Analysis of surface drifter tracks in the North Atlantic Ocean from the time period 1990 to 2007 provides evidence that warm subtropical waters have recently increased their penetration toward the Nordic seas. Prior to 2000, the warm water branches of the North Atlantic Current fed by the Gulf Stream turned southeastward in the eastern North Atlantic. Since 2001, these paths have shifted toward the Rockall Trough, through which the most saline North Atlantic waters pass to the Nordic seas. These surface drifters are able to overcome the Ekman drift, which would force them southward under the westerly winds dominating the subpolar Atlantic, yet the changes in path cannot be accounted for by changes in Ekman drift. Eddy kinetic energy from satellite altimetry shows increased energy along the shifted drifter pathways across the Mid-Atlantic Ridge since 2001. These near-surface changes have occurred during continual weakening of the North Atlantic subpolar gyre, as seen by altimetry. They are also consistent with the observed increase in temperature and salinity of the waters flowing northward into the Nordic seas. These findings suggest the changes in the vertical structure of the northern North Atlantic Ocean, its dynamics, and exchanges with the higher latitudes. Wind stress and its curl changes are discussed as a possible forcing of the changes in the pathways of the subtropical waters.
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.
 Saline Atlantic waters flow into the Nordic seas via three routes from the North Atlantic: through the Rockall Trough to the Faroe-Shetland Channel, through the Iceland basin over the Iceland-Faroe Ridge and via the Irminger Current passing west of Iceland (Figure 1). The Irminger Current also supplies saline waters to the western subpolar gyre where they participate in the intermediate and deep water formation. We focus on the surface expression of the routes leading to the first two channels, which carry most of the volume flux into the Nordic seas [Hansen and Østerhus, 2000]. There are indications that the water masses in these channels change over time [Holliday et al., 2008; Hátún et al., 2005; Pollard et al., 2004], and particularly striking increases in the upper ocean temperature and salinity have been observed in the recent decade in the upper northeastern Atlantic, where Atlantic Water flows across the Greenland-Scotland Ridge into the Nordic seas [Hátún et al., 2005; Holliday et al., 2008; Sarafanov et al., 2008]. The salinization of upper and deep waters can also be followed westward toward the Irminger and Labrador seas [Yashayev et al., 2008]. During the last decade also other major climatic fluctuations have been reported: significant decrease of deep convection in the Labrador Sea [Dickson et al., 1996, 2002]; the westward movement of the subarctic front in the eastern subpolar gyre [Bersch, 2002] which appears simultaneously with a northward surge of the highly saline Mediterranean waters at the intermediate depths along the eastern margin [Johnson and Gruber, 2007] and the weakening of the subpolar gyre associated with increased sea surface height [Hakkinen and Rhines, 2004]. Model results [Hátún et al., 2005] showed that the sea surface height variability is associated with a westward contraction of the subpolar gyre allowing the high-salinity subtropical waters to flow northeastward toward the Nordic seas. In their view the source of the saline waters was in the eastern Atlantic.
 These phenomena suggest a changing circulation and several hydrographic monitoring programs have begun to address these changes (the OVIDE project, the Extended Ellet Line (www.noc.soton.ac.uk/obe/PROJECTS/EEL/latestresults.php), and repeat occupations of key WOCE lines (for example AR7E). The important question as to whether this variability of the upper ocean subpolar circulation is related to the Atlantic meridional overturning circulation (AMOC), particularly its upper limb, remains unresolved. Model simulations suggest strong linkages between subpolar gyre circulation, Labrador Sea Water production and the Atlantic MOC [e.g., Hakkinen, 2001; Bentsen et al., 2004; Böning et al., 2006] and yet determination of boundary current transports from moored current meters is challenging [Bryden et al., 2005; Fischer et al., 2004]. Early results from the RAPID program, which aims at monitoring the Atlantic MOC at 26°N show significant interannual variability, while decadal structure is still emerging (http://www.nocs.co.uk/rapidmoc/). The MOC combines both deep- and intermediate- depth water masses, and some models [e.g., Kohl and Stammer, 2008] argue that the dense overflows from the Nordic seas are more influential in the net MOC transport than is the Labrador Seawater production. Bailey et al.  have argued that the balance between deep (Nordic seas) and intermediate (Labrador Sea and entrained subpolar gyre water) sources of the MOC can be affected in circulation models by numerical mixing in the descending overflows.
 The upper ocean waters flowing to the Nordic seas through the Faroe-Shetland Channel and Rockall Trough are a mixture of the North Atlantic Current (NAC) waters (having components of both Gulf Stream and fresh subpolar waters), the saline waters of the upper eastern Atlantic, and Mediterranean overflow waters [Rossby, 1996; McCartney and Mauritzen, 2001; Pollard et al., 2004]. The determination of the source waters, for example, in the Rockall Trough is further complicated by the fact that those water masses undergo significant long-term changes in temperature and salinity through both air-sea interaction in winter, and lateral mixing with prolonged recirculation. Yet Holliday et al.  argue that air-sea interaction variability is insufficient to explain the evolving changes in the northward subpolar water mass, suggesting instead changing advective pathways. Thus, the rising temperature and salinity of the eastern subpolar gyre is strong evidence for an increasing inflow from the subtropics. Our interest is to explore which component, direct access by the Gulf Stream waters or the eastern Atlantic waters, is creating the increased salinization since the late 1990s and particularly since 2002 as depicted by Holliday et al. .
 The surface water pathways can be followed by surface drifters, which have been deployed since the 1980s in the North Atlantic Ocean. Using this Lagrangian approach, Reverdin et al.  and Brambilla and Talley  show that surprisingly few subtropical surface drifters are diverted to the subpolar gyre between 1989 (the initiation of the large-scale surface drifter program) and 2002. A rather different depiction of the surface currents is provided by the Eulerian average of surface drifter fields which shows the surface waters from the western Atlantic crossing the Mid-Atlantic Ridge (MAR) and continuing toward north, guided by the Reykjanes Ridge, to the Iceland Basin and the Rockall Trough [Krauss, 1986; Brügge, 1995; Fratantoni, 2001]. The NAC crossings over MAR tends often to be locked to the topographic fracture zones as has been discussed by Bower and von Appen , but east of MAR the NAC structure is diffuse and highly variable as established by the early drifter experiment of Krauss . Thus, the sources of warm and saline subtropical water reaching the Nordic seas and feeding the overflows have several interpretations which may have depended on the time period investigated. Our objective here is to show from surface drifter movement and satellite altimetry that the saline water pathways can change with time particularly east of MAR. Data sources of this study are discussed in section 2. Analyses of the surface drifter and altimetric data are presented in section 3 and section 4, respectively. Possible forcing mechanisms for the changes in pathways are discussed in section 5 and conclusions are presented in section 6.
2. Data Sources
 Our first objective is to explore current changes for which we use surface currents from NOAA/AOML Global Lagrangian Drifting Buoy Database. The data are available since 1989, although somewhat sparse in the early years. The surface drifters are drogued to 15 m depth below sea surface, and we will use only drogued drifter observations. The database provides the drifter location and temperature every 6 h. The lifetime of the drogues vary considerably but the average life span of the drifters (∼271 days) [Brambilla and Talley, 2006] has not changed significantly over the years. The distribution of drifter launches for three different periods (1991–95, 1996–2000 and 2001–2005) are shown in Figure 2a where the drifter lifetime is expressed in color. The first period has some sparse areas but overall has a good coverage of the longest-living drifters, a total of 466 drifters launched in the area 25°N–70°N, 90°W–5°E. The second period has 451 drifter launches, but it appears more sparse than the first period because the launches were clustered off Gulf of Maine, Labrador Sea, and eastern North Atlantic. From 2001 to 2005, there are 459 drifters. The number of drifters in different lifespan categories is shown in Figure 2b for the three periods. To update the drifter movement to recent times, we also show results from an overlapping 5-year period 2003–2007 which had 691 drifter launches. Seasonal and spatial statistics for the period 1989–1999 have been published by Fratantoni , who found the data seasonally balanced and adequate to describe the North Atlantic circulation. Lumpkin  also discussed seasonal bias for the North Atlantic (north of 20°N and away from the coast lines) showing it to be near zero for drifters binned in 1° squares and launched before 1 August 2002.
 In order to support findings from surface drifters we will use altimetry data from TOPEX/Poseidon, ERS-1 and -2, Jason-1, and ENVISAT missions which are merged into a 1/3° data and are available from Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO). Sea surface height and geostrophic current data are available at 7 day time intervals. All of these quantities are anomalies relative to the mean of 1992–2006 with no filtering of seasonal variability. The geostrophic current anomalies are used to compute eddy kinetic energy (EKE) hence EKE mainly reflects eddy activity associated with the instabilities of the current. There is some underestimation of EKE in the subpolar gyre because of the coarse resolution compared to the baroclinic radius deformation. Despite this we expect to detect relative changes in EKE which should be an indication of the changes in the current location or enhancement of the current [Heywood et al., 1994].
 The surface drifters are imbedded in the surface Ekman layer, thus wind stress changes are anticipated to have an impact on the drifter tracks. To visualize the changes of the surface currents in the Eulerian frame, we take advantage of OSCAR surface current analysis (http://www.oscar.noaa.gov) which combines altimetric geostrophic velocities and Ekman drift computed from remotely sensed wind products (SSM/I, NSCAT, QuikSCAT). At the same time our goal is to explore the importance of wind stress and wind stress curl changes to the Atlantic meridional overturning upper limb, especially to the limb that reaches the Nordic seas and the Arctic Ocean. To analyze the forcing of the surface currents, we use monthly wind stress data from NCEP/NCAR Reanalysis [Kalnay et al., 1996] which has been used to compute wind stress curl and its variability. We also use QuikSCAT wind 1/2° data (available since 1999) (http://winds.jpl.nasa.gov/missions/quikscat/index.cfm) for the most recent years to resolve finer resolution features in the wind stress curl.
3. Surface Current Paths From Drifter Tracks
 First we start with drifter data which shows the most apparent shift in the paths of the North Atlantic surface waters. We focus on three periods, 1991–1995, 1996–2000, and 2001–2005 because these periods have distinctly different behavior of the North Atlantic Oscillation (NAO) (e.g., from http://www.cgd.ucar.edu/cas/jhurrell/indices.info.html#naostatseas) which is the dominant variability mode of the atmospheric circulation [Hurrell, 1995]. The first period represents years with high NAO index (the highest values of the 20th century), the second period starts with an intense reversal of the NAO index in 1995/1996 winter, and the 3rd period represents years of fluctuating weak positive and negative NAO index. The drifter distribution in the three periods could potentially produce biases in describing the drifter path changes. Instead of relying heavily on statistics, we want to consider the totality of the changes as presented by Figures 3–6 which depict upstream and downstream drifter paths through specific subtropical and subpolar regions as was done by Brambilla and Talley  (hereinafter referred to as BT) together with the North Atlantic drifter track coverage in each of the three periods.
 As in the earlier work of BT, we first consider a limited subtropical region where the drifters are found during each time period, and their path prior and after entering this specific region. The drifter paths before and after leaving the subtropical (or the subpolar) box can extend to the adjacent periods, hence the timing of circulation changes is slightly blurred by this technique (which is why we will resort to altimetry to determine more accurately the timing of the circulation change). The drifter lifetime can limit whether subtropical drifters reach the NE corner of the subpolar gyre (as noted in BT) even when conditions would be favorable for such drifter paths. To distinguish the differences in paths of subtropical drifters over the three periods, and particularly the changes since 2001, we choose the same box as BT over the Gulf Stream (GS), 35°N to 47°N, 78°W to 48°W. The drifter tracks entering and leaving the box are shown in Figures 3a–3c during the three periods. The first period appears to be rather sparsely occupied with tracks in the eastern Atlantic, however sparseness of drifter tracks is not a likely reason for the contracted tracks in Figure 3a. Moving the box to the central Atlantic where it could catch more drifter launches (Figure 2a) does not improve the eastern extent of the tracks (not shown). This issue will be discussed later in this section. The most apparent conclusion from the Figures 3a–3c is the expansion of tracks northward and northeastward across the basin: The 3rd period shows a major shift in the surface water path from the subtropics, the western Atlantic waters that were previously (prior to 2001) feeding the Bay of Biscay waters, have turned northeastward toward the Rockall Trough. Some of these NE drifters were launched in the slope waters but passing through the Gulf Stream box. Figures 3a–3c show that drifters exiting through the eastern side of the box form a northern wall curving around the Flemish Cap with no drifters straying into the central Labrador Sea. So all drifters passing the eastern side of the box have to be inside the warm GS waters, and hence emulate the behavior of drifters that could have been launched within the Gulf Stream. We also show an additional sequence of drifter tracks for years 2003–2007 in Figure 3d to highlight the persistence of the northeastward shift of the surface circulation after 2001.
 The above results for the three periods are summarized in Table 1. The number of drifters found in the GS box in each period fluctuates somewhat but the trend in the fraction of the GS drifters that end (loose their drogue or are retrieved) north of 50°N and 53°N increases triplefold over the three (four) periods. (Note that the ending may occur in the next period.) Another aspect is the apparent contracted state of the subtropical gyre in the first and second period which we address by counting drifters crossing the MAR and specifically crossing to east of 30°W. Again the percentage of drifters that are able to move east of 30°W more than doubles over time. If we count how far north these drifters reach (based on their northernmost latitude), the fraction of drifters in the 3rd (and 4th) period reaching latitudes 50°N and 53°N doubles compared to the second period. The fraction for the 1st period with a single drifter is unlikely to hold any significance. In the 3rd (and 4th) period about 30% of the GS drifters that reached 30W move northeastward across 50°N.
Table 1. Drifters Passing Through the Gulf Stream Box
Timing of drifter's end may occur in the next period.
Number of drifters reaching a given latitude means that sometime after the drifter left the GS box (having been the box in the stated period); it achieves the northernmost location but may or may not end south of that location.
 We can reverse track the drifters that reached the NE corner as in BT, however we chose a slightly different region, 20°W to 0°E, 53°N to 63°N, to emphasize the Rockall Trough which carries the most saline surface waters northward. The tracks for drifters that were in the box in the given time period and their tracks before entering and leaving this box (again the paths outside the box may fall into adjacent periods) are shown in Figures 4a–4c during the three periods. Again it may appear that there are too few tracks in the first period (Figure 4a) but in fact the northern region is well covered by (long living) drifter launches (Figure 2a) and tracks (shown later). Besides the obvious lack of any drifters south of 50°N entering the northern box, Figure 4a suggests that the subpolar drifters entering or launched in the northern Iceland Basin exited to westward, as one would expect the cyclonic subpolar circulation to operate. The middle period (Figure 4b) shows again that no drifters originated south of 45°N. The drifters from the southern Labrador Sea are embedded in the NAC which takes them nearly zonally across the MAR at the CGFZ into the Rockall Trough. Another branch of tracks originates in the central subpolar gyre (50°W–40°W, 55°N) and moves into the Iceland Basin. The southern boundary of the drifters that entered the northern box is well defined and almost parallel to 47°N. The middle period has numerous drifter launches between 45°N and 50°N, west of 30°W (Figure 2a), but south of the subpolar front, however only two end up in the northern box. Figure 4b, as well as Figures 4c and 4d, shows that the northern box is a destination for a few drifters launched in the eastern basin between 45°N and 50°N but east of 30°W, which likely represents movement of the eastern subtropical Atlantic surface waters.
 In the third period, 2001–2005, a “new” path passing through CGFZ toward the Rockall Trough has opened up, with drifters originating from the Gulf Stream waters (Figure 4c), south of 47°N. These drifters are embedded in the Gulf Stream Extension, some looping around the “Northwest Corner” at the mouth of the Labrador Sea and then zonally across to the Rockall Trough and some other tracks cross the MAR at more southern latitudes. The arrival of the drifters in the Rockall Trough originating south of 47°N is an indication of increased salt flux to the region. The salinity records from Faroe Shetland Channel and Rockall Trough [Holliday et al., 2008] and west European Basin [Bersch et al., 2007] show quite abrupt increases of salinity around 2002 after a moderate increase since 1994–1996. Besides the drifter tracks south of 47°N, there are also drifters originating elsewhere within subpolar and modified subpolar waters entering the NE corner as obvious from the Figure 4c. The Labrador Sea is not well covered by the drifter launches in the 3rd period, so we cannot assess surface water routes from the western subpolar region. Figure 4d as an additional 5-year period, 2003–2007 (and overlapping with the period 3) shows persistence of the subtropical path to the Rockall Trough with those drifters having an enhanced track coverage in the Gulf Stream box.
 A summary of the drifter movement passing through the northern box is given in Table 2. The fraction of drifters deployed south of 45°N (47°N) abruptly increases in the 3rd period to about 9% (14%) of the total number of drifters entering the northern box. (Again the drifters may have been deployed in the earlier period.) It is difficult to address the significance of the fraction because of the uneven coverage of the drifter deployments. However, on the basis of Figures 5 and 6 (to be discussed next) the overall density of the drifter tracks appears to be sufficient enough to warrant the change in tracks as a significant result.
Table 2. Drifters Passing Through the Northern Box
Drifters may have been launched in the earlier period, but were located in the northern box in the given period.
Total number of drifters found in the northern box
 To address the changes in the subtropical and subpolar drifter tracks and their representativeness, we plot the tracks for all drifters launched south and north of 45°N for all three periods. The sequence of these plots is shown in Figures 5 and 6 (the overlapping period 2003–2007 is omitted as it is indistinguishable from the 2001–2005 period in both cases). The Figure 5a shows that even in the early period the subtropical Atlantic was quite well covered by drifter tracks. In the first period the Bay of Biscay is particularly striking in its lack of drifters launched south of 45°N (Figure 3a). On the other hand, Figure 6a shows that the Bay of Biscay is filled with drifter tracks originating north of 45°N, hence Figures 6a and 5a are complementing fields filling the eastern North Atlantic with drifter tracks. Furthermore, comparison of Figure 5a with Figure 3a suggests that the strongly positive NAO with strongly anticyclonic curl severely limited the eastern and northeastern extent of the drifters originating in the Gulf Stream. During this time period the warm branches of the NAC were predominantly forced southeastward in the eastern basin.
 In the second period (Figure 5b) when NAO shifted sign in the 1996 winter, drifter track coverage expanded northward in the eastern basin. Since 2001 the subtropical tracks have significantly expanded northward and northeastward, as much as 5–10° latitude across the whole basin (Figure 5c).
 Regarding the representativeness of tracks leading to our chosen northern box (Figures 4a–4c), the seeming sparseness of Figure 4a can be contrasted with Figure 6a which shows the drifter tracks launched north of 45°N during 1991–1995. Figure 6a shows a dense coverage of tracks in all of the subpolar gyre (except the center of the Labrador Sea), including several tracks starting west of 40°W along the NAC path, but apparently only the drifters closest to the northern box manage to enter the box. The lack of drifters entering our northern box is surprising, since the subpolar latitudes are occupied by drifters living much longer than 200 days (Figure 2a). We can only speculate of the cause for this being the intense subpolar gyre circulation (as can be gathered from the geostrophic altimetric currents) competing with intense southward Ekman drift due to strong westerlies as defined by a positive NAO phase. The tracks in the two following periods (Figures 6b–6c) show increasing overlap with the subtropical waters. We anticipate that this overlap in the eastern basin is related to increased branching of the NAC, but the exact cause have to be left for a later study.
4. Eddy Kinetic Energy and Surface Current Changes From Altimetry
 Surface drifters are not evenly distributed in space or time, thus conclusions of major shifts in pathways and their timing need to be supported from other sources. The key characteristic of the major currents is their eddy shedding and meandering due to baroclinic instability leading to enhanced eddy kinetic energy (EKE) along the current paths. We invoke satellite altimetry to investigate the current fluctuations based on EKE. Altimetric EKE has been found to be an excellent way of identifying major current paths and their changes, for example clearly defining both the NAC, the Gulf Stream and intense jet-like structure of the Antarctic Circumpolar Current [Heywood et al., 1994; Menard, 1983; White and Heywood, 1995]. EKE time series can also lend support in pinpointing temporal transitions in the current paths. EKE is computed from geostrophic velocity anomalies relative to the whole period October 1992 to September 2006 (annual cycle is included). Instead of showing EKE for the three different periods we show the linear trend in EKE from 1992 to 2006 based on 1/3° AVISO data set in Figure 7a and the areas where the trend is 95% significant in Figure 7b. The areas of significance are patchy because the EKE is highly fluctuating (shown later by plots along 25°W and 52°N in Figure 8). In the trend Figure 7a, the EKE increase/decrease can be interpreted as a shift in the current path, although the intensity of the current (that would likely increase eddy production) can be equally possible if the current location did not change. Our main interest is not in changes in current intensity, but its location which we can then compare to the drifter tracks. We note first that, EKE has increased east of the MAR (east of 30°W) in several regions with values exceeding 30 cm2 s−2, and with significance 95% level, along the northward-northeastward shifted drifter pathway shown in Figure 3c. Specifically, the dense drifter track clusters in the eastern Atlantic, 42°N–55°N, 15°W–25°W, and 37°N–47°N, 25°W–35°W, overlay the increased EKE from altimetry. EKE has also increased in the Iceland Basin (Maury Channel) and in the northern end of Rockall Trough, i.e., in the two main branches of the NAC, and also in the Irminger Basin (west of the Reykjanes Ridge as in Figure 4d). Since these EKE changes are along the “traditional” routes of the NAC, we suggest that the increased energy have to relate to increased intensity of the currents, except in the case of the Irminger Sea, which marks the maximum in the subpolar pattern of downtrending altimetric sea-surface height (SSH) and subpolar gyre intensity between 1992 and the present (Hakkinen and Rhines , and the update of SSH EOF in Figure 11). It is also worth noting the regions where EKE has decreased: on the northern side of the Gulf Stream indicating a significant southward shift of the stream, in the Azores Current (between 30°N and 35°N). All of these shifts in EKE are significant at 95% level as seen from Figure 7b. This southward shift is also apparent in the long-term trend of altimetric SSH from 1992 to present.
 For a more detailed timing of the EKE changes we display latitude and longitude cross sections of EKE anomalies over time in Figures 8a–8b at 25°W and at 52°N. These two sections were chosen to describe the NAC activity in the eastern side of the MAR (25°W) and at the latitude of the northward turn of the NAC (52°N). There are two latitude ranges of high EKE activity in the 25°W section (Figure 8a). The high EKE anomaly at 32°N–34°N corresponds to the Azores Current which is well defined at this longitude (while being less distinct further east). Farther north, between 48°N–52°N, where NAC is crossing the MAR at several locations through Maxwell Fracture Zone (MFZ) at 48°N, Faraday Fracture Zone at 50°N and through Charlie Gibbs Fracture Zone (CGFZ) at 52°N, EKE increased significantly during 1999 and 2000 and has stayed elevated at least through 2005 [see also Bower and von Appen, 2008]. This EKE activity has weakened somewhat in 2006. The NAC crossing at around 40°N (Kurchatov Fracture Zone; KFZ) shows elevated EKE from 2000 to 2004. The northernmost elevated EKE zone at 57°N–60°N in Figure 8a lies in the central Iceland Basin and shows frequent EKE fluctuations but appears to have no trend. This band of high EKE is likely to represent the NAC path toward the Irminger Sea. EKE at 52°N section (Figure 8b) the NAC has not split into the Maury Channel and Rockall Trough branches. At this section NAC shows greatly increased EKE beginning in 2000, which continues to the end of the record. An intense and sustained EKE directly south of Rockall Trough occurs in 2004.
 The tendencies shown in the surface drifters and altimetric EKE are also evident in the ocean surface current analysis real time (OSCAR) product by NOAA where the altimetric geostrophic currents have been combined with Ekman drift. In Figures 9a–9b we show surface current differences between the period 2001–2005 and the first and the second periods, respectively. The first period was (from necessity) truncated to 3 years instead of 5 since TOPEX/Poseidon did not start until October 1992, nevertheless Figure 9a shows the developing strong anomaly in the midbasin along 40°N turning toward NE east of 30°W. The anomaly is more apparent and fully developed in Figure 9b using complete the 5-year periods. These NE current anomalies from the midbasin to Rockall Trough are enhanced in the same areas as the positive EKE trend in Figure 7a. The OSCAR product serves as an independent evidence of the NE current anomalies that are exhibited by surge of the subtropical origin surface drifters northeastward since 2001.
5. Potential Forcing of Surface Current Changes
 The large EKE changes in CGFZ, MFZ and KFZ took place about 4 years since the NAO underwent a major shift to a negative phase (in 1996), albeit lasting only 2 winters. Following this negative period the NAO returned briefly to strongly positive values in winter 2000, but NAO behavior since 1996 shows fluctuations of decreasing amplitude with a decline in amplitude. Hence the NAO changes do not directly explain the behavior of the surface drifters. Instead, we explore changes in wind stress curl as a potential forcing mechanism, because of the importance of wind stress curl to the basin-wide circulation. Addressing contributions from baroclinic effects driven locally or remotely are beyond the scope of this study, but Bersch et al.  show increased baroclinic transport in the upper (1000 m) ocean between Newfoundland Basin and the West European Basin along the A2 WOCE Hydrographic line.
 The behavior of the wind stress curl during the three periods is highlighted best if data from the winter months (January–March) are used. The average winter wind stress curl from NCEP/NCAR Reanalysis is displayed in Figures 10a–10c for the three periods. The most pronounced change is that the most recent years have had a positive curl extending from Cape Hatteras to the Rockall Trough. This feature is distinctly different from the two earlier periods where the zero-curl line crosses nearly zonally from Labrador to British Isles, allowing the negative wind stress curl to isolate the small area of positive curl off the eastern seaboard including the Gulf Stream region. In effect the increased meridional tilt of the wind stress curl zero line allows formation of a positive wind curl region connecting the Gulf Stream area and the Nordic Sill region. To show that this continuous positive-curl region is independent of resolution, higher-resolution (1/2°) QuikSCAT winds (available since 1999) support also the existence of the same pattern (Figure 10d). The increasing meridional tilt of the curl (as pictured in Figure 10), at least in idealistic numerical simulations of the North Atlantic, has been shown to allow the upper layer of the subtropical gyre to spread northeastward [Rhines and Schopp, 1991]. The northeastward expansion of the subtropical drifter tracks (launched south of 45°N) by 5–10° latitude (Figure 5c) can be interpreted to result from this expanding gyre effect.
 Observations show a reversal of subpolar and polar freshening since 1996 with a further abrupt increase in salinities since 2002 both in the Faroe-Shetland Channel and in Rockall Trough [Holliday et al., 2008]. Also water temperatures have abruptly reached new highs since 2002. These sudden large changes around 2002 cannot be explained by changes in the precipitation and evaporation, but require changes in circulation that brings more saline and warmer water masses to the region. To understand the recent changes, we analyzed surface drifter data for changes in saline water pathways. We also investigated altimetric EKE data for supporting evidence of the NAC changes to compensate for any biases in the drifter fields. We find that surface currents in the northern North Atlantic have undergone major changes, opening new pathways since 2001 from the western subtropics to the subpolar latitudes and toward Nordic Sills. Since 2001 the drifters originating south of 45°N–47°N and near the northern boundary of the subpolar front, but inside the warm water domain, have peeled off toward the Rockall Trough instead of turning toward Bay of Biscay. Hence the drifters are marking a subpolar front movement northwestward allowing expansion of subtropical, more saline waters northward. This drifter movement is consistent with Bersch et al. , who show the increased salt flux to the NE corner based on hydrography. Johnson and Gruber  report more saline waters existing down to 1500 m in 2003 at 20°W, between 40°N and Iceland which they attribute to an enhanced northward penetration of Mediterranean overflow waters. Supporting evidence of the northward flush of the subtropical surface waters is also recorded in the unexpectedly great increase of the sea surface partial pressure of CO2 between 1994/1995 and 2004/2005 which is associated with a very large (∼60%) reduction in the oceanic sink of atmospheric CO2 sink, especially in the eastern North Atlantic [Schuster and Watson, 2007]. The increased salinity of the waters flowing to the Nordic seas is likely to strengthen the meridional overturning and ultimately, possibly to oppose the widely predicted slowing of the global overturning induced by global warming in IPCC climate model simulations [Intergovernmental Panel on Climate Change, 2001].
 The pathways reported here have likely been opened and closed before but with the surface drifter observations since the late 1980s we have been able to observe these surface current changes. Opening of this pathway explains why the waters entering the Nordic seas have become much more saline since the mid-1990s [Holliday et al., 2008], and particularly after 2000. The earlier saline periods in the Rockall Trough occurred during the late 1950s lasting to the arrival of the Great Salinity Anomaly in the early 1970s and in the early 1980s [Belkin et al., 1998]. Prior to 1969 (1960–1969), also deeper waters at the Mediterranean overflow level were more saline which implies stronger northward transport of Mediterranean waters [Lozier and Stewart, 2008]. The presence of high-salinity deeper waters was established again at years 1998 and 2003 [Johnson and Gruber, 2007]. A longer time series exists from the Faroe-Shetland Channel which shows that even higher upper ocean salinities were encountered in the 1930s until about 1940 (before a 5-year gap), when a major warming occurred in the Atlantic subpolar zone [Delworth and Knutson, 2000; Belkin et al., 1998]. All of these epochs were characterized by a negative NAO index during or prior to the event. Considering this past history, the enhanced surface/upper ocean pathways from the western subtropics to the Nordic Sills are not stable, and there is no guarantee that they will remain equally active in the future years. In fact this pathway appears to be weakening based on the Ellett Line data (www.noc.soton.ac.uk/obe/PROJECTS/EEL/latestresults.php) which show slightly lower salinities in 2007 than in 2003 but higher than in the mid-1990s.
 Our investigation focused on the early 1990s with strong positive NAO years (1991–1995), transitioning to highly fluctuating NAO years with weak amplitude, divided into periods for 1996–2000 and 2001–2005. We invoke the changing pattern of wind stress curl as a possible forcing for the surface current change. A distinctive difference between the three periods emerges in the period 2001–2005. This last period has a positive wintertime curl extending from Cape Hatteras to Nordic Sills forming a continuous, tilted zero-curl line, unlike the two other periods when the zero-curl was almost zonal from Newfoundland to British Isles and had a gap near 50°N. Theoretical studies of wind driven circulation show that an increasingly meridional tilt of the wind stress curl zero line allows the northeastward expansion of the subtropical gyre [Rhines and Schopp, 1991]. This gyre effect was apparent from the northeastward expansion of the subtropical drifter tracks (launched south of 45°N) by 5–10° latitude.
 Altimetry can verify the drifter path changes from EKE fields which follow closely the locations of major currents such as the branches of NAC. From EKE patterns we can determine space-time behavior of the extended NAC, as in the steep increase in NAC EKE east of the Mid-Atlantic Ridge during year 2000. NAC EKE stayed elevated at least until 2005, consistent with the shift in the drifter tracks from eastward to northeastward. On the other hand, the sea surface height from altimetry indicates that as a whole, the upper subpolar gyre of the North Atlantic continues to weaken through 2005 in opposition to the surface current/EKE behavior. This weakening can be inferred from Figure 11 which displays the updated altimetric SSH principal component time series of Hakkinen and Rhines  using AVISO SSH data. In our view these two phenomena are not in disagreement, but instead point to changes in the vertical structure of the North Atlantic Ocean.
 S.H. gratefully acknowledges the support from NASA Headquarters Physical Oceanography Program for this work. P.B.R. is supported by NASA through the OSTST Science Team. We thank Denise Worthen for the invaluable technical assistance in data set analysis and graphics.