statistical analyses presented here. 294 2.6 The Sub-Polar Gyre (SPG) index

295
The index of SPG strengths used in this study was constructed from altimetry observations (Häkkinen 296 and Rhines, 2004)         While temperature has a clear seasonal cycle (Figure 5c), salinity does not ( Figure 5b). However, the 377 highest and lowest salinities typically occur in winter, and consequently the standard deviation of 378 the Tiree Passage time series varies seasonally, rising from 0.14 in August to 0.30 in January. 379 where is the input from the North Channel, + is the on-shelf component of cross-shelf flow, 458 is the contribution from runoff, and from precipitation. The outflow is: 459 Where ℎ is the flow past the Hebrides towards the North Sea, − the off-shelf component of 461 cross-shelf flow, and the evaporation. As , and are all less than 10 3 m 3 s -1 (Jones, 2016) 462 they are discounted for this approximation leaving: 463 commonplace, though with a 12 year record no trend is apparent (Figure 2). The 40 year TPC time 518 series shows that variability during the winter months (DJFM) dominates the inter-annual salinity 519 signal, whereas during the summer (JJAS) the salinity remains relatively stable at 34-34.5 (Figure 4c). 520 The winter months often feature very high or very low salinities, but these extremes largely cancel 521 out between years, so on average there is little seasonality in Tiree Passage salinity.  (Figure 7a) is an indicator that changes in Atlantic salinity are unlikely to be 531 detectable at the Scottish coastline. This de-correlation may be exacerbated because the Ellett Line 532 transect bisects the SCC as it recirculates around the southern tip of the Outer Hebrides, which may 533 limit the eastward influence of Atlantic water more at this latitude than if the experiment was 534 repeated further south. However it is probable that oceanic signals are largely masked by energetic 535 shelf processes, which will be examined in more detail in Section 6.1. Malin Shelf SS is present (Figure 7c). Whilst it might be speculated that this lack of shared variance is 539 because the SCC has its origins in the eastern, rather than central Irish Sea (where Port Erin is 540 located), we suspect that this is not the case. Allen et al., (1998) found Port Erin to be 541 representative of variability in the wider Irish Sea area, particularly in winter. In addition, Jones 542 We propose that the complex hydrography of the Malin Shelf can be largely characterised by four 563 states, and that the strong variability observed on the inner shelf can be explained in part by a 564 switching between these states. 565

State1: Buoyancy only; thermally stratified 566
If it is accepted that the underlying driver of the SCC is the salinity differential between the Irish Sea 567 and the Malin Shelf (Hill et al., 2008) this infers a permanent mechanism for buoyancy forcing for the 568 SCC regardless of season. This process occurs because of the tendency for fresher Irish Sea water to 569 spread out on top of denser Atlantic water as they meet on the southern Malin shelf. However due 570 to the Earth's rotation the lighter Irish Sea water is deflected to the right of its direction of travel, 571 and the system can therefore reach equilibrium with Irish Sea water forming a wedge against the 572 west Scottish coastline and flowing with the land on its right (a ROFI system, Simpson (1997)). The 573 theoretical structure is shown in Figure 8d, and the path of the buoyant plume, which becomes the 574 SCC (Hill, 1983;Hill et al., 1997;McKay and Baxter, 1985;McKay et al., 1986;Simpson and Hill, 1986) 575 is shown in Figure 8e. In its buoyancy-driven state, the SCC has many similarities with the coastal 576 current systems highlighted in other regions. Its typical flow rate of 10 4 -10 5 m 3 s -1 is somewhat 577 larger than many river plume systems (Münchow and Garvine, 1993b;Souza and Simpson, 1997;578 Wiseman et al., 1997) but is substantially smaller scale than, for instance the large coastal currents 579 off Greenland and Alaska (Bacon et al., 2002;Whitney and Garvine, 2005).  periods. The Ellett Line transect in Figure 9 shows the Malin shelf during a brief calm period 608 following a month of strong winds, predominately from the south. The shelf is close to fully mixed 609 and temperature is uniform with depth. Vertical density gradients are low, as evidenced by the 610 potential energy anomaly ϕ in Figure 9a, however horizontal density gradients are moderate. 611 Salinity in the eastern Minch is relatively high at 34.8, but still significantly lower than the oceanic 612 water to the west (35.4). Note that an Ellett Line section with a more pronounced haline 613 stratification (an example of the ROFI structure) is shown in Figure 10  While baroclinicity may be the underlying driver of flow on the Malin shelf, we argue that flow due 624 to wind stress can easily override this mechanism. Here, a simple order of magnitude comparison 625 between wind and buoyancy suggested by (Hill, 1983) is informative. 626 To estimate the poleward flow due to density differential on the Malin Shelf, a thermal wind 627 balance is assumed so that vertical density gradients are neglected: 628 Where is acceleration due to gravity, density, the change in density per unit distance offshore 630 and the Coriolis parameter. Substituting values of = 1027 kg m 3 and = 1 kg m -3 per 40 km, the 631 density-driven transport is estimated to be 5 m 2 s -1 . 632 26 A contrasting approximation is used to estimate wind-driven flow. In this scenario the sea surface 633 moves with a wind-driven friction velocity * in a shallow sea of depth H, with linearised bottom 634 friction = ( / ) 2 in which is the drag coefficient and the resultant net northward 635 transport. In this case the balance between transport and bottom friction is 636 Using a value of * = 1 cm s -1 (roughly corresponding to a 10 m wind speed of 10 m s -1 , (Csanady, 638 1981)) = 70 m and = 10 -3 , the wind-driven transport is 22 m 2 s -1 . Thus even moderate wind 639 forcing would be expected to temporarily replace buoyancy as the dominant driver of flow on the 640 shelf. In the above approximations, the wind speed required to balance the density-driven transport 641 is roughly 7 m s -1 ; a value which is regularly exceeded throughout the winter months in the NE 642 Atlantic. To gauge the short-lived but substantial influence of winter storms, we can use the 643 approximate linear relationship between wind and current speed in shallow seas suggested by 644 Whitney and Garvine (2005) Malin shelf, and that it may reside on the shelf until it is diluted or advected out of the region. Thus 666 the supply of the majority of freshwater onto Malin Shelf (≈ 90 %, Barnes and Goodley (1958), 667 Edwards and Sharples (1986)) would be dependent on local wind patterns. Figure 10  Westerly winds are often associated with stormy conditions in the NE Atlantic, and tend to be 682 strongest during the winter. The model studies of Davies and Xing (2003) and Xing and Davies 683 (2001b) indicate that westerly winds result in an intensified inflow from the outer shelf along the 684 north coast of Ireland and a blocking surface slope which limits or prevents the buoyancy-driven 685 outflow from the Irish Sea. This would have the effect of increasing salinity at the inner stations by 686 replacing the SCC with water from the outer shelf. Figure 11 shows To achieve this, the ESS observation dates were first divided into summer (April-September, state 1) 732 and winter (October-March, state 2) to broadly characterise the thermally stratified and thermally 733 mixed seasons. To define observations made after a period of unusual wind stress, a 60 day flushing 734 time was assumed as in previous sections. Note that as inferred by Figure 12, the shelf may be 735 flushed more rapidly during storm events but a more advanced modelling study would be required 736 to estimate how this may affect advection of salt and freshwater. Integrating over a reduced period 737 for westerly winds does not significantly improve correlations, and the possible reasons for this are 738 subsequently discussed. 739 The standard deviation of 60 day cumulative easterly and westerly stress on Malin shelf was 740 calculated for the full duration of the ECMWF record (1940-present). These were found to be 2.16 N 741 m -2 and 5.00 N m -2 respectively. Observations in the ESS time series which were preceded by 742 easterly or westerly winds exceeding 1.5 standard deviations from the mean were defined as state 3 743 or state 4 respectively. For easterly winds, this additive stress is approximately equivalent to 10 days 744 during a 2 month period with average wind speeds of more than 12 m s -1 , or 20 days exceeding 9 m 745 s -1 . For westerly winds, the threshold value is equivalent to 10 days which exceed 17 m s -1 (a 746 statistically unlikely scenario) or 20 days greater than 13 m s -1 . Where these criteria are fulfilled, 747 they override states 1 and 2. In the event that the criteria for both states 3 and 4 are met, state 4 748 takes priority due to the greater energy associated with westerly wind stress. 749 The time-mean of ESS observations assigned to each state is shown in Figure 13

757
The mean potential energy anomaly ϕ of each station grouped by state is shown in Figure 14. Note 758 that these averages are comprised of fewer observations than SS as ϕ is an integral value 759 necessitating a full-depth CTD measurement. States 1 and 2 (stratified and mixed, Figure 14a) 760 exhibit a similar cross-shelf pattern which is due to the regions of shallow and deep bathymetry 761 bisected by the Ellett Line. States 3 and 4 (wind-driven) are shown separately due to the much lower 762 values of ϕ in the absence thermal stratification (Figure 14b). Easterly wind stress (state 3) is 763 associated with higher stratification in the eastern SoH and west of the Hebrides, which is consistent 764 32 with the presence of a ROFI-type structure adjacent to the land. Westerly wind stress is associated 765 with greater stratification in the Sound of Mull (Station 1G) which, in accord with other measures in 766 this study, we take as evidence of local freshwater runoff residing at the surface, as the passage of 767 Atlantic storms also results in higher precipitation. The higher stratification and lower SS east of the 768 Hebrides (Station 10G) is noteworthy as to reach this station, freshwater must be advected 769 southwards along the eastern coast of the Hebrides (see inset in Figure 1). This suggests that a more 770 localised coastal current containing (though not necessarily driven by) freshwater runoff is active 771 during stormy periods when the SCC (in its buoyancy-driven form) is largely inactive.

777
We can gain some insight on the implications of these states for discrete observations by examining 778 time series of SS and ϕ at Station 7G ( Figure 15). This station exhibits some of the highest variance 779 on the shelf, and is better sampled than other stations in the SoH (Figure 6a shows ESS sample 780 frequency). 781 In this location both summer and winter observations feature high variability (Figure 15a), though on 782 average SS is slightly higher in the winter (e.g. Figure 12). Easterly wind stress has the clearest effect 783 on SS, as all instances are associated with very low salinities. While the availability of full-depth CTD 784 profiles is limited, 3 of the 4 observations made following easterly wind stress show above average 785 levels of stratification for winter (Figure 15b). There is not a clear relationship between SS and 786 westerly wind stress at this location; while some of the high SS observations in the 1990s were 787 preceded by westerly winds, many high SS events are not linked to westerly wind stress. However

794
There is substantial evidence in the literature that wind stress enhances residual flow on Malin Shelf, 795 and that westerly winds produce favourable conditions for the import of oceanic water to the 796 coastline Hill and Simpson, 1988;Inall et al., 2009;Xing and Davies, 1996). It is 797 therefore surprising that this pattern of wind stress does not show a strong relationship with SS 798 observations. We suggest several factors which may contribute to this finding. The pulses of high 799 salinity recorded by the TPM (and occasionally by the Ellett Line) are short-lived and are often 800 followed by an unusually fresh episode, presumably due to the enhanced contribution from 801 freshwater runoff associated with winter storms. The discrete observations of the ESS time series 802 (with irregular sampling averaging 3 times per year) are likely to under-sample or alias this mode of 803 variability which occurs on timescales of days or weeks. 804 As indicated by the TPM current meter observations in Figures 12 c and f, poleward residual flow on 805 the inner shelf during and following a wind event can exceed 25 cm s -1 for 10 days, which would 806 equate to a tracer transport of over 200 km. For comparison the distance between the shelf edge 807 and Tiree Passage via the accepted flow pathway for oceanic water is 130 km (Figure 1), so it seems 808 probable that water can be advected rapidly from the shelf edge during storms. The assumption 809 implicit in the 4-state model is that this process will be additive, with the impacts of individual wind 810 events being summed over time. However the events examined in Figure 12 suggest that the inner 811 In Figure A.1, the salinity measured by TPM at 20 m depth is compared to independent observations. 935 The maximum, minimum and mean salinities measured by each calibration CTD are shown in red. In 936 addition, the other datasets comprising the TPC time series are included to provide improved 937 temporal coverage. 938 939 Figure  In general, the Microcats were found to perform well with most salinity data falling within the error 945 bars of the independent observations. The only period of concern was between 2005 and 2007 946 which appeared to show a systematic decline in the salinity reported by the mooring, which is not 947 corroborated by any other datasets. Whilst this decline spans multiple mooring deployments, it was 948 found that the same instrument was redeployed repeatedly during this period due to a lack of 949 alternatives. Therefore it is suspected that this salinity decline was a systematic sampling error due