Interactive comment on “ Towards an improved organic carbon budget for the Barents Sea shelf , marginal Arctic Ocean ” by I . Pathirana

Introduction Conclusions References


Introduction
Despite the undisputed role of the Arctic Ocean in the modern climate system, the Arctic has only recently attracted significant attention as the public has become aware that ongoing, fundamental change in the Arctic cryosphere could be a response to global warming (IPCC, 2007).The changes in the cryosphere are shown in enhanced loss in multi-year sea ice, snow cover and permafrost thawing.The effects of these dramatic changes on the biogeochemical cycle in the Arctic Ocean and particularly on its adjacent shelf areas are currently a matter of intense discussion (Serreze et al., 2007;Wassmann et al., 2011;Arrigo et al., 2012).For instance, the continental shelves of the Arctic Ocean are important components of the global carbon cycle and may be responsible for 7-11 % of the global carbon sequestration in the ocean (Hedges and Keil, 1995;Stein and Macdonald, 2004b).
A proper quantitative understanding of the modern organic carbon storage capac- therefore essentially important.The Barents Sea, for example, is regarded as one of the most productive Arctic Ocean shelf seas that supports one of the world's richest fisheries (Sakshaug and Kovacs, 2009).Average annual gross primary production estimates range from 20 gC m −2 yr −1 in the seasonally ice-covered northern part to > 150 gC m −2 yr −1 in the Atlantic Water influenced southern Barents Sea (Wassmann et al., 2010).Numerical modelling of gross primary productivity with variable sea ice coverage in the Barents Sea reveals that a decrease in ice cover and increase in surface water temperature will lead to an increase in production in the northern Barents Sea area up to 100 gC m −2 yr −1 while production in the southern Atlantic Water region will decrease by 15-25 % (Ellingsen et al., 2008;Slagstad et al., 2011).To date, the highest variability is found in the marginal ice zone (MIZ, 50 to > 100 gC m −2 annually, Wassmann et al., 2010).Vertical carbon fluxes are highly variable (Olli et al., 2002;Reigstad et al., 2008Reigstad et al., , 2011) ) and pelagic-benthic coupling in the region is strong (Wassmann et al., 2006a).Annually approximately 32 gC m −2 yr −1 (Arctic Water) to 44 gC m −2 yr −1 (Atlantic Water) is exported below 90 m (Reigstad et al., 2008).
Maps of organic carbon content in surface sediments (Knies and Martinez, 2009) show the highest concentration of total organic carbon in the MIZ (> 2 wt.%) and lowest organic carbon concentration in the southern Barents Sea (< 1 wt.%).The organic carbon content in the ice-free region is mainly controlled by in situ produced marine organic matter.However, due to the proximity to the MIZ and in turn transfer of large amounts of land-derived inorganic and organic matter through melting sea ice, organic matter deposited in the shelf sediments below the MIZ comprises mixtures of marine (autochthonous) and terrestrial (allochthonous) sources (Vetrov and Romankevich, 2004;Winkelmann and Knies, 2005;Knies and Martinez, 2009).Hence, in order to provide a robust mass balance and eventually inferences on CO 2 sequestration for modern and future environmental scenarios, a geochemical characterization of its multiple potential sources is essential.
Additionally, the maps of organic carbon content in surface sediments and organic carbon accumulation in the central and eastern Barents Sea presented by Vetrov and  Romankevich (2004) were based on a large sample database and constructed accounting for topography, grain size and hydrographic conditions.This approach, however, lacks direct links to the sources and supply of the organic matter.Linking sedimentary data to physical and biological parameters on a regional scale can be done with a numerical model.In geology and petroleum related research, models exist that are used to back-calculate possible ranges of paleoproductivity in relation to organic carbon content in the sediment and are used to identify and quantify potential petroleum source rocks (e.g.Schwartzkopf, 1993;Knies and Mann, 2002).One of these models is OF-Mod 3D, a predictive, process-based, forward-modelling tool to calculate organic matter deposition and preservation in a 3-D grid throughout the modelled domain (Mann and Zweigel, 2008).One caveat is, however, that for ancient deposits not all input parameters are well-constrained and calibration of the models and evaluation of the results is difficult (Tommerås and Mann, 2006;Mann and Zweigel, 2008).With modern sedimentary data, close calibration can be achieved and information about the input and modelled parameters inferred.
The purpose of this paper is to introduce OF-Mod 3D as a tool for (sub-) recent sediment studies to provide a regional picture of the (marine and terrigenous) organic carbon fractions and marine paleoproductivity beyond core control and to provide regional calculations of organic carbon burial.Unlike other efforts providing estimates of organic carbon mass balance based on organic source estimates from low-resolution sample grids (Stein and Macdonald, 2004b;Kuzyk et al., 2009), the present study builds on a well-constrained organic facies model with a grid size of 625 km × 1000 km resolving 10000 yr in 15 vertical layers that is calibrated against a comprehensive and analytically consistent characterization of sedimentary organic matter across the MIZ in the western Barents Sea.By applying different chronological constraints on a selected set of sediment cores across the MIZ, we (1) validate a Holocene thickness map inferred from seismic interpretation (Gurevich, 1995), (2) calibrate the organic facies model for estimation and quantification of terrigenous and marine organic matter supply as well Introduction

Conclusions References
Tables Figures

Back Close
Full as marine primary productivity, and (3) finally discuss the accumulation rates of various organic matter sources and their burial on the western Barents shelf.

Study region
This study is carried out in the western Barents Sea between northern Norway and Svalbard including the shelf edge region.Figure 1 gives an overview over the study A detailed description of the water masses and circulation regime can be found in Loeng (1991).The North Atlantic Drift brings warm, saline Atlantic water (AW) into the Barents Sea from the southwest flowing northward along the shelf and branching out eastward into Bear Island Trough.Cold, fresh Arctic water (ArW) enters the Barents Sea from the northeast and flows southwestward along the flanks of Spitsbergen Bank.
The ArW from the northeast and the AW from the southwest are separated by a density barrier, the Polar Front (PF).Its position is mainly topographically controlled following the 250 m isobaths (Loeng, 1991) but also depends on the relative strengths of the two water masses.The northern part of this region is partially covered by sea ice in the winter.Melting of the ice in spring and summer together with increased insolation and heat leads to a stratified water column that induces a phytoplankton bloom that follows the receding ice edge northward (Sakshaug and Skjoldal, 1989).
Also shown in Fig. 1 and all following maps is the maximum southernmost ice extent in the western Barents Sea over the last 250 yr to distinguish between the iceinfluenced northern and ice-free southern part of the study region.3 Material and methods

Geochemical and sedimentological analysis
The calibration dataset consists of 190 surface sediment samples (described in Knies and Martinez (2009) + additional samples) and 6 short sediment cores (ca.30-40 cm) representing the last 50-3000 yr (locations shown in Fig. 1, data available at http: //doi.pangaea.de/10.1594/PANGAEA.817232).The material was collected during various cruises between 2001 and 2006 (Winkelmann and Knies, 2005;Knies et al., 2006;Jensen et al., 2007;Knies and Martinez, 2009;www.mareano.no).All cores were taken with multicorer equipment.Undisturbed surfaces of all short cores (first centimeter of core depth) were sampled and stored at −20 • C until analysis.Thereafter, all samples were freeze-dried and homogenized prior to analyses.Details on the applied analytical techniques are described in Knies et al. (2006) and Jensen et al. (2007).A brief overview over the geochemical methods for new samples in this study is given below.Grain size was measured on freeze-dried samples by wet sieving (size fraction > 2 mm diameter) and Coulter counter laser diffraction (< 2 mm diameter) on a Coulter LS 2000.Grain size distribution was determined as volume percent assuming uniform density.The grain size is reported here as sand fraction, where sand was defined as any material with a diameter > 63 µm, so sand fraction = 1 means only coarse material > 63 µm while sand fraction = 0 means only material < 63 µm.
Total organic carbon (TOC in weight percent, wt.%) was determined using a LECO CS 244 analyzer.Aliquots (200 mg or 500 mg) of the samples were treated with 10 % Introduction

Conclusions References
Tables Figures

Back Close
Full (volume) hydrochloric acid (HCl) at 60 • C to remove carbonate and washed with distilled water to remove excess HCl.Possible loss of organic material due to acid leaching is not taken into account.
Stable isotope ratios of the organic carbon fraction (δ 13 C org ) and the nitrogen fractions were determined by elemental analyzer isotope ratio mass spectrometry (EA-IRMS) on a Europa Scientific RoboPrep-CN elemental analyzer by Iso-Analytical, Crewe, UK, following the procedure described in Knies et al. (2007).δ 13 C org was determined on decarbonated samples.Total nitrogen was determined on aliquots of freezedried, homogenized samples, while inorganic nitrogen was determined on KOBr-KOH treated aliquots following Silva and Bremner (1966).Twenty percent of the samples were measured in duplicate.Organic nitrogen (N org ) was calculated as the difference between total nitrogen (N tot ) and inorganic nitrogen (N inorg ).

Endmember mixing model
To distinguish between marine and terrestrial organic matter, marine and terrestrial endmembers need to be assigned to a two-endmember mixing model typically using total organic carbon (TOC) and nitrogen content, kerogen microscopy, stable isotopes of organic matter (δ 13 C org ), Rock Eval pyrolysis (hydrogen index), or various biomarker data in the sediments (see e.g.Jasper and Gagosian, 1990;Stein, 1991;Stein and Macdonald, 2004a;Knies et al., 2007;Mann and Zweigel, 2008).In this study we use δ 13 C org and the percentage of organic nitrogen contained in a sample to quantify the proportions of marine and terrestrial organic carbon.δ 13 C org has been shown to be a reliable proxy to determine the proportion of terrestrial organic matter in Arctic marine sediments (e.g.Schubert and Calvert, 2001;Knies and Martinez, 2009).Typical values for C3 plant derived terrestrial organic δ 13 C org in high latitudes are −25.5 % to −29.3 % with an average of −27 % while the addition of C4 organic matter is less important in these regions (see e.g.Stein and Macdonald, 2004a;Knies and Martinez, 2009 for further references).The marine δ 13 C org -derived Introduction

Conclusions References
Tables Figures

Back Close
Full endmember values in the Spitsbergen/Barents Sea region ranges between −20.3 % and −21 % (Winkelmann and Knies, 2005;Knies and Martinez, 2009).Knies and Martinez (2009) showed that the marine nitrogen endmember is represented by its organic fraction, i. e. %N org (of total) = 100 %, whereas the terrestrial nitrogen component is likely composed of a mixture of organic and inorganic nitrogen bound as ammonium in the clay matrix and/or supplied by soil (terrestrial) organic matter (Winkelmann and Knies, 2005;Knies and Martinez, 2009).The latter is supported by Knies et al. (2007) showing that the inorganic fraction of the total nitrogen content in surface sediments off Spitsbergen can be used as proxy for terrestrial organic matter input.Further, Mann et al. (2009) compared the percentage of inorganic nitrogen (%N inorg = 100 − %N org ) to hydrogen index and maceral data in central Arctic Ocean Paleogene deposits under the assumption that N inorg in marine sediment is non-local, i.e. allochthonus and of terrestrial origin, and found that soil (terrestrial) organic matter contains ca.30-100 % N inorg .
This study uses the same δ 13 C org endmembers as Knies and Martinez (2009).A linear regression analysis of δ 13 C org vs. N org /TOC gives a terrestrial δ 13 C org endmember of −26.1 % .A linear regression analysis of %N org and δ 13 C org gives a marine δ 13 C org endmember of −20.1 % at 100 % N org (Fig. 2a and b).
For %N org , the marine endmember is defined as 100 %.To obtain the terrestrial %N org endmember we follow a procedure analogous to Jasper and Gagosian (1990).
According to Jasper and Gagosian (1990), typical values of N tot /TOC for sediments of terrestrial origin are 0.01-0.05,and 0.13-0.20 for marine origin.In a regression of N tot /TOC vs. %N org , we obtain the terrestrial %N org endmember at N tot /TOC = 0.05 resulting in %N org,terr = 17 % organic nitrogen (Fig. 2c).The latter supports the overall inference by Knies et al. (2007) that the inorganic proportions of the total nitrogen are applicable as proxy for quantifying the terrestrial organic matter in the European sector of the Arctic.Introduction

Conclusions References
Tables Figures

Back Close
Full

Age models and sedimentation rates
Age models and sedimentation rates were determined by 210 Pb, 137 Cs and/or AMS 14 C measurements.A list of cores pertaining to this study (BASICC1, BASICC8, St1245, R87MC006, R87MC006, R1MC85 and St20) and other published age data in the study region is provided in Table 1, together with the dating methods used to obtain the sedimentation rates.The age models for cores BASICC1 and BASICC8 are based on 210 Pb measurements and are described in Vare et al. (2010).The age model for St1245 is based on 137 Cs and AMS 14 C measurements and is described in Winkelmann and Knies (2005). 210Pb measurements of cores R87MC006 and R1MC85 are described in Jensen et al. (2007Jensen et al. ( , 2008)).AMS 14 C measurements on cores R87MC006 and St20 were done on shells, shell fragments and foraminifera.The measurements were performed by 14 Chrono Centre, Queens University, Belfast, UK, and calibrated using the CALIB 6.02 software (Stuiver and Reimer, 1993) with the Marine09 (R87MC006) and IntCal09 (St20) calibration curves (Reimer et al., 2009) and a ∆R of 0. Table 2 summarizes these radiocarbon ages.

OF-Mod 3D model set-up
OF-Mod 3D (Organic Facies Model 3D) simulates the deposition and burial of organic carbon on a basin scale, and is based on the interaction between inorganic and organic basin fill, as well as preservation of organic material.The organic part of the model is described in more detail in Mann and Zweigel (2008), while the basin fill modelling approach used here is different than in that paper.The inorganic basin fill is modelled based on the present-day depth and bathymetry maps.The lithology (sand fraction) distribution is calculated based on the spatial distribution of sedimentary facies (Felix et al., 2012).The sedimentary facies are determined with a set of fuzzy logic rules, where the facies are defined based on water depth and distance to shore (from the bathymetry maps).A sand fraction is assigned to each facies and the spatial distribution is calculated using Sugeno rules (e.g.Demicco and Klir, 2004).The facies rules Introduction

Conclusions References
Tables Figures

Back Close
Full are adjusted so that the modelled sand fraction values fit with the measured values.Sedimentation rate is determined through decompaction of the present-day deposits, whose thickness is given in OF-Mod using present-day depth maps.
To build the stratigraphic infill model, two bounding surfaces (top and bottom) were used.For the top surface the present day IBCAO bathymetry, Version 2.23 (Jakobsson et al., 2008), in this region was used.The bottom surface was constructed by combining the present day bathymetry with the thickness of Holocene sediments in this area (map by Gurevich, 1995), defining the Holocene to cover the last 10 000 yr. Holocene sediment thickness in the southwestern Barents region (not covered by the Gurevich, 1995, map) was inferred by extending the Gurevich (1995) map through calculating sediment package thicknesses from published sedimentation rates in the SW corner of the study region.The model grid consists of 125 × 200 cells and 15 vertical layers between 0 kyr (top layer) and 10 kyr (bottom layer) with a grid cell size of 5000 m.
The organic carbon (OC) is split into three different fractions: marine, terrigenous, and residual organic carbon.This results in a slightly different allocation of MOC and C terr than the two-endmember approach used for the measured values because part of both fractions has been assigned to the residual fraction.The residual fraction needs to be taken into account in OF-Mod because otherwise it would not be possible to model the low hydrogen index values associated with degraded material.Input of terrigenous and residual organic carbon is given directly in weight percentage.Marine organic carbon deposition and burial is calculated from the carbon flux from primary productivity at the sea surface, combined with an equation for the burial efficiency of organic carbon at the seafloor.where BE = burial efficiency (fraction), and LSR = sedimentation rate (cm kyr −1 ).Burial of all three types of organic carbon is lithology dependent: MOC and residual OC have higher preservation in finer grained deposits, while terrigenous OC tends to be preserved better in coarser grained deposits (Bergamaschi et al., 1997;Keil et al., 1998).
In OF-Mod the input of all three types of organic material is a combination of a basinwide trend and local increased input.Here a low background PP (35 gC m −2 yr −1 ) is used throughout the model region, and the processes related to the ice margin in the MIZ are represented by additional PP local input in the northern part (50 gC m giving a total PP of 85 gC m −2 yr −1 in this region.These PP values were calibrated to reproduce the TOC content of the surface sediments.A summary of the relevant input parameters is given in Table 3. To evaluate the performance of the model in replicating the calibration data, a simple goodness-of-fit procedure between the measured sediment samples and the closest model grid points is employed on the sand fraction and TOC data.A linear regression between the residuals (absolute difference between the model and the measurements) and the measured data is minimized.

Validation of the Holocene thickness map in the Barents Sea
Figure 3 shows the extended Holocene sediment thickness map and sediment thicknesses calculated for a number of core locations in the western Barents Sea (see Table 1 for sedimentation rates and references).Using the published sedimentation rates, Introduction

Conclusions References
Tables Figures

Back Close
Full expected sediment accumulated during 10 000 years was calculated.Most calculated sediment thicknesses agree well with the map (Fig. 3).There are, however, a number of locations with very high calculated thickness values (> 10 m).These seem unrealistic especially when taking compaction into account.When looking at the different dating methods, it becomes clear that overall those cores with AMS 14 C dates agree with the map and the reported thickness, whereas cores dated with other isotopes do not fit.This study includes one core, i.e.R87MC006 (see Table 1) that was dated with both AMS 14 C and 210 Pb.The sedimentation rate estimate of R87MC006 from 210 Pb is 80 cm kyr −1 based on the top 5 cm.The sedimentation rate estimate from an AMS 14 C date at 18.5 cm depth is 4 cm kyr −1 .These two estimates differ by an order of magnitude and the question arises how to interpret and if possible reconcile these two rates.
14 C has a much longer half-life (ca.5600 yr) and is the suitable isotope for this date range.However, many short box and multicores, presumably representing the most recent past, are routinely dated only with 210 Pb (half-life of 22.3 yr) or 137 Cs (half-life of ca. 30 yr) (e.g.Carroll et al., 2008;Kuzyk et al., 2009;Vare et al., 2010).The 210 Pb content of a sediment is usually measured on the fine fraction of the sediment since it is particle reactive (Soetaert et al., 1996).On the other hand 14 C is measured on shells, shell fragments or foraminifera, which are generally much larger in diameter than the fine sediment.Thus 210 Pb is comparatively more susceptible to processes happening within the sediment column, e.g.sediment mixing and bioturbation, than When comparing 210 Pb and AMS 14 C age models of a core from the Sea of Japan, Crusius and Kenna (2007) concluded that under low sediment accumulation conditions, bioturbation and mixing may dominate the processes in the sediment column.
In that case, 210 Pb may be more indicative of mixing and bioturbation than sediment accumulation over a longer period of time.
The Quaternary sediment cover in the Barents Sea is generally thin (Elverhøi and Solheim, 1983;Vorren et al., 1989;Gurevich, 1995).The seafloor is hard and glacial features like ridges, moraines, or glacial lineations can be readily identified on highresolution backscatter images (e.g.Vorren et al., 1989;Bellec et al., 2010;Winsborrow et al., 2010;Rüther et al., 2011).In many cases the reported thickness of Holocene sediments deposited since the last glaciation is less than 4 m (see examples in Table 2 and e.g.Elverhøi and Solheim, 1983;Gurevich, 1995;Rüther et al., 2011).Taking into account potentially missing core tops, 5 m of accumulated sediments in 10 kyr results in a sedimentation rate of 50 cm kyr −1 or 0.05 cm yr −1 and only 10 cm of accumulated sediments in 200 yr, the applicable time span for 210 Pb.We believe that sedimentation rates estimated with 210 Pb in the Barents Sea are exaggerated and the sedimentation conditions are better represented by AMS 14 C estimates.
As a result, only AMS 14 C sedimentation rates are considered reliable in this study.
The sediment thicknesses calculated from AMS 14 C sedimentation rates and the sediment thickness map (Gurevich, 1995) agree well.The extended Holocene sediment thickness map is accurate for the study region and sufficient for the purpose of this study.

Inorganic sediment characteristics
Sedimentation rates for the Holocene estimated from seismic data in the Barents Sea generally range from 1 to 100 cm kyr −1 , with higher rates (up to 500 cm kyr Romankevich , 2004).This pattern corroborates AMS 14 C based LSR calculations in various sediment cores in the Barents Sea (Table 1) and confirms previous inferences of a relatively good agreement between seismic interpretation and AMS 14 C-based chronologies of Holocene sequences (Elverhøi et al., 1989;Lebesbye, 2000).The core data indicate the highest LSR (> 75 cm kyr −1 ) occurring preferably in the glacial troughs (e.g.Bear Island Trough, Storfjorden Trough), in fjords (e.g.Storfjorden), and below the MIZ.Relatively low LSR is generally observed on shallow banks (e.g.Spitsbergen Bank, Tromsøflaket). Figure 4 shows the OF-Mod modelled LSR compared to LSR from sediment cores in the region.The modelled LSR agree particularly well with the 14 C dated cores in the southwestern Barents Sea.The mismatch between calculated and modelled LSR in northern Norwegian fjords is due to the large grid cell size for the coastal regions.For each grid cell one average value for the respective region is calculated.Grid cell size determines the area over which the average is calculated and therefore the resolution of the model.This model has a grid cell size of 5000 m × 5000 m.Large variability over a region smaller than that, including many fjords, is not reflected in the model outcome.Towards the MIZ, calculated LSR from 14 C dated cores are slightly higher (30-45 cm kyr −1 ) compared to the modelled LSR (0-30 cm kyr −1 ), however still on the same order of magnitude.Generally, the modelled LSR follows the bathymetry in the Barents Sea, and is not controlled by the MIZ.Similar to the interpreted seismic data (Gurevich, 1995), the modelled LSR is highest in morphological depressions and fjords, the natural sediment depocenters in the western Barents Sea (Faleide et al., 1993).Hence, the resulting OF-Mod 3D map (and thus the construction method of the Holocene bottom surface map used in OF-Mod 3D) can be considered to represent a realistic view on the Holocene sedimentation rate for the entire study region.
Figure 5a shows the distribution of the sedimentary sand fraction (proportion of coarse material > 0.63 µm, 0 ≤ SF ≤ 1) compared to the modelling results.The measured sediment data show that most of the study region (east and north) consists of very fine material (SF < 0.2), with the bulk of the coarser material found along the shelf break (SF > 0.4) and in the coastal region north of Norway (SF 0.2-0.8).The sand-rich Introduction

Conclusions References
Tables Figures

Back Close
Full deposits along the shelf edge are believed to be relic facies originating from sediment supplied by the Barents ice sheet to the shelf edge during sea level low stands during the last glaciation and erosion and re-deposition during deglaciation (Vorren et al., 1989;Bøe et al., 2009).The deeper parts of Bear Island Trough are highly variable (SF 0.2-0.8)whereas mainly fine-grained material accumulates in Storfjorden Trough.
There are some single higher values (SF 0.4-0.6) in the southeast region and east of Svalbard.Spitsbergen Bank consists mainly of coarse carbonates (Elverhøi and Solheim, 1983;We ¸sławski et al., 2012).The modelled sand fraction replicates the calibration data quite well.Figure 5b shows the regression analysis of the difference between the modelled SF and the measurements against the calibration data.The low correlation of the residuals compared to the measured values (R 2 = 0.35) indicates a good fit between the model and the measured data.Figure 5c shows the spatial distribution of the residuals plotted on top of the modelled sand fraction.No spatial trend is visible in the distribution of the residual magnitudes.The fine sediment in the Storfjorden Trough, coarser sediment in Bear Island Trough and coarse shelf break are represented well in the sand fraction model.The relic sand deposits are modelled through a separate facies but slightly overestimate the sand content.The model provides less detail north of the coastline of the Norwegian mainland and the modelled values are finer than the measured ones.This is due to resolution.The model resolution is coarser than the small scale variations in the point samples and the latter cannot be replicated in detail.
The resulting modelled grain size map also agrees well with the sediment distribution map by Elverhøi (1984) which is based on seismic and sparker data and sediment samples.Overall the inorganic, stratigraphic part of the model provides a consistent framework for the organic modelling efforts.

Organic matter characteristics
In the following only the results for the top layer, i.e. the present day/sediment surface, are discussed and compared to the surface sample dataset.Figure 6a shows the TOC distribution of the surface sediment data compared to the model results.shows the results of the goodness-of-fit test and Fig. 6c the spatial distribution of residual magnitudes plotted on top of the modelled TOC distribution.In general, the sediments exhibit a lower TOC content south of the MIZ (0.2-1.8 wt.% TOC) than within the MIZ, especially along the flanks of Spitsbergen Bank, and in the coastal regions (1.8-3 wt.% TOC).The TOC content of the sediments is also low (< 0.8 wt.% TOC) along the shelf break.There are some higher values (1.3-1.8 wt.%) upslope of Bear Island Trough.The OF-Mod results are in the same range (0-< 4 wt.%TOC) as the sediment samples and agree well with the observed (this study and Knies and Martinez, 2009) and previously published data (Stein et al., 1994;Vetrov and Romankevich, 2004).The regression between the residuals and the model results shows only a weak correlation (R 2 = 0.2) indicating a good model fit.The higher TOC content in the north and in the MIZ is captured, as well as the highs along coasts and maximum values along the flanks of Spitsbergen Bank.The low TOC content south of the maximum ice extent and along the shelf break is represented well.The model slightly overestimates the TOC content north of the Norwegian coast compared to the calibration data due to coarse resolution.In addition the model underestimates the TOC content of the upslope part of the Bear Island Trough.This is partly due to overestimates of sand content in this region and thus less modelled organic deposition.Another reason could be lateral transport of sediments downslope Bear Island Trough after deposition, which OF-Mod 3D does not include.This has recently been suggested as a possibility to explain the occurrence of the novel organic geochemical biomarker (IP25), likely biosynthesized by a limited number of sea ice diatoms during the spring bloom (Brown et al., 2011), in surface sediments of the Barents Sea south of the maximum ice extent (Navarro-Rodriguez et al., 2013).
According to Vetrov and Romankevich (2004), TOC in Barents Sea sediments is mainly composed of marine organic matter but with ∼ 30 % of terrigenous origin.A predominantly marine source of the organic matter has been further confirmed by bulk geochemical analyses (Tamelander et al., 2006;Zaborska et al., 2008).The present study, however, reveals distinct spatial variability in both the amount and composition Introduction

Conclusions References
Tables Figures

Back Close
Full of the sedimentary organic carbon.The sediments exhibit low marine organic carbon (MOC) content south of the MIZ (0-1 wt.%, Fig. 7), that represents > 90 % of the total sedimentary organic fraction.In contrast, high MOC content (1-2 wt.%) is found in the MIZ and in Storfjorden Trough.Maximum MOC occurs along the flanks of Spitsbergen Bank (1.3-2 wt.%).OF-Mod reproduces the pattern of low MOC content in the south compared the north and the maximum MOC content along Spitsbergen Bank well (Fig. 7).A goodness-of-fit analysis is not performed on the MOC data because of the different organic matter fraction allocation approaches (see Sect. 3.4).OF-Mod 3D predicts the highest MOC content (> 3 wt.%)along the bottom of the southern flank of Spitsbergen Bank.
Figure 8 shows the primary productivity distribution used as input in OF-Mod 3D compared to 3 sediment cores.Calculation of primary productivity (in gC m −2 yr −1 ) for the core positions is based on the amount of MOC (wt.%), dry bulk density (DBD in g cm −3 ), linear sedimentation rate (LSR in cm kyr −1 ) and water depth (z in m) via the formula in Knies and Mann (2002), and references therein:  ice containing significant amounts of inorganic and terrigenous organic matter (Nürnberg et al., 1994;Reimnitz et al., 1994;Stein et al., 1994;Pfirman et al., 1997).Indeed, observations from sediment traps suggest that the release of terrestrial organic matter from melting sea ice in the Barents Sea provides significant contributions to the vertical export of organic matter (Andreassen et al., 1996;Wassmann et al., 2004).This picture is consistent with other Arctic shelf regions such as the Laptev and Kara seas (Stein and Macdonald, 2004c) and reflects the significance of terrigenous components for the organic carbon cycle in the Arctic Ocean.The pronounced gradient from the shore to the open ocean is attributed to glacio-fluvial and riverine transport of eroded sediments enriched in terrestrial organic matter from a densely vegetated drainage area off Spitsbergen (e.g.Knies et al., 2007).The distribution of C terr in the study area is apparently independent of water depth, re-mineralisation processes and various sedimentation rates.Having already experienced some degradation and microbial attack in soils and during transport processes before entering the marine system, the terrestrial organic matter might be resistant to further extensive degradation at sea (Hedges and Keil, 1995).The OF-Mod results for C terr distribution are lower than the sedimentary data in general, but still on the same order of magntitude (0-2.5 wt.%).The general pattern of low C terr in the southern part and higher C terr in the north with maximal values in the fjords of Svalbard is, however, captured well, as is the distribution in Bear Island Trough.OF-Mod also predicts some C terr accumulation on the NE flank of Spitsbergen Bank.The higher C terr content in the samples upslope of Bear Island Trough and Hopen Deep is not reproduced.In addition to supply and the mixing model, terrestrial organic matter distribution in OF-Mod 3D also depends on the distribution of the sand fraction of the sediment -with higher SF linked to higher C terr (up to a maximum SF value, set here to 0.75).Lower modelled SF (see Sect. 4.2.1)thus also results in lower modelled C terr .Additionally the mixing model in OF-Mod is threefold and allocates TOC into the marine, terrestrial and residual fractions in contrast to the two-endmember approach used on the sedimentary data.A mismatch between MOC modelled by OF-Mod and MOC Introduction

Conclusions References
Tables Figures

Back Close
Full calculated for the sediments, and similarly for C terr from OF-Mod and C terr for the sediments, is expected because part of the total organic matter content in the sediments is in the residual fraction.The C terr model results indicate that part of the terrestrial fraction in the model region indeed lies in the residual fraction.

Organic carbon budget of the western Barents Sea
The full 3-D model simulates organic carbon deposition throughout the Holocene -in the model defined as the last 10 000 yr.With the current set-up, the modelled temporal changes are minimal (e.g.only ±0.005 wt.% TOC, ±0.0015 wt.% MOC). Figure 10 shows a profile of TOC deposited in Storfjorden over the modelled time period as an example.There are no vertical, i.e. temporal, changes in organic carbon content.This implies stable depositional conditions in Storfjorden since the last deglaciation.The horizontal, i.e. spatial, variations are the only changes and are reflected in the maps discussed in the previous section.Using modelled sedimentation rates (LSR) and dry bulk density (DBD), total sediment mass accumulation rates (MAR) in the study region can be derived via MAR = LSR • DBD.Subsequently accumulation of organic matter (TOC, MOC, C terr ) can be calculated by multiplying MAR with the respective quantity (whose units are wt.%after all).The results are shown in Table 4 and Fig. 11.With the results from OF-Mod 3D, accumulation rates and budgets can be calculated for each organic matter fraction and for each model grid cell.Regional variations in organic matter accumulation can be studied and hotspots of carbon accumulation and regions of non-deposition can be identified.Additionally using the position of the maximum ice extent, the study region is divided into a northern, ice-influenced part and a southern ice-free part and separate budgets are calculated for these two areas.Total amounts of sediment and organic matter buried throughout the Holocene are re-calculated for the last 11 000 yr in order to be able to compare the results to values calculated by Stein and Macdonald (2004b) for the Barents Sea and other Arctic Ocean shelf seas (Stein and Fahl, 2000;Stein and Macdonald, 2004c;Kuzyk et al., 2009;Kivimäe et al., 2010).Introduction

Conclusions References
Tables Figures

Back Close
Full Total bulk sediment accumulation is maximal in the depressions like Storfjorden, Bear Island Trough and north of Norway.In total 35.4 × 10 6 t of sediment accumulate per year, corresponding to a total sediment input of 390 × 10 9 t throughout the last 11 000 yr.In comparison, Stein and Macdonald (2004b) provided estimates for bulk sediment accumulation for the whole Barents Sea (1597 × 10 3 km 2 ) based on Vetrov and Romankevich (2004), however leaving out the southwestern Barents Sea shelf edge.These rates (listed in Table 4) are approximately 7 times higher than our estimates for an area 3 times larger than our study region.The eastern Barents Sea includes deeper depressions and thicker sediment packages than the western part of the shelf (Gataullin et al., 1993;Gurevich, 1995;Polyak et al., 1995) that account for higher accumulation in comparison.This study predicts the highest accumulation rates of total organic carbon (TOC) in Storfjorden (> 500 mgC cm −2 kyr −1 ) whereas there is almost no accumulation on Spitsbergen Bank (Fig. 11a).This results in > 120 tC yr −1 TOC buried in Storfjorden annually.Vetrov and Romankevich (2004) presented accumulation rates of total organic carbon for the Holocene in the central and eastern Barents Sea, leaving out the southwestern Barents Sea region and the western shelf edge.Their approach was based on a large sediment sample dataset and mapping was done taking into account topography, grain size, and hydrology but excluding links to organic matter supply.They used the sediment thickness map by Gurevich (1995) and sedimentation rates based on that map and on seismic data.The model in this study is based on the same sediment thickness map with the addition of the southwestern Barents region (see Sects.3.4 and 4.1).In comparison, this study's TOC mass accumulation rate map has more detail and extends the Vetrov and Romankevich map in the western area towards the shelf edge.
Both approaches predict negligible organic carbon accumulation on Spitsbergen Bank.Introduction

Conclusions References
Tables Figures

Back Close
Full The total annual accumulation of organic carbon based on this study is 0.33 tC yr −1 yielding approximately 3.65 × 10 9 tC buried throughout the last 11 000 yr. Compared to Stein and Macdonald (2004b), based on Vetrov and Romankevich (2004), these rates account for approximately 10 % of the whole Barents Sea region (see Table 4).
The largest accumulation of marine organic carbon (MOC, Fig. 11b) also occurs in Storfjorden (> 300 mgC cm −2 kyr −1 corresponding to a burial flux of 60-90 tC yr −1 ) and in the northeastern part of the study region (Hopen Deep, 200-250 mgC cm −2 kyr −1 corresponding to 60 tC yr −1 of MOC buried annually).Both areas are in the ice-and phytoplankton bloom influenced northern MIZ region.Overall MOC accumulation is estimated at 0.23 × 10 6 tC yr −1 yielding a total of 2.49 × 10 9 tC buried throughout the Holocene.These numbers also represent ca. 10 % of the total MOC accumulation calculated by Stein and Macdonald (2004b) (Table 4).MOC accumulation is nearly absent on Spitsbergen Bank.Spitsbergen Bank lies in the middle of the MIZ and primary productivity models show that marine primary production in the waters above Spitsbergen Bank is high (Wassmann et al., 2006b;We ¸sławski et al., 2012).We ¸sławski et al. ( 2012) postulate that Spitsbergen Bank may be a significant sink for marine organic carbon and a source of regenerated nutrients through recirculation and pumping of water through the coarse substrate.Our model shows some accumulation of marine organic matter on the southern flank of the Bank.But according to our modelling results, overall, Spitsbergen Bank is not a significant sink for organic material.
With the exception of coastal Storfjorden, C terr accumulation in the western Barents Sea is very low (< 100 mgC cm −2 kyr −1 on average, Fig. 11c) corresponding to an overall burial rate of 0.1 × 10 6 tC yr −1 and a total of 1.16 × 10 9 tC buried throughout the last 11 000 yr.The highest values of C terr accumulation in Storfjorden are a combination of input from land and substrate and may be in part overestimates by the model.
The western Barents shelf edge region missing in Vetrov and Romankevich (2004) and Stein and Macdonald (2004b) is included in this study and indicated in Fig. 3.The total area of this region is 103 × 10 3 km 2 .The region lies in the southern Atlantic Water Introduction

Conclusions References
Tables Figures

Back Close
Full influenced Barents Sea and receives ca.3.0×10 6 t of bulk sediment per year (Table 4).
However the burial flux of organic matter in this region is low (0.022 × 10 6 tc yr −1 for TOC, 0.015 × 10 6 tC yr −1 for MOC and 0.007 × 10 6 tC yr −1 for C terr ).Adding the results for the SW shelf edge to the total Barents Sea budget by Stein and Macdonald (2004b) (Table 4) provides a refinement of the Barents Sea carbon budget by including all relevant areas of the entire Barents Sea while only slightly increasing the overall budget.
Compared to other Arctic Ocean shelf seas (Table 4), the Laptev and Chukchi Seas have similar areal extents as the western Barents shelf in this study.The western Barents shelf receives less sediment (ca.half) than the Laptev Sea and less TOC (ca.one third) is buried annually.However, on the western Barents shelf ca. 3 times more MOC is buried annually (Stein and Macdonald, 2004b).Compared to the Chukchi Sea, the western Barents shelf receives ca. 2 times more sediment and similar amounts of C terr are buried annually, whereas the Barents shelf receives more TOC and more MOC is buried here (Stein and Macdonald, 2004b).The total area of Hudson Bay is 841 000 km 2 whereas only 125 000 km 2 receives active sedimentation (Kuzyk et al., 2009).The smaller area, however, receives 3 times more sediment than the western Barents shelf and more TOC and MOC is buried in Hudson Bay (Kuzyk et al., 2009).Kivimäe et al. (2010) give an estimate of modern (last 150 yr) organic carbon accumulation in the whole Barents Sea region of 9.2×10 6 tC yr −1 based on Carroll et al. (2008).This work is based on 210 Pb sedimentation rates in the Hopen Deep area, where our model also predicts higher organic carbon, especially MOC, accumulation.However, 9.2 × 10 6 tC yr −1 is a much higher value than any of the Holocene age studies of the Arctic Ocean shelf seas predict.
Comparing the southern, permanently ice-free region (area south of the maximum ice extent) to the seasonally ice-covered and ice-edge bloom influenced region (north of the maximum ice extent) yields the following results (Table 4).Accumulation of organic carbon in the ice-free southern region is low (< 80 tC yr −1 TOC, < 40 tC yr ice-covered northern region.In total there is more bulk sediment accumulated annually south of the maximum ice extent (19.4 × 10 6 tC yr −1 ) in a slightly larger area (288 × 10 3 km 2 vs. 262 × 10 3 km 2 ) than in the north (16 × 10 6 tC yr −1 ) but less organic material stored in the south (0.14 × 10 6 tC yr −1 compared to 0.19 × 10 6 tC yr −1 in the north).
This distribution reflects our modelled primary productivity (Fig. 8).The higher PP in the north also results in higher storage even though there is no accumulation on Spitsbergen Bank.However, our modelled PP differs from ocean-ecosystem primary productivity models (Ellingsen et al., 2008;Wassmann et al., 2010).These models predict highest PP in the southern, Atlantic Water influenced region, higher production on top of Spitsbergen Bank and overall low PP in the northern, Arctic Water influenced region.Their predicted PP values are also higher than our reconstructed PP: > 100 gC m −2 yr −1 predicted in the south compared to < 50 gC m −2 yr −1 reconstructed; < 100 gC m −2 yr −1 predicted in the north compared to up to 90 gC m −2 yr −1 reconstructed.On the other hand, modelling of PP in a future, ice-free scenario with higher air and sea surface temperatures in the Barents Sea as compared to today, suggests an increase in PP in the northern, Arctic Water domain opposed to a decrease of production in the southern, Atlantic Water domain (Ellingsen et al., 2008;Slagstad et al., 2011).This PP distribution pattern seems on a first glance similar to our Holocene-average PP distribution.Since our sedimentary model averages over considerably longer time spans than oceanecosystem models (thousands of years compared to monthly/seasonal) and covers the last 10 000 yr, our PP distribution could be reflecting an overall warmer scenario.
On the whole the Holocene in the Barents Sea region has been warmer than today with a the mid-Holocene thermal maximum and a cooling trend with a decrease in sea surface temperatures and an increase in sea ice cover in the last 5000 yr BP until the start of the industrial period (Sarnthein et al., 2003;Renssen et al., 2005;Wanner et al., 2008;Risebrobakken et al., 2010).
On the other hand, the reconstructed PP presented here is a reflection of the bottomup approach in OF-Mod 3D back-calculating necessary input of organic matter from Introduction

Conclusions References
Tables Figures

Back Close
Full the material that was actually deposited.Organic carbon content of the sediments and primary productivity in the surface waters are linked by vertical export of organic matter out of the productive zone (see Sect. 3.4).This is taken into account in OF-Mod 3D (see Sect. 3.4, Eqs. 1 and 2).However, the western Barents shelf is relatively shallow on average (200 m) and the links to vertical export out of the euphotic zone may be stronger than currently implemented in OF-Mod 3D.Vertical carbon fluxes are highly variable (Olli et al., 2002;Reigstad et al., 2008Reigstad et al., , 2011) ) and pelagic-benthic coupling in the region is strong (Wassmann et al., 2006a).The higher organic carbon content of the northern, ArW and ice-influenced region could be a reflection of a highly variable primary production regime with efficient vertical export and less recycling of nutrients in the water column than in the southern Barents Sea and more efficient burial than currently implemented.

Conclusion
This study shows that OF-Mod 3D is a valuable tool for regional modelling of the distribution of the marine and terrestrial organic carbon fractions and for reconstructing of organic matter and support observations of various shelf areas in the Arctic Ocean (Stein and Macdonald, 2004c).Reconstructed primary productivity is highest in the MIZ region and low throughout the southwestern Barents Sea.
Accumulation of organic carbon in the ice-free southern region is lower than in the seasonally ice-covered northern region.In total there is more bulk sediment accumulated annually south of the maximum ice extent but less organic material.Adding the results for the SW shelf edge to the total Barents Sea budget by Stein and Macdonald (2004b) provides a refinement of the Barents Sea carbon budget by including all relevant areas of the entire Barents Sea while only slightly increasing the overall budget.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full ity and future changes under variable physical conditions (less or no sea ice) is Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Based on the dataset March through August ice edge positions in the Nordic Seas 1750-2002 by Divine and Dick (2007), the ice edge is defined as the outer boundary for 30 % ice Discussion Paper | Discussion Paper | Discussion Paper | concentration.Monthly mean ice edge positions for the years 1750-2002 were compared and the maximum southernmost extent determined.The location of the maximum southernmost ice extent constructed in this way corresponds closely to the maximum sea ice extent in Navarro-Rodriguez et al. (2013) based on NSID and Met Office Hadley Centre ice charts.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | The flux is described by the equation of Betzer et al. (1984) CF = 0.409 • z −0.628 • PP 1.41 , where CF = organic carbon flux to the sediment surface (gC m −2 yr −1 ), PP = primary productivity (gC m −2 yr −1 ), and z = water depth (m).The burial efficiency at the seafloor is calculated using the equation of Betts and Holland (1991) Discussion Paper | Discussion Paper | Discussion Paper | log 10 (BE) = 1.39 • log 10 LSR log 10 (LSR + 7.9) + 0.34, Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 14 C (Soetaert et al., 1996).Mixing of the sediment and bioturbation spread the particle reactive radioisotopes ( 210 Pb and also 137 Cs) over a larger part of the sediment column than would be the case without disturbance and reduces the overall concentration of the isotope in the sediment (Johannessen and Macdonald, 2012).The depth at which the apparent 210 Pb background concentration is determined is deeper and the resulting sedimentation rate higher than without any mixing.Thus 210 Pb (and 137 Cs) age and sedimentation rate estimates are always maximum limits (Johannessen and Macdonald, 2012).Discussion Paper | Discussion Paper | Discussion Paper | −1 ) occurring in bathymetric deeper depressions, e.g.glacial troughs (Gurevich, 1995; Vetrov and Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 1) Primary productivity calculated from the core data reveals values of 80-110 gC m −2 yr −1 within the MIZ and < 20 gC m −2 yr −1 south of the MIZ.The model input productivity ranges from 30 gC m −2 yr −1 south of the MIZ to 90 gC m −2 yr −1 within the MIZ.

Figure 9
Figure 9 shows the terrestrial organic carbon (C terr ) content of the sediment data compared to the model results.The distribution pattern of C terr exhibits clear spatial trends with the highest values (up to 1.82 wt.%) in the fjords and gradually lower values towards the ice-free, open-ocean environment in the southwestern Barents Sea.The lowermost values occur along the shelf break.The higher C terr values in sediments below seasonally ice-covered areas are explained by melting of sediment laden sea 4955 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 40 tC yr −1 C terr compared to 80-> 180 tC yr −1 TOC, 40-90 tC yr −1 MOC, > 40 tC yr −1 C terr in the north).More organic matter is buried annually in the seasonally Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | primary productivity beyond core control on (sub-) recent time scales.The model is calibrated and the model results represent the surface sediment data well.Modelled sedimentation rates agree with published data and modelled lithology closely reflects the sediment data.Modelled total organic carbon content reproduces the calibration data well and captures the higher carbon content in the MIZ region and low carbon content in the southern part of the study region.The marine and terrestrial organic carbon fractions determined from sediment samples are separated by a new approach using δ 13 C org and the percentage of organic nitrogen contained in the samples (%N org ) in the mixing model.The model predicts the highest marine organic carbon content of the sediments along the southern flank of Spitsbergen Bank.The distribution of the organic carbon fractions is consistent with the presence of a spatially variable mixture of autochthonous and allochthonous sources Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .
Fig. 1.Surface circulation, after Loeng (1991), (red = Atlantic Water, blue = Arctic Water), Polar Front (dashed line), and maximum ice extent (solid line) in the western Barents Sea.The modelled region (dash-dot line), locations of the surface samples (circles), and sediment cores (triangles) in the region (dark triangles = used in the model) are indicated.
region, including the surface currents, ice extent, and locations of the sediment samples.There are two bathymetric highs: Bear Island and Spitsbergen Bank, where water depth is < 30 m at its shallowest point.There are two deeper channels in this region, Bear Island Trough (ca.500 m deep) south of Bear Island and Storfjorden Trough (ca.250 m deep) south of Svalbard.

Table 4 .
Total sediment and organic carbon (total, terrestrial and marine) mass accumulation rates and total Holocene (11 000 yr) sediment and organic carbon burial in the study region compared to other Arctic Ocean shelf seas.