Modelling the CO2 dynamics in the Laptev Sea, Arctic Ocean: Part I

Journal of Marine Systems 102–104 (2012) 29–38

Contents lists available at SciVerse ScienceDirect

Journal of Marine Systems

j ourna l homepage: www.e lsev ie r .com/ locate / jmarsys

Modelling the CO2 dynamics in the Laptev Sea, Arctic Ocean: Part I

Iréne Wåhlström a, Anders Omstedt b, Göran Björk b, Leif G. Anderson a,⁎a Department of Chemistry, University of Gothenburg, Swedenb Department of Earth Sciences, University of Gothenburg, Sweden

⁎ Corresponding author.E-mail address: [email protected] (L.G. Anderson)

0924-7963/$ – see front matter © 2012 Elsevier B.V. Alldoi:10.1016/j.jmarsys.2012.05.001

a b s t r a c t

Article history:Received 16 December 2011Received in revised form 25 April 2012Accepted 1 May 2012Available online 11 May 2012

Keywords:Carbon dioxideArctic OceanLaptev SeaBiogeochemistry

Global temperature observations during the last century show that the largest increase over the last decadeshas been manifested in the Arctic, particularly within the Siberian region. The Arctic Ocean is a harsh regionwith few field studies, resulting in limited temporal and spatial resolution of hydrographical data. One way tocircumvent this deficit is to utilise a model that represents processes which are known to possibly impactclimate. This has been done for the carbon system in the Laptev Sea of the Arctic Ocean by utilising a one-dimensional, time dependent coupled physical–biochemical model. This model was validated by observationaldata of temperature, salinity, phosphate, oxygen and the carbon system.The model simulation reveals that wind pattern is essential for the exchange of dissolved inorganic carbon withthe surrounding seas and the carbon dioxide exchange with the atmosphere. The latter is largely driven by thesurface water partial pressure of carbon dioxide that is impacted by primary production, water temperature,vertical mixing and river runoff. Themodel shows that the timing of these factors is critical for the flux of carbondioxide as is the sea ice coverage. Modelled primary production starts after the disappearance of sea ice and thespring flood has reached the area in June, it peaks in a short time and decreases slowly to negligible levels inmidSeptember. This primary production causes an undersaturation of carbon dioxide with up to 200 μatm duringthe productive season after which the partial pressure of carbon dioxide increases as the carbon dioxide richdeep water mixes up into the surface layer. However, surface water partial pressure of carbon dioxide is under-saturated all through the year, except for some years when there is a short period of outgasing in the beginningof June. This outgasing occurs when the ice breaks up late and river runoff accumulates under the ice.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

Climate change is amplified in the Arctic where the warming is onaverage twice that of the global mean (Serreze and Francis, 2006;Serreze et al., 2009). The most obvious effect of this warming is thedecline in seasonal sea ice coverage that has resulted in open waterduring summer in large areas of both the shelves and the deep centralbasins that earlier where ice-covered (Maslanik et al., 2007; Serrezeet al., 2007). This change in sea ice coverage feeds back into the climatesystem, not only by changes in albedo but also through changes in theexchange of greenhouse gases between the atmosphere and ocean.Several processes affect this air–sea greenhouse gas exchange and thenet effect of these is not straightforward to assess. For instance, moreopen-water will increase the air–sea transfer velocity and thus fasterequilibrate the surface water with the atmosphere. Open water willalso promote net primary production as more light will penetrate intothe upper water layers. This primary production is reinforced by themelting of sea ice that strengthens stratification, which hampers themixing of primary producers down below the photic zone. On the

.

rights reserved.

other hand melting of sea ice also restricts the supply of nutrientsfrom below to the photic zone that determine the maximum possibleprimary production. Furthermore, less sea ice results in more waveactivity, especially in the fall when storms are frequent. This increasedwave activity enhances vertical mixing with its impact on primaryproduction, but also adds to an increase in coastal erosion that bringslarge amount of organic matter into the shelf seas, where it is exposedto microbial degradation.

Most climate related processes in the Arctic Ocean are highlyimpacted by the general circulation and formation of water masses.Warm water enters from the Atlantic Ocean, both through the FramStrait and the Barents Sea. The Fram Strait branch follows the conti-nental slope counter clockwise around the Arctic Ocean. Parts of theBarents Sea branch enter the central basin through the St Anna troughand join the same circulation path as the Fram Strait Branch (Rudelset al., 1994). The remaining waters of the Barents Sea branch flowinto the Kara Sea,mix with river runoff mainly from the Ob and Yeniseyand continue into the Laptev Sea. Some of this water leaves the LaptevSea and flows into the central Arctic Ocean but some continues intothe East Siberian Sea, further mixed with river runoff, where it meetswater from the Pacific Ocean (Jones et al., 1998). Low salinity watersfrom the Kara, Laptev, East Siberian, Chukchi and Beaufort Seas buildup the low salinity surface waters of the central Arctic Ocean. This

30 I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

upper water leaves the Arctic Ocean through both the Canadian ArcticArchipelago and the western Fram Strait (Jones et al., 2008). The shelfseas properties are seasonally impacted by sea ice melting and freezingwhere freezing is larger thanmelting during the year and thus sea ice isexported into the central Arctic Ocean.

The Laptev Sea is a relatively shallow shelf area with mean depthof about 20–30 m in the southern part and deepening northward toabout 100 m before the bottom slopes down to the deep centralbasin (Fig. 1). The Lena River is the dominating source of freshwaterwith an annual mean discharge of 490 km3 that mainly arrives as apulse during summer with 4–5 times higher outflow than the annualmean. The transport of freshwater on the shelf is highly controlled bythe wind pattern during summer with eastward flow along the coastfor years with predominantly northerly and westerly winds, whilesoutherly to south-easterly winds tend to give an offshore transportof surface waters (Dmitrenko et al., 2005). The major part of theice production occurs as dynamical ice production in the flaw leadpolynya that opens for offshore wind during winter and is typicallytwo times larger than the thermodynamic growth (Dmitrenko et al.,2010). The river water mixes with saline water from the continentalslope boundary current, which generates an estuarine circulation withoutflows of low salinity water at the surface and inflows of high salinityslope water at depth.

Net primary production influences the seasonal signal of severalconstituents in the photic zone, also those that are impacted by theair–sea gas exchange. The concentration of dissolved inorganic carbon(DIC, i.e. the sum of H2CO3, HCO3

−, CO32−, CO2) and nutrients, nitrate

(NO3) and phosphate (PO4), decrease during primary production

Fig. 1. The Arctic Ocean with the Laptev

and oxygen (O2) increases. However, the water temperature alsoinfluences the solubility of gases, i.e. O2 and the partial pressure ofcarbon dioxide (pCO2). Therefore, when the seawater gets warmerin summer, the solubility decreases. This countervails the uptake ofCO2 in summer but amplifies the outgasing of O2. In the autumn,remineralisation of the organic matter restores the concentration ofDIC and nutrients togetherwith a decrease in O2 concentration. As car-bon dioxide is consumed during primary production pH will increasein the summer and decrease in the fall whenmineralisation of organicmatter dominates.

The Laptev Sea is one of the Siberian shelf seas that are stronglyaffected by atmosphere–land–ocean interaction and thus impactedby environmental changes in the river drainage basin. In order toassess the effects of environmental changes on the Arctic marinecarbon cycle, a one-dimensional, time dependent coupled physical–biochemical model has been developed for the Laptev Sea. In thiscontribution, the physical and biochemical model is presented andshown to reproduce the annual cycle of biochemical elements. Themodel results are analysed over an 18 year period from 1992 to 2009in order to show the basic biochemical interactions.

2. Methods

2.1. Measured data

Historical data from the Laptev Sea of salinity, temperature, PO4,NO3, O2 and DIC was used to validate the model computations aswell as to set the initial conditions for the depth profile and for the

Sea noted by the red, dotted square.

31I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

conditions of the inflowing waters. The largest data set, covering thelongest time period, is the Russian–American Hydrochemical Atlasof Arctic Ocean (Colony et al., 2002) but this does not include any car-bon parameters. Thus, complementary data from the expeditionsTUNDRA 1994 (Olsson and Anderson, 1997) and ISSS-08 (Andersonet al., 2009) was included together with pCO2 data from Semiletovet al. (2011) and Pipko et al. (2011). The carbon system data fromthe TUNDRA 1994 expedition comprises total alkalinity (TA) andDIC, while the ISSS-08 in addition includes pH. On both cruises certi-fied reference materials (CRM), supplied by A. Dickson, Scripps Insti-tution of Oceanography 119, USA, were used to assure the accuracy ofTA and DIC and it is on the order of 0.1%.

The partial pressure of CO2 was computed from DIC and TA(TUNDRA 1994) and from pH and TA (ISSS-08), respectively, usingthe software CO2SYS (Lewis andWallace, 1998). The carbonate dissoci-ation constants (K1 and K2) used were those of Roy et al. (1993) as theyshow the best internal consistency in the low temperaturewaters of theArctic Ocean when using any two of pH, DIC or TA as input parametersto the software CO2SYS.

2.2. Model description

2.2.1. Model introductionThe one dimensional (1d) model approach offers an efficient way

to describe and analyse the basic physical and biogeochemical pro-cesses in a particular area. A 1d model can be seen as an advancedbudget calculation, if wanted, but with the large advantage that itresolves the vertical stratification and have a rapid and realisticdynamic adaption to surface fluxes. The water exchange with thesurroundings needs however some assumptions in order to be as re-alistic as possible. In the present model application there is a dynamicexchange based on a geostrophic flow assumption and density differ-ences between the model water column and the outside column. Thesurface water exchange is further controlled by wind forcing throughEkman dynamics. By introducing such processes the 1d model will bedynamically active and adapt to changes of the external forcing suchas river water supply or wind stress. By comparing the model with TSstratification observations within the model domain one can test ifthe dynamical assumptions are realistic. Introducing biogeochemicalprocesses is relatively easy and a great advantage is that many config-urations can be tested and evaluated against chemical data (e.g. nutri-ents) due to the fast computations under the 1-d approach.

A general drawback with a 1-d model is that it is not possibleto address any variations in a horizontal direction since only a singlecolumn is described which should represent the average (horizontal)properties of the real system. It should be mentioned that the 1dapproach have been shown to give good results in enclosed basinssuch as the Baltic Sea and the entire Arctic Ocean with well definedstraits and sub-basins (Björk and Söderkvist, 2002; Omstedt, 2011).For a more open area as the Laptev Sea it will be more difficult tomatch with data since there are larger horizontal variations of prop-erties but the model should still give an approximate but realisticdescription of the system on a budget level.

A schematic illustration of the model is presented in Fig. 2. Thecoupled physical–biochemical model adapted to the Laptev Sea is basedon a one-dimensional, time dependent model approach (Omstedt,2011). The general differential equation that forms the base for themodel reads:

∂ϕ∂t þW

∂ϕ∂z ¼ ∂

∂z Γϕ∂ϕ∂z

� �þ Sϕ

where ϕ is the dependent variable, t is the time, z is the vertical coordi-nate, W is the water transport, Γϕ is the exchange coefficient and Sϕ isthe source and sink term for the dependent variable. The first term tothe left is the change in time for the dependent variable, the second

vertical advection and the first to the right is vertical turbulent diffusion.For further details, see Appendix. To simulate the physical part, six equa-tions are applied and for the biochemical part additional six equations areadded. The model-run starts 1989-09-01 and runs for just over 20 yearswith a spin-up time of 2 years. Other versions of themodel have been ap-plied for simulation of different areas: the Mackenzie estuary (Omstedtet al., 1994) and the Baltic Sea (Omstedt et al., 2009).

2.2.2. Physical model componentA technical description of the model is given in the Appendix but

the following gives an overview. The physical model includes equa-tions for momentum in east and north directions, heat, salinity andtwo equations for turbulence (turbulent kinetic energy and its dissi-pation rate). The model covers 50 m depth and the vertical resolutionis 48 seawater layers (grid cells) with the sediment and water surfaceas the boundary layers. The sediment layer acts as a sink for phyto-plankton, but no mineralisation occurs in this layer. The surfacelayer interacts with the atmosphere and the ice, including physicaland chemical parameters of relevance. The area/depth distributionwas reduced to reflect the central part of Laptev Sea.

The system is forced by meteorological data; air temperature,horizontal wind components (u and v), total cloudiness and relativehumidity. Data at 77.5° N, 125° E for every sixth hour were providedby the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from theirwebsite at http://www.esrl.noaa.gov/psd (Kalnay et al., 1996).

The estuarine circulation is modelled by adding fresh river runoffto the surface grid cell, which mixes with the underlying layers,where the salinity source is the high saline deep water that flowsinto the bottom grid cell. The mixed surface water flows out of thesea at the surface and is modelled through geostrophic outflow andEkman transport where the latter is dominating and the outflow isthus mainly wind driven. All of the outflows carry properties suchas heat, salinity, phytoplankton, O2, NO3, PO4, DIC, TA and dissolvedorganic carbon (DOC). The lateral conditions for corresponding deepwater inflows were prescribed based on available deep water obser-vations, where the DIC and TA concentrations were recalculated tocorrespond to the salinity. The exception was for phytoplankton,which content was set to zero in the deep water inflow.

River runoff is included in the model with the data of dischargerepresented by the Lena River. Monthly average discharge over theyears 1976–1994 obtained from R-ArcticNet (Lammers et al., 2001)(www.r-arcticnet.sr.unh.edu/v3.0/index.html) were included, withthe mean annual being 490 km3 y−1. The properties of the runoffwere for temperature and oxygen the same as in the surface water,for NO3 and PO4, 7.8 μM and 0.18 μM (Cauwet and Sidorov, 1996),for DIC and TA, 977 μM and 776 μM (Pipko et al., 2011), and for DOC841 μM (Cooper et al., 2008).

The Lena River discharge exhibits a large variation with lowdischarge between November and May and a large peak in June(Fig. 3). This spring flood in June emerges from the melting of iceand snow in the river and drainage basin. The river ice melts fromthe south and the water level increases as the melting continuesnorthward. When the northernmost river ice melts a strong pulse offreshwater flushes into the sea and creates a low salinity plume. Inthe model, this pulse peaks on June 1 each year. After the maximumin June, the discharge decreases again until it reaches its low levelsin November.

2.2.3. Biochemical model componentTo simulate the carbon system seven additional equations are

applied to compute the concentrations of DIC, NO3, PO4, O2, DOCand the concentration of phytoplankton. The equations to computeprimary production are based on those of Erlandsson (2008) and iscalculated as the primary production fuelled only by new nutrientsas no mineralisation occurs in the surface mixed layer (depth definedby the thermocline). Hence, the computed primary production equals

Fig. 2. A schematic illustration of the model principle. Double arrows indicate water transport that also carries different constituents; temperature, salinity, phytoplankton, O2, NO3,PO4, DIC, DOC and TA. On the left, the models grid cells are illustrated together with the vertical mixing of water that also transport the different constituents between the 48 gridcells. Q is water volume flux, RRO is the river runoff, POM is the particulate organic matter, and [X] is the concentration for different constituents.

32 I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

new primary production. This primary production is simplified in thismodel, as it is only represented by one phytoplankton type that islimited by nutrients (NO3 or PO4) and by light.

The dominating phytoplankton in the Arctic Ocean is the diatoms(Vinogradov et al., 2000) that are consistent with the phytoplanktonparameter in the model. The concentration of phytoplankton is calcu-lated in phosphorous unit and is then converted to nitrate throughthe Redfield ratio, P:N 1:16 (Redfield et al., 1963). In reality primaryproduction is also fuelled by organic nutrients that are released byrespiration (e.g. Martínez-García et al., 2010; Rubin, 2003). This isaccounted for in the model (Omstedt et al., 2009) by utilising a higherC:nutrients ratio (303 for C:P) than the classical Redfield ratio ofphytoplankton (106:1 for C:P). In the model, primary production con-sumes CO2, NO3 and PO4, while O2 is released (Fig. 2).

The produced phytoplankton sinks under the thermocline wheremineralisation starts that add CO2, NO3 and PO4 back to the water.

Fig. 3. Lena River discharge averaged from 1976 to 1994.

Some of these constituents mix back to the surface water and arefree to be utilised again. Consequently, the new primary productionis computed. DOC is mineralised all through the water column at arate equal to a one year half time (Alling et al., 2010; Letscher et al.,2011). The concentration of CO2 and O2 is not only affected by theprimary production and the mineralisation but also by the exchangewith the atmosphere. The atmospheric pCO2 values are downloadeddata from the National Oceanic and atmospheric Administration(NOAA), Point Barrow, Alaska (Thoning et al., 2010). For furtherdetails about the model biochemical part, see Appendix.

3. Results and discussion

3.1. Model overview

When it comes to comparing with observations a one-dimensionalmodel is likely to be more successful in matching observations forquantities that are driven at the sea surface, such as surface tempera-ture and biological production. Quantities driven through the sideborders, such as surface salinity impacted by river inflow at thecoast, are harder to match since observations tend to be much morevariable in the horizontal direction. As the modelling represent hori-zontal means the observed data need to be horizontally averagedover the model domain.

3.1.1. Surface waterThe model output of surface water properties at 4.5 m depth are

compared to observed data collected at 4–5 m depth and averagedin an area limited by 115 to 135° E and 74 to 77° N in the Laptev Sea.Themodel output shows the annual cycle of the physical and chemicalconstituents (Fig. 4). The agreement with the measurements is bestfor the more surface controlled quantities (direct or indirect) suchas temperature, PO4 and O2 (Fig. 4). The salinity observations showmore deviation since they are much more dependent on position

Fig. 4. Observed (red dots) and surface values of temperature, salinity, DIC, PO4, O2 and pCO2. Observations are horizontal averages over the depths interval 4–5 m in the areabetween 115 to 135° E and 74 to 77° N.

33I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

relative to the freshwater source and it is harder to judge howwell themodel performs with respect to surface salinity. The model gives,however, a realistic annual cycle with salinities just above 30 duringwinter and summer minimum salinities between 10 and 20. This an-nual cycle matches relatively well with the observations, keeping inmind that the measured values represent relatively scarcely sampleddata from specific locations in the Laptev Seawhile themodel representsan average (horizontal) water column represented by one depth profile.

DIC has a strong correlation to salinity and thus has the sameuncertainties as the salinity. Furthermore, it is impacted by primaryproduction, consuming CO2, and by air–sea exchange. Consequently,pCO2 is also impacted by these processes s well as by runoff. Themodelled pCO2 is substantially higher than the few available observa-tions, possibly a result of missing processes or incorrect boundaryconditions. However, it was necessary to increase the level of pCO2

and DOC in the runoff to unrealistic values in order to achieve theobserved pCO2.

3.1.2. Depth profilesCritical physical processes in the model are the net estuarine

circulation and vertical mixing, which also controls much of the bio-chemical processes. An example of comparison between model andobserved profiles is shown for 1993 which is the year with mostobservations (Fig. 5). The observed surface salinities in August havea considerable spread, but the mean correspond well with the mod-elled value of about 15. The model profiles also have a realisticshape and approaches the deep water salinity at around 25 m, similarto the observations, showing that the mixing and vertical advectionin the model are realistic. Deviations are larger in September with atendency that more freshwater is stored in the area than the modelshow. This can be an indication that the model has somewhat toolarge export of the low salinity surface water but it can also be dueto the fixed annual cycle of river discharge leaving the possibilitythat the discharge in September 1993 was larger than the long termmean. The data is however very close to the single observed stationat September 14 at the end of the period. Comparing with surfacedata for all years (Fig. 4) it is however hard to see any systematicbias towards to high surface salinity at the end of summer. For ex-ample in 1999, the model shows a tendency to keep a low surfacesalinity somewhat longer into the autumn than what is seen in theobservations.

The model temperature profiles in August are also realistic withsurface temperature (around 4 °C) and a decrease with depth, whichis similar to the observed profiles. The model temperature-profiles inSeptember are close to freezing over the entire depth range whilesome of the observed are still a few degrees above freezing. This ten-dency of autumn cooling occurring too rapidly indicates that therecan be some imperfections in the atmospheric forcing data for thisarea since the near surface temperature is heavily controlled by theatmospheric forcing.

3.1.3. Time seriesThe model gives a distinct seasonal cycle with pronounced winter

and summer periods for the upper 20–30 m (Fig. 6). The stratificationstarts in the beginning of June and depends mostly on the increase offreshwater from the river discharge. This freshwater, together withthe sea ice melt, decreases salinity and a halocline is established thathampers vertical mixing of the water column. Surface water has thelowest salinity in June to August when most of the sea is ice-free.

In September–October salinity increases as a combined effect ofsea ice formation that adds brine to the water column as well as bya decrease in river discharge. From November to April, the watercolumn is well mixed and the salinity is stable around 32. The ther-mocline forms in May when the surface water absorbs solar radiationafter the ice has disappeared. The surface temperature increases to amaximum of 8 °C in July–August and decreases again in Septemberwhen the atmospheric cooling begins.

3.2. Biochemical modelling

3.2.1. Primary productionIn the model, primary production starts in late May or in the be-

ginning of June (Fig. 7a) and peaks in June when the sea ice breaksup that allows enough light to penetrate into the surface water andthe river runoff pulse has reached the sea. The melting ice stratifiesthe surface water and stratification increases further by the pulse offreshwater from river runoff, which hampers mixing of primary pro-ducers down below the photic zone. After a short time, the modelledprimary production reaches its peak and decreases again when thenutrient concentration starts to limit productivity (Fig. 7b). Afterthe peak, primary production decreases close to linearly with timeuntil the beginning of September when light starts to be a limiting

Fig. 5. a–b) Observed (red) and modelled (black) temperature and salinity depth-profiles for August (solid) and September (solid with circles) in 1993. c) Positions for the obser-vations in August (filled symbols) and September (open symbols). August data are from the period 9-17/8 and September data from 1-14/9. Observed depth profiles from the samedate are averaged together for clarity of the plots. Model profiles are shown for the same dates as the observations. d) Temperature–salinity plot for all the observations and themodel data from the same dates.

34 I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

factor (Fig. 7a). This decrease is caused both by low sun elevation andby the start of sea ice formation. The model predicts one bloom withthe limiting nutrient being PO4, consistent with observations in 2008(Anderson et al., 2009).

Fig. 6. Modelled time-series of DIC, NO3, PO4, O2, pCO2, pH

The river brings large amounts of DIC and nutrients to the sea(Gordeev et al., 1999)which is also included in themodel. The nutrientscoming with the river water stimulate primary production (Sorokinand Sorokin, 1996) and this agrees well with the model output since

, temperature and salinity for the years 2000–2009.

Fig. 7. a) Modelled daily total primary production over the eighteen-year period 1992–2009 (black lines), with the average value in thick red line. b) Modelled nutrients, NO3 (blue)and PO4 (black), for the same period with the average in red lines.

35I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

primary production starts after the freshwater pulse flows into the seaand the ice has melted. Also the stratification that river freshwaterdevelops favours phytoplankton grows (Sakshaug, 2004) and hampersvertical mixing.

The modelled average–maximum primary production was1.5 gC m−2 d−1 (Fig. 7a), while the maximum primary productionduring the modelled eighteen years was 2.8 gC m−2 d−1 and occurredin the years 2002 and 2005 (Fig. 7a). These two years with maximumprimary production were years with longer period of less cloudinessin the summer. More solar insulation then penetrates into the surfacewater that gives energy to the phytoplankton and increases the watertemperature, both affecting the primary production. However, a highpeakdoes not imply a larger annual primary production as the nutrientsare the limiting factor and they are then consumed faster resulting ina sharper decrease in primary production after the peak. From observa-tions in the open south-east Laptev Sea Sorokin and Sorokin (1996)reported primary productivity rates of 0.03–0.11 gC m−2 d−1 in Sep-tember 1991, which is in the range of our modelled rates, less than0.2 gC m−2 d−1, for the same time period (Fig. 7a). The modelledprimary production earlier in the season is larger than in September,stressing the need to cover a whole season in order to get an accurateestimate of primary production.

The average, integrated modelled primary production over theeighteen years was 70±7 gC m−2 y−1. Sakshaug (2004) reportedtotal primary production in the European shelves to range between15–20 gC m−2 y−1 and up to 70 gC m−2 y−1 in the north, near themultiyear ice. Themodelled total carbon consumption within the LaptevSea was computed to 26±3 Tg Cy−1 using the area 370,000 km2

shallower than 50 m, estimated from Jakobsson (2002). Vinogradovet al. (2000) utilised primary production maps based on both satelliteand field data to deduce a carbon consumption of 10–15 Tg Cy−1.Considering the observed data scarcity, the uncertainties in the model,and that the model gives annual new primary production, one cannotdistinguish the model result from the observed estimates.

3.2.2. The CO2 sea–air fluxThe modelled pCO2 shows a seasonal variability with low values

during the productive summer and high, but still under-saturated,values in the winter (Fig. 8a). The CO2 supersaturated river runoff isadded to the upper layer of the model, starting on 1 June each year.Some years a peak in pCO2 develops that makes the surface water su-persaturated relative to the atmosphere and this results in an out-gasing of CO2 from the sea to the atmosphere. In reality, as in the

model, the river is supersaturated with respect to CO2 all seasons(Semiletov, 1999) and the large pulse of freshwater in June producesthis maximum of pCO2. The peak of pCO2 declines rapidly after a fewdays when primary production starts that consumes CO2. In the endof the productive season pCO2 increases to values close to or slightlyabove those of the atmosphere when the CO2 from mineralisationof organic matter bellow the thermocline is mixed up into the surfacelayer. At the same time, the decrease in temperature increases thesolubility of pCO2 and this restrains the increase of pCO2 from min-eralisation of organic matter. The winter pCO2 values also vary be-tween years, but much less than in the summer, as the sea ice coverhampers sea–air exchange as well as vertical mixing.

A negative sea–air flux of CO2 dominates in themodel, i.e. an uptakeby the surface water (Fig. 8b) in accordance with the general situationof other studies (Nitishinsky et al., 2007; Semiletov, 1999). Exceptionsfrom the negative flux are in June certain years (e.g. 1996, 2004 and2008, Fig. 8a) when the pulse of river runoff of high pCO2 enters thesea and there are short periods of outgasing (Fig. 8b). This outgasingis associated with a late ice break-up, which makes the river dischargeof supersaturated pCO2 runoff accumulate under the ice and an out-gasing to the atmosphere occurs when the ice disappears. Such an es-cape off CO2 is dependent on the wind condition, where strong windsdrive the outgasing before primary production draw down the surfacewater pCO2. The maximum outgasing was ~35 mmol m−2 d−1 in1996 (Fig. 8b) that was approximately 5 times higher than the otheryears with positive flux. 1996 had the latest ice break-up during themodel period and this gave a large accumulation of supersaturatedfreshwater under the ice before sea ice break-up.

In the other years of the eighteen-year model-run, the sea wasice-free when the river runoff flowed into the sea. These years canbe divided into two categories: one with less outgasing than above(e.g. 2002, 2006, and 2009, Fig. 8a) and one with no outgasing but aflux of CO2 into the seawater the whole year (e.g. 2000, 2003, 2005and 2007, Fig. 8a). In the latter, there is less wind than in the formerand this reduces the outgasing of CO2 to the atmosphere.

The variation in the daily summer seawater uptake of CO2 is sub-stantial between the years,with an average of about 20 mmol m−2 d−1

and a maximum of 84 mmol m−2 d−1 (Fig. 8b). Nitishinsky et al.(2007) calculated the CO2-flux based on a two-layer box model forthe summer of 1994, resulting in an uptake of CO2 by the sea of2.1 mmol m−2 d−1. This is an order of magnitude lower than themodel's average flux (20 mmol m−2 d−1) but is equivalent to themodel's fluxes in the end of the productive seasons.

Fig. 8. a) Modelled pCO2 in the surface layer (green line) and observed pCO2 in the atmosphere (blue line) at Point Barrow, Alaska, versus time. b) Modelled sea–air flux of CO2 over theeighteen-year period 1992–2009 (black lines), with the average value shown by the red, thick line. Positive values are sea-to-air flux and negative air-to-sea flux.

36 I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

3.2.3. DIC-budgetThe largest terms of the modelled DIC budget in the sea are due

to exchange of waters in- and out of the area (Fig. 9a). The exchangeof seawater is far from constant over the years and water exchangeis typically smaller in winter, a result of the sea ice hampering the en-ergy input by the wind. The transport of surface water in the model ismainly wind-driven; an eastward wind drives the Ekman transport topush the surface water northward, out of the sea. That the surfacewater circulation in the sea is dominantly wind driven is consistentwith observations (Guay et al., 2001). The average modelled in- andoutflow to the Laptev Sea was calculated to 0.08 Sv with a maximumat 1.6 Sv. Anderson et al. (1998) estimated the net outflow of surface

Fig. 9. Modelled sources and sinks of DIC; a) by inflow of deep water (blue) and outflows(black), and c) by river runoff (blue) and sea ice (green). Positive values are into the Lapte

water from the Laptev Sea to 0.16 Sv based on a budget calculationutilising salinity conservation. This is twice as much as the averagein the model but considering the uncertainties in these estimatesthe agreement is reasonable. Walsh et al. (2007) computed the pro-duction rate of deep, cold, salty shelf waters that flow out of theshelf and into the Arctic Ocean to be ~0.03 Sv, which is half of themodelled maximum outflow of deepwater.

The Lena River runoff has a strong seasonal signal and contributestogether with CO2 uptake from the atmosphere to the second largestmodelled source of DIC to the sea (Fig. 9b). Even if they are the secondlargest sources of DIC, it is one order of magnitude lower than theDIC added by the inflow of deepwater each year (Table 1). Burial of

of surface (green) and deep water (magenta), b) by sea–air (red) and sediment burialv Sea and negative out.

Table 1Modelled sources and sinks for DIC and associated volume fluxes over an eighteen-yearperiod, 1992–2009. Positive values are sources and negative sinks.

18 years DIC, 1012 mol y−1 Volume flux, km3 y−1

River runoff 0.48±0 490Net deep water 4.84±0.9 2,259Sea–air exchange 0.63±0.1 −2,488Surface water −5.24±0.9Ice −0.06±0 −261Sedimentation burial −0.60±0.1Phytoplankton out with water −0.03±0Balance −0.02±0.1

37I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

organic matter in the sediment is the next largest sink of DIC and is ofthe same order as the atmospheric and runoff sources (Table 1). Thisoccurs in the summer when there is primary production and the phy-toplankton sinks out of the bottom grid cell, which is identical withthem being buried in the sediment.

In wintertime, there is also a sink for DIC when the modelledwind-driven sea ice is transported out of the sea. When the sea iceis formed, DIC is trapped into the brine-pockets and is transportedout of the sea (Fig. 9b). This transport occurred all through the timewhen sea ice was present, i.e. from September to late May or the be-ginning of June. The modelled average sea ice-flux over the eighteen-year period was calculated to 261 km3 y−1. This is at the same orderof magnitude as earlier estimates of first-year ice flux, 670 km3 y−1

(Eicken, 2004). Phytoplankton also advects out of the area with thewater outflow making this another sink of DIC. This sink is lowbut in the same order as the export by sea ice. The modelled annualsinks and sources are summarised in Table 1 together with the rele-vant volume fluxes. Adding all the fluxes balances the modelled bud-get over the years and thereby DIC does not accumulate or leak withtime.

4. Conclusions

A 1-D coupled physical–biochemical model for the Laptev Sea,Siberian Arctic, was applied to increase the knowledge of the annualLaptev Sea carbon cycle. The model output shows that wind is criticalfor the exchange of DIC with the surrounding seas and CO2 exchangewith the atmosphere. Primary production starts after the disappear-ance of sea-ice and the spring flood have flushed into the sea. Thisbrings light into the seawater and stratifies the water, which hampersthe water mixing and therefore favours primary production. The pri-mary production peaks in a short time (less than a week) after whichit decrease close to linear until mid September when it is negligible.The limiting nutrient for primary production is phosphate.

In the surface water, an undersaturation of CO2 with up to 200 μatmis the effect of the productive season. After this season, the surfacewaters pCO2 increase to about atmospheric level. This rise in pCO2 oc-curs as the deep water, rich in decay products (one being CO2), mixesup into the surface layer. CO2 is undersaturated in the surface waterall year around; except for some years when there is a short period ofoutgasing in the beginning of June. This happens when the ice breaksup late and the river runoff has accumulated under the ice. This modellargely represents the biochemical cycle of the Laptev Sea and is thebasis for testing the sensitivity of the carbon dynamic to changes inthe forcing: e.g. increasing atmospheric temperature with its conse-quences for sea ice coverage, river discharge and permafrost thawing.

Acknowledgments

Financial support was received from the Swedish Research Council(contract no. 621-2010-4084), the European Union project EPOCA(contract 211384) and CarboChange (project reference 264879) and

Tellus, the Centre of Earth Systems Science at the University ofGothenburg.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.jmarsys.2012.05.001.

References

Alling, V., Sanchez-Garcia, L., Porcelli, D., Pugach, S., Vonk, J., van Dongen, B., Mörth, M.,Anderson, L.G., Sokolov, A., Humborg, C., Semiletov, I., Gustafsson, Ö., 2010. Non-conservative behavior of dissolved organic carbon across the Laptev and East SiberianSeas. Global Biogeochem. Cycles 24, GB4033, http://dx.doi.org/10.1029/2010GB003834.

Anderson, L.G., Olsson, K., Chierici, M., 1998. A carbon budget for the Arctic Ocean.Global Biogeochem. Cycles 12, 455–465.

Anderson, L.G., Jutterstrom, S., Hjalmarsson, S., Wahlstrom, I., Semiletov, I.P., 2009. Out-gassing of CO2 from Siberian Shelf seas by terrestrial organic matter decomposition.Geophys. Res. Lett. 36, 6.

Björk, G., Söderkvist, J., 2002. Dependence of the Arctic Ocean ice thickness distributionon the poleward energy flux in the atmosphere. J. Geophys. Res. 107 (C10), http://dx.doi.org/10.1029/2000JC000723.

Cauwet, G., Sidorov, I., 1996. The biogeochemistry of Lena River: Organic carbon andnutrient distribution. Mar. Chem. 53, 211–227.

Colony, R., Nikiforov, Y.G., Pivovarov, S.V., Pokrovsky, J.M., Priamikov, S.M., Timokhov,L.A., 2002. Data base and atlas of hydrochemistry of the Arctic Ocean. OceanScience Meeting, Honolulu, Hawaii, USA.

Cooper, L.W., McClelland, J.W., Holmes, R.M., Raymond, P.A., Gibson, J.J., Guay, C.K.,Peterson, B.J., 2008. Flow-weighted values of runoff tracers (delta 18O, DOC, Ba,alkalinity) from the six largest Arctic rivers. Geophys. Res. Lett. 35, 5.

Dmitrenko, I., Kirillov, S., Eicken, H., Markova, N., 2005. Wind-driven summer surfacehydrography of the eastern Siberian shelf. Geophys. Res. Lett. 32, 5.

Dmitrenko, I.A., Wegner, C., Kassens, H., Kirillov, S.A., Krumpen, T., Heinemann, G.,Helbig, A., Schroder, D., Holemann, J.A., Klagge, T., Tyshko, K.P., Busche, T., 2010.Observations of supercooling and frazil ice formation in the Laptev Sea coastalpolynya. J. Geophys. Res. Oceans 115, 9.

Eicken, H., 2004. The role of Arctic sea ice in transporting and cycling terrigenousorganic matter. In: Macdonald, R.S.a.R.W. (Ed.), The Organic Carbon Cycle in theArctic Ocean. Springer, Heidelberg, Germany, pp. 45–53.

Erlandsson, C.P., 2008. Vertical transport of particulate organic matter regulated by fjordtopography. J. Geophys. Res. Biogeosci. 113, 13.

Gordeev, V.V., Dgamalov, R.G., Zekcep, I.S., Julidov, W.W., Brizgalo, W.A., 1999. Estima-tions of inflow of nutrient with river runoff and ground water in Marginal Seas ofRussian Arctic. Water Source 26 (2), 206–211 (in Russian).

Guay, C.K.H., Falkner, K.K.,Muench, R.D.,Mensch,M., Frank,M., Bayer, R., 2001.Wind-driventransport pathways for Eurasian Arctic river discharge. J. Geophys. Res. Oceans 106,11469–11480.

Jakobsson, M., 2002. Hypsometry and volume of the Arctic Ocean and its constituentseas. Geochem. Geophys. Geosyst. 3, 18.

Jones, E.P., Anderson, L.G., Swift, J.H., 1998. Distribution of Atlantic and Pacificwaters in theupper Arctic Ocean: implications for circulation. Geophys. Res. Lett. 25, 765–768.

Jones, E.P., Anderson, L.G., Jutterstrom, S., Mintrop, L., Swift, J.H., 2008. Pacific freshwa-ter, river water and sea ice meltwater across Arctic Ocean basins: results from the2005 Beringia Expedition. J. Geophys. Res. Oceans 113, 10.

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha,S., White, G., Woollen, J., Zhu, Y., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak,J., Mo, K.C., Ropelewski, C., Wang, J., Leetmaa, A., Reynolds, R., Jenne, R., Joseph, D.,1996. The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc. 77,437–471.

Lammers, R.B., Shiklomanov, A.I., Vorosmarty, C.J., Fekete, B.M., Peterson, B.J., 2001.Assessment of contemporary Arctic river runoff based on observational dischargerecords. J. Geophys. Res. Atmos. 106, 3321–3334.

Letscher, R.T., Hansell, D.A., Kadko, D., 2011. Rapid removal of terrigenous dissolved or-ganic carbon over the Eurasian shelves of the Arctic Ocean. Mar. Chem. 123, 78–87.

Lewis, E., Wallace, D.W.R., 1998. Program developed for CO2 system calculations, ORNL/CDIAC-105. Carbon Dioxide Information Analysis Center, Oak Ridge NationalLaboratory, U.S. Department of Energy, Oak Ridge, Tennessee.

Martínez-García, S., Fernández, E., Álvarez-Salgado, X.-A., González, J., Lønborg, C.,Marañón, E., Morán, X.-A.G., Teira, E., 2010. Differential responses of phytoplanktonand heterotrophic bacteria to organic and inorganic nutrient additions in coastalwaters off the NW Iberian Peninsula. Mar. Ecol. Prog. Ser. 416, 17–33, http://dx.doi.org/10.3354/meps08776.

Maslanik, J., Drobot, S., Fowler, C., Emery,W., Barry, R., 2007. On theArctic climate paradoxand the continuing role of atmospheric circulation in affecting sea ice conditions.Geophys. Res. Lett. 34, 4.

Nitishinsky, M., Anderson, L.G., Holemann, J.A., 2007. Inorganic carbon and nutrientfluxes on the Arctic Shelf. Cont. Shelf Res. 27, 1584–1599.

Olsson, K., Anderson, L.G., 1997. Input and biogeochemical transformation of dissolvedcarbon in the Siberian shelf seas. Cont. Shelf Res. 17, 819–833.

Omstedt, A., 2011. Guide to process based modelling of lakes and coastal seas. Springer-Praxis Books in Geophysical Sciences. Springer-Verlag, Berlin, Heidelberg.

Omstedt, A., Carmack, E.C., Macdonald, R.W., 1994. Modeling the seasonal cycle ofsalinity in the Mackenzie shelf estuary. J. Geophys. Res. Oceans 99, 10011–10021.

38 I. Wåhlström et al. / Journal of Marine Systems 102–104 (2012) 29–38

Omstedt, A., Gustafsson, E., Wesslander, K., 2009. Modelling the uptake and release ofcarbon dioxide in the Baltic Sea surface water. Cont. Shelf Res. 29, 870–885.

Pipko, I., Pugach, S.P., Semiletov, I.P., Salyuk, A.N., 2011. Carbonate characteristics ofwaters of the Arctic Ocean continental slope. Dokl. Earth Sci. 438, 858–863.

Redfield, A.C., Ketchum, B.H., Richards, F.A., 1963. In: Hill, M.N. (Ed.), The influence oforganisms on the composition of sea-water: The Sea, 2. Interscience, New York,pp. 26–77.

Roy, R.N., Roy, L.N., Vogel, K.M., Portermoore, C., Pearson, T., Good, C.E., Millero, F.J.,Campbell, D.M., 1993. The dissociation-constants of carbon-acid in seawater at salin-ities 5–45 and temperatures 0-degrees-C to 45 degrees C. Mar. Chem. 44, 249–267.

Rubin, S.I., 2003. Carbon and nutrient cycling in the upper water column across thePolar Frontal Zone and Antarctic Circumpolar Current along 170°W. Global Bio-geochem. Cycles 17 (3), 1087, http://dx.doi.org/10.1029/2002GB001900.

Rudels, B., Jones, E.P., Anderson, L.G., Kattner, G., 1994. On the intermediate depth waters ofthe Arctic Ocean. In: Johannessen, O.M., Muench, R.D., Overland, J.E. (Eds.), The PolarOceans and Their Role in Shaping the Global Environment. American GeophysicalUnion, Washington, D.C., pp. 33–46.

Sakshaug, E., 2004. Primary and secondary production in the Arctic Seas. In: Stein, R.,Macdonald, R.W. (Eds.), The Organic Carbon Cycle in the Arctic Ocean. Springer,Heidelberg, Germany, pp. 59–81.

Semiletov, I.P., 1999. Aquatic sources and sinks of CO(2) and CH(4) in the polar regions.J. Atmos. Sci. 56, 286–306.

Semiletov, I.P., Pipko, I., Shakhova, N.E., Dudarev, O.V., Pugach, S.P., Charkin, A.N., McRoy,C.P., Kosmach, D., Gustafsson, O., 2011. Carbon transport by the Lena River from itsheadwaters to theArctic Ocean,with emphasis on fluvial input of terrestrial particulateorganic carbon vs. carbon transport by coastal erosion. Biogeosciences 8, 2407–2426.

Serreze,M.C., Francis, J.A., 2006. The Arctic amplificationdebate. Clim. Chang. 76, 241–264.Serreze, M.C., Holland, M.M., Stroeve, J., 2007. Perspectives on the Arctic's shrinking

sea-ice cover. Science 315, 1533–1536.Serreze, M.C., Barrett, A.P., Stroeve, J.C., Kindig, D.N., Holland, M.M., 2009. The emer-

gence of surface-based Arctic amplification. Cryosphere 3, 11–19.Sorokin, Y.I., Sorokin, P.Y., 1996. Plankton and primary production in the Lena River

estuary and in the south-eastern Laptev Sea. Estuarine Coastal Shelf Sci. 43, 399–418.Thoning, K.W., Kitzis, D.R., Crotwell, A., 2010. Atmospheric Carbon Dioxide Dry Air

Mole Fractions from Quasi-continuous Measurements at Barrow, Alaska; MaunaLoa, Hawaii; American Samoa; and South Pole, 1973–2009, Version: 2010-07-14,Path: ftp://ftp.cmdl.noaa.gov/ccg/co2/in-situ/ed.

Vinogradov, M.E., Vedernikov, V.I., Romankevich, E.A., Vetrov, A.A., 2000. Componentsof the carbon cycle in the Russian Arctic Seas: primary production and flux ofC(org) from the photic layer. Oceanology 40, 204–215.

Walsh, D., Polyakov, I., Timokhov, L., Carmack, E., 2007. Thermohaline structure andvariability in the eastern Nansen Basin as seen from historical data. J. Mar. Res.65, 685–714.