Annual net community production and the biological carbon flux in the ocean



The flux of biologically produced organic matter from the surface ocean (the biological pump), over an annual cycle, is equal to the annual net community production (ANCP). Experimental determinations of ANCP at ocean time series sites using a variety of different metabolite mass balances have made it possible to evaluate the accuracy of sediment trap fluxes and satellite-determined ocean carbon export. ANCP values at the Hawaii Ocean Time-series (HOT), the Bermuda Atlantic Time-series Study (BATS), Ocean Station Papa (OSP) are 3 ± 1 mol C m−2 yr−1—much less variable than presently suggested by satellite remote sensing measurements and global circulation models. ANCP determined from mass balances at these locations are 3–4 times particulate organic carbon fluxes measured in sediment traps. When the roles of dissolved organic carbon (DOC) flux, zooplankton migration, and depth-dependent respiration are considered these differences are reconciled at HOT and OSP but not at BATS, where measured particulate fluxes are about 3 times lower than expected. Even in the cases where sediment trap fluxes are accurate, it is not possible to “scale up” these measurements to determine ANCP without independent determinations of geographically variable DOC flux and zooplankton migration. Estimates of ANCP from satellite remote sensing using net primary production determined by the carbon-based productivity model suggests less geographic variability than its predecessor (the vertically generalized productivity model) and brings predictions at HOT and OSP closer to measurements; however, satellite-predicted ANCP at BATS is still 3 times too low.

1 Introduction

Net biologically produced organic matter in the upper ocean integrated over an annual cycle is equal to annual net community production (ANCP), which is defined as the difference between net primary production (NPP) and respiration by heterotrophs (animals and bacteria). ANCP is limited by the rate of delivery of essential nutrients (N, P, and Fe) to the sunlit surface ocean and the efficiency with which these nutrients are either metabolized or returned to deeper waters. ANCP is regarded as equal to the biologically produced organic carbon flux that escapes the upper ocean on an annual basis (the “biological pump”).

It is important to know the geographic distribution and the mechanisms that control ANCP for three main reasons. First, the net transfer of carbon from the surface ocean and atmosphere to depth results from uptake of preformed (as opposed to regenerated) nutrients in the euphotic zone. Since the availability of preformed nutrients in the euphotic zone is greater in higher latitudes, atmospheric pCO2 levels and thermocline O2 concentrations should be more sensitive to ANCP changes in these regions. The impression given in global maps of ANCP determined from satellite color and in ocean global circulation models (GCMs) is that organic matter export is much higher in high latitudes, but this has not been observationally confirmed (see later).

The second reason is that the mean respiration depth of organic matter once it exits the euphotic zone may depend on the fraction of the export that is by particulate organic matter and dissolved organic matter since respiration rates of these different phases are probably not the same. Deeper organic matter respiration enhances the drawdown of atmospheric pCO2 [e.g., Yamanaka and Tajika, 1997; Kwon et al., 2010] and reduces oxygen in the upper thermocline.

Finally, the mechanism(s) influencing the transfer of particulate organic matter to depth, either association with CaCO3 [Armstrong et al., 2002; Klaas and Archer, 2002] or ecosystem structure [Henson et al., 2012; Francois et al., 2002] are latitude dependent. A change in the production of CaCO3 or ecosystem structure resulting from a progressively warmer and more acidic ocean will have a regional impact on organic carbon export. It is essential to understand the geographic relationships among ANCP, CaCO3 production, and ecosystem structure to predict future changes in ocean hypoxia [Hoffmann and Schellnhuber, 2009].

The goal of this paper is to show the relationships among experimental measurements of ANCP from metabolite mass balance in the upper ocean, carbon export as determined from sediment trap fluxes, and satellite predictions of the biological pump. This type of analysis will hopefully promote calibrations that will eventually yield accurate estimates of ANCP from remote-sensing measurements. Experimental determinations of ANCP have occurred in relatively few locations because it involves measurements that resolve seasonal changes. For this reason I focus the discussion here on three locations where there have been enough measurements to resolve seasonal variability: The Hawaii Ocean Time-series (HOT) in the subtropical North Pacific [Church et al., 2013], the Bermuda Atlantic Time-series Study (BATS) in the subtropical North Atlantic [Church et al., 2013] and Ocean Station Papa (OSP) in the subarctic North Pacific [Timothy et al., 2013]. The take home messages of this paper are (1) mass balances of oxygen and dissolved inorganic carbon (DIC) (and its carbon isotopes) result in the same estimates of ANCP within the errors of these measurements at the time series sites, (2) upper ocean carbon fluxes and ANCP cannot be effectively determined using sediment trap fluxes, even if they do derive an accurate particulate carbon flux, because of the varying and important role of DOC and biological transport mechanisms, and (3) predictions of ANCP from satellite measurements and recent algorithms for net primary production (NPP) and the NCP:NPP ratio are close to experimentally determined ANCP at HOT and OSP but not at BATS.

2 Background

2.1 Modes of Organic Carbon Export

Net community production of organic carbon in the euphotic zone of the ocean is exported to depths where it is respired by bacteria and animals. The depth where net organic matter production gives way to net respiration (the compensation depth) is variable depending on the availability of nutrients and light penetration. Net respiration rates decrease exponentially below the compensation depth as measured by both the decrease with depth in particulate organic matter caught in sediment traps [e.g., Martin et al., 1987; Buesseler and Boyd, 2009] and observations of oxygen utilization rate (OUR) [e.g., Stanley et al., 2011; Jenkins, 1998; Martz et al., 2008]. The critical depth for defining the ANCP in the upper ocean is associated with the depth of the mixed layer in winter because organic matter that escapes the summer mixed layer but is respired above the winter mixed layer will be returned to the ocean surface by entrainment as the mixed layer deepens. Thus, part of the net oxygen and carbon production in summer is returned to the mixed layer in winter rather than being exported to the deep ocean. This has been demonstrated using seasonal measurements of pCO2 [Kortzinger et al., 2008] and oxygen [Quay et al., 2012] in the North Atlantic and by mass balances at the time series sites [e.g., Emerson et al., 2008; Emerson and Stump, 2010].

The three main modes of organic matter export from the upper ocean are particle fluxes, mixing of dissolved organic matter along a gradient that decreases with depth, and active transport by animals. Studies of dissolved organic carbon (DOC) in global circulation models (GCMs) indicate that measured gradients both horizontally in surface waters and with depth below the euphotic zone can be reproduced if about two thirds of the organic matter produced in the euphotic zone is DOC with a respiration lifetime of about 6 months [Yamanaka and Tajika, 1997]. This model yields a global DOC export that is about 20% of the total, which agrees with global observations [Hansell and Carlson, 2001]. However, it has also been shown that in the subtropics the fraction of organic carbon export due to DOC transport can be greater than one half of the total carbon export [Doval and Hansell, 2000; Abell et al., 2000].

Biologically mediated carbon export is mostly by diurnally migrating zooplankton which feed in the surface at night and then exit the euphotic zone for greater depths during the day. This process can be up to a third of the flux in regions where the migrators are an important component of the ecosystem [e.g., Steinberg et al., 2008]

2.2 Global Estimates of ANCP From GCMs and Satellites

Globally averaged values of ANCP have been determined from two main sources: GCMs that include a biogeochemical component and remote sensing by satellites. A summary of these estimates (Table 1) indicates remarkably consistent values. Twelve different GCM determinations result in a mean and standard deviation of 12 ± 2 Pg C yr−1 (Table 1). This result depends on the depth chosen for the base of the compensation layer in the models because of the exponential decrease of respiration below this depth. Najjar et al. [2007] derive an NCP of 14 Pg C yr−1 in five separate models using 75 m as the compensation depth. This value would be ~11 Pg C yr−1 at 100 m based on the depth dependence of respiration used in the models they describe. Lack of conformity in the export depth would cause some difference among model results in Table 1, but even with this caveat, they are quite consistent. I suspect the reason for the global model similarity in GCM-determined ANCP is that in models the dissolved flux of nutrients from deeper waters is controlled by water transport, which is tuned to agree with observed circulation-tracer distributions (i.e., bomb-produced carbon-14 and CFCs). Uniformity in global model-derived ANCP, or carbon export from the euphotic zone, was also pointed out by Boyd and Trull [2007], where they noted that model-based regional differences in ANCP were more variable than the global flux.

Table 1. Model and Satellite-Based Estimates of Annual Net Community Production (ANCP, in Moles C m−2 yr−1)a
 Zonally Averaged Annual NCPSource and Comments
 0–15° 15–30° 30–45° 45–60°
Global ANCP (Pg C yr−1)(mol C m−2 yr−1)
  1. a

    Global annual values are in column one; columns 2–5 contain zonally weighted averages for four latitude ranges (equatorial, 0–15°; subtropical, 15–30°; subtropical/subpolar, 30–45°; and subpolar, 45–60 °). Satellite-based values are described in the text. The results of Dunne et al. [2007] are presented in units of Pg C deg−1 yr−1 which we converted to units of mol C m−2 yr−1 using ocean area for each degree of latitude (m2 deg−1). The mean and standard deviation of the estimates are in bold numbers.

  2. b

    Two different estimates for total dissolved phosphorus in the ocean.

  3. c

    The different atmosphere models.

  4. d

    The mean of five models with circulations that yield acceptable circumpolar deep water 14C content. Values zonally averaged for 10°N–10°S, 10°–40°, and 40°–90°.

  5. e

    Mean and standard deviation of carbon fluxes in the above models. Results from Najjar et al. [2007] are not included in the 15°–30° and 30°–45° averages.

  6. f

    NPP from Behrenfeld and Falkowski [1997].

  7. g

    NPP from Marra et al. [2003].

  8. h

    NPP from Carr [2002].

10, 13b2.5 1.2 2.0 2.2Anderson and Sarmiento, 1995 (Nutrient restoring)
10 Yamanaka and Tajika, 1996 (Nutrient restoring)
112.4 1.1 3.2 2.3Six and Maier-Reimer, 1996 (Nutrient and T model)
13, 10c3.0 1.9 3.4 4.6Bopp et al., 2001 (Nutrient and T model) (Ecosystem model)
113.4 2.9 3.0 1.8
122.0 1.0 3.0 2.9Moore et al., 2002 (Ecosystem)
13 Schlitzer, 2004 (Nutrient restoring, inverse model)
14 ± 5d5.7 2.9 2.4Najjar et al., 2007 (Mean of five models) (Nutrient restoring)
12 ± 2e4.2 ± .5 1.6 ± .8 2.9 ± .5 2.6 ± .8 
Satellite Based
11 Laws et al., 2000
10f1.7 1.3 3.1 3.2Dunne et al., 2007
11g2.0 2.3 2.0 3.1(NCP/NPP from Dunne et al., 2005)
12h2.6 1.5 2.9 3.8 
9–13 Laws et al., 2011
9.4 Westberry et al., 2012
11 ± 12.1 ± .5 1.7 ± .5 2.7 ± .6 3.4 ± .4 

Estimates of carbon export from the surface ocean using satellite data (Table 1) are determined by relating ocean color and particle backscatter from satellites to net primary production (NPP) [e.g., Behrenfeld and Falkowski, 1997; Carr, 2002; Marra et al., 2003; Behrenfeld et al., 2005; Westberry et al., 2008]. It is then assumed that carbon export is equal to NPP times a euphotic zone respiration efficiency (NCP:NPP), which has been derived from both models [Laws et al., 2000] and empirical observations [Dunne et al., 2005, 2007]. We refer to this method as “satellite-based ANCP”. There are many fewer estimates of the global biological ANCP from satellite research; however, those that have been done suggest a value of 11 ± 1 Pg C yr−1 (n = 6), which is quite consistent with values determined from GCMs.

Zonally averaged values of carbon export for the equatorial (15°N–15°S), subtropical (15°–30°), subtropical/subpolar transition (30°–45°), and subpolar (45°–60°) regions are presented along with the global values in Table 1. These values were determined by simply integrating under the curves of ANCP versus latitude presented in these papers. Integration over latitude bands smooths more extreme latitudinal variations indicated in the papers. Zonally averaged values in the subtropics are about half those at the equator, transition, and subpolar regions. Ocean color in the surface ocean is lower in regions with nutrient-poor surface waters, and this is probably the reason for lower satellite-derived ANCP in the subtropical latitudes. Global circulation models also predict lower ANCP in the subtropics probably because these low-resolution models do not have the mechanisms necessary to transport nutrients into the euphotic zone [e.g., Lévy et al., 2012].

2.3 Experimental Determination of ANCP by Mass Balances of Carbon, Oxygen, and Nitrate

There are no standards for flux measurements so the only way to judge absolute accuracy is to determine the value by a variety of methods and accept those that agree. Annual net community carbon production at ocean time series sites has been determined by mass balances of oxygen, nitrate, and the stable isotopes of dissolved inorganic carbon (DI12C and DIC13C). A summary of the annual mass balance results is presented in Table 2, and a comparison of these results with measurements of particle fluxes by sediment traps is discussed in the next section. Error estimates by the authors for individual annual determinations in Table 2 range from ±20 to 50%. The standard deviation of the individual mean annual values is up to ± 50% reflecting both real interannual variability and measurement/model error. The most clear geographical differences in these data are that the mean value for ANCP measured in nearshore waters of the California Coast at the California Cooperative Oceanic Fisheries Investigations (CalCOFI) grid is 2–3 times those measured in the open ocean. The CalCOFI mean, 6.4 ± 1.9 mol C m−2 yr−1, is itself a composite of values that differ by a factor of 5 among nearshore, upwelling, and outer stations in the grid [Munro et al., 2013]. ANCP values in the northeast subarctic Pacific (station P; 2.1 ± 0.4 mol C m−2 yr−1) and northeast subtropical Pacific (HOT; 2.5 ± 0.7 mol C m−2 yr−1) are the same to within errors of the measurements. Values are presented for the equatorial Pacific in Table 2 even though they are not annual estimates because there is probably little seasonality in this location. Eastern equatorial Pacific mean values are about twice those in the western equatorial Pacific, but the mean for the whole region is not significantly different from values to the north in the subtropical ocean given the variation of the estimates. Comparisons of ANCP determined by nitrate mass balance between western and eastern side of the subarctic North Pacific [Wong et al., 2002; Goés et al., 2004] indicate values up to 50% higher in the western basin. Oxygen and DIC mass balance determinations of ANCP in the Northwest Atlantic (BATS) are higher than other open ocean stations probably because this is in a region of deep winter mixing.

Table 2. Annual Net Community Production (ANCP) Determined by Oxygen, δ13C-DIC Mass Balance, and NO3- Drawdowna
LocationAnnual NCP (mol C m−2 yr−1)
  1. a

    Values in parentheses are the mean and standard deviation of the individual values.

  2. b

    O2 mass balance [Emerson, 1987; Emerson et al., 1991, and Emerson and Stump, 2010].

  3. c

    NO3 drawdown in surface waters [Wong et al., 2002].

  4. d

    O2 mass balance [Emerson et al., 1995; Hamme and Emerson, 2006; Emerson et al., 2008].

  5. e

    Carbon isotope mass balance [Quay and Stutsman, 2003].

  6. f

    Carbon isotope mass balance [Keeling et al., 2004].

  7. g

    Carbon isotope mass balance [Brix et al., 2006].

  8. h

    Lee [2001] based on summertime DIC change determined from global DIC, pCO2, and Alk estimates in the mixed layer. Values are in the vacinity of the time series stations.

  9. i

    Carbon isotope mass balance (150°W–170°W) [Quay et al., 2009]. The range in the equatorial Pacific represents values before and after correction for horizontal δ13C gradients.

  10. j

    Carbon isotope mass balance (1.2°W, 170°W; 1.7°W; and 4.4°W, 140°W) [Zhang and Quay, 1997].

  11. k

    O2/Ar mass balance (110 and 95°W) [Hendricks et al., 2005].

  12. l

    Summary of results from Joint Global Ocean Flux Study, Equatorial Pacific [Quay, 1997]. The lower values are for El Niño conditions and the higher value for non–El Niño. A representitative value of 5 was used in the mean calculation.

  13. m

    Stanley et al. [2010].

  14. n

    Oxygen mass balance from the CARbon dioxide IN the Atlantic Ocean data [Quay et al., 2012].

  15. o

    O2 mass balance [Jenkins and Goldman, 1985; Spitzer and Jenkins, 1989; Jenkins and Doney, 2003].

  16. p

    Carbon isotope mass balance [Gruber et al., 1998].

  17. q

    Oxygen utilization rate (OUR) using 3He ventilation times [Stanley et al., 2011].

  18. r

    O2/Ar mass balance [Munro et al., 2013]. The range includes both nearshore and offshore values. The mean is for the entire CalCOFI grid.

E. Subarctic N. Pac.2.1b, 1.6b, 2.5b,3.0c(2.3 ± 0.6)
(OSP) 50°N, 145°W
E. Subtropical N. Pac.2.7d, 1.4d, 3.3d, 2.5e, 2.3f, 3.1g,1.7–2.2h, 2.3i(2.5 ± 0.7)
(HOT) 23°N,158°W
E. Equatorial. Pacific3.2–5.2i, 1.7j, 1.2j, 4.4j, 2.5k, 2–7l(3.3 ± 1.8)
W. Equatorial Pacific1.5m 
Subtrop/Subarc. N. Atl.2.8 ± 2.7n(2.8 ± 2.7) 
(40°N–65°N, 10°W–60°W)
W. Subtropical N. Atl.3.4o, 3.9o, 5.6o, 3.8p,4.9q,2.1r,2.6–3.5h(3.8 ±1.2)
(BATS) 32°N,64°W
Coastal3.3–17.0r(6.4 ± 1.9) 
CalCOFI (~30°N–34°N, 118°W–125°W)

The two first-order observations of the Table 2 compilation are that (1) experimentally determined, open ocean ANCP measurements are in the range of 2–4 mol C m−2 yr−1, and the values in the subtropical oceans are about the same as in other areas [see also Juranek et al., 2012]; and (2) nearshore ANCP values are at least 3 times the open ocean means. I use data from the time series stations to evaluate the satellite-based ANCP in the section 3.

2.4 Sediment Trap Organic Carbon Fluxes

The largest component of the organic carbon flux out of the euphotic zone is particulate and has been measured in hundreds of field experiments using sediment traps (see Lutz et al. [2002] for a global summary). The accuracy of these values however has been challenged by comparison of the fluxes with those determined by thorium mass balance [e.g., Benitez-Nelson et al., 2002; Buesseler et al., 2007]. Sediment trap particulate organic matter flux at BATS and HOT have been measured since 1988 and have been recently summarized by Church et al. [2013] who indicated nearly consistent annually averaged values of 0.9 ± 0.5 and 0.8 ± 0.3 mol C m−2 yr−1 at 150 m, respectively. At OSP shallow sediment trap results from 200 m were recently summarized by (D. A. Timothy et al., submitted manuscript, 2013) and indicate an annual average flux of 0.5 ± 0.4 mol C m−2 yr−1. A summary of the annually average particulate organic carbon fluxes is presented along with ANCP determined by mass balance for the period 1992 to 2008 in Figure 1. We choose this time interval because it encompasses the period when there are estimates of NPP from satellite remote sensing measurements of Sea-viewing Wide Field-of-view Sensor (SeaWIFS) (1997–2007; see later). Clearly, the sediment trap fluxes are lower than mass balance estimates of ANCP. This should be the case because of DOC export and active transport by zooplankton, which do not show up in the traps, and because the trap depths are 50–100 m below the base of the depth interval used to calculate ANCP. In the next section I evaluate the importance of these factors and present a comparison of the sediment trap fluxes and ANCP determined from metabolite mass balance.

Figure 1.

Experimental estimates of ANCP (mol C m−2 yr−1) as a function of year from 1992 to 2008. Large symbols indicate ANCP determined from mass balances of O2, NO3, and the stable carbon isotopes of dissolved inorganic carbon (δ13-DIC). The method is indicated next to the symbols, and references are given in Table 2. Smaller symbols are annually averaged particulate organic carbon fluxes from sediment trap deployments at HOT, BATS (150 m), and OSP (200 m). Data were downloaded from the compilations on the HOT and BATS websites and are from [Timothy et al., 2013] for OSP.

3 Discussion

3.1 The Relationship Between ANCP and Sediment Trap Flux

3.1.1 The Mass Balance Expressions

A quantitative comparison of experimental measurements of ANCP and sediment trap fluxes requires considering the roles of DOC flux, transport by migrating zooplankton, and the amount of respiration between the base of the ANCP layer (here the winter mixed layer depth) and the shallow sediment trap. I use a simple one-dimensional model to approximate the relationship between ANCP and the particle flux at the sediment trap depth.

At steady state over a period of 1 year, net community production should equal the downward flux of organic carbon, Fh (mol C m−2 yr−1), at the base of the ANCP-production region, z = zh (m). If we assume also that vertical processes dominate carbon production and respiration in the upper ocean (more about this later), then ANCP also equals the net-integrated organic carbon degradation, J (mol m−3 yr−1), below the depth, zh:

display math(1)

The flux terms can be divided into the three main modes of transport: particles, FP, dissolved organic carbon, FD, and zooplankton migration, FZ.

display math(2)

Buesseler and Boyd [2009] reviewed sediment trap data from eight locations where particle flux was determined by the 234Th method (the time series locations at HOT and OSP were included in this compilation, but not BATS). They adopt an exponential model for the particle flux at depth below the euphotic zone:

display math(3)

where Fo is the flux at the base of the “euphotic zone”, zo, and z* is the characteristic respiration length scale derived from 234Th and particulate organic carbon measurements. The scaling in equation (3) is independent of the depth chosen for zo, and as argued earlier, mass balance-determined ANCP refers to biological processes that occur above the depth of the winter mixed layer, so here we replace zo with zh. In the North Pacific at HOT and OSP the mixed layer maximum depth is about 110 m because of the halocline in that region of the ocean. Assigning the depth, zh, to regions of deep winter mixed layers is not as easy because the surface waters are sometimes not totally ventilated to the depths of the winter mixed layer. If they were, gases would be equilibrated with the atmosphere to this depth each year. An example of this complication is the BATS location where winter mixed layers are frequently 300–500 m, but the annually averaged 3H-3He ages reach ~2 years at depths as shallow as 150–200 m [Stanley et al., 2011]. For this reason I assign 150 m (zh = 150 m) as the base of the layer over which ANCP is determined at BATS. We shall see that uncertainty in this depth does not greatly affect the results of the comparison at this site.

Density surfaces below the winter mixed layer crop out poleward of the time series sites, so there is a natural disconnect between what happens in the surface and its relationship to deeper waters. To retain a sense of one-dimensionality in our treatment of the relationship between ANCP and respiration below, I couple the ANCP with only the top 100 m of the upper thermocline (zlzh = 100 m). The fraction of particulate organic matter that degrades in this region is defined as follows:

display math(4)

Using z* values determined by Buesseler and Boyd [2009] for HOT and OSP (216 m and 77 m) and (l h) = 100 m in equation (4) results in 40 to 70%, respectively, of the particle flux being degraded in the top 100 m below the winter mixed layer depth at these locations. This means that about half of the particulate organic matter exiting the upper ocean at these locations degrades in the top 100 m of the net respiration zone, which is similar to that observed in earlier respiration depth dependencies [Martin et al., 1987; Jenkins, 1998].

In our simple model organic carbon fluxes, F, and respiration rates, J, in the 100 m region below the winter mixed layer depth are related by as follows:

display math(5)

I determine the fraction of organic matter degradation due to DOC respiration in this depth interval by comparing the changes in DOC and apparent oxygen utilization (AOU). The change in dissolved organic carbon, ∆DOC, is the difference between the value in the ocean mixed layer and the measured value, and similarly, AOU is equal to the difference between the saturation and observed concentrations (AOU = [O2]sat−[O2]). Assuming a “Redfield” ΔO2/Δ C ratio during organic matter degradation, rO2:OC = 1.45 [Hedges et al., 2002] allows one to present the fraction of organic carbon respired due to DOC as follows:

display math(6)

If we assume that the depth dependence of the fluxes of all three components of the carbon flux—POC, DOC, and biological transport are the same—then the ratio on the left side of equation (6) is also equal to the ratios of the fluxes:

display math(7)

This assumption of equal degradability may not be true, but the relative lability of reactive DOC and POC during respiration is not well known, and it is probably a respectable first approximation. There is no basis for assuming this depth dependence is correct for biological migrations; however, we shall see that this mode of transport is the least important at the time series locations and plays a significant role only at OSP. The uncertainty introduced by these assumptions is difficult to determine, but it clearly increases the element of approximation in this calculation. Because we are dealing with data sets that are approximately one-dimensional, there are compromises between maintaining a believable local calculation and assumptions about relative respiration rates deeper in the water column. Combining equations (7) and (2) yields

display math(8)

Rearranging and substituting ANCP for Fh (equation (1)) results in an expression for the particle flux in terms of known quantities:

display math(9)

The anticipated particle flux at the depth of the sediment traps can be calculated by correcting the results of this expression for the sediment trap depth using the relationship in equation (3) with zo = zh.

3.1.2 The Role of DOC

The relationship between ΔDOC and AOU in the depth regions of 100–250 m at the time series stations are presented in Figures 2-5. ΔDOC is equal to the difference between measured DOC and the value present on this density horizon when it was at the surface (the “preformed DOC”). Because there are not a lot of seasonal surface ocean DOC measurements, I have used the surface values at the time series sites as the value for the surface outcrop. In the subtropical and subarctic North Pacific at HOT and OSP the local surface water assumption is probably correct as there is little indication that there are surface DOC gradients in the regions where densities in the range σθ = 24.0–25.5 (HOT) [see Abell et al., 2000] and σθ = 26.4–26.9 (OSP) outcrop. This may not be the case in the western North Atlantic at BATS. There is a decreasing trend in surface water DOC concentration at higher latitudes where waters in the upper 250 m at BATS outcrop (the density range σθ = 26.0–26.5). Data from Climate Variability and Predictability (CLIVAR) A-22 ( indicate a decrease in surface water DOC of about 5 µmol C kg−1 from 33°N, 65°W (σθ = 25.5; DOC ~ 60 µmol C kg−1) to 40°N, 70°W (σθ = 26.5; DOC = 55 µmol C kg−1). Using this trend as the preformed DOC value would decrease the ΔDOC/AOU measured value and the importance of DOC as the respiration substrate. I have decided not to make this correction because the available surface water data are summertime values, and it is the winter outcrop that is the end-member. Furthermore, we shall see that our calculation of the importance of DOC in carbon export at BATS is in the lower range of previous estimates.

Figure 2.

Apparent oxygen utilization (AOU = [O2,sat]−[O2]) versus ΔDOC (µmol kg−1) at the Hawaii Ocean Time-series (HOT, depth interval of 100–250 m) for the period 2001–2011. Oxygen and DOC data were taken from the HOT online data server. AOU was calculated from the oxygen data and O2 saturation values compiled by Garcia and Gordon [1997]. ΔDOC = (DOCsurf–DOC), where DOCsurf is the DOC measured at the ocean surface (see text). Color coding represents (a) depth (in meters) and (b) density (σθ).

Figure 3.

The same as in Figure 2 except for oxygen and DOC data from the Bermuda Ocean Time-series Study (BATS, depth interval = 100–250 m).

Figure 4.

The same as in Figure 2 except for oxygen and DOC data from CLIVAR P16 between 40°N and 50°N and 100–250 m.

Figure 5.

Comparison of ANCP determined by measurements at the Hawaii Ocean Time-series (HOT) station and by satellite-based determinations over the period of 1992 to 2008 (SeaWIFS satellite date exists for the period 1997–2007). VGPM means vertically generalized productivity model [Behrenfeld and Falkowski, 1997], and CbPM means carbon-based productivity model [Behrenfeld et al., 2005; Westberry et al., 2008]. The line labeled “Laws00” indicates NCP was determined from the satellite-derived NPP and the NCP:NPP determined by the ecosystem model of Laws et al. [2000]. The line labeled “Dunne07” indicates the NCP:NPP ratio was determined using the empirical relationship of Dunne et al. [2005]. Large symbols are annual estimates determined from oxygen and carbon isotope mass balances. Smaller symbols are from sediment trap fluxes at 150 m.

There is some scatter in the ΔDOC/AOU trend at HOT (Figure 2) which we attribute to occasional inaccuracy in the measurements. Ignoring the flyers on either side of the trend the ΔDOC/AOU ~ 0.45 (µmol C kg−1/ µmol O2 kg−1), which results in a DOC contribution to organic matter degradation of JD/J = 0.66 (equation (7))—two thirds of the organic matter degraded in the 110–210 m depth interval is degraded by DOC rather than particle flux. While this value is higher than the global average value of ~ 20%, it is consistent with previous results from this region of Abell et al. [2000] who used data independent of those at HOT. The ΔDOC/AOU trend at HOT crosses both depth and isopycnal surfaces between σθ = 24.0–25.0 (Figure 2) indicating that the waters in this depth range are vertically mixed on the time scale of the respiration process (CFC ages of water above σθ = 25.0 are less than 5 years, [Warner et al., 1996]).

The ΔDOC/AOU trend at BATS is very different—indicating a lesser importance for DOC degradation during respiration, and the trends fall along both depth and density surfaces rather than across them (Figure 3). The ΔDOC/AOU ratios must represent seasonal changes observed on different density horizons at BATS without significant mixing across them. The ΔDOC/AOU trend along density surfaces is ~ 0.09, yielding a JD/J ratio of 0.13 (equation (7)). This value is somewhat lower than the range (0.24–0.47) determined by Hansell and Carlson [2001] for the period between 1992 and 1998. These authors assumed all the DOC export to the 100–250 m depth range occurs in the months of November–February and that the DOC is then oxidized from February to August. By decoupling in time the changes in DOC and AOU and not considering changes on density surfaces, their estimate may be more vulnerable to biases of horizontal advection.

Dissolved organic carbon is not a routine measurement at the OSP time series and there are, to our knowledge, no tabulated data as part of the time series database. There are measurements in this region however along CLIVAR P16 ( which crosses the subarctic gyre along longitude 152°W—7° to the east of OSP. The trend in ΔDOC versus AOU for the 100–250 m range between 45°N and 50°N (Figure 4) from these data also indicate that changes along density gradients are different than the entire ensemble of data, but in this case the ΔDOC/AOU trends are stronger along density gradients rather than weaker as was the case at BATS. The importance of isopycnal transport between the surface ocean and top of the thermocline in this region has been demonstrated previously for dissolved inorganic carbon and alkalinity [Emerson et al., 2011]. Interpreting the gradients along isopycnals centered on σθ = 26.45 gives a ΔDOC/AOU ratio of ~0.09, yielding a JD/J ratio of 0.14, which is similar in magnitude to the results from BATS but much less than the gradient at HOT. Less than 20% of the AOU in the upper pycnocline at OSP is created by degradation of DOC.

3.1.3 Prediction of Particle Flux From ANCP

A summary of particulate organic carbon fluxes predicted from equation (9) is presented in Table 3. Experimentally determined ANCP values are those reported in Table 2. Carbon fluxes due to zooplankton diurnal migrations have been measured to be 0.06 mol C m−2 yr−1 at BATS [Steinberg et al., 2000] and 0.1 mol C m−2 yr−1 at HOT [Al-Mutairi and Landry, 2001], indicating that this process at these locations is less than 10% of the total carbon flux. There have been no similar zooplankton flux estimates at OSP; however, they have been determined in the western subarctic Pacific [Steinberg et al., 2008] to be much larger than those at HOT and BATS (0.6 mol C m−2 yr−1). Since diurnally migrating copepods are abundant in both the western and eastern subarctic Pacific [Harrison et al., 2004], I assume similar carbon fluxes by this mechanism at OSP. This results in the biological component being about one third of the total estimated NCP (Table 3).

Table 3. Comparison of the Particulate Organic Carbon Flux Calculated From the Annual Net Community Production, FPztrap, and Particulate Organic Carbon Fluxes Measured in Sediment Traps, Fpmeas, at the Three Time Series Sitesa
Locationzhztrapz*math formulaANCPFZhFPcalcFPmeas
  1. a

    Depths are in meters, and fluxes are in mol C m−2 yr−1. Calculated particle fluxes are from equation (9) using values for the depths of the winter mixed layer, zh, and sediment traps, ztrap, and the characteristic respiration depth scale (equation (3)), z*. The fraction of the carbon flux by DOC is math formula (equation (7)), and flux of carbon by migrating zooplankton is FZh.

  2. b

    Steinberg et al. [2000].

  3. c

    Al-Mutairi and Landry [2001].

  4. d

    Steinberg et al. [2008]. Measured in the western subarctic Pacific.

  5. e

    Church et al. [2013].

  6. f

    Timothy et al. [2013].


Comparison of predicted and measured organic carbon fluxes requires a correction for the difference in depths of the base of the winter mixed layer and the sediment traps (compare columns 8 and 9 in Table 3). At HOT and BATS this correction is small, but at OSP the depth difference is 90 m and the characteristic respiration depth, z*, is only 77 m resulting in the flux at the trap depth being about one third of that at the base of the winter mixed layer. The final comparison of the predicted and measured values (the last two columns in Table 3) indicates that they are within the errors of the calculation and measurements at HOT and OSP. At HOT the threefold difference between ANCP and particle fluxes is due almost entirely to the export of DOC. At OSP the fourfold difference is due partly to the export of carbon by migrating animals and partly because the traps were deployed ~ 90 m deeper than the winter mixed layer depth. The threefold to fourfold difference between ANCP and particulate organic carbon flux at BATS is unexplained. All of the factors that would cause this in our simple model are insufficient to explain this result.

One of the important uncertainties in my estimates of the particle flux from ANCP measurements at BATS is choice of the depth of the ANCP layer in this region of deep mixed layers and the depth dependence of respiration below it. Kadko [2009] determined shallow aphotic-zone respiration rates at BATS to be 2.2 mol C m−2 yr−1 using 7Be as a ventilation tracer. If this value is representative of an annual flux, then it would bring the sediment trap fluxes and ANCP estimates at BATS in Table 3 much closer to balancing. Kadko's measurements were made over one summer (July–November, 2007), and he assumed the ocean in this region is a one-dimensional system during this period. Stanley et al. [2011] recently determined the depth dependence of the oxygen utilization rate at this location using the 3H-3He dating method. They define two regions of exponential decay in oxygen utilization rate with the shallowest having a characteristic respiration depth using equation (4) of greater than 1000 m—much deeper than any of the values in the compilation of Buesseler and Boyd [2009] and not consistent with the suggestion of very intense shallow respiration rates at this location [Kadko, 2009].

The results of this analysis indicates that measured ANCP and particulate organic matter fluxes at HOT and OSP are explainable. This does not rule out inefficiencies in trap fluxes that are in the range of a factor of 2 or so [Buesseler et al., 2007; Benitez-Nelson et al., 2002] as our analysis has uncertainties that probably could not distinguish these differences. The comparison at BATS is different because predictions from ANCP and measured particulate organic carbon fluxes simply do not agree. This must be because fluxes at this location are dominated by horizontal processes, sediment traps are particularly ineffective in this region, or (and) shallow respiration rates are much greater than assumed. An important conclusion of this comparison is that even if sediment trap fluxes are accurate in some locations, it is not possible to scale them up to derive ANCP without thoroughly evaluating the geographically variable roles of both the DOC flux and carbon transported by migrating animals.

3.2 Comparison of Experimental and Satellite Estimates of ANCP

Net primary production and net community production have been estimated from satellite remote sensing measurements [e.g., Westberry et al., 2012; Laws et al., 2011], and in recent years these values have been tested using upper ocean metabolite and isotope mass balances [see Juranek and Quay, 2013; Juranek et al., 2012]. Estimating NCP from satellite data involves evaluating NPP from satellite remote sensing and then multiplying by separate estimates for the NCP:NPP ratio. Satellite remote sensing estimates of NPP are compiled for the world's ocean on the ocean productivity website at Oregon State University (, where there are two ways of calculating NPP. The first is from the vertically generalized productivity model (VGPM), which uses ocean color to estimate chlorophyll concentration, and this is assumed to be a measure of phytoplankton biomass. This model is tuned to 14C measurements of primary production, mostly from the North Atlantic Ocean [Behrenfeld and Falkowski, 1997]. A more recent method of calculating NPP from satellite remote sensing is the carbon-based productivity model (CbPM), which uses both color to estimate chlorophyll concentrations and optical backscatter to determine phytoplankton biomass [Behrenfeld et al., 2005; Westberry et al., 2008]. These quantities are combined to determine plankton growth rate. The latter method is believed to be more accurate because it does not rely on empirical relationships describing phytoplankton assimilation efficiencies (i.e., carbon fixed per unit chlorophyll per unit time), which are required when chlorophyll concentration is assumed to reflect biomass concentration [Behrenfeld et al., 2005].

Monthly averages values for NPP, determined by both VGPM and CbPM, sea surface temperature, and chlorophyll concentration, were downloaded from the Oregon State website. These values were averaged over an area that incorporated 1–2° of latitude and longitude in each direction of the time series sites. Average monthly values of NPP were multiplied by the NCP:NPP ratio calculated from chlorophyll concentration, temperature, and NPP (see below) and then annually averaged. The NCP:NPP ratio was determined using two methods. The ecosystem model of Laws et al. [2000] determines the NCP:NPP, or “e” ratio as a function of temperature and NPP. A more recent compilation of observations by Dunne et al. [2005, 2007] compared NPP determined from 14C primary production incubations with NCP estimated from both mass balances and sediment trap fluxes to evaluate what they call the particle export ratio “ep”. We will show that these two estimates differ by at most 30% at the time series sites with the empirical result [Dunne et al., 2005] being lower. A more complete comparison of these two methods for determining the NCP:NPP ratio is presented in Dunne et al. [2005].

Comparison of experimentally determined ANCP estimates with those from the satellite-based ANCP at HOT (Figure 5) indicate that values based on the CbPM method for determining NPP are nearly twice those determined from VGPM. Westberry et al. [2008] demonstrated that the CbPM method does a much better job than VGPM of reproducing experimentally measured 14C primary production at HOT suggesting that this satellite algorithm is an improvement over VGPM in the subtropical Pacific Ocean. The NCP:NPP ratio determined from the ecosystem model of Laws et al. [2000] is about 25% greater than the “ep” ratio of Dunne (0.16 compared to 0.12, respectively). I prefer the higher, more theoretical values because they are independent of sediment trap data which underestimate the organic carbon export. ANCP determined from CbPM and Laws et al. [2000] matches the experimental results at HOT to within the error of the measurements.

Experimental and satellite-based ANCP for the area in the vicinity of OSP (Figure 6) reveal opposite VGPM-CbPM differences to those described at HOT. In this case NPP results from VGPM were much higher than those determined using the CbPM method, but once again the latter procedure brings the satellite estimates closer to the experimental measurements. NPP determined from both VGPM and CbPM at OSP are strongly seasonal (results not presented), but the summertime values of the CbPM are lower, and more in line with measurements reviewed by Harrison et al. [2004]. The NCP:NPP ratio in this area determined from the model of Laws et al. [2000] varies from 0.2 in winter to 0.45 in summer, whereas it is about 0.30 without much seasonal variation from the Dunne et al. [2007] prediction. This difference results in an annually averaged NCP:NPP value that is again about 30% higher for the Laws method (Table 3).

Figure 6.

As in the caption for Figure 5 except here for the region surrounding the Canadian time series location at OSP in the subarctic Pacific Ocean. Satellite-based ANCP estimates are plotted using the NCP:NPP ratio from Laws et al., 2000 only, but results using both the Laws et al. [2000] and Dunne et al. [2005] methods are presented in Table 4.

Table 4. Summary of ANCP Values Determined by Oxygen and Carbon Mass Balance and by Satellite-Based Methods at Three Locations Where Annual Experimental Data Exista
LocationANCPmeas (mol C m−2 yr−1)NPPsat (mg C m−2 d−1)NCP:NPPANCPsat (mol C m−2 yr−1)
  1. a

    ANCP values are the product of NPP from the CbPM algorithm and NCP:NPP ratios from Laws et al. [2000]. Values in parentheses indicate seasonal ranges for VGPM and CbPM. Measured and predicted ANCP values are in bold.

OSP2.3 ± 0.6446 (130–900)340 (70–650)0.31 (0.18–0.49)0.31 (0.28–0.35)4.6 ± 0.53.2 ± 0.6
HOT2.5 ± 0.7265 (170–330)503 (420–550)0.16 (0.16–0.17)0.13 (0.12–0.15)1.4 ± 0.02.5 ± 0.1
BATS3.8 ± 1.2321 (150–600)281 (100–350)0.17 (0.16–0.18)0.14 (0.09–0.19)1.5 ± 0.11.4 ± 0.1

It has been previously noted [Emerson et al., 2008; Emerson and Stump, 2010] that satellite-determined ANCP using the VGPM algorithm predicts much greater values in the subarctic North Pacific than in the subtropical area, which is not consistent with observations at OSP and HOT. This problem appears to have gone away with the upgrade to the CbPM algorithm.

Comparison of experimental and satellite-based ANCP estimates at BATS (Figure 7) indicates that inconsistencies between experimental and satellite-based results still exist. Both VGPM and CbPM estimate NPP to be lower in the spring than 14C measurements at BATS (data not presented). Underprediction of NPP in the spring results in a low annually averaged NPP and thus ANCP. This problem is discussed by Westberry et al. [2008] and is, in part, due to underestimation of spring chlorophyll concentrations in the satellite data. Another potential problem is overestimation of photoacclimation resulting from differences between physiologically relevant mixing layer and the seasonal thermocline.

Figure 7.

As in Figure 5 except here for the region of the Bermuda Atlantic Time-series Study (BATS). Satellite-based ANCP estimates are plotted using the NCP:NPP ratio from Laws et al., 2000 only, but results using both the Laws et al. [2000] and Dunne et al. [2005] methods are presented in Table 4.

4 Conclusions

Annual net community production determined by euphotic zone mass balance at open ocean time series sites is 3 ± 1 mol C m−2 yr−1. Mass-balances estimates of ANCP from oxygen, carbon isotopes, and nitrate agree to within the error of these determinations where they have been done at the same location. Coastal values measured at a single location, CalCOFI, are about 3 times greater than the open ocean values. While there are too few ANCP measurements to be confident in geographical variability, there is presently little detectable meridional variation. This result is in stark contrast to ANCP values determined in global circulation models and some satellite-based methods where there are variations of at least a factor of 2 between high latitudes, equator, and the subtropics.

Particulate fluxes calculated from ANCP at the time series stations, after considering carbon export due to DOC flux and zooplankton migration and adjusting for the trap depth, are within the errors of the sediment-trap-determined fluxes at both HOT and OSP but about a factor of 3 greater than the values measured at BATS. About three quarters of the organic matter flux out of the upper ocean at HOT is via DOC and perhaps a third of the flux at OSP is due to animal migrations. At BATS neither of these mechanisms account for more than 20% of the measured ANCP flux. Based on just these three locations, geographic variations in mechanisms of carbon export are highly variable. Sediment traps may sometimes be collecting a representative organic carbon flux, but the values at these locations are 3–4 times less than ANCP because of the importance of other modes of transport and the depth of sediment trap deployment. Thus, it is presently not possible to derive accurate estimates of ANCP using sediment trap collections.

Satellite-based ANCP from the chlorophyll-only algorithm (VGPM) is inaccurate at all three time series locations based on comparisons to experimentally determined values. VGPM-based ANCP values at HOT were too low and at OSP too high. The newer algorithm that includes both optical backscatter and chlorophyll to determine NPP (CbPM) increases the estimate at HOT and decreases the value at OSP so that they are nearer experimentally determined ANCP values. Thus, the satellite-based contrast in organic matter export between the subtropics and subarctic ocean is now smaller and more in line with experimental observations. Prediction of ANCP from space at BATS is still too low by both methods indicating that there is still much work to do. Calibration of the satellite-based method for determining ANCP and organic matter export to the level where it captures known geographic variability will require a much richer global distribution of experimentally determined annual-estimates of the flux.


The author would like to thank David Munro for teaching me to analyze satellite productivity data. Ken Buesseler, Matt Church, and Mike Behrenfeld made helpful comments on the original draft. This research was supported by NSF grants OCE-0850286 and OCE- 1129112.