Recent independent lines of evidence suggest that the dissolution of calcium carbonate (CaCO3) particles is substantial in the upper ocean above the calcite 100% saturation horizon. This shallow-water dissolution of carbonate particles is in contrast with the current paradigm of the conservative nature of pelagic CaCO3 at shallow water depths. Here we use more than 20,000 sets of carbon measurements in conjunction with CFC and 14C data from the WOCE/JGOFS/OACES global CO2 survey to estimate in situ dissolution rates of CaCO3 in the Atlantic Ocean. A dissolution rate is estimated from changes in alkalinity as a parcel of water ages along an isopycnal surface. The in situ CaCO3 dissolution increases rapidly at the aragonite 100% saturation horizon. Estimated dissolution rates north of 40oN are generally higher than the rates to the south, which is partly attributable to the production of exported CaCO3 being higher in the North Atlantic than in the South Atlantic. As more CaCO3 particles move down the water column, more particles are available for in situ dissolution. The total water column CaCO3 dissolution rate in the Atlantic Ocean is determined on an annual basis by integrating estimated dissolution rates throughout the entire water column and correcting for alkalinity input of approximately 5.6 × 1012 mol C yr−1 from CaCO3-rich sediments. The resulting water column dissolution rate of CaCO3 for the Atlantic Ocean is approximately 11.1 × 1012 mol C yr−1. This corresponds to about 31% of a recent estimate (35.8 × 1012 mol C yr−1) of net CaCO3 production by Lee  for the same area. Our calculation using a large amount of high-quality water column alkalinity data provides the first basin-scale estimate of the CaCO3 budget for the Atlantic Ocean.
 The marine carbonate system affects the long-term fate of anthropogenic CO2 in the oceans and the rate of atmospheric CO2 increase by controlling the rate of oceanic CO2 uptake. Therefore many studies have focused on the fundamental processes controlling the distribution of the four inorganic CO2-system parameters in the oceans: fugacity of CO2 (fCO2), total dissolved inorganic carbon (CT), total alkalinity (TA), and pH (−log10[H+]). Knowledge of the rate of CO2 removal from the surface ocean via biogenic CaCO3 and of the ultimate delivery rate of CaCO3 to regions of the deep ocean that are corrosive to CaCO3 particles is incomplete. Since the settling time of CaCO3 particles is thought to be short compared to dissolution rates, it was generally believed that much of the carbonate dissolution takes place on or just beneath the surface of the sediments. However, evidence from a variety of sources suggests that as much as 60−80% of net CaCO3 production is dissolved in depths that are above the chemical lysocline, the depth below which the rate of CaCO3 dissolution distinctly increases [Milliman et al., 1999; François et al., 2002].
 It is well known that >80% of sinking organic matter is oxidized within the upper 1000−2000 m of the water column [Lutz et al., 2002], but much less is known about the amount of CaCO3 that dissolves in the water column and about the underlying processes. The biological production and dissolution of CaCO3 in the ocean result in changes in TA in the water column according to the following reaction:
A change in the balance of this reaction in the ocean would have a significant impact on atmospheric CO2 concentration [Zondervan et al., 2001]. Dissolution of CaCO3 particles increases TA in seawater and thus the capacity of the ocean to absorb CO2 from the atmosphere, whereas the production of CaCO3 leads to the opposite consequence.
 Two papers published in 1984 provide substantial evidence that the dissolution of aragonite occurs in the upper waters (where the depth is <1000 m) of the Pacific Ocean [Betzer et al., 1984; Byrne et al., 1984]. Milliman et al.  also presented several independent lines of evidence supporting the dissolution of CaCO3 particles in the upper 1000 m of the water column. If this upper-water dissolution of CaCO3 is significant, we may need to modify the long-held paradigm of the conservative nature of pelagic CaCO3 at upper water depths. In this paper, we use data from the World Ocean Circulation Experiment/Joint Global Ocean Flux Study/Ocean Atmosphere Carbon Exchange Study (WOCE/JGOFS/OACES) CO2 survey of the Atlantic Ocean to estimate in situ dissolution rates of CaCO3 in the water column by examining changes in TA along isopycnal surfaces. We also provide a CaCO3 budget for the Atlantic Ocean using the net CaCO3 production from Lee , the water column dissolution of CaCO3, and the dissolution of CaCO3 from sediments. This paper, in conjunction with the analyses of Feely et al.  for the Pacific Ocean and Sabine et al. [2002a] for the Indian Ocean, provides a first look at the global CaCO3 budget based on water column TA distribution.
2. Source of Data and Calculation Methods
2.1. WOCE/JGOFS/OACES Data
 Most of the carbon measurements used in this study were collected as part of the WOCE/JGOFS/OACES Atlantic survey between 1990 and 1998 (Figure 1). A total of 23 U.S. and European cruises were included in this analysis. TA values calculated from CT−pH or CT−fCO2 using thermodynamic constants were also included in this dataset when measured TA values were not available. The carbonic acid dissociation constants of Mehrbach et al.  as refitted by Dickson and Millero  were used in this calculation. Data from different sources were compiled, and the consistency of inorganic carbon measurements (fCO2, CT, TA, and pH) made on different cruises was quality checked following the procedures used by Lamb et al. . The consistency checks for the Atlantic data focused on CT and TA, because these parameters are used in the calculation of anthropogenic CO2 concentrations and in large-scale biogeochemical carbon cycling studies.
 Four independent methods were used to determine if there were systematic offsets between the various cruises: (1) Inorganic carbon-system values in deep water were compared where cruise tracks crossed, which are referred to as “crossover analyses”; (2) multiparameter linear regressions of CT (or TA) with potential temperature, salinity, oxygen, silicate, and nitrate were created for cruises that followed the same cruise track, and the calculated values were then compared with the measured parameters for individual cruises; (3) on cruises where more than two carbon-system parameters were measured, the internal consistency between parameters was determined from known thermodynamic relationships between the parameters; and (4) regional multiparameter linear regressions of CT (or TA) with potential temperature, salinity, oxygen, silicate, and phosphate were created using data that were deemed accurate based on the previous checks. The CT and TA values calculated from the regressions were then compared with individual cruises that showed significant offsets in the crossover analysis. These analyses suggest that the systematic cruise-to-cruise differences for CT and TA are small except for two cruises [Wanninkhof et al., 2003]. Adjustments of +14 and −7 μmol kg−1 were recommended for TA values on A01W and A09, respectively, whereas no specific adjustments for CT were required. After adjusting the A01W and A09 TA data, the entire data set is believed to be internally consistent to ±4 and ±6 μmol kg−1 for CT and TA, respectively. The final data set comprises 28,639 measurements of CT and 18,771 measurements of TA, and is available at http://cdiac.ornl.gov/oceans/datameta.html.
 Various tests were also carried out to evaluate the quality of the chlorofluorocarbon (CFC) data, including comparing air measurements with global atmospheric CFC trends [Walker et al., 2000] and with surface CFC concentration measurements, and examining the consistency of CFC concentration and CFC-11/CFC-12 ratio measurements along vertical profiles. CFC concentration and ratio measurements were also compared along sections and at crossover stations. A data-quality flag was assigned to each CFC measurement based on WOCE guidelines, and the CFC data are reported as picomole per kilogram seawater on the SIO98 calibration scale. These data are available at http://cdiac.ornl.gov/oceans/datameta.html.
2.2. Seawater TA
 The TA of seawater is defined as the number of moles of hydrogen ions equivalent to the excess of proton acceptors (bases formed from weak acids with pK ≥ 4.5 at 25oC and zero ionic strength) over proton donors (acids with pK < 4.5) in 1 kg of sample [Dickson, 1981],
where brackets represent concentrations in seawater (μmol kg−1) and [H+]F is the concentration of free hydrogen ions. The concentrations of NH3 and HS− are not included in equation (2) because they are generally very low in open-ocean water.
 The dominant processes that modify surface-seawater TA are evaporation and fresh-water inputs, which manifest themselves as variations in salinity. In deep waters, where salinity variations are considerably smaller, the formation or dissolution of biogenic CaCO3 becomes important in determining the variations in TA. The net addition of carbonate ion (CO32−) increases seawater TA, whereas the release of protons during the remineralization of organic matter decreases seawater TA (this latter process is less significant). Thus the latter two processes result in a compensatory effect on the change in TA, and so the change in measured TA in a particular water parcel provides a lower limit on the amount of CaCO3 that dissolves. However, in our work the release of protons by organic matter remineralization is corrected for in the calculation by assuming a constant stoichiometric ratio between dissolution and organic remineralization (see equation (6) in section 2.4).
2.3. Saturation State of Seawater With Respect to Calcium Carbonate
 Most of the upper water is supersaturated with aragonite and calcite, while much of the deep water is undersaturated. The degree of saturation of seawater with aragonite or calcite is defined as the ratio of the ion product of the concentrations of calcium ([Ca2+]) and [CO32−] at the in situ temperature, salinity, and pressure, divided by the stoichiometric solubility product with respect to aragonite (K*sp-arg) and calcite (K*sp-cal) at the same conditions,
 The values of K*sp-arg and K*sp-cal at 1 atmosphere as a function of temperature and salinity were determined by Mucci . The effect of pressure on the solubility of aragonite or calcite was estimated from the measurements of Ingle . When Ω > 1, seawater is supersaturated with respect to aragonite (or calcite); conversely, when Ω < 1, seawater is undersaturated. Since the calcium-to-salinity ratio in seawater does not vary by more than ∼1.5%, variations in the ratio of [CO32−] to the stoichiometric solubility product primarily govern the degree of saturation of seawater with respect to aragonite or calcite. In situ [CO32−] in seawater was calculated from a pair of measured inorganic carbon parameters (e.g., TA + CT, or pCO2 + CT) using the set of pressure-corrected thermodynamic constants, which was previously shown to be the most consistent with a calibrated field dataset compiled from a global carbon survey [Lee et al., 2000]. The constants we used include the carbonic acid dissociation constants of Mehrbach et al.  as refitted by Dickson and Millero , along with equilibrium constants of other ancillary components (e.g., boric acid dissociation, solubility of CO2, water hydrolysis, and phosphoric- and silicic-acid dissociation) necessary to characterize the carbonate system in seawater as summarized by Millero . The pressure effect on these thermodynamic constants was estimated from molal volume and compressibility data [Millero, 1983, 1995],
where R is the gas constant, P is the applied pressure, and ΔVi and ΔKi are changes in molal volume and compressibility for dissociation constants of various acids. All calculations were made using the QuickBasic CO2 program developed by K. Lee, which was based on the earlier version by F.J. Millero (University of Miami). Results calculated using this program were the same as those calculated using a CO2 program developed by Lewis and Wallace , which is available at http://cdiac.esd.ornl.gov/oceans/home.html. The concentrations of calcium, borate, sulfate, and fluoride were estimated using the equations of Riley and Tongudai , Uppström , Morris and Riley , and Riley , respectively.
 We also calculated aragonite and calcite saturation states of seawaters in the preindustrial era using the same program and equations, and measured TA and estimates of preindustrial levels of CT (CT°) for all water samples taken during the WOCE/JGOFS/OACES Atlantic survey. We assumed in this calculation that TA has not changed due to oceanic uptake of anthropogenic CO2 during the industrial era. The CTo for each sample was obtained by subtracting out the anthropogenic CO2 concentration (K. Lee et al., An updated anthropogenic CO2 inventory in the Atlantic Ocean, submitted to Global Biogeochemical Cycles, 2003) (hereinafter referred to as Lee et al., submitted manuscript, 2003) using the modified version of the ΔC* approach developed by Gruber et al. . The extension of the method includes the more accurate treatment of nonlinear mixing between the air-sea disequilibria of end-member water types on an isopycnal surface using an optimum multiparameter analysis (for details see Sabine et al. [2002b] and Lee et al. (submitted manuscript, 2003)). A preformed TA estimate different from that of Gruber et al.  is also used. Our estimation is based on high-quality Atlantic surface (<100 dbar) TA data.
 The probable errors in the Ωarg = 1 and Ωcal = 1 saturation horizons due to uncertainties in equilibrium constants and in measured TA and CT values are given in Table 1. Uncertainties for measured thermodynamic constants were taken from the original works [Ingle, 1975; Mucci, 1983; Millero, 1995], and those for the measured parameters were from Wanninkhof et al. . Estimated probable errors are ±0.067 in Ωarg for water at 2500 m and ±0.11 in Ωcal for water at 4500 m, which equate to uncertainties of ±300 m in the Ωarg = 1 saturation depth and of ±470 m in the Ωcal = 1 saturation depth. The probable error is the square root of the sum of the squared errors due to uncertainties in thermodynamic constants and measured parameters.
Table 1. Estimated Errors in the Calculated Values of Ωarg and Ωcal Caused by Uncertainties in Thermodynamics Constants and Measured Parametersa
Input Parameters and Uncertainties
Estimated Error of Ωarg (2500 m)
Probable error for Ωarg: ±0.067; probable error for Ωcal: ±0.11; Ki*: dissociation constants of carbonic and boric acids.
2.4. Determination of the In Situ Dissolution Rate of CaCO3
 The amount of CaCO3 dissolved in a subsurface water parcel is estimated from changes in TA by subtracting out the preformed TA concentration and correcting for the TA decrease resulting from the release of protons during the remineralization of organic matter. The contribution from organic matter is determined by using apparent oxygen utilization (AOU = O2 (saturated values at a given temperature and salinity) − O2 (measured)) as an indicator. These corrections introduce a new tracer ΔTACaCO3 that is used in this paper to quantify the amount of CaCO3 dissolved in the water mass in question,
where TAMEAS is the measured TA, and TAo is the preformed TA. The second term on the right-hand side accounts for the decrease in TA resulting from the oxidation of organic matter by using AOU, and the N/O2 ratio(= 0.0941) [Anderson and Sarmiento, 1994] is used rather than nitrate concentration to avoid having to directly estimate the preformed nitrate values [Chen, 1978]. A coefficient of 0.63 proposed by Kanamori and Ikegami  was also used to account for TA contributions from the oxidation of organic nitrogen, phosphorus, and sulfur. This coefficient was derived from the assumption that reduced sulfur in organic matter is completely oxidized to sulfate and that the contribution of resulting sulfate to TA would be about 20% of the total contribution from nitrate and phosphorus. Therefore the coefficient of 0.63 is about 20% higher than a commonly used value of 0.53, which accounts only for the change in TA caused by the oxidation of organic nitrogen and phosphorus. Equation (6) differs slightly from those in the companion papers of Feely et al.  and Sabine et al. [2002a], in which NTA (NTA = TA × 35/S, where S is the salinity) was used instead of TA. Also, here the term ΔTACaCO3 is used, which is comparable to their TA* term. The results using either TA or NTA are nearly identical for a salinity of 33−37.
 The TA° of a water parcel in the interior of the ocean is an estimate of TA that the water had when it was last at the surface. We estimated TA° from a multilinear regression model using conservative tracers such as S and NO as independent variables. NO is defined as NO = O2 − RO2/N × N [Broecker, 1974]; we used RO2/N = −10.625 [Anderson and Sarmiento, 1994]. The TA data (<100 dbar) from the WOCE/JGOFS/OACES Atlantic CO2 survey gave the following relationship:
where S is on the practical salinity scale, and NO is in μmol kg−1. The standard error (1σ) of the estimated TA° is ±10.3 μmol kg−1 based on 2345 data points. Only half of this error is reflected in estimating ΔTACaCO3, because half of the TA change equates to the CaCO3 change. We used NO rather than PO (PO = O2 − RO2/P × P) because significant data gaps exist in the Atlantic phosphorus data set. The loss of nitrate due to water column denitrification is insignificant in the Atlantic Ocean [Gruber and Sarmiento, 1997] and should not affect our calculations. The probable error of the estimated ΔTACaCO3 due to uncertainties in measured and preformed TA and in the N/O2 ratio is ±8 μmol kg−1 based on an uncertainty of ±6 μmol kg−1 in measured TA [Wanninkhof et al., 2003], of ±10.3 μmol kg−1 in preformed TA [see equation (7)], and of ±0.0218 in the N/O2 ratio [Anderson and Sarmiento, 1994].
 If CaCO3 dissolution occurs in a given water parcel, values of ΔTACaCO3 increase as the water parcel ages. If the effects of mixing between different water masses are taken into account accurately, the slope between values of ΔTACaCO3 and corresponding ages of water parcels can be used as the in situ dissolution rate of CaCO3 particles (this assumes that changes in ΔTACaCO3 along isopycnal surfaces are solely due to water column dissolution of CaCO3 particles). An alternative method for estimating the dissolution rate is to divide each value of ΔTACaCO3 by its age. However, the rate determined by this method might include a large error if there is systematic age biasing as discussed in section 2.5. Therefore, in this paper we estimated the in situ CaCO3 dissolution on isopycnal surfaces by determining the slope between values of ΔTACaCO3 and ages of water parcels. Systematic age biasing may not significantly affect our slope-based results as long as its magnitude is constant over the period of analysis.
 To estimate in situ CaCO3 dissolution rates in waters where values of ΔTACaCO3 are positive, we plotted them against water parcel ages derived from CFC-11 for the upper ocean and 14C for deep waters where CFC-11 is not detected (Figure 2). The CFC-11 age was calculated by converting the CFC-11 concentration (in pmol kg−1) in the subsurface water to partial pressure (pCFC-11) at the potential temperature and salinity [Doney and Bullister, 1992] and then matching pCFC of the subsurface water with the pCFC of the atmosphere for the appropriate year. In this calculation, the subsurface water parcel is assumed to have been in solubility equilibrium when it was in contact with the overlying atmospheric pCFC-11. Thus the age of the subsurface water is defined as the time difference between the measurement date and the date when the water parcel was last in contact with the atmosphere. The use of CFC-11 age is limited to upper waters with CFC-11 ages less than 35 years (corresponding to ∼0.1 pmol kg−1) because systematic biases in CFC ages due to dilution and nonlinear mixing effects tend to be larger for older waters [Warner et al., 1996; Doney et al., 1997; Sonnerup, 2001]. Some of the potential uncertainties associated with the use of the tracer to date water masses are discussed further in section 2.5.
 For waters with CFC-based ages greater than 35 years, we used age estimates from natural Δ14C. For waters that contained bomb-generated 14C, this was subtracted from the total radiocarbon (Δ14C) to derive the natural Δ14C component. Rubin and Key  proposed a separation method based on the strong correlation between natural Δ14C and potential alkalinity (PTA = (TA + nitrate) × 35/S). We used their PTA−Δ14C algorithm along with PTA data calculated from the WOCE/JGOFS/OACES dataset to estimate naturally occurring Δ14C. Resulting estimates of natural Δ14C were then used to calculate the age of water parcels. The natural Δ14C is a good method for determining the age of intermediate and deep waters because this isotope has a half-life of 5730 ± 40 years [Godwin, 1962]. The use of natural Δ14C is confined to waters deeper than 1500 m in most of the Atlantic Ocean, except for the northern North Atlantic where the CFC penetrates to waters deeper than 1500 m due to deep convective mixing.
2.5. Uncertainty in Estimated Dissolution Rates
 Estimated dissolution rates of CaCO3 are subject to change due to two important sources of errors: (1) the uncertainty in estimating ΔTACaCO3: we do not estimate dissolution rates if waters have ΔTACaCO3 less than 8 μmol kg−1, which is close to the probable error in estimating ΔTACaCO3; and (2) the uncertainty in the pCFC-based age: the tracer age is not necessarily identical to the true or ideal ventilation age. Because of the nonlinear atmospheric-CFC history, mixing of waters with different pCFCs can introduce significant bias in the resulting pCFC ages, compared to the true age of the water. Age biasing due to mixing has been examined using observational data and simple models [e.g., Doney et al., 1997; Sonnerup, 2001]. Because of the quasi-linear increase in atmospheric CFC-11 during 1965–1990, mixing between waters ventilated during this period should produce relatively small age biases. Age biasing can be significantly greater for mixing between younger and older waters, which tends to bias the age of the mixture toward the younger (higher-CFC-bearing) component. In this study we limit the use of pCFC ages for the calculation of CaCO3 dissolution rates to waters with CFC ages <35 years. The rates we obtained might also be influenced by bidirectional mixing of the assumed steady state signal of excess TA along isopycnal surfaces, while the CFC ages are determined assuming a unidirectional penetration of CFC along the same isopycnals. However, the latter effect is probably small, as it is largely accounted for by the increasing preformed TA values when moving toward the poles.
 Initial undersaturation of CFCs in an outcrop region will make CFC-based ages older than the true age [Wallace et al., 1994; Doney et al., 1997]. Because CFCs were increasing relatively rapidly in the atmosphere during the period of the study and because the dissolution rates are integrated along the path of the isopycnal over several years, a small degree of CFC disequilibrium in high-latitude waters should not significantly affect our results.
3.1. Distribution of Alkalinity
 The distribution of TA in the surface mixed layer of the Atlantic Ocean is mainly controlled by the factors that govern salinity [Broecker and Peng, 1982; Millero et al., 1998]: A variation in salinity of 1 results in a change of approximately 56 μmol kg−1 in TA (see equation (7)). Other nonconservative processes, such as precipitation and dissolution of biogenic CaCO3, also contribute to the variability of TA, albeit to a much lesser extent [Brewer et al., 1975; Brewer and Goldman, 1976; Millero et al., 1998]. The highest concentrations of TA are observed in the surface mixed layer waters at 30°N and 20°S where salinity maxima as high as S = 37.2 are found. From here, the salinity and TA decrease to S = 35 and TA = 2300 μmol kg−1 in high-latitude waters. This is in contrast with the Pacific Ocean where the highest TA values are generally found in deep waters [Feely et al., 2002]. The conservative behavior of surface TA is particularly true in low-latitude regions of the Atlantic Ocean (between 40oN and 40oS). Much of the spatial and seasonal TA variability in these regions can be removed by normalizing the result to a constant salinity (S = 35) [Lee et al., 1997; Millero et al., 1998]. Consequently, the NTA of surface waters remains nearly constant (NTA ∼ 2290 μmol kg−1) from 40°S to 40°N, and increases with latitude, largely due to convective mixing of deep waters that have accumulated excess TA from CaCO3 dissolution. Detailed analysis of the surface TA distribution in the Atlantic Ocean and in the other major basins are given by Millero et al. .
Figures 3 and 4 show the meridional distributions of salinity, TA, and NTA in the eastern and western Atlantic. The main features of the deep-water characteristics and circulation of the Atlantic Ocean are shown in the general structures of the TA, NTA, and salinity sections. The Antarctic Intermediate Water (AAIW) originates south of the Polar Frontal Zone and is evident as a low-salinity tongue extending to 20°N centered at about 800 m depth. The TA section also shows local minima in this region. Between the AAIW and the abyssal Antarctic Bottom Water (AABW) is the North Atlantic Deep Water (NADW), which originates in the far North Atlantic and is most evident in the salinity section. The TA section also shows a local minimum in this region, but this minimum extends southward to only 40°N due to TA increasing from CaCO3 dissolution as the NADW moves to the south. The third major feature found in the intermediate water is the effect of the Mediterranean Water (MW) as a salinity maximum above the NADW. The MW maximum is clear in the eastern North Atlantic but less conspicuous in the west and in the south as it gradually loses its unique characteristics by mixing with the waters above and below it. The MW is shown as a TA maximum centered at about 1000 m depth and 35°N. Below the TA minimum layer centered at about 800 m depth in the South Atlantic, TA values gradually increase with depth. The same is true for waters deeper than 2000 m in the North Atlantic. The overall similarity between the TA and salinity sections suggests that deep-water circulation plays a critical role in the TA distribution in the deep Atlantic.
 The main features of the deep-water circulation of the Atlantic Ocean are not as clear in the TA sections as in the NTA and salinity sections. The NTA concentration is generally lower in waters shallower than 500 m. The lower NTA concentration in shallow warm waters is generated by a net reduction in TA resulting from the biological production of CaCO3, while the higher NTA values for deep waters result from the net effect of in situ dissolution of CaCO3 and oxidation of organic matter. Part of the NTA increase due to dissolution of CaCO3 is offset by the release of protons from the oxidation of organic matter. This effect is generally <20% of the total NTA change in the upper Atlantic, but could be up to 40% in the deep waters of the South Atlantic. The NTA values for the NADW are lower than those for the AABW and AAIW; older water masses have more time to accumulate TA from the dissolution of CaCO3.
3.2. Degree of Saturation of CaCO3
 Previous studies have shown that the dissolution rates of CaCO3 in the interior of the ocean are nonlinearly influenced by the degree of undersaturation [Morse and Berner, 1972; Keir, 1980]. It is, therefore, important to have an accurate knowledge of the saturation state of seawater with respect to aragonite and calcite. Figure 5 shows the meridional distribution of [CO32−] and the 100% saturation horizons for aragonite and calcite in the eastern and western Atlantic. The upper waters shallower than 1000 m in the South Atlantic and 2500 m in the North Atlantic are supersaturated by as much as 400% with respect to aragonite; the waters below these depths are undersaturated. The depth variations are primarily due to a difference in [CO32−] [Morse and Berner, 1972; Pytkowicz, 1973; Broecker and Takahashi, 1978]. An interesting feature in the eastern basin is the undersaturated water centered at 800 m between 20°S and 10°N that is sandwiched by supersaturated waters above and below. This undersaturated water is the northern extension of AAIW and may be attributed to the [CO32−] decrease caused by reaction with protons resulting from the oxidation of organic matter during the journey of AAIW from the surface waters of the Southern Ocean to about 1000 m in the tropical Atlantic. However, the mechanism responsible for the formation of this undersaturated water has not yet been identified.
Figure 6 shows the 100% saturation depths for aragonite (Figure 6a) and calcite (Figure 6b) in the Atlantic Ocean. Both horizons are deepest in the North Atlantic and generally become shallower toward the south. The saturation depths are set by the in situ [CO32−] of waters and the pressure dependence of K*sp-arg and K*sp-cal. The degree of saturation decreases with depth because the solubility of these minerals generally increases with depth, which is attributable to several factors. First, the effect of pressure on the dissociation constants of carbonic and boric acids results in the pH decreasing and consequently in [CO32−] also decreasing. Second, remineralization of organic matter falling from the surface releases CO2 into the water, decreasing [CO32−] and pH, and increasing the solubility of CaCO3. The first effect is generally more important in deep waters, whereas the second effect is in shallow waters. Third, CaCO3 becomes slightly more soluble as temperature drops with depth. However, the temperature effect is small. These combined effects cause the solubility of CaCO3 to increase significantly with depth.
 The 100% saturation depth for aragonite is about 2500 m in areas between 20°S and 60°N, and decreases to about 1000 m in areas south of 30oS. The saturation depths for aragonite are generally deeper in the western basin than in the eastern basin (Figure 6) because the older waters in the eastern basin have lower [CO32−] due to accumulation of protons from the oxidation of organic matter. The 100% saturation depth for calcite is significantly deeper than that for aragonite. It is slightly deeper than 4000 m in the North Atlantic and decreases to 3000−4000 m in the South Atlantic. About 58% of the water in the Atlantic Ocean is supersaturated with respect to aragonite, while 80−90% of the water is supersaturated with respect to calcite. Overall, the waters in the Atlantic Ocean are much less corrosive to CaCO3 particles than those in the other major basins.
3.3. In Situ Dissolution Rates of CaCO3
 The dissolution rates presented in this paper were separately determined for three different regions (70°N−40°N, 40°N−40°S, and 40°S−70°S). For each region the rates were also separately estimated for the upper waters (<1500 m) and intermediate and deep waters (>1500 m).
 The meridional distribution of ΔTACaCO3 in the Atlantic Ocean is shown in Figure 7. The major circulation features of the Atlantic Ocean are evident in the structures of the ΔTACaCO3 section. However, the general increase in the ΔTACaCO3 concentration with depth suggests that the alkalinity signal in deep waters is produced mainly by in situ dissolution of CaCO3 particles. The ΔTACaCO3 concentration ranges from <10 μmol kg−1 in the upper ocean to 20−30 μmol kg−1 in deep waters. The discernable ΔTACaCO3 increase (>10 μmol kg−1) begins at about 800 m in the South Atlantic but at about 2500 m in the North Atlantic (north of 40°N). The depth at which the significant ΔTACaCO3 increase occurs is approximately consistent with the aragonite 100% saturation horizon. Below this depth, the values of ΔTACaCO3 increase rapidly over several hundred meters and then increase gradually from there to the bottom of the ocean. Much of the ΔTACaCO3 increase occurs in waters between the aragonite and calcite 100% saturation horizons, suggesting that the degree of saturation of seawaters is a primary factor controlling the extent of in situ dissolution of CaCO3.
 Dissolution rates for waters shallower than 1500 m in Figure 8a were estimated using CFC-11 apparent ages. Rates for the intermediate North Atlantic waters between σ4 = 45.4 and σ4 = 45.8 in Figure 8b were also estimated using CFC-11 ages, while rates for the deep North Atlantic (σ4 > 45.8) and South Atlantic waters were obtained using Δ14C-derived ages. This analysis helps us identify where and to what extent CaCO3 dissolution occurs in the water column.
 Rates in the South Atlantic (south of 40°S) are nearly zero in waters less than σθ = 27.4 and increase to a maximum rate of ∼0.28 μmol kg yr−1 at σθ = 27.6. Rates in the North Atlantic (north of 40°N) are not statistically significant at σθ < 27.8 because values of ΔTACaCO3 are typically less than 10 μmol kg−1, which is within the uncertainty in estimating ΔTACaCO3. Maximum dissolution rates in the North Atlantic occur in waters near the mean aragonite saturation horizon of approximately 2500 m (Figure 8b). The reduction in depth where the maximum rate occurs is consistent with the corresponding change in the aragonite 100% saturation horizon. Below σθ = 27.8 in the South Atlantic the dissolution rates sharply decrease to <0.1 μmol kg yr−1, whereas the rates in the North Atlantic increase (Figure 8b). The rates at greatest densities of the upper waters (σθ ∼ 27.8) in the South Atlantic are close to the deep-water values that are estimated using the Δ14C method, suggesting no bias between the two methods of determining dissolution rates. The dissolution in low latitudes (between 40°S and 40°N) occurs in waters with densities greater than σθ = 27.4, but observed rates are lower than 0.05 μmol kg yr−1 and significantly lower than the rates at higher latitudes.
 Dissolution rates in the intermediate and deep waters of the South Atlantic (>1500 dbar) are in the range of 0.030−0.061 μmol kg−1 yr−1 (Figure 8b). These rates are up to an order of magnitude lower than those found in shallower waters. In contrast, the rates in the North Atlantic for densities σ4 (referenced to 4000 dbar) between 45.4 and 45.8 are in the range of 0.24−0.54 μmol kg−1 yr−1 and then sharply decrease to <0.1 μmol kg yr−1 at σ4 = 45.83. The dissolution rates south of 40°S are lower than the values in the rest of the Atlantic Ocean at densities σ4 between 45.6 and 45.9, but then increase to a maximum of 0.061 at σ4 = 46.05.
4.1. Upward Migration of the Aragonite 100% Saturation Horizon
 The protons ([H+]) formed by dissolution of anthropogenic CO2 in seawater lower the pH so that some of them combine with [CO32−] to form [HCO3−]. Thus addition of anthropogenic CO2 into the ocean decreases [CO32−], which in turn lowers saturation states of seawater with respect to aragonite or calcite. Figure 9 shows profiles of anthropogenic CO2 concentration in 20° latitude belts between 70°S and 70°N (Lee et al., submitted manuscript, 2003). The surface concentrations of anthropogenic CO2 (typically 40−60 μmol kg−1) are highest in the subtropical waters and decrease toward higher latitudes (20−40 μmol kg−1). Anthropogenic CO2 generally penetrates to shallower depths in the tropical and subtropical regions, and to deeper water as latitude increases. However, the symmetrical feature of vertical penetration of anthropogenic CO2 between the two hemispheres breaks down toward the poles: Anthropogenic CO2 penetrates all the way down to the bottom in the northern high-latitude regions, whereas in sharp contrast to this the shallowest penetrations are observed in the high-latitude Southern Ocean. The distribution of anthropogenic CO2 in the Atlantic Ocean shows that a significant part of the Atlantic Ocean has been affected by the vertical penetration of anthropogenic CO2.
 In regions between 30°N and 30°S in the Atlantic Ocean the aragonite 100% saturation horizon for the preindustrial era is nearly the same as that for the present day (Figure 10, solid and dashed lines, respectively). By contrast, the aragonite 100% saturation horizon has migrated upward by approximately 100−150 m in the South Atlantic and in the western North Atlantic. The penetration of anthropogenic CO2 in the eastern North Atlantic is not sufficiently deep to affect the saturation horizon. The calcite 100% saturation depths in the Atlantic Ocean are typically deeper than 4000 m and are not affected by the penetration of anthropogenic CO2; therefore they are not shown in Figure 10. Waters contaminated by anthropogenic CO2 have experienced changes in saturation states with respect to aragonite. Upward migration of the aragonite saturation horizon in the Atlantic Ocean during the industrial era suggests that CaCO3 particles falling from the surface may begin to dissolve at shallower depths, modifying the in situ dissolution of CaCO3 and the supply of CaCO3 to the sediments.
4.2. Possible Mechanisms Responsible for In Situ Dissolution of CaCO3
 Many of the currently available in situ dissolution rates of CaCO3 particles have been measured at different saturation states (or depths) by determining the decrease in mass of CaCO3 particles [Byrne et al., 1984]. Our dissolution-rate estimate differs from previous estimates in that the estimate not only depends on the decrease in the mass of sinking CaCO3 particles but also on the total mass flux. That is, for a similar decrease in the mass of sinking CaCO3 particles, a greater dissolution rate of CaCO3 will be inferred for areas with greater CaCO3 production in the overlying water column.
 The spatial variability of the dissolution rate shown in Figure 8 could be attributed partly to regional variability in the rain rate of CaCO3 [see Lee, 2001, Figure 4]. The net CaCO3 production estimated by integrating seasonal decreases in PTA in the mixed layer is virtually zero in subtropical areas between 40°and 40°S [Lee, 2001], as PTA is constant throughout the year [Bates et al., 1996; Millero et al., 1998]. The estimated net CaCO3 production in the North Atlantic (>40°) is much higher than in the South Atlantic, which is consistent with the contention that ballasting of organic matter by CaCO3 is more important in the North Atlantic than in the rest of the world [Berger, 1992; Armstrong et al., 2002]. The higher dissolution rates in the North Atlantic are attributable to higher CaCO3 rain rates. Insignificant dissolution rates in the subtropical Atlantic are consistent with insignificant rain rates. In waters with the same degree of saturation with aragonite or calcite, the amount of CaCO3 exported from the surface will determine the extent of the in situ CaCO3 dissolution.
 The significant increase in ΔTACaCO3 along isopycnal surfaces begins at 2500 m in the North Atlantic and 800 m in the South Atlantic, at or below the aragonite 100% saturation horizon. This ΔTACaCO3 increase implies discernable rates of CaCO3 dissolution, suggesting that a significant portion of CaCO3 production in the North Atlantic is dissolved in the water column. However, we cannot rule out that other mechanisms contribute to the excess ΔTACaCO3 near the aragonite saturation horizon. One mechanism is dissolution of CaCO3 in the guts of zooplankton and in fecal pellets. This mechanism was originally proposed by Takahashi  and later by others [Bishop et al., 1980; Pond et al., 1995; Milliman et al., 1999; Jansen and Wolf-Gladrow, 2001]. Several laboratory experiments, however, have shown that the pH in guts of grazers ranges from neutral to alkaline, which is an unfavorable condition for dissolution of CaCO3 [Harris, 1994; Pond et al., 1995]. Dissolution therefore may occur during the early feeding stages [Pond et al., 1995]. The other proposed mechanisms include increased dissolution of CaCO3 in microenvironments where microbial oxidation of organic matter occurs [Jansen and Wolf-Gladrow, 2001] and dissolution of the more soluble phases of CaCO3 such as pteropods and high-magnesium calcite [Betzer et al., 1984; Byrne et al., 1984; Morse and Mackenzie, 1990; Sabine et al., 1995]. An additional mechanism recently proposed by Chen  is that TA generated from the decomposition of organic matter occurring in shelf sediments could be a significant source of excess TA for subsurface waters in the open ocean. In oxygen-depleted shelf sediments, manganese, iron, and sulfate are used as electron accepters to decompose organic matter, all of which increase TA. It is, however, not possible to quantify the effect of shelf waters high in NTA because the horizontal extent of the shelf waters is not known very well. More than one mechanism may collectively contribute to the positive ΔTACaCO3 in super-saturated waters with respect to calcite, although one of them may dominate in certain environments [Milliman et al., 1999].
4.3. CaCO3 Budget in the Atlantic Ocean
 To construct the basin-scale CaCO3 budget we need to know the net CaCO3 production in the euphotic zone and what fraction of it dissolves within the water column before reaching the seafloor for burial. To quantify the amount of CaCO3 dissolved in the water column, the ΔTACaCO3 must be corrected for TA input from the dissolution of sedimentary CaCO3. Existing estimates of global net CaCO3 production are based on direct measurements (e.g., calcification rates and sediment-trap fluxes) or models that combine information about ocean circulation with PTA variations. Estimates range from 42 × 1012 mol C yr−1 to 170 × 1012 mol C yr−1 [see Milliman et al., 1999; Sarmiento et al., 2002]. Recently, Lee  estimated a global net CaCO3 production of [92 ± 25] × 1012 mol C yr−1 by integrating seasonal decreases in PTA in the mixed layer. This value is approximately three times higher than a recently revised trap-based estimate of 34 × 1012 mol C yr−1 at 2000 m depths [Iglesias-Rodriguez et al., 2002]. In our calculation of the CaCO3 budget we used the value of [35.8 ± 9.5] × 1012 mol C yr−1 obtained by Lee  for the Atlantic Ocean.
 The total amount of CaCO3 dissolved in the Atlantic Ocean was estimated by integrating dissolution rates for all the isopycnal surfaces. The inventory of the total ΔTACaCO3 (TOTAL-ΔTACaCO3) for each of three latitude belts (70°N−40°N, 40°N−40°S, and 40°S−70°S) was determined by integrating the mean profile (f-mean) of dissolution rates from surface (SFC) to a mean bottom depth (MD) (Table 2),
Table 2. Estimated Water Column Inventory of ΔTACaCO3 in the Atlantic Oceana
We equate this to the inventory of in situ dissolution of CaCO3.
The CaCO3 dissolution rate for 70°N–40°S does not include integrated rates shallower than 1500 m, because values of ΔTACaCO3 for these waters are <10 μmol kg−1, which is close to the uncertainty in estimating ΔTACaCO3.
 Since our rates were estimated on isopycnal surfaces, a mean depth-density relationship was derived for each of the three latitude blocks (see Figure 8) and applied to the respective density-dissolution-rate profile, to derive a mean depth-dissolution-rate profile for each latitude block. This method yields a basin-scale dissolution rate of [16.7 ± 3.3] × 1012 mol C yr−1, which is probably an overestimate because part of the excess ΔTACaCO3 may be derived from sulfate reduction [Chen, 2002] and dissolution of CaCO3 occurring in sediments.
 Biogeochemical processes within the sediment alter the sedimentation and burial of CaCO3 particles once they reach the seafloor. A major benthic process influencing CaCO3 preservation and dissolution in deep-sea sediments is the oxidation of organic matter [Emerson and Bender, 1981]. Several previous studies have confirmed that CaCO3 dissolves in response to metabolic CO2 produced during degradation of organic matter. Different techniques yield CaCO3 dissolution rates in sediments with mixed results: In situ microelectrode profiles have shown the occurrence of CaCO3 dissolution above the calcite saturation horizon [Archer et al., 1989; Martin and Sayles, 1996; Hales and Emerson, 1997], whereas benthic-chamber incubation experiments suggest that the metabolic CO2-driven dissolution is not as important as suggested by microelectrode measurements [Jahnke, 1994; Jahnke et al., 1994]. The ΔTACaCO3 contribution from the dissolution of sedimentary CaCO3 was estimated here from a limited number of in situ measurements. The entire Atlantic basin was given a mean CaCO3 dissolution rate of 7 ± 4 μmol cm−2 yr−1, as obtained from in situ electrode measurements in sediments of the Ceara Rise (5°N, 45°) of the western tropical Atlantic [Martin and Sayles, 1996; Hales and Emerson, 1997]. This method yields a rate of sedimentary CaCO3 dissolution of [5.6 ± 3.2] × 1012 mol C yr−1, accounting for 34% of the in situ dissolution rate of [16.7 ± 3.3] × 1012 mol C yr−1 for the entire Atlantic basin.
 The resulting water column dissolution rate for the entire Atlantic corrected for the sedimentary dissolution of CaCO3 is 11.1 × 1012 mol C yr−1, which equates to a significant fraction (31%) of the total net CaCO3 production of [35.8 ± 9.5] × 1012 mol C yr−1 for the same area. Our budget calculation would also imply a CaCO3 accumulation rate of 25.6 × 1012 mol C yr−1 for the Atlantic, which is significantly greater than the accumulation rate of 5.4 × 1012 mol C yr−1 for the same area determined by Milliman and Droxler  and Catubig et al.  based on the measured CaCO3 content in numerous deep-sea sediment samples. The estimate of 5.4 × 1012 mol C yr−1 is close to 50% of the global deep-sea CaCO3 accumulation rate. Although the discrepancy between these two independent estimates of the CaCO3 accumulation rate appears large, the difference may be insignificant considering that estimations of the production and water column dissolution of CaCO3, and sedimentary inputs of CaCO3 have potential uncertainties of at least 100% [see Iglesias-Rodriguez et al., 2002, Table 1], which are considerably greater than those used in our budget calculation. Therefore the closure of the CaCO3 budget is presently far from complete and requires a global synthesis of data on alkalinity and sediment trapping. This will also require time series measurements of CaCO3 production along with satellite and ground-based data.
 The ΔTACaCO3 values calculated using WOCE/JGOFS/OACES Atlantic CO2 survey data suggest that significant CaCO3 dissolution occurs in undersaturated waters of 2500−3000 m depth in the North Atlantic and of 500−1000 m depth in the South Atlantic. The excess ΔTACaCO3 rapidly increases at the aragonite 100% saturation horizon. Our analysis suggests that the excess ΔTACaCO3 in intermediate waters results from water column dissolution of CaCO3 particles. Our results generally support the conclusion, drawn from particle-flux data obtained during the North Atlantic Bloom Experiment, that a significant fraction (30−40%) of the total CaCO3 flux dissolves in the water column at 1000−4000 m depth [Yu et al., 2001]. Our results are also consistent with the suggestion by Byrne et al. , Betzer et al. , and Milliman et al.  that more soluble forms of CaCO3 such as aragonite and high-magnesium calcite are a source of excess ΔTACaCO3 in the upper water column. This shallow-water dissolution of CaCO3 creates excess TA in the upper ocean, which is important in the short-term buffering of fossil-fuel CO2 taken up by the ocean.
 This work would not have been possible without the effort of many scientists particularly responsible for the carbon and CFC measurements on the ships during the WOCE/JGOFS/OACES global CO2 survey conducted between 1990 and 1998. We wish to thank all of the U.S. and European scientists who contributed to the Atlantic data set that was compiled for the Global Ocean Data Analysis Project (GLODAP), and Andrew Dickson and an anonymous reviewer for constructive suggestions on the manuscript. This work was partially supported by grant R01-2002-000-00549-0 (2002) from the Basic Research Program of the Korea Science and Engineering Foundation (K. L.), by the Brain Korea 21 Project in 2003 (K. L.), by the National Oceanic and Atmospheric Administration Office of Oceanic and Atmospheric Research (C. S., R. A. F., R. W., J. L. B., and F. J. M.; contract GC 90–220), by the National Science Foundation (C. S.; contract OCE-0137144, R. M. K., and F. J. M.), by the Joint Institute for the Study of the Atmosphere and Ocean, and by the Cooperative Institute of Marine and Atmospheric Studies(K. L.).