Modeling the marine aragonite cycle: changes under rising carbon dioxide and its role in shallow water CaCO3 dissolution

The marine aragonite cycle has been included in the global biogeochemical model PISCES to study the role of aragonite in shallow water CaCO 3 dissolution. Aragonite production is parameterized as a function of mesozooplankton biomass and aragonite saturation state of ambient waters. Observation-based estimates of marine carbon- 5 ate production and dissolution are well reproduced by the model and about 60% of the combined CaCO 3 water column dissolution from aragonite and calcite is simulated above 2000 m. In contrast, a calcite-only version yields a much smaller fraction. This suggests that the aragonite cycle should be included in models for a realistic representation of CaCO 3 dissolution and alkalinity. For the SRES A2 CO 2 scenario, production 10 rates of aragonite are projected to notably decrease after 2050. By the end of this century, global aragonite production is reduced by almost one third and total CaCO 3 production by 19% relative to pre-industrial. Geographically, the e ﬀ ect from increasing atmospheric CO 2 , and the subsequent reduction in saturation state, is largest in the subpolar and polar areas where the modeled aragonite production is projected to 15 decrease by 65% until 2100.


Introduction
The ocean calcium carbonate (CaCO 3 ) budget is a topic of long standing interest in marine geochemical research (e.g. Berger, 1978;Milliman, 1993;Milliman and Droxler, 1996;Milliman et al., 1999;Iglesias-Rodriguez et al., 2002;Berelson et al., 2007). 20 The growing awareness of the impacts of rising atmospheric CO 2 concentrations on the chemistry of the ocean and related consequences on marine biota in general and calcifying organisms in particular has recently revived this area of study (e.g. Caldeira and Wickett, 2003;Feely et al., 2004;Orr et al., 2005;Royal Society, 2005, Kleypas et al., 2006. In a recent review of the marine CaCO 3 cycle, Iglesias-Rodriguez et (60-80%) of pelagic CaCO 3 production in the upper 500 to 1000 m of the water column (e.g. Milliman and Droxler, 1996;Berelson et al., 2007). This contradicts the general paradigm of the conservative nature of pelagic CaCO 3 at shallow depths (Sverdrup et al., 1942). The latter is directly derived from the evolution with depth of the saturation product for CaCO 3 minerals which increases with decreasing temperature and 10 increasing pressure/depth (Zeebe and Wolf-Gladrow, 2001). As a consequence surface ocean waters are at present oversaturated with respect to CaCO 3 (e.g. Orr et al., 2005) precluding dissolution from a thermodynamic point of view.
Mechanisms proposed to explain the occurrence of CaCO 3 dissolution above the saturation horizon invoke biological mediation. It was proposed that CaCO 3 particles 15 would dissolve after ingestion by pelagic grazers during gut passage (Harris, 1994). Until recently, dissolution during gut passage had, however, not been unequivocally demonstrated (Honjo and Roman, 1978;Pond et al., 1995). Antia et al. (2008) have shown that microzooplankton grazing may explain the reported amplitude of dissolution fluxes, while its exact quantitative contribution to shallow-water dissolution remains 20 uncertain. Alternatively, the oxic remineralization of particulate organic C in aggregates could be at the origin of undersaturated microenvironments triggering the dissolution of CaCO 3 particles associated to POC within the same aggregate (Milliman et al., 1999). However, Jansen et al. (2002) concluded from a modeling study that CaCO 3 dissolution in marine aggregates in response to CO 2 production by POC remineralisation is 25 unlikely to be at the origin of observed shallow-water dissolution fluxes.
In the absence of a clear identification of processes driving shallow water CaCO 3 dissolution, we might address the topic from the side of methods used to infer the order of magnitude of dissolution fluxes. On one hand, we have the evolution of CaCO 3 Introduction fluxes recorded in sediment traps in comparison to upper ocean CaCO 3 production estimates (e.g. Milliman et al., 1999;Feely et al., 2004;Wollast and Chou, 1998). This approach has its own set of caveats (e.g. Milliman et al., 1999;Berelson et al., 2007) the complete discussion of which is beyond the scope of this paper. On the other hand, we have tracer based approaches, in particular the alkalinity based approach TA * 5 (Feely et al., 2002;Sabine et al., 2002;Chung et al., 2003). Conceptually, it relays on the breakdown of total alkalinity (TA) into three components: the preformed alkalinity TA 0 , the alkalinity of the water parcel as it leaves the surface ocean; the contribution from organic matter remineralization at depth, TA AOU , and the contribution from CaCO 3 dissolution at depth, TA * . Friis et al. (2006) conclude from a model based assessment 10 of the TA * method, that this method probably overestimates the magnitude of CaCO 3 dissolution at shallow water depth because this approach does not take into account the influence of water mass transport.
In the pelagic realm CaCO 3 is produced mainly as two polymorphs of differing solubility: calcite (coccolithophores, foraminifera) and aragonite (mostly pteropods), the 15 latter being 50% more soluble than calcite. The contribution of aragonite to shallow water CaCO 3 dissolution is mentioned by several authors (e.g. Milliman et al., 1999;Berelson et al., 2007), but judged to be insignificant on the basis of its supposed only minor contribution to total pelagic CaCO 3 production (10% according to Fabry, 1990). Estimates of aragonite production and fluxes in the modern ocean are however scarce. 20 Moreover these estimates cover a wide range, spreading from 10 to 50% of the total global CaCO 3 flux (Berner, 1977;Berger, 1978;Berner and Honjo, 1981;Fabry and Deuser, 1991). A significant contribution of aragonite to the reported shallow water dissolution can thus not be ruled out a priori. Our study explores the potential contribution of aragonite to shallow water CaCO 3 dissolution by means of a model study. 25 As a follow-up of Gehlen et al. (2007), we implemented aragonite production and dissolution to the global biogeochemical ocean model PISCES. In the first section of this paper, we present a global open ocean CaCO 3 budget for calcite and for aragonite. Model improvements are discussed with reference to the study by Gehlen et al. (2007) which was also based on PISCES, but did only consider calcite. In the second section, we assess potential changes to the global CaCO 3 budget in response to increasing atmospheric CO 2 levels following the IPCC SRES A2 scenario.

The PISCES model
2.1 General model description 5 The model used in the current study is the PISCES global ocean biogeochemical model (Aumont et al., 2003;Aumont and Bopp, 2006;Gehlen et al., 2006) which simulates the biogeochemical cycle of oxygen, carbon and the main nutrients controlling marine biological productivity: nitrate, ammonium, phosphate, silicate and iron. Biological productivity is limited by the external availability of nutrients. The phosphorus and 10 nitrogen cycles in the model are decoupled by nitrogen fixation and denitrification.
The model distinguishes two phytoplankton size classes (nanophytoplankton and diatoms) and two zooplankton size classes (microzooplankton and mesozooplankton). The C/N/P ratios are assumed constant for all species and fixed to the Redfield ratios. For phytoplankton, the prognostic variables are total biomass, iron, chlorophyll and 15 silicon contents. The internal ratios of Fe/C, Chl/C and Si/C are predicted by the model. The only prognostic variable for zooplankton is total biomass. The bacterial pool is not modeled explicitly.
There are three non-living compartments for organic carbon in PISCES: small particulate organic carbon (POC s ), big particulate organic carbon (POC b ) and semi-labile 20 dissolved organic carbon (DOC). The C/N/P composition of dissolved and particulate matter is also coupled to Redfield stochiometry. However, the iron, silicon and calcite pools of the particles are fully simulated and their ratios relative to organic carbon are allowed to vary. The production of calcite is assigned to nanophytoplankton as a function of temperature, saturation state, nutrient and light availability. The particu-Introduction aggregation from nanophytoplankton and diatoms, fecal pellet production and grazing. The model simulates dissolved inorganic carbon and total alkalinity. The carbon chemistry follows the OCMIP protocol (http://www.ipsl.jussieu.fr/OCMIP). A description of the model including model equations and parameters can be found as auxiliary material in Aumont and Bopp (2006). 5

Calcium carbonate geochemistry
The two-fold objective of this paper is (1) exploring the potential for aragonite dissolution to significantly contribute to shallow water CaCO 3 dissolution and producing a first evaluation of the impact of future rising atmospheric CO 2 levels on aragonite production and dissolution. We adopt a geochemical approach to the question. This is justified by the limited observations available on aragonite production and dissolution in the ocean. Aragonite production and dissolution was implemented to PISCES in line with the general model structure and previous developments around the CaCO 3 geochemistry in the model (Gehlen et al., 2007). This choice allows a straight-forward comparison between the model results of this study and output of PISCES in its calcite-only version 15 (referred to as PISC/CAL) (Gehlen et al., 2007).

Aragonite production
In the modern ocean pelagos, cosmopolitan shelled pteropods are the main producers of aragonite (Lalli and Gilmer, 1989). Information on pteropod biogeography and ecology is scarce and largely summarized in Lalli and Gilmer (1989). Pteropods are mostly 20 found in near-surface waters down to depths of about 200 m. Only a few species are restricted to deeper waters. We did not implement pteropods as a distinct functional type in the model, but assigned aragonite production to the corresponding zooplankton size class, mesozooplankton. Orr et al. (2005) report dissolution of pteropod aragonite shells in living organisms during exposure to water undersaturated with respect 25 to aragonite. We might thus assume that the vertical distribution of pteropods across 1660 the water column is controlled by carbonate chemistry, e.g. that the organisms, and thus aragonite production, are restricted the water masses oversaturated with respect to aragonite.
We are unaware of published data on the response of calcification in pteropods to the saturation state. Previous studies on the effect of carbonate chemistry on the cal-5 cification in marine calcifiers focused on coccolithophores (e.g. Sciandra et al., 2003, Riebesell et al., 2000Zondervan et al., 2001Zondervan et al., , 2002Delille et al., 2005), foraminifera (e.g. Wolf-Gladrow et al., 1999;Bijma et al., 1999) and corals (e.g. Gattuso et al., 1998;Kleypas et al., 1999;Langdon et al., 2003Langdon et al., , 2005. All these studies report a decrease in calcification with increasing pCO 2 or decreasing saturation state. We expect that the 10 calcification response will display a tolerance over a range of oversaturation typical of the natural variability experienced by the organism until a critical threshold is reached. Passing the threshold value, calcification is expected to decrease rapidly. The mathematical expression corresponds to a saturation curve, such as the Michaelis Menten equation derived for calcite production in Gehlen et al. (2007) from a compilation of experimental results for E. huxleyi.
Aragonite production is expressed as the ratio of particulate inorganic C in the form of aragonite (PIC A ) to particulate organic C (POC), as a function of the saturation state of seawater and mesozooplankton biomass: where: MESO, mesozooplankton biomass (mole C l −1 ); (PIC A /POC) max =0.8, maximum ratio reached under optimal growth conditions; K max =0.4, with analogy to the half saturation constant; and K sp,A corresponds to the stoichiometric ion concentration product for aragonite after Mucci (1983). The pool of 25 sinking aragonite particles is fuelled by mortality. The sum of calcite and aragonite production routed to the sinking particles is referred to as net CaCO 3 production. Introduction The particles sink with a prescribed sinking speed according to: where: w min , minimum sinking speed of 50 m d −1 ; w max , maximum sinking speed at 2000 m below the mixed layer of 200 m d −1 ; z, depth (m); z m , mixed layer depth (m).
The export flux is defined as the amount of sinking CaCO 3 particles that falls through the depth horizon of 100 m. There is no ocean-sediment interaction included in our model version. From the flux of CaCO 3 reaching the deepest model boxes (=lower boundary) an alkalinity equivalent of totally 0.18 PgC y −1 corresponding to river input 10 is removed (burial flux). The rest is re-dissolved instantaneously in the deepest grid cells. The model is strictly mass conserving.

Aragonite dissolution
Compared to the numerous studies addressing the dissolution kinetics of calcite (e.g. Morse and Berner, 1972;Berner and Morse, 1974;Morse, 1978;Honjo and 15 Erez, 1978;Plummer et al., 1978;Keir, 1980;Walter and Morse, 1985;Arakaki and Mucci, 1995;Morse and Arvidson, 2002;Gehlen et al., 2005a, b;Morse et al., 2007), only a modest number of investigations focused on the dissolution behavior of aragonite (e.g. Morse et al., 1979;Keir, 1980;Busenberg and Plummer, 1986;Morse et al., 2007). These studies converge to assign a similar rate law to aragonite and calcite 20 dissolution. The corresponding rate expression, R, for CaCO 3 dissolution writes: where: k, dissolution rate parameter, d −1 ; n, dimensionless reaction rate constant. The dissolution of CaCO 3 is in general described as a higher order reaction with estimates of n ranging from 2.3 to 4.5 (e.g. Morse and Arvidson, 2002 information available specifically on aragonite dissolution, we adopted the same rate description and reaction rate constant as for calcite in Gehlen et al. (2007). Aragonite dissolution is thus described as a first order reaction (n=1) with respect to undersaturation with a dissolution rate constant of 10.9 d −1 . The latter was derived from the evolution with depth of CaCO 3 fluxes recorded in traps deployed below 1000 m (Gehlen et 5 al., 2007). The quantity of CaCO 3 particles that re-dissolves in the deepest grid cells of the model is not included in our definition of pelagic CaCO 3 dissolution. The ratio between pelagic dissolution and net production is here denoted as relative CaCO 3 dissolution. 10 The physical fields that are used to force PISCES are based on a climatological simulation with the ORCA2 global ocean model configuration of OPA version 8.2 . The mean horizontal resolution of the model is approximately 2 • by 2 • cos latitude and the meridional resolution is enhanced to 0.5 • at the equator. The model has 30 vertical levels; with an increment that increases from 10 m at the surface to 15 500 m at depth (12 levels are located in the first 125 m). The physical/dynamical simulation with OPA employs climatological atmospheric forcing from various data sets. These include NCEP/NCAR 2m atmospheric temperature (averaged between 1948 and 2003) and relative humidity, ISCCP total cloudiness (averaged between 1983 and 2001), precipitation (averaged between 1979 and 2001), weekly wind stress based on 20 ERS and TAO observations and creates a representation of ocean circulation/mixing that is forced by observational climatologies. Please see Aumont and Bopp (2006) for more details and the associated references.

Model simulations
Model alkalinity field was initialized with the global mean alkalinity value from Goyet et al. (2000). The new parameterizations of aragonite production and dissolution were BGD 5,2008 Modeling the marine aragonite cycle R. Gangstø et al. Interactive Discussion cation of an approximate upper limit of which aragonite may contribute to the estimated shallow depth CaCO 3 dissolution. The consequent changes in the CaCO 3 cycle call for a long spin-up; hence the model was integrated over 5000 years using the degradationintegration tool DEGINT (Aumont et al., 1998) to reach quasi-equilibrium. The tracer output was used to re-initialize the PISCES model at the original resolution (2×2 • cos 5 (latitude), 31 vertical levels) and the model was then integrated over 300 years using a constant climatological ocean circulation field without interannual variability before reaching final equilibrium. Atmospheric CO 2 was all the time kept constant and equal to 278 parts per million (ppm).
To study the effect of rising pCO 2 on the carbonate cycle, a model experiment  horizon, there is a very good match between model and observations. The modeled aragonite saturation horizon follows the estimated saturation depth characteristics: it is deepest in the Atlantic, and shoals through the Indian to the Pacific Oceans following the water circulation along the deep conveyor belt, where the water masses get more and more enriched in DIC the older they get due to the remineralization of organic 5 matter.
4.1.2 CaCO 3 sources and sinks Table 1 compares global pelagic pre-industrial CaCO 3 production, export and dissolution obtained from the present model study, from a similar study (PISC/CAL) which only considered calcite (Gehlen et al., 2007) and from observation-based estimates. The  (Table 1). Finally the budget is closed by the sinking CaCO 3 particles that reach the model lower boundary. This flux does not correspond to the burial flux reported in literature and may not be directly compared to observations. 20 The modeled CaCO 3 export from the upper 100 m matches well the result by Sarmiento et al. (2002) of 0.6 PgC y −1 . The export is smaller than the net production due to a quantity of CaCO 3 production below, and a small amount of dissolution above, 100 m. The different environmental controls on calcite and aragonite production in the model lead to different global distributions of calcite and aragonite production (Fig. 2). For the calcite production (Fig. 2a), the spatial variability is well reproduced by the model: low values in central ocean gyres, high values in upwelling zones of the eastern bound- A total open ocean CaCO 3 production of 6.9-7.9 mgC m −2 d −1 was calculated by Morse and MacKenzie (1990). The model yields an average total CaCO 3 production of 6.7 mgC m −2 d −1 .

BGD
The distribution of aragonite production in the model follows ocean productivity with 15 maximum values in areas where the latter is high (e.g. upwelling zones, subpolar waters) (Fig. 2b). This pattern reflects the occurrence of mesozooplankton which is tightly coupled to the availability of food. The model simulates an average aragonite production of 2.4 mgC m −2 d −1 . Fabry (1989Fabry ( , 1990 (Betzer et al., 1984). Of the total net CaCO 3 production, aragonite production in the model (0.30 PgC y −1 ) amounts to 35%.
While the model predicts aragonite production in high latitudes of the right order of 25 magnitude, it overestimates the production in equatorial areas. About 40% of the modeled total aragonite production takes place between 0 • and 20 • while only 10% occurs at latitudes greater than 40 • . While the calcification rates in equatorial upwelling ar-eas are considered high, most of the aragonite production is assumed to take place in subpolar and polar areas (Lalli and Gilmer, 1989). This indicates a general overestimation by our model of the aragonite production in the equatorial areas, possibly in combination with an underestimation in higher latitudes. The latter is supported by an underestimation of the modeled mesozooplankton biomass in ocean gyres and subpo-5 lar areas compared to observations compiled by Buitenhuis et al. (2006), especially in the north Pacific. The low mesozooplankton biomass in ocean gyres in turn reflects the tendency of the model to underpredict ocean productivity in these regions (Gehlen et al., 2006). However, in agreement with observations, the model presents higher aragonite production in the western North and eastern equatorial Pacific and lower values in 10 the Central Pacific, which is consistent with the mesozooplankton distribution in these areas (Fabry, 1990;Buitenhuis et al., 2006).

Calcite and aragonite dissolution
Although total CaCO 3 dissolution in both the present model version and in PISC/CAL match literature estimates, we see an essential difference between the two model ver- 15 sions when studying the vertical distribution of the CaCO 3 dissolution in the water column. The profiles of CaCO 3 water column dissolution for the Atlantic, Pacific and Indian Ocean (Fig. 3) show how the water column dissolution of aragonite (Fig. 3b) starts at distinctly shallower depths than dissolution of calcite (Fig. 3a), which is directly related to the shallower location of the aragonite saturation horizon compared to 20 the calcite saturation horizon in the three oceans (e.g. Broecker and Takahashi, 1978;Feely and Chen, 1982;Feely et al. 2004). It results in shallow depth CaCO 3 dissolution in the Pacific Ocean, mid-depth water dissolution in the Indian Ocean and deep water dissolution in the Atlantic Ocean basin (Fig. 3c). Maxima in open water aragonite dissolution are at 400 m, 1000 m, and 3200 m depth in the Pacific, Indian and Atlantic, 25 respectively, while the basin-mean peak rates for water column calcite dissolution are simulated at 1000 m, 2300 m, and 4200 m.  Milliman and Droxler (1996) compared the dissolution in the upper 1000 m to the total CaCO 3 production and reported that between 60-80% of the total production dissolved at shallow depths. Feely et al. used the 2000 m horizon to separate the upper and lower part of the water column and compared the dissolution 5 in the upper part to the total water column dissolution. Combining the sediment trap flux of CaCO 3 at 2000 m (0.41 PgC y −1 ) and the CaCO 3 accumulation in continental shelf sediments (0.13-0.17 PgC y −1 ), they conclude that up to 60% or more of the total dissolution occurs in the water column or sediments in the upper 2000 m, with most of the CaCO 3 dissolving before it reaches the sediment-water interface. Whereas our model yields a dissolution flux in the upper 1000 m accounting for only 14% of the total CaCO 3 produced, it simulates a dissolution flux above 2000 m corresponding to 58% of the total pelagic water column CaCO 3 dissolution, comparable to the findings of Feely et al. (2004). In contrast, the equivalent percentages for the model version PISC/CAL which only considered calcite are no more than 2.5 and 38%, respectively. 15 Our modeled aragonite contribution to the shallow depth CaCO 3 dissolution represents an upper limit estimate. Thus, the discrepancy between model output and estimates (Milliman and Droxler, 1996) in the upper 1000 m supports the findings that additional mechanisms (e.g. Friis et al., 2007, Antja et al., 2008 are needed in order to explain the overall estimated loss of CaCO 3 or excess of alkalinity in the upper 20 part of the water column. However, our results maintain that the underestimation of the modeled CaCO 3 dissolution in the upper 0-2000 m in PISC/CAL may to a great extent result from an exclusion of aragonite in the PISCES model as proposed by Gehlen et al. (2007). The increased shallow depth dissolution obtained by implementing aragonite yields support to the hypothesis that high dissolution fluxes at depths 25 above 2000 m may largely be attributed to more soluble CaCO 3 phases like aragonite (Iglesias-Rodriguez et al., 2002;Kleypas et al., 2006). Furthermore, the vertical distribution of the CaCO 3 dissolution fluxes in the present model version is considerably improved compared to the PISC/CAL version.

Transient simulation
Projections of future changes of the CaCO 3 budget with increasing atmospheric CO 2 will next be presented and discussed. Figure 4 presents a) timeseries of the mean saturation state of surface waters with respect to aragonite and corresponding maps for year b) 1861, c) 2000, d) 2050, e) 2075 and f) 2100 for the SRES A2 scenario. 5 In addition to the Global Ocean, results are shown for two critical areas: 1) the high latitudes or polar and subpolar regions (here defined as the area >40 • S and >40 • N) where the largest change in saturation state occurs, and 2) the equatorial area (here restricted within 20 • S-20 • N) which include important upwelling areas with high calcification rates. Time series of a) and b) net calcite and aragonite production, c) and d) 10 calcite and aragonite dissolution and e) and f) relative calcite and aragonite dissolution are shown in Fig. 5. Model results for CaCO 3 production, export, sinking fluxes, dissolution and relative dissolution for year 1861, 2000 and 2100 are summarized in Table 2.
The mean global saturation state of surface ocean waters (0-100 m) with respect to 15 aragonite decreases from Ω a >3.5 in year 1861 to less than 2 in year 2100 (Fig. 4a).
In the equatorial area the mean saturation state decreases rapidly from Ω a >4 to Ω a =2 at the end of the scenario, whereas at high latitudes where the saturation state is already very low, it reaches values below 1 before year 2100. According to our parameterization, calcification may occur everywhere where Ω a >1. Maps of Ω a averaged 20 over the top 100 m of the water column ( Fig. 4b-f) show that in year 1861 the criteria for aragonite production is met everywhere across the global surface ocean. In year 2000 the areas of Ω a >4 are somewhat reduced, while in year 2050 such areas do no longer exist and small regions of the surface water in the Arctic Ocean are now already projected to be undersaturated with respect to aragonite for the SRES A2 scenario. 25 Following the decreasing saturation values, in year 2075 large areas of both the Arctic and the Southern Ocean are undersaturated with respect to aragonite, whereas in year 2100, compared to pre-industrial times, the conditions for calcification have altered dra-BGD 5,2008 Modeling the marine aragonite cycle R. Gangstø et al. matically. Most of the Southern and the Arctic Ocean experience at this time Ω a <1, indicating unfavorable conditions for aragonite production and matching results of Orr et al. (2005) for the Southern Ocean and of Steinacher (2007) for the Arctic. The change in saturation state with respect to calcite follows a similar pattern (see Gehlen et al., 2007), however as the saturation horizon of aragonite is located shal-5 lower in the water column than the one of calcite (e.g., Broecker and Takahashi, 1978;Feely and Chen, 1982) the surface waters get undersaturated with respect to aragonite at a much earlier stage.
The predicted reduction in saturation state is further reflected in the modeled production and dissolution rates of CaCO 3 (Fig. 5 and Table 2). Both calcite and aragonite 10 calcification rates have decreased considerably at the end of the SRES A2 scenario (Fig. 5a). While the reduction in production rates is insignificant during the historical period up to year 2000, it is large after year 2050. For calcite the total global production rate per year has decreased by 13% at the end of the scenario, from 0.57 (pre-industrial times.) to 0.49 PgC y −1 (year 2100). For aragonite the corresponding 15 reduction in global production rate is of considerably 29%, i.e. of almost 1/3 since preindustrial times, with global aragonite production decreasing from 0.31 to 0.22 PgC y −1 . The modeled decrease in total CaCO 3 production is 19%. The calcite and aragonite export flux at 100 m decreases by 13 and 41%, respectively, providing a decrease in total CaCO 3 export of 23%. 20 Gehlen et al. (2007) simulates a decrease in CaCO 3 production of 27% at the end of their scenario, where the initial production was 0.8 PgC y −1 . The export was reduced by 29%. Their final atmospheric CO 2 concentration was however 4 times the preindustrial concentration, while with the SRES A2 scenario used in the present study the atmospheric value of CO 2 in year 2100 is closer to 3 times the pre-industrial value. 25 On the other hand, the duration of their scenario was shorter and the annual rate of pCO 2 was higher. At 3 times the pre-industrial value a reduction in calcification of approximately 15% was predicted with PISC/CAL, not far from our predicted reduction in calcite production of 13% with the present model version.
In addition to the decrease in carbonate production with increasing CO 2 concentrations, the modeled amount of water column calcite and aragonite dissolution decreases (Fig. 5b). The fact that this occurs, although one might expect higher dissolution in more acidic waters, can be explained by the reduced abundance of particulate CaCO 3 due to lower CaCO 3 productivity. The relative amount of pelagic CaCO 3 dissolution, 5 however, expressed as the ratio of calcite and aragonite dissolution versus the respective rate of production (dissolution/production) in percent, is increasing. A slight increase from 67 to 71% of calcite material is dissolved in the water column, whereas for aragonite the ratio increases by 12%, from 60 to 72%, i.e. 3 times more than for calcite. This leads to a total increase in relative CaCO 3 dissolution from 64 to 71%. Due 10 to the gradually lower saturation state value in surface waters, the predicted increase in relative calcite and especially aragonite dissolution is highest in the upper part of the water column. In the upper 2000 m the relative dissolution of calcite, aragonite and total CaCO 3 increase by 10, 16 and 11%, respectively ( Table 2).
As the waters within year 2100 mainly become undersaturated with respect to cal- 15 cite and aragonite in the high latitudes, we would expect the largest effects on CaCO 3 production in this area. In fact, the calcite production in the high latitudes is reduced by as much as 24%, while the aragonite production is reduced dramatically by 65% (from 0.03 to 0.009 PgC y −1 ). The relative calcite dissolution has increased by 4%, whereas the relative aragonite dissolution has experienced a drastic increase of 72%, from 41 to 20 113%. The latter value implies that in the end of the SRES A2 scenario, more aragonite is actually dissolved than produced in this area, indicating transport of some aragonite particles from lower to higher latitudes. The changes in aragonite production and dissolution suggest the potential for detrimental impacts of acidification on pteropods in the high latitudes, where shells of pteropods dominate the flux of CaCO 3 (Collier et 25 al., 2000;Honjo et al., 2000;Urban-Rich;Accornero et al., 2003;Tsurumi et al., 2005;Manno et al., 2007).
In the equatorial area, the calcite and aragonite production is reduced by 9.5 and 18%, whereas the relative dissolution has increased by 4% and 5%, respectively. This change is small compared to the percentages presented for the high latitudes, however the quantitative impacts on calcite and aragonite producers may be large due to the moderate to high calcification rates observed in equatorial regions (e.g. Deuser and Ross, 1989;Fabry and Deuser, 1991;Kalberer et al., 1993;Fischer et al., 1996). In year 2100 the model still predicts a large area with aragonite saturation values above 5 2 in the gyres (Fig. 4). However, near the coast and in the equatorial upwelling area the saturation state is already close to 1 in year 2100. In areas with upwelling of DIC enriched deep-water the saturation state may decrease rapidly with increasing CO 2 concentrations. We would thus expect that shortly after year 2100 the surface waters in the equatorial and especially upwelling areas would become undersaturated, which 10 would probably prevent large amounts of aragonite from being produced in this area.

Conclusions
The work presented here is intended as a first step towards understanding the role of aragonite production and dissolution in the global CaCO 3 budget and its evolution with increasing atmospheric pCO 2 . Aragonite was included in the global biogeochem- 15 ical model PISCES, where the parameterized production and dissolution were implemented with analogy to the corresponding terms for calcite. The production was assigned to mesozooplankton as a function of mesozooplankton biomass and aragonite saturation state of ambient waters. The marine carbonate cycle is well reproduced by the model with net global carbonate production rates corresponding to recent data-20 based estimates. Except from an overestimation by the model of produced aragonite in the equatorial zones, the model predicts a global spatial variability in aragonite production and dissolution consistent with literature. With aragonite included in the model, the total CaCO 3 water column dissolution above 2000 m constitutes 58% of the modeled dissolution corresponding to estimates.
In contrast, the corresponding value for the calcite-only version is 38%. The model provides less CaCO 3 dissolution in the upper 1000 m than reported, indicating that other mechanisms in addition to aragonite dissolution are needed to explain the entire loss of CaCO 3 in the upper part of the water column. Compared to the calcite-only model version however, taking into account production and dissolution of aragonite in addition to calcite in PISCES significantly improves the vertical distribution of dissolution rates and fluxes by allowing water column CaCO 3 dissolution above the calcite saturation 5 horizon. Our results support the hypothesis that high dissolution fluxes in the upper part of the water column may largely be attributed to aragonite.
Effects of increasing atmospheric CO 2 on the CaCO 3 production and dissolution were studied using the SRES A2 scenario. Following the increasing values of atmospheric CO 2 , year 2050 appears to be a critical threshold moment; at this time already 10 the entire ocean surface is predicted to have suboptimal conditions for aragonite production and some surface water areas in the Arctic are undersaturated with respect to aragonite. As a result of the decreasing saturation state, the global modeled aragonite production is reduced by almost one third in year 2100 compared to pre-industrial times. The global relative aragonite dissolution (dissolution/production) has increased 15 by 12%. At latitudes above 40 • , where production rates of aragonite are assumed to be highest, a decrease in aragonite production of substantially 65% is projected. Here the relative aragonite dissolution increases by 72%. Total global modeled CaCO 3 production has decreased by 19% by the end of the SRES A2 scenario, compared to approximately 15% for the model version without aragonite. Most of the increase in rel-20 ative dissolution takes place in the upper part of the water column. While the feedback to atmospheric CO 2 of reduced CaCO 3 production and increased relative dissolution is assumed to be small (Gehlen et al., 2007), the predicted changes in aragonite production and dissolution highlight potential impacts of changing carbonate chemistry on aragonite producing organisms and the surrounding ecosystem of the pelagic ocean, 25 especially in subpolar and polar areas.
The inclusion of 1/3 aragonite is in the upper range of most estimates, leading to a probable overestimation of the quantitative changes in aragonite production and dissolution with increasing atmospheric pCO 2 . Alternatively, if the decrease in production of pteropod aragonite follows a linear parameterization as found by  for corals instead of the curve fitted to experiments with coccolithophores, a faster response to the increasing CO 2 is expected. This would point towards an underestimation of the quantitative reduction in aragonite production suggested by our model. Thus further exploration is needed to quantify the contribution of aragonite to the total 5 CaCO 3 budget and the evolution of aragonite production and dissolution in the future ocean.  10.1029/2002GB002001, 2003 The vertical flux of biogenic and lithogenic material in the Ross Sea: moored sediment trap observations , Deep-Sea Research II, 47, 3491-3520, 2000. Delille, B., Harlay, J., Zondervan, I., Jacquet, S., Chou, L., Wollast, R., Bellerby, 5 R. G. J., Frankignoulle, M., Borges, A. V., Riebesell, U., and Gattuso, J.-P.: Response of primary production and calcification to changes of pCO 2 during experimental blooms of the coccolithophorid Emiliania huxleyi, Glob. Biogeochem. Cy., 19, GB2023, doi:10.1029/2004GB002318, 2005. Deuser, W. G. and Ross, E. H.: Seasonally abundant planktonic foraminifera of the Sargasso BGD Table 1. The carbonate cycle in PISCES: Comparison between model output and data in Pg CaCO 3 -C y −1 . PISC/CAL=results from an equivalent model study without aragonite (Gehlen et al., 2007). (14%) (2.5%) (60-80%) (Milliman and Droxler, 1996)