The calcite lysocline as a constraint on glacial/interglacial low-latitude production changes

Transcription

1 GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 12, NO. 3, PAGES , SEPTEMBER 1998 The calcite lysocline as a constraint on glacial/interglacial low-latitude production changes Daniel M. Sigman, Daniel C. McCorkle, and William R. Martin Woods Hole Oceanographic Institution, Woods Hole, Massachusetts Abstract. We investigate the response of the calcite lysocline to changes in the export production of the low-latitude surface ocean (the combined equatorial, tropical, and subtropical regions). We employ different CaCO3 throughput schemes in a time-dependent ocean carbon cycle model to separate the CaCO3 production/lysocline balance from the other components of the model response and to estimate the effect of dissolution driven by organic matter decomposition in surface sediments. With this model, we carry out three experiments based on previous hypotheses for the cause of glacial/interglacial atmospheric CO2 variations: (1) a simple increase in low-latitude production, (2) a decrease in its CaCO3/organicarbon rain ratio, and (3) an increase in the depth of organic carbon regeneration. None of these changes, when taken alone, can lower atmospheric CO2 to glacial levels without violating observations of the glacial lysocline depth. If a low-latitude production increase were accompanied by a decrease in the CaCOJorgani carbon rain ratio, then these simultaneous changes could lower atmospheric CO2 to observed glacial levels without causing a significant change in the ocean-average depth of the lysocline. However, even with these simultaneous changes, the required increase in production is 50% or more of the modern rate of production. 1. Introduction The alkalinity of seawater reflects a balance between the input of alkalinity from rivers and the removal of calcium carbonate in sediments and coral reefs [Edmond, 1974; Milliman, 1974, 1993; Opdyke and Walker, 1992; Keir, 1995]. This balance is maintained by the distribution of calcium carbonate accumulation in the deep sea, which is largely controlled by the depth of the "calcite lysocline," the transition separating shallower seafloor sediments in which some calcite is preserved from deeper seafloor sediments in which almost all of the calcite rain dissolves [Berger, 1970; Archer, 1991a]. One of the primary controls on the depth of the lysocline is the calcite saturation depth, which is set by the carbonate ion concentration of bottom water and the pressure dependence of calcite solubility [Broecker and Takahashi, 1978]. Because of its effect on the calcite burial rate, the response of the lysocline to a change in the deep ocean carbonate ion concentration represents a negative feedback which drives deep-sea calcite burial toward a balance with the input of alkalinity from rivers. A change in either the river input rate of alkalinity or the seafloor burial rate of calcium carbonate changes the whole ocean alkalinity inventory. This alkalinity change shifts the ocean-average calcite saturation depth and thus the depth of the lysocline. Changing the depth of the lysocline, in turn, causes a change in the calcium carbonate Now tat Department of Geosciences, Princeton University, Princeton, New Jersey. Copyright 1998 by the American Geophysical Union. Paper number 98GB /98/98GB burial rate, bringing the ocean carbon cycle to a new balance between river input and seafloor burial. Open ocean biological production is one potential driver of changes in the oceanic alkalinity balance. For example, an increase in the open ocean calcium carbonate (CaCO 3) rain rate increases the rate of CaCO, burial on the seafloor above the lysocline. As a result, the whole ocean inventories of alkalinity (ALK) and dissolved inorganic carbon (DIC) decrease in a 2:1 ratio (the stoichiometry of "dissolved CaCO3"). This inventory change causes a decrease in the deep ocean carbonate ion concentration ([CO3'2] oc ALK- DIC, neglecting borate speciation [Broecker and Peng, 1982]), which, in turn, causes a shoaling of the calcite saturation horizon. This shoaling of the saturation horizon causes a shoaling of the lysocline, which decreases the area of seafloor over which CaCO3 is buried, lowering global CaCO3 burial back to its original rate. The response of the lysocline to a change in open ocean CaCO3 production is important for two reasons. First, if river input and shelf burial remain constant, variability in the lysocline depth reflects changes in the rate and composition of the biogenic rain. Second, the whole ocean change in ALK and DIC which drives the lysocline change also affects the concentration of atmospheric CO 2 [Keir, 1995], with a 2:1 increase in ALK and DIC causing a decrease in atmospheric pco 2 [Broecker and Peng, 1982]: (2 * DIC-ALK) 2 pco 2 o ALK-DIC Because of this effect, variations in the lysocline have direct significance to the observation of glacial/interglacial variations in atmospheric CO2, which is ~80 ppm lower during glacial maxima than during peak interglacials [Barnola et al., 1987]. 409

2 410 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT Much paleoceanographic work has addressed the role of biological production in glacial/interglacial atmospheric CO2 variability. The low-latitude ocean (the combined equatorial, tropical, and subtropical regions) is responsible for most deep-sea calcite burial [Milliman, 1993; Catubig et al., 1994, 1998]. Therefore a change in low-latitude production may cause a significant lysocline response. Any hypothesis regarding changes in low-latitude production must be consistent with paleoceanographic data regarding the lysocline, and the expected CO2 impact of a production change must include the effect of the oceanic alkalinity input/output balance. The change in the ocean-average lysocline depth over the last glacial/interglacial cycle is a subject of current research. The Atlantic lysocline was apparently about 1 km shallower during the last glacial [Crowley, 1983, 1985; Curry and Lohmann, 1986], and Howard and Prell [1994] estimate that the glacial lysocline was 0.5 km shallower in the Atlantic and Indian sectors of the Southern Ocean. No clear change in lysocline depth is observed in the Indian Ocean, although preservation and dissolution events have occurred on glacial/interglacial transitions [Peterson and Prell, 1985]. While equatorial and northeast Pacific data suggest that the glacial lysocline was 0.6 km deeper [Farrell and Prell, 1989; Karlin et al., 1992], there is an active debate as to whether this deepening was due to a local increase in CaCO3 production or to an increase in the saturation depth and lysocline of the entire lowlatitude Pacific [Archer, 199 lb; Anderson et al., 1995]. Compiling a global ocean data set, Catubig et al. [1994, 1998] find that the ocean-average lysocline depth of the last glacial was the same as or shallower than that of the Holocene. In this study, we investigate the response of the lysocline to changes in low-latitude production to learn if and by what mechanisms low-latitude biological production changes could have lowered atmospheric CO2 to observed glacial levels without causing large glacial/interglacial changes in the oceanaverage lysocline depth (i.e., 1 km or more). Using a timedependent ocean-atmosphere carbon cycle model with architecture after the model of Keir [1988] ("CYCLOPS"), we have run three model experiments which involve changes in lowlatitude biological production: (1) an increase in low-latitude export production, (2) a decrease in the calcium carbonate/organic C ratio (CaCO3/Corg) of low-latitudexport production, and (3)an increase in the fraction of Corg export production which reaches the deep ocean (i.e., below 1.5 km). Similar experiments have been carried out previously, with a variety of assumptions being made about the role of deep-sea CaCO 3 preservation [Broecker and Peng, 1984; Dymond and Lyle, 1985; Broecker and Peng, 1987; Boyle, 1988; Keir, 1988; Archer and Maier-Reimer, 1994]. We have performed the model experiments using a set of CaCO 3 input/output models which cover the range of models used by earlier workers, and we have incorporated parameterizations /br sediment respiration-driven dissolution [Emerson and Bender, 1981 ] in order to evaluate the potential role of this process. We frame our results in the context of previous hypotheses regarding the role of low-latitude biological export production in glacial/interglacial atmospheric C02 variations. In section 3.1, we address the potential for a simple increase in lowlatitude biological production to lower atmospheric C02 [Broecker, 1982; Broecker and Peng, 1987]. In section 3.2, we investigate the effect of changes in the composition of the low-latitude biogenic rain, focusing on the role of sediment respiration-driven dissolution in the deep-sea CaCO 3 preservation response [Archer and Maier-Reimer, 1994]. Our findings suggesthat neither a simple increase in low-latitude production nor a change in its composition, when taken alone, can explain glacial/interglacial atmospheric CO 2 changes without violating paleoceanographic data on the calcite lysocline. In section 3.3, we consider a coupling of these different hypotheses. These simultaneous changes can lower atmospheric CO 2 while maintaining a constant ocean-average lysocline depth, but the magnitude of the change in export production required to reach glacial observationseems unlikely. 2. Model Description The results presented in this study are from a time-dependent box model of the ocean-atmosphere system which simulates the cycles of carbon, alkalinity, phosphate, 02, 3C, and 4C [Sigman, 1997]. Architecture, circulation, and production/remineralization parameters are taken from the CYCLOPS model of Keir [1988]. We focus our model description on the production/regeneration cycles and the CaCO 3 input/output dynamics, as these aspects of the model are most importanto this study. The model treatment of gas exchange processes and isotope systematics is described by D.M. Sigman and S.J. Lehman (manuscript in preparation, 1998). Interglacial control distributions are given in Table 1. Sigman [1997] compares the distributions of phosphate, DIC, and ALK with those of Keir [1988] and with basin averages of data from the Geochemical Ocean Sections Study Circulation (Figure 1) Water is cycled through the low-latitude surface boxes by mixing with and advection from the intermediate-depth boxes. The intermediate boxes are fed largely from the surface Antarctic, which is, in turn, fed by the upwelling of Circumpolar Deep Water. North Atlantic Deep Water advects into the Southern Ocean Circumpolar deep box, where it mixes with water from other deep boxes and is upwelled into the surface Antarctic, from which it is advected into the intermediatedepth boxes. The Southern Ocean Circumpolar deep box has large mixing terms with the surface Antarctic and with the lowlatitude deep boxes, so that deep water has effectively both North Atlantic and Southern Ocean sources in the model. The details of the circulation are discussed by Keir [1988] Export Production/Regeneration Cycle (Figure 2) We assume that new production in the low-latitude surface ocean completely consumes the available phosphate supply. Low-latitude new production is thus controlled by (1) the phosphate concentration of subsurface waters and (2) the amount of water exchanged between subsurface and surface waters. The C:N:P ratios of exported organic matter in our model are the standard Redfield values of 106:16:1, for consistency with Keir [1988]. In this version of the model, the roles of nitrate and phosphate as major nutrients are not distinguished.

3 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT 411 Table 1. Model Conditions for the Interglacial Control Surface Temperature Salinity DIC O2 PO4 '3 ALK 5'3C A'4C pco2 CO3 '2 øc %,!zmol kg" %, %, ppm!zmol kg" Atlantic Indian South Pacific North Pacific Boreal Atlantic Antarctic Intermediate Atlantic Indian South Pacific North Pacific Deep Boreal Atlantic Atlantic Antarctic Indian South Pacific North Pacific Export Co CaC03 mol m -2 yr 'l Whole Ocean Atmosphere Terrestrial atmosphere surface Atlantic surface Indian m intermediate Atlantic intermediate Indian deep Atlantic deep Indian surface S. Pacific ]( I surface N. Pacific I intermediate S. Pacific intermediate N. Pacific Figure 1. The model architecture and circulation (after the CYCLOPS model of Keir [1988] and following his depiction). Water fluxes are in sverdmps. Single arrows denote advective fluxes, while double arrows denote fluxes in both directions (coupled advective fluxes or mixing). Volumes and areas of the boxes are given by Keir [ 1988].

4 412 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT riverine low latitude dissolved [P04 '3] -- O, Ca/Corg = 0.4 CaC03. Corg CaCO rn 1500 rn... CaCO 3 rain buried CaCO 3 rain dissolved hioh latitude [P04-3] > 0 I Corg 100% v Figure 2. The model production/regeneration cycle (after Keir [1988]). The Ca/Corg ratio is molar. See Table 2 for a comparison with other models and for the alternative parameter values used in the sensitivity tests. For the model experiments described in this study, we focus on the case in which high-latitude production is held constant, so that the phosphate concentration of the high-latitude surface ocean may vary. Runs of the model experiments with other rules for high-latitude production indicate that the choice of the high-latitude production rule is significant only for model experiment 1, a 30% increase in the oceanic phosphate reservoir. This dependence is discussed in the context of the results from model experiment 1. The CaCO3/Cor g rain ratio of export production is different for the low- and high-latitude surface boxes. The export production of the low-latitude surface boxes has a rain ratio of 0.4 in molar C units (ALK/DIC rain ratio = 0.46), while that of the high-latitude boxes has a rain ratio of 0 (ALK/DIC rain ratio = ). The production-normalized whole ocean rain ratio is 0.31 (ALK/DIC rain ratio = 0.36). The Co and CaCO3 water column regeneration profiles are also different for the low- and high-latitude regions. Of the Co raining out of the low-latitude boxes, 84% is degraded in the intermediate-depth boxes, and the remaining 16% is degraded in the deep boxes, so that no Co is buried. The CaCO rain associated with low-latitude production penetrates more deeply into the ocean interior, with 20% of the rain dissolving at intermediate depths and the remaining 80% raining onto the seafloor. Given a rain ratio of 0.4 and the above regeneration profiles, the CaCO3/Co flux ratio into the low-latitude deep ocean is 2, which is consistent with but perhaps in the upper range of observations from deep sediment traps and seafloor studies [Archer, 1991a; Honjo and Manganini, 1992; Honjo et al., 1995; Honjo, 1996]. The high-latitude subsurface is not separated into intermediate and deep boxes, so that Cor production from the high-latitude sin face is completely regenerated in a single subsurface box, which mixes with the low latitude deep boxes. The CYCLOPS production/regeneration parameters are comparable to those of other models (Table 2). The most notable discrepancy with the model of Broecker and Peng [! 986, 1987] ("PANDORA"), which uses a CaCO /Co g rain ratio of 0.2 in its low-latitude and subpolar boxes and a ratio of 0.4 in its polar surface boxes. As discussed in section 3.1 with regard to the results for model experiment 1, the use of different production/regeneration parameters is one potential cause for disagreements among models. In order to check the sensitivity of our results to the production/regeneration parameters, we have also carried out the model experiments using different values for these parameters (Table 2, "Sensitivity Tests"), as described in section Models of CaCO3 Throughput The model experiments are carded out under three different cases of CaCO 3 input/output dynamics: (1) "closed system," (2) "compensated," and (3) "open system." These different cases allow us to resolve the three different components of the carbon cycle response to a change in biological production: (1) the redistribution of ALK and DIC within the ocean, which directly affects atmospheric CO2 and may also cause a change in deep ocean [CO '2]; (2) compensation for any such changes in deep ocean [CO '2] by a transient change in deep-sea calcite preservation; and (3) the migration of the lysocline to a new steady state depth in response to any changes in the steady state CaCO3 rain rate. In the closed system case, there is neither river input of DIC and ALK nor seafloor burial of biogenic CaCO v As a result, production changes simply redistribute DIC and ALK within the ocean-atmosphere system, without changing the whole ocean reservoir of these components. In the compensated and open system cases, whole ocean reservoir changes in DIC and ALK occur via loss or gain of "dissolved CaCO," which represents two equivalents of ALK per mole of DIC. The compensated case includes the potential for the closed system redistribution of ALK and DIC to drive a Table 2. Low-Latitude Carbon Cycle Parameters Model This work, Interglacial Control [Keir, 1988] Ca/Con of Low- Con Exported to CaCO3 Exported Latitude Deep Boxes, to Deep Boxes, Production % % 0.4 a Broecker and Peng 0.2 b [1986,1987] Boyle [1988] 0.16 b or 42 Archer and Maier- Reimer [1994] 0'25a Sensitivity tests c c a Roughly 25% of all calcite production is buried and thus lost from the deep box, so the effective Ca/Co,g is lower, about 0.3 for our model. Broecker and Peng [1987] use a higherain ratio for high-latitude production (0.4); Boyle [1988] does not separate high-and low-latitude production. * River input of dissolved CaCO3 must also be halved in these tests to compensate for the lysocline deepening caused by the halving of CaCO3 production associated with these parameter changes.

5 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT 413 transient preservation or dissolution event in the deep sea ("CaCO compensation") [Broecker and Peng, 1987; Boyle, 1988]. In the compensated case, any change in the deep ocean [CO -2] due to the redistribution of ALK and DIC is removed b y a transient preservation or dissolution event, which changes whole ocean ALK and DIC in a 2:1 ratio. This whole ocean ALK and DIC change, in turn, drives a change in atmospheric CO2. The compensated case does not explicitly model the feedbacks between river input and deep-sea calcite preservation which lead to the oceanic alkalinity balance. Theretore this case does not capture the link between CaCO production and the steady state lysocline depth. In the open system case, deep ocean [CO '2] is explicitly controlled by the balance between the river input of alkalinity and the seafloor burial of biogenic CaCO.. As a result, the open system case includes the steady state lysocline effect as well as the transient lysocline effect of CaCO. compensation and the closed system effect of ALK and DiC Iedistribution. Thus the steady state lysocline response can be estimated by the difference between the open and compensated cases. We use two open system CaCO models: "rain-based" and "mixed layer." The rain-based open system results yield a hypothetical end-member case for the lysocline response, in which bottom water saturation state alone controls the dissolu- tion of biogenic CaCO. This case has no sediment memory. Therefore it includes neither "chemical erosion" [Keir, 1984] nor sedimentary respiration-driven dissolution. The goal of the rain-based case is to provide a simplified view of the lysocline changes which occur in the model experiments, assuming the traditional picture of bottom water undersaturation control on CaCO preservation [Broecker and Takahashi, 1978]. With the mixed layer open system case, we consider the more realistic aspects of the lysocline. For this model, we have developed a parameterization of seafloor dissolution using results from an explicit sediment geochemistry model [Martin and Sayles, 1996]. The parameterization, described in the appendix, includes the effects of both bottom water-driven dissolution and sediment respiration-driven dissolution. In section 3.1, the rain-based nodel is used for the comparison of the open system results with those of the closed system and compensated cases. In section 3.2, the mixed layer open system model is used to study the effect of sedimentary respiration-driven dissolution on the production/lysocline link. In a separate manuscript, we use the mixed layer open system model to investigate the effect of sediment mixed layer dissolution and chemical erosion on the timing of lysocline responses [Sigman, 1997; D.M. Sigman et al., manuscript in preparation, 1998] CaCO3 compensation model. From the open system case final output for the response to a given model experiment, the average [CO3-2] of the deep ocean is calculated. In the compensation step, ALK and DIC are changed in a 2:1 ratio in all boxes equally until the average deep ocean [CO '2] equals that of the interglacial control. Annospheric CO2 is then recalculated given this whole ocean ALK and DIC change. The CaCO 3 compensation procedure is similar to that of Broecker and Peng [1987]. The only significant difference is our inclusion of the Atlantic in our calculation of deep ocean [CO. '2] change [Sigman, 1997]. The significance of this differ- ence is discussed by Broecker and Peng [1987] and is men- tioned in section 3 where relevant Rain-based open system model. In the rain-based open system model, the only calcite vulnerable to dissolution is the yearly calcite rain; the sediments have no memory of previous calcite accumulation. The calcite saturation depth is calculated from the bottom water carbonate concentration ([CO -2]bw) using the equation for the carbonate concentration at calcite saturation ([CO '2].,at)of Broecker and Takahashi [1978] (for consistency with Keir [1988]). Below the saturation depth, dissolution occurs as a linear function of bottom water undersaturation, dissolution = Krb *([CO 2 ]sat-[co 2 ]bw) with a rate coefficient Iqb of 1.2 gmol CaCO cm -2 yr '1 per gmol kg - of [CO '2] undersaturation. K ("rb" tbr rain based) was calculated from a calcite dissolution rate constant (k ) of 30 day - [Kelt, 1980], assuming a linear dependence on undersaturation. The depth at which the dissolution rate equals the C'aCC) 3 rain onto the qeafioor is the rain-based model calcite compensation depth (CCD). Under standard conditions, the rain-based lysocline, as defined by the saturation depth and the CCD, is approximately 800 m thick. We approximate the hypsometry of each basin as a set of ramps, based on the results of Menard and Smith [ 1966] Mixed layer open system model. In the mixed layer open system model, there are 18 levels distributed between 1.5 and 6.5 km in each of the deep basins, with closer spacing in the depth range with the most seafloor area [Menard and Smith, 1966], each level representing a 10 cm deep sediment mixed layer. The dissolution model used by each mixed layer is derived by generating multivariate polynomial fits to output from an off-line sediment geochemistry model [Martin and Sayles, 1996] and using the resultant polynomial expressions in the ocean box model. In this study, we treat bottom water-driven dissolution and respiration-driven dissolution as additive components of the total dissolution; these two components are parameterized separately, as described in the appendix. For the mixed layer open system model, the calcite saturation equation of Ingle [1975] is used for consistency with the off-line sediment geochemistry model [Martin and Sayles, 1996]. The rain-based and mixed layer open system responses differ slightly because of the different calcite saturation equations used in these two versions of the open system model (from Broecker and Takahashi [1978] and Ingle [1975], respectively [see Sigman, 1997]). For the experiments employing the mixed layer open system model (sections 3.2 and 3.3), the Atlantic, Indian, and Pacific low-latitude basins are collapsed into a single set of low-latitude surface, intermediate, and deep boxes. This is done to simplify the explanation of our results. The rain-based open system results (reported in section 3.1) give an adequate sense of the interbasin differences we observe when basin resolution is included in the ex- periments [Sigman, 1997]. In addition, the significance of these interbasin differences is uncertain, given the evidence for major differences in deep circulation during the last ice age [e.g., Boyle and Keigwin, 1982]. 3. Results In this study, we discuss results from three different model experiments, each of which involves a change in low-latitude

6 , 414 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT export production. In model experiment 1, we increase the whole ocean phosphate reservoir by 30%, which drives a simple increase in low-latitude production [Broecker, 1982; Broecker and Peng, 1987; Keir, 1988]. In model experiment 2, we halve the CaCO,/Corg rain ratio of low-latitude production [Broecker and Peng, 1984; Dymond and Lyle, 1985; Archer and Maier-Reimer, 1994]. In model experiment 3, we increase in the fraction of Corg export production which reaches the deep boxes [Boyle, 1988; Archer and Maier-Reimer, 1994]. In section 3.1 we compare the closed system, compensated, and open system effects for model experiment 1. The comparisons for model experiments 2 and 3 are discussed by Sigman [1997]. In section 3.2, we consider all three model experiments with regard to sediment respiration-driven dissolution. In section 3.3, we consider the effects of combining the different model experiments Comparison of Closed, Compensated, and Open System Responses to an Increase in Low- Latitude Production For the three model experiments, we compare the different magnitudes of the atmospheric CO2 response due to (1) the closed system redistribution of ALK and DIC within the ocean, (2) the compensation for the effect of this redistribution on deep ocean [CO3 '2] by a transient calcite dissolution or preservation event in the deep ocean, and (3) the change in the steady state saturation depth driven by changes in the CaCO3 rain rate (Figure 3). The time-dependent results for the closed and open system cases help to illustrate these three different effects in model experiment 1, a 30% increase in the wholeocean phosphate reservoir (Figure 4). Using the open system results, we assess whether this experiment can lower atmospheric CO2 to glacial levels while being consistent with the relative constancy of the lysocline depth through glacial/interglacial cycles. Broecker's [1982] hypothesis of an increase in the phos- phate content of the ocean provides a mechanism to increase low-latitude production without the auxiliary effects associated with changing circulation. Our model experiment follows that of Broecker and Peng [1987], in which the whole-ocean phosphate reservoir is increased by 30%, with low-latitude produc- tion increasing so as to maintain phosphate-free surface water. However, we focus on the case in which the high-latitude productive flux is held constant. We choose this high-latitude fixed production rule for two reasons. First, production in most high-latitude regions is not limited by phosphate or nitrate [e.g., Cullen, 1991; Mitchell et al., 1991 ], so that we would not expect production in these regions to vary with the phosphate supply. Second, allowing high-latitude production to change obscures the effects of low-latitude production. In the comparison of our results with those of Broecker and Peng [1987], we return to this issue, exploring the effect of their "constant surface ocean phosphate residence time" rule, which causes high-latitude production to increase in proportion to the phosphate reservoir increase. Increasing the phosphate reservoir by 30% yields a 34 ppm CO2 decrease for the closed system case, a 29 ppm decrease for the compensated case, and only an 18 ppm decrease for the 0 experiment 1 experiment 2 experiment 3 I i -2 experiment 1 experiment 2 experiment 3 i I ß -10- ( o._ 4- g -30- ' O -40- o o. _.c -50- o E -60..c_ e- 0- o N O.c 0.5- o._ ( -70-._c g 1.5- o -80 o 2 30% CaJCorg ratio deeper Corg phosphate halved regeneration increase (a) closed system compensated 30% CaJC ratio deeper C org org phosphate halved regeneration increase (b) A open system Figure 3. Comparison of the model experiment steady state responses in the closed system case (squares with dashed lines), compensated case (circles with dotted lines), and open system case (triangles with solid lines) for (a) atmospheric CO2 and (b) ocean-average calcite saturation depth. Responses are plotted as differences from the interglacial control. Model experiment 1 is a 30% increase in the whole ocean phosphate reservoir, which causes a 32% increase in low-latitude export production. Model experiment 2 is a halving of the low-latitude CaCO3/Co g ratio of the rain out of the surface ocean (from 0.4 to 0.2, molar ratio). Model experiment 3 is an increase the fraction of Co g export production which enters the deep boxes (from 16% to 40%). For the experiments shown here, high-latitude biological production is held constant.

7 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT (a) 3.5 (c) 280 open.,:: (: 260 closed -o e OO model years 1.5 (b) 2 (d) I I I model years ' I I I I I I model years model years Figure 4. Time-dependent results for model experiment 1, a 30% increase the whole ocean phosphate reservoir. (a) Atmospheric CO2, for both closed and open system scenarios. The compensated case is not shown because it lacks time dependence. The closed system output has been offset by 9 ppm to start at the same initial value as the open system (297 ppm). (b) The instantaneous ratio of CaCO 3 burial to riverine input of "dissolved CaCO3." If the ratio is <1, then the ocean is gaining ALK and DIC in a 2:1 ratio. If the ratio is >1, then the ocean is losing them in the same ratio. The calcite saturation horizons in the (c) Atlantic and (d) North Pacific, which have the youngest and oldest deep water, respectively, are shown. Increased CaCO3 rain associated with the production increase causes CaCO 3 burial to exceed fiver input (Figure 4b), so that the ocean begins to lose ALK and DIC in a 2:1 ratio. As a result, the pco2 of surface water increases, causing atmospheric CO2 to increase from the closed system case (Figure 4a). The whole ocean loss of ALK and DIC also causes deep ocean [CO3' ] to decrease, so that the saturation horizon shoals (Figures 4c and 4d, note the interbasin difference). This shoaling decreases the area of CaCO 3 burial, so that CaCO 3 burial decreases back to equality with river input (Figure 4b). Note that no transient dissolution event is evident. open system case (Figure 3a). Thus lysocline processes reduce As a result, the difference in the lysocline depths between the the magnitude of the CO2 decrease resulting from the produc- Atlantic (Figure 4c) and Pacific (Figure 4d) increases, although tion increase. The time-dependent output plotted in Figure 4 the lysoclines of all deep boxes shoal so as to bring CaCO 3 shows the opposing effects of the closed system and open burial back into balance with the river input. system processes on atinospheric CO 2. Increased production The quantitative link between CaCO3 production and the immediately lowers atmospheric CO2 by pumping D1C out of depth of the steady state lysocline is illustrated in Figure 5, the surface, leading to the initial drop in atmospheric CO2 seen which shows the ocean-average depth of the steady state calfor both closed and open system scenarios (Figure 4a). How- cite saturation horizon over a range of low-latitude production ever, the increase in the CaCO 3 rain associated with increased rates (controlled by varying the whole ocean phosphate reserlow-latitude production effuses a temporary excess in CaCO 3 voir while keeping the CaCO3/C,,rg rain ratio fixed at 0.4). At burial relative to the river input of dissolved CaCO 3 (Figure the standard case production value, the saturation horizon 4b). This causes a loss of ALK and DIC from the ocean in a 2:1 shoals by 175 m for a 10% increase in CaCO 3 rain rate. The ratio, which raises the pco2 of surface water and results in sensitivity of the saturation depth to the CaCO 3 rain rate behigher atmospheric CO2 for the open system scenario (Figure comes greater as the lysocline becomes deeper. This is due 4a, [Keir, 1988]). The loss of ALK and DIC also causes the both to the ocean hypsometry and to an increase in the depth deep [CO3 '2] to decrease, so that the saturation horizon shoals sensitivity of the lysocline at low levels of CaCO 3 production. an average of 540 m throughouthe ocean (Figures 3b, 4c, 4d). That is, as the CaCO 3 rain rate is decreased in the model, the As the saturation horizon and lysocline shoal, the rate of area of seafloor burial required to match the fivefine alkalinity CaCO 3 burial decreases, driving the burial/river ratio back to input increases in a hyperbolic fashion. unity (Figure 4b). An additional reason for the small magnitude of the CO2 The Atlantic-Pacific difference in lysocline changes is due decrease in the open system case is the lack of a transient disto deep ocean circulation [Keir, 1988]. The deep Atlantic re- solution event associated with CaCO 3 compensation. In fact, ceives surface waters which have a higher [CO3 '2] than in the CaCO3 compensation causes a minor preservation event, restandard case, while the old waters of the Pacific have gained sulting in a small (5 ppm) atmospheric CO:. increase, as is more DIC from the regeneration of the increased biogenic rain. evident from comparison of the closed and compensated cases

8 416 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT c 3.5-._o o,) 4- low latitude CaCO 3 rain rate (factor* standard) 0.75 I I Table 3. Atm. CO2 for the Interglacial Control Sensitivity Tests Case Standard Test 1 a Test 2 b Test 3 c Closed Compensated Open Values are given in parts per million of atmospheric volume. a Deep Cor regeneration is 60%. b Low-latitude CaCO3/Cor = 0.2 (fiver input of CaCO3 halved). c Tests 1 and 2 are combined Figure 5. The steady state, ocean-average saturation depth in the rain-based open system model as a function of the imposed CaCO 3 rain rate. The CaCO_ rain rate is varied by altering the low-latitude production rate, using the mechanism of model experiment 1. As the CaCO. rain rate is increased, the saturation horizon shoals to maintain a constant global CaCO. burial rate. (Figure 3a). The compensation effect is driven by a 1.5-gM increase deep ocean [CO3'2] which occurs in the closed system model, representing a -0.1 km deepening of the saturation horizon (Figure 3b). In the compensated case, this [CO3 '2] increase causes a minor excess in deep-sea calcite preservation, causing the ocean to lose ALK and DIC in a 2:1 ratio. This small transient preservation event is not apparent in the open system time-dependent results because of the dominant effect of the steady state lysocline change (Figure 4). By contrast, model experiments 2 and 3 both cause a large dissolution response associated with CaCO3 compensation (Figure 3). This dissolution response is evident in the time-dependent output as a transient shoaling of the saturation horizon [Sigman, 1997]. As demonstrated by Keir [1988], increasing nutrient utilization in the high-latitude ocean drives a significant transient dissolution event, which is responsible for roughly 25% of the CO2 decrease which a high-latitude nutrient change causes. The difference between the transient effects of low- and highlatitude biological production on deep ocean [CO3 '2] is due to the different production/regeneration parameters of low- and high-latitude biological production. The presence of CaCO 3 in the low-latitude biogenic rain and the importance of Corg remineralization in the intermediate-depth boxes both make low-latitude production less effective at lowering deep ocean [CO3'2], relative to high-latitude production. For every mole of DIC which the low-latitude CaCO 3 rain transfers into the deep sea, it transfers 2 eq of alkalinity; these changes, when taken together, increase deep ocean [CO3'2]. The Corg rain pumps DIC into the deep ocean and causes a small decrease in deep alkalinity (associated with No g oxidation), driving a decrease in deep ocean [CO3'2]; however, only 16% of the low-latitude biogenic rain out of the surface boxes survives into the deep boxes, limiting the resultant [CO3 '2] decrease. There are uncertainties in the values which should be used for the production/regeneration parameters of low-latitude biological production, so that it is possible that the chosen parameter values lead to an inaccurate sensitivity of the model to low-latitude biological production. In order to address this concern, we have repeated the model experiments using different values for these parameters. The alternative parameter sets used, referred to as "sensitivity tests," are summarized in Table 2. These cases are not to be thought of as model experiments themselves. Rather, each provides a new initial steady state from which the model experiment is carried out. Sensitivity case 1 assumes a much higher proportion of Co g oxidation in the deep box (60%) than in the standard case (16%). Sensitivity case 2 assumes a low-latitude CaCO3/Co g rain ratio of 0.2 rather than 0.4. Note for this case that the assumed river input must also be halved; otherwise, the "interglacial" steady state lysocline would be far too deep, and atmospheric pco 2 would be far too low. Sensitivity case 3 combines cases 1 and 2. We have not attempted to tune the model to interglacial conditions using these different parameter sets (the change in the river input rate for sensitivity cases 2 and 3 being the one exception). The initial atmospheric CO2 concentration for each of the sensitivity cases is given in Table 3. The atmospheric CO2 change which occurs for model experiment 1 for each of the three sensitivity cases is given in Table 4. Because the standard interglacial conditions vary among the sensitivity test cases, 5 ppm or smaller differences in the response of atmospheric pco 2 to model experiment 1 should not be considered significant. The main reasons for the modest CO2 decrease in model experiment 1, as described above, are (1) the atmospheric CO2 effect of the shoaling of the steady state lysocline and (2) the lack of a dissolution event associated with CaCO3 compensation. The sensitivity cases do not show significant differences in the steady state lysocline response. Moreover, for none of the sensitivity cases do we observe CaCO compensation to significantly enhance the CO2 decrease associated with model experiment 1. That is, over a wide range of assumed production/regeneration parameters, an increase in low-latitude production fails to cause a CaCO 3 compensation-driven transient dissolution event in our model. Table 4. Atm. CO2 Change for Experiment 1 Sensitivity Tests Case Standard Test 1 Test 2 Test 3 Closed Compensated Open Experiment #1 is a 30% increase ocean phosphate. Changes are in parts per million. Deep Co regeneration is 60%. Low-latitude CaCO /Co = 0.2 (river input of CaCO halved). Tests 1 and 2 are combined.

9 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT 417 Broecker and Peng [1987] have carded out a model experiment analogous to our experiment 1 closed system and compensated cases. Their results suggesthat the lysocline re- In summary, switching to the high-latitude "constant phosphate residence time" production rule of Broecker and Peng [1987] changes the compensation response from a small pressponse would significantly enhance the pco 2 drop associated ervation event to a small dissolution event. However, it still with this experiment, from a 33 ppm decrease in the closed system case to a 52 ppm decrease in the compensated case. Our contradictory results are due to several important differences in our models. The first major difference is their assumption of no change in the steady state lysocline in response to the production increase. Their assumption is motivated by the observation of no large glacial/interglacial variations in the ocean-average lysocline depth. However, the hypothesized production increase would, indeed, have caused a lysocline shoaling. We does not generate the large (19 ppm) compensation-driven CO2 decrease which they observe. The difference in high-latitude production rules is fundamental to the difference between the results presented here and the analogous experiment of Broecker and Peng [1987], but we cannot duplicate their results simply by assuming their production rule. Differences in model architecture and production parameters, in conjunction with variable high-latitude production, are responsible for the remaining difference. First, the intermediate-depth boxes of the PANDORA model outcrop directly at cannot disregard the coupling between CaCO3 production and high-latitudes [Broecker and Peng, 1986, 1987]; such the lysocline. Because of it, the lack of an ocean-averaged "subpolar" surface boxes are absent from the CYCLOPS model. lysocline change in the paleoceanographic record is an important constraint. It provides us with information by which to judge whether the hypothesized production change actually occurred, and it must be taken into account when assessing its potential effect on the carbon cycle. The second major difference is the role of high-latitude production in the model experiment. While we hold high-latitude production constant, Broecker and Peng [1987] assume a "constant surface ocean phosphate residence time" rule for Production in the subpolar outcropping region of the intermediate-depth boxes, like that of the high-latitude surface boxes, is remineralized only in the deep boxes of PANDORA and thus efficiently pumps DIC into the deep boxes of their ocean model. Thus enhanced production in these subpolar regions is an additional driver of the deep ocean [CO3 '2] decrease in the PANDORA response to model experiment 1. As with the polar regions, the subpolar regions (such as the Subantarctic) are not currently limited by the major nutrients. Therefore we argue biological production, which causes high-latitude production that subpolar regions, like polar regions, would not respond to to increase in proportion to the phosphate reservoir increase. When we run model experiment 1 assuming the constant residence time rule of Broecker and Peng [ 1987], we find that there an increase in the oceanic phosphate or nitrate supply with a proportional increase in production. Second, the CaCO3/Corg rain ratio is 0.4 in the PANDORA is a small (1!.tM)deep ocean [CO. '2] decrease in the closed polar surface boxes (the Antarctic and the subarctic Pacific), system case. CaCO3 compensation for this decrease causes a 3 ppm CO2 decrease, in contrast to the 5 ppm CO2 increase which compensation causes when high-latitude production is held constant (Figure 6). The CO2 decrease due to CaCO3 compensation is amplified to 7 ppm if only the IndoPacific basin is used to calculate the size of the dissolution event as in Broecker and Peng [1987]. As a result, the net open system response i s slightly less of a CO2 source than when the "constant flux" high-latitude production rule is used (Figure 6) higher than the value of 0.2 used in the low-latitude and subpolar surface boxes (Table 2) [Broecker and Peng, 1986, 1987]. This high rain ratio is inconsistent with sediment trap data from nutrient-rich polar regions [Wefer et al., 1982, 1988; Fischer et al., 1988; Honjo and Manganini, 1990; Honjo, 1996]. Given the PANDORA distribution of rain ratios, an increase in the importance of biological production in the lowlatitude and subpolar ocean relative to that of the polar ocean would drive a decrease in the effective global CaCO3/Corg rain closed. _ open constant flux 280- open C ,z ' 01os_ed _ o compensated constant residence time (a) (b) ,..., ,...,......, model years model years Figure 6. The time course of atmospheric CO2 for both the closed system (dashed line) and open system (solid line) response to model experiment 1, assuming two different export production rules for the highlatitude boxes: (a) constant flux, with no change in high-latitude export production in response to the 30% phosphate reservoir increase, identical to Figure 4, and (b) constant surface phosphate residence time, which causes high-latitude production to increase in proportion to the phosphate supply (as used by Broecker and Peng [1987]). Assuming constant phosphate residence time, the CaCO 3 compensation response is a small transient dissolution event, which makes the net CO2 increase due to open system processe smaller than in the constant flux case. The greater CO2 decrease for the closed system response in Figure 6b than in Figure 6a is due to increased pumping of DIC out of the high-latitude surface boxes.

10 418 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT ratio in the PANDORA model, whereas it would cause an increase in the effective global rain ratio in all other models [Bolin et al., 1983; Keir, 1988; Maier-Reimer, 1993]. Thus we would expecthis assumption of a high polar CaCO3/Corg rain ratio to make lower-latitude production a more efficient driver of deep ocean undersaturation in the PANDORA model than in other models. We have previously noted the lack of a compensationrelated, transient dissolution response to increased lowlatitude production, using a simpler box model including intermediate/deep ocean resolution, similar to that of Lyle and Pisias [1990]. By contrast, we and other workers do observe a transient dissolution event when high-latitude production is increased simultaneously with the low-latitude production increase (Figure 6b) [Broecker and Peng, 1987; Boyle, 1988]. Taken together, these results suggesthat high-latitude production is the major driver of the transient dissolution event observed in previous model experiments of enhanced productivity [e.g., Broecker and Peng, 1987]. In an open syste n ocean, even if an increase in low-latitude production could drive a transient dissolution event of the amplitude suggested by Broecker and Peng [ 1987], its effect on atmospheric CO 2 would be largely canceled by the shoaling of the steady state lysocline. As a result. a simple increase in low-latitude export production is an inefficient mechanism by which to lower atmospheric CO2. This conclusion is insensitive to the mechanism used to increase low-latitude production. We have also carried out a model experiment in which low-latitude production is increased by increasing the mixing between the intermediate boxes and the low-latitude surface boxes [Boyle, 1988; Broecker and Peng, 1987]. The closed system response to this model experiment is different from that of experiment 1, with atmospheric CO 2 increasing because of temperatureffects on CO2 exchange (D.M. Sigman, unpublishe data, 1994). However, the lysocline responses, both transient and steady state, are very similar to those of model experiment Effect Dissolution of Sedimentary Respiration-Driven on the Production-Lysocline Link The model of CaCO 3 dissolution used in the comparison of the closed, compensated, and open system cases in section 3.1 follows the traditional picture of bottom water undersaturationdriven dissolution [Broecker and Takahashi, 1978]. However, two observations from studies of seafloor dissolution suggest that the traditional picture is incomplete: (1) pore water undersaturation due to sediment respiration appears to be a significant driver of calcite dissolution, and (2) for a given degree of undersaturation, calcite dissolution appears to be slower than previously thought [e.g., Emerson and Bender, 1981; Sayles, 1981; Archer et al., 1989; Berelson et al., 1994; Hales et al., 1994; Jahnke et al., 1994; Martin and Sayles, 1996]. In this section, we consider how these findings, when incorporated into a model of sediment mixed layer dissolution, affect the carbon cycle response to our model experiments. Each model experiment is repeated for the open system case, using a sediment model developed from an off-line sediment geochemistry nodel (see appendix and section 2.3.3). Each model experiment is run twice. In the first run, only bottom water undersaturation causes dissolution, and a dissolution rate constant of 30 day ' is assumed [Keir, 1980]. In the second run, our parameterization of sediment respiration-driven dissolution (or "respiratory dissolution") is included, and the dissolution rate constant is changed from 30 day ' to 1 day '. These two changes in the second run must be made together to produce a realistic model of the lysocline [Emerson and Archer, 1990]. We compare the model experiment responses for these two dissolution models, one with respiratory dissolution and one without it (as done by Archer and Maier-Reimer [ 1994]), to investigate the relationship between the saturation horizon and the lysocline. For our parameterization of respiratory dissolution, we have chosen a Co g oxidation profile within the sediments which probably overestimates the importance of respiratory dissolution in the open ocean (see appendix). On the other hand, the CaCO s rain rate of 18.5 g m -2 yr - is probably in the high end of estimates for the CaCO 3 rain rate [Honjo, 1996]. Perhaps the most pertinent point is that 32% of the CaCO3 rain dissolves at the saturation horizon in our standard interglacial case. This value is consistent with estimates from open ocean studies and modeling work, although it may be at the low end of the range of estimates [Archer, 1991 a; Martin and Sayles, 1996]. One of the important determinants of the saturation horizon/lysocline relationship is the depth dependence of the Corg rain within the deep ocean. Sediment trap data and benthic oxygen flux data show a decrease with depth in the supply of Cot g to the sediments, both in the open ocean [Suess, 1980: Martin et al., 1987; Berger et al., 1988] and on continental slopes [Smith, 1978; Hales et al., 1994]. This depth dependence of the Co s rain, because it enhances the relative importance of dissolution on the shallow seafloor, may explain the lack of depth dependence in measured rates of respiratory dissolution above the saturation horizon [e.g., Hales et al., 1994], since the downward decrease in the Co g rain would counteracthe effect of increased calcite solubility with depth. We first show the model experiments under the assumption that the Cot s rain is independent of depth within the deep boxes ( km, Figure 8). Then we show the result of including a relatively minor Co s rain depth dependence, focusing on model experiment 3 (Figure 9). Experiment 1, the-30% increase in low-latitude production, was described in section 3.1. This experiment is essentially unaffected when respiratory dissolution is incorporated into the dissolution model, with the saturation depth shoaling by 400 m and atmospheric CO2 decreasing by 16 ppm (Figure 7). Experiment 2, halving the low-latitude CaCO3/Co g rain ratio, causes a 54-ppm CO2 decrease without the effects of respiratory dissolution. Roughly half of this CO2 decrease is due to a deepening of the steady state saturation horizon [Sigtnan, 1997]. The inclusion of respiratory dissolution results in a 14 ppm greater CO2 decrease (68 ppm rather than 54 ppm) and a 600 m greater deepening of the saturation horizon (Figure 7). Experiment 3, deepening the Co s water column regeneration profile, results in a small (16 ppm) CO2 decrease without the effects of respiratory dissolution. Most of the CO 2 decrease is due to a transient dissolution event associated with CaCO3 compensation [Sigman, 1997]. When respiratory dissolution is incorporated into the dissolution model, there is a much greater CO2 decrease of 55 ppm (Figure 7a)'. This large en-

11 SIGMAN ET AL.' THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINW 419 experiment I experiment 2 experiment 3 experiment I I I I I experiment 2 experiment _ i o_ o c_ 2 30% Ca/c ratio deeper C 30% Ca/C ratio deeper C org org org org phosphate halved regeneration phosphate halved regeneration increase increase (a) ß -.B... w/out resp. diss'n, kd=30/day + w/resp. diss'n, kd=l/day (b) Figure 7. The effect of respiratory dissolution on changes in (a) atmospheric CO2 and (b) the saturation horizon depth for each of the model experiments. The two cases shown are (case 1, squares with dotted lines) no respiratory dissolution and a "fast" dissolution rate constant k d of 30 day ' and (case 2, circles with solid lines) active respiratory dissolution and a slower dissolution rate constant of 1 day ', as used by Archer and Maier- Reirner [1994]. The minor differences between the "no respiratory dissolution" results and the open system results from section 3.1 (Figure 3) are due mostly to the switch to the sediment mixed layer model, which uses the calcite solubility equation of Ingle [1975]. hancement of the CO2 decrease is driven by a 1.4-km increase in the steady state, ocean-average saturation depth (Figure 7b). These results are consistent with those of Archer and Maier- Reimer [1994] for similar experiments. Because of the large CO 2 decreases which model experiments 2 and 3 cause when respiratory dissolution is included, Archer and Maier-Reimer [1994] have suggested that these types of changes represent viable mechanisms for lowering atmospheric CO 2 during the last ice age. While information about the glacial lysocline is incomplete, the compiled data indicate no large difference in the ocean-average lysocline depth between the modem and glacial oceans [Catubig et al., 1994, 1998]. Given the large increases in the saturation depth associated with experiments 2 and 3 (Figure 7b), a key requirement of the Archer and Maier-Reimer [1994] hypothesis, as recognized by Sanyal et al. [1995], is that an increase in the importance of respiratory dissolution would decouple the glacial saturation horizon from the lysocline, so that the saturation horizon could deepen to a degree that would not be evident in the depth distribution of CaCO3 weight percentage from glacial-age sediments. We evaluate this prediction by comparing the changes in saturation depth with the coincident changes in the CaCO3 percentage of surface sediments on the deep seafloor. Figure 8 shows the change in surface sediment CaCO. percentage and CaCO. burial rate for each of the three model experiments, with respiratory dissolution active. For model experiments 2 and 3, the lysocline deepens by 800 m or more. This lysocline deepening can be understood as a result of the importance of bottom water saturation state to respiratory dissolution [Archer, 1991a]; in these experiments, respiration-driven dissolution has not become so great as to dissolve all CaCO. above the saturation horizon. The lysocline deepening that occurs in model experiments 2 and 3 can also be understood from the perspective of the CaCO3 riveffburial balance. These experiments enhance the importance of respiratory dissolution relative to dissolution due to bottom water undersaturation. This makes CaCO3 burial rates less depth-dependent, with proportionally more of the dissolution occurring on the shallow seafloor, removing the relatively sharp depth transition in CaCO3 burial rate which occurs near the saturation horizon in the interglacial control (Figures 8e and 8t). With less CaCO3 burial in the shallow ocean, there must be an increase in CaCO3 burial rates in the deeper ocean to balance the river input. The percentage of CaCO 3 in open ocean sediments does not decrease significantly until the burial rate is 25% or less of the CaCO3 rain rate to the seafloor, because of the relatively low rate of noncarbonate sedimentation [Broecker and Peng, 1982]. Thus the "smearing out" of CaCO3 burial, with lower burial rates over a larger depth range, results in a significant deepening of the lysocline. In model experiment 3, the amount of the CaCO 3 rain that dissolves at the saturation horizon increases from 30% to 80% (Figure 8f). The depth dependence of CaCO 3 burial is greatly reduced in this case, leading to about 800 m of lysocline deepening (Figure 8c). Only once CaCO 3 dissolution is dominated by organic matter respiration is it possible to deepen the saturation horizon without causing a further deepening of the lysocline. In order to achieve these conditions, we have added experiments 2' and 3' as more extreme versions of experiments 2 and 3, respectively (Figure 8). In experiment 2', the CaCO3/Cof rain ratio is lowered to 0.15, rather than to 0.2 for experiment 2. In experiment 3', the fraction of Col reaching the deep box is increased to 60% rather than to 40% in experiment 3. Only in experiment 3', the most extreme case, does it appear that the

12 420 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT CaCO 3 weight fraction CaCO 3 weight fraction I i i! i I i i i I i 2 CaCO 3 weight fraction i i i i I i I i model experiment 1: 30% increase model e:xd. eriment 2: half the Ca/Corg whole ocean [PO4'3 ] rain ratio from 0.4 to 0.2 (2': to 0.15) (a) (b) model experiment 3: increase in fraction of Corg reaching the deep boxes from 16% to 40% (3': to 60%) (c) -e- standard case - - model experiment model experiments2',3' CaCO 3 burial (g m '2 yr 'l) CaCO 3 burial (g m '2 yr 'l) / I I I 2 - -, -', ,.1,, CaCO 3 burial (g m '2 yr 'l) al model experiment 1' 30% increase model experiment 2: half the Ca/Corg whole ocean [PO4'3 ] rain ratio from 0.4 to 0.2 (2': to 0.15) (d) (e) model experiment 3: increase in fraction of Corg reaching the deep boxes from 16% to 40% (3': to 60%) (f) -e-- standard case model experiment model experiments 2',3' Figure 8. Depth profiles (a)-(c) CaCO 3 weight traction and (d)-(f) CaCO3 burial rate (g m '2 yr ']) for each of the model experiments, using the open systemixed layer model with active sediment respiration-driven dissolution and a k d of 1 day ']. In each frame, the open circles and the solid bar at 3.5 km are the depth profile and saturation depth, respectively, for the standard interglacial control. The solid circles and the dashed bar are the new profile and saturation depth, respectively, for the model experiment. Figures 8b, 8c, 8e, and 8f of model experiments 2 and 3 also show the response to experiments 2' and 3' (triangles and dotted bar). saturation horizon can be substantially deeper than the lysocline. Model experiment 3' causes a 0.3 km greater deepening of the lysocline than occurs in experiment 3, although the saturation horizon deepens by an additional 1.2 km (Figure 8c). The depth dependence of CaCO3 burial is already very gradual in model experiment 3, lacking a clear depth transition near the saturation horizon. The higher rate of respiratory dis= solution in model experiment 3' does not significantly change this pattern (Figure 8f), so that there is no additional mass balance feedback driving the lysocline to deepen further. As a result, the increase the Cot rain rate from experiment 3 to experiment 3' counteracts the effect of an increase in the saturation depth, resulting in little net change in the distribution of CaCO3 burial and the depth of the lysocline. The conditions of model experiment 3' are extreme. The CaCO3/Cor ratio of the rain onto the seafloor (to be distinguished from the rain out of the surface ocean) is 0.5 in experiment 3', whereas typical values for this ratio in the modem equatorial, tropical, and subtropical ocean, including productive regions, range between 1 and 2 [Archer, 1991a; Honjo and a,... :.; nn,. ra... :..., 1995; Honjo, 1996]. The transition from modem conditions to the conditions at which the saturation horizon and lysocline start to separate requires a major deepening of the ocean-average lysocline (Figure 8c). While decreasing the CaCO rain rate (model experiments 2 and 2') has the same general effect on the saturation depth as increasing the Co,s rain rate (model experiments 3 and 3'), these two mechanisms do not have identical effects on the

13 SIGMAN ET AL.: THE LYSOCLINE AS A PAI OCEANOGRAPI-1IC CONSTRAINT 421 relationship between the saturation horizon and the lysocline. Model experiment 2', unlike experiment 3', does not separate the lysocline from the saturation horizon. The low-latitude CaCO 3 rain cannot be significantly less than 0.15 in our model; below this value, there is not enough CaCO3 rain to balance the rivedfie input of alkalinity. Thus, in our model, a simple decrease in the CaCO3 rain rate cannot cause a separation of the lysocline from the saturation horizon. To this point, we have assumed that the amount of Corg being oxidized in sediments is independent of depth below 1.5 km (i.e., in the deep boxes). Here we repeat model experiments 3 and 3', in this case assuming a depth dependence of the Corg rain to deep-sea sediments. We use the Co g rain depth dependence of Archer and Maier-Reimer [1994], (1) adjusted for the depth dependence in the CaCO 3 rain assumed in their model and (2) normalized to give the same Co g rain rate at 3.5 km as in the case of no Co g rain depth dependence (see appendix). The Co g rain depth dependence yields a 30% lower Corg rain rate at 6.5 km than at 1.5 km, which is gradual in comparison to the decrease defined by the equations of Suess [1980], Martin et al. [1987], and Berger et al. [1988] (Figure A3). Even with this relatively subtle depth dependence of the Co rain, there is a significantly stronger coupling between the saturation horizon and the lysocline. When the depth dependence is included, the lysocline deepening for both model experiments 3 and 3' is only slightly less than the deepening of the saturation horizon (Figure 9a). This stronger coupling between the saturation horizon and the lysocline, compared with the results withouthe depth dependence in the Co g rain, is due to an increase in calcite dissolution on the shallow seafloor relative to deeper dissolution near the saturation horizon (Figure 9b). Respiratory dissolution well above the saturation horizon causes both the saturation horizon and the lysocline to deepen because it increases the area of seafloor burial required to match the river input of alkalinity. In contrast, respiratory dissolution near the saturation horizon tends to shoal the lysocline relative to the saturation horizon. By adding a depth dependence to the Co g rain, we increase the amount of dissolution on the shallow seafloor relative to dissolution near the saturation horizon. This enhances the tendency for respiratory dissolution to deepen the lysocline, while it reduces its capacity to decouple the lysocline from the saturation horizon. This effect is particularly important for the hypothetical case in which the saturation horizon is shifted to very great depths [Sanyal et al., 1995], resulting in a comparable deepening of the steady state lysocline. In summary, our model experiments suggest that deepening of the saturation horizon due to an increase in respiratory dissolution leads to a significant deepening of the lysocline as well. For what we consider to be our most realistic model of respiration-driven dissolution (i.e., including a depthdependent Co g rain), the lysocline/saturation horizon decoupling is insignificant relative to the overall lysocline deepening (Figure 9). Even under the most favorable conditions, we would expect the ocean-average lysocline to deepen by at least a kilometer in the hypotheses explored by Archer arwl Maier- Reimer [1994] and Sanyal et al. [1995]. Lysocline deepening of this magnitude is not observed. Therefore our results argue against large production-driven changes in the depth of the saturation horizon over the last glacial/interglacial transition. This constraint is reinforced if shelf CaCO 3 burial was reduced during the last glacial [Opdyke and Walker, 1992], since this alone would drive a deepening of the lysocline [Archer and Maier-Reirner, 1994] Coupled Changes in the Quantity and Characteristics of Low-Latitude Production Neither an increase in low-latitude production, a change in its CaCO3/Cor g rain ratio, nor a deepening of Co g regeneration, when occurring alone, could have lowered atmospheric CO2 to CaCO 3 weight fraction , I I I I I I I i,,,.,i i 2 CaCO 3 burial (g m '2 yr '1) (a) (b) standard case - -- model experiment at- model experiment 3' Figure 9. Depth profiles of (a) CaCO 3 weight fraction and (b) CaCO 3 burial rate (g m '2 yr ' ) for the interglacial control, experiment 3, and experiment 3', analogous to the output shown in Figures 8c and 8f. In this case, a subtle depth dependence in the Co g rain is included (see text). This more realistic treatment results in an even stronger coupling between the saturation horizon and the lysocline.

14 422 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT \. ': '".. t CaCO3/Corg rain ratio (molar) (a) ß " "...,/',.,',,/",;. "x, CaCO3/Corg rain ratio (molar) (b) Figure 10. Changes in (1) atmospheric CO2 (solid contours, in parts per million) and (2) the average depth of the saturation horizon (dotted contours, in kilometers, positive for deepening) due to combined changes in low-latitude production (controlled by varying the oceanic phosphate reservoir, as in experiment 1) and the low-latitude CaCO3/Corg rain ratio (as in experiment 2). (a) Respiratory dissolution not included the sediment mixed layer model, and k d is set at 30 day '.(b) Respiratory dissolution included, and kd is set at 1 day '. If we assume that the ocean-average steady state saturation horizon depth does not change on glacial/interglacial transitions, then the combined changes which could have lowered atmospheric CO, are defined by the solid bold line. glacial levels without causing unobserved changes in the depth of the steady state lysocline. As we have shown, inacceptable parameter values would shift accordingly. Nevertheless, the fundamental requirement of a balance remains. creasing low-latitude production causes the steady state lysocline to shoal (experiment 1), while decreasing the rain ratio of this production causes the lysoc[ine to deepen (experiment 2). Here we combine enhanced Froduction with a decreased rain 4. Conclusions In this study, we have investigated the response of the ratio to address whether the combination of these different ocean carbon cycle to various changes in low-latitude producchanges causes a significant CO2 decrease without violating the paleoceanographic observation of a relatively constant lysocline depth. Such a simultaneous change is certainly possible given current knowledge of the controls on the characteristics of export production. The CaCO3/Cor g ratio of the biogenic flux is anticorrelated with production rates in the modem ocean [Tsunogai and Noriki, 1991 ]. As a result, an increase in lowlatitude production, such as that resulting from model experition by deconvolving the response into its three parts: (1) the "closed system" redistribution of ALK and DIC within the ocean, (2) the transient response of deep-sea calcite preservation to this redistribution ("CaCO. compensation"), and (3) the adjustment of the steady state lysocline depth to changes in the biogenic CaCO 3 rain rate. Our results indicate important constraints on low-latitude biological production changes which have been previously hypothesized to drive observed glacial/interglacial atmosment 1, may be associated with a decrease in the CaCO3/Co, pheric CO2 variations. Broecker [1982] and McElroy [1983] rain ratio [Keir and Berger, 1983]. The quantitative aspects of the predicted steady state balance depend on the dissolution model. Without respiratory dissolution (Figure 10a), the model requires roughly proportional complementary changes in the rate of low-latitude production and the CaCO3/Co, rain ratio of this production. That is, a doubling in production requires roughly a halving of the rain ratio. With respiratory dissolution included in the dissolution model (Figure 10b), a given change in the CaCO3/Cor s rain ratio causes a greater deepening of the saturation horizon (and a greater decrease in atmospheric COO, requiring a greater increase in low-latitude production to maintain a constant saturation depth. With or without respiratory dissolution, large production changes, corresponding to a 50% increase in oceanic phosphate or more, are required to lower atmospheric CO2 by 80 ppm without changing the saturation depth. If the glacial saturation depth were taken to be different than the modern depth or if there was a component of saturation depth change due to river input or shelf burial changes, then the range of described possible mechanisms by which the major nutrient reservoirs might increase during glacial times, leading to an increase in low-latitude biological production. However, it was recognized that this basic mechanism would require large changes in nutrient reservoirs to produce the observed amplitude of CO2 change. Broecker and Peng [1987] presented model results suggesting that CaCO3 compensation would significantly increase the CO2 drawdown associated with a nutrientdriven production increase. This would make the nutrient reservoir hypothesis a viable explanation for lower atmospheric CO,. during the last ice age. Since only low and middle latitude regions are limited by nitrate and phosphate today, it seems most reasonable to expect only those regions to respond to an increase in the oceanic inventory of these nutrients. Because CaCO is an important component of the biogenic rain in the low and middle latitudes, a simple increase in this rain causes an increase in the area-normalized rate of CaCO burial, which is countered by a shoaling of the lysocline to a new steady state depth. This

15 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT 423 shoaling of the lysocline to a new steady state is associated with a decrease in oceanic alkalinity, which works to increase atmospheric CO2. In addition, when only low and middle latitude biological production are allowed to increase, we find that CaCO 3 compensation, the transient component of the lysocline response, does not have a significant effect on atmospheric CO2 because an increase in the low-latitude biogenic rain pumps alkalinity as well as dissolved inorganic carbon into the deep ocean, causing little change in the [CO3 '2] of deep water. Since the steady state lysocline response works to increase atmospheric CO2, while CaCO 3 compensation has very little effect (and may even work to increase COO, the low CO2 levels of the last ice age could not have been due to a simple increase in low-latitude biological production. This represents an important limitation on the nutrient reservoir hypotheses for glacial/interglacial atmospheric CO, change [Broecker, 1982; McElroy, 1983; Ganeshrarn et al., 1995]. Decreasing the CaCO3/Corg rain ratio of low-latitude production and increasing the fraction of the Corg rain which reaches abyssal depths both drive a decrease in atmospheric CO:, partly through their effect on deep-sea CaCO 3 burial [Broecker and Peng, 1984; Dymond and Lyle, 1985; Boyle, 1988; Sigman, 1997]. Archer and Maier-Reimer [1994] showed that sediment respiration-driven dissolution increases the capacity of these changes in low-latitude biological production to lower atmospheric CO2. The greater CO2 decrease is due to enhanced deepening of the calcite saturation horizon when respiratory dissolution is included in the seafloor dissolution model [Archer and Maier-Reimer, 1994]. However, our results demonstrate that respiratory dissolution does not effectively decouple the lysocline from the saturation depth in the oceanic response to production changes, so that the deepening of the lysocline is also enhanced. A large glacial deepening of the whole-ocean lysocline is not observed in paleoceanographic CaCO percentage data, arguing against large productiondriven changes in the depth of the steady state saturation horizon. In a separate manuscript, we also consider the potential effect of chemical erosion on the timing of lysocline migration, concluding that chemical erosion similarly cannot decouple the lysocline depth from CaCO3 production on glacial/interglacial timescales [Sigman, 1997; Sigman et al., manuscript in preparation, 1998]. Given these model results and the available data on glacial/interglacial lysocline variability, the only low-latitude production mechanism which could explain lower atmospheric pco 2 during the last glacial involves the coupling of changes which have opposite effects on the lysocline. As an example, we have considered an increase in low-latitude export production, which would cause the steady state lysocline to shoal, coupled with a decrease in its CaCO3/Cor g rain ratio, which would cause the lysocline to deepen. Even with this coupling, a 50% increase or more in low-latitude biological production would be required to cause an 80-ppm decrease in atmospheric pco 2. Such high production rates in the glacial low-latitude ocean seem unlikely. Because of the very low CaCO3/Corg rain ratio of export production in nutrient-rich polar regions, hypotheses regarding glacial/interglacial changes in highlatitude production and nutrient status are not similarly constrained by the history of the steady state lysocline. Thus our results provide additional motivation to look to the polar re- gions for the cause of glacial/interglacial atmospheric CO2 variations. Appendix: Parameterizing Calcite Dissolution in the Sediment Mixed Layer Model Rather than incorporating an explicit model of dissolution into an ocean model (as done by Archer and Maier-Reimer [1994]), we have generated multivariate polynomial fits to output from a sediment geochemistry model [Martin and Sayles, 1996] and used these parameterizations in the ocean box model. Since we wish to run the model both with and without the effects of sediment respiration-driven dissolution, we treat bottom water undersaturation-driven dissolution and sediment respiration-driven dissolution as additive components of total dissolution [Martin and Sayles, 1996], with independent parameterizations for these two types of dissolution. Bottom water-driven dissolution is defined as the dissolu- tion which occurs when there is no Corg respiration within the sediments. Sediment respiration-driven (or "respiratory") dissolution is defined as the difference between the total dissolu- tion and bottom water-driven dissolution. The variables incorporated into the parameterization for respiratory dissolution are (1) the bottom water saturation state, (2) the Co rain rate, and (3) the weight fraction of CaCO 3 in the sediment. Variables which were considered in our study but which are not included in the multivariate fit are (1) the calcite dissolution rate constant ka [Keir, 1980] and (2) the oxidation profile of Co within the sediment column. The stone variables, except for those relating to C o:, are considered in the parameterization of bottom water-driven dissolution. The explicit sediment geochemistry model is described by Martin and Sayles [ 1996]. For bottom water driven dissolution, we assume an equation of the form dissolution = Kbw *([CO 2 ]sat-- [CO '2]bw) n * Fca m (after Keir, [1988]), where K w is the coefficient relating bottom water undersaturation to dissolution rate, [CO3'2]sat is the carbonate concentration at saturation for calcite (after Ingle [1975]),[CO3'2] w is the carbonate concentration of the bottom water, n is the exponent relating bottom water undersaturation to dissolution rate, F, is the weight fraction of calcite in the sediment mixed layer, and m is the exponent relating the weight fraction of calcite to dissolution rate. By running the sediment geochemistry model over a range of undersaturation (for approximately 100% CaCO3), we estimate n and K w. Running the model over a range of CaCO 3 percentages (by varying the CaCO 3 rain rate) for several degrees of undersaturation, we find that m is close to 0.5. This has proven to be independent of variation in other parameters. The same square root dependence of dissolution rate on CaCO 3 weight fraction was used by Keir [1988]. The bottom water dissolution rate varies logarithmically with the calcite dissolution rate constant ka. We consider two different dissolution rate constants in our investigation of respiratory dissolution. For the case with respiratory dissolu- tion, we use a dissolution rate constant ka of 1 day 4 for consistency with Archer and Maier-Reirner [1994]. For the case without respiratory dissolution, we use a ka of 30 day 'l [Keir,

16 424 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT... / / / / respiration-driven calcite / / dissolution (gmol cm-2 yr-1) / / ' 10- ' 0 I o 20 r,.) Corg rain rate (pmol cm '2 yr '1) Figure A1. The sediment respiration-driven calcite dissolution rate as a function of bottom water saturation state and Corg rain rate, calculated using the explicit sediment geochemistry model of Martin and Sayles [1996]. In the cases where bottom water is undersaturated with respect to calcite, the model is run both with and without respiration-driven dissolution, and the difference is calculated. The dissolution rate constant k a is 1 day ' as used by Archer and Maier-Reimer [1994], and the calcite rain rate is in excess to keep the calcite fraction at a constant value of approximately 100%. 1980], rather than the value of 100 day ' used by Archer and Maier-Reimer, which generates an unrealistically sharp lysocline in our model. We have parameterized bottom water-driven dissolution for both of these cases. For a k d of 1 day 'l, with [CO3'2] in units of mol kg ', the equation is dissolution = 3.76 * 1012 *([CO '2]sat - [CO '2]bw) 2'35 * Fca 0'5 with dissolution in units of gmol CaCO 3 cm -2 yr '. For a kd of 30 day 'i, the equation is dissolution = * 1012 *([CO '2 ]sat- [CO '2]bw) 2'35 * Fca 0'5 The value of 2.35 estimated for n is similar to the value of 2.75 used by Keir [1988]. This is encouraging, as we have determined these values independently. Table A1. Polynomial for Dissolution Coefficient Term a'b ACO3>0 ACO3<0 Intercept ACO Rorg ACO E-06 ACO E-06 Rorg Rorg ACO3*Rorg ACO3*Rorg ACO3*Rorg ACO32*Rorg ACO3 2*ROrg E-06 ACO32*Rorg E-08 ACO33*Rorg E-06 ACO33*Rorg E E-07 ACO33*Rorg E E-09 a Units are as follows: ACO3, gmol kg 't (positive for supersaturation); Rorg, gmol C cm - yr' ; calculated dissolution, gmol CaCO3 cm '2 yr -t. b Boundaries are as follows: ACO3, <-19.8, >38.2; and Rorg, >30. Extrapolate using the first derivative. To parameterize respiratory dissolution, we have run the dissolution model over a range of (1) saturation state and (2) Corg rain rate. The dissolution rate constant used is 1 day - after Archer and Maier-Reimer [1994], and the CaCO 3 weight percentage is close to 100% (see below). The Co g oxidation profile is taken from a site of relatively deep sedimentary Cofg oxidation (site H) in the Ceara Rise study of Martin and Sayles [1996]. A deep Co g oxidation profile places DIC production far enough below the sediment surface to limit the buffering effect of carbonate ion diffusing in from bottom water, so that our E' 2'5 CaCO 3 dissolution (% of rain rate) i,.,j, i i i i i i i w/resp. diss'n, no Corg profile, kd=l day 'l + w/resp. diss'n, w/corg profile, kd=l day 'l no resp. diss'n, kd=30 day 'l, o 5-._ Figure A2. Depth profiles of calcite dissolution (as a percentage of the rain) in the interglacial control for the mixed layer open system model with (1) no respiratory dissolution and an assumed ka of 30 day ' (triangles), (2) active respiratory dissolution and an assumed ka of 1 day 'z, with the Co rain to the seafloor independent of water column depth below 1.5 km (open circles), and (3) the same as condition 2 but with a Co rain depth dependence adopted from Archer and Maier-Reimer [ 1994] (solid circles).

17 ß SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT 425 fraction of Corg rain remaining (relative to 1.5 km) I I adopted here (Archer and Maier- Reimer [1994], adjusted for CaCO 3 rain)... Archer and Maier-Reimer [1994] Berger et al. [1988]... Martin et al. [1987] Suess [!980] Figure A3. The depth dependence of the Corg rain adopted from Archer and Maier-Reimer [1994], before (dashed line) and after (solid line) adjustment for the depth dependence in their CaCO3 rain and normalization t o give the same Cot g rain rate at 3.5 km as in the depth-independent case. The fraction of Co rain remaining relative to 1.5 km is shown. The depth-dependent case is compared with the depth-independent case in section 3.2. The equations of Suess [1980], Martin et al. [1987], and Berger et al. [1988] are plotted for comparison to illustrate that the depth dependence adopted here is relatively minor. use of a deep oxidation profile represents a significant bias Acknowledgements. We thank R. Keir for his help with the CYCLOPS model conditions. The manuscript benefited greatly frown toward higher respiratory dissolution rates (W.R. Martin, unreviews by D. Archer, E. Boyle, and R. Toggweiler and from the compublished data, 1996). nents of J. Adkins, B. Hales, R. Oxburgh, A. Sanyal, and J. Sarmiento. From these results, we subtract the dissolution rate calcu- This work was funded by the NSF Graduate Fellowship Program, the lated assuming no Co rain. This difference is taken as the res- JOIFUSSAC Ocean Drilling Graduate Fellowship Program, and NSF piratory dissolution component (Figure AI)and is fitted to a grant OCE to D.M. This is WHOI contribution term polynomial function of saturation state and Co rain rate (Table A1). When this parameterization is incorporated into the ocean box model, dissolution at the saturation hori- References zon is 32% of the calcite rain (Figure A2). We find that the respiratory dissolution rate, unlike that of Anderson, R.F., Y. Lao, M.Q. Fleisher, and F. Marcantonio, New evidence for dissolution control of carbonate cycles in equatorial Pacific bottom water-driven dissolution, is essentially independent of sediments, paper presented at 5th International Conference on Pa- CaCO3 percentage until the CaCO3 percentage is below ap- leoceanography, Halifax, Nova Scotia, Canada, proximately 10%. Thus no dependency on CaCO 3 weight per- Archer, D., Modeling the calcite lysocline, J. Geophys. Res., 96, 17,037- centage, CaCO 3 rain rate, or noncarbonate rain rate is needed 17,050, 1991a. for the component of dissolution driven by sediment respira- Archer, D., Equatorial Pacific calcite preservation cycles: Production or dissolution?, Paleoceanography, 6, , 1991 b. tion. Archer, D., A data-driven model of the global calcite lysocline, Global The polynomial fit was made for the range of saturation Biogeochern. Cycles, 10, , states from +40 to -20 gmol/kg (Figure A!). For more super- Archer, D., and E. Maier-Reirner, Effect of deep-sea sedimentary calsaturated and undersaturated conditions, the model results are cite preservation on atmospheric CO 2 concentration, Nature, 367, , extrapolated using the first derivative of the polynomial fits at Archer, D., S. E nerson, and C.R. Smith, Dissolution of calcite in deepthe boundary. sea sediments: ph and 0 2 microelectrode results, Geochirn. Cosrno- In our study of respiratory dissolution, we consider two chirn. Acta, 53, 2,831-2,845, cases for the Co rain rate to the seafloor. In the first case, we Barnola, J.M., D. Raynaud, Y.S. Korotkevitch, and C. Lorius, Vostok ice assume no depth dependence for the amount of Cot which rains core provides 160,000-year record of atmospheric CO2, Nature, 329, , onto the seafloor (within the deep boxes). In the second case, Berelson, W.M., D.E. Hammond, J. McManus, and T.E. Kilgore, Dissowe use the Cot g rain depth dependence of Archer and Maier- lution kinetics of calcium carbonate in equatorial Pacific sediments, Reimer [1994], adjusted for a depth dependence which they Global Biogeochern. Cycles, 8, , assume for the CaCO 3 rain rate and normalized to give the same Berger, W.H., Planktonic foraminifera: Selective solution and the lysocline, Mar. Geol., 8, , Corg rain rate at 3.5 km as the case without depth dependence Berger, W.H., K. Fisher, C. Lai, and G. Wu, Ocean carbon flux: Global (Figure A3). The resultant Co g rain depth dependence yields a maps of primary production and export production, in Biogeochemi- 30% lower Co g rain rate at 6.5 km than at 1.5 kin. This Co g cal Cycling and Fluxes Between the Deep Euphotic Zone and Other rain decrease is gradual in comparison to the decrease defined Realms, Res. Rep 88-1, edited by C.R. Agegian, pp , NOAA Undersea Res. Prog., Silver Spring, Md., by the equations of Suess [1980], Martin et al. [1987], and Bolin, B., A. Bjorkstrom, K. Holmen, and B. Moore, The simultaneous Berger et al. [1988]. Its effect on the depth dependence of use of tracers for ocean circulation studies, Tellus, Ser. B, 35, 206- seafloor CaCO. dissolution in the standard case is shown in 236, Figure A2. A relatively gradual Co rain depth dependence was Boyle, E.A., The role of vertical chemical fractionation in controlling chosen to yield a conservative estimate of the coupling be- late quaternary atmospheric carbon dioxide, J. Geophys. Res., 93, 15,701-15,714, tween the saturation depth and lysocline in tlie deep ocean, Boyle, E.A., and L.D. Keigwin, Deep circulation of the North Atlantic since a greater depth dependence causestronger coupling be- for the last 200,000 years: Geochemical evidence, Science, 218, 784- tween the saturation horizon and lysocline (see section 3.2). 787, 1982.

18 426 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT Broecker, W.S., Ocean chemistry during glacial time, Geochim. Cosmochim. Acta, 46, , Broecker, W.S., and T.-H. Peng, Tracers in the Sea, 690 pp., Eldigio, Palisades, N.Y., Broecker, W.S., and T.-H. Peng, The climate-chemistry connection, in Climate Processes and Climate Sensitivity, Geophys. Monogr. Ser., vol. 29, edited by J.E. Hansen and T. Takahashi, pp , AGU, Washington, D.C., Broecker, W.S., and T.-H. Peng, Carbon cycle 1985: Glacial to interglacial changes in the operation of the global carbon cycle, Radiocarbon, 28, , Broecker, W.S., and T.-H. Peng, The role of CaCO 3 compensation in the glacial to interglacial atmospheric CO 2 change, Global Biogeochem. Cycles, 1, 15-29, Broecker, W.S., and T. Takahashi, The relationship between lysocline depth and in situ carbonate ion concentration, Deep Sea Res., 25, 65-95, Catubig, N.R., D.E. Archer, R. Francois, P.B. demenocal, W. Howard, and E.-F. Yu, Global deep-sea burial rate of calcium carbonate during the last glacial maximum, Paleoceanography, 13, , Catubig, N.R., D.E. Archer, W.R. Howard, P.B. demenocal, and P.N. Froelich, an atlas of LGM calcium carbonate concentrations in deep sea sediments and implications for the deep sea burial rate (abstract), Eos, 75(44), Fall Meet. Suppl., 367, Crowley, T.J., Calcium carbonate preservation patterns in the central North Atlantic during the last 150,000 years, Mar. Geol., 51, 1-14, Crowley, T.J., Late Quaternary carbonate changes in the North Atlantic and Atlantic/Pacific comparisons, in The Carbon Cycle and Atmospheric C02: Natural Variations Arcbean to Present, Geophys. Monogr. Set., vol. 32, edited by E.T. Sundquist and W.S. Broecker, pp , AGU, Washington, D.C., Cullen, J.J., Hypotheses to explain high-nutrient conditions in the open sea, Limnol. Oceanogr., 36(8) , Curry, W.B., and G.P. Lohmann, Late Quaternary carbonate sedimentation at the Sierra Leone Rise (eastern equatorial Atlantic Ocean), Mar. Geol., 70, , Dymond, J., and M. Lyle, Flux comparisons between sediments and sedimentraps in the eastern tropical pacific: Implications for atmospheric CO 2 variations during the Pleistocene, Limnol. Oceanogr., 30(4), , Edmond, J.M., On the dissolution of carbonate and silicate in the deep ocean, Deep Sea Res. Oceanogr. Abstr., 21, , Emerson, S., and D. Archer, Calcium carbonate preservation in the ocean, Philos. Trans. R. Soc. London, Ser. A, 331, 29-40, Emerson, S., and M.L. Bender, Carbon fluxes at the sediment-water interface of the deep sea: Calcium carbonate preservation, J. Mar. Res., 39, , Farrell, J.W., and W.L. Prell, Cli nate change and CaCO 3 preservation: An 800,000 year bathymetric reconstruction from the central equatorial Pacific Ocean, Paleoceanograp hy, 4, , Fischer, G., D. Fuetterer, R. Gersonde, S. Honjo, D.R. Ostermann, and G. Wefer, Seasonal variability of particle flux in the Weddell Sea and its relation to ice cover, Nature, 335, , Hales, B., S. Emerson, and D. Archer, Respiration and dissolution in the sediments of the western North Atlantic: Estimates from models of in situ microelectrode measurements of porewater oxygen and ph, DeepSea Res., Part 1, 41, , Honjo, S., Fluxes of particles to the interior of the open oceans, in Particle Flux in the Ocean, edited by V. Ittekkot et al., Wiley Interscience, New York, Honjo, S., J. Dymond, R. Collier, and S.J. Manganini, Export production of particles to the interior of the equatorial Pacific Ocean during the 1992 EqPac experiment, Deep Sea Res. Part II, 42, , Honjo, S., and S.J. Manganini, Particle fluxes and modern sedimentation in the polar oceans, in Polar Oceanography, Part B: Chemistry, Biology, attd Geology, edited by J.W.O. Smith, pp , Academic, San Diego, Calif., Honjo, $., and S.J. Manganini, Particle fluxes to the interior of the North Atlantic Ocean studied at two JGOFS stations: 34N 21W and 47N 20W, Deep Sea Res., Part A, 40, , Howard, W.R., and W.L. Prell, Late Quaternary CaCO 3 production and preservation in the Southern Ocean: hnplications for oceanic and atmosphericarbon cycling, Paleoceanography, 9, , Ingle, S.E., Solubility of calcite in the ocean, Mar. Chem., 3, , Jahnke, R.A., D.B. Craven, and J.-F. Gaillard, The influence of organic matter diagenesis on CaCO 3 dissolution at the deep-sea floor, Geochim. Cosmochim. Acta, 58, , Karlin, R., M. Lyle, and R. Zahn, Carbonate variations in the northeast Pacific during the late Quaternary, Paleoceanography, 7, 43-61, Keir, R.S., The dissolution kinetics of biogenic calcium carbonates in seawater, Geochim. Cosmochim. Acta, 44, , Keir, R.S., On the late Pleistocene ocean geochemistry and circulation, Paleoceanography, 3, , Keir, R.S., Recent increase in Pacific CaCO 3 dissolution: a mechanism for generating old 4C ages, Mar. Geol., 59, , Keir, R.S., Is there a component of Pleistocene CO 2 change associated with carbonate dissolution cycles?, Paleoceanography, 10, , Keir, R.S., and W. Berger, Atmospheric CO 2 content in the last 120,000 years: The phosphat extraction model, J. Geophys. Res., 88, , Lyle, M., and N. Pisias, Ocean circulation and atmospheric CO 2 changes: Coupled use of znodels and paleoceanographic data, Paleoceanography, 5, 15-41, Maier-Reimer, E., Geochemical cycles in an ocean general circulation model: Preindustrial tracer distributions, Global Biogeochem. Cycles, 7, , Martin, J.H., G.A. Knauer, D.M. Karl, and W.W. Broenkow, VERTEX: carbon cycling in the northeast Pacific, Deep Sea Res.. Part A, 34, , Martin, W.R., and F.L. Sayles, CaCO 3 dissolution in sediinents of the Ceara Rise, western equatorial Atlantic, Geochim. Cosmochim. Acta, 60, , Menard, H.W., and S.M. Smith, Hypsometry of ocean basin provinces, J. Geophys. Res., 71, , Milliman, J.D., Marine Carbonates, 375 pp., Springer-Verlag, New York, Milliman, J.D., Production and accumulation of calcium carbonate in the ocean: Budget of a nonsteady state, Global Biogeochem. Cycles, 7, , Mitchell, B.G., E.A. Brody, O. Holm-Hansen, C. McCLain, and J. Bishop, Light limitation of phytoplankton biomass and macronutrient utilization in the Southern Ocean, Limnol. Oceanogr.. 36(8), , Opdyke, B.N., and J.C.G. Walker, Return of the coral reef hypothesis: Basin to shelf partitioning of CaCO 3 and its effect on atmospheric C02,Geology, 20, , Peterson, L.C., and W.L. Prell, Carbonate preservation and rates of climate change: An 800 kyr record from the Indian Ocean, in The Carbon Cycle and Atmospheric CO : Natural Variations Arcbean to Present, Geophys. Monogr. Set., vol. 32, edited by E.T. Sundquist and W.S. Broecker, pp , AGU, Washington, D.C., Sanyal, A., N.G. Hemming, G.N. Hanson, and W.S. Broecker, Evidence for a higher ph in the glacial ocean from Boron isotopes in foraminifera, Nature, 373, , Sayles, F.L., The composition and diagenesis of interstitial solutions-ii. Fluxes and diagenesis at the water-sediment interface in the high latitude North and South Atlantic, Geochim. Cosmochim. Acta, 45, , Sigman, D.M., The role of biological production in Pleistocene atmospheric carbon dioxide variations and the nitrogen isotope dynamics of the Southern Ocean, Ph.D. thesis, 385 pp., Joint Program in Oceanogr., Mass. Inst. of Technol./Woods Hole Oceanogr. Inst., Woods Hole, Mass., Smith, K.L., Jr., Benthic community respiration in the N.W. Atlantic Ocean: In situ measurements from 40 to 5200 m, Mar. Biol., Berlin, 47, , Suess, E., Particulate organic carbon flux in the oceans- Surface productivity and oxygen utilization, Nature, 288, , Tsunogai, S., and S. Noriki, Particle fluxes of carbonate and organic carbon in the ocean: Is the marine biological activity working as a sink of atmosphericarbon?, Tellus, Ser. B, 43, Wefer, G., E. Suess, W. Balzer, G.'Liebezeit, P.J. Muller, A. U ngerer, and W. Zenk, Flux of biogenic components from sediment trap de-

19 SIGMAN ET AL.: THE LYSOCLINE AS A PALEOCEANOGRAPHIC CONSTRAINT 427 ployment in circumpolar water of the Drake Passage, Nature. 299, , Wefer, G., G. Fischer, D. Fuetterer, and R. Gersonde, Seasonal particle flux in the Bransfield Strait, Antarctica, Deep Sea Res., Part A, 35, , W.R. Martin, Depart nent of Marine Chemistry and Geochemistry, Woods Hole Oceanographic Institution, MS 8, Woods Hole, MA ( D.C. McCorkle, Department of Marine Geology and Geophysics, Woods Hole Oceanographic Institution, MS 8, Woods Hole, MA ( dmccorkle@whoi.edu) D.M. Sigman, Department of Geosciences, Princeton University, Guyot Hall, Princeton, NJ ( sigman@ geo.princeton.edu) (Received December 18, 1996; revised March 20, 1998; accepted April 8, 1998.)