Low planktic and benthic δ18O and δ13C values in sediments from the Nordic seas of cold stadials of the last glaciation have been attributed to brines, formed similar to modern ones in the Arctic Ocean. To expand on the carbon isotopes of this hypothesis, I investigated benthic δ13C from the modern Arctic Ocean. I show that mean δ13C values of live epibenthic foraminifera from the deep Arctic basins are higher than mean δ13C values of upper slope epibenthic foraminifera. This agrees with mean high δ13C values of dissolved inorganic carbon (DIC) in Arctic Bottom Water (ABW), which are higher than mean δ13CDIC values from shallower water masses of mainly Atlantic origin. However, adjustments for oceanic 13C Suess depletion raise subsurface and intermediate water δ13CDIC values over ABW δ13CDIC ones. Accordingly, during preindustrial Holocene times, the δ13CDIC of ABW was as high or even higher than today but lower than the δ13CDIC of younger subsurface and intermediate water. If brine-enriched water significantly ventilated ABW, brines should have had high δ13CDIC values. Analogously, high-δ13CDIC brines may have been formed in the Nordic seas during warm interstadials. During cold stadials, when most of the Arctic Ocean was perennially sea ice covered, a cessation of high-δ13CDIC brine rejection may have lowered δ13CDIC values of ABW, and ultimately the δ13CDIC in Nordic seas intermediate and deep water. So in contrast to the idea of enhanced brine formation during cold stadials, the results of this investigation imply that a cessation of brine rejection would be more likely.
Planktic and benthic stable isotopes in Nordic seas sediment cores from the last glacial period are high during interstadials and low during stadials [Rasmussen et al., 1996a; Dokken and Jansen, 1999; Meland et al., 2008]. It is widely accepted that low stadial planktic δ18O values may reflect meltwater input from icebergs [Rasmussen and Thomsen, 2009; Thornalley et al., 2010a]. However, the cause for low benthic stable isotopes is still debated. So it was suggested that analogously to processes in the Recent polar oceans, sea ice and large-scale brine formation transferred the stable isotope signal of the surface water to intermediate and deep water masses [Vidal et al., 1998; Dokken and Jansen, 1999; Meland et al., 2008]. Others suggest that low benthic δ18O values may reflect the influence of warm intermediate and deep water masses [Rasmussen et al., 1996b; Bauch and Bauch, 2001].
To test whether deep water stable isotopes modified by brines are recorded in benthic foraminiferal δ18O and δ13C, I analyzed live (Rose Bengal stained) epibenthic foraminifera from the Arctic Ocean, where brine formation takes place today [Aagaard et al., 1985; Shapiro et al., 2003; Ivanov et al., 2004]. I assume that modern hydrography and sea ice conditions in parts of the Arctic Ocean are similar and comparable to conditions in parts of the Nordic seas during stadial times [Meland et al., 2008]. Since the determination of benthic δ13C is robust and its use as δ13CDIC proxy is well described, I here focus on the discussion of epibenthic calcite δ13C. The benthic δ18O will be discussed elsewhere.
1.2 Oceanographic Settings
The Arctic Mediterranean Sea, north of Fram Strait, contains only 1% of the world ocean's water but represents 3% of its area. This is because about 70% of the surface area of the Arctic Ocean is occupied by shallow continental shelves with average water depths between about 40 and 200 m, the width of which along the Eurasian coast exceeds 800 km in most places [e.g., Tomczak and Godfrey, 1994; Stein, 2008]. Three principal water masses can be distinguished: Arctic Bottom Water (ABW), Atlantic Water (AW), and Arctic Surface Water (ASW).
ABW fills the deep basins up to about 900 m water depth [Tomczak and Godfrey, 1994]. In the Eurasian basins, ABW sometimes is called Eurasian Basin Bottom Water to differentiate from Canadian Basin Bottom Water, which is warmer (Figures 1 and 2). Furthermore, an intermediate water mass, shallower than about 2000 m water depth in the Amerasian basins and <1500 m in the Eurasian basins, can be distinguished from ABW, which is called upper ABW (uABW) or upper Polar Deep Water or Arctic Intermediate Water [e.g., Jeansson et al., 2008]. ABW is a mixture of water from (i) the inflow of Atlantic water over the Barents Sea shelf in the Eurasian basins, (ii) contributions from the Arctic shelves, and (iii) recirculated Greenland Sea Deep Water (GSDW) and Norwegian Sea Deep Water (NSDW) entering the central Arctic Ocean through Fram Strait [Jones et al., 1995].
The formation of ABW generally involves GSDW and brine-enriched shelf water (BSW) from the Arctic shelves [Aagaard et al., 1985; Rudels and Quadfasel, 1991; Jones et al., 1995; Jones, 2001]. GSDW is formed during winter in the central Greenland Sea in short-lived and small-scale events, during which intense vertical convection occurs and surface water sinks to the bottom. Formation of GSDW occurs in mainly open water [Rudels and Quadfasel, 1991; Tomczak and Godfrey, 1994]. BSW is formed on the Arctic shelves [Aagaard et al., 1985]. There surface water is depleted in 18O and 13C by riverine fresh water, which freezes in coastal polynyas rejecting dense but low δ18O brine. Brines from low-salinity shelves, however, just ventilate the Arctic Ocean halocline [Aagaard et al., 1985; Cavalieri and Martin, 1994; Bauch et al., 2005, 2011a, 2011b]. Only on high-salinity shelves, such as the salty Barents and Kara Seas, brines are dense enough to contribute to deep water formation and its stable isotope signal [Bauch et al., 1995; Shapiro et al., 2003; Ivanov et al., 2004]. Similarly, an investigation of benthic δ18O records from Storfjorden (South Svalbard) suggests that in the Arctic only brines, formed from cold and salty waters with high δ18O values, are sufficiently dense to contribute to deep water formation [Rasmussen and Thomsen, 2009]. Bauch and Bauch , based on deep profiles of δ18O in the Arctic Ocean from Bauch et al. , conclude that today only very little low δ18O freshwater via brines reaches Arctic Bottom Water (ABW). Jones et al. , however, conclude that dense water, triggered by brine-enhanced waters formed on shelves, flows down the continental slopes. This dense water, consisting of brines and primarily of waters entrained from the Atlantic layer, contributes most to ABW. The second most important contributor to ABW is inflow of Atlantic Water over the Barents Sea shelf and least important is inflow of NSDW through Fram Strait [Jones et al., 1995]. The residence time of ABW is estimated at between 250 and 450 years [Schlosser et al., 1994; Macdonald and Bewers, 1996] with highest ages in the fairly isolated deep Canada Basin.
Above ABW, AW enters the Arctic through Fram Strait and occupies the depth range between about 900 and 150 m. The AW, also known as Atlantic Layer, typically is defined by the 1°C isotherm [Rudels et al., 2004] (Figure 1). The uppermost layer of about 200 m of the water column in the Arctic Ocean is ASW. Within this layer, it can be distinguished between a 25–50 m thick surface layer and a subsurface layer of Arctic Ocean halocline water below [Aagaard et al., 1985, 1987; Tomczak and Godfrey, 1994]. The renewal age of the Atlantic layer and the halocline water is given with about 25 and about 10 years, respectively [Östlund, 1982; Macdonald and Bewers, 1996]. At present, the central Arctic Ocean is permanently covered by sea ice, but the vast area of continental shelves surrounding the central basins is free of sea ice during summer. On the shelves, most of the sea ice is formed during winter [Aagaard et al., 1985].
1.3 The Distribution of δ13C in the Ocean
The distribution of δ13C in oceanic dissolved inorganic carbon (DIC) mainly depends on the biological cycle of 13C-depleted organic matter and the effects of gas exchange at the air and sea interface. Ocean surface water δ13CDIC is influenced by carbon isotopic fractionation during exchange between the species of DIC and atmospheric CO2, such that equilibrium fractionation increases with 0.1‰ per 1°C temperature decrease [Zhang et al., 1995]. This process is particularly important at high latitudes with temperatures close to freezing point and high wind speeds [Charles and Fairbanks, 1990; Broecker and Maier-Reimer, 1992; Lynch-Stieglitz et al., 1995]. In regions of bottom water formation in the inner Weddell Sea of the Antarctic Ocean, even deep and intermediate water δ13CDIC is influenced by air and sea gas exchange [Charles et al., 1993; Mackensen et al., 1993; Mackensen, 2012]. However, on the Arctic shelves where air sea gas exchange is strong [e.g., Anderson et al., 2004], no high-δ13CDIC water masses have been described so far. There, and in the northern North Atlantic in general, relatively low surface δ13CDIC values have been explained by enhanced uptake of atmospheric low-δ13C CO2 into the surface ocean [Lynch-Stieglitz et al., 1995]. To reach carbon isotopic equilibrium during air and sea gas exchange, it takes about 10 times longer than for CO2 itself, i.e., about 10 years in a 50 m thick mixed layer [Tans, 1980; Lynch-Stieglitz et al., 1995]. In addition, ocean surface waters are replaced faster than this time span by subsurface water or from other sources. So ocean surface DIC generally is not in complete isotopic equilibrium with atmospheric CO2 [Broecker and Maier-Reimer, 1992; Lynch-Stieglitz et al., 1995].
An additional lowering of surface water δ13CDIC during air and sea gas exchange by anthropogenically decreased atmospheric δ13CCO2, commonly referred to as the oceanic 13C Suess effect [Quay et al., 1992], has to be considered when comparing Recent δ13CDIC water values with stadial benthic calcite. Olsen et al.  calculated an anthropogenic δ13CDIC reduction between 1981 and 2003 of −0.3‰ in the core of Atlantic water inflow into the Arctic Ocean west of Spitsbergen and of −0.2 to −0.5‰ in the middepth Norwegian Sea. Bauch et al.  estimated an oceanic 13C Suess effect of about −0.9 ±0.2‰ in halocline waters and −0.6 ±0.1‰ in Atlantic-derived waters of the Nansen Basin. The δ13CDIC of source waters for brine formation additionally is lowered due to entrainment of Arctic river freshwater. From the Ob and Yenisey River estuaries, for example, δ13CDIC values of −12 to −4‰ have been reported [Galimov et al., 2006].
Since the formation of brines in the Arctic Ocean takes place in coastal polynyas on shallow continental shelves during wintertime, and downslope cascading of dense water happens episodically and only at slopes off the so-called high-salinity shelves [Quadfasel et al., 1988; Ivanov et al., 2004], direct observations and documentation of such events are scarce. Therefore, I use δ13C values from calcite tests of live epibenthic foraminifera, compared to some water δ13CDIC measurements, to investigate the bottom water δ13CDIC distribution in the Arctic Ocean and the Nordic seas. The use of epibenthic δ13C as a proxy of δ13CDIC is based on the assumption that benthic foraminiferal species Cibicidoides wuellerstorfi, Cibicides lobatulus, and Cibicides refulgens fractionate carbon isotopes between water DIC and shell carbonate in a one-to-one relationship during calcification of their tests [Duplessy et al., 1984; McCorkle and Keigwin, 1994; Mackensen et al., 2000; Eberwein and Mackensen, 2008]. This assumption is warranted as long as strong seasonal productivity, ultimately leading to the development of a phytodetritus layer at the seafloor, is not influencing the sites under investigation [Mackensen et al., 1993; Zarriess and Mackensen, 2011].
2 Material and Methods
2.1 Water Samples
On cruises with PRV Polarstern and RV Heincke between 1991 and 2008 at 262 positions between latitudes 67° and 90°N and between longitudes 19°W and 164°E, water was sampled to determine its δ13CDIC (Figure 2). Immediately after subsampling of Niskin bottles, water samples were poisoned with a saturated solution of mercury chloride, sealed with wax, and stored at 4°C temperature until further treatment. On shore, 1 mL of water was injected through a septum into a vial with 70 μL concentrated phosphoric acid flushed with pure helium. All water samples were treated in duplicate. After storage at room temperature for complete reaction, the resultant CO2 then was transferred via a Finnigan Gas Bench II to a Finnigan MAT 252 gas mass spectrometer for determination of its stable carbon isotope ratio. Results are given in δ notation versus VPDB (Vienna Pee Dee belemnite). The precision of δ13C measurements based on an internal laboratory standard measured over a 1 year period together with samples was better than ±0.1‰.
Additional seven δ13CDIC values from the central Arctic Ocean were from stations occupied by icebreaker Oden in 1991 (T. Johannessen, personal communication, 1994). Also included are data of bottom water samples from seven and five stations occupied by RV Lance and RV G.O. Sars, respectively, on cruises in 1990 [Nydal et al., 1991], as well as data from four Geochemical Ocean Sections Study (GEOSECS) stations in the Greenland, Iceland, and Norwegian (GIN) Sea sampled in 1972 [Bainbridge, 1981].
2.2 Surface Sediment Samples
Between 1987 and 2012 at 556 sites in the Arctic ocean and Nordic seas, surface sediment samples were recovered on 12 cruises with PRV Polarstern, 3 cruises with PRV Araon, and 1 cruise with RV Heincke for benthic foraminifera analyses, 305 of which were from below 250 m water depth and yielded foraminifera species suitable for isotopic analysis in this study (Figure 3 and Table S1 in the supporting information). Immediately after recovery on board ship, all samples except those from PRV Araon were preserved in Rose Bengal-stained ethanol (1 g/L). On shore, samples were wet sieved through 63 µm, 125 µm, and 2 mm sized meshes. After drying at temperatures less than 60°C, between 1 and 10 epibenthic foraminifera were selected from the larger than 125 µm fraction for stable isotope analyses. If available, one to three specimens of Cibicidoides wuellerstorfi were separated and transferred, via the automated carbonate preparation device Kiel Carbo, into Finnigan MAT251 or MAT253 mass spectrometers. If C. wuellerstorfi specimens were not available, specimens of closely related species Cibicides lobatulus and Cibicides refulgens were analyzed. No other species were included in this study. The mass spectrometers were calibrated via international standard NBS-19 to the PDB scale, and results are given in δ notation versus VPDB. The precision of δ13C measurements, based on an internal laboratory standard (Solnhofen limestone) measured over a 1 year period together with samples, was better than ±0.06‰. In addition, I included 12 C. wuellerstorfi δ13C values from unstained surface sediment samples taken in 1994 in the Mendeleyev Ridge area of the Amerasian basins [Poore et al., 1999].
Bottom water δ13CDIC data (Figure 2) and live (stained) and all (live and dead) benthic foraminiferal δ13CCib data (Figure 3 and Table S1) were gridded with Data Interpolating Variational Analysis (http://modb.oce.ulg.ac.be/projects/1/diva) and mapped with Ocean Data View (ODV)[Schlitzer, 2011]. For a closer look, I subdivided the area of interest into four geographic and hydrographic regions, which again are divided into a shallow and deep water box each (Tables 1 and 2 and Figures 4-6): (i) the GIN Sea (GIN), including surrounding continental slopes; (ii) the Eurasian basins and slopes, divided at 90°E into a southwestern (SEU) and a northeastern (NEU) part; and (iii) the Amerasian basins and eastern slopes (AME). Deep water boxes were chosen such that bottom water masses below a “permanent pycnocline” were covered, i.e., well below Atlantic and intermediate water masses. Alternative calculations confining shallow boxes just to water depths characterized by Atlantic-derived water did not change the results of this study and are not given.
Table 1. Mean Bottom Water δ13CDIC According to Coarse Hydrographic, Bathymetric, and Geographic Distribution (Figure 5)
Mean δ13CDIC (‰ VPDB) ± Std Dev
GIN Sea (65°–80°N, 20°W–40°E)
Greenland, Svalbard slopes 250–1500 m
0.99 ± 0.19
GIN Sea >1500 m
1.18 ± 0.14
SW Eurasian Basins (80°–90°N, 20°W–90°E)
Greenland, Svalbard slopes 250–1500 m
0.89 ± 0.28
SW Eurasian basins >1500 m
1.14 ± 0.21
NE Eurasian Basins (70°–90°N, 90°–140°E)
Laptev Sea slope 250–1500 m
0.83 ± 0.22
NE Eurasian basins >1500 m
0.95 ± 0.27
Amerasian Basins (70°–90°N, 140°E−170°W)
East Siberian Sea slope 250–2000 m
1.04 ± 0.29
Makarov Basin >2000 m
1.26 ± 0.12
Table 2. Mean Epibenthic δ13C According to Coarse Hydrographic, Bathymetric, and Geographic Distribution (Figure 5)
Live δ13C (‰VPDB) ± Std Dev
All δ13C (‰VPDB) ± Std Dev
GIN Sea (65°–80°N, 20°W–40°E)
Greenland, Svalbard slopes 250–1500 m
1.02 ± 0.31
1.03 ± 0.30
GIN Sea basins >1500 m
1.21 ± 0.13
1.22 ± 0.14
South Eurasian Basins (80°–90°N, 20°W–90°E)
Barents and Kara Seas slopes 250–1500 m
1.15 ± 0.20
1.17 ± 0.22
South Eurasian basins >1500 m
1.34 ± 0.20
1.41 ± 0.24
North Eurasian Basins (70°–90°N, 90°–140°E)
Laptev Sea slope 250–1500 m
1.20 ± 0.24
1.23 ± 0.29
North Eurasian basins >1500 m
1.31 ± 0.29
1.27 ± 0.24
Amerasian Basins (70°–90°N, 140°E−130°W)
East Siberian/Chukchi Sea slope 250–2000 m
1.44 ± 0.22
1.44 ± 0.22
Amerasian Basins >2000 m
1.44 ± 0.10
1.43 ± 0.12
3.1 Water δ13CDIC
A mean δ13CDIC of 0.98 ± 0.23‰ is calculated from 285 water samples from close to the seafloor. The samples recovered over a time period of almost 20 years plus four samples from GEOSECS stations in 1972 give a general idea on the deep and bottom water δ13CDIC distribution and deep circulation (Figures 2, 5, and 6). Ninety-four out of 285 samples were from below 1500 m water depth and thus about the depth range covered by unmodified ABW in the Arctic Ocean and by NSDW and GSDW in the GIN Sea. The mean δ13CDIC of these water masses is 1.13 ± 0.20‰. A more detailed calculation reveals in the GIN Sea, GSDW and NSDW with mean δ13CDIC values of 1.18‰, and in the adjacent southern Eurasian basins, ABW with mean 1.14‰. This is in contrast to low values of 0.95‰ in the deep northern Eurasian basins and high values of 1.26‰ in the deep Amerasian basins (Table 1 and Figures 2, 5, and 6). It has to be noted, however, that the latter two means both in numbers and distribution are not as representative as the former two. So are the Amerasian basins in these data represented by just six values from the Makarov Basin (Figure 2).
A total of 138 out of 285 samples were from water depths between 250 and 1500 m, thus covering water masses as different as Atlantic Water inflow in the eastern GIN Sea and the Eurasian basins and Arctic water outflow in the west via the East Greenland Current. However, all of these subsurface water masses are characterized by a mean δ13CDIC of 0.96 ± 0.21‰. A more detailed calculation reveals significant differences only between the shallow GIN Sea and the shallow Eurasian basins and between shallow Eurasian and Amerasian basins (Table 1 and Figures 2 and 4-6).
3.2 Epibenthic Calcite δ13C
We determined the stable carbon isotopic composition of C. wuellerstorfi or Cibicides spp. from 305 surface sediment samples, which resulted in a mean δ13C of 1.24 ± 0.28 (Figure 3). In 246 samples, a sufficient number of live specimens were found to determine a most reliable mean δ13C of 1.22 ± 0.27‰ (Figure 3).
A total of 140 out of 305 samples are from below 1500 m water depth and covered by ABW, NSDW, and GSDW. The mean benthic foraminiferal calcite from these ABW-bathed grounds including areas covered by GSDW/NSDW is 1.34 ± 0.20‰. If analyzing only live specimens from 109 samples, a mean of 1.32 ± 0.19‰ is obtained. The detailed calculation reveals increasing means from the deep GIN Sea via the Eurasian basins to Amerasian basins from 1.21 to 1.44‰ and 1.22 to 1.43‰ in the live and all data sets, respectively (Table 2 and Figures 3-5).
A total of 165 out of 305 surface sediment samples were from water depths between 250 and 1500 m, including water masses of most different origin and genesis. So the mean δ13C of 1.15 ±0.30‰ is given here just to complete statistics. However, the area-specific and roughly water mass-confined calculation reveals increasing means from the shallow GIN Sea via the shallow Eurasian basins to shallow Amerasian basins from 1.02 to 1.44‰ (Table 2 and Figures 3-5). Furthermore, mean δ13C of the shallow GIN Sea and shallow southern Eurasian basins are significantly lower than the means of the corresponding deep basins. In contrast, differences between shallow and deep water means in the northern Eurasian and the Amerasian basins are not significant, i.e., they plot within the overlapping range of standard errors (Table 2 and Figure 4).
4.1 Water δ13CDIC
Low δ13CDIC values on the Svalbard continental slope between 200 and 900 m water depth reflect Atlantic water inflow, whereas very low bottom water δ13CDIC values in Svalbard's fjords and on uppermost slopes and shelves are partly associated with methane release [Damm et al., 2005; Westbrook et al., 2009]. Low δ13CDIC values on the wide continental shelf off East Greenland may result from glacier meltwater runoff and melted landfast ice [Funder et al., 2011]. Low δ13CDIC values in Barents Sea troughs and on Kara Sea continental slopes (Table 2, Figure 2) may partly reflect dissipating influence of old Atlantic-derived water. For another part, low δ13CDIC in Barents and Kara Seas troughs may result from dense and low δ13CDIC BSW, which was formed by entrainment of river freshwater and salinization during sea ice formation [Rudels et al., 2004; Bauch et al., 2011a]. BSW formed during winter freezing in the Barents and Kara Seas is dense enough to feed water masses deeper than the halocline and potentially modify AW bathing Kara and Laptev Seas continental slopes [Aagaard et al., 1985; Shapiro et al., 2003; Ivanov et al., 2004; Rudels et al., 2004].
Below 1500 m water depth in the GIN Sea and southern Eurasian basins, and below 2000 m in the Amerasian basins, mean δ13CDIC is 0.2‰ higher than mean δ13CDIC values from shelves and upper slopes. Only in the northern Eurasian basins, i.e., mainly at the Laptev Sea continental slope, the difference is within standard errors (Table 2). However, generally high mean values in the depth corroborate the spontaneous impression provoked by Figures 2 and 6 that the DIC of ABW in the deepest layers of the Eurasian basins and of GSDW/NSDW in the GIN Sea is enriched in 13C. So we end up with the paradox that water masses adding to ABW in the Eurasian basins and the Makarov Basin are carbon isotopically by about 0.2‰ lighter than the final product they are part of and admixed to, and moreover, the older water mass appears to be the carbon isotopically heavier one.
Here it is worthwhile to recall the possible sources of ABW before considering the δ13CDIC distribution in more detail: Jones et al.  suggest that dense BSW, which flows down the continental slopes consisting of brines and waters entrained from the Atlantic layer, contributes most to ABW. Especially, brines from polynyas in the relatively high-salinity Barents and Kara Seas significantly contribute to Arctic deep water formation [Bauch et al., 1995], whereas brines from low-salinity shelves may only ventilate the halocline water [Bauch et al., 2005; Rasmussen and Thomsen, 2009]. The second most important contributor to ABW is inflow of Atlantic Water over the Barents Sea shelf and least important is inflow of NSDW through Fram Strait [Jones et al., 1995]. Having this in mind, three principle reasons and processes can explain the observed distribution: (i) an additional source of high δ13CDIC BSW not recognized in the limited area of this investigation, notably on the Chukchi Sea shelves; (ii) the composition of a major part of ABW of NSDW entering through Fram Strait; and (iii) a lowering of young surface water δ13CDIC during air and sea gas exchange by anthropogenic decrease of atmospheric δ13CCO2, commonly referred to as the oceanic 13C Suess effect [Quay et al., 1992].
A possible high δ13CDIC source is snow and glacier meltwater from areas without soil cover and no vegetation and bacterial decay of organic matter, the DIC of which equilibrates with atmospheric CO2 and not with depleted soil CO2 and thus carry a δ13C signal close to a high marine surface δ13CDIC at low temperatures. Originally, such a pathway was suggested to explain high δ13C of secondary calcites in high-mountain regions [Magaritz, 1973]. However, the present δ13CDIC data coverage and knowledge of the Eurasian hinterland do not give any hint in support of this hypothesis and recognition of such supply areas. On the contrary, δ13CDIC data from the Ob and Yenisey River estuaries in the Kara Sea range between −12 and −4‰ [Galimov et al., 2006].
Since the δ13CDIC of NSDW is virtually the same as the one of ABW in the southern Eurasian basins (Table 1 and Figures 2 and 4-6), advective supply of NSDW, as well as recirculation and exchange with ABW may explain similar δ13CDIC values of ABW and NSDW. This, however, is in conflict with current understandings of the origin of ABW [Jones et al., 1995], although it is in support of a much earlier, classic oceanographic view [Nansen, 1915].
Anthropogenically enriched 12C of atmospheric CO2 has lowered δ13CDIC values by about −0.9 ± 0.2‰ in halocline waters and −0.6 ± 0.1‰ in Atlantic-derived waters of the Nansen Basin [Bauch et al., 2000], both of which have residence times of about 10 and 25 years, respectively [Macdonald and Bewers, 1996]. This strong 13C depletion was estimated by comparing planktic foraminiferal δ13C from the water column and sediment surfaces [Bauch et al., 2000]. By comparing water data from 1981 and 2002/2003 in the Nordic seas, Olsen et al.  calculated an oceanic 13C Suess effect of −0.3‰ west of Spitsbergen in the core of AW inflow into the Arctic Ocean, −0.2 to −0.5‰ in the middepth Norwegian Sea, but no significant change in the depth range bathed by NSDW and GSDW, i.e., below about 1500 m. Olsen et al.  calculated the Suess effect only for a period of about 20 years. A scaling up to the full industrial period using the approach of Tanhua et al.  as applied by Olsen and Ninnemann  increases the anthropogenic depletion of δ13CDIC of AW inflow to about −1.0‰. Since there was no anthropogenic-induced change detected in deep GSDW [Olsen et al., 2006], which is younger than ABW, no correction of ABW δ13CDIC is required. Also, an ABW still reflecting a preindustrial δ13CDIC is in agreement with residence times of 250 to 300 and 350 to 450 years in the Eurasian and Amerasian basins, respectively, as calculated from radiocarbon and 39Ar data [Schlosser et al., 1994]. So in the end, a correction for an oceanic 13C Suess effect in the Arctic Ocean of +1‰ may be reasonable to adjust inflowing AW δ13CDIC to a preindustrial level. Even a correction of about 0.6‰ [Bauch et al., 2000] is more than sufficient to increase the mean δ13CDIC, calculated for shallow water boxes in this paper, to values clearly higher than deep water values (Table 1 and Figure 4). The above estimated correction for an oceanic Suess effect would make the preindustrial ABW δ13CDIC lower than its sources. This, of course, complies with the concept of a nonconservative behavior of δ13CDIC.
4.2 Epibenthic Calcite δ13C
The distribution of epibenthic δ13C is based on a much larger data set and as such extends and widens that of δ13CDIC considerably. Generally, the distribution pattern of δ13C values of live and all (live plus dead specimens) epibenthic foraminifera well reflects the distribution of δ13CDIC of the specific deep and bottom water masses bathing the corresponding areas of the seafloor. Again, the paradox appears that the oldest and deepest water mass, i.e., ABW in the central Arctic basins is represented by highest δ13C calcite values, surrounded by shallower and younger waters with lower values at the continental slopes, ultimately feeding ABW (Figures 2 and 3). However, in contrast to the limited δ13CDIC data, the denser and more evenly distributed epibenthic δ13C suggests East Siberian and Chukchi Sea slopes to be bathed by waters with high δ13CDIC (Figure 3). Moreover, if less reliable δ13C values of empty tests are included, even lower values in the deep Canada Basin may be indicated (Figure 3), although epibenthic δ13C in the deep Makarov Basin remains high. In summary, calculation of mean values in the NE Eurasian and the Amerasian basins shows no statistically significant difference between deep and shallow boxes, though Amerasian basin values are higher than NE Eurasian ones (Table 2 and Figure 3). The latter might simply reflect an even lower sedimentation and remineralization of organic carbon in the deep Amerasian basins.
If a one-to-one relationship or possibly a constant offset is assumed between δ13C and δ13CDIC, the observed distribution of epibenthic mean δ13C in the shallow boxes may reflect three processes: (i) the 13C depletion of ASW and AW due to the oceanic Suess effect decreases from Fram Strait, along Siberia into the Beaufort Sea with increasing age of the water masses; (ii) due to laterally and seasonally extended sea-ice coverage, the excess uptake of atmospheric CO2 and thus kinetic depletion of 13CDIC may decrease from Fram Strait into the Beaufort Sea and thus raise surface water δ13CDIC; and (iii) the 13C enrichment due to thermodynamic fractionation increases with decreasing surface water temperatures on this route. However, the latter process probably is less effective, since surface water temperatures are mostly near freezing point in most areas.
4.3 Epibenthic δ13C Versus δ13CDIC
Mean δ13C values of live C. wuellerstorfi from the deep Arctic Ocean below 1500 m water depth are about 0.2‰ higher than mean δ13CDIC values of corresponding bottom water, as are the benthic foraminiferal δ13C means from the continental shelves and upper slopes (Figures 4 and 5). In paleoceanographic studies, an epibenthic foraminiferal δ13C typically is considered as reliable δ13CDIC proxy, if deviations are within a ±0.2‰ range [Duplessy et al., 1984]. Since a systematic offset of +0.2‰ has been observed in the Weddell Sea as well [Mackensen, 2012], such a constant positive deviation may reflect specific environmental conditions at high latitudes, and may be worthwhile to be discussed elsewhere. New results derived from sensitivity experiments with a reaction-diffusion model for calcification seem to suggest a possibly more significant impact of water temperature on benthic foraminiferal δ13C than previously acknowledged [Hesse et al., 2012].
However, at the Siberian continental margin, the offset between benthic foraminiferal δ13C and bottom water δ13CDIC is about as twice as high (Tables 1 and 2, and Figures 4 and 5). Possible reasons for such a high discrepancy between bottom water δ13CDIC and epibenthic δ13C as encountered off the Laptev and East Siberian Sea shelves include artifacts due to data quantity and coverage. So clearly the number of δ13CDIC measurements supposed to be representative of Amerasian basins δ13CDIC is too low and the few stations are restricted to the Makarov Basin. Furthermore, these stations geographically and hydrographically do not sufficiently overlap with epibenthic δ13C stations from Chukchi and East Siberian Seas slopes. In addition, differences between δ13CDIC and epibenthic δ13C may be due to calcite dissolution, which would leave benthic foraminiferal tests depleted in 12C dependent on the carbonate ion concentration of the water mass, and if no authigenic calcite is formed in the test's microenvironment. However, virtually the same δ13C of stained and empty tests strongly argues against such a scenario (Table 2). Also, most of the Arctic Ocean floor lies above the lysocline for calcite [Jutterström and Anderson, 2005].
4.4 Brine-Enriched Shelf Water and 13C Imprint
Planktic and benthic foraminiferal stable isotope records from the North Atlantic and Nordic seas show high δ18O and δ13C values during interstadials close to modern ones but low values during stadials [Rasmussen et al., 1996a; Dokken and Jansen, 1999]. It is suggested that in stadials during sea ice formation, brines are formed that bring low δ18O and maybe low δ13C values of surface waters to deeper intermediate water [Vidal et al., 1998; Dokken and Jansen, 1999; Waelbroeck et al., 2006; Meland et al., 2008; Thornalley et al., 2010b]. The Recent δ13CDIC and benthic foraminiferal δ13C distribution indicates low δ13C values on the Arctic shelves and the upper Laptev Sea slope (Figures 2-4). However, as suggested in the present study, dense brine-enriched shelf water with entrained Atlantic Water that cascades down the slope is strongly affected by a 13C Suess effect. A correction for this anthropogenic effect would raise carbon isotope ratios to levels higher than those in preindustrially formed ABW. So if using today's high Arctic conditions and bottom water formation processes as modern analogue for stadial conditions during the last glaciation in the more southern GIN Sea, modern Arctic low-δ13C brines that are biased by a strong 13C Suess effect should be interpreted as high-δ13C brines.
Therefore, I here suggest that low δ13C values during stadials in the North Atlantic and GIN Sea may not result from brine rejection on North Atlantic and GIN Sea shelves but from suppressed brine formation on Arctic shelves instead, resulting in a reduced thermodynamic imprint and lower δ13C values compared to interstadial ones. Since most of the Arctic Ocean was perennially ice covered during stadials [Aagaard-Sørensen et al., 2010; Polyak et al., 2010], and many of the shallow continental shelves fell dry, no or less dense brine-enriched waters were produced that could modify intermediate and subsurface water or episodically cascade downslope into the deep basins. Consequently, the contribution of high-δ13C water to deep and bottom water circulation was low compared to interstadial times. So not only possibly low-δ13C water from north Atlantic and Nordic seas shelves but less high-δ13C water from the deep Arctic Ocean might be attributed to low benthic stable isotope excursions during stadials in the Nordic seas. Also during stadials, an intensified bottom water exchange and recirculation through Fram Strait [cf. Bauch and Bauch, 2001] might have added to the difference between low stadial and high interstadial δ13C values in addition to a varying proportion of Arctic BSW. The above reasoning is in line with the confirmation of a reduced thermodynamic imprint in the Southern Ocean δ13CDIC when during the Last Glacial Maximum, sea ice coverage was expanded [Mackensen, 2012]. It is also supported by an investigation of benthic foraminiferal isotope records from Storfjorden (South Svalbard), which suggests that only brines formed from cold and salty waters with high stable isotope values are sufficiently dense to contribute to deep water formation [Rasmussen and Thomsen, 2009].
Mean δ13C of ABW in the Amerasian basins is higher than that of Eurasian basins' ABW, and overall by 0.2‰ higher than that of GIN Sea bottom water masses (Figures 3-5). Highest ABW δ13C in Amerasian basins may reflect lowest primary production and deep water organic matter decay, due to extensive sea ice coverage. Obviously, this overcompensates the fact that Canadian Basin Bottom Water is fairly isolated and the oldest bottom water in the Arctic Ocean.
I propose that in the deep Nordic seas below 2000 m water depth, low stadial benthic δ13C values were probably not caused by production of large-scale BSW on adjacent shelves, as was inferred from low δ18O values, but rather by a cessation of BSW formation in the Arctic Ocean, possibly accompanied by intensified deep and bottom water exchange through Fram Strait. If so, relatively low benthic δ13C values in stadials may not be attributed to brine formation but to less formation of high-δ13C brine-enriched water than in interstadials. This hypothesis is supported by high epibenthic δ13C values from the modern Weddell Sea continental slope and rise, which were recently documented and attributed to strong thermodynamic fractionation during bottom water generation off the Antarctic ice shelf [Mackensen, 2012]. However, further evaluation of oxygen isotope data of BSW and Recent epibenthic calcite from the Amerasian basins of the Arctic Ocean is needed to help clarifying whether Arctic BSW today carries a low-δ18O signature similar to brine-enriched high-salinity shelf water of the Weddell Sea [Mackensen et al., 1996; Dokken and Jansen, 1999; Mackensen, 2001].
It is worth noting that this study suggests a general positive deviation of >0.2‰ of epibenthic δ13C values from δ13CDIC values in the Arctic Ocean, in contrast to the well-known one-to-one relationship in the GIN Sea. If further investigation and more data on the δ13CDIC distribution in the Arctic Ocean confirmed a systematic deviation, this would have implications for deep water circulation reconstructions based on epibenthic δ13C.
Thanks are due to Heike Röben and Susanne Wiebe for sediment sample preparation and selection of benthic foraminiferal individuals. I thank Günter Meyer and Lisa Schönborn for running and maintaining our mass spectrometers, Seung-Il Nam (Incheon, Korea) for benthic foraminiferal samples from the East Siberian continental margin taken in the course of project PP11070, Truls Johannessen (Bergen, Norway) for providing some δ13CDIC determinations from the central Arctic Ocean, as well as Seung-Il Nam, Jens Matthießen, and Reiner Schlitzer for discussion. Advice and helpful suggestions for improvement of reviewers and the Editor are gratefully acknowledged.