Regional variability in the vertical flux of particulate organic carbon in the ocean interior

Abstract

[1] Carbon transport within sinking biogenic matter in the ocean contributes to the uptake of CO₂ from the atmosphere. Here we assess the extent to which particulate organic carbon (POC) transport to the ocean’s interior can be predicted from primary production or export flux. Relationships between POC flux and depth are generally described by a uniform power law or rational decrease with depth, scaled to new or total primary production of POC. While these parameterizations of flux are used in most quantitative biogeochemical models, they are based on data sets from a limited geographic and depth range. We examine these relationships through a review of parameters derived from ¹⁴C uptake experiments, regional remote sensing, ²³⁴Th studies, nitrogen balances, and sediment trap records. Ocean regions considered include sites studied by the Joint Global Ocean Flux Study, Hawaii Ocean Time-series, and Bermuda Atlantic Time-series Study programs and involve observed and radiochemically corrected flux to depth. We demonstrate regional variability in the efficiency of the biological pump to transport organic carbon from surface waters to the ocean’s interior. Commonly applied flux relationships, while representative of some areas of the ocean, generally overestimate flux to depth. We estimate that the fraction of carbon transported as POC to depths greater than 1.5 km ranges between 0.10 and 8.8% (1.1% average) of primary production and between 0.28 and 30% (5.7% average) of export from the base of the euphotic zone. We develop empirical parameterizations of flux to depth using region-specific constants. Using a one-dimensional ocean model, we predict that the residence time of biogenic carbon may vary by up to 2 orders of magnitude depending on the regional efficiency of export and vertical transport.

1. Introduction

[2] The flux of organic matter from surface waters to the deep ocean has a direct influence on the partitioning of CO₂ between the ocean and atmosphere. This biological carbon pump includes three interrelated processes: primary production, export, and flux to depth. Primary production in surface waters, fueled by sunlight and nutrients, converts dissolved inorganic carbon to particulate organic matter (POM) and enhances the ocean’s ability to take up CO₂ from the atmosphere. POM is exported from surface waters through a variety of processes, including consumption and repackaging by zooplankton and fish into fecal pellets, particle aggregation, and zooplankton migration and excretion at depth. As exported POM sinks within the ocean’s interior, mesopelagic bacteria and zooplankton oxidize the organic components back into their dissolved inorganic constituents. A small portion of sinking organic matter reaches the seafloor, where it is either remineralized, supporting benthic biological activity, or buried in the geologic record.

[3] Determining the fate of organic matter sinking below the photic zone is crucial to understanding the role of the biological pump in the global carbon cycle and may have a role in determining oceanic sequestration of anthropogenic CO₂, especially in the context of changing ocean circulation [Kurz and Maier-Reimer, 1993; Sarmiento et al., 1998; Pahlow and Riebesell, 2000]. In general, the greater the depth at which sinking organic carbon is remineralized, the longer time it takes to return to the photic zone as dissolved CO₂, where it may reenter atmospheric carbon cycle. Organic carbon which reaches the deep ocean is entrained in water masses that have longer flow pathways back to the surface and smaller advective water velocities than in the upper ocean. The ventilation of disphotic and aphotic ocean waters occurs on timescales ranging from annual to hundreds years in the upper ocean (∼0.1–1 km) to up to 1000 years in the deep ocean (>1.5 km). Thus, to describe the residence time of biogenic carbon in the ocean, the depth of remineralization must be known.

[4] The intensity of this biological pump depends on several variables, including the level of photosynthetic production, the amount of zooplankton grazing, and the degree of oxidative remineralization at different depths in the water column. Furthermore, photosynthetic production is controlled by the availability of light, nutrients, and trace metals, as well as phytoplankton speciation, temperature, and grazing. At steady state, nutrients removed from surface waters in the form of descending particulate matter are balanced by the upward advective and diffusive supply of dissolved nutrients. The upwelled nutrients support new production [Eppley and Peterson, 1979] in surface waters and are supplied by the remineralization of sinking organic matter.

[5] In this paper we reevaluate relationships between primary production, export, and flux to depth. We present evidence for variability in the vertical flux of organic matter within the ocean that is not accounted for by currently applied flux algorithms. We develop empirical algorithms with varied constants to predict flux to depth in different ocean regions.

2. Previous Flux to Depth Relationships

[6] The two most commonly applied descriptions of oceanic particulate organic carbon (POC) flux are that of Suess [1980] (equation (1)) [see also Six and Maier-Reimer, 1996; Dymond et al., 1997; Paerl, 1997; Soltwedel, 1997; Petsch and Berner, 1998; Sigman et al., 1998; Tyrell, 1999] and that of Martin et al. [1987] (equation (2)) [see also Sarmiento and Le Quere, 1996; Brewer et al., 1997; Emerson et al., 1997; Druffel et al., 1998; Lee et al., 1998; Sarmiento et al., 1998; Hedges et al., 1999; Buesseler et al., 2000; Fischer et al., 2000].

Flux to depth C_flux(z) is described as a function of the primary production of organic carbon in surface waters C_prod or the export of organic carbon C_export from the base of the photic zone z₀, scaled to depth below the sea surface z.

[7] The Suess [1980] rational equation was determined from ¹⁴C-based primary production and sediment trap flux measurements, collected from the subtropical eastern Pacific and northwestern Atlantic at depths between 50 and 5400 m (Figures 1 and 2). The specific relationship between primary production and particle flux out of the euphotic zone is unclear [Knauer et al., 1984a]. The inability of primary production to predict flux in many areas [Bishop, 1989; Boyd and Newton, 1995; Karl et al., 1996; Lampitt and Antia, 1997] suggests that the magnitude of primary production may not be the most important factor in determining flux to depth [Boyd and Newton, 1999]. The flux to depth relationship of Martin et al. [1987], based on export, is thought to be a more accurate parameterization of flux to depth [Bishop, 1989; Boyd and Newton, 1999; Lampitt and Antia, 1997].

Figure 1

Locations of flux to depth measurements derived from sediment traps used in this study (open circles). The plus signs and crosses indicate the sample locations used by Suess [1980] and Martin et al. [1987], respectively. OSP, Ocean Station Papa; HOT, Hawaii Ocean Time-Series; BATS, Bermuda Atlantic Time-Series Study; NABE, North Atlantic Bloom Experiment.

Figure 2

Global ocean annual sediment trap POC flux versus depth used by Suess [1980] and Martin et al. [1987] and in this paper (solid, shaded, and open circles, respectively; references in Table 1). Dotted horizontal lines indicate depth bins used in regional correlation analysis.

[8] The Martin et al. [1987] normalized power function is a “best fit” derived from sediment trap data collected in the low-latitude to midlatitude east Pacific from depths between 100 and 2000 m (Figures 1 and 2). The exponent (−0.858) has been shown to vary within and between ocean basins [e.g., Banse, 1994; Karl et al., 1996; Usbeck, 1999], suggesting that flux cannot be described by variability in export alone. However, the Martin et al. [1987] relationship was adopted by the Ocean Carbon Model Intercomparison Project (OCMIP) project as a component of the “standard” OCMIP biology model. Global ocean models commonly describe flux to depth using the fixed parameters shown in (1) and (2). The global variability of the relationships between POC flux to depth and export or primary production has not previously been assessed.

3. Methods

3.1. Data Selection

[9] We assess the ability of the currently applied Suess [1980] and Martin et al. [1987] algorithms to predict flux to depth. We apply primary production estimates predominantly derived from ¹⁴C uptake experiments, and we apply export estimates derived from several methods including thorium isotope uptake experiments, f ratios, and mass balances. Flux predicted by the Suess [1980] and Martin et al. [1987] relationships is compared to sediment trap measurements from different ocean regions. We use the data outlined above to generate new empirical region-specific flux algorithms. The variability of ocean carbon storage predicted by the region-specific flux algorithms is assessed in a one-dimensional ocean model.

[10] Regions were selected on the basis of the availability of primary production, export, and multiple flux to depth estimates to include a variety of physical and biogeochemical provinces. Process studies include the North Atlantic Bloom Experiment (NABE), equatorial Pacific, and Arabian Sea Joint Global Ocean Flux Study (JGOFS) programs. Time series studies include those from the Sargasso Sea/Bermuda Atlantic Time-Series Study (BATS), northeast subarctic Pacific/Ocean Station Papa (OSP), and north central Pacific gyre/Hawaii Ocean Time-Series (HOT). Data selection and filtering are outlined in sections 3.2 and 3.3.

3.2. Estimating Flux to Depth Using Sediment Trap Data

[11] Problems associated with using sediment traps to characterize flux include hydrodynamic biases [Lorenzen et al., 1981; Baker et al., 1988; Buesseler, 1991; Gust et al., 1992, 1994; Siegel et al., 1990; Siegel and Deuser, 1997], zooplankton migration [Longhurst and Harrison, 1988; Walsh et al., 1988; Dam et al., 1995], sample contamination by swimmers [Lee et al., 1988; Karl and Knauer, 1989; Michaels et al., 1990], sample degradation [Knauer et al., 1984b; Honjo, 1990; Honjo et al., 1995], and brine addition [Macintyre et al., 1995; Gardner, 2000]. Hence the accuracy of sediment traps is debated [Jurg, 1996; Gust and Kozerski, 2000]. Reported errors associated with these caveats are variable; some arguably decrease with increasing depth below the photic zone [Gardner, 2000].

[12] The use of sediment traps for measuring the flux of settling particles to the deep ocean (>1.5 km) has been validated by ²³⁰Th and ²³¹Pa calibration studies [Scholten et al., 2001; Yu et al., 2001]. These radionuclide studies suggest that sediment traps may often undersample fluxes within the mesopelagic zone (<1.5 km). With the exception of the California margin, Yu et al. [2001] found that a trapping efficiency of 40% is a typical minimum value for the pelagic upper ocean. To account for the potential undertrapping error, we perform our analyses both with and without this correction factor applied to samples from within the mesopelagic zone. In particular, radiochemically calibrated fluxes are calculated to be the observed flux divided by 0.4. This trapping efficiency estimate may be uncertain in part because of the variable incorporation of radionuclides on particles of different sizes [Gardner, 2000; Yu et al., 2001].

[13] Data from ocean sediment trap deployments of the past 20 years are compiled in Table 1 and Figures 1 and 2. The following criteria are used in selecting sediment trap data for this study: (1) All values are from depths greater than the local photic zone and mixed layer depth maximum, (2) data from within 200 m of the seafloor are excluded to avoid contamination by sediment resuspension, (3) sediment traps located within coastal/shelf regions (generally, total water depths <500 m) are excluded to avoid input of terrigenous organic matter, and (4) data are included only if flux to depth is measured throughout an entire year. Exceptions to criterion 4 occur occasionally where samples were collected for a periods shorter than a full year. Accordingly, taking into account seasonal bias, the calculated flux is reported as a minimum if the samples were collected during low-flux periods or as a maximum if they were collected during high-flux periods. The Suess [1980] and Martin et al. [1987] relationships are each based on less than 50 sediment trap data points, of which >80% do not meet the criteria listed above and are not included. Our analysis includes 180 new sediment trap annual flux estimates.

Table 1. Sediment-Trap-Derived Annual Flux to Depth Rates^aa MIZ, marginal ice zone; SMT, Salzgitter Electronics, Kiel; OSU, Oregon State University; ACC, Antarctic Circumpolar Current; APFZ, Antarctic Polar Frontal Zone; and SAF, Subantarctic Front.

Region	Trap ID	Collection, Interval, Years	Latitude	Longitude	Water Depth, m	Trap Depth, m	POC Flux, mg m^−2 d^−1	Trap Type	Reference
Polar Arctic
Subarctic Atlantic^bb Data used in regional comparisons.	LB-1	1983–1984	69.5	10	3161	2761	1.37	Parflux 5	Honjo et al. [1987]
Subarctic Atlantic^bb Data used in regional comparisons.	BI-1	1984–1985	76	11	–	2800	2.85	Parflux 5	Honjo et al. [1987]
Subarctic Atlantic^bb Data used in regional comparisons.	FS-1	1984–1985	78.9	1.4	2823	2442	0.41	Parflux 6	Honjo et al. [1987]
Subarctic Atlantic^bb Data used in regional comparisons.	SP1	1987–1988	78.9	6.7	1618	1087	13.7	Kiel SMT	Hebbeln [2000]
Subarctic Atlantic^bb Data used in regional comparisons.	SP2	1988–1989	78.9	6.7	1661	1110	9.0	Kiel SMT	Hebbeln [2000]
Subarctic Atlantic^bb Data used in regional comparisons.	SP3	1989–1990	78.9	6.7	1676	1125	21.1	Kiel SMT	Hebbeln [2000]
Subarctic Atlantic^bb Data used in regional comparisons.	NA-1	1985–1986	65.5	1	3058	2630	0.59	Parflux 6	Honjo et al. [1987]
Norwegian Sea^bb Data used in regional comparisons.	–	1986–1987	67.8	5.5	1300	500	6.66	funnel	Bathmann et al. [1990]
Subarctic Atlantic^bb Data used in regional comparisons.	NB-1	1985–1986	70	−2	3269	2749	0.53	Parflux 6	Honjo et al. [1987]
Greenland Sea^bb Data used in regional comparisons.	OG	1988–1989	72.5	−9.5	2700	500	10.4	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	OG	1988–1989	72.5	−9.5	2700	1000	3.56	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	OG	1988–1989	72.5	−9.5	2700	2200	0.9	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	OG	1989–1990	72.5	−9.5	2700	500	10.1	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	OG	1989–1990	72.5	−9.5	2700	1000	3.86	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	OG	1990–1991	72.5	−9.5	2700	500	2.93	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	OG	1990–1991	72.5	−9.5	2700	1000	2.05	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	OG	1990–1991	72.5	−9.5	2700	2200	0.99	Kiel	Bodungen et al. [1995]
Greenland Sea^bb Data used in regional comparisons.	GB-21	1985–1986	74.6	−6.7	2588	1966	0.94	Parflux 5	Honjo et al. [1987]
Greenland Sea^bb Data used in regional comparisons.	GB-23	1985–1986	75.6	−6.7	3445	2823	0.4	Parflux 5	Honjo et al. [1987]
Atlantic Ocean
NE Atlantic^bb Data used in regional comparisons.	–	1989–1990	47.8	−19.5	4555	3100	5.5	Parflux 7	Newton et al. [1994]
NABE^bb Data used in regional comparisons.	S	1989–1990	33.8	−21	5100	1160	2.74	Parflux 7	Honjo and Manganini [1993]
NABE^bb Data used in regional comparisons.	S	1989–1990	33.8	−21	5100	1980	2.82	Parflux 7	Honjo and Manganini [1993]
NABE^bb Data used in regional comparisons.	S	1989–1990	33.8	−21	5100	4480	2.36	Parflux 7	Honjo and Manganini [1993]
NABE^bb Data used in regional comparisons.	N	1989–1990	47.7	−21.7	4435	1110	4.05	Parflux 7	Honjo and Manganini [1993]
NABE^bb Data used in regional comparisons.	N	1989–1990	47.7	−21.7	4435	2110	3.78	Parflux 7	Honjo and Manganini [1993]
NABE^bb Data used in regional comparisons.	N	1989–1990	47.7	−21.7	4435	3730	2.74	Parflux 7	Honjo and Manganini [1993]
Sargasso Sea^bb Data used in regional comparisons.	S2	1977^cc Sample time periods less than 1 year.	31.5	−55.9	5581	976	2.43^dd Trap deployments less than an entire year: July–October (110 days), may be overestimated.	Parflux	Honjo [1980]
Sargasso Sea^bb Data used in regional comparisons.	S2	1977^cc Sample time periods less than 1 year.	31.5	−55.9	5581	3694	0.087^dd Trap deployments less than an entire year: July–October (110 days), may be overestimated.	Parflux	Honjo [1980]
Sargasso Sea^bb Data used in regional comparisons.	S2	1977^cc Sample time periods less than 1 year.	31.5	−55.9	5581	5206	0.07^dd Trap deployments less than an entire year: July–October (110 days), may be overestimated.	Parflux	Honjo [1980]
Sargasso Sea/BATS	BATS	1988–1998	31.7	−64.2	4400	150	26.0	MultiPIT	BATS online data
Sargasso Sea/BATS	BATS	1989–1998	31.7	−64.2	4400	200	20.2	MultiPIT	BATS online data
Sargasso Sea/BATS^bb Data used in regional comparisons.	BATS	1989–1998	31.7	−64.2	4400	300	14.1	MultiPIT	BATS online data
Sargasso Sea/BATS^bb Data used in regional comparisons.	BATS	1989–1990	31.8	−64.2	4400	400	12.2	MultiPIT	BATS online data
Sargasso Sea^bb Data used in regional comparisons.	SCIFF	1979–1985	31.8	−64.2	4400	3200	2.0	Parflux	Deuser et al. [1990]
SW Africa	NU2	–	−29	13	–	768	18.9	–	Usbeck [1999]
SW Africa	NU2-l	1992–1993	−28.6	14.6	3055	2516	4.4	–	Fischer et al. [2000]
SW Africa	WR1	1988–1989	−20.1	9.2	2217	1640	16.7	Kiel SMT	Wefer and Fischer [1993]
SW Africa	WR2u	1989–1990	−20	9.2	2196	599	14	Kiel SMT	Wefer and Fischer [1993]
SW Africa	WR2l	1989–1990	−20	9.2	2196	1654	10.4	Kiel SMT	Wefer and Fischer [1993]
SW Africa	WR3	1990–1991	−20	9	2208	1648	7.73	Kiel SMT	Usbeck [1999]
NW Africa^bb Data used in regional comparisons.	CV1u, 2u	1992–1994	11.5	−21	4968	1000	7.36	Kiel SMT	Usbeck [1999]
NW Africa^bb Data used in regional comparisons.	CV1l, 2l	1992–1994	11.5	−21	4968	4500	3.48	Kiel SMT	Usbeck [1999]
NW Africa^bb Data used in regional comparisons.	CB1-l	1988–1989	20.8	−19.7	3646	2195	3.3	Parflux 6	Fischer et al. [1996]
NW Africa^bb Data used in regional comparisons.	CB2-l	1989–1990	21.1	−20.7	4092	3502	4.4	Parflux 5	Fischer et al. [1996]
NW Africa^bb Data used in regional comparisons.	CB3-u	1990–1991	21.1	−20.7	4094	730	5.5	Kiel SMT	Fischer et al. [1996]
NW Africa^bb Data used in regional comparisons.	CB3-l	1990–1991	21.1	−20.7	4094	3557	4.7	Kiel SMT	Fischer et al. [1996]
NW Africa^bb Data used in regional comparisons.	CB4-u	1991^cc Sample time periods less than 1 year.	21.1	−20.7	4108	733	9.3^ee Trap deployments less than an entire year: March–November (200 days), may be overestimated.	Kiel SMT	Fischer et al. [1996]
NW Africa^bb Data used in regional comparisons.	CB4-l	1991^cc Sample time periods less than 1 year.	21.1	−20.7	4108	3562	5.5^ee Trap deployments less than an entire year: March–November (200 days), may be overestimated.	Kiel SMT	Fischer et al. [1996]
Equatorial Atlantic	WA3u	1993–1994	−7.5	−28.0	5570	671	2.6	–	Fischer et al. [2000]
Equatorial Atlantic	WA3l	–	−8	−28	5570	5031	0.81	–	Usbeck [1999]
Equatorial Atlantic	EA8u	1991–1992^cc Sample time periods less than 1 year.	−5.8	−9.4	3450	598	7.45	–	Usbeck [1999]
Equatorial Atlantic	EA8m	1991–1992^cc Sample time periods less than 1 year.	−5.8	−9.4	3450	1833	6.47	–	Usbeck [1999]
Equatorial Atlantic	EA8l	1991–1992^cc Sample time periods less than 1 year.	−5.8	−9.4	3450	2890	3.92	–	Usbeck [1999]
Equatorial Atlantic	WA4u	–	−4	−26	–	808	5.1	–	Usbeck [1999]
Equatorial Atlantic	WA4l	–	−4	−26	–	4555	2.67	–	Usbeck [1999]
Equatorial Atlantic	GBZ4	1989–1990	−2.2	−9.9	3912	696	3	Kiel SMT	Wefer and Fischer [1993]
Equatorial Atlantic	GBZ5u	–	−2	−10	–	597	8.22	–	Usbeck [1999]
Equatorial Atlantic	GBZ5l	–	−2	−10	–	3382	6.3	−	Usbeck [1999]
Equatorial Atlantic	GBN3u	1989–1990	1.8	−11.1	4481	853	8.2	Kiel SMT	Wefer and Fischer [1993]
Equatorial Atlantic	GBN3l	1989–1990	1.8	−11.1	4481	3921	6	Kiel SMT	Wefer and Fischer [1993]
Subtropical Atlantic	E	1977–1978^cc Sample time periods less than 1 year.	13.5	−54	5288	389	6.73^ff Trap deployments less than an entire year: November–February (98 days), may be underestimated.	Parflux 2	Honjo [1980]
Subtropical Atlantic	E	1977–1978^cc Sample time periods less than 1 year.	13.5	−54	5288	988	3.94^ff Trap deployments less than an entire year: November–February (98 days), may be underestimated.	Parflux 2	Honjo [1980]
Subtropical Atlantic	E	1977–1978^cc Sample time periods less than 1 year.	13.5	−54	5288	3755	1.73^ff Trap deployments less than an entire year: November–February (98 days), may be underestimated.	Parflux 2	Honjo [1980]
Subtropical Atlantic	E	1977–1978^cc Sample time periods less than 1 year.	13.5	−54	5288	5068	1.7^ff Trap deployments less than an entire year: November–February (98 days), may be underestimated.	Parflux 2	Honjo [1980]
Pacific Ocean
Ocean Station P	OSP	1989–1993	50	−145	4250	200	18.2	Parflux 5–7	Wong et al. [1999]
Ocean Station P^bb Data used in regional comparisons.	OSP	1983–1993	50	−145	4250	1000	7.42	Parflux 5–7	Wong et al. [1999]
Ocean Station P^bb Data used in regional comparisons.	OSP	1982–1993	50	−145	4250	3800	3.10	Parflux 5–7	Wong et al. [1999]
Subarctic Pacific^bb Data used in regional comparisons.	Aleutian Islands	–	49	−174	5400	4800	3.4	–	Takahashi [1995]
Subarctic Pacific^bb Data used in regional comparisons.	Aleutian Islands	–	49	−174	5400	4800	5.1	–	Takahashi [1995]
Subarctic Pacific^bb Data used in regional comparisons.	Bering Sea	–	53.5	−177	3800	3200	8.1	–	Takahashi [1995]
Subarctic Pacific^bb Data used in regional comparisons.	Bering Sea	–	53.5	−177	3800	3200	10.5	–	Takahashi [1995]
North central Pacific Gyre^bb Data used in regional comparisons.	P1	1978^cc Sample time periods less than 1 year.	15.4	−152	5792	378	3.56^gg Trap deployments less than an entire year: September–November (61 days), may be underestimated.	Parflux 2	Honjo [1980]
North central Pacific Gyre^bb Data used in regional comparisons.	P1	1978^cc Sample time periods less than 1 year.	15.4	−152	5792	978	0.55^gg Trap deployments less than an entire year: September–November (61 days), may be underestimated.	Parflux 2	Honjo [1980]
North central Pacific Gyre^bb Data used in regional comparisons.	P1	1978^cc Sample time periods less than 1 year.	15.4	−152	5792	2778	1.09^gg Trap deployments less than an entire year: September–November (61 days), may be underestimated.	Parflux 2	Honjo [1980]
North central Pacific Gyre^bb Data used in regional comparisons.	P1	1978^cc Sample time periods less than 1 year.	15.4	−152	5792	4280	0.88^gg Trap deployments less than an entire year: September–November (61 days), may be underestimated.	Parflux 2	Honjo [1980]
North central Pacific Gyre^bb Data used in regional comparisons.	P1	1978^cc Sample time periods less than 1 year.	15.4	−152	5792	5582	0.66^gg Trap deployments less than an entire year: September–November (61 days), may be underestimated.	Parflux 2	Honjo [1980]
North central Pacific Gyre/HOT	ALOHA	1988–1999	22.8	−158.0	4800	150	14^hh POC calculated as 50% of the total particulate carbon reported based on simultaneous inorganic and organic carbon determinations of sediment trap material from the North Pacific [Wong et al., 1999].	MultiPIT	HOT online data
North central Pacific Gyre/HOT	ALOHA	1994–1996	22.8	−158.0	4800	200	9^hh POC calculated as 50% of the total particulate carbon reported based on simultaneous inorganic and organic carbon determinations of sediment trap material from the North Pacific [Wong et al., 1999].	MultiPIT	HOT online data
North central Pacific Gyre/HOT	ALOHA	1988–1995	22.8	−158.0	4800	300	7.7^hh POC calculated as 50% of the total particulate carbon reported based on simultaneous inorganic and organic carbon determinations of sediment trap material from the North Pacific [Wong et al., 1999].	MultiPIT	HOT online data
North central Pacific Gyre/HOT^bb Data used in regional comparisons.	ALOHA	1989–1995	22.8	−158.0	4800	500	5.5^hh POC calculated as 50% of the total particulate carbon reported based on simultaneous inorganic and organic carbon determinations of sediment trap material from the North Pacific [Wong et al., 1999].	MultiPIT	HOT online data
South China Sea^bb Data used in regional comparisons.	SCS-C	1990–1995	14.6	115.1	4310	1200	4.20	Parflux 6	Jianfang et al. [1998]; Wiesner et al. [1996]
South China Sea^bb Data used in regional comparisons.	SCS-C	1990–1995	14.6	115.1	4310	2240	3.51	Parflux 6	Jianfang et al. [1998]
South China Sea^bb Data used in regional comparisons.	SCS-C	1990–1995	14.6	115.1	4310	3770	2.52	Parflux 6	Jianfang et al. [1998]; Wiesner et al. [1996]
South China Sea^bb Data used in regional comparisons.	SCS-N	1987–1988	18.5	116	3750	1000	3.92	Parflux 6	Jianfang et al. [1998]; Wiesner et al. [1996]
South China Sea^bb Data used in regional comparisons.	SCS-N	1987–1988	18.5	116	3750	3350	2.02	Parflux 6	Jianfang et al. [1998]; Wiesner et al. [1996]
North Equatorial Current	NEC-T	1988–1989	12.0	134.3	5300	1200	0.38	Parflux 6	Kempe and Knaack [1996]
North Equatorial Current	NEC-B	1988–1989	12.0	134.3	5300	4300	0.43	Parflux 6	Kempe and Knaack [1996]
Equatorial Counter Current	ECC-T	1988–1989	5.0	138.8	4130	1130	1.78	Parflux 6	Kempe and Knaack [1996]
Equatorial Counter Current	ECC-B	1988–1989	5.0	138.8	4130	3130	0.67	Parflux 6	Kempe and Knaack [1996]
Equatorial Pacific^bb Data used in regional comparisons.	S	1983–1984	11	−19140	4800	700	3.12	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	S	1983–1984	11	−140	4800	1600	2.38	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	S	1983–1984	11	−140	4800	3400	1.62	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	9N	1992–1993	9	−140	5100	2150	1.51	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	9N	1992–1993	9	−140	5100	2250	1.52	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	9N	1992–1993	9	−140	5100	4400	0.96	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	5N	1992–1993	5	−140	4493	1191	6.02	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	5N	1992–1993	5	−140	4493	2091	4.50	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	5N	1992–1993	5	−140	4493	3793	3.84	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	2N	1992–1993	2	−140	4397	2203	3.96	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	C	1983–1984	1	−139	4400	1095	2.96	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	C	1983–1984	1	−139	4400	1895	4.25	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	C	1983–1984	1	−139	4400	3495	3.26	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	C	1984–1985	1	−139	4400	1083	5.36	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	C	1984–1985	1	−139	4400	1883	6.78	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	C	1984–1985	1	−139	4400	2908	5.18	OSU trap	Dymond and Collier [1988]
Equatorial Pacific^bb Data used in regional comparisons.	EQ	1992–1993	0	−140	4358	880	4.64	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	EQ	1992–1993	0	−140	4358	2284	4.49	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	EQ	1992–1993	0	−140	4358	3618	4.38	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	2S	1992–1993	−2	−140	4293	3593	3.61	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	5S	1992–1993	−5	−140	4198	1216	2.72	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	5S	1992–1993	−5	−140	4198	2099	2.72	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	5S	1992–1993	−5	−140	4198	2316	2.80	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	12S	1992–1993	−12	−135	4294	1292	1.52	Parflux	Honjo et al. [1995]
Equatorial Pacific^bb Data used in regional comparisons.	12S	1992–1993	−12	−135	4294	3594	0.72	Parflux	Honjo et al. [1995]
Peru-Chile Current	CH-3–4	1993–1994	−30	−73.2	4360	2323	7.67	Kiel SMT	Hebbeln et al. [2000]
Panama Basin^bb Data used in regional comparisons.	PB	1979^cc Sample time periods less than 1 year.	5.35	−81.9	3856	667	12.6^ii Trap deployments less than an entire year: August–December (112 days), may be underestimated.	Parflux	Honjo et al. [1982]
Panama Basin^bb Data used in regional comparisons.	PB	1979^cc Sample time periods less than 1 year.	5.35	−81.9	3856	1268	8.95^ii Trap deployments less than an entire year: August–December (112 days), may be underestimated.	Parflux	Honjo et al. [1982]
Panama Basin^bb Data used in regional comparisons.	PB	1979^cc Sample time periods less than 1 year.	5.35	−81.9	3856	2265	9.06^ii Trap deployments less than an entire year: August–December (112 days), may be underestimated.	Parflux	Honjo et al. [1982]
Panama Basin^bb Data used in regional comparisons.	PB	1979^cc Sample time periods less than 1 year.	5.35	−81.9	3856	2869	10.9^ii Trap deployments less than an entire year: August–December (112 days), may be underestimated.	Parflux	Honjo et al. [1982]
Panama Basin^bb Data used in regional comparisons.	PB	1979^cc Sample time periods less than 1 year.	5.35	−81.9	3856	3769	11.4^ii Trap deployments less than an entire year: August–December (112 days), may be underestimated.	Parflux	Honjo et al. [1982]
Panama Basin^bb Data used in regional comparisons.	PB	1979^cc Sample time periods less than 1 year.	5.35	−81.9	3856	3791	10.6^ii Trap deployments less than an entire year: August–December (112 days), may be underestimated.	Parflux	Honjo et al. [1982]
Panama Basin^bb Data used in regional comparisons.	–	1979–1980	5.37	−85.6	3860	890	9.7	Parflux 2	Honjo [1982]
Panama Basin^bb Data used in regional comparisons.	–	1979–1980	5.37	−85.6	3860	2590	10.9	Parflux 2	Honjo [1982]
Panama Basin^bb Data used in regional comparisons.	–	1979–1980	5.37	−85.6	3860	3560	13.7	Parflux 2	Honjo [1982]
Indian Ocean
Arabian Sea, oceanic^bb Data used in regional comparisons.	MS5	1994–1995^cc Sample time periods less than 1 year.	10	65	4411	800	5.7^jj Trap deployments less than an entire year: November–April (170 days), may be underestimated.	Parflux 7	Honjo et al. [1999]
Arabian Sea, oceanic^bb Data used in regional comparisons.	MS5	1994–1995	10	65	4411	2363	3.8	Parflux 7	Honjo et al. [1999]
Arabian Sea, oceanic^bb Data used in regional comparisons.	MS5	1994–1995	10	65	4411	3915	3.3	Parflux 7	Honjo et al. [1999]
Arabian Sea, oceanic^bb Data used in regional comparisons.	CAST	1987	14.5	64.6	3904	2913	3.0	Parflux 6	Haake et al. [1993]
Arabian Sea, oceanic^bb Data used in regional comparisons.	CAST	1986	14.5	64.6	3906	2907	5.2	Parflux 6	Haake et al. [1993]
Arabian Sea, oceanic^bb Data used in regional comparisons.	CAST	1988	14.5	64.6	3908	3021	7.1	Parflux 6	Haake et al. [1993]
Arabian Sea, oceanic^bb Data used in regional comparisons.	EAST	1989	15.6	68.6	3807	2938	6.0	Parflux 6	Haake et al. [1993]
Arabian Sea, oceanic^bb Data used in regional comparisons.	EAST	1987	15.6	68.6	3776	2772	3.8	Parflux 6	Haake et al. [1993]
Arabian Sea, oceanic^bb Data used in regional comparisons.	EAST	1986	15.6	68.6	3776	2776	7.7	Parflux 6	Haake et al. [1993]
Arabian Sea, oceanic^bb Data used in regional comparisons.	EAST	1990	15.6	68.6	3862	2928	5.2	Parflux 6	Haake et al. [1993]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS1	1994–1995	17.7	58.9	1448	808	10.5	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS1	1994–1995^cc Sample time periods less than 1 year.	17.7	58.9	1448	999	10.5^kk Trap deployments less than an entire year: November–August (276 days), may be overestimated.	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS2	1994–1995	17.4	58.8	3650	828	13.5	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS2	1994–1995	17.4	58.8	3650	903	17.2	Parflux 7	Haake et al. [1993]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS2	1994–1995	17.4	58.8	3650	1974	17.4	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS2	1994–1995	17.4	58.8	3650	3141	13.2	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS3	1994–1995	17.2	59.6	3470	764	13.2	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS3	1994–1995	17.2	59.6	3470	858	17.5	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS3	1994–1995	17.2	59.6	3470	1857	16.3	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS3	1994–1995	17.2	59.6	3470	2871	12.8	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS4	1994–1995	15.3	61.5	3980	821	8.9	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS4	1994–1995	15.3	61.5	3980	2229	11.1	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	MS4	1994–1995	15.3	61.5	3980	3478	8.9	Parflux 7	Honjo et al. [1999]
Arabian Sea, coastal^bb Data used in regional comparisons.	WAST	1986	16.3	60.3	4024	3023	7.1	Parflux 6	Haake et al. [1993]
Arabian Sea, coastal^bb Data used in regional comparisons.	WAST	1987	16.3	60.3	4024	3036	6.9	Parflux 6	Haake et al. [1993]
Arabian Sea, coastal^bb Data used in regional comparisons.	WAST	1988	16.3	60.3	4024	3034	9.0	Parflux 6	Haake et al. [1993]
Arabian Sea, coastal^bb Data used in regional comparisons.	WAST	1990	16.3	60.3	4024	3016	12.3	Parflux 6	Haake et al. [1993]
Bay of Bengal	South-s	1987–1988	4.4	87.3	4017	1040	6.49	Parflux	Ittekkot et al. [1991]
Bay of Bengal	South-d	1987–1988	4.4	87.3	4017	3006	5.59	Parflux	Ittekkot et al. [1991]
Bay of Bengal	Central-s	1987–1988	13.2	84.4	3259	906	7.23	Parflux	Ittekkot et al. [1991]
Bay of Bengal	Central-d	1987–1988	13.2	84.4	3259	2282	7.15	Parflux	Ittekkot et al. [1991]
Bay of Bengal	North-s	1987–1988	17.4	89.6	2263	809	9.84	Parflux	Ittekkot et al. [1991]
Bay of Bengal	North-d	1987–1988	17.4	89.6	2263	1727	7.26	Parflux	Ittekkot et al. [1991]
Polar Antarctic
ACC, Atlantic^bb Data used in regional comparisons.	VIII-u	1992	−62.1	−40.6	3280	2453	6.62	Parflux	Pudsey and King [1997]
ACC, Atlantic^bb Data used in regional comparisons.	VIII-l	1992	−62.1	−40.6	3280	3259	2.81	Parflux	Pudsey and King [1997]
ACC, Atlantic^bb Data used in regional comparisons.	I-l	1990	−63.2	−42.7	3798	3777	2.07	Parflux	Pudsey and King [1997]
ACC, Atlantic^bb Data used in regional comparisons.	I-u	1992	−63.2	−42.7	3793	2966	6.53	Parflux	Pudsey and King [1997]
ACC, Atlantic^bb Data used in regional comparisons.	I-1	1992	–63.2	–42.7	3793	3766	1.52	Parflux	Pudsey and King [1997]
ACC, Atlantic^bb Data used in regional comparisons.	II-l	1990	−63.5	−41.7	4552	4531	5.73	Parflux	Pudsey and King [1997]
ACC, Atlantic^bb Data used in regional comparisons.	III-u	1990	−64	−40.9	4537	3710	5.11	Parflux	Pudsey and King [1997]
ACC, Atlantic	WS2-l	1987^cc Sample time periods less than 1 year.	−64.9	−2.5	5000	4456	0.47^ll Trap deployments less than an entire year: January–November (304 days), may be underestimated.	Parflux	Wefer and Fischer [1991]
ACC, Atlantic^bb Data used in regional comparisons.	WS3-u	1988–1989	−64.9	−2.5	5053	360	6.47	–	Fischer et al. [2000]
Bouvert Island^bb Data used in regional comparisons.	BO1-u	1990–1991	−54.3	−3.4	2734	450	7.37	funnel	Fischer et al. [2000]
Bouvert Island	BO2-u	1987^cc Sample time periods less than 1 year.	−54.3	−3.4	2965	456	1.32^mm Trap deployments less than an entire year: May–December (200 days), may be underestimated.	funnel	Fischer et al. [2000]
Polar Front^bb Data used in regional comparisons.	PF1-u	1987–1988	−50.1	−5.9	3750	700	8.93	–	Fischer et al. [2000]
Polar Front^bb Data used in regional comparisons.	PF3-u	1989–1990	−50.1	−5.9	3785	614	10.5	–	Fischer et al. [2000]
MIZ, Atlantic	KG1-u	1983–1984	−62.3	−57.5	1952	494	5.56	Parflux 6	Fischer et al. [2000]
MIZ, Atlantic	KG2-u	1984–1985	−62.3	−57.5	1650	693	0.99	Parflux 6	Fischer et al. [2000]
MIZ, Atlantic	KG3-u	1985–1986	−62.3	−57.5	1992	687	3.07	Parflux 6	Fischer et al. [2000]
MIZ, Atlantic	KG1-l	1983–1984	−61.3	−57.5	1952	1588	1.6^nn POC estimated assuming that the lower trap has the same combustible flux %C as the upper trap [Fischer et al., 2000; Wefer et al., 1988].	funnel	Wefer et al. [1988]
MIZ, Pacific	MS5	1996–1997	−66.2	−170	3016	937	5.21	Parflux 6	Honjo et al. [2000]
ACC, Pacific	MS4	1996–1997	−63.1	−170	2886	1031	6.03	Parflux 6	Honjo et al. [2000]
APFZ, Pacific	MS3	1996–1997	−60.3	−170	3958	1103	6.3	Parflux 6	Honjo et al. [2000]
SAF, Pacific	MS2	1996–1998	−56.9	−170	4924	982	4.66	Parflux 6	Honjo et al. [2000]
SAF, Pacific	MS2	1996–1998	−56.9	−170	4924	4224	1.71	Parflux 6	US JGOFS online data
SAF, Pacific	MS1	1996–1997	−53.0	−175	5441	986	1.1	Parflux 6	Honjo et al. [2000]

• a MIZ, marginal ice zone; SMT, Salzgitter Electronics, Kiel; OSU, Oregon State University; ACC, Antarctic Circumpolar Current; APFZ, Antarctic Polar Frontal Zone; and SAF, Subantarctic Front.
• b Data used in regional comparisons.
• c Sample time periods less than 1 year.
• d Trap deployments less than an entire year: July–October (110 days), may be overestimated.
• e Trap deployments less than an entire year: March–November (200 days), may be overestimated.
• f Trap deployments less than an entire year: November–February (98 days), may be underestimated.
• g Trap deployments less than an entire year: September–November (61 days), may be underestimated.
• h POC calculated as 50% of the total particulate carbon reported based on simultaneous inorganic and organic carbon determinations of sediment trap material from the North Pacific [Wong et al., 1999].
• i Trap deployments less than an entire year: August–December (112 days), may be underestimated.
• j Trap deployments less than an entire year: November–April (170 days), may be underestimated.
• k Trap deployments less than an entire year: November–August (276 days), may be overestimated.
• l Trap deployments less than an entire year: January–November (304 days), may be underestimated.
• m Trap deployments less than an entire year: May–December (200 days), may be underestimated.
• n POC estimated assuming that the lower trap has the same combustible flux %C as the upper trap [Fischer et al., 2000; Wefer et al., 1988].

[14] The analytical determination of POC in sediment trap samples has been refined during the past few decades. Not all data sources (Table 1) specify the methodology used to separate particulate inorganic carbon (PIC) from POC. When methods are cited, both fuming with acid and acid rinsing are most commonly used to remove PIC. These methods may lead to the overestimations and underestimations of POC concentrations [Grasshoff et al., 1999]. Conversely, the treatment method used to remove PIC may be within the error of sediment trap sample processing [Honjo, 1980]. While the lack of methodological consistency introduces uncertainty into the comparison of sediment trap records, this problem, as well as other trap technology uncertainties, would have also been present in the data underlying the development of the Martin et al. [1987] and Suess [1980] parameterizations.

[15] To standardize analysis between regions where the frequency of sediment trap measurements at different depths varies, formulations of new flux to depth equations incorporate data binned between depth ranges. The flux estimates at each locale are averaged within the following depth ranges: 0.5 to 1 km, 1 to 2 km, 2 to 3 km, 3 to 4 km, and >4 km. Average flux values are assigned a nominal depth of the mean depth of observations in each depth range.

3.3. Estimating Primary Production and Export

[16] Table 2 includes regional estimates of primary production and export primarily obtained during JGOFS process, time series, and other field-based studies. In most cases, primary production was equated to net photosynthesis as estimated by ¹⁴C uptake experiments. The one exception is the Southern Ocean/Atlantic sector region. Here, primary production is estimated using an algorithm based on monthly climatological phytoplankton concentrations from the coastal zone color scanner (CZCS) and in situ ¹⁴C-based primary productivity data from throughout the Southern Ocean [Arrigo et al., 1998]. We made this exception because of the lack of ¹⁴C-based primary production results for this region.

Table 2. Regional Estimates of Primary Production and Particulate Organic Carbon (POC) Export From the Base to the Photic Zone^aa Export rates derived from POC/²³⁴Th uptake experiments. Primary production estimates derived from ¹⁴C uptake incubation experiments.

Region	Latitude	Longitude	Z0, m	f Ratio	Export, mg C m^−2 d^−1	Primary Production, mg C m^−2 d^−1	Reference: Export	Reference: Primary Production
Greenland and Norwegian Seas	65° to 79°	−10° to 11°	50	0.3	70^bb Derived from f ratios (assuming that export equals primary production times f ratio).	240	Bodungen et al. [1995]^cc Suggested annual mean (Table 2).	Bodungen et al. [1995]^cc Suggested annual mean (Table 2).
NE Atlantic/NABE	34° to 48°	−21° to −22°	40	0.45	280^bb Derived from f ratios (assuming that export equals primary production times f ratio).	700	Bury et al. [2001]^dd Monthly weighted average April–August, 32°–50°N, 20°W data compilation (Tables 6 and 7; references cited within), may be overestimated.	Bury et al. [2001]^dd Monthly weighted average April–August, 32°–50°N, 20°W data compilation (Tables 6 and 7; references cited within), may be overestimated.
Sargasso Sea/BATS	32°	−56° to −64°	140	–	30, 44^ee Derived from nutrient mass balances.	360	Buesseler [1998]^ff Average March–October, may be overestimated. ; Michaels et al. [1994]^e,e Derived from nutrient mass balances. ^gg Annual average 1989–1993.	Michaels et al. [1994]^gg Annual average 1989–1993.
Subarctic Pacific/OSP	50°	−145°	70	–	70	1000	Charette et al. [1999]^hh Suggested annual mean.	Boyd and Harrison [1999]^hh Suggested annual mean.
North Central Pacific gyre/HOT	15° to 23°	−151° to −158°	150	–	66^ee Derived from nutrient mass balances.	460	Emerson et al. [1997]^hh Suggested annual mean.	Karl et al. [1996]^hh Suggested annual mean.
South China Sea	15° to 18°	115° to 116°	100	–	230	420	Huang et al. [1996]^ii Winter, may be underestimated.	Huang [1988]^jj Average April–July, may be overestimated.
Arabian Sea, oceanic	10° to 16°	65° to 69°	90	–	70	1000	Buesseler et al. [1998] (stations n11, s11, and s15)^kk Average NE Monsoon, Spring Intermonsoon, Mid-SW Monsoon, and Late-SW Monsoon periods.	Barber et al. [2001] (stations n11, s11, and s15)^kk Average NE Monsoon, Spring Intermonsoon, Mid-SW Monsoon, and Late-SW Monsoon periods.
Arabian Sea, coastal	16° to 18°	59° to 62°	65	–	130	1250	Buesseler et al. [1998] (stations s2, s3, s4, and s7)^kk Average NE Monsoon, Spring Intermonsoon, Mid-SW Monsoon, and Late-SW Monsoon periods.	Barber et al. [2001] (stations s2, s3, s4, and s7)^kk Average NE Monsoon, Spring Intermonsoon, Mid-SW Monsoon, and Late-SW Monsoon periods.
Equatorial Pacific (±2° latitude)	−2° to 2°	−139° to −140°	114	–	90	1035	Murray et al. [1996] (stations 6–10)^ll Average February–March and August–September (Table 5).	Murray et al. [1996] (stations 6–10)^ll Average February–March and August–September (Table 5).
Equatorial Pacific (±5° latitude)	5°, −5°	−137° to −140°	130	–	76	780	Murray et al. [1996] (stations 4 and 12)^ll Average February–March and August–September (Table 5).	Murray et al. [1996] (stations 4 and 12)^ll Average February–March and August–September (Table 5).
Equatorial Pacific (±9°–16° latitude)	9° to 16°, −9° to −16°	−134° to −152°	130	–	42	344	Murray et al. [1996] (stations 1, 2, and 15)^ll Average February–March and August–September (Table 5).	Murray et al. [1996] (stations 1, 2, and 15)^ll Average February–March and August–September (Table 5).
Panama Basin	5°	−82° to −86°	70	0.43	100^bb Derived from f ratios (assuming that export equals primary production times f ratio).	230	Bishop et al. [1986]^mm Average summer and winter (Table 9).	Bishop et al. [1986]^mm Average summer and winter (Table 9).
Southern Ocean/Atlantic sector	−50° to −65°	−2° to −43°	40	0.15	60^bb Derived from f ratios (assuming that export equals primary production times f ratio).	380^nn Derived from regional remote sensing.	Sambrotto and Mace [2000]^oo Average November–March, 55°–65°S, 170°W (Table 4), export may be overestimated.	Arrigo et al. [1998]^pp Annual average Weddell Sea sector, pelagic and marginal ice zone (MIZ) (Table 2).
NW Africa	12° to 21°	−20° to −21°	50	0.6	1200^bb Derived from f ratios (assuming that export equals primary production times f ratio).	2000	Jewell [1994]^qq Average coastal upwelling, may be underestimated [Jewell, 1994]. ; Minas et al. [1986]^qq Average coastal upwelling, may be underestimated [Jewell, 1994].	Huntsman and Barber [1977]^rr Average March–May, may be underestimated [see Minas et al., 1986].

• a Export rates derived from POC/²³⁴Th uptake experiments. Primary production estimates derived from ¹⁴C uptake incubation experiments.
• b Derived from f ratios (assuming that export equals primary production times f ratio).
• c Suggested annual mean (Table 2).
• d Monthly weighted average April–August, 32°–50°N, 20°W data compilation (Tables 6 and 7; references cited within), may be overestimated.
• e Derived from nutrient mass balances.
• f Average March–October, may be overestimated.
• g Annual average 1989–1993.
• h Suggested annual mean.
• i Winter, may be underestimated.
• j Average April–July, may be overestimated.
• k Average NE Monsoon, Spring Intermonsoon, Mid-SW Monsoon, and Late-SW Monsoon periods.
• l Average February–March and August–September (Table 5).
• m Average summer and winter (Table 9).
• n Derived from regional remote sensing.
• o Average November–March, 55°–65°S, 170°W (Table 4), export may be overestimated.
• p Annual average Weddell Sea sector, pelagic and marginal ice zone (MIZ) (Table 2).
• q Average coastal upwelling, may be underestimated [Jewell, 1994].
• r Average March–May, may be underestimated [see Minas et al., 1986].

[17] Estimates of primary production based on ¹⁴C uptake experiments, as in the case of sediment traps, may include errors. Estimates of primary production may differ significantly if they are derived from carbon fixed in POC or in both dissolved organic carbon (DOC) and POC [Sakshaug et al., 1997]. The duration of the incubation experiment is another potential source of error. Long incubation times may yield lower carbon uptake rates than short incubations (∼1 hour) because of the likelihood of the recycling of labeled carbon [Dring and Jewson, 1982; Sakshaug et al., 1997]. The organic C to chlorophyll a ratio is highly variable and is another potential source of error [Sakshaug et al., 1997]. Hence ¹⁴C-based estimates of primary production may include significant uncertainties.

[18] We employ several approaches to estimate export of particulate organic carbon from surface waters on regional and annual scales: ²³⁴Th, mass balances, and new production (Table 2). The ²³⁴Th-based export estimates use ²³⁴Th activity measurements to calculate vertical ²³⁴Th fluxes, which are multiplied by the organic C to ²³⁴Th ratio of sinking particulate material to quantify POC export. While this approach has been applied to a wide range of oceanographic settings [Buesseler, 1998] and is a preferred methodology [Buesseler, 1991], associated errors have not been quantified. Calculating the ²³⁴Th flux involves biases and uncertainties including steady state effects, time variability, and transport terms [Buesseler et al., 1992; Wei and Murray, 1992; Lee et al., 1993; Kim et al., 1999]. Additionally, estimates of the export of organic components are associated with large errors because of uncertainty in the ratio of C:²³⁴Th on sinking particles [Michaels et al., 1994; VanderLoeff et al., 1997; Gardner, 2000]. Careful elemental mass balances relying on annual budgets of dissolved and particulate material may offer the most accurate estimates of export and are incorporated where available.

[19] The export of organic carbon from surface waters is often estimated using f ratios by equating new production with export, assuming steady state on an annual basis. The f ratio is the ratio of new production to total (new plus regenerated) production. We apply this approach where annual ²³⁴Th and mass balance export estimates are unavailable. A significant and variable portion of total export calculated by this method may be in the form of DOC [e.g., Quay, 1997; Stoll et al., 1996] and unsuitable for comparison to the POC flux recorded by sediment traps. At high latitudes, where DOC is a minor component of flux and ²³⁴Th-based export rates are available during only a few months, rates of POC export are derived by assuming that new production equals POC export.

4. Results and Discussion

4.1. Flux to Depth

[20] A review of the locations of annual sediment trap sites reveals large areas with little or no data (Figure 1). The Northern Hemisphere contains almost 3 times the number of flux to depth estimates as the Southern Hemisphere contains, despite having less than 40% of the global ocean area. Few sediment trap data are available from the western sides of main ocean basins. The central ocean gyres and large portions of the Southern Ocean are poorly characterized. Areas of low productivity are relatively undersampled compared to more productive coastal margins. Oligotrophic and eutrophic portions of the ocean are relatively undersampled, compared to mesotrophic areas (70% of total samples). Furthermore, the moderately productive Southern Hemisphere Subtropical Convergence (∼40°S) has received little attention. Globally, the bulk of flux to depth data is from depths greater than 1 km.

[21] Annual carbon flux to depth for all global locations is shown in Figure 2. Above 1 km, global variability of annual flux to depth is more than an order of magnitude larger than below that depth (Table 3). In the intermediate and deep ocean, variability decreases with increasing depth. Here, the vertical flux of POC follows an irregular pattern of both increasing and decreasing flux with increasing depth (Table 4). In almost half of the POC flux rate change estimates the flux to deeper traps exceeds the flux observed in shallower traps in the same region. The irregular pattern may be due to a variable availability of sinking material to decomposition, differential particle sinking trajectories [Siegel et al., 1990; Siegel and Deuser, 1997], trapping inaccuracies, or heterogeneity of particle production and export and flux to depth within the regions considered. The variability of flux to depth estimates in the intermediate and deep ocean constitutes a minor fraction of particle production and export. Most of the modification of flux to depth occurs in the upper 1 km of the water column.

Table 3. Descriptive Statistics of Flux Rates at Different Depth Ranges in the Global Ocean^aa Flux rates are given in units of mg C m⁻² d⁻¹.

Depth, km	Average	Standard Deviation	Minimum	Maximum	N
0.1–0.25	17.5	6.4	9.0	26.0	5
0.25–0.5	7.2	3.6	1.3	14.1	14
0.5–1	7.4	4.7	0.55	18.9	40
1–2	6.8	5.4	0.38	21.1	30
2–3	4.5	3.3	0.40	12.8	38
3–4	5.4	3.7	0.09	13.7	38
>4	2.0	1.7	0.07	5.7	15

• a Flux rates are given in units of mg C m⁻² d⁻¹.

Table 4. Average Rate of Change of the Vertical POC Flux to Depth in the Intermediate and Deep Ocean Derived From Sediment Trap Data^aa Positive and negative values indicate the percent by which flux observed in deeper traps exceeds or lags flux at shallower traps within the depth ranges indicated.

	Percent Δ Flux km^−1
	1–2 km	2–3 km	>3 km
Greenland and Norwegian Seas	−79	76	–
NE Atlantic/NABE	−3.1	63	−54
Sargasso Sea/BATS	−79^bb Average over 1 to 3 km range.	−79^bb Average over 1 to 3 km range.	−61
Subarctic Pacific/OSP	12^bb Average over 1 to 3 km range.	12^bb Average over 1 to 3 km range.	−46
North Central Pacific gyre/HOT	55	55	−16
South China Sea	−12	−38	60
Arabian Sea, oceanic	−	68	−39
Arabian Sea, coastal	41	−32	–
Equatorial Pacific (±2° latitude)	17	7.6	42
Equatorial Pacific (±5° latitude)	−24	9^cc Average over 2 to >3 km range.	9^cc Average over 2 to >3 km range.
Equatorial Pacific (±9°–16° latitude)	−29	−17^cc Average over 2 to >3 km range.	−17^cc Average over 2 to >3 km range.
Panama Basin	1.2	43	9.8
Southern Ocean, Atlantic sector	–	−45	−39
NW Africa	−46	35	−30

• a Positive and negative values indicate the percent by which flux observed in deeper traps exceeds or lags flux at shallower traps within the depth ranges indicated.
• b Average over 1 to 3 km range.
• c Average over 2 to >3 km range.

[22] Where flux data are available for both the shallow and deep ocean, fluxes to depths between 0.5 and 1 km are highly correlated with fluxes to depths greater than 1 km (Figure 3). In contrast to Yu et al. [2001] and Scholten et al. [2001], these correlations suggest that sediment trap data from depths between 0.5 to 1 km may be as “valid” as those from below 1 km. The variability of upper ocean fluxes indicates that the conditions and forcings that serve to create and attenuate variability in flux to depth are concentrated within 1 km below the base of the photic zone. While sediment-trap-derived flux to depth data may be internally consistent, primary production and export are poorly correlated with flux (Table 5). This lack of correlation suggests that using primary production or export alone may not allow for accurate predictions of flux to depth.

Figure 3

Relationships between sediment-trap-derived POC flux to depth estimates within different depth ranges for regions listed in Table 2.

Table 5. Correlation Matrix for Primary Production, Export From the Base of the Photic Zone (Table 2), and Flux to Various Depth Ranges (Table 1)

	Primary Production	Export	Flux (0.5–1 km)	Flux (1–2 km)	Flux (2–3 km)	Flux (3–4 km)	Flux (>4 km)
Primary production	1
Export	0.74	1
Flux (0.5–1 km)	0.16	0.25	1
Flux (1–2 km)	0.03	0.32	0.93	1
Flux (2–3 km)	−0.10	0.15	0.85	0.79	1
Flux (3–4 km)	0.02	0.20	0.91	0.78	0.87	1
Flux (>4 km)	0.41	0.62	0.99	0.95	0.78	0.91	1

4.2. Predicted Versus Observed Flux to Depth

[23] We compare observed flux to depth, derived from sediment trap data, to predicted flux to depth using the flux algorithms of Suess [1980] and Martin et al. [1987] derived from primary production and export beneath the base of the euphotic zone, respectively (Figure 4). Both the Suess [1980] and Martin et al. [1987] relationships describe flux as decreasing more rapidly with depth in the upper ocean and less rapidly with depth in the deep ocean. Disagreement between predicted and observed flux to depth is generally greatest in regions where export and primary production are larger. Disagreement is also larger in the shallow subphotic ocean (0.1–2 km) and approaches observed values with increasing depth. Within this depth range, overall disagreement is typically largest with the Suess [1980] relationship, which overestimates observed flux to depth by an average of 8 times and a maximum of 36 times. Here, the Martin et al. [1987] relationship overestimates observed flux by an average of 6 times and a maximum of 23 times. Worldwide, the Suess [1980] and Martin et al. [1987] relationships both overestimate and underestimate flux to the deep ocean (>1.5 km). The Suess [1980] and Martin et al. [1987] relationships predict flux to the deep ocean by factors ranging from 0.2 to 6 (3.1 average) and from 0.4 to 14 (3.3 average) times observed values, respectively.

Figure 4

Regional observed (solid circles; Table 1) and radiochemically corrected (open circles) sediment-trap-based flux to depth, export (diamonds; Table 2), and primary production (squares; Table 2). Regional empirical algorithms describing the observed flux of particulate organic carbon to depth (equations (8), solid black line, and (9), dashed black line) are compared to those of Martin et al. [1987] (solid red line) and Suess [1980] (dashed red line).

[24] The Martin et al. [1987] relationship predicts observed flux to depth well in some regions (Arabian Sea, Southern Ocean/Atlantic sector, and subarctic Pacific/OSP) and overestimates flux to depth in other regions studied. In regions with export values >200 mg C m⁻² d⁻² (NW Africa, South China Sea NE, and Atlantic/NABE), disagreement is larger. Here, predictions range between 5 and 14 times the observed values, increasing with increasing export. These disagreements suggest that organic carbon flux varies less with depth in the deep ocean than predicted by the Martin et al. [1987] relationship. Furthermore, in all regions, the use of the Martin et al. [1987] relationship with a variable exponent (instead of a fixed −0.858) underestimates observed flux to depths greater than 2 km. This implies that the use of a power law equation to describe flux to depth may overestimate the rate of change of flux with depth in the deep ocean. The inaccuracy of flux to depth exhibited by the Martin et al. [1987] and Suess [1980] relationships is probably due, in part, to the refinement of POC export and primary production methodologies over the past two decades.

4.3. Storage Efficiency p and s Ratios

[25] In order to compare flux of organic carbon between ocean regions, flux to depth is normalized to local primary production and export:

The p and s ratios describe the vertical transport of organic carbon normalized relative to surface water primary production and export, respectively (Figure 5). These ratios gauge the efficiency of the retention of sinking organic carbon, the storage efficiency, within the deep ocean.

Figure 5

Observed flux to depth normalized to (a) primary production (p ratio: flux/primary production) and (b) export (s ratio: flux/export) for the major time series, JGOFS, and other oceanographic studies.

[26] In general, different ocean regions exhibit different groupings of p and s ratios. These groupings are more easily distinguished with s ratios than with p ratios. Large storage efficiencies of exported carbon (s ratios) are found in the Panama Basin, subarctic Pacific/OSP, Arabian Sea, and Southern Ocean/Atlantic sector, regions. The NW Africa, north central Pacific gyre/HOT, NE Atlantic/NABE, and South China Sea regions commonly exhibit low exported carbon storage efficiencies. Fluxes per unit of carbon produced during primary production (p ratios) are largest in the Panama Basin, Southern Ocean/Atlantic sector, and Arabian Sea regions. Smaller storage efficiencies of carbon derived from primary production are found in the NW Africa, north central Pacific gyre/HOT, and NE Atlantic/NABE regions. Worldwide, the fraction of carbon fluxed to the deep ocean (>1.5 km) ranges from 0.10 to 8.8% (1.1% average) of primary production and from 0.28 to 30% (5.7% average) of export.

[27] There is little direct relationship between primary production and export and corresponding p and s ratios. Correlations between primary production or export and the regional p and s ratios are poor (all r² < 0.5). Overall, p and s ratios tend to be larger at regions with larger rates of primary production and export, suggesting that flux to depth may be more efficient during periods of low primary production and export.

4.4. Regional Empirical Algorithms

[28] The flux of POM to depth is mediated by complex physical and biological interactions that are not well understood. These interactions control the depth-dependent rates of POM sinking and remineralization. In order to better relate flux to depth to the relevant controls, we propose describing flux to depth using a minimum of parameters that include sinking and remineralization rates. Thus, flux to depth (z) can be described by the exponential equation:

where D is the instantaneous rate of decay and W is the sinking rate, both in the same units of time [Banse, 1990, 1994].

[29] POM fluxing to depth is composed of many individual particles with a continuum of individual decay and sinking rates. As POM sinks, labile organic fractions are remineralized rapidly, allowing refractory material to become more important with depth [Asper and Smith, 1999; Bishop, 1989]. We recognize both labile and refractory fractions of sinking POM and approximate this process by modifying (5) to describe flux as the sum of two fractions of POM (s₁ and s₂), with decay rates (D₁ and D₂) and sinking rates (W₁ and W₂) normalized to export (equation (4)):

We rely on several assumptions to minimize the number of parameters needed to describe this process. Flux to depth typically decreases rapidly with increasing depth in the upper ocean. The magnitude of change of flux with increasing depth is greatly diminished in the deep ocean relative to the upper ocean. We assume that the rapid upper ocean decrease of flux to depth constitutes the rapid decay of fresh labile organic material and the near-constant fluxes approached in the deep ocean correspond to rapidly sinking organic material and/or the refractive resistance organic material to regeneration. Hence we approximate flux to depth where a fraction of POM sinks quickly or remineralizes slowly enough that decay is essentially zero (when D₂ approaches zero or W₂ approaches infinity) and (6) reduces to

For simplicity, we define k_s = (D₁ / W₁) m⁻¹ and rewrite (7):

Equation (8) describes the relationship between export and flux to depth in various oceanographic regions. To describe the relationship between primary production and flux to depth, we substitute the p ratio (equation (3)) for the s ratio in (6), (7), and (8), yielding

The parameter k_p is defined in a similar manner to k_s. Limitations of this approximation of the flux to depth process include decay and sinking rates, which change with depth (i.e., a variable k with depth). This approximation may be insensitive to gradual decreases in the rate of flux with increasing depth, as observed in the deep ocean [Martin et al., 1987; Suess, 1980]. The use of constant depth-independent terms (p₂ and s₂) is based on our regional analysis, which indicates that flux in the deep ocean increases with depth almost as often as it decreases with depth (Table 2).

[30] The results of the exponential fits of (8) and (9) applied to observed and radiochemically corrected regional sediment-trap-derived flux to depth, export, and primary production rate estimates are described in Tables 6a and 6b and Figures 4 and 6. These equations typically predict flux to the deep ocean within 20% of the average observed values. The p₂ and s₂ parameters describe the more rapidly sinking and/or more refractive portions of flux to the deep ocean and sediment surface. These parameters generally approximate minimum p and s ratios. The p₂ refractive/rapidly sinking fractions of primary production are above average in the Panama Basin and Southern Ocean/Atlantic sector regions. The s₂ refractive/rapidly sinking portions of export are above average in the Panama Basin, Southern Ocean/Atlantic sector, Subarctic Pacific/OSP, and Arabian Sea regions. The p₂ and s₂ values indicate that from 0.21 to 4.9% (1.2% average) of primary production and from 0.37 to 11% (4.6% average) of export reaches the deep ocean (>1.5 km), similar to calculated p and s ratios.

Figure 6

Regional profiles of (a and b) observed and (c and d) radiochemically corrected POC flux to depth normalized to primary production (PP) and export below the photic zone (equations (1), (2), (8), and (9); Tables 6a and 6b) for the upper 2000 m depth. Regions shown include the following: 1, Greenland and Norwegian Seas; 2, NE Atlantic/NABE; 3, Sargasso Sea/BATS; 4, Subarctic Pacific/OSP; 5, north central Pacific gyre/HOT; 6, South China Sea; 7, Arabian Sea (regional average); 8, equatorial Pacific (regional average); 9, Panama Basin; 10, Southern Ocean, Atlantic sector; and 11, NW Africa. Profiles and model results derived from the Suess [1980] and Martin et al. [1987] relationships are included for comparison (lines 12 and 13 (thick lines), respectively).

Table 6a. Regional Empirical Parameters Describing Particulate Organic Carbon Flux to Depth Normalized to Primary Production Using Unadjusted and Radiochemical Calibrated Sediment Trap Fluxes^aa See text. The e-folding length scale of remineralization is given by 1/k_p. All r² values are >0.93, except for the Greenland and Norwegian Seas region. Equation: p(z) = + p₂.

	Observed			Radiochemical Calibration
	p1	1/kp, m	p2	p1	1/kp, m	p2
Greenland and Norwegian Seas	1.53	116	0.00440	1.00	550	0.00104
NE Atlantic/NABE	1.37	125	0.00485	1.19	223	0.00481
Sargasso Sea/BATS	1.83	221	0.0298	3.89	102	0.00617
Subarctic Pacific/OSP	10.8	29.4	0.00627	5.80	39.7	0.00813
North central Pacific gyre/HOT	61.4	36.4	0.00210	21.7	48.8	0.00259
South China Sea	1.88	157	0.00688	1.53	230	0.00779
Arabian Sea (regional average)	1.89	119	0.00747	1.52	181	0.00744
Equatorial Pacific (regional average)	3.20	107	0.00437	3.21	105	0.0109
Panama Basin	1.69	122	0.0463	1.43	170	0.0491
Southern Ocean, Atlantic sector	1.99	56.8	0.0148	1.73	70.4	0.0172
NW Africa	1.61	104	0.00238	1.44	136	0.00301

• a See text. The e-folding length scale of remineralization is given by 1/k_p. All r² values are >0.93, except for the Greenland and Norwegian Seas region. Equation: p(z) = + p₂.

Table 6b. Regional Empirical Parameters Describing Particulate Organic Carbon Flux to Depth Normalized to Export Using Unadjusted and Radiochemical Calibrated Sediment Trap Fluxes^aa See text. The e-folding length scale of remineralization is given by 1/k_s. All r² values are >0.93, except for the Greenland and Norwegian Seas region. Equation: s(z) = + s₂.

Region	Observed			Radiochemical Calibration
	s1	1/ks, m	s2	s1	1/ks, m	s2
Greenland and Norwegian Seas	1.28	178	0.0145	1.00	529	0.00398
NE Atlantic/NABE	1.32	137	0.0121	1.14	275	0.0119
Sargasso Sea/BATS	6.55	74.1	0.00372	1.32	599	0.0112
Subarctic Pacific/OSP	2.24	77.6	0.0896	1.00	439	0.0939
North central Pacific gyre/HOT	8.15	70.9	0.0132	3.09	122	0.0149
South China Sea	1.74	177	0.0124	1.42	278	0.0135
Arabian Sea (regional average)	1.40	183	0.0813	1.10	455	0.0617
Equatorial Pacific (regional average)	2.02	167	0.0443	1.53	243	0.0832
Panama Basin	1.44	146	0.101	1.21	228	0.112
Southern Ocean, Atlantic sector	1.33	105	0.0904	1.15	167	0.0938
NW Africa	1.55	117	0.00393	1.38	153	0.00487

• a See text. The e-folding length scale of remineralization is given by 1/k_s. All r² values are >0.93, except for the Greenland and Norwegian Seas region. Equation: s(z) = + s₂.

[31] The p₁, s₁, 1/k_p, and 1/k_s parameters describe the more labile and/or more slowly sinking portion of flux. Together they approximate the remineralization length scale, or e-folding depth of flux. Smaller p₁ and s₁ values decrease the portion of flux entering the ocean at depth z₀. Larger 1/k_p and 1/k_s values deepen the inflection point of the curve describing flux, below which flux approaches p₂ or s₂ values and becomes constant with depth. The net effect is to diminish the supply of carbon to the upper ocean.

[32] These parameters are poorly constrained because of a lack of and potential unreliability of flux data in the mesopelagic zone. Parameterizations are highly sensitive to whether or not upper ocean sediment trap data are deemed valid. Applying the radiochemical correction generally lowers p₁ and s₁ values and increases 1/k_p and 1/k_s values, with the net result of decreasing flux to depth in the upper ocean. In this case, parameters derived may be considered to estimate maximum flux to depth. For both the observed and radiochemically corrected flux data, values for p₂ and s₂ greater than 0.01 are only found where primary production and export are less than 400 mg C m⁻² d⁻¹ and 150 mg C m⁻² d⁻¹, respectively. These results suggest that where primary production and export are low, a greater portion of the transported material is refractory and/or rapidly sinking.

4.5. Effect on Carbon Storage

[33] We performed a set of simulations in a one-dimensional ocean model to assess the influence of the various remineralization profiles (equations (1), (2), (8), and (9); Tables 6a and 6b) on the retention of CO₂ generated by the remineralization of POC below the photic zone. We allowed a pulse of POC to remineralize in the ocean interior according to each regional profile and simulated how long it would take for the ΣCO₂ generated at depth to be transported back up to the photic zone.

[34] The ocean model used here is described by Caldeira et al. [1998]. The ocean is represented by a box diffusion model [Oeschger et al., 1975; Siegenthaler, 1983], which is essentially a one-dimensional column representing mean oceanic vertical transport. We make no attempt to simulate regional differences in ocean circulation. The diffusion coefficients for the vertical transport of dissolved matter vary with depth. The diffusion coefficients were chosen [Caldeira et al., 1998] such that the change in ocean ¹⁴CO₂ inventory between 1945 and 1975 matches the estimated 1975 bomb radiocarbon inventory [Broecker et al., 1995] of 305 × 10²⁶ atoms and the modeled 1975 ocean mean and surface ocean Δ¹⁴CO₂ matches the basin-volume-weighted mean of the natural plus bomb Δ¹⁴CO₂ values measured in the Geochemical Ocean Sections Study (GEOSECS) program [Broecker et al., 1985]. The Δ¹⁴C is a normalized and ¹³C-adjusted ¹⁴C/¹²C ratio, and δ¹³C is a normalized ¹³C/¹²C ratio [Broecker and Peng, 1982]. This tuning yields a vertical eddy diffusion coefficient of 8820 m² yr⁻¹ at the base of the mixed layer diminishing with an e-folding length scale of 500 m to a minimum of 2910 m² yr⁻¹ at the ocean bottom. Similar parameter values were used in previous simulations [Oeschger et al., 1975; Siegenthaler, 1983; Hesshaimer et al., 1994].

[35] Analysis using this one-dimensional ocean model indicates that the residence time of biogenic carbon may vary up to 2 orders of magnitude depending on the regional location of carbon fixation and export. The Suess [1980] and Martin et al. [1987] relationships both yield residence times similar to the maximum of regional relationships derived in this study (Figure 7). Parameters derived from the radiochemically corrected flux to depth data typically yield longer residence times than the observed flux to depth data. Simulations based on regional profiles indicate that the residence time of POC exported below the euphotic zone in the deep ocean may span 2 orders of magnitude.

Figure 7

Predictions of (a and b) observed and (c and d) radiochemically corrected carbon storage below the photic zone using a one-dimensional ocean model (see text). Regions shown include the following: 1, Greenland and Norwegian Seas; 2, NE Atlantic/NABE; 3, Sargasso Sea/BATS; 4, Subarctic Pacific/OSP; 5, north central Pacific gyre/HOT; 6, South China Sea; 7, Arabian Sea (regional average); 8, equatorial Pacific (regional average); 9, Panama Basin; 10, Southern Ocean, Atlantic sector; and 11, NW Africa. Profiles and model results derived from the Suess [1980] and Martin et al. [1987] relationships are included for comparison (lines 12 and 13 (thick lines), respectively).

5. Conclusions

[36] We demonstrate that the commonly applied Martin et al. [1987] and Suess [1980] relationships approximate flux to depth in several locations. However, these relationships overestimate flux to depth in most ocean regions. The overestimation of flux to depth is more pronounced with the Suess [1980] equation. Our observations indicate that these constant power law and rational relationships used in current large-scale oceanographic models generally overestimate deepwater POC fluxes and hence underestimate particle regeneration in the water column. For a given ocean circulation model, the use of these relationships could result in unreliable estimates of new production and residence time of exported carbon in the deep ocean.

[37] We illustrate regional variability in the biological pump’s ability to store carbon in the ocean below the photic zone. Flux to depth normalized to primary production and export indicates from 0.1 to 8.8% (1.1% average) of primary production and from 0.28 to 30% (5.7% average) of export enters the deep ocean (>1.5 km). The use of new region-specific empirical flux to depth algorithms, which in part parameterize the lability of settling particulate organic matter, in a one-dimensional ocean model suggests up to 2 orders of magnitude of variability in the efficiency of carbon storage in the ocean. Applying a radiochemical correction to the observed POC flux to depth tends to lengthen remineralization length scales and subsequently increase the predicted residence time of carbon in the ocean. Global relationships between primary production and export out of the euphotic zone and flux to depth remain unclear.

[38] From our analysis it is not clear that export is any more directly related to flux to depth than is primary production. Consideration of other processes which modify the efficiency of the biological pump may improve this relationship. Modifications to the sinking flux include zooplankton swimming and excretion at depth [Bishop, 1989; Bishop et al., 1986; Vinogradov, 1970], the association of fluxing material with hard parts and mineral surfaces [Ittekkot et al., 1991; Lee et al., 1999], variability of the mixed layer depth [Fischer et al., 1996], mass sedimentation events [DiTullio et al., 2000; Kemp et al., 2000; Smetacek, 2000], plankton community structure [Boyd and Newton, 1999], primary production seasonality [Berger and Wefer, 1990; Lampitt and Antia, 1997], the rate of microbial decomposition [Arnosti et al., 1998; Cho and Azam, 1988; Laws et al., 2000], and non-Redfield uptake of nutrients and decomposition of organic matter [Arrigo et al., 1999; Pahlow and Riebesell, 2000].

Acknowledgments

[39] We thank the following researchers for their inspiration and advice: James Bishop (Laurence Berkeley National Laboratory), Kevin Arrigo (Stanford University), and Donald Olson (RSMAS, University of Miami). This research was conducted as a portion of M. Lutz’s Ph.D. dissertation and was supported by a number of sources, including the NSF ROAVERRS program, DOE Center for Research on Ocean Carbon Sequestration, Stanford University McGee Foundation, and International JGOFS Program (Ocean Biogeochemical Modeling Course, Bangalore, India).