Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 The upper ocean of the Rockall Trough exhibits coherent interannual variations in temperature and salinity over the past 26 years, with highs in the mid-1980s and late 1990s and lows in the late 1970s and early 1990s, and with ranges of ±0.5°C and ±0.05 in salinity. The origins of the interannual changes are discussed, covering three potential influencing factors: the propagation of anomalies developed upstream of the basin, the effect of local air-sea interaction, and the result of changes of regional circulation bringing different water masses into the region. The changes in heat and freshwater content of the upper ocean are directly compared to observed variations in air-sea heat and freshwater fluxes over the period of the time series. It is shown that the role of the atmosphere in locally altering the oceanic properties, particularly salinity, is relatively small and insufficient to explain the changes. Two recent hydrographic surveys are analyzed to ascertain how the distribution of water masses to the south of the basin may influence the properties of the northern Rockall Trough upper ocean, and the results are reviewed in the context of historical analyses. It is found that the critical factor in determining the properties is the varying amount of relatively cool and fresh North Atlantic Current water mixing with the dominant water mass, the warm saline Eastern North Atlantic Central Water at the entrance to the basin. Variations of inflowing water masses are caused by east-west changes in the location of the subpolar front, and the relationship of these changes in regional circulation to the North Atlantic wind stress field is discussed.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The upper ocean of the Rockall Trough exhibits multiyear variability in its properties that can be distinguished from the seasonal effects of summer stratification and winter mixing [Ellett et al., 1986]. Since 1975 the mean temperature of the upper ocean has varied by ±0.5°C and the salinity by ±0.05, once the seasonal cycle has been removed [Holliday et al., 2000]. A 26 year time series of full depth hydrographic measurements at 8°–15°W 57°N shows strong interannual variability and suggests the existence of a 15–20 year cycle, with minima of temperature and salinity in 1976 and 1994, and maxima in 1982 and 1998 [Holliday et al., 2000, Reid et al., 2001]. The mechanisms that cause the variability of the Rockall Trough upper ocean temperature and salinity on interannual to decadal timescales are investigated.
 The hydrography of the Rockall Trough was reviewed by Ellett et al. . There is a general northward drift of warm saline upper waters with anticyclonic recirculation and enhanced flow in the shelf edge current at the continental shelf break. The circulation in this region is complex in mesoscale detail with many eddies and anticyclonic recirculation features around banks and seamounts. The upper waters lie within the temperature-salinity range of Eastern North Atlantic Central Water (ENAW) and are substantially more saline than the subpolar mode waters of the Iceland Basin. Ellett and Martin  and Ellett et al.  showed that the ENAW in the Rockall Trough had originated from the south (the intergyre Biscay region), but also contained influences from Sub-Arctic Intermediate Water (SAIW) and Mediterranean Water (MEDW).
 Various theories to explain the variability in the Rockall Trough have been put forward, but none have been satisfactorily proved. The three main theories are local atmospheric exchange, propagating anomalous events from elsewhere, and lateral movement of major fronts. The first discussion of Rockall Trough subsurface and surface variability was presented by Ellett and Martin . While not directly attempting to determine the origin of the variability, they did note a mismatch in the timing of surface temperature and salinity anomaly extremes and attributed that to varying amounts of cooling of the upper water by the atmosphere. They also highlighted the observations made by Tulloch and Tait  of water entering the Rockall Trough from the west in the early 1950s, contrasting that to their own observations in the mid 1960s where all the upper ocean water appeared to be coming from the south. The topic was revisited by Ellett et al.  in which it was noted that the origin of the dramatic decrease in temperature and salinity in the late 1970s was attributed to the Great Salinity Anomaly (GSA) [Dickson et al., 1988] but that other explanations include an increase of SAIW advected into the southern Rockall Trough by eddies [Ellett, 1982] and an eastward shift of the sub-Arctic (subpolar) front [Dooley et al., 1984]. More recently a reanalysis of North Atlantic time series by Reverdin et al.  examined a collection of subpolar temperature and salinity time series from 1948 to 1990, and looked for spatial coherence of heat and salt content across the North Atlantic (in the Rockall Trough only surface temperature and salinity were examined). They concluded that 70% of the variance in the basin-wide North Atlantic upper ocean salt content was caused by anomalous features propagating from the west of the subpolar gyre to the east. However they conceded that the study was dominated in time by the GSA period, and in space by the salt content variations in the western Atlantic (western variations being 3 times the magnitude of eastern). During their discussion of mechanisms to explain the anomalies of heat and salt content, Reverdin et al.  preferred the hypothesis of propagating events to the hypothesis of changes in position of the Sub-Arctic Front. However they conclude that in the eastern subpolar region, atmospheric forcing “must play a significant role in modifying the low frequency temperature variability”, and further that mixing and advection could explain the salinity variations.
 In this paper we analyze new hydrographic and atmospheric flux data to investigate the mechanisms behind the interannual to decadal variability of the upper ocean temperature and salinity in the Rockall Trough. The processes that determine the variability fall into three main categories; (1) local air-sea interaction, (2) changes in the regional circulation, and (3) propagating anomalies. Holliday et al.  has already shown that the variability was not simply a reflection of propagating anomalies in the ENAW. Here we examine the effect of local air-sea interaction and changes in regional circulation, presenting evidence of the relative contribution of each one to the Rockall Trough water mass properties.
 The oceanographic data analyzed here are the north Rockall Trough time series of full depth CTD data collected from 1975 to 2000 (the “Ellett Line”). The time series and the variability of temperature, salinity and transport were described by Holliday et al. . The location of the time series is given in Figure 1.
2. Variations in Local Air-Sea Interaction
 With the recent publication of improved atmospheric flux products we are now in a position to look at variations in the fluxes across the ocean-atmosphere boundary and to examine the hypothesis that local air-sea exchange was the origin of interannual variations in temperature and salinity. If this hypothesis was correct then the magnitude and pattern of variations in the oceanic heat and freshwater content would match the magnitude and pattern of variations in the atmospheric fluxes over the same period. The model behind this hypothesis is that the overall steady state in the Rockall Trough is achieved by the input of heat and salt by advection balanced by output of heat to the atmosphere and input of freshwater from the atmosphere. The approximate mean budget is an annual net heat loss of −40 W/m2 and a net gain of freshwater of 0.43 m/yr [Josey et al., 1998], (the sign convention being positive from atmosphere to ocean) balanced by 1.5 Sv of water at a mean of 9.0°C, 35.33. Any year-to-year variations in the atmospheric fluxes would result in changes in heat or freshwater content reflected in the mean temperature and salinity assuming the advective fluxes were constant.
 Heat and freshwater content anomalies for each section can be calculated referenced to an arbitrary value of temperature and salinity. Heat content anomaly H = ∫ρCpTdz where ρ is density, Cp is specific heat capacity of saltwater, T is the temperature (reference temperature of 0°C) and z is depth in meters. Freshwater content anomaly where Sref is the reference salinity (here 36.00), S is salinity, and z is depth in meters. In order to remove any bias introduced by a predominance of summer and autumn cruises in the time series, a seasonal model of both heat and freshwater content needs to be removed from the anomalies. For this purpose the upper ocean was divided into 3 layers (0–300, 300–600, and 600–900 m) and heat and freshwater content anomalies calculated for each section in the time series. To focus on the interannual variations the time series mean was subtracted. A sinusoidal curve was fitted to the anomalies sorted on month and for fits that were statistically significant at the 95% confidence level the modeled monthly values were subtracted.
 The heat content anomalies are shown in Figure 2. The anomalies in the surface layer (0–300 m) plotted against year reflect the relatively random sampling of the seasons and hence provide little information regarding decadal-scale patterns. However sorted by month they show the expected clear seasonal cycle, with maximum heat content anomaly in September/October and the minimum at the end of winter in March/April. The seasonal model in the surface layer is statistically significant at the 95% confidence level (r2 = 0.83 for n = 45) and can be subtracted to show the true interannual variability. The middle layer (600–900 m) also has a statistically significant seasonal cycle (r2 = 0.42 for n = 45), though it takes a rather different form with the highest temperatures in December–January, decreasing through to June, then slowly increasing to the early winter maximum again. The explanation for this pattern is probably the winter convection initially bringing warmer water to the intermediate depth (December–January), then temperatures decrease as the convective process cools the mixed layer (February–April), restratification occurs in May halting the cooling process, the temperature increases slightly as the mixed layer breaks down in the summer and advection brings warmer water into the region again, then the temperature remains fairly stable until the initial convection occurs in early winter. The fitted seasonal cycle for the lower layer (600–900 m) is not statistically significant (r2 = 0.13 for n = 45).
 For those seasonal models with a significant fit (the layers 0–200 and 400–600 m), the monthly model values were subtracted from the heat content anomalies. The anomaly time series are replotted in Figure 3 to examine the interannual variability. The surface layer exhibits more noise than the lower layer, but indicates relatively stable heat content from 1975 to 1995, followed by clear warming from 1995 onward. The middle and lower layers both show evidence of slightly increasing heat content from 1980 to 1985, and subsequent decreasing heat content to a minimum in 1995, and a rise from 1995 to 2000 of the same order of magnitude as the surface layer. An important feature of this figure is that the lower layer shows the same interannual variations as the middle layer particularly, even though the analysis of the seasonal cycle suggests that it is not affected by seasonal changes in atmospheric heat fluxes. This is the first small piece of evidence that local atmospheric flux variations may not be responsible for the changes in heat content anomaly.
 The freshwater content anomalies are shown in Figure 4. Unlike the heat content anomalies, none of the models shown are statistically significant at the 95% confidence limit (surface layer r2 = 0.02, middle layer r2 = 0.09, lower layer r2 = 0.04, n = 45 for all layers). Thus the modeled seasonal cycle can be rejected and the anomalies from the 1975–2000 mean taken as representing the interannual variability. A clear seasonal cycle in the surface salinity has previously been reported in the literature; Ellett and Jones  reported a typical amplitude of salinity ∼0.05. A decrease in salinity of 0.05 in the summer months would result in increase in freshwater of 0.07 m in 50 m, an order of magnitude smaller than the interannual variations in the upper 300 m. In fact fitting a sinusoidal curve to the monthly values of the top 50 m also results in a statistically insignificant fit (r2 = 0.07, n = 45), so we conclude that the seasonal surface salinity cycle is negligible in terms of the freshwater content.
 The most striking impression from Figure 4 is that all three layers show the same interannual to decadal pattern, and all contribute the same magnitude of freshwater anomaly to the water column. It is noticeable that the coherence of the patterns is far stronger than the heat content anomalies. The highest freshwater levels occur in the later 1970s when the Great Salinity Anomaly passed through the region; a feature which originated in an unusually large influx of cold fresh water to the subpolar gyre. High freshwater content anomaly was prevalent again in the early 1990s and it decreased to a low in 1998 (upper layer) and in 2000 in the middle and lower layer.
 The oceanic heat and freshwater content anomalies can be directly compared to atmospheric heat and freshwater flux anomalies. The main source of atmospheric flux data for this analysis was the time series used to generate the SOC Flux Climatology [Josey et al., 1998]. The product contains monthly files from January 1980 to December 1997 for each 1° × 1° box in the global ocean. In addition, data from the CPC Merged Analysis of Precipitation (CMAP) [Xie and Arkin, 1997] are used for the freshwater budget analysis. The SOC climatology was constructed from the COADS database of ship observations, but with corrections applied for errors due to reporting procedures on individual ships. The disadvantage of the ship-based data set is that it is not homogeneous in sampling distribution, though the North Atlantic is relatively well sampled. The main weakness of the SOC climatology is generally considered to be the precipitation; ship-based observations can suffer from biases due to avoidance of major storm tracks, and because of errors in visual estimates of rainfall and/or errors in ship-borne rain gauges. The CMAP data set is a relatively recent precipitation product based on satellite estimate of rainfall and provides an alternative estimate to use in this analysis (2.5° × 2.5° grid size).
 The net heat flux (Qnet) is the sum of sensible (QH), latent (QE), longwave (QLW) and shortwave (QSW) fluxes. Net heat flux (positive values represent heat gain by the ocean). The data in the SOC climatology are in monthly bins which can be averaged to obtain a yearly mean net heat flux. A “year” here was chosen as April to March so that it runs approximately between the end of winter one year and the next. The convention used is to name the April–March year by its start date, for example, April 1980 to March 1981 is named “1980”. The area selected over which the averaging was carried out is 55°–57°N, 10°–15°W and 50°–55°N, 12°–17°W (Figure 1). This region effectively covers the main part of the Rockall Trough with the Ellett line at the northern edge, and adequately covers the distance the sampled Rockall Trough water may have traveled in one year. The boundaries were based on the general source of Rockall Trough water given in (inter alia) Ellett et al. , Pollard et al. , McCartney and Mauritzen , and on baroclinic transport estimates (mean and range) given by Holliday et al. . The variance of heat flux values in the region can gives an estimate of uncertainty in each annual mean. By attempting to close the global heat budget it is estimated that the Climatology has a global error of around 30 W/m2, though Josey  argue that the error is greater in some areas than others, and less in the North Atlantic. Josey  makes a comparison with in situ data from meteorological buoys west of the Iberian Peninsula, and the estimated error was 10% of the latent heat flux. In the Rockall Trough there are no in situ data for an equivalent error estimate, so here the standard error is shown while acknowledging there is an unknown bias in the absolute heat loss.
 The annual means of precipitation (P) and evaporation (E) were calculated over the region from where water in the Rockall Trough originates. Precipitation means were calculated from the SOC climatology and from CMAP to assess the relatively validity of the means which were expected to be significantly different because of the possible errors in the SOC climatology. Precipitation is recorded as a variable in both data sets, but evaporation has been derived from the SOC climatology from observed latent heat (QE) and sea surface temperature (SST), following the bulk formula given by da Silva et al. : . The P and E data were averaged into annual values covering the whole area and P-E and the standard error calculated. (Table 1 and Figure 5). Rather surprisingly the magnitude and interannual variations of the precipitation agree rather well between SOC climatology and CMAP. This is an encouraging result and suggests that the pattern of variability they describe is reasonably robust, though it is still possible that both absolute magnitudes may have large biases. In the following analysis the CMAP precipitation data are used in preference over the SOC climatology precipitation data for the reasons outlined above.
Table 1. Mean and Standard Deviations of Atmospheric Heat and Freshwater Fluxes for the Area 55°–57°N, 10°–15°W and 50°–55°N, 12°–17°W
Annual mean heat flux, W/m2
SOC annual mean precipitation, m/yr
SOC annual mean evaporation, m/yr
CMAP annual mean precipitation, m/yr
Annual mean freshwater flux (P-E), m/yr
 If the “local air-sea interaction” hypothesis was correct then the magnitude and pattern of these variations in atmospheric fluxes would be equivalent to the changes in heat and freshwater content. To make a direct comparison of the atmospheric and oceanic time series we have converted the atmospheric fluxes to effective annual changes in heat and freshwater content by multiplying by time (seconds) and plotted them with the integrated heat and freshwater content anomalies (Figure 6).
 By simply looking at the range of atmospheric flux variations in Figure 6 it can be seen that they are insufficient to explain the range of oceanic variations. The atmospheric flux variations were around half the size required to produce the changes in heat content anomaly so could potentially be a contributing factor. However the tendency over the time series is for the atmospheric fluxes to act in opposition to the oceanic heat content anomaly, and it is probable that this damping effect results in the relatively unchanging values in the layer most affected by the atmosphere, the upper 300 m. Most strikingly, the atmospheric freshwater flux variations were an order of magnitude smaller than the variations in freshwater content anomaly of the Rockall Trough upper ocean, and appear as nothing more than insignificant noise. The evidence from the heat and freshwater content anomalies thus disproves the hypothesis that local air-sea interaction is the major cause of the interannual to decadal Rockall Trough temperature and salinity variability.
3. Changes in Regional Circulation
 The final hypothesis to be tested is that the temperature and salinity variability was caused by changing proportions of different water masses advected into the region, in other words, changes in the regional circulation which controls the source of advected water for the region. The upper ocean water masses that lie to the south and west of the Rockall Trough are the warm saline ENAW and the relatively cool fresh Western North Atlantic Central Water (WNAW) in the upper layers, and the Sub-Arctic Intermediate Water (SAIW) and Mediterranean Outflow Water (MEDW) at intermediate depths [Ellett et al., 1986]. Potentially all four water masses could enter the basin, though the upper ocean is dominated by ENAW [Ellett and Martin, 1973, Holliday et al., 2000], and the intermediate water is relatively homogeneous with a tight θ-S relationship that lies somewhere between SAIW and MEDW [Ellett et al., 1986, Read and Ellett, 1991]. ENAW originates from the intergyre region to the south and is more saline than the water of the NAC frontal zone, gaining its salinity largely because of winter mixing and cooling which act to increase salinity at a given temperature [Pollard et al., 1996]. WNAW is essentially the warm saline water carried by the NAC, though is relatively cool and fresh compared to ENAW and will be frequently referred to as “cool and fresh” in the discussion below. The water carried between the fronts of the NAC is always fresher and cooler than ENAW, though it may be more saline than WNAW where it has mixed with ENAW to the southeast (eastern margin [Pollard et al., 1996]), or fresher where it has mixed with SAIW to the north (Mid-Atlantic Ridge [Harvey and Arhan, 1988]).
 The literature shows that not only do the properties of the 4 main water masses vary from year to year, but their regional extent does too. Harvey  showed the limit of the SAIW and WNAW to be around 20°W in data from 1957 to 1958 and 1962, but in contrast Read and Ellett  found them as far east as the southern Rockall Trough (16°W) in data from 1989. The interesting difference between these meridians is their relative location to the Rockall Trough; 20°W is the longitude of the southern tip of the Rockall Plateau at water depth of 1500 m (17°W at 1000 m) so water moving northward west of 17°–20°W will be forced to flow westward around the topography into the Iceland Basin. East of 17°–20°W, northward flowing water can move freely into the Rockall Trough. The hypothesis to be tested is that the temperature and salinity variations in the Rockall Trough can be explained as periods of greater or lesser influence of the relatively cool fresh SAIW and WNAW (the NAC water) in the dominant water mass, the warm saline ENAW.
 Two recent hydrographic surveys are analyzed to investigate the eastward extent of NAC water and how that relates to the properties in the Rockall Trough. The first is a survey from October 1996 completed with full depth CTD stations and a SeaSoar; a towed vehicle that undulates between the sea surface and 350–500 m depth, providing closely spaced inclined CTD profiles [Leach and Pollard, 1998]. The second is a CTD survey carried out in May 1998 [Smythe-Wright, 1999]. The surveys were selected for this analysis because they occurred at contrasting periods in terms of the Rockall Trough ENAW properties; October 1996 was relatively cool and fresh in the Rockall Trough, whereas May 1998 was warm and saline.
 Selected CTD stations from the October 1996 survey can serve to illustrate the relative properties of SAIW, WNAW, ENAW and MEDW because the survey included a series of stations around the entrance to the Rockall Trough. Figure 7a shows three potential temperature-salinity (θ-S) curves; the central curve is from north Rockall Trough time series (station 12943) showing the typical well-mixed curve. The curve for station 13005 (53°59.9′N, 24°34′W) shows the SAIW water mass south of Hatton Bank, and the curve for 13017 (50°15.0′N, 15°0.1′W) shows the MEDW water mass at the Porcupine Bight. Note that both water masses have their cores at density (σθ) 27.50 kg/m3 though that isopycnal is much shallower in the SAIW. Figure 7b shows θ-S curves from May 1998 and it can be seen that while SAIW was influencing the water at one station on 20°W (13492), there was no true SAIW minimum as seen in 1996. The CTD in the north Rockall Trough (13527) appears to have virtually no WNAW influence above approximately 500 dbar as it mixes toward and eventually overlays the upper θ-S curve (ENAW) of the station from the Porcupine Bank (13487). Below 500 dbar the north Rockall Trough station shows the characteristic θ-S curve which lies between the SAIW and MEDW types, but the whole curve (to 1200 dbar) was shifted toward the MEDW relative to the 1996 profile (plotted in gray) by 0.02 in salinity and 0.2°C in temperature. This results in the whole upper ocean, from the surface to 1200 dbar, being warmer and saltier in 1998 than in 1996.
 These two cruises were contrasting in properties in selected stations to the south of the Ellett line, as well as in the Rockall Trough itself. From the selected profiles the high temperatures and salinities of the Ellett line in 1998 appear to be due to the upper water column being entirely ENAW in origin, whereas the cooler fresher Ellett line in 1996 reflects the fact that the whole upper ocean was not just ENAW but also contained influence from SAIW and WNAW. The two semisynoptic surveys allow mapping of the spatial extent of the WNAW and ENAW in each year, which can be used to examine whether these contrasting scenarios were consistent across the whole region to the south and west of the Rockall Trough. By looking at the CTD θ-S relationships we can define a boundary between the two upper ocean water types and investigate how far west the NAC waters reach in the two different years (following the method of Pollard et al. , which mapped salinity anomaly on density surfaces). The temperature-salinity relationship of the water masses vary from year to year so we consider the 1996 cruise separately from the 1998 cruise.
 The SeaSoar survey in October 1996 provided very high horizontal resolution, shallow CTD data between the deep stations shown in Figure 7. SeaSoar θ-S curves for a track running east from station 13005 onto the continental shelf are shown in Figure 8a. There are three clusters in the shallow water (excluding the seasonally warmed surface layer) indicating the presence of three types of central water, the most saline being ENAW with salinity greater than 35.375 on isopycnal 27.20 kg/m3. The modified WNAW of the NAC forms a cluster of profiles with salinity range 35.275–35.375 on the isopycnal 27.20 kg/m3, and the freshest water type is the true WNAW. The depth of isopycnal 27.20 kg/m3 ranges from 150 m to more than 350 m across the area, but it is always below the seasonal thermocline. The SeaSoar profiles are relatively shallow (<400 m) but show the presence of SAIW at the deepest level in some profiles. Data from all the SeaSoar tracks and CTDs have been gridded, and the salinity mapped on isopycnal 27.20 kg/m3 is shown in Figure 8b. Station positions are overlaid in Figure 8b; the gridding was performed on the isopycnal, with no account of topography taken, and so is less reliable away from the profile positions, especially over the topography of the Rockall and Hatton Banks. (The grid was 0.25° latitude by 0.5° longitude, with a weighted search radius for each point being 2.5° latitude and 5° longitude.) Salinities between 35.275–35.375 (the NAC WNAW) are given medium gray shading to distinguish them from the dark grays of the ENAW, and the light grays of the WNAW. It is clear that NAC WNAW reaches east of 20°W to the south of the Rockall Trough where it is able to mix with the ENAW and influence the properties of the water flowing into the basin.
 The May 1998 survey took closely spaced CTD stations rather than SeaSoar data, so the horizontal resolution is lower but the data reach to the seafloor. There is a less clear distinction between WNAW and ENAW than the October 1996 survey, mostly because very few of the stations were far enough west to sample the true WNAW and the SAIW. Station 13492 is one that does, and it is possible to find a boundary between a type of NAC WNAW and the ENAW at salinity 35.325 on isopycnal 27.25 kg/m3 (Figure 9a). By mapping salinity on isopycnal 27.25 kg/m3 (Figure 9b) it can be seen that there is no true WNAW in evidence, just a limited amount of NAC WNAW at 20°W, and that almost all the water south of, and entering the Rockall Trough is ENAW.
 In summary, the two recent regional surveys have shown that in 1996 when the Rockall mode water was cooler and fresher, NAC frontal WNAW flowed east of 20°W and mixed with the ENAW, whereas in 1998 when the Rockall mode water was warmer and saltier the NAC WNAW was only present in small amounts at 20°W and did not significantly influence the ENAW. The evidence supports the hypothesis that regional circulation changes caused the temperature and salinity changes observed in the north Rockall Trough.
 Further evidence of the two scenarios (WNAW and NAC frontal waters east of 20°W, cool/fresh Rockall Trough; WNAW and NAC frontal waters west of 20°W, warm/saline Rockall Trough) can be found in the literature. In 1989 when the WNAW and SAIW was observed reaching all the way to the eastern margin [Read and Ellett, 1991], the Ellett line water was cool and fresh (actually the start of the cool fresh period that extended to 1996). Similarly, Read  showed that in the middle of that cool fresh period in 1991, WNAW was seen east of 17°W at 52°N. Ellett et al.  described data from that OWS J that showed WNAW and SAIW east of 20°W in 1974 and 1977, the latter year being the time of the cold fresh Great Salinity Anomaly (GSA) in this region.
 Other examples show WNAW and NAC frontal waters remaining west of 20°W at times when the Rockall Trough was warm and saline. Harvey's  description of ENAW was based on data from 1957 and 1958, and he noted at that time that WNAW and SAIW was only found west of 20°W, while Ellett and Jones  showed the water in the Rockall Trough to be unusually warm and saline in the late 1950s. In 1966 the Rockall Trough was cool but saline according to Ellett and Jones , and Wade et al.  note from OWS J data that in 1966 WNAW and SAIW were only found west of 20°W. The last example highlights the point that this hypothesis does not exclude the influence of changing properties of the source waters. Property variations in both ENAW and WNAW do influence the Rockall Trough temperature and salinity, and this is especially clear when one water mass is dominant over the other. In 1966 the cool Rockall Trough water probably reflected cooler than usual ENAW to the south. During the GSA period the Rockall Trough salinity was especially low, not just because it contained a greater amount of WNAW and SAIW than average, but also because those water masses were unusually fresh.
 There has been evidence of significant changes in position of the NAC at the eastern margin; White and Heywood  investigated interannual changes in location of the NAC as determined by zones of maximum eddy kinetic energy anomaly. They showed that while 1987, 1988 and 1991/1992 showed some small variations from the “average state” over the total period, in winter 1993/1994 the “NAC zone” moved from the Iceland Basin and entered the Rockall Trough, presumably bringing cool fresh water with it (1994 was a particularly cool fresh year at the Ellett line). They argued that this change was associated with the southward shift by about 10° of latitude of the line of zero wind stress curl which resulted in a southward change in the location of the NAC of about 2°–3° in the mid Atlantic, and about 4° eastward move of the NAC. Bersch et al.  showed that a westward retreat of the NAC in the Iceland Basin in 1995/1996 caused an increase in the temperature, thickness, salinity and density of subpolar mode waters in the east as a different water mass filled the region. They also attributed the movement of the NAC to the effects of the wind field described by the sudden change from predominantly positive North Atlantic Oscillation (NAO) index to negative in the winter of 1995/1996; the result of the decrease in the NAO index was a weakening of the westerlies and a subsequent weakening of the strength of the circulation of the subpolar gyre. Although October 1996 was a cool fresh year at the Ellett line, since 1996 the Rockall Trough has become progressively warmer and more saline as little WNAW contributes to its properties.
 Subsequent studies have shed more light on the nature of temporal changes in the latitude of the line of zero wind stress curl, its relationship to the NAO and the effect on meridional overturning. Lorbacher and Koltermann  find a strong positive correlation between the latitude of the line of zero wind stress curl and the winter NAO index (with a time lag of 1 year), and a subsequent positive correlation with the northward volume and heat transport of subpolar mode water across 48°N. Between 30° and 40°W the latitude of the zero wind stress curl line varies from 43°–49°N, with most northern positions during the high NAO periods of the early 1980s and early 1990s. This is clear indication of a fast dynamic response of the ocean to changes in atmospheric conditions. However this pattern of response does not simply translate to the changing conditions a little further to the east and north in the Rockall Trough, where it appears that an east-west shift of the NAC, relative to key topography and related perhaps to the shape of the subpolar gyre, is dominating the conditions.
 A measure of the strength of the NAC and the subpolar gyre as a whole has been developed by Curry and McCartney  and termed an ocean “transport index”. They argue that the baroclinic pressure difference across the Gulf Stream between Labrador Sea and Bermuda provides an index of the strength of the circulation of the subpolar gyre. When the index is high the NAC is strengthened, and there is evidence that in the northeast Atlantic the eastern boundary current is also stronger. The index suggests that the transport of the Gulf Stream and NAC was low during the low NAO years of the 1960s and peaked in the high NAO years of the 1990s, and similarly to the findings of Lorbacher and Koltermann , in general terms is related to the latitudinal shifts of the westerlies (the line of zero wind stress curl). While there is no statistical correlation between the Rockall Trough properties and the transport index, in the Curry and McCartney  results are tentative suggestions of a link. Peaks in the property anomalies and the transport index occurred in the mid-1980s and the late 1990s possibly because the higher strength of the eastern boundary current increased the advective flux of the warm saline ENAW. However we have shown that it is not just the strength of the eastern boundary current that is important; the location of the subpolar front relative to the Rockall-Hatton Bank is critical. There are periods when it is clear that the transport index does not represent the full scenario. In the late 1960s the Rockall temperature and salinity anomalies were high while the transport index was decreasing from a maximum in 1958. In the mid 1970s there was a peak in the transport index when the Rockall Trough properties were dominated by the GSA effect. Subsequently, in the early 1980s, the salinity and temperature of Rockall Trough were increasing rapidly while the transport index dipped to a low in 1981. In summary, however, neither the ocean transport index nor the atmospheric NAO appear to be the only influence on the flux of western surface waters entering the Rockall Trough.
 A model-based investigation of mechanisms of low-frequency variations in North Atlantic circulation has shown both a rapid (barotropic) ocean response to the atmosphere (the NAO) and a delayed (baroclinic) ocean response [Eden and Willebrand, 2001]. The interplay of these response times could explain the observed “similarity” but noncorrelation of the conditions in the Rockall Trough to the changes in the NAO. The non-eddy-resolving model was forced with realistic surface heat and wind stress fields, specifically the variability due to changes in the NAO index. They found that the NAO changes in wind stress caused a fast (time lag 0 year) Sverdrup-like ocean response in the form of an anticyclonic circulation anomaly near the subpolar front, reducing the net northward heat transport as NAO increased (contradicting the Lorbacher and Koltermann  results at 48°N). However they noted an opposite and delayed response (time lag 3 yr) leading to a strengthening of the subpolar gyre and a corresponding enhancement of the northward heat transport. A simultaneous response to high NAO-induced increase in surface heat fluxes resulted in further increased gyre strength and heat transport with a lag of 3–8 years. For the eastern subpolar gyre, the model results suggest that the fast response is more significant than the delayed response in terms of the barotropic stream function, but that the latter does nonetheless have an effect. The immediate effect of the short-term response in the east is the northwestward shift of the subpolar front as observed south of the Rockall Trough in the early 1980s and late 1990s. The period of freshening corresponding to an eastward movement of the NAC fronts to the south of the basin in the early 1990s, while occurring during rising NAO conditions, may have been the delayed response to the increasing NAO index 5–8 years earlier.
 The temperature and salinity of the upper ocean of the northern Rockall Trough has varied by ±0.5°C and by ±0.05 since 1975, with minima in both properties in the late 1970s and the early 1990s, and maxima in the mid-1980s and late 1990s. The interannual cycles in temperature and salinity can be viewed as changes in heat and freshwater content anomalies which can be directly compared to variations in local atmospheric fluxes for the same period. Those variations in air-sea transfer of heat and freshwater were much too small to account for the changes in the upper ocean. Instead it is argued that the temperature and salinity variations in the Rockall Trough were the result of changes in regional circulation, manifested as a greater or lesser contribution of the cooler and fresher WNAW and NAC frontal water to the dominant water mass of the basin, the warm saline ENAW. The flux of western surface waters into the Rockall Trough may be related to changes in the strength and shape of the subpolar gyre circulation which in turn is related to changes in the wind field; for example in the early 1980s and 1ate 1990s the periods of maximum temperature and salinity in the Rockall Trough (minimum flux of western water), occurred during periods of rising NAO index, with increased subpolar gyre circulation, a westward movement of the subpolar front, and the associated strengthening of the eastern boundary current; all fast oceanic responses to changes in the NAO index. However there are periods when the direct relationship between the Rockall water mass properties and the NAO does not hold; these may be a local manifestation of the delayed baroclinic response of the subpolar gyre to changes in the wind stress field.
 The author thanks Harry Bryden for helpful comments on the analysis and contents of the manuscript; Simon Josey, Andrew Coward and Bob Marsh for help in accessing various data sets. As always we acknowledge the tremendous effort by David Ellett and his colleagues for the years of observations on the Rockall Trough time series, the “Ellett line”. This research formed part of the author's PhD thesis for the University of Liverpool, supervised by Raymond Pollard and Harry Leach. This is a UK Contribution to the World Ocean Circulation Experiment.