Transient tracer distributions in the Fram Strait in 2012 and inferred anthropogenic carbon content and transport

. The storage of anthropogenic carbon in the ocean’s interior is an important process which modulates the increasing carbon dioxide concentrations in the atmosphere. The polar regions are expected to be net sinks for anthropogenic carbon. Transport estimates of dissolved inorganic 5 carbon and the anthropogenic offset can thus provide information about the magnitude of the corresponding storage processes. Here we present a transient tracer, dissolved inorganic carbon (DIC) and total alkalinity (TA) data set along 78 ◦ 50 (cid:48) N 10 sampled in the Fram Strait in 2012. A theory on tracer relationships is introduced which allows for an application of the Inverse Gaussian - Transit Time Distribution (IG-TTD) at high latitudes and the estimation of anthropogenic carbon concentrations. Mean current velocity measurements along 15 the same section from 2002-2010 were used to estimate the net ﬂux of DIC and anthropogenic carbon by the boundary currents through the Fram Strait above 840m. The new theory explains the differences between the theoretical (IG-TTD based) tracer age relationship and the speciﬁc 20 tracer age relationship of the ﬁeld data by saturation effects during water mass formation and / or the deliberate release experiment of SF 6 in the Greenland Sea in 1996 rather than by different mixing or ventilation processes. Based on this assumption, a maximum SF 6 excess of 0 . 5 − 0 . 8 fmolkg − 1 25 was determined in the Fram Strait at intermediate depths ( 500 m - 1600 m ). The anthropogenic carbon concentrations are 50 − 55 µmolkg − 1 in the Atlantic Water / Recirculating Atlantic Water, 40 − 45 µmolkg − 1 in the Polar Surface Water / warm Polar Surface Water and between 10 − 35 µmolkg − 1 30 in the deeper water layers, with lowest concentrations in the bottom layer. The net ﬂuxes through the Fram Strait indicate a net outﬂow of ∼ 0 . 4 PgC yr − 1 DIC and ∼ 0 . 01 PgC yr − 1 anthropogenic carbon from the Arctic Ocean into the North Atlantic, albeit with high uncertainties. 35

Abstract. The storage of anthropogenic carbon in the ocean's interior is an important process which modulates the increasing carbon dioxide concentrations in the atmosphere. The polar regions are expected to be net sinks for anthropogenic carbon. Transport estimates of dissolved inorganic 5 carbon and the anthropogenic offset can thus provide information about the magnitude of the corresponding storage processes.
Here we present a transient tracer, dissolved inorganic carbon (DIC) and total alkalinity (TA) data set along 78 • 50 N 10 sampled in the Fram Strait in 2012. A theory on tracer relationships is introduced which allows for an application of the Inverse Gaussian -Transit Time Distribution (IG-TTD) at high latitudes and the estimation of anthropogenic carbon concentrations. Mean current velocity measurements along 15 the same section from 2002-2010 were used to estimate the net flux of DIC and anthropogenic carbon by the boundary currents through the Fram Strait above 840m. The new theory explains the differences between the theoretical (IG-TTD based) tracer age relationship and the specific 20 tracer age relationship of the field data by saturation effects during water mass formation and / or the deliberate release experiment of SF 6 in the Greenland Sea in 1996 rather than by different mixing or ventilation processes. Based on this assumption, a maximum SF 6 excess of 0.5 − 0.8 f mol kg −1 Changes in the Arctic during the last decades stand in mutual relationship with changes in the adjacent ocean areas such as the Nordic Seas, the Atlantic and the Pacific Ocean. 40 The temperature of the Atlantic Water flowing into the Arctic Ocean through the Fram Strait has warmed since 1997 , which thus increased the heat flux into the Arctic. This has a significant influence on the perennial sea ice thickness and volume and thus on the 45 fresh water budget (Polyakov et al., 2005;Stroeve et al., 2008;Kwok et al., 2009;Kurtz et al., 2011). The exchange and transport of heat, salt and fresh water through the major gateways like the Fram Strait, Barents Sea Opening, Canadian Archipelago and Bering Strait are also directly 50 related to changes in ventilation of the adjacent ocean areas (Wadley and Bigg, 2002;Vellinga et al., 2008;Rudels et al., 2012). The ventilation processes of the Arctic Ocean have a significant impact on the anthropogenic carbon storage in the world ocean (Tanhua et al., 2008). Studying the 55 fluxes of anthropogenic carbon through the major gateways contributes to understand the integrated magnitude of such ocean-atmosphere interactions. It additionally provides information of a changing environment in the Arctic Mediterranean. The required flux data of the prevailing water masses, 60 i.e. current velocity fields, are obtained by time series of long-term maintained mooring arrays in the different gateways. The Fram Strait is the deepest gateway to the Arctic Ocean with highest volume fluxes equatorwards and polewards. A well-established cross-section mooring array is lo- Schauer et al., 2008) which provided the basis for heat transport estimates in the past (Fahrbach et al., 2001;Schauer et al., 2004Schauer et al., , 2008. However, the current data interpretation and analysis of this moor-70 ing array is complicated due to a recirculation pattern in the Fram Strait (Schauer and Beszczynska-Möller, 2009;Rudels et al., 2008;Marnela et al., 2013;de Steur et al., 2014) and strong mesoscale eddy activity (von Appen et al., 2015a). The spatial and temporal volume flux variability and the in-75 sufficient instrument coverage in the deeper water layers, i.e. below the West Spitsbergen Current (WSC) and East Greenland Current (EGC), lead to high uncertainties of the net flux through the Fram Strait. Hence, it is the most relevant but also the most challenging gateway with respect to transport 80 budgets in the Arctic Mediterranean. Estimating an anthropogenic carbon budget presupposes the knowledge of the concentration ratio between the natural dissolved inorganic carbon (DIC) and the anthropogenic part (C ant ) in the water column. An estimate of DIC trans-

Tracer and carbon data
A data set of CFC-12, SF 6 , DIC and TA was obtained during the ARK-XXVII/1 expedition from 14 June to 15 July 2012 110 from Bremerhaven, Germany to Longyearbyen, Svalbard on the German research vessel Polarstern (Beszczynska-Möller, 2013). Figure 1 shows the stations of the zonal section along 78 • 50 N, where measurements of CFC-12, SF 6 , DIC, and TA were conducted. The meridional section along the fast 115 ice edge was only sampled for CFC-12 and SF 6 and shows no differences in the horizontal tracer distributions compared to the corresponding longitude range of the zonal section. Therefore we have only used the zonal section for the following analysis.

120
Water samples for the determination of the transient tracers CFC-12 and SF 6 were taken with 250 ml glass syringes and directly measured on board, using a purge and trap GC-ECD system similar to Law et al. (1994) and Bullister and Wisegarver (2008). The measurement system is identical to 125 the "PT3" system described in Stöven and Tanhua (2014) except the cooling system and column composition. The trap consisted of a 1/16" column packed with 70 cm Heysep D and cooled to −70 • C during the purge process using a Dewar filled with a thin layer of liquid nitrogen. The 1/8" precol-130 umn was packed with 30 cm Porasil C and 60 cm Molsieve 5Å and the 1/8" main column with 180 cm Carbograph 1AC. Due to malfunctioning of the Electron Capture Detector (ECD) of the measurement system, the samples of 6 stations (between station 15 and 53) were taken with 300 ml 135 glass ampules and flame sealed for later onshore measurements at GEOMAR. The onshore measurement procedure is described in Stöven and Tanhua (2014). The precision for the onshore measurements is ±4.4 %/0.09 f mol kg −1 for SF 6 and ±1.9 %/0.09 pmol kg −1 for CFC-12. The precision for 140 onboard measurements is ±0.5 %/0.02 f mol kg −1 for SF 6 and ±0.6 %/0.02 pmol kg −1 for CFC-12. Water samples for DIC and total alkalinity (TA) were taken with 500 ml glass bottles and poisoned with 100 µl of a saturated mercuric chloride solution to prevent biologi-145 cal activities during storage time. The sampling procedure was carried out according to Dickson et al. (2007). The measurements of DIC and TA were performed onshore at the GEOMAR, using a coulometric measurement system (SOMMA) for DIC (Johnson et al., 1993(Johnson et al., , 1998) and a 150 potentiometric titration (VINDTA) for TA (Mintrop et al., 2000). The precision is ±0.05 %/1.1 µmol kg −1 for DIC and ±0.08 %/1.7 µmol kg −1 for TA. The data of all obtained chemical parameters will be available at the Carbon Dioxide Information Analysis Center (CDIAC) by the be-155 ginning of 2016. The physical oceanographic data (temperature, salinity, and pressure) from this cruise can be found at Beszczynska-Möller and Wisotzki (2012).

Water transport data
An array of moorings across the deep Fram Strait from 9 • E to 160 7 • W has been maintained since 1997 by the Alfred Wegener Institute and the Norwegian Polar Institute. Since 2002, it has contained 17 moorings at 78 • 50 N. Here we use the gridded data from the array from summer 2002 to summer 2010 as described in . The more re-165 cent data has either not been recovered yet or the processing is still in progress. The moorings contained temperature and velocity sensors at five standard depths: 75 m, 250 m, 750 m, 1500 m, and 10 m above the bottom. The hourly measurements were averaged to monthly values and then grid-ded onto a regular 5 m vertical by 1000 m horizontal grid using optimal interpolation. While the Atlantic Water in Fram Strait has warmed since 1997,  showed that there is a strong seasonal cycle in the Atlantic Water transport through the Fram Strait, but that there 175 is no statistically significant interannual trend between 1997 and 2010 in the volume transport. We consider the long term average volume flux of the following water masses: Atlantic Water advected in the West Spitsbergen Current defined as longitude ≥ 5 • E and depth ≤ 840 m; Polar Water flowing 180 southward in the East Greenland Current defined as mean temperature ≤ 1 • C and depth ≤ 400 m; and finally Recirculating and Arctic Atlantic Water which is both due to the recirculation of Atlantic Water in the Fram Strait (de Steur et al., 2014) and the long loop of Atlantic Water through 185 the Arctic Ocean (Karcher et al., 2012), defined as longitude ≤ 1 • E and depth ≤ 840 m, not Polar Water. The estimate of the volume transport across the Fram Strait below 840 m from the moorings is more complicated. The method of Beszczynska-Möller et al. (2012) which was developed 190 to study the fluxes in the West Spitsbergen Current predicts a net southward transport of 3.2 Sv below 840 m. This is unrealistic given that there are no connections between the Nordic Seas and the Arctic Ocean below the sill depth of the Greenland-Scotland Ridge (840 m) other than the Fram

TTD method 205
A transit time distribution (TTD) model (Eq. 1) describes the propagation of a boundary condition into the interior of the ocean and is based on the Green's function (Hall and Plumb, 1994).
Here, c(t s , r) is the specific tracer concentration at year t s and location r, c 0 (t s − t) the boundary condition described by the tracer concentration at source year t s − t and G(t) the transit time distribution of the tracer. The exponential term corrects for the decay rate of radioactive transient tracers.

215
Equation 2 provides a possible solution of the TTD model, based on a steady and one-dimensional advective velocity and diffusion gradient (Waugh et al., 2003).
It is known as the Inverse-Gaussian transit time distribu-220 tion (IG-TTD) where G(t) is defined by the width of the distribution (∆), the mean age (Γ) and the time range (t).
One can define a ∆/Γ ratio of the IG-TTD which represents the proportion between the advective and diffusive properties of the mixing processes as included in the TTD. The lower 225 the ∆/Γ ratio, the higher is the advective share. A ∆/Γ ratio of 1.0 is the commonly applied ratio (unity ratio) at several tracer surveys (e.g. Waugh et al., 2004Waugh et al., , 2006Tanhua et al., 2008;Schneider et al., 2010;Huhn et al., 2013;Schneider et al., 2014).

230
Another approach is based on a linear combination of two IG-TTDs which can be used to describe more complex ventilation patterns (Eq. 3) (Waugh et al., 2002). The variables of this model are ∆ 1,2 and Γ 1,2 of the two IG-TTDs and α which describes the ratio between both distributions.

235
The main problem of applying this method is that five free parameters need to be determined. Hence, this model combination can be constrained with five transient tracers with sufficiently different input functions. Alternatively, predefined parameters can be used (Stöven and Tanhua,240 2014).
Note that the use of CFC-12 as transient tracer is limited 245 to concentrations below the recent atmospheric level since the production of CFC-12 was phased out in the early 1990s so that the depletion rate exceeds the emission rate since the early 2000s. This causes indistinct time information of CFC-12 since one concentration describes two dates in the atmo-250 spheric history. To this end, the use of CFC-12 is restricted to water masses with concentrations below the current atmospheric concentration limit. The emission rate of SF 6 still exceeds the depletion rate so that the atmospheric concentration is still increasing. SF 6 thus provides distinct age information 255 of water masses over the complete concentration range.

Anthropogenic carbon and transport estimates
The IG-TTD model can be used to estimate the total amount of anthropogenic carbon in the water column (Waugh et al., 2004). For this purpose it is assumed that the anthropogenic 260 carbon behaves like an inert passive tracer, i.e. similar to a transient tracer. Then applying equation 1, the concentration of anthropogenic carbon in the interior ocean (C ant (t s )) is given by equation 4.
C ant,0 is the boundary condition of anthropogenic carbon at year t s − t and G(r, t) the distribution function (see equation 1). The historic boundary conditions are described by the differences between the preindustrial and modern DIC concentrations at the ocean surface. These anthropogenic off-270 sets can be calculated by applying the modern (elevated) partial pressures of CO 2 and then subtracting the corresponding value of the preindustrial partial pressure. In each case, the preformed alkalinity was used as second parameter to determine the specific DIC concentrations (calculated using the Here we assumed a constant pCO 2,water saturation in the surface. Since exact saturations are not well constrained, we present sensitivity calculations of different saturation states / disequilibria (see section 3.6 below). The atmospheric his-280 tory of pCO 2,atm is taken from Tans and Keeling (2015). The preformed alkalinity was determined by using the alkalinity / salinity relationship of MacGilchrist et al. (2014). This relationship is based on surface alkalinity and salinity measurements in the the Fram Strait which were corrected for sea-ice 285 melt and formation processes. The time-dependent boundary conditions (C ant,0 ) and Eq. 4 can then be used to calculate anthropogenic carbon concentrations (C ant (t s )) and the corresponding mean age. Finally, the mean age of Eq. 4 can be set in relation to the 290 transient tracer based mean age of the water and allows for back-calculating C ant (t s ), i.e. it provides the link between the tracer concentration and the anthropogenic carbon concentration. We then proceed to estimate transports of anthropogenic car-295 bon through the Fram Strait. Transports are the product of concentrations times velocities integrated over an area. We assume that the trace gas concentrations change relatively slowly between years and that there are no significant seasonal changes. Hence, we can take the concentrations from

315
To highlight the different transient tracer characteristics we defined the water mass type of each sample by using the water mass properties suggested by Rudels et al. (2000Rudels et al. ( , 2005 and the salinity and temperature data of this cruise from Beszczynska-Möller and Wisotzki (2012).

320
Water masses of the Arctic Ocean are the Polar Surface Water (PSW) which is the cold and low saline surface and halocline water; the warm Polar Surface Water, defined by a potential temperature (Θ) > 0, which comprises sea ice melt water due to interaction with warm Atlantic Water and due to solar ra-  Figure 3 shows the partial pressure of CFC-12 and SF 6 at the zonal section across the Fram Strait. Both tracers have significant concentrations through the entire water column 365 and show a similar distribution pattern. The Atlantic Water shows a relatively homogeneous distribution of both tracers with CFC-12 partial pressures > 450 ppt and SF 6 > 6 ppt. The Polar Surface Water at the shelf region shows a more distinct structure with partial pressures between 4 − 8 ppt of 370 SF 6 and 410−560 ppt of CFC-12. The smaller concentration gradient of CFC-12 in the surface compared to SF 6 is related to the recently decreasing atmospheric concentration of CFC-12, which causes only slightly varying boundary conditions at the air-sea interface (see section 2.3). The high-tracer 375 concentration layer of the Polar Surface Water extends further eastwards as overlaying tongue of the Atlantic Water between 2−6 • W. The intermediate layer between 500−1600 m is characterized by a clear tracer minimum along the continental slope of Greenland with partial pressures between

Transient tracers and the IG-TTD
The IG-TTD can be numerically constrained using transient tracer couples, CFC-12 and SF 6 in our case, which provides 420 information about the mean age and the parameters of the IG-TTD (Waugh et al., 2002;Sonnerup et al., 2013;Stöven and Tanhua, 2014). The method of validity areas, introduced in Stöven et al. (2015), is used to determine the applicability of the tracer couple. For this purpose, the tracer age is calcu-425 lated from the transient tracer concentrations (Waugh et al., 2003) which provides the tracer age relationship of the tracer couple. Figure 5 shows the tracer age relationship of our field data (colored by water mass) in relation to the range of theoretical tracer age relationships of the IG-TTD, i.e. for ∆/Γ 430 ratios between 0.1 − 1.8, which describe the range from advectively dominated to diffusively dominated water masses (grey shaded area). The black line in Fig. 5  ter. Note that the Arctic Atlantic Water and upper Polar Deep Water merge with the upper set for SF 6 tracer age larger than about 25 years. However, the upper set does not correspond to the unity ratio and, moreover, it is outside the validity area of the IG-TTD. Water masses related to the lower set can be 450 applied to the IG-TTD with tendencies towards higher ∆/Γ ratios (> 1.0) since the data is clearly above the black line indicating a dominance of diffusive processes. Another approach is provided by the linear combination of two IG-TTDs. Since we only have the data of two transient 455 tracers, we used the same predefined parameters as described in Stöven and Tanhua (2014) which includes one more diffusive water parcel (∆ 1 /Γ 1 = 1.4) and one very advective water parcel (∆ 2 /Γ 2 = 0.6). Figure 6 shows, similar to for different α of 0.2, 0.5 and 0.8. Although this combination describes several scenarios from highly advective to diffusive mixing of two distributions, it can be seen that most of the observed data points are still outside the validity area. Thus, the tracer age relationship between CFC-12 and SF 6 can neither 465 be described by the IG-TTD nor a linear combination of two IG-TTDs. Based on the raw field data, and on assumptions implemented in the IG-TTD (like constant mixing processes along the flow pathway as well as constant saturation of the gases at the 470 surface before entering deeper layers), the IG-TTD or lin-ear combinations of the IG-TTD can only partly describe the ventilation pattern of water masses in the Fram Strait. Nevertheless, by comparing the shape of the two field data sets with the shape of the black line in Fig. 5, it is noted that both sets show similar characteristics as the unity ratio or, generally, as IG-TTD based tracer age relationships. This opens up the possibility to use the IG-TTD the other way around, i.e. to assume a fixed ∆/Γ ratio to determine the deviation of transient tracer concentrations rather than using the tran-480 sient tracer concentration to determine the ∆/Γ ratio. Since several publications found the unity ratio of ∆/Γ = 1.0 to be valid in large parts of the ocean, we assumed that this is also true for water masses in the Fram Strait. Figure 7 shows the mean tracer age relationship of the upper set (red line) and the tracer age relationship of the unity ratio (black line / same as in Fig. 5). The offset of the field data related to the unity ratio suggests an undersaturation of CFC-12 and / or a supersaturation of SF 6 (see black box in Fig. 7). This uncommon coexistence of under-and supersaturated transient 490 tracers is discussed in the following section.

Saturations and excess SF 6
The surface saturations of transient tracers are influenced by sea surface temperature and salinity, ice coverage, wind speed, bubble effects, atmospheric growth rate of the tracer 495 and the boundary dwell time of the water parcel (i.e. the time the water parcel is in contact with the atmosphere). However, the saturation state of transient tracers at the air-sea interface before, during and after water mass formation is rarely known, since water mass formation generally occurs in win-500 ter at high latitudes, which renders it almost impossible to obtain measurements. Shao et al. (2013) provide modelled data of monthly surface saturations of CFC-11, CFC-12 and SF 6 from 1936 to 2010 on a global scale. This model output can be used to estimate the tracer saturation ratio of dif-505 ferent water masses by using the surface saturation of the specific formation area and yearly formation period. The formation types and areas are notably different for water masses that occur in the Fram Strait. The model output shows high variabilities in surface saturations at different formation sites, 510 namely the Greenland Sea, the Arctic shelf regions and the Arctic open water (Fig. 8). In contrast, the tracer age relationships of the two sets in Fig. 5 indicate relatively similar deviations in saturation. The complex boundary conditions in the Arctic, e.g. possible gas exchange through ice cover 515 and the changing extent of the ice cover, might bias the results of the saturation model. Therefore, we only used the surface saturation of the Greenland Sea (Area 1 in Fig. 8) which agrees with the findings of Tanhua et al. (2008) who used available field data to investigate historic tracer satu-520 rations. The IG-TTD based mean age provides the link between the observed tracer concentrations and the corresponding time-dependent saturation. Therefore, the saturation corrections were applied to the atmospheric history (boundary conditions) of each tracer. These new boundary conditions 525 are then applied to the measured tracer concentrations and the IG-TTD which then yields a saturation-corrected mean age. This mean age in turn can then be used to back-calculate the saturation-corrected tracer concentrations using the original (uncorrected) boundary conditions.

530
The SF 6 excess is estimated using the corrected CFC-12 concentrations and the IG-TTD (∆/Γ = 1.0) to calculate theoretical SF 6 concentrations of the water parcel, i.e. backcalculated SF 6 concentrations. The difference between the theoretical SF 6 concentration and the measured SF 6 concen-535 tration denotes the SF 6 excess in the water. Note that this SF 6 excess is based on the assumption that the IG-TTD and unity ratio describe the prevailing ventilation pattern of the water masses. Figure 9 shows the SF 6 excess in f mol kg −1 and ppt for depths below 200 m. This upper depth limit is invoked 540 by the fact that CFC-12 concentrations above the current atmospheric concentration limit cannot be applied to the IG-TTD. The SF 6 excess is much higher (0.5 − 0.8 f mol kg −1 / 1.0−1.6 ppt) for northwards propagating water masses compared to water masses of Arctic origin (0 − 0.4 f mol kg −1 / 545 0−0.8 ppt). There are at least two possible effects which can cause such significant supersaturations of SF 6 . One possibility refers to the deliberate tracer release experiment in 1996 where 320 kg (∼ 2190 mol) of SF 6 were introduced into the central Greenland Sea (Watson et al., 1999).

550
The patch was redistributed by mixing processes and entered the Arctic Ocean via the Fram Strait and Barents Sea Opening and the North Atlantic via Denmark Strait and the Faroe Bank Channel (Olsson et al., 2005;Tanhua et al., 2005;Marnela et al., 2007). Assuming that 50 − 80 % of the deliberatly 555 released SF 6 still remains in the Nordic Seas and the Arctic Ocean (1095 − 1752 mol) and that 10 − 50 % of the corresponding total water volume of 1.875 · 10 18 − 9.375 · 10 18 l (Eakins and Sharman, 2010) is affected, a mean offset of 0.12 − 0.93 f mol l −1 might be found. This mean offset is in 560 the range of the observed SF 6 excess concentrations. However, CFC-12 and SF 6 data of the Southern Ocean (Stöven et al., 2015) shows similar tracer age relationships compared to the Fram Strait data but with no influence of deliberately released SF 6 . This indicates that another source of excess 565 SF 6 may exists which is much larger than the source of the tracer release experiment. Liang et al. (2013) introduced a model which estimates supersaturations of dissolved gases by bubble effects in the ocean. This model predicted an increasing supersaturation 570 for increasing wind speed and decreasing temperature, i.e. the bubble effect becomes more significant at high latitudes. Furthermore, Liang et al. (2013) show that the magnitude of supersaturation depends on the solubility of the gas. The less soluble a gas, the more supersaturation can be expected. Sup-575 porting this, Stöven et al. (2015) describe surface measurements of SF 6 and CFC-12 directly after heavy wind conditions in the Southern Ocean where SF 6 supersaturations of 20 − 50 % could be observed. The CFC-12 concentrations were only affected to a minor extent which indeed can be ex-580 plained by the differences in solubility. This bubble-induced supersaturation can also be expected to occur during the process of water mass formation in the Greenland Sea which usually occurs during late winter, i.e. during a period with low surface temperatures and heavy wind conditions. Fur-585 thermore, the maximum SF 6 excess in the Arctic Intermediate Water layer in Fig. 9 and the generally elevated tracer concentrations of CFC-12 and SF 6 in the same area (see Fig.  3) reaffirm the assumption of bubble induced supersaturation of SF 6 . However, this hypothesis stands in opposition to the 590 current assumption that trace gases are generally undersaturated during water mass formation (Tanhua et al., 2008;Shao et al., 2013). Future investigations are necessary to determine the different impact of under-and supersaturation effects on soluble gases 595 at the air-sea interface. It can be expected that possible scenarios are not restricted to distinct saturation states anymore but rather comprise mixtures of equilibrated, under-and supersaturated states of the different gases.

Anthropogenic carbon and mean age 600
Since CFC-12 is not affected by tracer release experiments and possibly only to minor extent by bubble effects we used this tracer to calculate the mean age of the water and the corresponding anthropogenic carbon content. SF 6 was only used in the surface and upper halocline, i.e. where CFC-605 12 exceeds the atmospheric concentration limit of 528 ppt and where effects of SF 6 supersaturation are comparatively small. Saturation-corrected tracer data was applied for subsurface data below 100 m whereas surface data was found to be near equilibrium state with the atmosphere. Figure   610 10 shows the anthropogenic carbon distribution and Fig. 11 shows the mean age of the water masses. As expected from the relation between transient tracers, mean age and anthopogenic carbon, the distribution patterns are similar to that of transient tracers. The highest anthropogenic carbon con-  Table 1 shows the mean values and standard deviation of each specific water layer.

635
The determined values are comparable to the findings of Jutterström and Jeansson (2008) who used a similar method to determine anthropogenic carbon of the East Greenland Current in 2002. The Fram Strait section of their data set shows a similar distribution pattern of anthropogenic carbon but with 640 lower concentration levels compared to our data from 2012. The concentration differences indicate an increase of the anthropogenic carbon content between 25 − 35 % in the entire water column during the elapsed ten years. This corresponds to an increase of 2 µmol kg −1 yr −1 in the Atlantic 645 Water, 1 µmol kg −1 yr −1 in the Polar Water and between 0.5 − 1 µmol kg −1 yr −1 in the deeper water layers. Based on these current rates of increase it can be assumed that the import of anthropogenic carbon by Atlantic Water becomes more dominant compared to the export by Polar Water in the 650 future. Furthermore, when looking at the different gateways to the Arctic Ocean, it can be assumed that the Atlantic Water entering the Arctic Ocean via the Barent Sea has similar anthropogenic carbon concentrations as in the Fram Strait and that the outflow water through the Canadian Archipelago has 655 similar concentrations as the Polar Water in the Fram Strait. The inflow of Pacific Water transports ∼ 46 µmol kg −1 of anthropogenic carbon into the Arctic Ocean (Stöven, unpublished data 2014). This implies that the inflowing water masses transport more anthropogenic carbon into the Arctic 660 Ocean than the outflowing water masses since the water mass exchange must be balanced.

Sensitivities in anthropogenic carbon
The calculations presented above are based on the ideal case of pCO 2,atm =pCO 2,water at the sea surface before entering 665 the ocean interior, and the assumption that the saturation correction of the tracers and the unity ratio of the IG-TTD are true for water masses in the Fram Strait. Since these three parameters involved cannot be directly determined, it is very likely that deviations from the ideal case occur. Therefore, we 670 present the corresponding sensitivities in the following. The sensitivities are determined by changing only one parameter and keeping the others constant at ideal conditions. Figure 12a and 12b show the sensitivities of changes in tracer saturation using the example of CFC-12 since most of the 675 anthropogenic carbon calculations are based on this tracer. Small deviations of ±5 % in CFC-12 saturations cause only small deviations of anthropogenic carbon concentrations of ±1 µmol kg −1 / ±2 − 4 %. Furthermore, the sensitivity depends on the partial pressure range of CFC-12. The lower 680 the partial pressure the less sensitive are the anthropogenic carbon concentrations to changes in CFC-12 saturation. The maximum deviations are ±6 µmol kg −1 / ±11 − 16 % for partial pressure > 400 ppt. The white patches in Fig. 12a and 12b correspond to supersaturations which exceed the atmo-685 8 T. Stöven et al.: C ant in the Fram Strait spheric concentration limit of CFC-12. Figure 12c and 12d show the sensitivities due to changes in the ∆/Γ-ratio of the IG-TTD. The sensitivity is very low (< 1 µmol kg −1 / < 5 %) for most of the ratio and concentration range. Partial pressures below 100 ppt and ∆/Γ < 0.4 show the highest sensitivty with deviations between 5 − 10 µmol kg −1 / 50 − 200 %. The unusual sensitivity distribution is related to the indistinct boundary condition of CFC-12 in recent years and the distribution function of the TTD. For more detailed information, see Stöven et al. (2015).

695
The sensitivities of deviations in pCO 2 saturations are shown in Fig. 12e and 12f. The absolute error is characterized by a relatively steady change with changing saturation states. The absolute error is more or less independent of the partial pressure of CFC-12 and leads to maximum deviations of 700 ±20 − 25 µmol kg −1 . The relative error (0 − 200 %) shows an increasing sensitivity of anthropogenic carbon concentrations to changes in pCO 2 saturations and decreasing CFC-12 partial pressures. Note that a negative deviation of 100 % corresponds to anthropogenic carbon concentration of 705 0 µmol kg −1 which is also indicated by the turning-points where the contour lines continue parallel to the x-axis in Fig.  12e. This indicates that small uncertainties in pCO 2 saturations can cause large errors in anthropogenic carbon estimates for low tracer concentrations, i.e. for a high mean 710 age of the water. Furthermore it is unclear to what extent the time period and type of sea ice coverage as well as the sea ice formation and melting processes bias the pCO 2 and tracer saturations at high latitudes. The uncertainty of the pCO 2 saturation remains as the largest error source although the satu-715 ration of pCO 2 and CFC-12 counteract each other. Table 2 shows the transport estimates of DIC and anthropogenic carbon separated into northwards flowing (positive values) and southwards flowing (negative values) water 720 masses. The northwards flux comprises the Atlantic Water of the West Spitsbergen Current, the southwards flux comprises the Recirculating / Return Atlantic Water and the Polar Water of the East Greenland Current. The mean flux of deep water layers below 840 m was taken to be 0 Sv and therefore 725 not considered for this estimate. Furthermore, any net flux below 1500 m would not change the anthropogenic carbon inventory of the Nordic Seas or the Arctic Ocean due to the homogeneous distribution of anthropogenic carbon at these depths. The depth range between 840 m and 1500 m might 730 contribute to either the Arctic or the Nordic Seas reservoir but it is still an enclosed basin-basin interaction. The northwards flux transports 3592 ± 2612 T g C yr −1 (mean ± standard deviation) of DIC and 78 ± 57 T g C yr −1 of anthropogenic carbon into the Arctic Ocean. This inflow 735 is exceeded by an outflow of 2852 ± 1549 T g C yr −1 / 67 ± 36 T g C yr −1 by Recirculating and Return Atlantic Water and 1118 ± 639 T g C yr −1 / 23 ± 13 T g C yr −1 by Polar Water. The carbon transport uncertainties are relatively high and there is a lack of water transport data on the 740 Greenland shelf region, e.g. Belgica Bank. Thus we cannot with great confidence decide whether more anthropogenic carbon is transported into or out of the Arctic region through the Fram Strait.

Uncertainties
We showed that neither the IG-TTD nor linear combinations of the model can describe the tracer age relationships between CFC-12 and SF 6 in the Fram Strait. This means that either the models are not suitable to describe the prevail-750 ing ventilation pattern or that there are other reasons which lead to the specific concentration ratios. Here we focused on the second case which incorporates the assumptions that the tracer age relationships are related to different saturation states of the transient tracers and, furthermore, that the sim-755 ple IG-TTD model can describe the ventilation processes of all water masses in the Fram Strait. The uncertainties of our approach thus correspond to the chosen shape of the IG-TTD, i.e. the unity ratio of ∆/Γ = 1.0, and the uncertainties of the measurement precision of the 760 transient tracers and apparent transient tracers (see section 3.6 above). Further uncertainties are related to processes which influence the gas exchange and thus the boundary conditions of the tracers. This includes the important but yet rarely investigated impact of sea ice cover, sea ice forma-765 tion and sea ice melting processes as well as bubble effects during heavy wind conditions, see discussion in section 3.6. The flux estimates are based on transient tracer and DIC data of the ARK-XXVII/1 cruise which only show the specific distribution pattern during June / July 2012 and thus neglect 770 any interannual variabilities of the parameters. The determination of the preformed alkalinity highly depends on the used method. Here we used the linear relationship between surface alkalinity and salinity which is a commonly used method. However, other authors recommend the use of alkalinity / 775 salinity data from the subsurface layer (Vazquez-Rodriguez et al., 2012) or the surface temperature and salinity dependencies (Lee et al., 2006). The transport estimates are complicated by the fact that the flow field in the Fram Strait is dominated by small scale fea-780 tures. The Rossby radius is 4 − 6 km which means that the mooring spacing is only able to fully resolve the mesoscale near the shelfbreak in the West Spitsbergen Current. Otherwise, eddies may be aliased between the moorings. The velocities in the recirculation area in the center of the Fram 785 Strait are actually mostly westward  and thus along the mooring array line. Therefore, the meridional velocities in the center of the Fram Strait are only the small residuals of much larger zonal velocities. As a result the finite accuracy and precision of the current 790 direction measurements has a big impact on the meridional exchanges. Additionally, at depth the flow is topographically steered, but the topographic features are not fully resolved. Interannual variations are also neglected here, but they are small . The exchange flow 795 across the Fram Strait below 840 m (sill depth of Greenland-Scotland ridge) is assumed to be 0 Sv for the present purpose.