Nonsinking Organic Matter Production in the California Current

Abstract

Productive eastern boundary upwelling systems such as the California Current Ecosystem (CCE) are important regions for supporting both local and remote food webs. Several studies have reported on the temporal and spatial variability of primary production and gravitational export in the CCE. However, few studies have quantified the partitioning of net primary and new production into other reservoirs of detrital organic matter. This study tested the hypothesis that nonsinking detrital reservoirs are an exportable reservoir of new production in the CCE with samples collected by the California Cooperative Oceanic Fisheries Investigation survey between 2008 and 2010. Water column gradients in nitrate (NO₃⁻) and total organic carbon (TOC; which excludes sinking particulate organic carbon) were used to estimate potential rates of new production (P_New) and TOC production (P_TOC), respectively. The P_TOC:P_New varied between 0.16 and 0.56 and often increased with indicators of enhanced autotrophic production. At times, surface stratification was also correlated with elevated P_TOC:P_New. In the most productive, inshore region, P_TOC exceeded previously reported sinking export rates, which identified TOC as a quantitatively significant repository of exportable carbon in the CCE. However the sum of P_TOC and sinking export for these productive regions was less than both P_New and oxygen-based estimates of net community production. These results imply that nonsinking reservoirs alone are not sufficient to explain observed imbalances between production and export for the most productive CCE regions.

Key Points

• Production in the California Current Ecosystem generates 5–30 micromoles per liter of total organic carbon primarily in the dissolved phase
• Near the coast an average of 11 ± 3% of net primary production and 21 ± 4% of new production is partitioned into accumulating surface carbon
• Sinking plus nonsinking organic carbon reservoirs balance net community production in all but the most productive regions of the California Current Ecosystem

Plain Language Summary

The ocean's biological pump is typically quantified as the organic carbon that quickly sinks, that is, is “exported,” out of the surface lighted zone to be subsequently sequestered in the deep ocean. Recent studies have shown that other forms of organic matter produced by phytoplankton can also contribute to carbon export. In this study, we quantified how much new production and net primary production was channeled into nonsinking reservoirs such as dissolved organic carbon and suspended particulate organic carbon in the productive eastern boundary California Current Ecosystem. To match the data coverage provided by our organic carbon measurements we used satellite data to calculate net primary production and used measured depth profiles of nitrate together with model-derived upwelling velocities, to determine new production. We quantified the amount of nonsinking organic matter that accumulated in surface waters following production and found that the timescale of accumulation enabled this reservoir to participate in export. In some regions, as much carbon was present in the accumulated nonsinking reservoir as was quantified as sinking particulate carbon. We also found that lateral export from the productive coastal region was a potentially important pathway that could carry nutrients and carbon in organic matter to less productive waters.

1 Introduction

Under steady state conditions the export of organic matter from the surface ocean is expected to be equivalent to “new production” (Eppley & Peterson, 1979; Laws et al., 2000). Sinking particulate organic matter (POM), which can comprise intact phytoplankton cells, and the fecal pellets and remnants of zooplankton and planktivorous fish, is the most prominent form of export (Eppley & Peterson, 1979). In the California Current Ecosystem (CCE), enhanced Ekman transport and mesoscale variability can spatially and temporally decouple new and export production, where decoupling mechanisms include lateral offshore transport of inorganic nutrients and organic species such as biomass and detrital organic matter (Hales et al., 2005; Landry et al., 1997; Plattner et al., 2005; Stukel et al., 2011, 2013). Modeling efforts focused on the California Current have further shown that eddy-induced subduction of inorganic and organic nitrogen into the upper thermocline can reduce nearshore production at the site of upwelling but fertilize communities at the edge of the gyre (Gruber et al., 2011; Letscher et al., 2013).

Small particles that are suspended in seawater (~1–70 μm, Abramson et al., 2010) are isotopically and chemically distinct from fast-sinking particles typically associated with vertical export (Abramson et al., 2010; Altabet, 1988; Druffel et al., 1992). This suspended POM can represent a significant fraction of the carbon and nitrogen accumulating in coastal environments following periods of upwelling (Hill & Wheeler, 2002; Walker & McCarthy, 2012). Dissolved organic matter (DOM) contributes the highest concentration of organic carbon to the water column in most oceanic environments (Halewood et al., 2012; Hill & Wheeler, 2002; Walker & McCarthy, 2012; Wear et al., 2015). Both DOM and suspended POM are treated here as separate and distinct from fast-sinking particles, but both have been shown to participate in lateral export following subduction and advection (Abell et al., 2000; Carlson et al., 2004; Duarte et al., 2013; Letscher et al., 2013, 2016; Omand et al., 2015; Siegel et al., 2014). In temperate regions of high convective overturn, the subduction of nonsinking organic matter (OM) can account for as much carbon export as the sinking organic matter flux (Carlson et al., 1994). In situ measurements and models suggest that nonsinking OM can exceed 25% of total export production from the surface ocean (Hansell & Carlson, 1998; Najjar et al., 2007; Stukel et al., 2017), and support up to 67% of apparent oxygen utilization in the mesopelagic (Emerson, 2014). Assessing the magnitude of primary production that is channeled into nonsinking organic carbon could reconcile the observed regional decoupling of new and export production in the CCE, the potential for which will be addressed here.

In this study, we calculated seasonal new production (P_New) and total organic carbon (TOC) production (P_TOC) for four hydrographically distinct regions of the southern CCE. These calculations relied on measured concentrations of nitrate and TOC from samples collected between 2008 and 2010 on quarterly California Cooperative Oceanic Fisheries Investigations (CalCOFI) survey cruises. Here the TOC measurement primarily quantified organic carbon that was homogenously distributed in a small volume of whole seawater, which includes dissolved organic carbon (DOC) and some suspended particulate organic carbon (POC).

The relative partitioning of P_New and satellite-based net primary production (NPP) into the TOC reservoir was calculated and compared to regional rates of sinking POC flux to examine the relevance of nonsinking OM as a repository of newly fixed carbon. Variations in P_TOC:P_New and P_TOC:NPP were further examined in the context of available biological and physical data to identify potential mechanisms that influenced partitioning across the CCE.

2 Materials and Methods

2.1 Study Region and Sampling Period

Samples for this study were collected between 2008 and 2010 on quarterly CalCOFI cruises augmented by the CCE Long Term Ecological Research (LTER) program. Several of the measured and derived variables used throughout the manuscript can be found in supporting information Table S2. TOC and suspended POC samples were collected from up to 20 depths between 0 and 515 m for TOC and from 0 to 100 m for suspended POC. Hydrographic data and biological and inorganic chemical parameters were available for all relevant depths for ~75 stations (Figure 1). NPP was also estimated for this same time period from satellite data. Samples were collected on CalCOFI cruises from April 2008 (i.e., cruise 200804); August 2008 (200808); October 2008 (200810); January, July, and November 2009 (200901, 200907, and 200911, respectively); and January, April, and August 2010 (201001, 201004, and 201008, respectively). In total, 5,213 TOC and 1,966 POC samples were analyzed for this study.

Figure 1

Sampling locations in the CalCOFI domain. Colors depict the annual mean [TOC] (μM C) at the Chl a maximum depth (Chl a max) across the California Current Ecosystem. Greater [TOC] are observed in coastal upwelling dominated sites. California Current Ecosystem-CalCOFI subregions discussed in the text are identified in the figure as Inshore North (Chl a max depth of 14 ± 4 m), Inshore South (22 ± 6 m), California Current (33 ± 6 m), and Offshore (62 ± 12 m). CalCOFI = California Cooperative Oceanic Fisheries Investigation; Chl a = chlorophyll a; TOC = total organic carbon.

For ease of data analysis and discussion, sample collection locations within the typical CalCOFI sampling grid are designated as Inshore North (13 stations), Inshore South (28), California Current (18), and Offshore (18) regions (Munro et al., 2013; Taylor et al., 2015; Figure 1). Mean [TOC] distributions at the Chl a max depth suggest these designations appropriately differentiate regional influences on [TOC] (Figure 1). Generally, the California Current region represents the transition between the oligotrophic offshore and eutrophic inshore and loosely tracks the path of the California Current along the shelf break. However, the California Current meanders and does not always adhere to the above definition of fixed station location. Southern Inshore typically differs in surface circulation patterns depending on time of year but generally receives more surface water from the south (i.e., from Baja California) and the North Pacific Subtropical Gyre than the Northern Inshore region (Hayward & Venrick, 1998).

2.2 Hydrographic and Biogeochemical Analysis

As part of the ongoing CalCOFI program, continuous measurements of pressure, temperature, and conductivity were made from surface to 515 m, bottom depth permitting, via shipboard conductivity-temperature-depth-rosette-based sensors (Sea-Bird Electronics, Inc., Bellevue, WA, USA). Standard nutrient and hydrographic samples were collected via rosette casts at 20 standard depths. Measured variables include salinity (practical salinity scale, United Nations Educational, Scientific and Cultural Organization (Unesco), 1991), dissolved oxygen, phosphate, nitrate, nitrite, ammonium, silicate, and chlorophyll a (Chl a; see www.calcofi.org for analytical protocols). Chl a was measured via fluorescence on acetone extracts of POM (Goericke, 2002; Taylor et al., 2015). Nutrients were analyzed shipboard by continuous flow analysis within 2–16 hr of sample collection. Nitrite and nitrate are measured concurrently using a colorimetric assay on a cadmium reduction coil (after Armstrong et al., 1967, with modification). Accuracy of nitrate measured by this method is ±0.05 μM, as determined by standards.

2.3 TOC Analysis

Seawater for [TOC] analysis was collected into combusted 40-ml borosilicate vials and sealed using acid-washed vial caps with septa that had not been previously pierced. Seawater was immediately acidified to pH 2, where 40 ml of seawater received two drops (~200 μl) of trace metal grade 12N HCl (Fisher Scientific). Acidified samples were stored at room temperature in the dark until analysis via high-temperature combustion on a Shimadzu 500 V-CSN/TNM-1 (Shimadzu Corp., Kyoto, Japan) that was modified from the manufacturer's design. The Shimadzu combustion oven contained a quartz column filled with platinum (Pt) catalyst beads placed between two 0.5-cm (column height) plugs of glass wool, which improved peak shape (as recommended by other users). Combustion columns were preconditioned on 40–100 injections of filtered (0.2 μm) seawater until the baseline of measured carbon was stable. Columns and catalyst were generally changed every 120–150 samples. A magnesium perchlorate water trap was placed prior to the halogen trap and changed daily. CO₂-free carrier gas was used to precondition the column, and ultrahigh purity grade O₂ gas was delivered to the instrument as the carrier gas during sample analysis. Each acidified TOC sample was sparged for 2 min and measured following high-temperature combustion at 680 °C. During analysis, five 100-μl injections were made from a single sample reservoir, and samples were reanalyzed when the percent coefficient of variation (CV) of the best three injections was >5%. TOC measurements were calibrated using an 8-point calibration curve between 10 and 100 μM C of potassium phthalate in Milli-Q water. Milli-Q water and deep sea reference water were analyzed every 10 samples. National Science Foundation-supported deep Florida Strait Reference Standards (Batch 6FS—2006; Dennis Hansell, Rosenstiel School of Marine and Atmospheric Science, University of Miami) were used to determine machine accuracy. Reference standards measured here typically ranged between 41 and 44 μM C. If reference standard measurements were outside the expected range then TOC samples were reanalyzed. The reference standards were measured within a mean CV of 2.7%, which typically corresponded to an average TOC range of ±1.2 μM C. The expected concentration range for batch reference materials is provided at http://www.rsmas.miami.edu/groups/biogeochem/. Further details of sample analysis details and data verification protocols can be found at cce.lternet.edu/research. The relationship between TOC and density below the surface Ekman layer was examined to identify any outliers in the data set, and this served as a final quality check of each data point. This relationship is remarkably predictable in subsurface waters (i.e., >250 m), and so, samples that produced data points that strayed greater than one standard deviation from the best fit TOC:potential density relationship were reanalyzed to test whether the measured [TOC] was reproduced. As others have noted (e.g., Hansell et al., 1997), the small volume of the TOC measurement preferentially selects for the more homogeneously distributed DOC reservoir. As discussed in the results, when suspended POC concentrations ([POC]) are high, that is, during bloom conditions, [TOC] will also be influenced by [POC]. The amount of suspended POC captured by the TOC measurement cannot be predicted or easily quantified, but a comparison to suspended POC measurements (see below) unequivocally demonstrates that not all suspended POC is captured in the TOC measurements. Furthermore, the TOC measurement does not capture the sinking POC present during sampling (and any aggregation and settling after TOC sample collection is not included in the measurement). As such, our [TOC] measurement excludes fast-sinking organic carbon, consistently quantifies DOC, and sometimes includes suspended POC. As these are all operational definitions it is not possible to completely separate each fraction. It can be confidently stated that [DOC] is adequately captured by the TOC method but that in productive regions, the total nonsinking OM reservoir may be underestimated due to the exclusion of some suspended POC. For this reason, [TOC] is also compared to measured suspended [POC] in later discussions.

2.4 POC Analysis

Suspended [POC] and [PON] were determined on 1 to 2 L of seawater filtered through 25-mm precombusted (450 °C, 6 hr) Whatman GF/F filters (nominal pore size of 0.7 μm). Inorganic carbon was removed by acidifying the filters with HCl vapor, followed by oven drying, overnight. One half of the filter was used for PON and POC analysis on a Costech Elemental Analyzer (Costech Analytical Technologies, Valencia, CA, USA) according to standard protocols (see http://cce.lternet.edu/data/methods-manual).

2.5 Satellite-Estimated NPP

The low temporal and spatial resolution of shipboard ¹⁴C-based measurements of NPP (¹⁴C-PP) restricted the size of the TOC data set that could be compared to various indices of production. However, satellite-based estimates of NPP achieved the spatial coverage that matched our TOC data set. Satellite NPP was calculated using a transformed vertically generalized production model (VGPM) for the California Current region (termed VGPM-CAL) after (Kahru et al., 2009). The VGPM model estimates NPP using detected surface chlorophyll, photosynthetically active radiation, and temperature together with estimates of euphotic zone depth and photosynthetic efficiency (Behrenfeld & Falkowski, 1997). Briefly, monthly average VGPM data were obtained at 10-km grid spacing from Oregon State's Ocean Productivity website (http://www.science.oregonstate.edu/ocean.productivity/). Values were converted from the standard VGPM product to VGPM-CAL using the following formula: 10^log (VGPM-0.1924). The average of eight spatial estimates within a 20-km square of CalCOFI sampling location and within ±15 days of sampling was used to estimate the average satellite-based NPP value. Satellite-based estimates of NPP scale well with other methods of estimating primary producer biomass (e.g., [Chl a] at Chl a max depths; r² = 0.53 and p < 0.001). Munro et al. (2013) extensively discussed discrepancies between satellite NPP, ¹⁴C-PP (as estimated by CalCOFI's noon to dusk incubation) and geochemical gross oxygen production estimates for the region, and the reader is referred to that study for more information on the different methods used to estimate production.

2.6 ΔTOC and ΔPOC

“New” additions to the TOC and POC reservoir can be quantified by examining changes in stocks above background concentrations (i.e., ΔTOC and ΔPOC) as both DOC and POC include long-lived fractions that are unlikely to vary with indices of primary production (Druffel et al., 1992). For this calculation, [TOC] and [POC] were first binned into 0.25-kg/m³ potential density (σ_θ) intervals across the entire CalCOFI grid. Second, the minimum 10th percentile value of all measured [TOC] or [POC] within a density interval was identified, which resulted in a single density-based profile (Figure S1) that was designated as the background TOC or POC profile. We used the minimum 10th percentile value and not the minimum measured value to avoid biasing the background profile with outliers. Such an approach also accommodated the analytical error of our measurements. On average, the minimum 10th percentile value represented the 11th (±5) lowest [TOC] observation per isopycnal. To compute ΔTOC and ΔPOC for each density interval, the minimum TOC or POC profile was subtracted from the measured quarterly profile at each station. Studies at other locations computed this parameter using seasonal DOC profiles that constrained background concentrations (e.g., Sargasso and Ross Seas; Carlson et al., 1994, 1998). However, the CCE region is subject to high mesoscale and microscale eddy activity, which means that the seasonal resetting of surface ocean organic carbon concentrations is inconsistently detected.

Seasonal ΔTOC and ΔPOC presented for each subregion (Figure S2) represent compilations (in terms of median and range) of values within the surface σ_θ < 26.0 kg/m³. We determined the spread surrounding ΔTOC according to the following formulation:

(1)

The range (i.e., denoted by “±”) in ΔTOC encountered regionally or seasonally is determined as the 10th and 90th percentile values (e.g., Figure S2). The error for the 10th and 90th percentile ΔTOC values were propagated using the CV per measured sample and the CV for the region and season-specific 10th percentile value.

2.7 Biogeochemistry Estimation of the California Current Model Velocities

The Biogeochemistry Estimation of the California Current (BECCO) model currently under development at SIO (http://sose.ucsd.edu/CASE/) has been used here to estimate vertical and horizontal velocities for 2008–2010. The BECCO model has a 1/16-of-a-degree (~7 km) resolution and has 72 vertical depths of varying thickness. Five-day averages for those model locations within a 40-km radius of the relevant CalCOFI station were used to estimate vertical velocities. Validation of BECCO model velocities has been reported by Todd et al. (2011, 2012).

2.8 Principal Components Analysis

Data were divided spatially according to Figure 1 to provide statistically significant estimates of TOC production and new production. However, the complex hydrography within a CalCOFI subregion can blur signatures of ecological and biogeochemical processes. As such, consistent patterns in [TOC] variability were further examined by identifying other measurements that exhibited a significant correlation with [TOC]. The dominant variables driving [TOC] in each CalCOFI subregion were first identified using a stepwise multiple regression analysis using Minitab (v17.1.0). For regression analysis, surface mean values were first calculated from individual profiles of 18 variables measured by CalCOFI to be compared to surface mean [TOC]. Properties for each profile were averaged over σ_θ < 26.0 kg/m³, where 26.0 kg/m³ is defined as the depth of the pycnocline based on Jacox et al. (2016). In addition to the 18 near-surface variables available from the CalCOFI database, eddy kinetic energy (Kelly et al., 1998) and NPP were calculated separately using satellite data (see Table S1 for a list of variables tested). All 20 variables were then transformed by taking the cube root to normalize data sets.

For each subregion a subset of four predictor variables resulted in minimum derived Mallows' C_p values (Mallows, 1973) relative to the number of predictors selected across the region. For instance for the Inshore North region, surface mean [Chl a], apparent oxygen utilization, buoyancy frequency (0 over 200 m; Smith & Ferrari, 2009) and surface mean spiciness (Munk, 1981) were the best predictor variables. These correlations suggested that indices of surface stratification, upwelling and phytoplankton productivity could be driving [TOC] variability in this region. Overall, the combination of predictor variables explained 66% and 57% of the variance in TOC for the Northern and Southern Inshore regions, respectively. Less of the variance in the California Current (32%) and Offshore (15%) regions was explained by the four best region-specific predictor variables. Table S1 provides a list of the predictor variables selected for each subregion, and Text S1 includes a brief description of the physical, ecological, and biogeochemical significance of each variable.

Minitab was used again to perform a principal component analysis (PCA) for each subregion on the identified combination of four variables and [TOC], after normalizing each variable to its cube root value (Matlab [vR2013a] produced similar data separations by PCA and was used to produce biplots in Figure S3). Such an analysis separated sampling locations on a biplot based on their physical/biological signatures. The combined variance explained by the first two components ranged from 65% to 77% for all subregions. Results of the PCA were primarily analyzed here in terms of how stations separated across PC biplot space. For ease of discussion, the focus was on ecosystem characteristics that were easily differentiated across the PCA-quadrant space (see Text S2 for a brief description of PCA quadrants). Stations from the same cruise (season) that fell into a particular quadrant were also averaged to provide a mean PCA-based quadrant value for each cruise (season). Any quadrant could have up to nine, cruise-specific, mean values for each property (9 is the total number of cruises in this study). Such an approach reduced data complexity from 5,032 [TOC] data points to 121 values (e.g., Figure 4), which facilitated identification of broad ecosystem patterns associated with changes in [TOC].

2.9 Statistical Evaluation of Best Fit Lines

Figures generated in the following analysis include linear and logarithmic best fit lines using Sigma Plot (v10.0). The best fit lines include ±95% confidence intervals. All model best fit lines were determined using a Model II formulation using Matlab Code provided by Ed Peltzer (MBARI; www.mbari.org).

2.10 New Production

Phytoplankton production in the southern CCE is primarily limited by nitrate (Eppley, Renger, et al., 1979; Eppley & Peterson, 1979). For example, over the long-term, seasonal, inshore, mean [NO₃⁻] at the pycnocline (26.0 kg/m³; Jacox et al., 2016) was correlated with both [Chl a] at the depth of the Chl a maximum (R = 0.37, p < 0.001) and satellite estimates of NPP (R = 0.67, p < 0.001; Figure S4).

In the southern CCE, vertical transport is the major source of nitrate delivered to the euphotic zone (Messié et al., 2009; Pennington et al., 2010). Therefore, potential P_New can be estimated from the vertical gradient in measured nitrate and vertical advection rates, which in this case were determined from a regional model. Vertical diffusivity is only important during particular seasons (e.g., Haskell et al., 2016) and is a small component of the total vertical flux (Olivieri & Chavez, 2000; Pennington et al., 2010; Todd et al., 2012).

The [NO₃⁻] at the base of the nitracline was designated the upwelling [NO₃⁻], but the nitracline depth is described slightly differently than in previous studies (e.g., Hansell & Carlson, 1998; Haskell et al., 2016; Pennington et al., 2010; Romera-Castillo et al., 2016). For instance, researchers associated with the CalCOFI Program define the nitracline depth as the depth where [NO₃⁻] reaches a value of 1 μM (e.g., Bjorkstedt et al., 2010). Such a description is primarily used to designate a value that can be easily compared across time and space, whereas in the current study, the goal was to calculate potential new production from supplied [NO₃⁻]. As such, nitracline was defined here as the “depth of first nitrate use.”

Chemical gradients below the nitrate:temperature departure depth (125 m in Figure S5) are explained entirely by changes in temperature (and salinity). Therefore, any changes in [NO₃⁻] at deeper depths can be attributed to physical mixing. However, for depths shallower than the nitrate:temperature departure depth, changes in [NO₃⁻] are no longer adequately explained by physical parameters and instead correspond with the first instance of increased fluorescence over deep values. Based on this correlation, we interpret the decrease in [NO₃⁻] relative to its temperature-predicted value to be a result of biological uptake (i.e., <125 m in Figure S5). As such, our designation of nitracline was determined as the deepest depth where observed nitrate was 2.5 ± 0.5 μM lower than the [NO₃⁻] predicted by the temperature versus nitrate relationship. That is, we assumed that biological uptake was occurring at depths where [NO₃⁻] fell below the predicted value by 2.5 ± 0.5 μM or more. The magnitude of this departure for each profile was largely restricted by sampling resolution. Nitracline depths as defined in this study often corresponded with the deepest depth where conductivity-temperature-depth-rosette-based fluorescence increased over background values (Figure S5), consistent with the assumption that autotrophic production was responsible for the observed departure in [NO₃⁻] from the temperature-based prediction value. Certain groups of primary producers can grow at low light near the base of the euphotic zone (near the nitracline, as defined here), including the cyanobacterium prochlorococcus (Chisholm et al., 1988; Moore et al., 2002; Ting et al., 2002) and photosynthetic picoeukaryotes (Rodríguez et al., 2005). The mean isopycnal surface for our nitracline estimate was 25.6 ± 0.5 kg/m³, not significantly different from the isopycnal surface used to define the pycnocline (26.0 kg/m³; Jacox et al., 2016).

P_New within the euphotic zone was calculated according to equation 2. For this formulation, the euphotic zone was designated as the region of the water column shallower than the depth of first nitrate use. To calculate mean [NO₃⁻] in the euphotic zone (i.e., [NO₃⁻]_EZ), we only considered nonzero concentrations in the region shallower than the nitracline depth. This [NO₃⁻]_EZ was then subtracted from the [NO₃⁻] at the nitracline depth ([NO₃⁻]_first) to provide an estimate of the [NO₃⁻] that had been utilized in the euphotic zone. Finally, the vertical advection rate for each depth profile was extracted from the BECCO model to convert utilized [NO₃⁻] to a nitrate removal rate. Specifically, the vertical velocities used for this rate were an average of ±10 m within the nitracline depth and within ±10 days surrounding the CalCOFI cruise sampling date. These 20-day average upwelling velocities were representative of the typical event scale observed at the CCE-2 mooring for the chemically (and biologically) relevant species nitrate, oxygen and pCO₂ (Martz et al., 2014). For reference, the CCE-2 mooring is positioned on the shelf break off Point Conception along Line 080.0 of the CalCOFI domain. This is a slightly longer timescale than the 7.5 days of biomass turnover used elsewhere to estimate nitrate-based production (e.g., Aksnes & Ohman, 2009). Additionally, vertical velocities across the base of the fluorescence maximum and nitracline do not vary by more than 0.05 m/day, and with mean vertical velocities of 0.5 to 1.0 m/day, nitrate can be transported vertically by 10 to 20 m over typical event scales (20 days). At certain locations, this can result in significant transport of nitrate (e.g., ~10 μM in Figure S5) into higher light levels where fluorescence reaches maximum levels. The removal rate was subsequently converted to carbon equivalents using published C:N ratios (Martiny et al., 2013).

(2)

In the above equation (m/day) refers to the mean vertical transport within ±10 m of the nitracline depth, [NO₃⁻]_first (μM) refers to the concentration of nitrate at the nitracline depth, and refers to mean nitrate concentrations (μM) in the euphotic zone (the region between the nitracline and up to shallowest depth with observable [NO₃⁻] > 0 μM). We then had to convert P_New to carbon currency using C:N ratios in order to compare with other rates discussed in this study. Two different C:N values were chosen to represent the different CCE subregions. The Offshore and California Current communities of the CCE are typically comprised of small cyanobacteria, in particular Prochlorococcus (Taylor et al., 2015). These communities are expected to be similar to those gyre communities for which Martiny et al. (2013) determined a C:N of 195:28 (~7:1). The Inshore communities are typical of nutrient-rich upwelling zones with C:N of 137:18 (7.6:1). In formulating equation 2, we assumed that the nitrate supplied to the euphotic zone with a BECCO model-derived 20-day average vertical velocity was channeled into primary production at this same rate (i.e., utilization rate; Messié et al., 2009; Pennington et al., 2010; Romera-Castillo et al., 2016).

Other processes such as nitrification can influence observed [NO₃⁻] in the euphotic zone (e.g., Santoro et al., 2010) and physical processes other than vertical advection can also supply nitrate to the euphotic zone. Vertical advection was most often an order of magnitude greater than mean eddy diffusivity in the CCE (Gelpi & Norris, 2008; Todd et al., 2012). However, when winds are reduced, particularly inside the Southern California Bight, diapycnal diffusion of nitrate can sometimes be the dominant vertical supply route (Haskell et al., 2016). Nonetheless, in the absence of robust diffusion estimates for each [NO₃⁻] profile only vertical advection was considered in this formulation of nitrate-based production.

For equation 2 to yield a meaningful utilization rate, each location with a [NO₃⁻] profile had to be characterized by a positive vertical advection rate (i.e., upwelling) and a clearly distinguishable nitracline depth, as defined above. Of the 646 depth profiles that had both [NO₃⁻] and [TOC], 282 satisfied both conditions necessary for calculating P_New using equation 2. The sensitivity of the resulting P_New estimate to these omissions was examined and will be discussed later.

2.11 TOC Production

Potential P_TOC was estimated using a method similar to P_New as described above. [TOC] at the depth of [NO₃⁻]_first was subtracted from the mean “euphotic zone” [TOC]_EZ (note that TOC is produced in the euphotic zone while nitrate is consumed)_. The TOC production rate (mmol C m⁻² d⁻¹) was calculated by multiplying this excess [TOC] in the euphotic zone by the average upwelling velocity used in equation 2. The fraction of P_New that was partitioned into TOC was expressed as P_TOC:P_New and was determined as a profile specific value. To be consistent with P_New, calculated [TOC]_EZ did not include measured [TOC] at depths where [NO₃] was 0 even though there was measurable TOC. We assumed that if nitrate was not present then there was no net TOC production. Although not used in the P_TOC calculation, depths where TOC concentrations were positive but [NO₃⁻] was 0.0 were considered for the ΔTOC calculation described in section 2.6. At depths where [NO₃⁻] was 0.0, positive TOC was considered to be accumulating. If we include these “accumulating” TOC depths in the P_TOC calculation, sampling locations with a stratified water column, such as those in the Offshore region, P_TOC could increase by as much as 20–30%.

2.12 Important Differences Between Data Sets

When the goal was to examine stocks and dynamics of [TOC] and [POC], data from all 646 stations were used. These data sets were used to calculate density binned TOC and POC accumulations (i.e., ΔTOC [n = 5,032] and ΔPOC [n = 1,966]) as in section 2.6. When we sought to identify the physical and biological parameters that best explained [TOC] variability in each region, we calculated mean surface zone properties (σ_θ < 26.0 kg/m³) for each station during each cruise ([TOC]_Surface and ΔTOC_Surface). In this case, mean surface properties were compared by stepwise multiple regression followed by PCA to visualize ecologically relevant correlations that could be driving the variability in surface [TOC]. The PCA was divided by subregion, and both PCA quadrant-specific means (Table S2 and Figure 3) and cruise-specific means (Figures 4 and 6) within a PCA quadrant were calculated for comparative analysis (see Figure S6 for a schematic of the averaging methods used here). The two estimates of added TOC (the euphotic zone [TOC] difference used for P_TOC [section 2.11] and ΔTOC) were different from one another and varied with region and season (Figure S7). Finally, for several properties it was useful to have a spatially averaged value for the entire CalCOFI domain, and this was calculated by scaling each subregion-specific mean value by the areal contribution of that subregion to the total CalCOFI domain.

3 Results

3.1 Profiles and Regional Trends

[TOC] for the southern CCE between 2008 and 2010 were 36.0–210.5 μM C (n = 5,032) over depths 0–3550 m and potential densities (σ_θ) 22.5–27.8 kg/m³. [TOC] typically increased relative to the background with increasing [Chl a] (>1.0 μg/L; Figure 2). In Inshore regions of the CCE, shoaling of denser isopycnals (σ_θ: 25.5–26.0 kg/m³) increased nutrient concentrations in the euphotic zone and led to [Chl a] in excess of 1.0 μg/L and [TOC] in excess of 70 μM C (Figure 2). During the summer, surface waters warmed and their potential density and surface [Chl a] decreased (σ_θ < 25.0 kg/m³; [Chl a] < 1.0 μg/L), yet [TOC] sometimes persisted at elevated levels. [TOC] in excess of the upwelling concentration of ~50 μM C reflected either recent “addition” to the reservoir by primary production or a persisting pool of TOC from recent biological activity (Carlson et al., 1994). Along a specific isopycnal, median [TOC] at inshore sites exceeded offshore values by 3–10 μM (Figure 2). The Inshore North region exhibited the greatest additions, which were likely linked to persistent moderate upwelling at this location throughout the year (Huyer, 1983).

Figure 2

Colored dots in (a) represent individual [TOC] with associated [Chl a], and the flow of carbon discussed here is outlined in (b). Line profiles for each region in (a) represent median [TOC] for 0.25-kg/m³ density bins (interpolated black lines). Data points where [Chl a] > 1.0 μg/L are often associated with denser isopycnals (25.0–26.0 kg/m³) shoaling into the surface euphotic zone. Data points where [Chl a] < 0.5 μg/L and TOC = 60–75 μM C are associated with warmer, less dense (<24.5 kg/m³) summer and fall stratified waters. Black dots have no associated [Chl a] measure. In (b) [NO₃⁻]-rich waters upwell across the nitracline fueling primary production, the result of which can either accumulate as a combination of DOC and POC (measured as TOC) or be subject to export as sinking carbon or subducted nonsinking carbon. Chl a = chlorophyll a; DOC = dissolved organic carbon; POC = particulate organic carbon; TOC = total organic carbon.

Throughout the water column [TOC] was most strongly correlated with potential temperature (R² = 0.66, p < 0.0001, n = 5,032). As expected, [TOC] was also significantly correlated with log-transformed depth (R² = 0.58, p < 0.0001) and other chemical variables whose concentrations change with depth, such as [O₂] (R² = 0.56, p < 0.0001) and [PO₄] (R² = 0.62, p < 0.0001). These correlations point to the strong physical control on [TOC] profiles in this hydrographically complex region.

At the pycnocline (i.e., σ_θ ≈ 26.0 kg/m³), [TOC] was 50.0–55.0 μM C (Figure 2), and these concentrations were sometimes observed in surface waters during an upwelling event. In upwelling waters, TOC was primarily DOC (i.e., [DOC] ≈ [TOC]; from comparing filtered and whole seawater collected from similar locations but different studies; Figure S8a). In fact, throughout much of the Inshore South, California Current, and Offshore subregions [TOC] measurements primarily reflected [DOC]. For example, there was consistent overlap of surface measured [TOC] with surface [DOC] from the Santa Barbara Channel (SBC; Wear et al., 2015; Figure S8a). Additionally, [DOC] collected in 2013 for Climate and Ocean: Variability, Predictability and Change project line P02 (east of longitude 139.0W) had similar values to [TOC] by depth (Figure S8a). During other cruises conducted in the same regions (e.g., CCE LTER cruises based near Point Conception), we measured DOC and TOC separately and compared to POC (Figure S8b) and to other estimates of DOC for the region (Figure S8c). In addition, all local [DOC] data sets (Figure S8c) were used to generate a minimum 10th percentile [DOC] profile, which was then used calculate ΔDOC (as above for ΔTOC) from these same data sets. If ΔTOC from CalCOFI stations only captured DOC additions then ΔTOC should be equal to ΔDOC (see also Text S3 for further discussion). When [POC] < 5 μM C, ΔDOC was ~60–80% of ΔTOC (Figure S8d). Exact agreement between these data sets is not expected since DOC and TOC data sets are temporally and spatially decoupled, so we interpret these results to indicate that at low [POC], ΔDOC ≈ ΔTOC. However, during coastal bloom conditions when [POC] > 5 μM C, ΔDOC represents 29–73% of ΔTOC (51% on average; Figure S8d), which supports the conclusion that suspended POC impacts calculated ΔTOC at higher [POC]. Other studies in productive regions have made similar observations (e.g., Carlson et al., 1998).

3.2 Insights Into Regional TOC Variability

To identify key parameters that best explained the variance in surface [TOC] we carried out a stepwise multiple regression analysis using a suite of mean surface water properties (σ_θ < 26.0 kg/m³). In each subregion four variables related to water column stability and primary production most effectively correlated with surface [TOC] (see Text S1 for a description of the variables used for each subregion). The combination of those four variables and [TOC] were then used in a PCA to further separate out the subregional data sets into unique conditions (Figure S3 for PCA biplots). Since each of the five variables contributed to PC1 and PC2 of the PCA the resulting distribution of points in the PCA biplot reflects the combined contribution of those variables. However, there were many stations where selected variables were all relatively close to mean conditions and those points fell toward the middle of the PCA biplot. In other words, PCA could not always separate all stations in each subregion into distinct and easily identifiable clusters. To avoid biasing the analysis by removing stations that fell toward the center of the PCA biplot, we designated all sampling locations that fell into each quadrant in the PCA biplot as a “cluster.” As such, each geographic subregion (Figure 1) had four different clusters/quadrants representing a range of surface [TOC] and [Chl a] conditions (Figure 3).

Figure 3

Principal components divided stations in each subregion into four “quadrants.” Q1–Q4 refers to each quadrant of the principal component analysis biplot and are ordered here by decreasing mean quadrant [Chl a]. Principal component analysis was performed for each region using surface [TOC], and a combination of four other variables that best explained the variance in [TOC]. See section 2 for further details on quadrant designations and Table S2 for mean values for each quadrant. Inside the box plots, the solid and dotted lines are the median and mean, respectively. The edges of the box plots represent the 25th and 75th percentiles, and outside the box plots the whiskers are the 10th and 90th percentile values, and the dots identify “outliers.” Note change in scale between Inshore (top panels) and Offshore (bottom panels). Symbols denote approximate conditions as *Upwelling, ^Bloom, #Accumulation and †Low OM. Chl a = chlorophyll a; OM = organic matter; TOC = total organic carbon.

In the Inshore North and South regions, stations with elevated mean quadrant-specific [TOC] values, 75.7 ± 10.1 and 79.2 ± 13.7 μM, respectively, were associated with sampling locations that had high [Chl a] (Figure 3) and also high NPP (not shown). The highest [TOC] stations were typically closest to shore (<20 km from shore). However, [Chl a] alone did not explain the observed differences in surface mean [TOC]. For example, some sampling locations in the Inshore South (n = 55 out of 207), clustered based on relatively high [TOC] (73.4 ± 7.8 μM) but low [Chl a] (0.93 ± 0.06 μg/L; Figure 3). Other factors influencing [TOC] at these sampling locations can be inferred by examining PC biplots.

To simplify the PCA, we calculated profile-specific surface mean values (the mean of all values above the pycnocline) for all properties including ΔTOC, where ΔTOC provides an estimate of TOC that is added over and above background concentrations. To further simplify the comparison we combined all surface mean ΔTOC values within a quadrant that were from the same cruise to calculate a cruise-specific value. As expected, ΔTOC values increased with increasing [Chl a] (ΔTOC = 6.5(ln [Chl a]) + 11.1; R² = 0.64, p < 0.001; Figure 4 and Table S2); ΔPOC, calculated in a similar manner, was highly correlated with ΔTOC across the region (R² = 0.84, p < 0.001; Figure S9). Based on the relationship to [Chl a], a “Bloom” classification (Figure 4) was assigned to the quadrant in Inshore biplots containing sampling locations with the highest ΔTOC (20–35 μM) and highest [Chl a] (Figure 4).

Figure 4

Each point represents a cruise mean for identified principal component analysis quadrants (see quadrant description in section 2). The “Upwelling” condition was designated for the California Current (Q2) and Inshore North (Q3; see Text S2 and Figure S3 for quadrant designations). “Bloom” conditions were designated for the Inshore North (Q4) and South (Q4). “Accumulation” conditions were designated for Q1 of all subregions. And “Low OM” conditions were identified for Inshore North and South and California Current and Offshore: quadrants 2, 3, 3, and 4. The thinner dotted lines outside of model II best fit lines represent the ±95% confidence intervals. Chl a = chlorophyll a; OM = organic matter; TOC = total organic carbon.

In the Inshore North and California Current regions several sampling locations that had medium [Chl a] and ΔTOC (10–15 μM) but high [NO₃⁻] clustered together in each PCA biplot (Figure 4). Quadrants where these sampling locations were found were designated as “Upwelling” conditions. In the California Current region, the Upwelling quadrant was characterized by higher [Chl a] and higher wind speed (Figure S3 biplots) indicating that upwelling was likely mediated by wind stress curl (Rykaczewski & Checkley, 2008). Finally, some sampling locations had moderate ΔTOC but were primarily differentiated by high buoyancy frequency (N2; Figure S3), which indicated stratified conditions. Elevated ΔTOC (5–15 μM) at these sampling sites may have resulted from postproduction accumulation of TOC. Quadrants containing these sampling locations were assigned an “Accumulation” designation (Figure 4, last panel). The ΔPOC varied in a similar manner (Table S2).

3.3 New Production and TOC Production

Sampling locations that were classified as “Bloom” had the highest P_New estimates, where mean P_New was 139.0 ± 71.7 mmol C m⁻² d⁻¹ (Table S2). In contrast, mean P_New for the “low [TOC]” in the Offshore region was 4.7 ± 2.2 mmol C m⁻² d⁻¹. Net community production (NCP) estimated from oxygen:argon (O₂:Ar) in the euphotic zone exhibited values similar to P_New across the CalCOFI domain (Munro et al., 2013). Corresponding mean P_TOC values for these two conditions were 43.8 ± 37.7 mmol C m⁻² d⁻¹ and 0.8 ± 0.5 mmol C m⁻² d⁻¹, respectively (Table S2). Variations in P_New across the region primarily reflected changes in model-derived upwelling velocities (R² = 0.83) that converted NO₃⁻ removal to P_New (Figure S10a). Theses velocities varied over several orders of magnitude (e.g., 0.01 to 1.0 m/day). Upwelling influenced P_TOC in a similar way (r² = 0.88; data not shown).

As previously discussed, P_New and P_TOC were only calculated for depth profiles with both a clearly identifiable depth of first NO₃ use and active upwelling (i.e., positive vertical advection) at the time of sampling (i.e., 282 profiles out of 646 sampled during the time series). In the Inshore region, several profiles were omitted because they were too shallow to establish an accurate depth of first NO₃⁻ utilization (97 out of 328 Inshore profiles). As an alternative, we used [NO₃⁻] at a fixed depth of 60 m to replace [NO₃⁻]_first (Messié et al., 2009) to test whether omitting these stations affected the overall P_TOC:P_New calculation. P_New as calculated by the fixed 60 m method resulted in a similar mean P_New for the region; however, P_TOC increased by 13% on average (Figure S11). This impacted the inshore ratio of P_TOC:P_New, increasing by ~3% on average.

Additionally, some locations were omitted in the California Current and Offshore regions (116 of 318) because they had net downwelling conditions at the time of CalCOFI sampling. Although we could not test how P_New was directly impacted when these stations were omitted, we wanted to explore whether other productivity indicators that were testable could provide some insight. We assessed how both average [Chl a] at the chlorophyll max depth and satellite-based NPP for each cruise were altered in the offshore region when these stations were omitted. If the regional mean value for these parameters was under(over)estimated by omitting stations, then it would be logical to assume that P_New was also impacted. However, cruise-specific [Chl a] and NPP values were overestimated by only ~3% when stations experiencing net downwelling at the time of sampling were omitted.

Finally, a relatively small number of Inshore profiles (n = 13% or 4% of inshore stations) were omitted because there was no measurable gradient in [NO₃⁻] within the euphotic zone (i.e., no change in [NO₃⁻]). Overall, the number and location of omitted stations varied annually, and so, the spatial and temporal coverage afforded by this large data set was necessary for accurately estimating mean values for each property.

3.4 Satellite-Based NPP, the f-Ratio, and Carbon Partitioning

The relationship of P_New to NPP, or the “f-ratio”(Eppley et al., 1979), was examined using satellite-based NPP. Bottle incubation data for primary production were available for a small subset of sampling locations during each cruise, but satellite NPP achieved the broader spatial coverage appropriate for comparing to [NO₃⁻] and TOC data sets. Across all regions of the CCE, P_New was significantly correlated (R² = 0.60, p < 0.001) with satellite-based NPP (Figure S10b), and the f-ratio (P_New:NPP), decreased from 0.53 ± 0.23 Inshore to 0.37 ± 0.31 Offshore (Figure S12). Across the domain, the present study found P_New:NPP was a spatially averaged 0.44 ± 0.30. A similar spatial f-ratio relationship was determined for the CalCOFI region by dividing O₂:Ar-NCP (NCP from O₂:Argon measurements) by CalCOFI ¹⁴C-based primary production estimates (converted to a 24-hr value; Munro et al., 2013).

When separated by subregion and PCA space, P_TOC:NPP and P_TOC:P_New had values of 0.04 ± 0.02 and 0.12 ± 0.06 for the least productive oligotrophic Offshore region and up to 0.16 ± 0.08 and 0.28 ± 0.15 for coastal Bloom conditions (Table S2, Figures S13a, and S13b). For comparison, spatially averaged P_TOC:NPP was 0.06 ± 0.04, and P_TOC:P_New was 0.15 ± 0.07.

For any given “ecosystem” condition, the mean euphotic zone [TOC] term used in the P_TOC calculation was the main driver of P_TOC:NPP and P_TOC:P_New variability. Mean euphotic zone [NO₃⁻]_EZ and the upwelling [TOC] and [NO₃⁻] were much less variable.

4 Discussion

In the CalCOFI region there is little methodological agreement between different estimates of production and export (e.g., Gruber et al., 2011; Nagai et al., 2015; Plattner et al., 2005; Stukel et al., 2011, 2017). Inherent differences between the timescale and lengthscale captured by methods estimating new production and sinking particle export could give rise to these observed discrepancies. In addition, organic matter can be exported by phases other than gravitationally sinking particles; where DOM and suspended POM, both considered to be nonsinking, can be exported away from sites of production. For example, neither O₂:Ar-NCP nor new production agree well with direct estimates of sinking organic matter flux. To reconcile this imbalance, previous studies have invoked nonsinking organic matter as an additional pathway of export. In this study, thousands of [TOC] and [NO₃⁻] measurements from the southern CCE region were used to compare new production rates with export production estimates that included nonsinking reservoirs of detrital organic matter. In addition, the large supporting data set available through CalCOFI was used to examine potential controls on the size of the nonsinking reservoir that is available for export in the CCE.

4.1 NPP and Potential New Production in the CCE

To calculate new production we formulated a method that took advantage of the large CalCOFI data set of [NO₃⁻] measurements. Although our approach differed from previous studies, the results are in broad agreement with published work. For example, P_New in the Inshore North region varied from a high of 118.2 ± 72.2 mmol C m⁻² d⁻¹ during the two spring cruises to a low of 51.7 ± 15.9 mmol C m⁻² d⁻¹ during the two fall cruises (Figure S13). Averaged over the nine cruises, P_New was 73.6 ± 42.8 mmol C m⁻² d⁻¹ (Figure 5) and was similar to a previous estimate for the central CCE (<150 km from the coast) of 74 ± 9 mmol C m⁻² d⁻¹ (Messié et al., 2009). The larger range in the present CalCOFI data set (±42.8 mmol C m⁻² d⁻¹) reflects greater variability in the direct water column-based measurements as opposed to broader climatological-based production estimates (range of 9 mmol C m⁻² d⁻¹ in Messié et al., 2009). In the Inshore South, P_New was 57.3 ± 34.2 mmol C m⁻² d⁻¹ during spring cruises (Figure S13) and 28.1 ± 19.5 mmol C m⁻² d⁻¹ averaged seasonally (Figure 5). As an example, for the more productive Inshore South CalCOFI station (086.7 035.0) P_New was 134.3 and 113.9 mmol C m⁻² d⁻¹ during spring 2008 and 2010, respectively. These values were comparable to spring 2013 and 2014 data from the nearby San Pedro Ocean Time series (SPOT; 83 ± 42 to 103 ± 55 mmol C m⁻² d⁻¹), which were calculated using a similar nitrate removal method but with more direct measurements of vertical advection/diffusion velocities (Haskell et al., 2016). For comparison, the Peru and Benguela eastern boundary current systems exhibit mean P_New values of around 120 mmol C m⁻² d⁻¹ (Messié et al., 2009).

Figure 5

Annual average production values (in units of mmol C m⁻² d⁻¹) for the four subregions in this study. Subregions are arranged from left to right: Offshore, California Current, Inshore North (right top), and Inshore South (right bottom). Standard deviations represent the sum of seasonal squared errors. NPP = net primary production; P_New = new production; P_TOC = total organic carbon production.

Our spatially averaged P_New was 26.6 ± 21 mmol C m⁻² d⁻¹ across the CalCOFI domain and nearly identical to previous long-term (1984–1997) new production estimates (27.2 ± 13 mmol C m⁻² d⁻¹ [Bograd et al., 2001; Roemmich, 1989]). A mean O₂:Ar-NCP of 17.5 ± 5.3 mmol C m⁻² d⁻¹ is also in general agreement with our calculated P_New for the grid (Munro et al., 2013). The general agreement between NCP and estimates of nitrate-based new production was also observed at SPOT (Haskell et al., 2016). Overall, the agreement across different methods lends confidence to the P_New approach introduced in this study.

Previous annual mean f-ratio estimates for the Central CCE are between 0.59 and 0.66 (Messié et al., 2009; Pennington et al., 2010), similar to the f-ratio for the southern CCE during productive periods (0.58–0.69; Eppley et al., 1979; Munro et al., 2013). For comparison, our highest f-ratios (P_New:satellite-based NPP) were 0.58 ± 0.10 during “Bloom” periods in the Inshore North region and 0.59 ± 0.46 during “Upwelling” (wind-stress curl) periods in the California Current region (Table S2). Enhanced f-ratios estimated during bloom and upwelling conditions (as identified by PCA) are greater than seasonal mean estimates for all subregions (e.g., spring f-ratios of 0.53 ± 0.14 and 0.45 ± 0.26 for the Inshore North and California Current, respectively), suggesting that the PCA approach successfully isolated ecosystem states where nitrate made a greater contribution to net primary production. Previous estimates were made from much smaller data sets, which likely explains the smaller range reported in these studies. The variability within identified ecosystem conditions (e.g., deviations in Table S2) is consistent with the complex hydrography of the CCE region, which can advect and redistribute biomass and chlorophyll along and across shore (Barth et al., 2002; Chenillat et al., 2015; Haury et al., 1986; Olivieri & Chavez, 2000).

4.2 TOC Cycling in the CCE

The high f-ratio of 0.58 ± 0.10 during “Bloom” periods in the Inshore North accompanied high [TOC] and [Chl a] (Figures 3 and 4) in the region, which supported new, nitrate-driven production of [TOC]. Other upwelling regions show a similar relationship between nutrient supply and TOC or DOC production, such as the equatorial Pacific Ocean (Peltzer & Hayward, 1996), Iberian Peninsula (Álvarez-Salgado et al., 2001), Ross Sea (Carlson et al., 2000), eastern North Pacific Ocean off Oregon (Hill & Wheeler, 2002), East China Sea (Ogawa & Tanoue, 2003), Oyashio Current (Hasegawa et al., 2010), and SBC (Halewood et al., 2012; Wear et al., 2015). In the southern CCE, bottle incubation studies that have measured DOC and suspended POC production during phytoplankton grow out experiments have also confirmed the partitioning of new production into detrital reservoirs (Halewood et al., 2012; Wear et al., 2015).

In the current data set, when ΔTOC was compared with the abundances of different phytoplankton classes identified via microscopy (CalCOFI, Taylor et al., 2015), a positive correlation was observed with diatom and dinoflagellate (both heterotrophic and autotrophic) abundances during “Bloom” conditions (data not shown). These organisms dominate the >20-μm size fraction and constitute 50–70% of total [Chl a] at these locations (Goericke, 2011). As such, the primarily positive correlation between TOC and [Chl a] can be used to infer potential mechanisms of TOC production. For instance, extracellular release of DOM by these large phytoplankton (e.g., Fogg, 1983; Teira et al., 2001), sloppy feeding by grazers on these larger cells (Lampert, 1978; Saba et al., 2011), cell lysis (Breitbart, 2012; Bidle, 2015), and other processes (Carlson & Hansell, 2014) likely contributed high [TOC] and ΔTOC to Inshore regions.

Across the southern CCE study region, P_TOC normalized to P_New increased significantly with increasing [Chl a] (R² = 0.37, p < 0.001, Figure 6), and this trend was driven primarily by changes in [TOC]. Enhanced production of TOC during bloom conditions has been documented in several locations, including a coastal upwelling system in the eastern North Atlantic (42 vs. 15 mmol C m⁻² d⁻¹ during postbloom periods; Álvarez-Salgado et al., 2001). Partitioning of net and new production into DOC has also been documented for spring blooms in the oligotrophic Sargasso Sea (DOC was 0.59–0.70 of NCP; Hansell & Carlson, 1998), during coastal upwelling off the Iberian Peninsula (DOC was 0.20 of NPP; Álvarez-Salgado et al., 2001), and spring blooms in the western North Atlantic (DOC ranged between 0.10 and 0.40 of NCP; Romera-Castillo et al., 2016).

Figure 6

The ratio of P_TOC to P_New as a function of [Chl a]. Each point represents a unique cruise and condition as noted in Figure 4 and the model II regression fit is applied across all shown data points (n = 86; y = 0.07ln(x) + 0.20; R² = 0.37). Thinner dotted lines outside of natural log best fit line represent the ±95% confidence intervals. Chl a = chlorophyll a; OM = organic matter; P_New = new production; P_TOC = total organic carbon production.

The ratio of P_TOC:P_New varied from 0.12 ± 0.08 to 0.28 ± 0.15 between low production locations in the oligotrophic gyre and coastal bloom locations in the Inshore North. Both Inshore North “Bloom” (0.28 ± 0.15) and mean springtime P_TOC:P_New (0.24 ± 0.15) were significantly elevated (Mann Whitney, p < 0.01) over other PCA-based conditions and seasons. In the time period 2013–2014, spring partitioning of NCP:DOC was also significantly elevated (~0.28) at SPOT using a subeuphotic zone carbon mass balance approach that approximated the relationship between DOC and total oxidized carbon (decreasing to ~0.08 in the fall; Haskell et al., 2016). For comparison, P_TOC:P_New as estimated here for the nearest CalCOFI station to SPOT (086.7 035.0) was 0.16 and 0.23 during spring 2008 and 2010, respectively.

The PCA-based approach also appeared to distinguish some of the sampling locations as “TOC accumulation” sites (23% of total stations) based on a decoupling of [TOC] and [Chl a]. At these locations, surface ΔTOC increased between 5 and 15 μM C. DOC additions of similar magnitude (10–12 μM) have also been reported for the SBC region during warm stratified and postbloom conditions (Halewood et al., 2012; Wear et al., 2015). Other coastal regions, such as throughout the Mediterranean (Romera-Castillo et al., 2013; Santinelli et al., 2013) and Ross Seas (Carlson et al., 1998, 2000), show similar DOC accumulations (10–20 μM C) during postbloom, stratified conditions.

TOC accumulation sites often exhibited moderate P_TOC:P_New as well (0.15 up to 0.35; Figure 6). Sometimes, stations that exhibited strongly stratified conditions, such as those observed during the summer cruise of 2009 fell into the accumulation designation by PCA and exhibited elevated P_TOC:P_New (0.22 and 0.35 in the Inshore North and South regions, respectively). These findings are more consistent with bottle-based DOC percent extracellular release (PER) experiments near the Santa Barbara basin, which had consistently higher TOC partitioning under postbloom, nutrient-limited conditions compared to bloom conditions (Halewood et al., 2012).

Micronutrient limitation of heterotrophic bacteria that degrade DOC could also result in TOC accumulation (Kirchman et al., 2000). To determine the importance of iron limitation during our sampling period we used the diatom-specific Si_ex proxy (after Sarmiento et al., 2004, as modified by King & Barbeau, 2011). The Si_ex proxy refers to the difference between measured silicic acid concentrations and [NO₃⁻] in the euphotic zone when the upwelling ratio of the two macronutrients is ~1. Under iron replete conditions diatoms have a silica:nitrate uptake ratio of 1, and so, Si_ex is expected to be ≥0. However, when diatoms are iron limited they will draw down silicic acid at a faster rate than nitrate and Si_ex < 0.0 (i.e., when excess [NO₃⁻] is present in the euphotic zone iron limitation of diatoms is inferred [Hutchins & Bruland, 1998; King & Barbeau, 2011]).

About 13% of the ~646 profiles identified for the 2008–2010 CalCOFI time series had mean surface (Si_ex) <0.0, which was consistent with iron limitation of phytoplankton production, and perhaps, by extension, limitation of heterotrophic bacterial production. Several of the profiles characterized as iron limited were encountered in the Inshore South (50–100 km from shore) during summer and fall, and these locations may be perennially iron limited during these seasons (e.g., King & Barbeau, 2011). These sites were characterized by surface [TOC] >65 μM, [Chl a] <1.0 μg/L, and moderate P_TOC:P_New of 0.20 ± 0.11. Relatively low [Chl a] indicated that active TOC production was muted, and so, TOC accumulation likely resulted from iron limitation of TOC consumption by heterotrophic bacteria rather than enhanced TOC production. Most of the identified iron-limited stations also had low bacteria counts (<5 × 10⁵ cells per millimeter). We cannot rule out the possibility that Fe-stressed phytoplankton increased their partitioning of P_New into TOC.

Overall, we observed that surface mean [TOC] can become elevated (e.g., 70–90 μM) during coastal bloom conditions in the CCE corresponding to P_TOC:P_New ratios of 0.20–0.40. Some locations exhibit elevated [TOC] but low [Chl a], signaling postproduction accumulation of TOC. A subset of these locations can be characterized by conditions consistent with iron limitation. However, low bacteria counts and micronutrient limitation are not always linked to TOC accumulation. In general, elevated P_TOC:P_New represented enhanced partitioning of new production into TOC.

4.3 P_TOC Contributions to Export and New Production

Nonsinking reservoirs of organic carbon—specifically DOC and suspended POC—can contribute to carbon export through mechanisms such as subduction and/or biological repackaging (DeVries & Weber, 2017; Siegel et al., 2016). In fact, the export of nonsinking organic carbon has been suggested as a pathway to resolve the ~twofold offset between sinking export and new production or NCP (Bograd et al., 2001; Emerson, 2014; Munro et al., 2013; Roemmich, 1989). For each CCE subregion gravitational export flux has been estimated using either the ²³⁴Th method or by VERTEX-style sediment traps (Haskell et al., 2016; Stukel et al., 2011, 2013), and NCP has been estimated from surface ocean O₂:Ar measurements (Haskell et al., 2016; Munro et al., 2013). Subregion means of sinking export and NCP can be directly compared with P_TOC and P_New as estimated here (Figure 7).

Figure 7

The combination of sinking POC (Stukel et al., 2011, 2013) and P_TOC is compared with potential P_New and NCP estimated from net oxygen production (Munro et al., 2013). P_New and P_TOC were calculated using seasonal means over 2008–2010. NCP is the annual mean of that reported for 2005–2008 in Munro et al. (2013). Sinking POC for Southern Inshore are from Haskell et al. (2016) for 2013–2014 at San Pedro Ocean Time series, and mean estimates for other regions are from Stukel et al. (2011, 2013) over 2006–2008. Error bars represent standard deviations. NCP = net community production; P_New = new production; P_TOC = total organic carbon production; POC = particulate organic carbon.

As discussed in section 4.2, P_New estimates from the present study are in agreement with previous O₂:Ar-NCP (Figure 7; Haskell et al., 2016; Munro et al., 2013). In the most productive Inshore region, the calculated P_TOC exceeded existing estimates of sinking export. This implicates TOC as an important reservoir or exportable carbon in the CCE. However, even when P_TOC and sinking export rates are combined for this subregion, the “total” export is still lower than either P_New or NCP (Figure 7). Here other export terms including contributions from suspended POC or consumption of sinking organic matter by diel vertical migrators (DVMs) may play important roles. The contribution of suspended POC production is discussed below. Consumption by DVM can be 20–50% of sinking particles, suggesting that this process significantly impacts the organic carbon budget of the upper ocean (Davison et al., 2013; Steinberg & Landry, 2017). In contrast, total export appears to fall within the margin of error for the California Current and Offshore regions (Figure 7).

Suspended POC additions (ΔPOC) are similar to ΔTOC in Inshore regions and contribute to approximately 33–50% of ΔTOC Inshore (Figure S8d and Text S3). Including ΔPOC in nonsinking organic carbon production increases P_TOC in the most productive Inshore region by up to 7 mmol C m⁻² d⁻¹ but on average suspended POC contributes ~20–50% of P_TOC estimates presented here. Early studies in the Southern California Bight estimated residence times for suspended POM on the order of days to weeks (Eppley, 1992; Eppley et al., 1983), which indicates little exchange with the fast-sinking particle reservoir, which may sink several meters each day. However, more recently, Stukel et al. (2017) estimated that slowly sinking particles (<10 m/day) can contribute about 50% to the sinking flux in the southern California Current System (CCS) region. They also reported that the passive subduction of suspended POM could contribute between 7% and 80% to the total export, with a mean of 26%. Slowly sinking particles, if captured in our suspended POM measurement, would likely be much smaller than the suspended POC production rate (7 mmol C m⁻² d⁻¹) estimated here for Inshore regions, and so, are unlikely to contribute significantly to the local imbalance between export and new production. Including ΔPOC in the P_TOC calculation resolves a small portion of the inshore imbalance against P_New or NCP but still does not reach mean values; however, including maximal estimates of both suspended POC production and DVMs brings total export estimates to within 10% of mean NCP.

4.4 Offshore Gradients in [TOC]

Biogeochemical models invoke the nonsinking organic matter reservoir as a mechanism for transporting nutrients and carbon from coastal CCE to the ocean's interior (Gruber et al., 2011; Letscher et al., 2013). To test the possible importance of this process, lateral TOC transport was estimated by considering this pool to be a passive tracer (Letscher et al., 2013), where gradients in [TOC] between adjacent regions were assumed to result from transport only. Surface currents of the CCE are dominated by alongshore transport of 5–12 km/day but simultaneous cross-shore transport can average 0.5 km/day (Todd et al., 2011). Using BECCO model lateral velocities and a simple box model approach (Peltzer & Hayward, 1996), we found areal integrated cross-shore TOC export from the Inshore North region could be on the order of 10 ± 7 mmol C m⁻² d⁻¹ during more productive upwelling conditions. Such estimates are similar to those for lateral export away from sites of production in the equatorial upwelling region of the eastern tropical Pacific Ocean (Peltzer & Hayward, 1996). This estimated lateral export from the Northern Inshore represents ~50% of mean Northern Inshore P_TOC. If even half of this nonsinking reservoir persists on the order of 6–12 months (i.e., semilabile DOM; Hansell, 2013), then coastal TOC can be effectively transported to TOC-depleted regions like the California Current (50–100 km). Future modeling efforts based in the CCE should incorporate TOC transport and TOC turnover rates to better constrain potential lateral transport of inshore primary production.

4.5 DOC Production in Eastern Boundary Current Systems

Eastern boundary currents (EBCs), including the CCS, are among the most productive systems in the ocean (Chavez & Messié, 2009). These regions are also important sites of DOC production. For example, between 2% and 40% of NPP, with a mean around 20%, is partitioned into DOC in the Canary, Benguela, and Humboldt Currents (Álvarez-Salgado et al., 2001; Cuevas et al., 2004; Teira et al., 2001; Troncoso et al., 2003). Ekman-driven lateral transport can also export carbon and nutrients in the form of organic matter from EBCs to less productive gyre ecosystems (Letscher et al., 2013, 2016; Nagai et al., 2015; Reynolds et al., 2014; Torres-Valdés et al., 2009). Our temporally and spatially extensive data set enabled us to constrain production rates of TOC, which includes DOC, and identify ecosystem conditions that drive variability in TOC partitioning as reported in previous EBC studies. Our data indicate that lateral export of coastal DOC is an important process in the CCS, and with a TP:NPP similar to that reported for other EBCs the processes of lateral export may be relevant in those regions as well.

5 Summary and Conclusions

The surface production of total organic carbon (P_TOC) in the CCE correlates with indicators of primary production (e.g., [Chl a]). To provide a broader context, calculated P_TOC was compared with estimates of potential P_New and satellite-based production (NPP), and it was found that P_TOC:P_New ranged from 0.12 ± 0.08 to 0.28 ± 0.15 and P_TOC:NPP ranged from 0.04 ± 0.02 to 0.28 ± 0.15 between the low production periods in the oligotrophic gyre and the highly productive periods in the Inshore North, respectively. For most regions of the CCE, P_New falls within the range of previous estimates of O₂:Ar-NCP and agrees with previous findings for the region that the two methods are in general agreement. This observation lends support to the method used here for calculating P_New and, to some extent, P_TOC. Our calculations found that TOC represents a quantitatively important reservoir of reduced carbon in the CCE. For instance, P_TOC (mean of 18.3 ± 13.4 mmol C m⁻² d⁻¹) exceeded published estimates of sinking export (mean of 8.2 ± 5.2 mmol C m⁻² d⁻¹) in coastal regions of high production. Together, P_TOC and gravitational export nearly approximate P_New and NCP for most regions of the CCE but fell short in the most productive inshore region, where other forms of export or poorly constrained lateral export and subduction of organic matter should be more closely examined. A simple estimate of cross-shore transport from the Inshore North to the adjacent California Current region indicated that TOC transport could be up to 10 mmol C m⁻² day⁻¹, representing about half of mean P_TOC in the Inshore North region.

Ultimately, this data set allowed us to examine the role of nonsinking organic matter in carbon export, in line with a priority that was recently designated by the EXPORTS Science Plan (Siegel et al., 2016). To further enhance our understanding of the role of TOC as a repository of fixed carbon, we attempted to derive the flux of TOC produced as a fraction of new production. Such a parameter is also important for numerical models that are attempting to resolve the net effects of biological processes in the region. By modifying existing regional models to incorporate the proposed TOC partitioning coefficients (as a function of P_New and NPP), we can test the accuracy of these coefficients by observing whether these models reproduce [TOC] distributions encountered across the CCE. Together, the data and derived products discussed in this paper fill an important knowledge gap that brings us closer to a more accurate parameterization of the role of TOC as a pathway for carbon export and accumulation in the California Current (Nagai et al., 2015).

Acknowledgments

The authors wish to thank the captain, crew, and research staff associated with the CalCOFI Program making the high-quality data sets available. Some of the data sets presented here were supported in part by CCE-LTER augmented funding (NSF grant OCE-1026607). The data sets have been made publicly available as noted in the section 2 and can be accessed at http://cce.lternet.edu/data/ and www.calcofi.org. The authors wish also to thank Mati Kahru, Arianne Verdy, Matt Mazloff, Bruce Cornuelle, and Craig Nelson for useful discussions regarding the analysis of data presented here. The corresponding author also wishes to thank the Scripps Institution of Oceanography Graduate Division for financial support.