Particulate organic carbon (POC) sinking out of the sunlit euphotic zone at the surface of the ocean feeds the deep sea and alters the CO2 concentration of the atmosphere. Most of the sinking POC is reoxidized to dissolved inorganic carbon (DIC) before it hits the sea floor, but the mechanism for this is poorly understood. Here we develop a global model of the microbial loop in the aphotic zone based on new measurements of deep ocean bacterial metabolism. These together imply that a significant fraction of the decreasing POC flux with depth is converted to dissolved organic carbon (DOC) rather than directly to DIC as is commonly assumed, thereby providing the substrate for free-living bacteria in the deep ocean. The model suggests the existence of a substantial DOC-pool with a relatively fast turnover time in the deep sea. By implementing the microbial loop in a model of the global ocean circulation, we show that the observed gradient of DOC in the deep North Atlantic can be explained by the temperature dependence of bacterial metabolic activity in conjunction with the formation of deep-water at high latitudes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 DOC in the ocean is mainly of marine origin despite the relatively large fluxes of terrigenous derived organic carbon into the ocean from runoff (0.4 Pg/yr) and aeolian sources [Opsahl and Benner, 1997]. This also holds for the deep-sea DOC in the Greenland Sea where low concentration of lignin oxidation products and high δ13C-values of the DOC and the optical properties of the chromophoric dissolved organic matter all indicate a marine biogenic origin of the DOC [Opsahl et al., 1999; Stedmon and Markager, 2001]. The observed concentration of DOC below 1000 m depth is significantly higher in the northern North Atlantic (48–50 μM) than in the rest of the deep ocean (∼43 μM, except for the northern North Pacific where DOC<40 μM) [Hansell and Carlson, 1998a]. The deep-sea DOC distribution is also characterized by an old 14C-age of about 4000–6000 year, which indicates a very low turnover rate of the deep DOC pool [Bauer et al., 1992]. Increased primary production during the spring and summer season causes an accumulation of DOC in the euphotic zone where the surface concentration has been observed to increase by up to 10 μM in the subtropical Atlantic [Carlson et al., 1994] and by 20–30 μM in the Norwegian Sea [Børsheim and Myklestad, 1997]. Subsequently the accumulated DOC is remineralized or transported into the aphotic zone during the autumn and the winter seasons.
 These observations justify a separation of the marine DOC-pool into a “semilabile” and a “refractory” pool with a typical turnover timescale of months and millennia, respectively. This separation of the complex DOC pool into distinct fractions has been used to model the distribution of DOC in the water column. Yamanaka and Tajika  consider a semilabile and a refractory pool of DOC in a global ocean circulation model and they find that the DOC distribution in the upper part of the main thermocline (0–400 m depth) is well described by an e-folding time for the decay of the semilabile DOC of about 0.5 yr. Due to the short timescale for the decomposition of semilabile DOC into DIC they find that the deep-sea semilabile DOC concentration is only about 1 μM in the deep ocean with a maximum of about 3 μM in the deep northern Pacific. A similar timescale of less than 1 yr for the semilabile DOC is determined in a study of the global carbon cycle where the role of the plankton dynamics in the euphotic zone is explicitly resolved [Six and Maier-Reimer, 1996]. Archer et al.  estimate a timescale of about 6 months for the decay of newly produced DOC in the equatorial Pacific which is transported into the seasonal thermocline. In a one-dimensional study of the cycling of labile, semilabile and refractory DOC, the transport by vertical advection and diffusion of DOC was shown to be important in the upper part of the main thermocline whereas the DOC released from the sinking flux of POC was shown to be important for providing the substrate for the bacteria in the deeper part of the water column [Anderson and Williams, 1999]. In accordance with the studies above, Anderson and Williams  estimate a turnover timescale of 0.4 year for the semilabile pool of DOC.
 The deep-sea thermohaline circulation in the North Atlantic is dominated by the southward transport of North Atlantic Deep Water (NADW) which originate in the Greenland Sea area [Reid and Lynn, 1971]. Interactions between the thermohaline circulation and the DOC-pool could explain the observed deep-sea gradient of DOC in the North Atlantic, but would imply a timescale for the degradation of DOC of several decades, as this is the timescale for the deep circulation in the eastern North Atlantic [Smethie et al., 2000]. This is much longer than the estimated short timescale of about 6–12 months for the decay of the semilabile fraction of DOC in the tropical and subtropical seasonal thermocline, and significantly less than the timescale for the old “refractory” deep-sea DOC pool. Therefore these models can not explain the observed deep-sea DOC distribution in the North Atlantic reported by Hansell and Carlson [1998a].
 Here we show that a simple microbial loop based on the bacterial biomass and the temperature and substrate dependent bacterial metabolic activity can account for the different turnover timescales ranging from months to millennia and that the influence of the bacterial activity is reflected in the deep-sea DOC distribution, in particular in areas close to deep-water formation regions. The microbial loop is based on new measurements from the Greenland Sea which provide evidence for the importance of the particle-associated transport of DOC in the water column. Our results supports the view that a major part of the organic carbon cycle in the ocean may be characterized as: POC → DOC → Bacteria → DIC [Cho and Azam, 1988; Nagata et al., 2000].
2. Measurements of Labile DOC, Bacterial Biomass and Production
 We measured the concentration of labile organic carbon and the production and biomass of bacterioplankton during two cruises to the Greenland Sea in June and August 1999 (see Appendix A). Labile DOC (DOCl) is here defined as the amount of DOC which can be utilized by bacteria within 1 week of incubation at 20°C [Søndergaard and Middelboe, 1995]. DOCl is not an estimate of the total pool of biodegradable DOC (DOCb), but provides an estimate of the fraction of DOCb that is readily utilized by bacteria. Labile DOC was found to be present at significant concentrations of 0.04–0.76 μM below 1000 m depth. We cannot exclude that the measured concentration of labile DOC may be slightly overestimated due to DOC contamination from water filtration and handling of incubation bottles (M. Middelboe and C. Lundsgaard, Microbial activity in the Greenland Sea: The role of DOC lability, mineral nutrients and temperature, submitted to Aquatic Microbial Ecology, 2002). However, the fact that we found indications for a well-functioning microbial loop in the deep sea (i.e., significant bacterial production, Table 1) and significant abundance of both viruses and heterotrophic protozoa, supported the observed presence of a pool of labile DOC in the deep waters. The presence of a labile DOC-pool with a turnover time of weeks to months can only be explained by transport via fast sinking (of the order 100 m d−1) particulate organic matter, as any advective or diffusive transport of surface derived DOC to this depth would require several months (see section 5). The bacterial biomass decreases from 11–20 mg C m−3 in the upper 100 m to about 4 mg C m−3 at 500 m depth and the Production/Biomass (P/B)-ratio decreases correspondingly (Table 1).
Table 1. Bacterial Biomass, Production, P/B-Ratios and Labile DOC in the Greenland Seaa
B, mgC m−3
P, 10−5 gC m−3 d−1
μ, 10−2 d−1
T, temperature; N, number of samples; B, bacterial biomass and standard error (SE); P, bacterial production and SE; μ, specific growth rate and SE; n, number of DOCl samples; 〈DOCl〉, range of DOCl-values; DOCl, mean DOCl-values and SE. P and B were measured in the Greenland Sea in June and August 1999. DOCl was measured in August 1999.
20.80 ± 1.3
124.57 ± 9.3
9.25 ± 0.79
5.39 ± 0.86
11.47 ± 0.7
30.11 ± 5.0
3.38 ± 0.46
1.45 ± 0.25
4.35 ± 0.4
3.31 ± 0.8
1.10 ± 0.29
0.66 ± 0.21
3.75 ± 0.6
1.76 ± 0.4
0.76 ± 0.18
0.44 ± 0.10
3.24 ± 0.5
1.16 ± 0.6
0.80 ± 0.44
0.27 ± 0.09
3.35 ± 0.6
1.75 ± 0.5
0.93 ± 0.37
0.07 ± 0.04
1.22 ± 0.2
1.52 ± 0.6
1.15 ± 0.60
3. Microbial Loop
 Together with earlier reported observations of the deep-ocean DOC distribution [Hansell and Carlson, 1998a] the observed labile deep-sea DOC-pool in the Greenland Sea motivated the formulation of a model for a microbial loop working in the deep ocean based on the bacterial uptake of particle-transported DOC (Figure 1). The model assumes that all of the sinking POC is dissolved to biodegradable DOCb through hydrolytic enzymatic activity [Smith et al., 1992; Middelboe et al., 1995], viral lysis of particle-associated bacteria [Proctor and Fuhrman, 1991] or by grazing zooplankton. The dissolution of POC provides the substrate for the free-living bacteria in the deep ocean [Cho and Azam, 1988; Nagata et al., 2000]. Furthermore we assume that the subsequent remineralization of DOCb ultimately is controlled by the bacterial activity. The deep sea is resource-limited with respect to the supply of organic carbon [Billen et al., 1990; Dufour and Torréton, 1996; Nagata et al., 2000]; therefore we neglect the effects of viruses, bacterivores and other grazers in controlling the uptake of DOCb in our model of the mesopelagic zone. Bacterial carbon uptake is determined from our observations of bacterial growth rates and bacterial growth efficiency, and bacterial mortality is constrained by previously reported observations of bacterial abundance and the dissolution of POC to DOCb in the deep-ocean (see below).
3.1. Model Description
 The microbial loop working in the water column below the euphotic zone is driven by the dissolution (D) of the sinking POC to DOCb (Figure 1). The growth of the bacterial biomass (B) is regulated by the bacterial carbon uptake (U) and the mortality (M). The bacterial mortality contributes to the DOCb and the refractory DOC (DOCr) pools with the fractions ϵb and ϵr respectively. The remaining part of the mortality loss and a fraction of the bacterial carbon uptake determined by the yield (ω) are remineralized to DIC. DOCr is transformed to DOCb at the surface through photo-dissociation (L). The conservation equations for the different carbon pools in the microbial loop are given by:
In addition to the sinks and sources formulated in equations (1)–(4) the contribution from advective and diffusive transports also affect the distribution of the four state variables. The influence from these transports on shaping the global deep-sea DOC distribution is studied in section 4.
3.2. Bacterial Uptake
 The bacterial production (P), related to the bacterial uptake via the yield, can be described in terms of the specific growth rate (μ) and the bacterial biomass:
The temperature dependence of the bacterial metabolic activity can be characterized as:
where T0 = 0°C is a reference temperature and μ0 is a temperature normalized specific growth rate which can depend on other limiting factors, as for example the substrate availability. The temperature dependence of bacterial activity is found to be characterized by a Q10-factor within the range of 2–3 in large areas of the ocean [Pomeroy and Wiebe, 2001], and our growth experiments with mixed bacterial samples from the Greenland Sea gives a value of Q10 = 3.0 (M. Middelboe and C. Lundsgaard, Microbial activity in the Greenland Sea: The role of DOC lability, mineral nutrients and temperature, submitted to Aquatic Microbial Ecology, 2002). We use this value as representative for the temperature dependence of the bacterial uptake in the microbial loop. The temperature dependence of the bacterial carbon uptake in the model is based on our observations in the Greenland Sea. This temperature dependence has been found in previous studies of the bacterial activity in cold waters [Rivkin et al., 1996] and the bacterial production has also been shown to correlate with both temperature and chlorophyll in the surface layers of the global ocean [Rivkin and Legendre, 2001]. From these experiments we also determine a yield factor of ω = 0.33 which lies within the range of earlier reported values for the marine environment [del Giorgio and Cole, 1998]. We apply this value in the microbial loop.
 The specific growth rate decreases from 0.03 d−1 at 100 m depth to a low value of about 0.01 d−1 below 500 m (Table 1). The relative small vertical temperature gradient in the Greenland Sea can not account for this decrease in μ, and therefore we suggest that this change reflects a decrease in the substrate value of the released DOC. To account for this increase of the specific growth rate in the upper ocean we model μ0 as:
where α is a substrate affinity of the bacterial uptake of DOCb. In general, the bacterial uptake can be modeled by a Monod expression involving a maximum uptake rate and a half saturation constant, but as bacterial uptake in the aphotic zone in the watercolumn is resource limited this expression can be linearized assuming a low concentration of DOCb. The bacterial uptake can then finally be described as:
The yield and temperature dependence of the bacterial uptake are well studied components of the bacterial metabolism, whereas the substrate affinity of deep-sea DOC is largely unknown. Therefore we consider the sensitivity of the microbial loop to this parameter below.
3.3. Bacterial Mortality
 The bacterial abundance (N) below 1000 m depth in the equatorial, subtropical and subarctic areas in the North Pacific is well described as:
where N1000 is the abundance at zd = −1000 m depth and z is the vertical coordinate [Nagata et al., 2000]. A uniform value of a = −0.9 characterizes the abundance in all these different locations, and this finding has implications for modeling the bacterial mortality in equations (1)–(4).
 The parameterization of the bacterial mortality in the deep ocean is not well constrained so we assume that the general form of this term can be described by:
where m is a mortality constant and n is the power dependence of M on the bacterial biomass. The steady state solution of equations (1) and (2), which can be solved without considering the DIC and the DOCr pools when the contribution from photodissociation (L) in the surface layer is neglected, is then given by:
As the source term for the microbial loop in the aphotic zone (D) we use the convergence of the vertical flux of POC found from sediment trap studies:
where F0 is the particulate flux at the bottom of the surface mixed layer at the depth z0, and we use a value of b = −0.86 which has been found to fit observations in large areas of the ocean [Martin et al., 1987]. Together with the observed power dependence of bacterial abundance from Nagata et al. , this parameterization of D constrains the mortality term to be related to the bacterial biomass with n = 2 resulting in a power dependence of the bacterial biomass given by: , close to the observed value of a = −0.9.
 The steady state solution obtained by using the estimated global mean value of new production (F0) of 1.53 mol C m−2 yr−1 [Martin et al., 1987] in equation (12) is fitted to the mean bacterial biomass distribution in the deep ocean observed by Nagata et al. . Using a representative value of N1000 = 6.0 · 1010 cells m−3 and assuming a carbon content of ν b = 20 fg C cell−1 the best fit solution for the bacterial biomass is obtained, by minimizing the residual
giving a best fit value of m = 0.41 M−1 s−1 (Figure 2).
 The bacterial mortality is regulated both by viral lysis and grazing from higher trophic levels and these terms are assumed to amount equally to the mortality in equation (10) [Fuhrman and Noble, 1995]. We furthermore estimate that 95% of the bacterial mortality due to viral lysis is recycled as DOCb whereas the uptake by bacterivores and the subsequent carbon flow through higher trophic levels only results in a recycling of 45% of the bacterial biomass as DOCb. This gives a value of the recycling coefficient ϵb = 0.5 (0.95 + 0.45) = 0.7 in the microbial loop, whereas the remaining part of the mortality is respired as DIC or degraded into the refractory DOC-pool [Ogawa et al., 2001] through the term ϵr, which is constrained by the observed DOC-Δ14C (see section 4).
3.4. Steady State Solutions
 The steady state solution of the bacterial biomass and DOCb in the microbial loop (equations (1) and (2)) can be calculated for a given vertical temperature distribution and using the dissolution profile of POC given in equation (12) (Figure 3). Here we use a value for the substrate affinity of α = 0.6·10−3 (μM d)− 1. The sensitivity of the temperature dependence of the bacterial biomass is illustrated by solving the equations for two different temperature profiles which represents the cold conditions at high latitudes and the global averaged temperature conditions in the watercolumn, respectively. The deep-sea DOCb concentration of about 1 μM increases towards the values near the surface of about 6 μM and 29 μM in the warm and cold case, respectively.
 The turnover time of DOCb increases from 3.0 yr at 100 m depth to 68 yr at 3800 m depth when the temperature is 0°C in the whole water column, whereas the solution for a case using the global mean vertical temperature distribution [Levitus et al., 1994] results in a turnover time for DOCb of 0.6 yr at 100 m depth increasing to 58 yr at 3800 m depth, showing the significance of the temperature dependence of the bacterial carbon demand in the upper part of the water column. These long timescales in the deep ocean imply that the DOCb-pool can interact with the large scale ocean circulation, in particular in cold areas at high latitudes where the mixing between surface and deep waters is dominated by convection during winter time. Adding vertical mixing to the solutions would result in higher deep-sea DOCb levels for the cold solution and therefore this temperature dependence could explain the observed high deep-sea concentration of DOC in the North Atlantic.
 The steady state solution for B and P are independent of the temperature as shown in equation (11). The solutions for B, P and μ fall within the ranges reported by Nagata et al. , whereas our measured values from the Greenland Sea are about an order of magnitude higher (Figure 3d–3f). This discrepancy could be explained by a higher level of DOCb in the deep ocean at high latitudes, as shown in section 4, because this would result in higher values of B, P and μ.
 The steady state solution of DOCb depends inversely on the substrate affinity (α) as:
For a substrate affinity of about 10−3 (μM)−1 the mean concentration of DOCb in the watercolumn between 100 and 3800 m depth is 1.7 μM and 2.7 μM for the warm and cold case, respectively (Figure 4). For low values of α of about 10−4 (μM)−1 the total pool of DOCb for the two temperature distributions increase to more than 10 μM whereas the total DOCb-pool is reduced to about 1 μM when α approaches 2·10−3 (μM)−1. This sensitivity of the total DOCb pool to the value of α is used to constrain α by the modeled global distribution of DOCb, as discussed below.
4. Global Model Results
 To examine the possible interaction between the deep-sea DOC-pool and the large-scale circulation, the microbial loop was implemented in a general ocean circulation model HAMOCC2 [Maier-Reimer, 1993; Archer et al., 2000] which has a resolution of about 3.5° latitude × 3.5° longitude and 11 vertical layers with increasingly resolution from the bottom towards the 50 m thick surface layer. The tracer field is integrated with a time step of 1 year and transported by the annual mean circulation. New production is regulated by surface nutrients, light and iron. A fraction of 17% of new production accumulates in the surface ocean as DOCb and is subsequently mixed out of the euphotic zone [Hansell and Carlson, 1998b] and the rest is exported as sinking POC. The bacterial biomass is assumed stationary as the turnover time for the bacterial carbon pool is about 1 year in the model, in accordance with estimates of the turnover time of the deep-sea bacterial biomass (Dufour and Torréton, 1996). The equations in the microbial loop are integrated within each time step with a fourth order Runge-Kutta algorithm. The model is integrated for 20,000 years to reach deep ocean DOC-Δ14C steady state.
4.1. DOC and Bacterial Biomass
 The global averaged distribution of DOC increases from a deep value of about 43 μM below 3000 m to about 48 μM at the surface, whereas the refractory component is relatively constant throughout the water column with a value of about 41 μM (Figure 5a). At the surface the DOCr decreases due to the transformation of DOCr to DOCb through photodissociation. The vertical distribution of DOCb increases from a deep value of about 1 μM below 3000 m to a surface value of about 6 μM. The sources of DOCb in the surface layer only includes a small fraction of the primary production (∼17% of NP) and this is the reason for the relatively low DOCb concentration in the mixed layer (Figure 5b). The majority of this surface DOC pool has a residence time of less than one year and therefore it will not influence the deep sea DOC distribution but only the upper few hundred meters of the water column.
 The globally averaged bacterial carbon biomass increases from a low concentration of about 3·10−2 μM below 4000 m depth to a surface value of about 0.4 μM at 100 m depth (Figure 5c). This vertical distribution reflects the distribution of DOCb which is the substrate for the bacterial biomass. The global averaged value is in good accordance with the global mean vertical reference profile (Figure 5c, dashed line) based on the observations of Nagata et al.  and described by equation (9). The observed range of the bacterial abundance in the North Pacific ranges within a factor of ten, including the influence from seasonal changes, so the global averaged model results of the bacterial biomass are well within the observed values of Nagata et al. .
4.2. DOC-Δ14C Distribution
 The observed deep ocean DOC concentration and the DOC-Δ14C place a strong constraint on the fraction of the bacterial mortality (ϵr) which contributes to DOCr. The only sink of refractory DOC in the model is assumed to be due to photodissociation in the sunlit mixed layer where DOCr is dissociated into DOCb [Mopper et al., 1991; Benner and Biddanda, 1998; Cherrier et al., 1999]. The photodissociation is parameterized as a first-order decay process with a decay constant of 150 years [Yamanaka and Tajika, 1997], scaled with an incident light factor which also regulates new production [Bacastow and Maier-Reimer, 1990], thereby taking the latitudinal dependence of solar insolation into account [Moran and Zepp, 1997]. A value for ϵr of 2.4% leads to a global mean deep-ocean value below 500 m depth of 43 ± 1.4 μM for DOC and −466 ± 19 ppt for DOC-Δ14C. The latter is within the observed range between −450 ppt and −500 ppt [Bauer et al., 1992].
 The global averaged DOC-Δ14C profile has a minimum value of about −480 ppt below 3000 m depth to a surface value of −410 ppt (Figure 6a). The refractory pool has a deep minimum value of about −490 ppt whereas the influence from newly produced organic matter in the main thermocline causes the DOCr-Δ14C to increase to about −460 ppt at the bottom of the surface layer. The gradual release of DOC from the sinking flux of POC has a 14C-value which is determined by the DIC-Δ14C value in the surface layer because this carbon pool is about two orders of magnitude larger than the DOC-pool. This influence causes the DOCb-Δ14C to be relatively constant throughout the water column with a value of about −60 ppt, which is close to the global averaged DIC-Δ14C value in the surface layer (Figure 6b). The surface values of DOCb-Δ14C are relatively old because of the small amount of photodissociated DOCr which is transformed into DOCb. Both old and newly produced DOC are incorporated into the bacterial biomass [Cherrier et al., 1999]. So, in conclusion, the model is able to reproduce the observed 14C age of bulk DOC in the deep ocean.
4.3. Global Deep-Sea DOC Distribution
 The mean DOCb distribution below 1000 m depth in the model has the highest values of about 7 μM located in the Nordic Sea area where convective mixing is strongest in the model, and a general low concentration of DOCb in the rest of the ocean between 1 and 3 μM (Figure 7a). The large new production in the strong upwelling zone in the eastern equatorial Pacific leads to a high deep-sea DOCb concentration of about 4 μM. The vertical mean concentration of DOC below 1000 m in the Atlantic shows significantly elevated levels in the northern North Atlantic of 48.0 ± 1.1 μM (Figure 7b), within the errorbars of earlier reported values of 48.1 ± 0.3 μM [Hansell and Carlson, 1998a]. Throughout the rest of the ocean the modeled deep-sea DOC concentration has a relatively constant level of about 43 μM. The DOCr concentration is almost constant with a mean value of 41.2 μM due to its long residence time. In the northern North Atlantic the deep DOCr-levels are slightly lower due to the large vertical mixing with low photo-dissociated surface DOCr. This model cannot explain the low values of about 39 μM and 34 μM (not shown) observed in the subtropical and northern North Pacific respectively [Hansell and Carlson, 1998a; Druffel and Bauer, 2000]. Additional sinks or sources would have to be included to account for the low deep-sea DOC concentrations in these areas (see section 5).
 The concentration of the bacterial biomass and the bacterial production is significantly higher in the northern North Atlantic than in the North Pacific (Figure 8). The modeled B and P distributions in the North Pacific falls within the ranges reported by Nagata et al. . The observations in the Greenland Sea of B and P was made during the productive summer season and therefore the annual averaged values would be lower by a factor of 2 or more, in consistency with the model results. The higher deep-sea specific growth rate in the North Atlantic can not be explained by the about 3°C lower deep-sea temperature there but is due to the larger concentration of DOCb.
 The bacterial carbon uptake depends upon the quality of the available substrate which is regulated by the substrate affinity α in the microbial loop. A value of α = 0.6·10−3 (μM d)−1 is used here and gives a DOCb concentration in the North Atlantic in agreement with observations. The DOCb distribution is sensitive to lower values of α (compare Figure 4), such that a 50% lower value of α would result in a DOCb pool about twice as large, whereas the sensitivity is less pronounced for higher values of α. An integration of the general circulation model was carried out with a doubling of α and this resulted in a decrease of the global deep-sea DOCb concentration of about 1 μM and a drop of the elevated northern North Atlantic concentration to about 4 μM. Few long-term incubation experiments with marine deep-sea DOC samples have been carried out [Hopkinson et al., 1997], but such experiments could provide information about the substrate dependent bacterial uptake and could constrain the unknown parameter α.
 The significance of the temperature dependence and circulation for the mean deep-sea DOCb distribution below 500 m is reflected in the high correlation (r∼0.74) between new production (NP) and the DOCb when the northern North Atlantic is disregarded compared to the case when this area is included (r∼0.54) in the analysis of the model data (Figure 9). To illustrate the influence of the temperature on the DOCb distribution two integrations were carried out with the temperature in the microbial loop fixed to the global mean vertical temperature distribution (as shown in Figure 3a, solid line) and with the temperature fixed to 0°C (Figure 3a, dashed line). In the case with the global mean vertical temperature profile, the global DOCb concentration correlates well with NP (r∼0.78) whereas the extreme case with a constant temperature of 0°C results in high concentrations of DOCb in the northern North Atlantic and thereby a low correlation between DOCb and NP (r∼0.46).
 The implications of the model results can be summarized as: Low surface temperatures at high latitudes cause a relatively long residence time of DOCb in the upper part of the water column (0–1000 m). Large vertical mixing and the formation of deepwater in the northern North Atlantic transport DOCb deeper into the water column and this leads to elevated concentrations of DOCb in the North Atlantic region. The high deep-sea concentration of DOCb also causes a higher concentration of B and a larger P than seen elsewhere in the ocean.
 The focus of this study is the biodegradable part of the DOC-pool in the aphotic zone, which we estimate to account for 2–15% of the total DOC in the water column with the highest fraction in the subpolar areas and in areas with a high export production. DOC released from sinking aggregates through bacterial mediated enzymatic hydrolysis is an important substrate source for the freeliving bacteria in the ocean interior [Cho and Azam, 1988; Simon et al., 1992; Nagata et al., 2000]. Here we present observations of labile DOC from depths below 1400 m in the Greenland Sea which provide direct evidence for the particle associated transport of DOC. The incubation experiments were carried out during one week at 20°C. Using the temperature dependence of the bacterial uptake, characterized by Q10=3.0, this would correspond to the bacterial uptake during 9 weeks at 0°C. A possible source of this deep-sea labile DOC-pool could be the transport associated with convective activity which reaches 1000–2000 m depth in the late winter [Lherminier et al., 1999]. However, the measured turnover rates of DOCl imply that any labile DOC transported to 1000–2000 m depth by wintertime convection would have been consumed by bacteria in early spring, and therefore not be present below 1400 m in August. Therefore it is unlikely that the labile DOC is transported to intermediate depth levels by convective processes. Alternatively, the DOC could be transported to depth through diffusive or advective processes, but even the relatively high vertical diffusion observed in the central Greenland Sea of about 10−4 m2 s−1 [Watson et al., 1999] would only lead to vertical transports of less than 100 m during a 10-week period. The observation of deep-sea labile DOC can only be explained as DOC released from the sinking flux of POC.
 Convective activity during the cold winter season deepens the mixed layer in the Greenland Sea, and therefore the accumulation of DOCb during the spring and summer seasons in the surface layer can potentially contribute to the total transport of DOCb deep into the water column. The surface accumulation of DOC can account for a significant fraction of the new production in the area [Noji et al., 1999], but the inverse correlation between the seasonal changes of DIC in the Greenland Sea and the corresponding changes of DOC in the Norwegian Sea indicates that most of the surface derived DOC is remineralized in the surface layer [Miller et al., 1999]. From measurements of the bacterial production the estimated turnover time of the organic carbon in the surface layer (0–50 m) ranges within 8–91 days [Børsheim, 2000], and this also supports the view that surface derived DOC only makes a minor contribution to the deep-sea DOC in the area.
 Due to the limited data base which is currently available to establish parameterizations of the processes in the microbial loop which are valid on a global scale, the model was intentionally kept as simple as possible with a minimum number of free parameters. The critical parameters in the model are related to the sources and sinks of the bacterial biomass. We assume that 100% of the POC flux enters the DOCb-pool, but this might be an overestimate of the actual source of DOCb. Simon et al.  estimated that the bacterial carbon uptake in the depth range of 80–600 m in the subarctic Pacific accounted for 41–172% of the POC flux measured in sediment traps, and Nagata et al.  found that the bacterial carbon uptake accounted for 31–153% of the estimated POC flux in the deep North Pacific below 1000 m depth. These studies show that the sinking POC flux provides the substrate for the free-living bacterial biomass, but also they indicate that the assumption of a 100% release of POC to DOCb could be an overestimation. Our results are not so sensitive to the fraction of the sinking POC-flux which enters the microbial loop because the mortality coefficient (m) in equation (10) is set by the level of the source term as seen from the steady state solution in equation (11) and the corresponding best fit solution defined from equation (13). A smaller fraction of the POC flux which enters the DOCb-pool would cause a correspondingly lower value of m.
 Recent observations from the subtropical north Pacific, show that archaea makes up a significant fraction of the microbial biomass below 1000 m depth [Karner et al., 2001]. The archaeal biomass is included in the observed estimates of the microbial biomass (which is simply referred to as the “bacterial biomass” above), but few experiments has been carried out on the archeal activity. Observations from the Pacific and the Mediterranean Sea show that archea coexist with the bacteria and that they incorporate dissolved amino acids [Ouverney and Fuhrman, 2000]. These findings support the validity of our estimates of microbial production based on the incorporation of 14C-leucine (Appendix A), but it is still unclear whether the archaeal metabolic activity can be described in terms of the parameterizations used for the bacterial activity.
 The nature of an additional deep-sea sink is presently unknown, but it could be a very slow decomposition of the DOCr pool during its deep-sea transit from the Atlantic to the Pacific, for example due to hydrolysis, bacterial consumption or adsorption onto particulate material [Druffel et al., 1996]. The DOC-Δ14C age difference between the Sargasso Sea and the North Pacific is about 2000 years which is longer than the corresponding DIC-Δ14C age difference of about 1500 years [Druffel et al., 1992]. This difference between the age of the DIC and the DOC pool could in turn be explained if a slow deep-sea sink of the DOCr preferentially removed the “youngest” part of the DOCr pool. Such an “age”-fractionation has been observed to occur on much shorter timescales on riverine DOC transported into the ocean [Raymond and Bauer, 2001].
 The temperature dependence of the bacterial uptake also makes the oceanic DOC-pool sensitive to major changes in the thermal distribution in the ocean, as could have prevailed during other climatic regimes as for example the cold glacial periods. Two integrations were carried out where the temperature in the Q10 factor of the bacterial uptake were subtracted and added by 4°C, respectively. The global DOCb concentration in the standard case is 2.8 μM and this increases (decreases) to 4.9 μM (2.4 μM) when 4°C is subtracted (added) corresponding to a change of 33 PgC (−6 PgC) of the global DOCb inventory.
 In conclusion, a microbial loop based on the bacterial biomass in the aphotic zone can explain the elevated deep-sea concentration observed in the North Atlantic. Previous treatments of DOC [Yamanaka and Tajika, 1997; Six and Maier-Reimer, 1996; Murnane et al., 1999] impose spatially uniform degradation timescales for DOC ranging from 0.5 to 11 years. In contrast, our results argue that the temperature-dependent, bacterially mediated DOC degradation has a significant impact on the distribution of DOC in the deep-sea. Through the workings of the bacterial biomass, the biological pump in the ocean achieves more degrees of freedom. Temperature dependent bacterial metabolism thus introduces a new degree of freedom to the global carbon cycle.
Appendix A.: Measurements of DOCl, B, and P
 DOCl was determined in regrowth experiments as described by Søndergaard and Middelboe . In principle, all organisms except an inoculum of bacteria were removed from a water sample and the bacteria were allowed to grow to reach the carrying capacity with respect to organic carbon in the sample. DOCl was estimated from the increase in bacterial biomass during incubation using an empirically determined growth efficiency of 0.24. The bacterial biomass was calculated from the bacterial abundance (measured by flow cytometry) using an empirically determined conversion factor of 23 fgC cell−1. Bacterial growth efficiency and cell carbon content was determined in separate experiments. Here bacterial respiration was measured as oxygen consumption (using Winkler titration) and bacterial carbon production was measured as the increase in POC collected on glass-fiber filters and measured as CO2 after combustion at 650°C. Bacterial production was estimated from the incorporation of 14C-leucine [Chin-Leo and Kirchmann, 1988] using empirical conversion factors of 2060 gC mol−1 leucine in June and 1280 gC mol−1 leucine in August. Average values of leucine based data are presented here. More details about methods are given by M. Middelboe and C. Lundsgaard (Microbial activity in the Greenland Sea: The role of DOC lability, mineral nutrients and temperature, submitted to Aquatic Microbial Ecology, 2002).
 This work was funded by the Danish Natural Science Research Council, the Carlsberg Foundation and from the EU-project ORFOIS (contract EVK2-CT-2001-00100).