This study evaluates the presence of intermediate water from the Greenland Sea in the Iceland Basin deduced from the observed excess of the tracer sulphur hexafluoride (SF6), released in the central Greenland Sea in 1996. The large tracer release experiment has served a unique opportunity to follow the spread of Greenland Sea intermediate water to the adjacent basins of the Nordic Seas and to the areas bordering this region. In the present study, using data from May–June 2001, the released tracer was detected at the sill in the Faroe Bank Channel and at several locations in the Iceland Basin of the North Atlantic, just downstream the sill and southeast of Iceland. The estimated excess of the released tracer at the Icelandic slope combined with reported values of the volume flow at this location suggest an annual transport rate of approximately 1.4 kg excess SF6. The results suggest an upper transit time from the central Greenland Sea to the area southeast of Iceland of approximately 4 years.
 The formation and circulation of dense waters from the Arctic Mediterranean, i.e., the Nordic Seas and the Arctic Ocean, has been of great interest since the early days of the 20th century. One of the main issues has been the contribution of these waters to the dense overflow into the North Atlantic, where it is incorporated in the North Atlantic Deep Water (NADW). This water mass is important for the global ocean circulation and changes in its properties could have an effect on the circulation in all the large oceans [e.g., Dickson and Brown, 1994]. The exchange of denser water between the Nordic Seas and the North Atlantic is much limited by the shallow Greenland-Scotland Ridge. The two deepest parts of the ridge are the 630-m deep Denmark Strait and the 850-m deep Faroe Bank Channel (FBC), which are the main sites of the overflows into the North Atlantic (Figure 1). In addition there are some overflows of denser water across the Iceland-Faroe Ridge and the Wyville-Thomson Ridge [e.g., Hansen and Østerhus, 2000].
 To be able to predict the sensitivity of the overflows to climate forcing it is important to deduce the main sources to the overflow waters, or more correctly, to identify the key formation areas where the source waters attain their hydro-chemical properties. This has recently been done for the overflow through the Denmark Strait [Jeansson et al., 2008; Tanhua et al., 2008] and a similar investigation of the overflow through the Faroe Bank Channel is planned. One site in the Nordic Seas that has attained much focus is the Greenland Sea, mostly due to the convective formation of very dense water. The importance of this area as a supplier to the dense overflows from the Nordic Seas into the North Atlantic have been debated during the last decades, and there is still no consensus in the matter, especially for the overflow through the Denmark Strait, recently high-lighted by Dickson et al.  and Tanhua et al. . Even though traditional Θ-S analysis has shown the presence of Arctic Intermediate Water in different parts of the Nordic Seas [e.g., Blindheim, 1990; Rudels et al., 2005] the large tracer release experiment in the summer of 1996 has certainly served an even stronger tool to identify this water mass. It commenced with the release of 320 kg of the tracer sulphur hexafluoride (SF6) in the central Greenland Sea, and the tracer was injected close to the density surface σ0 = 28.0492 kg m−3, which corresponded to approximately 300 m [Watson et al., 1999]. The release and the spreading of the tracer for the period 1996 to 2002 have recently been described by Messias et al. . The released tracer tagged the intermediate water in the central Greenland Sea, the Greenland Sea Arctic Intermediate Water (GSAIW), which is formed by winter convection in the Greenland Sea Gyre [e.g., Karstensen et al., 2005]. Studies dealing with the spreading of the released tracer have shown that this water mass has contributed to the Denmark Strait Overflow Water (DSOW) [Olsson et al., 2005b] as well as the Iceland-Scotland Overflow Water (ISOW) [Olsson et al., 2005a]. In addition, the tracer has been traced all the way to the Labrador Sea [Tanhua et al., 2005] and has been observed entering the Arctic Ocean [Marnela et al., 2008]. SF6 has a transient record in the atmosphere (Figure 2), starting to be produced in the mid-1950s [Maiss and Brenninkmeijer, 1998] and has a great potential to be used as a transient tracer in the oceans, in the same way as the chlorofluorocarbons (CFCs) has been for several decades [e.g., Law and Watson, 2001; Tanhua et al., 2004]. Since the production and release of the CFCs have been restricted by the Montreal Protocol their atmospheric evolution has slowed down considerably, and CFC-11 has even declined, which makes these compounds less useful as transient tracers. SF6, however, still shows a linear atmospheric increase, resulting in an increase also in the ratio between SF6 and the CFCs (Figure 2).
 We present the tracer concentrations in units of mixing ratios, which is calculated from the measured concentration and the solubility of the tracers, hereby removing the dependency on temperature and salinity [Warner and Weiss, 1985; Bullister et al., 2002]. We will identify the released tracer, and hence any contribution of the GSAIW, from elevated levels of the SF6/CFC-11 ratio compared to the atmospheric history (Figure 2). The ratio between SF6 and CFC-11 or CFC-12 is a stronger tool for identification of excess tracer than the measured tracer concentrations alone, especially when the tracer is markedly diluted. This holds since only SF6 was released, and no CFCs, and thus the ratio for the tagged water should be higher than that which can be found in the atmospheric record. The atmospheric SF6/CFC-11 ratio in the Northern hemisphere in 2001 was just below 19 × 10−3 and values above this level should then reflect a signal of the released tracer, which we will refer to as excess SF6. In this paper we will present the first evidence of released SF6, and hence of Greenland Sea water, downstream of the overflows across the Iceland-Scotland Ridge (Figure 1). For this purpose we have adopted the method used by Tanhua et al.  for the Denmark Strait overflow. We will identify excess SF6 from elevated levels of the SF6/CFC-11 ratio and estimate the transport of the excess using reported values of volume transports in the area.
2. Methods and Data
 The presented data was collected between May 27 and June 19 in 2001 during a cruise with R/V Håkon Mosby as a part of the EU project “Tracer and Circulation in the Nordic Seas Region” (TRACTOR). The cruise covered a large part of the Nordic Seas and water samples were collected for several chemical parameters, but this study will focus on SF6 and CFCs, together with hydrography. CTD measurements were performed using a SEABIRD with a General Oceanics Rosette sampler with 12 Niskin bottles.
 CFC-11 and CFC-12 were determined in water samples collected throughout the cruise. The analysis is based on purge-and-trap work-up of the water samples followed by gas chromatographic separation and electron capture detection. A description of the analytical technique is given by Fogelqvist . The analytical precision, given as the standard deviation for multiple samples taken at the same depth, is about 2% for both CFC-11 and CFC-12. Standardization was achieved by calibration gas prepared at Brookhaven National Laboratory [Happell and Wallace, 1997] and cross-calibrated against gas prepared at Scripps Institute of Oceanography. The standard gases were calibrated against the SIO-93 scale.
 The SF6 was analyzed using a purge-and-trap method followed by gas chromatography and electron capture detection, closely following the method described by Law et al. . The 1-σ precision estimated from the average of the standard deviation was better than 0.04 fM or 1.5% of background surface values [Messias et al., 2008].
 A comparison of the CFC data showed that there seemed to be an issue with CFC-12 during this cruise, especially in the surface layers; the CFC-12 data had generally a larger scatter and the mean surface saturation was ∼120%, whereas CFC-11 was only slightly supersaturated and appeared reliable from all aspects. From this we choose to use CFC-11 instead of CFC-12 throughout the work.
3. Estimate of Excess SF6 in the Iceland Basin
 We will here focus on the incorporation of the released tracer, and hence water originating from the Greenland Sea, in the ISOW. We will define the ISOW as water denser than σ0 > 27.8 kg m−3, following the definition by several authors [e.g., Saunders, 1996; Hansen and Østerhus, 2000]. For an estimate of the amount of excess SF6 in the ISOW samples we adopted the method used by Tanhua et al.  for the Denmark Strait overflow, but used the SF6/CFC-11 ratio vs. SF6 instead of SF6/CFC-12 vs. CFC-12. The approach is to make a fit of all non-overflow data (hence with σ0 < 27.8 kg m−3) and use this as a background level. Consistent with Tanhua et al.  we used a quadratic fit of the data. Then all overflow samples are evaluated with the fitting function, and it is assumed that all data points above the upper error limit (50% confident interval) of the fit had a contribution of excess SF6. Hence, the calculated excess corresponds to the difference between the measured ratio and the upper error limit [Tanhua et al., 2005]. The estimated excess SF6 is then calculated back to gravimetric units, using the solubility equations [Warner and Weiss, 1985; Bullister et al., 2002], for the mass transport calculations. We estimated the amount of excess SF6 in the ISOW, both in the FBC and on the three sections in the Iceland Basin. We will start to present the result from the sill in the FBC (Section I). As is obvious from Figure 3 there were clearly elevated ratio values in almost all of the samples within the ISOW density range, and the mean excess (and its standard deviation) amounted to 0.41 ± 0.17 fmol kg−1. In a previous study Olsson et al. [2005a] estimated the excess of the released tracer in the FBC, between summer 2000 and early 2002, to 0.9 ± 0.1 fmol kg−1, which then is approximately twice the amount we present here. However, there are several differences in the approach of the two estimates, including different definitions of the overflow water; Olsson et al. [2005a] used the more strict definition as water colder than 0.3°C, which roughly equals water denser than σ0 > 28.0 kg m−3 [Mauritzen et al., 2005]. We tested this by performing a new fit of only FBC data and this stricter ISOW definition, but arrived at approximately the same value as for the less strict definition. Even when disregarding the error bounds and hence assuming that all ratio values higher than the fitting line, and not the upper error limit, equaled excess of the released tracer (cf. Figure 3) the mean excess (0.50 ± 0.079) was clearly lower than the estimate of Olsson et al. [2005a]. We here would like to stress that the method we adopted from Tanhua et al.  gives a statistically strong evidence of the released tracer, but is assumed to give a low limit of the excess and the more diluted the tracer is the larger tendency has the method to underestimate the true number.
 Three sections were sampled “downstream” of the FBC, in the Iceland Basin (Section II–IV; see Figure 1). However, of these, only the most southern section (Section IV) captured a clear amount of the overflow water, while a very limited amount of ISOW samples were collected at Section II (2 samples) and Section III (4 samples); see Figure 3. The mean excess (and standard deviation) at these sections were 0.078 ± 0.24 fmol kg−1 at Section II, located approximately 125 km from the sill, and, 0.037 ± 0.15 fmol kg−1 Section III, situated at the southeastern continental slope of Iceland. These low numbers, compared with the excess in the FBC, could most likely be explained by a strong dilution of the excess via mixing with mostly excess-free water. This is supported by the findings of Mauritzen et al.  who observed the largest mixing of the overflow where the plume starts to descend into the deeper parts of the Iceland Basin, approximately 100 km downstream of the sill in the FBC.
 The most southern section (Section IV), however, showed a rather large amount of samples within the ISOW density range, and most of these had ratio values well above the non-ISOW samples (Figure 3), and the mean excess was 0.073 ± 0.11 fmol kg−1. The section extends quite far from the slope, and thus it might be more interesting to just look at the actual slope stations. The re-calculated excess at these stations were 0.094 ± 0.11 fmol kg−1.
 The most plausible explanation for the higher excess at this location further downstream of the ISOW, compared with Section II and III, is that the sampling at the “upstream” sections did not catch most of the overflow plume. This suggestion is strengthened by the low number of overflow samples at Section II and III compared with Section IV. Moreover, we assume that Section IV covered the vast part of the overflow plume, consisting of the different sources to the ISOW, hence both from the FBC and the overflow across the Iceland-Faroe Ridge [e.g., Meincke, 1983; Hansen and Østerhus, 2000]. Section III, on the other hand, is situated higher up on the Iceland slope, and is more likely to cover part of the overflow, but not the core of the plume.
4. Discussion and Conclusions
 In an attempt to estimate the transport of excess SF6 in the FBC and at Section IV we combine our reported excess SF6 values at these sections with reported values of the volume transport. For the FBC we adopt the value from Hansen and Østerhus  of 1.9 ± 0.3 Sv (1 Sv = 106 m3 s−1), which is the average flux between 1995 and 2005, of water denser than σ0 > 27.8 kg m−3. This results in a transport of 3.5 ± 0.5 kg excess SF6 yr−1, which should be compared with the number of Olsson et al. [2005a] of ∼5 kg excess SF6 yr−1 using a flux of 1.2 Sv (based on the mean volume flux of water colder than 0.3°C in 2000 [Hansen et al., 2001]). Our estimated transport of excess SF6 is hence ≥20% lower than what Olsson et al. [2005a] reported, but as discussed above this difference originates in the different approaches.
 For an estimate of the transport at Section IV we used the number reported by Saunders  for the flux of dense water (σ0 > 27.8 kg m−3) southeast of Iceland of 3.2 ± 0.5 Sv. This value comes from data taken in 1990–1991, at depths 1000–2300 m using current meters located roughly at the position of our Section IV. Since we here want to focus on the transport along the Icelandic slope we omitted the two most eastern stations in Section IV (see Figure 1). Combining the flux value from Saunders  with our estimated amount of excess SF6 at the slope on Section IV (∼0.09 fmol kg−1), results in a transport of 1.4 ± 0.2 kg excess SF6 yr−1.
 The released tracer was first detected in the FBC in winter 1998/1999, approximately 2.5 years after the release in the Greenland Sea [Olsson et al., 2005a]. The present study is the first observation of the released tracer south of the Iceland-Scotland Ridge. This gives an upper transit time of the tracer from the central Greenland Sea to the location southeast of Iceland of approximately four years. However, we here want to stress that the transit time was most likely shorter since the spread of the released tracer into the Iceland Basin has not been evaluated before.
 The presented data give clear evidence that the released tracer had entered the Iceland Basin of the North Atlantic, something that has not been reported before. The highest estimated excess in the overflow water (σ0 > 27.8 kg m−3) along the continental slope southeast of Iceland was in the order of ∼0.1 fmol kg−1 with an annual transport of 1.4 kg. As mentioned above the tracer-tagged water has previously been traced from the Denmark Strait, within the DSOW, and into the Labrador Sea [Tanhua et al., 2005]. These two sources, the DSOW and the ISOW, are important contributors to the North Atlantic Deep Water; it now seems clear that the dense GSAIW contributes to both these overflows.
 We wish to thank the Captain and Crew on M/S Håkon Mosby for excellent support during the cruise. We also acknowledge Johanna Balle for her contribution to the analysis of the SF6 samples and Toste Tanhua for kindly sharing Matlab scripts and time for discussions. The EU FP 5 project TRACTOR (Tracer and circulation in the Nordic Seas Region), contract EVK2-2000-00080, was the main funding source for the field work part of these studies although the initiation of the release experiment was financed inside EU/MAST III project ESOP-2. Comments from two anonymous reviewers clearly improved the manuscript. This is publication A 224 from the Bjerknes Centre for Climate Research.