Coastal ocean CO 2 carbonic acid carbonate sediment system of the Anthropocene

GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 20,, doi:10.1029/2005gb002506, 2006 Coastal ocean CO 2 carbonic acid carbonate sediment system of the Anthropocene Andreas J. Andersson, 1 Fred T. Mackenzie, 1 and Abraham Lerman 2 Received 7 March 2005; revised 4 December 2005; accepted 5 January 2006; published 2 March 2006. [1] There is little doubt that human activities such as burning of fossil fuels and land use practices have changed and will continue to change the cycling of carbon in the global coastal ocean. In the present study, two biogeochemical box models were used to investigate the consequences of increasing atmospheric CO 2 and subsequent ocean acidification and increasing riverine transport of organic matter and nutrients arising from human activities on land on the global coastal ocean between the years 1700 and 2300. Numerical simulations show that the net flux of CO 2 between coastal ocean surface water and the atmosphere is likely to change during this time from net evasion to net invasion owing to increasing atmospheric CO 2, increasing net ecosystem production arising from increasing nutrient loading to this region, and decreasing net ecosystem calcification due to lower carbonate ion concentration and subsequent lower surface water saturation state with respect to carbonate minerals. Model calculations show that surface water saturation state with respect to calcite will decrease 73% by the year 2300 under a business-as-usual scenario, which in concert with increasing temperature will cause overall biogenic calcification rate to decrease by 90%. Dissolution of carbonate minerals increased by 267% throughout the model simulation. This increase was in part due to increased invasion of atmospheric CO 2, but mainly due to greater deposition and remineralization of land-derived and in situ produced organic matter in the sediments, producing CO 2 that caused pore water ph and carbonate saturation state to decrease. This decrease, in turn, drove selective dissolution of metastable carbonate minerals. As a consequence, the relative carbonate composition of the sediments changed in favor of carbonate phases with lower solubility than that of an average 15 mol% magnesian calcite phase. Model projected changes in surface water carbonate saturation state agree well with observations from the Hawaiian Ocean Time series and the calculated air-sea CO 2 exchanged agrees well with a recent independent estimate of this flux derived from measurements from diverse coastal ecosystems scaled up to the global coastal ocean area. Citation: Andersson, A. J., F. T. Mackenzie, and A. Lerman (2006), Coastal ocean CO 2 carbonic acid carbonate sediment system of the Anthropocene, Global Biogeochem. Cycles, 20,, doi:10.1029/2005gb002506. 1. Introduction [2] In recent years there has been considerable discussion concerning the effects of rising atmospheric CO 2 and temperature on coral reef and other carbonate ecosystems, as well as the effects of lower saturation state, due to absorption of atmospheric CO 2 (ocean acidification ), of surface ocean waters with respect to carbonate minerals on carbonate secreting organisms [e.g., Gattuso et al., 1999; Kleypas et al., 1999, 2001; Langdon et al., 2000, 2003; 1 Department of Oceanography, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, Honolulu, Hawaii, USA. 2 Department of Geological Sciences, Northwestern University, Evanston, Illinois, USA. Copyright 2006 by the American Geophysical Union. 0886-6236/06/2005GB002506 Andersson et al., 2003; Buddemeier et al., 2004; McNeil et al., 2004; Feely et al., 2004; Orr et al., 2005]. This decrease in carbonate saturation state has led several investigators, among those cited above, to infer that carbonate secreting organisms may have difficulty in maintaining calcification at present rates and that coral reef ecosystems may be subject to enhanced biological and mechanical erosion in the future. The increase in atmospheric CO 2 is not the only factor behind the lower saturation state of ocean water with respect to carbonate minerals; the saturation state is also affected by the production and remineralization of organic carbon in coastal ocean water and sediments, as is documented in the subsequent sections of this paper. The latter processes depend in turn on the inputs of the nutrient elements nitrogen and phosphorus and organic carbon to the surface ocean water of the coastal zone. In addition, temperature change affects both the rates of biological 1of13

calcification as well as primary production, and the chemical state of the carbonate system in seawater. [3] The fairly complex interactions between the environmental factors mentioned above in the shallow-water coastal ocean were studied by means of two established models, TOTEM (Terrestrial Ocean atmosphere Ecosystem Model) [Ver, 1999a] and SOCM (Shallow-water Ocean Carbonate Model) [Andersson, 2003], as described in detail by Andersson and Mackenzie [2004] and Andersson et al. [2005]. The SOCM using inputs from TOTEM calculations analyzes the relationships between saturation state and calcification rate, storage of organic carbon and carbonate minerals in sediments, as well as the overall behavior of the CO 2 carbonic acid carbonate sediment system in the shallow-water ocean throughout the past 300 years of the Industrial Age, also known as the Anthropocene, and into the continuing industrial future to the year 2300. Here we emphasize the long-term, centuryscale changes in the CO 2 -carbonic acid-carbonate sediment system of the coastal ocean during 600 years of the past and future history of Earth s surface environment under the influence of rising atmospheric CO 2 concentrations and temperature, and increasing riverine inputs of nutrients and organic carbon to the coastal ocean. 2. Structure of a Global Coastal Ocean Model [4] The conceptual model of the shallow-water coastal carbonate system (the model SOCM) is shown in Figure 1 as consisting of two major domains, a surface water domain and a pore water-sediment domain. The global shallowwater ocean environment, including coastal zones, reefs, banks and continental shelves extending over an area of 28.3 10 6 km 2 [Milliman, 1974], exchanges water with the open ocean at residence times of several years. Substantial uncertainties are associated with any generalizations of the global coastal region, including the water residence time, which is highly variable and heterogeneous in both time and space. Accordingly, SOCM permits adjustment of residence times from the shorter global residence time of 2 to 3 years, based on estimates of continental margin and coastal upwelling rates [Chavez and Toggweiler, 1995], to those on the order of a decade. [5] The surface water domain contains the reservoirs of dissolved inorganic carbon (DIC) and organic matter consisting of dissolved organic carbon (DOC), organic detritus, and living biota. The sediment and pore water domain contains the reservoirs of organic matter, produced in part in situ in the coastal waters and in part transported from land; refractory particulate inorganic carbon originating from continental erosion; in situ produced calcite, aragonite, and magnesian calcite represented by an average composition of 15 mol% MgCO 3 ; and a pore water reservoir of average composition based on data from the Bahamas, Bermuda, and elsewhere (Morse and Mackenzie [1990], with references to other data sources). [6] The following processes control the inorganic and organic carbon cycles in the shallow coastal ocean model [e.g., Mackenzie et al., 1998]: (1) inputs of dissolved and particulate inorganic and organic carbon from land; (2) inputs of dissolved inorganic and organic carbon by upwelling from the deep ocean; (3) export of dissolved and particulate inorganic and organic carbon to the open ocean; (4) pelagic and benthic primary production; (5) remineralization of organic matter that is in part produced in situ and in part imported from land; (6) biological calcification and inorganic deposition of carbonate minerals calcite, aragonite, and magnesian calcites that depend on the DIC concentration, degree of saturation of ocean water with respect to each mineral, and temperature; (7) carbonate mineral dissolution due to an increase in dissolved CO 2, caused by a rising atmospheric concentration and remineralization of organic carbon; the rates of dissolution depend on temperature and degree of saturation with respect to the mineral; and (8) CO 2 exchange between the atmosphere and surface ocean water that is controlled primarily by the CO 2 partitioning in response to the changing CO 2 concentration in the atmosphere and surface ocean water. [7] The standard model of SOCM calculates the airsea CO 2 exchange and changes in the chemical and mineralogical state of the water-sediment system that are driven by biogenic calcification, inorganic precipitation and dissolution of carbonate phases, and the production, import, export and remineralization of organic matter, as given in points 1 8 above. Inputs of carbon from land and primary production that is driven by the nutrient N and P inputs from land, and upwelling were taken from the results of TOTEM. The mathematical relationships behind these processes in SOCM and their links to the global model TOTEM are described by Andersson et al. [2005], with references to the earlier work. Because the CO 2 flux across the sea-air interface determines to some extent the biogeochemical behavior of the coastal carbonate system and it is an environmentally important transfer process, we summarize below the mathematical relationship for this important flux in SOCM. [8] For a coastal ocean water reservoir, as shown in Figure 1, the balance of dissolved inorganic carbon (DIC) is the difference between its inputs and outputs as fluxes F 1 yr, where fluxes are in units of mol C per year, DDIC ¼ F DIC inflow F DIC outflow F CaCO3 ppt þ F CaCO3 diss F GPP þ F R þ F CO2 ; where F DIC inflow is input from rivers and upwelling, F DIC outflow is output to the open ocean, F CaCO3 ppt and F CaCO3 diss are carbonate precipitation and dissolution, respectively, F GPP and F R are gross primary production and total community respiration, respectively, and F CO2 is air-sea CO 2 exchange. The organic carbon balance is, similarly to that of DIC, DC org ¼ F Corg inflow F Corg outflow þ F GPP F R F Corg S ; ð1þ ð2þ 2of13

where F Corg S is the permanent burial and net organic carbon storage rate in sediments. The sum of (1) and (2), with the notation of DDIC = [DIC eq ] [DIC 0 ], is DIC eq ½ DIC0 Š ¼ F DIC inflow F DIC outflow F CaCO3 ppt þ F CaCO3 diss þ F Corg inflow F Corg outflow F Corg S DC org þ F CO2 : ð3þ [DIC eq ] is the DIC concentration at equilibrium with atmospheric CO 2 after all the other input and removal fluxes have been accounted for. [DIC 0 ] is the initial DIC concentration (2000 mmol kg 1 ). The individual terms in equations (1) (3) and their initial values at the end of preindustrial time, taken as the year 1700, are given by Andersson et al. [2005]. [9] The CO 2 air-sea flux (F CO2 ) in equation (3) is negative when CO 2 is emitted from coastal water, making the coastal water a source of CO 2 to the atmosphere, and it is positive when the coastal zone is a CO 2 sink. With this notation, equation (3) can be written as an algebraic sum of the difference between the DIC outflow and inflow, net calcification in the system that is the difference between the CaCO 3 precipitation and dissolution rates (NEC), and a net change in the organic carbon content of the system: F CO2 ¼ ðf DIC outflow F DIC inflow Þ þ F CaCO3 ppt F CaCO3 diss þ F Corg outflow þ F Corg S F Corg inflow þ dc org =dt þ dc DIC =dt: [10] In the precipitation reaction of CaCO 3 in seawater or fresh water, written as Ca 2þ þ 2HCO 3 ¼ CaCO 3 þ CO 2 þ H 2 O; two moles of HCO 3 are removed from the aqueous solution for every one mole of precipitated CaCO 3 and one mole of CO 2 is released back to the solution. Primary production can consume CO 2, such that the net effect of CaCO 3 and organic matter formation is Ca 2þ þ 2HCO 3 ¼ CaCO 3 þ CH 2 O þ O 2 : ð4þ ð5þ ð6þ The stoichiometry of the preceding two reactions is only correct for a relatively low ph fresh water solution, but not for seawater. Depending on environmental factors such as temperature, salinity, atmospheric CO 2 concentration, DIC speciation, and rate at which CaCO 3 precipitates, a certain fraction (q) of the CO 2 released to the aqueous solution will evade to the atmosphere [Smith, 1985; Frankignoulle et al., 1994; Lerman and Mackenzie, 2005]. At the environmental conditions since the end of pre-industrial time, taken as the year 1700, to the early 2000s, q was calculated as 0.51 to 0.66 mol/mol between 15 and 25 C, and atmospheric CO 2 concentrations between 280 to 375 ppmv [Lerman and Mackenzie, 2005]. On the basis of the preceding considerations, the model SOCM partitions the carbon released to seawater upon the precipitation of CaCO 3 as 60% gaseous release (q = 0.6) to the atmosphere and 40% incorporation into the DIC pool of seawater. The rates of CaCO 3 precipitation and dissolution as functions of the degree of carbonate saturation of ocean water and temperature were calculated from kinetic relationships given by Andersson et al. [2005]. [11] A change in the organic carbon balance of the system is represented by the net ecosystem metabolism (NEM*), defined in a nonsteady state from equation (2) as [Andersson and Mackenzie, 2004], NEM* ¼ F GPP F R ¼ F Corg outflow þ F Corg S F Corg inflow þ dc org =dt: Calculation of the NEM* and its component terms is detailed by Andersson et al. [2005]. [12] Because the surface water is assumed to attain equilibrium with the atmosphere instantaneously, the last term in equation (4), dc DIC /dt, is the time rate of change of total dissolved inorganic carbon content (C DIC ) of the surface water as a function of changes in the inorganic carbon content of the atmosphere according to the Revelle- Munk function [Bacastow and Keeling, 1973; Revelle and Munk, 1977]. Finally, the CO 2 flux across the air-sea interface is from equations (4) and (7), F CO2 ð7þ ¼ ðf DIC outflow F DIC inflow ÞþNEC þ NEM* þ dc DIC =dt: ð8þ At the onset of simulation in the year 1700, equation (8) yields from the data shown in Figure 1 (in 10 12 mol Cyr 1 ), F CO2 ¼ ð1503 f32 þ 1504:2gÞþð24:5 6Þ þ ð18 þ 9 26 8Þþ0 ¼ 21:7: ð9þ [13] Bearing in mind the uncertainties of the flux estimates that are given in the literature and in Figure 1, the calculated CO 2 emission of 21.7 10 12 mol C yr 1 is in very good agreement with the result of 21 ± 5 10 12 mol C yr 1, obtained from the CO 2 air sea transfer model of Lerman and Mackenzie [2005] over a range of temperature and other environmental conditions that takes into account inflows and outflows of DIC and reactive C org, as well as the formation and net storage of CaCO 3 and C org in sediments. [14] The major forcings on the natural initial steady state condition in SOCM are atmospheric CO 2 concentration, variations in global mean surface temperature, and organic matter and nutrient inputs to the coastal ocean [see Andersson et al., 2005]. Atmospheric CO 2 concentrations and temperature are according to historical records and the IPCC IS92a, business-as-usual (BAU) scenario until the year 2100 [Enting et al., 1995; Intergovernmental Panel on Climate Change, 1996, 2001]. Inputs of organic and inorganic carbon and nutrients (N and P) via rivers and runoff and upwelling are based on observational data [Meybeck, 3of13

Figure 1. Schematic of the Shallow-water Ocean Carbonate Model (SOCM) and preindustrial carbon estimates adopted in the current simulations [Andersson, 2003]. SOCM consists of two major domains representing the water column (surface water and organic matter) and the pore water sediment system (pore water, sediment organic matter, river-derived particulate inorganic carbon (PIC), calcite, aragonite, and 15 mol% magnesian calcite). Reservoir masses (shown in italics) are in units of 10 12 mol C. Arrows denote carbon fluxes between reservoirs (shown inside parentheses) in units of 10 12 mol C yr 1. For twoway arrows the direction of the net flux is shown next to the flux estimate. The dashed lines indicate carbon flux owing to CaCO 3 production (equation (5)) symbols O, R, C, A, and M represent fluxes to and from sediment reservoirs organic matter, river-derived PIC, calcite, aragonite, and magnesium calcite, respectively. 1982; Meybeck and Ragu, 1995] and model results from TOTEM until the year 2025, and linearly extrapolated thereafter. At the initial condition of the model simulation in the year 1700, the input and output fluxes of C, N, and P were assumed to be balanced and the shallow-water ocean environment was considered to be in a quasi-steady state. At this time, the carbon material balance is consistent with a coastal water residence time of 4 years. The trends from 1700 to 2025 are discussed extensively by Ver et al. [1999a, 1999b], Mackenzie et al. [2002], and Lerman et al. [2004]. The following trends from 2000 to 2300 should be noted: (1) atmospheric CO 2 increases from about 370 ppmv in the year 2000 to 1730 ppmv in 2300; this represents an increase of almost 2900 Gt of carbon solely in the atmosphere, which is equivalent to approximately 58% of the global fossil fuel reserves of coal, oil, and gas [Sundquist, 1985; Kvenvolden, 1988]; (2) global mean temperature increases about 6.7 C from 2000 to 2300; and (3) riverine inputs of dissolved inorganic carbon, dissolved organic carbon, labile and nonlabile particulate organic carbon, particulate inorganic carbon, each increase by approximately 195%, whereas total nitrogen and total phosphorus increase 277% and 294%, respectively. Since these increases are based mainly on linear extrapolations into the future, they probably represent the near upper bounds in the ranges of estimates. In the following, the forcings are used as inputs to SOCM to examine the behavior of CO 2 carbonic acid carbonate system dynamics and cycling in the coastal ocean. 4of13

[15] We recognize the problems inherent in any model predictions for a physically and chemically heterogeneous environment as that of the coastal ocean, but we are encouraged to do so because both the TOTEM and SOCM model calculations are supported by the available historical data [Neftel et al., 1985; Friedli et al., 1986; Sarmiento and Sundquist, 1992; Barnola et al., 1995; Etheridge et al., 1996; Keeling et al., 1996; Winn et al., 1998; Keeling and Whorf, 2002; Borges, 2005] (see also Hawaiian Ocean Time Series (HOTS), 2004, available at http://hahana.soest.hawaii.edu/hot/hot_jgofs.html), and a similar atmospheric CO 2 forcing was employed by Archer et al. [1998] in a model study of the long-term fate of fossil fuel CO 2 and the role of the ocean and deep-sea carbonate sediments in the neutralization of anthropogenic CO 2. 3. Six Hundred Years of Changing Conditions in the Coastal Ocean [16] In this section we present SOCM results for changing conditions in the coastal ocean for 600 years of the Anthropocene under rising temperatures and atmospheric CO 2 concentrations and loading of coastal zone waters with organic carbon and nutrients, including important aspects of its marine carbon chemistry, carbonate saturation state, NEM*, NEC, carbon flows, and sediment-pore water system. This sets the temporal framework for discussion in further sections of some issues related to these changes. 3.1. Changing Marine Inorganic Carbon Chemistry [17] As the atmospheric concentration of CO 2 increases, the flux of this gas into the surface ocean will also increase following Henry s law. The dissolving CO 2 reacts with water and forms aqueous CO 2 and carbonic acid, which dissociates into bicarbonate, carbonate, and hydrogen ions. The net result of this homogeneous chemical process can be illustrated by the simplified reaction CO 2 þ H 2 O þ CO 2 3 ¼ 2HCO 3 ; ð10þ which demonstrates how an increase in CO 2 essentially titrates carbonate ions to form bicarbonate ions with a concomitant shift to lower ph values and higher DIC concentrations of the solution. Near the present-day atmospheric CO 2 concentration, the relative increase in total DIC is approximately only 10% of the relative atmospheric CO 2 increase [Bacastow and Keeling, 1973]. As the partial pressure of CO 2 increases, the relative increase in DIC becomes smaller and consequently the CO 2 buffer capacity of seawater also decreases. In contrast to this simple homogeneous reaction involving the dissolution of CO 2 in seawater, because dissolution at undersaturation of CaCO 3 minerals by CO 2 -charged water is a heterogeneous reaction and produces carbonate ions, the CO 2 buffer capacity of seawater can be enhanced by this process. [18] Because increasing atmospheric CO 2 lowers the ph and carbonate ion concentration in seawater, it also results in a decrease in the seawater saturation state (W) with respect to carbonate minerals, such as calcite and aragonite. The degree of saturation is determined by the following relationship: W ¼ Ca2þ CO 2 3 ; ð11þ K sp * where K* sp denotes the stoichiometric solubility product, and [Ca 2+ ] and [CO 3 2 ] are total concentrations of calcium and carbonate ions, respectively. Because the calcium concentration is nearly constant in the ocean and is by a factor of about 30 to 50 greater than the concentration of the carbonate ion in surface water, the saturation state is mainly controlled by the abundance of this anion. Although in a supersaturated solution (W > 1) a carbonate phase can precipitate, this is strictly not always the case because of kinetic constraints and inhibition by various components of seawater, such as magnesium and phosphate ions and dissolved organic matter [e.g., Berner et al., 1978; Morse, 1983; Morse and Mackenzie, 1990]. As an example, most surface waters are supersaturated with respect to calcite by 5 to 6 times, but no significant abiotic precipitation of carbonate minerals is observed other than in a few environments such as the Great Bahama bank and the Persian Gulf [Morse and Mackenzie, 1990]. Instead, most calcium carbonate in the modern ocean is of biological origin, produced by marine organisms that make their skeletons, shells, and tests out of CaCO 3. [19] Figure 2 shows the results of SOCM calculations for changing coastal ocean inorganic carbon chemistry and carbonate saturation state from the year 1700 to 2300. As atmospheric CO 2 and temperature rise during this period of time, the DIC content of coastal zone seawater increases about 17% as the CO 2 solubility increase due to a higher atmospheric CO 2 partial pressure exceeds its lower solubility at a higher temperature. As increasing amounts of CO 2 are absorbed by the coastal ocean, the ph falls and the distribution of marine carbon species changes, resulting in an increase in HCO 3 and dissolved CO 2 relative to CO 3 2 (Figure 2a), as follows from equation (10). In concert with this decrease in ph and rise in DIC concentration is a decline in the saturation state of seawater with respect to carbonate minerals, mainly due to the decline in the concentration of CO 3 2 (equation (11); Figure 2b). The saturation state with respect to calcite falls by 73% during the 600 years of numerical simulation where most of the decrease occurs after the year 2000. At the end of the simulation, coastal ocean surface seawater is approximately saturated with respect to 15 mol% magnesian calcite (solubility curve from Bischoff et al., [1987, 1993]) but about 174% and 126% oversaturated with respect to calcite and aragonite, respectively. However, the seawater at this stage is undersaturated with respect to many magnesian calcite compositions of higher magnesium content. These calculations are for the coastal ocean as a whole; in the higher latitude, colder surface ocean waters, despite warming sea surface temperatures, these waters will be undersaturated with respect to all these phase compositions at this time. As an effect of decreasing carbonate saturation state and increasing temperature in the model simulation, biogenic carbonate production decreased 5of13

the system. Figure 2c illustrates how these features of the coastal ocean have changed and will change during the 600 years of the SOCM simulation. As the coastal ocean receives increasing amounts of organic carbon and nutrients, the balance between gross primary production (GPP) and total aerobic and anaerobic respiration, that is net ecosystem metabolism (NEM*), changes. Prior to about the year 2000, total respiration exceeds GPP and the system is net heterotrophic. After the year 2000, the system becomes increasingly autotrophic, thus acting as a sink of atmospheric CO 2 solely due to organic metabolism where GPP, driven by the higher nutrient inputs from land, exceeds total respiration. The magnitude of the CO 2 flux due to the organic carbon imbalance (NEM*) in the year 2300 is equivalent to 52 10 12 mol C yr 1 from the atmosphere to coastal waters. The total coastal ocean air sea flux at this time is about 172 10 12 mol C yr 1, reflecting the rise in atmospheric CO 2, the changing organic metabolism imbalance, the declining formation of carbonate minerals in the coastal ocean, and the increasingly positive difference between the DIC outflow and inflow, the term (F DIC outflow F DIC inflow ) in equation (8) (Figure 2c). Figure 2. Surface water dissolved inorganic carbon chemistry between 1700 and 2300 calculated by SOCM at 25 C and a salinity of 35 psu. (a) Total dissolved inorganic carbon concentration [DIC], [CO 2 ], [HCO 3 ], and [CO 2 3 ]. (b) Surface water saturation state (W) with respect to calcite, aragonite, and 15 mol% magnesian calcite. (c) Air-sea CO 2 exchange (equation (8)) owing to net ecosystem metabolism (NEM*; equation (7)), net ecosystem calcification (NEC; CaCO 3 precipitation less dissolution), atmospheric CO 2 concentration (dc DIC /dt), and the imbalance of DIC between export to the open ocean and input from rivers and upwelling (F DIC outflow F DIC inflow ) between 1700 and 2300. by 90% from 24.5 10 12 mol CaCO 3 yr 1 in year 1700 to 2.5 10 12 mol CaCO 3 yr 1 in year 2300. 3.2. Changes in Variables Related to Air-Sea Exchange of CO 2 [20] As atmospheric CO 2 and temperatures rise and nutrient and organic carbon loading of the global coastal ocean continues into the future, there will be considerable changes in the direction and/or magnitude of air-sea CO 2 exchange, NEM*, NEC, and flow of carbon through 3.3. Sediment Pore Water System [21] The sediment pore water system of the coastal ocean reacts somewhat differently to rising temperatures and increasing atmospheric CO 2 than its surface waters. The changes in the marine inorganic carbon chemistry and carbonate saturation state of the sediment-pore water system are mainly driven by the increasing organic carbon flux to the sediment-water interface. This is due to increasing inputs of organic carbon to the coastal zone and increasing new production in the coastal zone from greater river and groundwater discharges of nutrients during the time period of the model simulation. Figure 3a shows how the saturation state of the pore waters changes through time for various carbonate minerals. Just prior to the year 2065, the pore waters go undersaturated with respect to the 15 mol% magnesian calcite of the model and hence this phase can potentially dissolve. At the end of the simulation, the pore waters are approximately saturated with respect to aragonite and still about 140% oversaturated with respect to calcite. Figure 3b illustrates the change in the organic carbon flux for the 600 years of the model simulation and the consequent production of CO 2 in the sediment-pore water system resulting from the microbial decay of the sedimented organic carbon which drives the dissolution of carbonate phases in the sediment. It should be pointed out that prior to the pore water becoming undersaturated with respect to the 15 mol% magnesian calcite, the calcite phases of higher magnesium content were already dissolving as the pore water saturation state fell through time. During the 600 years of the simulation, the carbonate dissolution flux increases from 6 to 22 10 12 mol C yr 1, or 267% (Figure 3a). This dissolution flux, mainly involving high magnesian calcite phases, could increase the CO 2 buffer capacity of the coastal ocean according to equation (10). In fact, such dissolution has been proposed to act as a buffer to neutralize anthropogenic CO 2 uptake by the ocean and 6of13

Figure 3. (a) Pore water saturation state with respect to calcite, aragonite and 15 mol% magnesian calcite calculated at 25 C and a salinity of 35 psu as well as total carbonate dissolution between the years 1700 and 2300. The carbonate ion concentration [CO 2 3 ] is shown for year 1700, 2065 and 2300. (b) Organic matter sediment flux and CO 2 production owing to decay or remineralization of organic matter. possibly counteract any negative effects of rising atmospheric CO 2 on calcareous organisms and carbonate systems [Barnes and Cuff, 2000; Halley and Yates, 2000]. However, the results of SOCM suggest that increased dissolution of metastable carbonate minerals will not produce sufficient surface water alkalinity to restore to a significant extent any changes in ph and carbonate saturation state in the global coastal ocean arising from increasing atmospheric CO 2 [Andersson et al., 2003] (see also discussion in section 6). An exception to this may be observed in some reef and carbonate ecosystems that have long water residence times. [22] In the model simulations, the observed increase in carbonate dissolution is mainly the result of a decrease in pore water carbonate saturation state and not that of the surface water, which remains supersaturated with respect to the average magnesian calcite phase of 15 mol% MgCO 3 by the year 2300. Ultimately, decreasing carbonate saturation state of surface water will reach the upper layers of the sediment-pore water system and could lead to enhanced dissolution of carbonate minerals, particularly under the shallow oxic conditions of microbial respiration of organic carbon where the initial seawater buried in the sediments starts out with a lower ph and carbonate saturation state than before the perturbation. However, the modeled decrease in pore water carbonate saturation state is mainly due to increasing input and subsequent remineralization of organic matter in the global coastal zone sediments, rather than due to increasing atmospheric CO 2. This result is in agreement with the role of organic matter remineralization in the dissolution of carbonate sediment that has been well demonstrated as occurring under aerobic conditions in the pore waters of sediments in the Gulf of Calvi, Corsica [Moulin et al., 1985]. Selective dissolution of magnesian calcite also occurs in the water column of Devil s Hole, Bermuda [Balzer and Wefer, 1981], as a result of extensive remineralization of organic matter below the seasonal thermocline during summer, producing elevated levels of CO 2 and low carbonate saturation state. [23] Differential dissolution of carbonate mineral phases will affect the average carbonate mineral composition and the rate of formation of carbonate cements in the pore water-sediment system [Garrels and Wollast, 1978; Mackenzie and Agegian, 1989; Morse and Mackenzie, 1990; Andersson et al., 2003]. Because carbonate cements function as the glue for reef structures, any changes to their content and/or composition could affect the strength and ability of a reef to withstand environmental stresses. The starting mineral composition of the coastal sediments in SOCM (Figure 1) is taken as calcite 13 weight%, aragonite 63 wt%, and 15 mol% Mg-calcite 24 wt%. Since the start of simulation in the year 1700 to 2000, the average composition of the sediment has not changed owing to the changes in atmospheric CO 2 and seawater chemistry. However, a projection to the year 2300 shows that the calcite weight fraction is unchanged (13 wt%), the aragonite fraction increases to 66 wt%, and the Mg-calcite fraction decreases to 21 wt%. These relatively small changes in the sediment reflect the large mass ratio sediment/water that is taken in the model. 4. Model Results Compared With Observational Data [24] In this section we demonstrate that the SOCM calculations for air-sea CO 2 exchange and surface water carbonate chemistry accord well with observational data available from measurements of exchange in coastal zone ecosystems and from time series studies of the open ocean. Because the databases are somewhat limited in geographical coverage and time, we limit the model simulations to the past history of the marine carbon system and its projections to the end of the 21st century. In addition, the results of the comparisons between model output and observational data also provide us with some test of the validity and robustness of the model and its ability to predict at least trends, if not magnitudes, of changes in the marine carbon system of the coastal ocean on into the future. 4.1. Air-Sea CO 2 Exchange in Coastal Ecosystems [25] The results of SOCM suggest that the global coastal ocean has served as a net source of CO 2 to the atmosphere throughout most of the past 300 years (Figure 4) owing to the process of calcification and a state of net heterotrophy, which leads to generation of CO 2 [Andersson and Mackenzie, 2004; Mackenzie et al., 2004]. However, the role of this region has recently switched, or soon will do so, to a net sink of CO 2 because of rising atmospheric CO 2 and potentially increasing inorganic nutrient loading from land that stimulates new production and net autotro- 7of13

Figure 4. Net air-sea CO 2 exchange (10 12 mol C yr 1 ) between 1700 and 2100 calculated by SOCM adopting an average coastal water residence time (t) of 4 years and 12 years. The solid line indicates the average flux of the two scenarios. The data points are the recent net air-sea CO 2 exchange estimates by Borges [2005] based on observational data from coastal zone ecosystems: coastal zone including estuaries (solid circle) acts as a net source of CO 2, and excluding estuaries (open circle) the coastal region acts as a net sink of CO 2. Note that estuaries (1 10 6 km 2 ) constitute less than 4% of the global coastal region (28 10 6 km 2 ). phy. The calculated CO 2 flux significantly depends on the term (F DIC outflow F DIC inflow ) in equation (8), which represents a small difference between two very large fluxes that are not well known, but can be defined on the basis of a material balance. As a consequence, the flux of CO 2 is strongly affected by the initial carbon material balance and the residence time of water in the coastal ocean. In previously published results of SOCM [Andersson and Mackenzie, 2004], the difference between these large flux terms was taken as approximately zero, because of the uncertainty associated with the terms, and it did not affect the air-sea CO 2 exchange substantially. [26] The current model projections of the net air-sea CO 2 exchange agree well with the estimates of Borges [2005], based on CO 2 flux data from different coastal environments and extrapolated to the global coastal region. Including estuaries, Borges concluded that the global coastal region currently serves as a net source of CO 2 to the atmosphere of approximately 10 10 12 mol C yr 1. This value agrees relatively well with the estimate by SOCM of 6 10 12 mol C yr 1 for the year 2000 (Figure 4), adopting a relatively long coastal ocean water residence time of 12 years (Figure 1, DIC mass/f DIC outflow = 6000 10 12 mol C/480 10 12 mol C yr 1 12 yr). However, if estuaries were not included, Borges result is that the global coastal region serves as a net sink of CO 2 equivalent to approximately +30 10 12 mol C yr 1 (Figure 4), which agrees well with the results of SOCM adopting a shorter, and probably more reasonable, residence time of approximately 4 years (Figure 1, F DIC outflow = 1503 10 12 mol C yr 1 ). 4.2. Open Ocean Marine Carbon Time Series Observations [27] Owing to the increase in atmospheric CO 2, the dissolved inorganic carbon concentration of seawater has already been observed to increase in surface waters at the locations of the Hawaiian Ocean Time series (HOT), located in the North Pacific subtropical gyre [Winn et al., 1998] (see also Hawaii Ocean Time series at http://hahana.soest. hawaii.edu/hot/hot_jgofs.html), and the Bermuda Atlantic Time-series Study (BATS), in the North Atlantic subtropical gyre [Bates et al., 1996; Bates, 2001]. The surface water carbonate saturation state has also decreased at these locations, causing the carbonate saturation horizon to become shallower, which has been observed in several regions of all major ocean basins [Feely et al., 2004]. Figure 5 illustrates the observed decrease in surface water calcite saturation state at station ALOHA between 1989 and 2001 (Hawaii Ocean Time series) compared to single measurements taken at nearby stations during the CO 2 dynamics cruise in 1981 (ENP) [Chen et al., 1986], the GEOSECS cruise in 1973 [Takahashi et al., 1980], and the calcite saturation state projected by SOCM, which agrees well with the trend observed at station ALOHA. A decreasing saturation state trend with respect to calcite is also apparent in inshore data from Bermuda (Figure 5), although caution is advised because data are scarce and the CO 2 system was measured and analyzed by different investigators. The calculated carbonate saturation state of the surface water of the global coastal ocean may decrease by 46% between the year 1700 and year 2100 (Figure 5), although the surface water will remain supersaturated with respect to calcite, aragonite, and 15 mol% Mg-calcite (Figure 2b). 5. Effect of Calcium Carbonate Saturation State on Calcification Rates [28] Increasing atmospheric CO 2 and subsequently decreasing surface water carbonate saturation state may have a negative effect on marine calcifying organisms because their ability to calcify depends on the carbonate saturation state. Numerous experimental investigations have demonstrated the existence of a direct relationship between the rate of calcification and seawater carbonate saturation state that is related to pco 2 for a number of different calcifying organisms, such as coccolithophorids [Riebesell et al., 2000; Zondervan et al., 2001; Sciandra et al., 2003], foraminifera [Spero et al., 1997; Bijma et al., 1999], coralline algae [Smith and Roth, 1979; Borowitzka, 1981; Mackenzie and Agegian, 1989, Gao et al., 1993], and scleractinian corals [Gattuso et al., 1998; Marubini and Thake, 1999; Marubini et al., 2001, 2003; Reynaud et al., 2003]. In addition, a positive correlation between calcification rate and the degree of carbonate mineral saturation of ocean water was found in experiments conducted on typical calcareous communities in incubation chambers and mesocosms [Halley and Yates, 2000; Leclercq et al., 2000, 2002], and on the artificial reef of Biosphere 2 [Langdon et al., 2000, 2003]. On the basis of these experimental results and future projections of surface water carbonate saturation, marine biogenic calcification may decrease by 8of13

Figure 5. Average annual surface water calcite saturation state (W) at station ALOHA (Hawaiian Ocean Time Series (HOTS), 2004, available at http://hahana.soest.hawaii.edu/ hot/hot_jgofs.html) and single measurements taken at nearby stations during the CO 2 dynamics cruise in 1981 (ENP) [Chen et al., 1986] and the GEOSECS cruise in 1973 [Takahashi et al., 1980] (open circles). Inshore data from Mangrove Bay, Bermuda (solid circles) and the surface water calcite saturation calculated by SOCM between 1700 and 2100 (solid line) are also shown. The inset shows changes at station ALOHA and nearby stations between 1970 and 2005 highlight the trend on this timescale. The error bars indicate 1 standard deviation. Because the exact locations differ and the carbonate system was analyzed by different investigators, the linear regression is only for the data from HOT (dashed line: y = 0.0252t + 56.1284; R 2 = 0.82; p 0.01). The dissolved inorganic carbon system and calcite saturation state were calculated at in situ temperature and salinity in each case from either total alkalinity or pco 2 together with total DIC using the program CO2SYS [Lewis and Wallace, 1998] and the carbonate dissociation constants of Mehrbach et al. [1973], refit by Dickson and Millero [1987]. For details on the solubility constant of calcite adopted in the calculation see Lewis and Wallace [1998]. 11 to 44% owing to a doubling of the atmospheric CO 2 concentration relative to pre-industrial conditions. Since the onset of the industrial revolution, the rate of calcification of marine calcifying organisms is expected to have already declined by 6 to 14% [Gattuso et al., 1999; Langdon, 2002; Kleypas et al., 1999; Buddemeier et al., 2004]. It should be pointed out that the rate of calcification is also strongly related to other parameters such as temperature, light [Marubini et al., 2001], nutrients [Paasche and Brubak, 1994; Marubini and Atkinson, 1999; Ferrier-Pagès et al., 2000; Sciandra et al., 2003], and metabolic and photosynthetic activity of the organism if the organism is autotrophic or dependent on autotrophic symbionts. [29] Contrary to the foreseen decline in calcification rate, analyses of drill cores taken from Porites colonies along the Great Barrier Reef suggest that the rate of calcification of these corals has increased rather than decreased between 1880 and the later part of the twentieth century [Lough and Barnes, 2000]. Similar results showing increased calcification rates toward the present have also been obtained from a coral core taken in French Polynesia [Bessat and Buiges, 2001]. Lough and Barnes [2000] attributed the observed increase in calcification on the Great Barrier Reef to an increase in the average annual sea surface temperature during the twentieth century and concluded that calcification rates have significantly increased in response to global warming. In SOCM simulations, adopting different forms of the calcification rate dependence on the saturation state and temperature, an increase in DIC and temperature of seawater results in a higher calcification rate [Andersson et al., 2005]. However, once the surface water carbonate saturation state became lower than a certain threshold in the second half of the 21st century, the negative effect of this decrease becomes stronger than the positive effect of increasing DIC and temperature and the rate of calcification significantly decreases in the model simulations. In addition, it is difficult to defend an argument that increasing sea surface temperatures (SSTs) will continue to promote increased global calcification rates in the future. Corals have been observed to bleach in coral reef ecosystems of the Caribbean, Bermuda, Hawaii, and the South Pacific under conditions of only a 1 to 2 C rise in SST, leading to morbidity or mortality of the corals [Buddemeier et al., 2004]. Thus it is likely that increasing SSTs will only lead globally to more bleaching events and enhanced stress on coral reef ecosystems. 6. Buffering of Rising CO 2 in Surface Water by Carbonate Dissolution [30] In order to buffer the surface water ph and carbonate saturation state against the effects of rising atmospheric CO 2, dissolution of CaCO 3 minerals in the global surface ocean essentially has to balance the net invasion of anthropogenic CO 2 (170 10 12 mol C yr 1 ). Langdon [2002] pointed out that this condition can be met by a mean, but very high, global dissolution rate of 17 mmol CaCO 3 m 2 day 1 (170 10 12 mol C yr 1 ). Even if dissolution of CaCO 3 in the surface sediment layer were equal to the current rate of global CaCO 3 production, estimated as between 53 and 94.3 10 12 mol C yr 1 [Milliman, 1993; Wollast, 1994; Milliman and Droxler, 1996; Mackenzie et al., 2004], the surface water would still only be partially buffered. However, dissolution rates of this magnitude are very unlikely since most of the CaCO 3 produced in pelagic environments (65 10 12 mol C yr 1 ) sinks and dissolves at deeper depths because surface waters are oversaturated with respect to most carbonate minerals. Thus because the physical exchange of water between the shallow-water ocean environment and the open ocean is relatively fast, substantial dissolution of metastable carbonate minerals within the shallow-water region is necessary to produce sufficient alkalinity to counteract any changes in the surface water chemistry owing to increasing atmospheric CO 2. Current estimates of CaCO 3 dissolution in the global coastal ocean range from 6.7 10 12 mol C yr 1 [Milliman, 1993; Wollast, 1994] to 10 10 12 mol C yr 1 [Langdon, 2002], the higher estimate 9of13

corresponding to about 6% of the average net anthropogenic invasion of CO 2 into the global ocean during the 1980s (160 10 12 mol C yr 1 )[Sarmiento and Gruber, 2002]. A calculated dissolution rate increase in the coastal ocean only is given in section 3.3 as 6 to 22 10 12 mol C yr 1. According to Langdon [2002], direct measurements of carbonate dissolution range from 0 to 13.7 mmol CaCO 3 m 2 day 1, which if extrapolated to the entire shallow-water ocean environment corresponds to 0 to 140 10 12 mol C yr 1. A dissolution rate of the latter magnitude is more than 5 times the amount of CaCO 3 produced annually within the global coastal ocean (24.5 10 12 mol C yr 1 ) and implies a large loss of calcium carbonate minerals from reef structures, carbonate banks, and sediments. 7. Discussion of the Coastal Carbonate System to the Year 2300 and Beyond [31] Despite a decrease in the saturation state of coastal seawater with respect to carbonate minerals, the results of SOCM show that warm surface seawater will remain supersaturated with respect to a magnesian calcite phase of 15 mol% MgCO 3 until the final years of the simulation (Figure 2). However, it is important to realize that surface water in high latitudes will become undersaturated with respect to this phase, and also other carbonate phases with higher solubility, much earlier than this projection. According to Orr et al. [2005], Southern Ocean surface waters may be undersaturated with respect to aragonite by the year 2050, consequently exerting a significant negative effect on high-latitude calcareous organisms. [32] It would be anticipated that as the magnesium content of various organisms with magnesian calcite skeletons changed in the future, clasts and calcareous debris produced from their mechanical and biological disintegration would become less magnesium rich. It can also be seen in Figure 3 that, because reactive organic matter loading of coastal sediments is anticipated to increase in the future, enhanced decomposition of organic carbon in the interstitial waters of these sediments will result in a further decrease of their carbonate saturation state. Magnesian calcite cements in marine sediments would tend to have lower magnesium contents. As a result of these changes, carbonate sediment composition in the future would tend to be enriched in aragonite and calcite relative to magnesian calcite high in magnesium. Analogous to the calcite seas of the geologic past [Sandberg, 1983, 1985], several centuries into the future magnesian calcite and aragonite precipitates may become less abundant and calcites of lower MgCO 3 content more abundant. It is likely that the low saturation state predicted for the future might also affect the formation of carbonate ooids and whitings. The rate of formation of these carbonate deposits that are to some extent chemical precipitates from seawater may follow the general relationship of a slower precipitation rate at a lower degree of saturation of the solution [Morse and Mackenzie, 1990]. [33] About the time of the final years of the current model simulation, the atmospheric CO 2 concentration is anticipated to start to decrease slowly [Archer et al., 1998; Caldeira and Wickett, 2003]. At this time, the global reservoirs of conventional fossil fuels of approximately 5000 Gt C [Sundquist, 1985; Kvenvolden, 1988] would be very close to exhaustion and anthropogenic CO 2 emissions would have been on the decline for several decades. On a timescale of several centuries beyond our current model simulations, the ocean will absorb a substantial fraction of the anthropogenic CO 2 that has accumulated in the atmosphere. In part, the extent of CO 2 uptake will be dependent on the formation of deep and intermediate waters and the mixing time of the oceans, which is of the order of 1000 years. Current evidence suggests that the ocean has taken up about 46 to 50% of the total CO 2 emitted from burning of fossil fuels and cement manufacturing between 1800 and the late 1990s [e.g., Mackenzie et al., 2001; Sabine et al., 2004]. The uptake of anthropogenic CO 2 by the oceans, as mentioned previously, will result in a shoaling of the carbonate saturation horizon in the major ocean basins, a phenomenon already being observed [Feely et al., 2004]. According to Archer et al. [1998], dissolution of deep-sea carbonate minerals will increase owing to increasing CO 2 content of oceanic bottom waters, but will not play a significant role in the fate of anthropogenic CO 2 in the coming centuries. However, on a timescale of 5 to 6 ka, CO 2 neutralization by seafloor carbonates may account for 60 to 70% of the anthropogenic CO 2 that has entered the ocean. As the atmospheric CO 2 concentration starts to decline, the surface water ocean carbonate saturation state most likely will start to increase. Nevertheless, the results of the present simulations clearly demonstrate that by the year 2300, coastal carbonate environments may have changed significantly relative to today. The return toward the initial state will in part depend on the response of the open ocean to the perturbation, the residence time of global coastal waters, and the rates of calcification of the shallow-ocean biological communities at that time. 8. Conclusions [34] The results of the numerical model simulations demonstrate that the CO 2 carbonic acid carbonate system of the shallow-water coastal ocean will be significantly affected under future conditions of a higher CO 2 world and increased riverine transport of organic matter and nutrients to this region. The net flux of CO 2 between the surface water and the atmosphere is expected to reverse from net evasion from the coastal waters to net invasion owing to increased net ecosystem production, decreased net ecosystem calcification, and increased atmospheric CO 2 concentration from burning of fossil fuels and land-use activities. Surface water saturation state with respect to carbonate minerals (calcite, aragonite, and magnesian calcites) will decrease and, consequently, the rates at which calcareous organisms calcify will also decrease. It remains to be demonstrated what the ecological implications of such changes will be. It is also unlikely that the negative effect of decreasing carbonate saturation state could be counteracted by other environmental variables, such as increasing temperature and DIC concentration, which indirectly may have a positive effect on the rates of biogenic calcification. Dissolution of metastable carbonate minerals will increase, 10 of 13