Effect of temperature-dependent organic carbon decay on atmospheric pco 2

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112,, doi: /2006jg000187, 2007 Effect of temperature-dependent organic carbon decay on atmospheric pco 2 Katsumi Matsumoto, 1 Taketo Hashioka, 2 and Yasuhiro Yamanaka 2,3 Received 27 February 2006; revised 4 November 2006; accepted 19 January 2007; published 21 April [1] Extendingy an almost universal observation that the rate of microbial activity increases with temperature, we propose that marine microbial activity was suppressed during previous glacial periods and allowed proportionally more organic carbon to be exported out of the surface ocean. A stronger organic carbon pump and therefore lower rain ratios of CaCO 3 to organic carbon may have contributed to the low atmospheric CO 2 content during the Last Glacial Maximum. Previous study of temperature-dependent export production (Laws et al., 2000) and our map of data-based, global distribution of the rain ratios lend support to today s rain ratios being controlled at least partly by temperature. A close examination with a high-resolution regional ocean ecosystem model indicates that the correlation between rain ratio and temperature is caused indeed by preferential remineralization of organic matter, but a part of the correlation is also driven by temperature-dependent community composition. An extrapolation of these results to the globe using a global carbon cycle box model with a module for sediments indicates that the drawdown of atmospheric CO 2 by the proposed mechanism is approximately 30 ppm. While this estimate is subject to uncertainty, the fact that it represents nearly one third of the glacial-interglacial variation in atmosphere pco 2 suggests the potential importance of the new mechanism. Given the historical difficulty in explaining the full CO 2 amplitude with a single cause, we suggest that a set of multiple mechanisms were responsible and that the temperature-dependent POC degradation rate is one of them. We discuss two possible difficulties with our proposal that have to do with the potentially important role that ballasts play in organic carbon export and the possibility that enhanced biological pump is self limiting. Citation: Matsumoto, K., T. Hashioka, and Y. Yamanaka (2007), Effect of temperature-dependent organic carbon decay on atmospheric pco 2, J. Geophys. Res., 112,, doi: /2006jg Introduction [2] Vostok ice core measurements show that atmospheric CO 2 content was lower during the past four peak glacial periods by 80 to 100 ppm than during interglacial periods including the present Holocene [Petit et al., 1999]. This natural CO 2 fluctuation is large and occurred in tandem with changes in air temperatures over Antarctica and global continental ice volume. Since its first discovery in the 1980s, the fluctuation remains a mystery, although it is well understood that the deep ocean with its large carbon reservoir must have played a key role [Broecker, 1982]. As reviewed by Archer et al. [2000] and Sigman and Boyle [2000], a number of explanations has been proposed over the years to account for the low atmospheric CO 2 content during the Last Glacial Maximum (LGM) approximately 1 Department of Geology and Geophysics, University of Minnesota, Minneapolis, Minnesota, USA. 2 Graduate School of Environmental Earth Sciences, Hokkaido University, Sapporo, Japan. 3 Frontier Research System for Global Change, Yokohama, Japan. Copyright 2007 by the American Geophysical Union /07/2006JG000187$ ,000 years ago. Most explanations involve changes in either the ventilation of deep waters or surface biological productivity. Early studies using simple box models have shown that atmospheric CO 2 content is sensitive to those changes particularly in high-latitude surface regions where deep waters outcrop [Knox and McElroy, 1984; Sarmiento and Toggweiler, 1984; Wenk and Siegenthaler, 1985]. Biological productivity in low-latitude surface waters can also reduce atmospheric pco 2 by either enhancing the biological pump (i.e., the export of particulate organic carbon (POC) out of the mixed layer) [Broecker, 1982], or by diminishing the export of CaCO 3 relative to that of POC (i.e., the so-called rain ratio hypothesis [Archer and Maier-Reimer, 1994]). Despite all the work for over two decades, there is as yet no explanation for the mystery that is widely accepted. The mechanisms proposed thus far either violate some paleoclimatological constraint, such as the deep ocean lysocline reconstruction [Farrell and Prell, 1989], or are insufficient to account for the full glacialinterglacial CO 2 fluctuation (see reviews by Archer et al. [2000] and Sigman and Boyle [2000]). [3] The difficulty in explaining the mystery may be rooted in our effort to identify a single cause. It is tempting 1of9

2 to seek a single cause, because the fluctuation is so regular and highly correlated with global continental ice volume changes over the past 650,000 years [Siegenthaler et al., 2005]. If multiple mechanisms were responsible for the CO 2 fluctuation, they were likely interdependent and exerted similar amounts of influence on atmospheric CO 2 content in each of the last four glacial-interglacial cycles. Otherwise, the high regularity would be difficult to explain by multiple mechanisms save coincidence. [4] In spite of all the uncertainty that surrounds this mystery, we place relatively high confidence on three glacial boundary conditions and their implications. First, lower sea surface temperatures (SST) during peak glacial periods increased CO 2 solubility. Available paleoceanographic evidence suggests that cooling in low-latitude surface temperatures during the last glacial maximum compared to the Holocene was between 2 3 C [Broecker, 1986; CLIMAP Project Members, 1981; Rostek et al., 1993] and 5 C [Guilderson et al., 1994; Stute et al., 1992]. High-latitude cooling could not have exceeded 4 C to avoid seawater from freezing. Simple models show that a reduction in SST in low latitudes by 5 C and in high latitudes to near freezing point lowers atmospheric pco 2 by roughly 30 ppm [Matsumoto et al., 2002; Sigman and Boyle, 2000]. Second, a 3% increase in sea surface salinities (SSS) (equivalent to 1 PSU), corresponding to a 120 m drop in sea level [Fairbanks, 1989], decreased CO 2 solubility. The same models show that this increases atmospheric pco 2 by about 7 ppm. Finally, the land-to-ocean transfer of about a 500 Pg-C (Pg = g), inferred from a whole ocean d 13 C difference of about 0.3% between LGM and the Holocene [Boyle, 1992; Duplessy et al., 1988; Matsumoto and Lynch- Stieglitz, 1999], would raise atmospheric pco 2 by about 40 ppm before some of this rise is neutralized by ocean sediments [Archer et al., 2000]. The increase in atmospheric pco 2 would be even larger, if we use higher estimates of land-to-ocean transfer [Adams et al., 1990; Kaplan et al., 2002]. The combined effect of these three conditions raises the atmospheric CO 2 content, making the task of explaining the low glacial CO 2 more difficult. These conditions are deemed relatively reliable, because they are simple, largely abiotic, and do not call for unknown changes in ocean circulation or biological production. [5] Here we evaluate a new mechanism for lowering atmospheric pco 2 that shares nearly the same attributes. It calls for the dependence of the rate of organic matter decomposition on temperature. The dependence is based on well-established observations that biological activity increases with temperature: organisms grow faster and die younger under higher temperatures. In a pioneering work, Eppley [1972] showed that phytoplankton growth rates are roughly doubled for a 10 C increase in temperature (i.e., Q 10 = 2.0). It has been shown on land as well that microbial activity increases exponentially with increasing temperature [Edwards, 1975], and the relation often and likewise shows aq 10 of 2.0 [Singh and Gupta, 1977]. [6] All else being equal, this temperature dependence would enhance carbon export under glacial conditions, because the low ocean temperatures would retard the decay rate of sinking POC. Within the mixed layer, this effect would reduce the export rain ratio of CaCO 3 to organic carbon. This is consistent with the export production being strongly dependent on temperature [Laws et al., 2000]. Outside the mixed layer, this effect would lengthen the depth scale of POC remineralization or degradation into inorganic forms. Both effects will lower atmospheric pco 2 and produce a distribution of dissolved inorganic carbon (DIC) that is bottom-heavy [Boyle, 1988]. [7] The new mechanism is different from the more traditional triggers of strengthening the biological pump. The usual way is to increase primary production by invoking either a larger inventory of a macronutrient such as phosphorus [Broecker, 1982] or a greater availability of micronutrient such as iron [Martin, 1990]. By assuming constant remineralization rate, these forcing would indeed increase the export of POC. Our contention is that a similar effect can be realized by just slowing the organic carbon remineralization rate without changing the amount of primary production. [8] In this study, we first examine a data-based distribution of the export rain ratio for signs of temperature dependence. Then we use a prognostic ocean ecosystem model coupled to a high-resolution, regional circulation model to quantify the sensitivity of POC remineralization and rain ratio to temperature changes. In doing so, we consider the indirect effect that temperature can have on the rain ratio through community composition. Finally, we use a simple global ocean-sediment carbon cycle model to obtain a rough estimate of how much the proposed mechanism may affect atmospheric pco Models and Data [9] Following Sarmiento et al. [2002], we use an ocean biogeochemical-transport box model of the top 100 m of the water column and measurements of ALK and nitrate to revisit the modern distribution of export rain ratios. The model essentially assumes that a downward transport of ALK and DIC by particulate matter across the 100 m boundary is balanced primarily by an upward transport of ALK and DIC by upwelling. This model is therefore applicable to regions where the horizontal gradients of ALK and DIC are much smaller than the vertical gradient, so that lateral transport can be largely neglected. The vertical transport of CaCO 3 to organic carbon across the boundary is estimated with the mean values of potential ALK (P ALK ; ALK normalized by salinity and nitrate) and salinity-normalized nitrate (nno 3 )[Sarmiento et al., 2002, equation (11)]: J CaCO3 ¼ 1 J organicc 14:6 P ALK;d P ALK;s nno 3;d ; ð1þ nno 3;s where subscripts s and d correspond respectively to above and below the boundary. The constant 14.6 reflects both the elemental stoichiometry of C:N of 117:16 [Anderson and Sarmiento, 1994] and the fact that one CO 3 2 ion accounts for two alkalinity equivalents. [10] We use the gridded ALK and nitrate data from GLODAP (Global Ocean Data Analysis Project) [Key et al., 2004]. The data set is comprised primarily of measurements from WOCE (World Ocean Circulation Experiment) and JGOFS (Joint Global Ocean Flux Study) and is publicly 2of9

3 available at the GLODAP website ( gov/oceans/glodap/glodap_home.htm). [11] We investigate the mechanism of temperature dependence of the CaCO 3 to organic carbon rain ratio with a 15-compartment prognostic, ecological model coupled to an ocean general circulation model of the northwest Pacific [Hashioka and Yamanaka, 2007]. The model has a horizontal resolution of 1 1 and 54 vertical levels, and its domain reaches 60 N in the north and 20 N in the south. The ecological model includes two types of phytoplankton and three types of zooplankton. The larger phytoplankton type represents diatoms, the main agents of vertical transport of POC. The smaller phytoplankton type includes a fixed fraction (10%) of coccolithophorids, and this fraction is scaled to net primary production. The large and predatory zooplankton types represent copepods and euphausiids respectively. Ten percent of the small zooplankton type represents foraminifera. The sinking of coccolithophorids and foraminifera exports CaCO 3 out of the mixed layer. Other prognostic state variables of the model include ALK, DIC, silicic acid, nitrate, ammonium, and dissolved and particulate organic matter. [12] In the 15-compartment ecological model, the rate of phytoplankton growth depends on available nutrient concentrations with the classical Michaelis-Menton kinetics [Monod, 1949], available sunlight, ambient water temperature, and the concentration of phytoplankton itself. The rates of mortality of all organisms have the same quadratic dependence on their mass c e kt biomass 2, where c is a constant of proportionality, k = , and T is temperature in C. The value of for k gives the Q 10 = 2.0 temperature dependence [Eppley, 1972]. The rates of decomposition of all particulate matters have a linear dependence on their mass c e kt biomass. The value of k is again for all particles. The value of constant c is 1/4 as small for water column CaCO 3 dissolution as for particulate organic nitrogen decomposition. The relative magnitude of this important parameter (1/4 for CaCO 3 and 1 for POC) describes the different decomposition rates of different particles. That is, POC is more easily degraded than CaCO 3, as indicated by observations of vertical distribution of nutrients and alkalinity and of remineralization depth scales of the two particles. The values of this and other parameters in the ecosystem model have been determined by tuning the model to observations from the modern North Pacific [Hashioka and Yamanaka, 2007]. In our sensitivity study, the temperature of the ecological model was changed by +5 C and 5 C compared to our standard model until steady state was reached. The temperature in the physical model is unchanged to ensure that ocean dynamics remain unchanged for all simulations. [13] The effect of temperature-dependent POC remineralization on atmospheric pco 2 is assessed with the global atmosphere-ocean-sediment carbon box model of Matsumoto et al. [2002]. The surface ocean is divided into five regions: northern high latitudes, low latitudes, subantarctic, Antarctic polar, and Antarctic Divergence. This model s physical and biological parameters, including export production, were objectively determined by an inverse model, so that the misfit between observed and simulated nutrient and carbon isotope distributions in the ocean and atmosphere is minimized. The control run has an export rain ratio of that is consistent with found by Yamanaka and Tajika [1996]. In the sediment model, dissolution of CaCO 3 is driven by bottom water carbonate ion undersaturation and depends on the CaCO 3 content of the sediments. There is no contribution of respiration CO 2 to the dissolution. If anything, this omission will likely underestimate the response of atmospheric pco 2 [Sigman et al., 1998], as we discuss in section 3.2. Inclusion of the sediment model allows the depth of the calcite saturation horizon to change and thereby simulates carbonate compensation explicitly. Steady state solutions of the model are obtained after 3,000 years without sediments (i.e., closed system) and after 30,000 years with sediments (open system). 3. Results [14] We first discuss the spatial distribution of the rain ratio of CaCO 3 to organic carbon on the basis of modern ocean data. We then use our ecosystem model to investigate how the rain ratio depends on temperature. Finally we use our box model to estimate the temperature effect on atmospheric CO Temperature Dependence of Organic Carbon Decomposition [15] The distribution of the estimated rain ratios being exported at 100 m water depth is shown in Figure 1. The estimates at each GLODAP data grid point are quite variable in space (Figure 1a). However, it can be seen that the estimated ratios are generally higher in the warmer, lowlatitude waters than in colder, higher-latitude waters. This trend is consistent with our expectation and is clearer in the latitudinal distributions of the estimated rain ratio for each ocean basin (Figures 1b, 1c, and 1d). As pointed out by Sarmiento et al. [2002], the rain ratio in low latitudes is roughly about 0.1, which is much lower than the classical value of 0.25 [Li et al., 1969]. The scatter in the estimated rain ratio about the general trend can be quite large, because the simple vertical transport model used to derive the rain ratio can break down locally. Much of the large scatter is caused by the diminishing vertical gradient in nitrate across the 100 m boundary and the low signal of ALK relative to its measurement precision. For this reason, Sarmiento et al. [2002] estimate the rain ratio in much larger regions, where the assumptions of the model hold up better, and obtain the uncertainty in their estimates by a bootstrap method. As indicated by the bootstrap method and suggested by the scatter in the local rain ratio estimates (Figures 1b, 1c, and 1d), the estimation has the highest level of statistical significance in the warm, low latitudes, where the ratios are generally the highest. This is consistent with a more recent study [Jin et al., 2006] that also find higher ratios in the topics and subtropics and lower values in the high latitudes, although the values are somewhat higher overall than those of Sarmiento et al. [2002]. [16] Results from our prognostic ecosystem model show a much clearer temperature dependence of the export rain ratio for the standard run (Figure 2a). As expected, the rain ratios are lowest in the coldest waters of the model domain in the subarctic and are about The highest ratios can be higher than 0.15, and these are found in the warmest waters 3of9

4 Figure 1. Data-based estimates of the modern distribution of rain ratios. (a) Local rain ratio was estimated at each grid point of the GLODAP data set. Latitudinal distribution is shown for (b) Indian, (c) Atlantic, and (d) Pacific. See section 2 for the method. in the subtropics. Rain ratios with intermediate values fall between the extreme water temperatures. [17] However, only about half of this correlation between the model-predicted rain ratios and temperatures is actually caused by the temperature-dependent, preferential remineralization of organic matter over CaCO3. The other half is due to phytoplankton community composition, which also varies with temperature. As in the real ocean, the cold subarctic waters in the model domain are dominated by the large phytoplankton group consisting of diatoms. Warm waters have comparatively more coccolithophorids. The annual mean diatom fraction is much higher in the colder subarctic (over 60% of the total primary production) than in the warmer subtropics. This spatial dependence of the community composition does not necessarily indicate a direct causal relationship between it and temperature, because phytoplankton growth also depends on available nutrients and sunlight. The correlation between phytoplankton composition and temperature partly arises from the fact that available nutrients and sunlight are correlated to various degrees with temperature. [18] In Figure 2a, there is a break at around 21 C in the trend of the model-predicted rain ratio. The smooth and predictable dependence below this temperature is absent above. Above 21 C, the rain ratio is affected by the fact that there is very little nitrate and thus production. In these waters, the small phytoplankton type has significant competitive advantage over the large type, because the former has larger surface area to volume ratio. The smaller type is thus able to take up nutrients faster from the ambient waters than the larger type. This ecological competition is partly to blame for the behavior of the rain ratio above 21 C. [19] In our +5 C and 5 C sensitivity simulations (Figures 2b and 2c), we can largely remove the temperature effect of phytoplankton composition (i.e., diatoms versus coccolithophorids) and isolate more clearly the effect on preferential remineralization of organic matter over CaCO3. The community composition is essentially unchanged in these simulations, because the large and small phytoplankton types have the same Q10 = 2.0 dependence. For both cases, the rain ratio anomaly increases with temperature (in Figures 2b and 2c, rain ratios from the standard run were subtracted from both sensitivity runs). The anomaly is larger for the +5 C simulation than the 5 C simulation, because the temperature dependence of biological activity is exponential in the model following Eppley [1972]. Also, the temperature effect on the rain ratios can still be seen above the transition at 21 C. [20] Outside the mixed layer, the effect of temperature on POC degradation alters the length scale of remineralization (Figure 3). Our +5 C and 5 C sensitivity simulations show as expected that the cumulative degradation of POC referenced to its value at 100 m is larger for the +5 C experiment than the 5 C experiment at all depth (Figures 3c and 3d). However, in terms of absolute rates, temperature-dependent POC degradation is most pronounced in the upper 100 m (i.e., export rain ratio) than below it (i.e., remineralization depth scale) at both subtropical and subarctic sites (Figures 3a and 3b). For example, the subarctic decomposition rate in the +5 C experiment 4 of 9

5 (dashed line, Figure 3a) compared to the control experiment (solid line, Figure 3) is significantly greater (i.e., more positive) above 100 m. This leaves less POC to sink out in the +5 C experiment, so that the absolute degradation rate below 100 m is actually less (i.e., more negative) than the control. The opposite is true in the 5 C experiment, where more POC escapes degradation and this effect is again most pronounced in the upper 100 m (dash-dotted line, Figures 3a and 3b). Evidently the effect of temperature on POC degradation has a more dominant impact on the export rain ratio than on POC remineralization depth scale. [21] These results are entirely consistent with the study of Francois et al. [2002]. Because organic matter in warmer surface waters is more thoroughly processed by complex food webs and microbial loops [Laws et al., 2000], the exported organic matter is less (i.e., lower f-ratios) but is more refractory in nature and thus less likely to be degraded in the ocean interior [Francois et al., 2002]. [22] We can tentatively extrapolate the export rain ratio response from our regional model to the world ocean by multiple linear regression models. Predictions of rain ratio anomaly for the +5 C and 5 C simulations (Figures 2b and 2c) are modeled as functions of temperature and nitrate, which we identified as being the most important independent variables. We give separate regression models for the two simulations because the slopes are different (see amplitudes in Figures 2b and 2c): RainRatioAnomalyðþ5 Þ¼7: þ 1: T þ 4: ½NO 3 Š ð2þ RainRatioAnomalyð 5 Þ¼ 5: : T 1: ½NO 3 Š: ð3þ Figure 2. Rain ratios predicted by the ecological model of Hashioka and Yamanaka [2007]. (a) Rain ratios are shown for our standard simulation. Rain ratio anomalies are shown for (b) +5 C experiment and (c) 5 C experiment. In both cases, rain ratios from the standard experiment were subtracted from those of the sensitivity experiments. Horizontal axis is annual mean temperatures in the top 100 m. [23] In these equations, both temperature T ( C) and nitrate concentration (micromole-n/l) are annual mean values in the top 100 m. The correlation coefficients for these equations are 0.84 and 0.80 respectively. These equations, combined with annual mean climatology of temperature and nitrate [Levitus and Boyer, 1994; Levitus et al., 1993], predict a whole ocean change in the export rain ratio of for the +5 C simulation and for the 5 C simulation. These are clearly extrapolations, because the range of temperatures in our regional model does not span the world ocean temperatures Effect of the Temperature Dependence on Atmospheric pco 2 [24] Figure 4 shows the steady state response of atmospheric CO 2 content to changes in the global mean export rain ratio, as predicted by the box model of Matsumoto et al. [2002]. Two cases are shown. Figure 4a shows the more traditional way of changing the ratio by keeping the POC flux constant and changing the flux of CaCO 3. This is essentially a carbonate pump case, and it may be driven by a shift in ecosystem composition. Figure 4b shows the more relevant case to this study, in which the change in rain ratio is driven by POC flux. [25] The atmospheric pco 2 response without sediments (i.e., closed system) represents the response of the atmosphere to an oceanic redistribution of DIC and ALK (circles, Figure 4). In Figure 4a, the response to the entire range of perturbation ( rain ratio) is limited to within ±20 ppm. [26] The closed system response is much larger when organic carbon pump is the forcing (circles, Figure 4b). For a forcing of (i.e., a warming case), the response is in excess of +80 ppm. For a cooling forcing of 0.074, the response is greater than 60 ppm, which is more than half of the full glacial-interglacial amplitude in atmospheric pco 2. However, not all of this can be attributed to the temperaturedependent mechanism that we are evaluating. When the global rain ratio is reduced from the initial to by increasing the POC export flux everywhere, nitrate becomes nearly depleted in low-latitude surface waters. In other words, this is how much it takes to deplete the initial, low-latitude nitrate concentration of about 4 umol kg 1.Atthis point, atmospheric CO 2 drawdown is 30 ppm (vertical dashed line, Figure 4b). Any further reduction in the global rain ratio is thus achieved by increasing the export of organic carbon at higher latitudes. Eventually even high-latitude surface waters become depleted as organic carbon flux is increased. [27] The response with the effect of carbonate compensation included (i.e., open system) is large when the forcing is 5of9

6 Figure 3. Vertical profiles of particulate organic nitrogen remineralization rate predicted by the ecological model of Hashioka and Yamanaka [2007]. Absolute decomposition rate is plotted for sites in (a) subarctic and (b) subtropics. The decomposition rate below 100 m is normalized to that at 100 m at (c) subarctic site (45 N, 155 E) and (d) subtropics (30 N, 155 E). Solid line indicates the standard model. The +5 C and 5 C experiments are indicated with dashed and dash-dotted lines, respectively. carbonate pump (crosses, Figure 4a) but small when it is organic carbon pump (crosses, Figure 4b). Although our sediment model does not include respiratory CO 2 dissolution, we expect our conclusions drawn from Figure 4b to be reasonable, given that the omission of respiratory CO 2 dissolution has almost no effect on atmospheric CO 2 for an organic carbon pump forcing [Sigman et al., 1998]. The same omission for a carbonate pump forcing apparently reduces the response of atmospheric CO 2 [Sigman et al., 1998], but this is of limited relevance to this study, which calls for a change in POC flux (Figure 4b) rather than CaCO 3 flux (Figure 4a). 4. Discussion [28] Our analysis of data-derived rain ratios suggested and model-predicted export rain ratios showed clearly a correlation with temperature. Our ecological model indicated the importance of both ecological community composition and the degradation rate of organic matter. The dependence of POC decay rate on temperature in our ecological model (Figure 2) is entirely consistent with the results of Laws et al. [2000]. The underlying dependence of metabolic rates on temperature appears robust. We believe that the latitudinal distribution of data-derived modern rain ratios (Figure 1) can be explained at least qualitatively with our mechanism. Sarmiento et al. [2002] did not consider the temperature dependence and instead noted that some aspects of their distributions are somewhat at odds with expectations and are difficult to verify with observations. The correlation between temperature and community composition is difficult to characterize, as it simply may reflect a correlation of temperature with, for example, nutrients and light that may actually control the ecology. 6of9

7 Figure 4. Steady state response of atmospheric CO 2 content to rain ratio perturbations from the oceansediment carbon cycle model of Matsumoto et al. [2002]. Rain ratio change is driven by a change in either (a) CaCO 3 flux or (b) POC flux, while keeping the other flux constant. Bold circles indicate the standard run. Crosses and open circles indicate model runs with and without sediments. Vertical dashed line in Figure 4b indicates the point where the low-latitude surface nitrate is depleted (see text). Note that the rain ratios cannot be translated readily to temperature because they are also predicted using nitrate. [29] On the basis of these arguments, we propose that the export rain ratios in the glacial oceans were reduced, as a result of cold seawater retarding the rate of POC degradation. An extrapolation of our results indicated that a5 C cooling would reduce the export rain ratio by about Available paleoceanographic evidence suggests that a 5 C cooling may be the upper limit of SST change during LGM [Broecker, 1986; CLIMAP Project Members, 1981; Guilderson et al., 1994; Rostek et al., 1993; Stute et al., 1995]. However, it appears unlikely that the new mechanism alone can cause a 0.02 drop in the global rain ratio, because nitrate may become depleted in low-latitude surface waters. Achieving a global 0.02 drop requires increased POC export in high latitudes, and the increase must be large in order for the areally small polar waters to affect the global rain ratio. This suggests that once the low-latitude surface waters are depleted in nitrate, the mechanism that draws down atmospheric CO 2 switches to a high-latitude nutrient depletion [Knox and McElroy, 1984; Sarmiento and Toggweiler, 1984; Wenk and Siegenthaler, 1985] from the mechanism proposed here. This transition point is reached when the global rain ratio reaches (from to 0.084) and CO 2 is drawn down 30 ppm (dashed line, Figure 4b). An equivalent experiment with GCMs will very likely yield a larger drawdown. Matsumoto et al. [2002] showed that both their MOM-based Princeton model and the Hamburg LSG model [Archer et al., 2000] had about 3.5 times the response as simple box models to equivalent rain ratio forcings. The 30 ppm estimate provided here is subject to significant uncertainty, because we had to extrapolate results from our regional ocean model to the globe in order to obtain a global rain ratio change to a 5 C cooling. Then a simple box model was used to estimate the CO 2 drawdown. Under ideal situation, everything would be done consistently within a single model. We therefore argue that the 30 ppm drawdown, which represents a sizable fraction of the full ppm glacial-interglacial pco 2 amplitude, indicates the possibility that our mechanism has played an important role in explaining the full amplitude. It does not exclude other mechanisms from operating at the same time. In fact, additional mechanisms are necessary in light of the field s continuing difficulty explaining the full amplitude with a single mechanism. [30] We note two potential obstacles to an acceptance of our proposed mechanism. One is the apparently important role that mineral ballasts play in vertically transporting POC [Armstrong et al., 2002]. Taking a step further with the model of Armstrong et al. [2002], Klaas and Archer [2002] use a compilation of sediment trap data to argue that a substantial change in the rain ratio may be buffered against perturbations in surface production, given that CaCO 3 is the most important ballast in POC transport. The basic idea is that if there is more (or less) CaCO 3, then there should be more (less) POC export regardless of how much POC there is, and so their relative proportions (i.e., rain ratios) in the sinking aggregate would not change much. Some have even suggested that the rain ratio hypothesis may no longer be a viable explanation for glacial-interglacial pco 2 fluctuation [Ridgwell, 2003]. [31] The new ballast model is compelling, but we believe that any buffering of the rain ratio with regard to our proposed mechanism remains to be demonstrated. First, both Armstrong et al. [2002] and Klaas and Archer [2002] used sediment trap data from water depths below 1000 m in their analysis. Therefore the association between ballasts and POC is unclear in shallower waters including the surface mixed layer, where we expect to see the largest effect of temperature on POC decomposition (Figure 3). Second, as Armstrong et al. [2002] note, their data-based deep water rain ratios (or what they call the asymptotic transport ratios ) are statistically different in different regions. That is, the modern rain ratio is not buffered spatially. Third, the ballast model cannot explain data-based rain ratios in areas of high CaCO 3 export [Klaas and Archer, 2002]. Finally, it should be reminded that sinking aggre- 7of9

8 gates are often associated with not one type of mineral but multiple types, hence the word aggregate. In fact, it has been argued recently that it is the sinking POC that scavenges various suspended mineral particles and not the minerals scavenging POC [Passow, 2004; Passow and De la Rocha, 2006]. [32] The second potential obstacle is the possibility that enhanced biological pump by our mechanism is self limiting in nutrient-limited productivity regimes, because deeper POC remineralization would draw down nutrients from the surface and thus reduced primary and export production. We recognize this as a real possibility, which we are unable to evaluate quantitatively at this time. Productivity in many parts of the modern ocean is indeed limited by nutrients, where this self-limitation may apply. However, other parts like the high-nutrient low-chlorophyll regions are not limited by phosphate and nitrate. Iron fertilization experiments to date indicate that these regions may be iron limited [De Baar et al., 2005], but this limitation may have been relieved during peak glacial periods by increased dust flux [Wolff et al., 2006]. If so, these regions may not be susceptible to self-limitation. The extent to which self limitation in certain regions can cause the same for the world ocean is not obvious. Further complicating the issue is a finding by Palter et al. [2005] that horizontal advection of nutrients, in addition to the usual vertical mixing, may be important in supplying nutrients to the surface. Resolving the self-limitation issue requires more empirical studies that shed light on nutrients dynamics, which then would have to be incorporated into global ocean carbon cycle models for analysis. 5. Summary [33] We propose that marine microbial activity was suppressed by the low ocean temperatures during the last glacial period. This allowed proportionally more POC to escape degradation in the mixed layer and thereby strengthened the organic carbon pump. We believe that the idea is appealing because it is simple. It is based on almost universal observations that metabolic rates increase with temperature. It is almost abiotic, especially from a macroscopic view, in that tiny entities like bacteria are relatively simple and respond to temperature predictably. The mechanism does not call for undocumented changes in ocean circulation or biological production. Simulations with a global carbon cycle box model indicate a 30 ppm drawdown of atmospheric CO 2 by the proposed mechanism. This estimate is uncertain but suggests a potential role of the new mechanism in explaining part of the full glacialinterglacial atmospheric CO 2 fluctuations. Given the historical difficult in explaining the fluctuations with a single cause, we suggest that a set of multiple mechanisms were responsible and that the temperature-dependent POC degradation rate is one of them. [34] Under global warming, we would expect the opposite effect: higher temperatures would increase organic matter degradation particularly in the upper water column. With an inefficient organic carbon pump, this may lead to an increase in atmospheric pco 2. It is possible to have a positive feedback, as higher atmospheric pco 2 would increases ocean temperatures that in turn would raise the rain ratio and thus atmospheric pco 2. A global carbon cycle model with nutrient dynamics is needed to investigate properly this potentially important feedback. [35] Acknowledgments. We benefited greatly from constructive comments by anonymous reviewers and Associate Editor. K. M. acknowledges support by the University of Minnesota. T. H. was supported by the 21st Century COE program funded by MEXT. References Adams, J. M., H. Faure, L. Faure-Denard, J. M. McGlade, and F. I. Woodward (1990), Increases in terrestrial carbon storage from the Last Glacial Maximum to the present, Nature, 348, Anderson, L. A., and J. L. Sarmiento (1994), Redfield ratios of remineralization determined by nutrient data analysis, Global Biogeochem. Cycles, 8(1), Archer, D., and E. Maier-Reimer (1994), Effect of deep-sea sedimentary calcite preservation on atmospheric CO 2 concentration, Nature, 367, Archer, D., A. Winguth, D. Lea, and N. Mahowald (2000), What caused the glacial/interglacial atmospheric pco 2 cycles?, Rev. Geophys., 38(2), Armstrong, R., C. Lee, J. Hedges, S. Honjo, and S. G. Wakeham (2002), A new, mechanistic model for organic carbon fluxes in the ocean based on the quantitative association of POC with mineral ballasts, Deep Sea Res., Part 2, 49, Boyle, E. A. (1988), The role of vertical chemical fractionation in controlling late quaternary atmospheric carbon dioxide, J. Geophys. Res., 93, 15,701 15,714. Boyle, E. A. (1992), Cadmium and d 13 C paleochemical ocean distributions during the stage 2 glacial maximum, Annu. Rev. Earth Planet. Sci., 20, Broecker, W. S. (1982), Ocean chemistry during glacial time, Geochim. Cosmochim. Acta, 46, Broecker, W. S. (1986), Oxygen isotope constraints on surface ocean temperatures, Quat. Res., 26, CLIMAP Project Members (1981), Seasonal reconstructions of the Earth s surface at the Last Glacial Maximum, MC-36, Geol. Soc. of Am. Map Chart Serv., Boulder, Colo. De Baar, H., et al. (2005), Synthesis of iron fertilization experiments: From the Iron Age in the Age of Enlightenment, J. Geophys. Res., 110, C09S16, doi: /2004jc Duplessy, J.-C., N. J. Shackleton, R. G. Fairbanks, L. Labeyrie, D. Oppo, and N. Kallel (1988), Deepwater source variations during the last climatic cycle and their impact on the global deepwater circulation, Paleoceanography, 3(3), Edwards, N. T. (1975), Effects of temperature and moisture on carbon dioxide evolution in a mixed deciduous forest floor, Soil Sci. Soc. Am. Proc., 39, Eppley, R. W. (1972), Temperature and phytoplankton growth in the sea, Fish. Bull., 70, Fairbanks, R. G. (1989), A 17,000-year glacio-eustatic sea level record: Influence of glacial melting rates on the Younger Dryas event and deep-ocean circulation, Nature, 342, Farrell, J. W., and W. L. Prell (1989), Climatic change and CaCO 3 preservation: An 800,000 year bathymetric reconstruction from the central equatorial Pacific Ocean, Paleoceanography, 4(4), Francois, R., S. Honjo, R. Krishfield, and S. Manganini (2002), Factors controlling the flux of organic carbon to the bathypelagic zone of the ocean, Global Biogeochem. Cycles, 16(4), 1087, doi: / 2001GB Guilderson, T., R. G. Fairbanks, and J. L. Rubenstone (1994), Tropical temperature variations since 20,000 years ago: Modulating interhemispheric climate change, Science, 263, Hashioka, T., and Y. Yamanaka (2007), Seasonal and regional variations of phytoplankton groups by top-down and bottom-up controls obtained by a 3-D ecosystem model, Ecol. Modell., 202, 68 80, doi: / j.ecolmodel Jin, X., N. Gruber, J. Dunne, J. Sarmiento, and R. Armstrong (2006), Diagnosing the contribution of phytoplankton functional groups to the production and export of particulate organic carbon, CaCO 3, and opal from global nutrient and alkalinity distributions, Global Biogeochem. Cycles, 20, GB2015, doi: /2005gb Kaplan, J. O., I. C. Prentice, W. Knorr, and P. J. Valdes (2002), Modeling the dynamics of terrestrial carbon storage since the Last Glacial Maximum, Geophys. Res. Lett., 29(22), 2074, doi: /2002gl Key, R., A. Kozyr, C. Sabine, K. Lee, R. Wanninkhof, J. L. Bullister, R. A. Feeley, F. Millero, C. Mordy, and T.-H. Peng (2004), A global ocean 8of9

9 carbon climatology: Results from GLODAP, Global Biogeochem. Cycles, 18, GB4031, doi: /2004gb Klaas, C., and D. Archer (2002), Association of sinking organic matter with various types of mineral ballast in the deep sea: Implications for the rain ratio, Global Biogeochem. Cycles, 16(4), 1116, doi: / 2001GB Knox, F., and M. B. McElroy (1984), Changes in atmospheric CO 2 : Influence of the marine biota at high latitudes, J. Geophys. Res., 89, Laws, E. A., P. G. Falkowski, W. O. J. Smith, H. Ducklow, and J. J. McCarthy (2000), Temperature effects on export production in the open ocean, Global Biogeochem. Cycles, 14(4), Levitus, S., and T. P. Boyer (1994), World Ocean Atlas 1994, vol. 4, Temperature, NOAA Atlas NESDIS 4, 129 pp., Natl. Oceanic and Atmos. Admin., Silver Spring, Md. Levitus, S., M. E. Conkright, J. L. Reid, R. G. Najjar, and A. Mantyla (1993), Distribution of nitrate, phosphate and silicate in the world oceans, Prog. Oceanogr., 31, Li, Y. H., T. Takahashi, and W. S. Broecker (1969), Degree of saturation of CaCO 3 in the oceans, J. Geophys. Res., 74, Martin, J. H. (1990), Glacial-interglacial CO 2 change: The iron hypothesis, Paleoceanography, 5(1), Matsumoto, K., and J. Lynch-Stieglitz (1999), Similar glacial and Holocene deep water circulation inferred from southeast Pacific benthic foraminiferal carbon isotope composition, Paleoceanography, 14(2), Matsumoto, K., J. L. Sarmiento, and M. A. Brzezinski (2002), Silicic acid leakage from the Southern Ocean: A possible explanation for glacial atmospheric pco 2, Global Biogeochem. Cycles, 16(3), 1031, doi: /2001gb Monod, J. (1949), The growth of bacterial cultures, Annu. Rev. Microbiol., 3, Palter, J. B., M. S. Lozier, and R. T. Barber (2005), The effect of advection on the nutrient reservoir in the North Atlantic subtropical gyre, Nature, 437, Passow, U. (2004), Switching perspectives: Do mineral fluxes determine particulate organic carbon fluxes or vice versa?, Geochem. Geophys. Geosyst., 5, Q04002, doi: /2003gc Passow, U., and C. L. De la Rocha (2006), Accumulation of mineral ballast on organic aggregates, Global Biogeochem. Cycles, 20, GB1013, doi: /2005gb Petit, J. R., et al. (1999), Climate and atmospheric history of the past 420,000 years from Vostok ice core, Antarctica, Nature, 399, Ridgwell, A. J. (2003), An end to the rain ratio reign?, Geochem. Geophys. Geosyst., 4(6), 1051, doi: /2003gc Rostek, F., G. Ruhland, F. C. Bassinot, P. J. Muller, L. Labeyrle, Y. Lancelot, and E. Bard (1993), Alkenone paleotemperatures for the last 150,000 years for an equatorial Indian Ocean core, Nature, 364, Sarmiento, J. L., and J. R. Toggweiler (1984), A new model for the role of the oceans in determining atmospheric pco 2, Nature, 308, Sarmiento, J. L., J. Dunne, A. Gnanadesikan, R. Key, K. Matsumoto, and R. D. Slater (2002), A new estimate of the CaCO 3 to organic carbon export ratio, Global Biogeochem. Cycles, 16(4), 1107, doi: / 2002GB Siegenthaler, U., et al. (2005), Stable carbon cycle-climate relationship during the late Pleistocene, Science, 310, Sigman, D. M., and E. A. Boyle (2000), Glacial/interglacial variations in atmospheric carbon dioxide, Nature, 407, Sigman, D. M., D. C. McCorkle, and W. R. Martin (1998), The calcite lysocline as a constraint on glacial/interglacial low-latitude production changes, Global Biogeochem. Cycles, 12(3), Singh, J. S., and S. R. Gupta (1977), Plant decomposition and soil respiration in terrestrial ecosystems, Bot. Rev., 43, Stute, M., P. Schlosser, J. F. Clark, and W. S. Broecker (1992), Paleotemperatures in the southwestern United States derived from noble gases in ground water, Science, 256, Stute, M., M. Forster, H. Frischkorn, A. Serejo, J. F. Clark, P. Schlosser, W. S. Broecker, and G. Bonani (1995), Cooling of tropical Brazil (5 C) during the Last Glacial Maximum, Science, 269, Wenk, T., and U. Siegenthaler (1985), The high-latitude ocean as a control of atmospheric CO 2,inThe Carbon Cycle and Atmospheric CO 2 : Natural Variations Archean to Present, edited by E. T. Sundquist and W. S. Broecker, pp , AGU, Washington, D. C. Wolff, E. W., et al. (2006), Southern Ocean sea-ice extent, productivity and iron flux over the past eight glacial cycles, Nature, 440, Yamanaka, Y., and E. Tajika (1996), The role of the vertical fluxes of particulate organic matter and calcite in the oceanic carbon cycles: Studies using an ocean biogeochemical model, Global Biogeochem. Cycles, 10(2), K. Matsumoto, Department of Geology and Geophysics, University of Minnesota, 310 Pillsbury Drive, Minneapolis, MN 55455, USA. (katsumi@ umn.edu) T. Hashioka and Y. Yamanaka, Graduate School of Environmental Earth Sciences, Hokkaido University, N10, W5, Sapporo , Japan. (hashioka@ees.hokudai.ac.jp; galapen@ees.hokudai.ac.jp) 9of9