Mesoscale ocean fronts enhance carbon export due to gravitational sinking and subduction

Significance

Transport of organic carbon from the sunlit surface ocean to deeper depths drives net oceanic uptake of CO₂ from the atmosphere. However, mechanisms that control this carbon export remain poorly constrained, limiting our ability to model and predict future changes in this globally important process. We show that the flux of sinking particles (typically considered the dominant form of downward transport of organic carbon) is twice as high at a frontal system, relative to surrounding waters or to nonfrontal conditions. Furthermore, downward transport by subduction leads to additional carbon export at the front that is similar in magnitude to the sinking flux. Such enhanced C export at episodic and mesoscale features needs to be incorporated into biogeochemical forecast models.

Abstract

Enhanced vertical carbon transport (gravitational sinking and subduction) at mesoscale ocean fronts may explain the demonstrated imbalance of new production and sinking particle export in coastal upwelling ecosystems. Based on flux assessments from ²³⁸U:²³⁴Th disequilibrium and sediment traps, we found 2 to 3 times higher rates of gravitational particle export near a deep-water front (305 mg C⋅m⁻²⋅d⁻¹) compared with adjacent water or to mean (nonfrontal) regional conditions. Elevated particle flux at the front was mechanistically linked to Fe-stressed diatoms and high mesozooplankton fecal pellet production. Using a data assimilative regional ocean model fit to measured conditions, we estimate that an additional ∼225 mg C⋅m⁻²⋅d⁻¹ was exported as subduction of particle-rich water at the front, highlighting a transport mechanism that is not captured by sediment traps and is poorly quantified by most models and in situ measurements. Mesoscale fronts may be responsible for over a quarter of total organic carbon sequestration in the California Current and other coastal upwelling ecosystems.

The magnitude of plankton-mediated primary production (PP) that is removed annually from the surface ocean−atmosphere system and transported to depth remains poorly constrained, with estimates varying from 5 Pg C⋅y⁻¹ to 21 Pg C⋅y⁻¹ (1–3). Inadequate resolution of the many mechanisms that drive export flux—sinking particles and aggregates, active transport by vertically migrating organisms, advection and diffusion of particles and dissolved organic compounds—is also a major challenge for parameterizing ocean models that seek to predict future responses to climate impacts. Although sinking material is generally assumed to dominate the export of organic carbon in the oceans, sinking flux is often significantly lower than simultaneously measured new or net community production (4–6) and insufficient to meet the metabolic requirements of deep-sea and benthic organisms (7). This has led to the notion of mesoscale ocean features (fronts and eddies) as sites where locally enhanced vertical advection may stimulate production and gravitational (sinking) export (8–10), or move bulk suspended organic matter to depth during subduction events (11–13). Due to the complex 3D structure and temporal variability of these features, however, simultaneous quantification of sinking and subduction has not been achieved previously in any observational field study.

The southern California Current Ecosystem (CCE) is a productive eastern boundary current biome representative of coastal upwelling ecosystems worldwide. Nearshore waters off of Point Conception are typically cold, salty, and nutrient-rich due to upwelling, whereas the southward-flowing California Current forms a low-salinity band that separates the coastal upwelling region from oligotrophic subtropical waters further offshore. Although the coastal CCE has high primary productivity (14) and nitrate uptake (15–17), the vertical export of carbon as sinking particles (assessed by both sediment traps and ²³⁸U−²³⁴Th disequilibrium) is comparatively low (18, 19). Both models and in situ data suggest that this production−export imbalance results, in part, from lateral transport of particles produced in the coastal area to the offshore region where net export is expected (20–23). However, the expected high export ratio in offshore waters of the CCE is not supported by existing in situ measurements (19, 24).

Submesoscale and mesoscale fronts are common features in the southern CCE and are increasing in frequency (25). These features are often locations with enhanced nutrient input to the surface layer and elevated biological standing stocks and particle concentrations (26–28). To understand the potential roles of frontal systems in carbon export via both gravitational sinking and subduction, we studied a frontal region inshore of the California Current off Southern California in August 2012 using a combination of transect sampling and Lagrangian experiments (i.e., tracking the temporal evolution of water parcels). In the results presented in Results and Discussion, we (i) quantify phytoplankton carbon production; (ii) determine vertical carbon transport due to sinking particles using sediment trap and ²³⁴Th methods; (iii) evaluate potential mechanisms driving enhanced gravitational flux, including mesozooplankton fecal pellet production and diatom trace metal limitation; and (iv) estimate subduction of particles to depth using a data assimilative Regional Ocean Modeling System (ROMS) model.

Results and Discussion

We found that gravitational flux was amplified approximately twofold at the front relative to surrounding waters or typical nonfrontal regions of the CCE, and that subduction of organic matter contributed additional export of comparable magnitude. The former finding is based on high ²³⁸U−²³⁴Th deficiency and large particle fluxes into sediment traps in the frontal region. The latter finding is supported by vertical sections of ²³⁴Th, particulate organic carbon (POC), and total organic carbon (TOC) across the front as well as a physical model assimilating the results of satellite remote sensing products and numerous temperature and salinity profiles in the frontal region. In Gravitational Flux and Subduction of Organic Matter, we explain this evidence in detail.

Gravitational Flux.

The study site was a stable eddy-related frontal region (hereafter E-Front; Fig. 1) characterized by sloping isopycnals (density surfaces), with the 1,024.5 kg⋅m⁻³ to 1,024.9 kg⋅m⁻³ isopycnal surfaces outcropping near the core of the front and descending to a depth of ∼50 m on the western (offshore) side of the front (Fig. 2). The strongest surface expression of E-Front was an east−west salinity gradient, with salinity of 33.35 at the front center (Fig. S1). For further analyses, we define the eastern boundary (coastal side) of E-Front as the location where the 33.5-salinity isopleth outcrops to the surface, and the western boundary (offshore) as the location where the 1,024.2 kg⋅m⁻³ isopycnal shoals to a depth of 30 m.

Fig. 1.

(A) Study region in the CCE with satellite-tracked drifter trajectories (black tracks within rectangle) and Archiving, Validation and Interpretation of Satellite Oceanographic data (AVISO) sea-surface height anomalies (SSHa) on 9 August (cruise midpoint). Thin white tracks are locations of Lagrangian cycles on previous CCE LTER cruises (19, 22). (B and C) The ²³⁸U−²³⁴Th deficiency measured in surface seawater on (B) SeaSoar Survey 1 and (C) SeaSoar Survey 2. Thick black lines show the locations of front crossings 1 and 2. Thin black lines are sea surface salinity. (D) Probability density function of surface layer ²³⁴Th deficiency (from SeaSoar surveys). Dashed line is median value on CCE-P1208. High deficiency is indicative of prior gravitational flux.

Fig. 2.

The ²³⁸U−²³⁴Th deficiency measured on vertical sections across the front for (A) section 1 and (B) section 2. Black lines are density contours. Note the subduction feature along the 1,025 isopycnal in A (from 15 to 30 km along transect) and its absence from B. (C and D) Particulate organic carbon concentrations on (C) front crossing 1 and (D) front crossing 2. White lines show depth of the euphotic zone (0.1% light level). Vertical dashed white lines show front boundaries as defined in Gravitational Flux.

Fig. S1.

Physical structure of E-Front. (A and C) Surface salinity measured during SeaSoar (A) Survey 1 and (C) Survey 2. Drifter trajectories are overlain in color, illustrating the mean surface salinity on the numbered cycles (colored tracks are experimental arrays, and gray tracks are sediment traps). White lines are locations of front transects. (B and D) SST during SeaSoar (B) Survey 1 and (D) Survey 2. Drifter trajectories are overlain on the figure with colors showing the mean SST measured on the cycles. Colored tracks are experimental array, and thick gray tracks are sediment traps. (E and F) Vertical sections of salinity along (E) transect 1 and (F) transect 2. Black lines are isopycnals, and gray dots show sampling locations.

Surface sampling during an initial mesoscale survey (30 July to 5 August) showed high ²³⁴Th deficiency (relative to secular equilibrium with ²³⁸U) on the coastal side of the front, indicating high particle export (Fig. 1B). The E-Front core acted as a barrier separating this high-deficiency water from offshore low-deficiency waters. Three weeks later, a repeat survey showed high surface ²³⁴Th deficiency throughout the study region (Fig. 1C). Including both surveys, surface ²³⁴Th concentrations were lower than typically found in nonfrontal regions of the CCE (Fig. 1D, P < <0.01, two-sided Mann−Whitney u test), indicating high ²³⁴Th deficiency and enhanced particle export. In fact, the median deficiency was higher than all but one previous surface measurement of ²³⁴Th made in nonfrontal regions on three previous month-long cruises in the CCE. This finding suggests that export near the front was substantially higher than would be expected in the region when fronts are absent. Although enhanced vertical flux at the front is not conclusively demonstrated by the ²³⁴Th results alone due to the long half-life of ²³⁴Th (∼24 d) relative to the short residence time of water in the front (3 to 4 d, based on drifter trajectories), this conclusion is also supported by the sinking material collected in shallow drifting sediment traps.

We deployed sediment traps during five Lagrangian experiments (hereafter “cycles”) in the core of and to either side of E-Front. Temperature−salinity (T-S) plots (Fig. S2A) showed that cycle 1 sampled the portion of E-Front with outcropping isopycnals, whereas cycles 2 and 5 sampled the western portion of E-Front where front expression was predominantly subsurface (with cycle 5 slightly closer to the center of the front). Cycles 3 and 4 sampled nonfrontal water on the coastal and offshore sides, respectively, of E-Front. Sediment traps showed significantly higher carbon export near the base of the euphotic zone at the front (cycle 1, 437 ± 25 mg C⋅m⁻²⋅d⁻¹) than measured in the offshore (cycle 4, 133 ± 18 mg C⋅m⁻²⋅d⁻¹) or coastal (cycle 3, 150 ± 32 mg C⋅m⁻²⋅d⁻¹) locations. Cycle 1 export was twofold greater than the highest sediment trap flux previously measured in nonfrontal waters of the CCE (Fig. 3A), corroborating the ²³⁴Th inferences of high local export in the frontal region. Enhanced export was also found for cycle 5 (328 ± 24 mg C m⁻² d⁻¹), conducted in waters with features most closely resembling the region of subsurface expression of the front, but not for cycle 2 (150 ± 68 mg C⋅m⁻²⋅d⁻¹).

Fig. S2.

(A) T-S plot with offshore locations from the SeaSoar surveys (dots) and transects (+) in cyan, coastal regions in green, regions with surface expression of the front (blue), and regions with subsurface expression of the front (magenta, only diagnosed on transects). Filled circles indicate surface temperature and salinity measured during the five cycles. Note that cycle 1 has T-S conditions closely matching the region with surface expression of the front, whereas cycles 2 and 5 have conditions matching the region with subsurface expression of the front. (B) Mean Si_ex on Lagrangian cycles. Siex (= [H₄SiO₄] – [NO₃⁻] × R_Si:NO3). Negative values indicate Fe stress. (C) Chl a (green) and nitrate (red) concentrations in control (dashed lines) and +Fe treatments (solid lines). Bars are uptake rates in daily 5- to 6-h incubations with ¹⁵NO₃⁻ added to 500-mL aliquots from each treatment. (D) Mesozooplankton biomass (red) and grazing (green) on the Lagrangian cycles.

Fig. 3.

(A) Comparison of CCE-P1208 sediment trap-based export measurements (red) with prior sediment trap measurements made in nonfrontal regions on CCE LTER cruises (blue). Note high export for frontal cycles 1 and 5. (B) Vertical carbon flux at E-Front during transects (Left) and Lagrangian cycles (Right). Gravitational POC flux was estimated using two approaches: ²³⁸U−²³⁴Th deficiency with a simple steady-state equation without upwelling (blue, Left) and sediment traps (purple, Right). Advective carbon flux is an additional form of carbon export estimated from vertical velocities and POC (red) or ΔTOC concentrations (orange; ΔTOC is the difference between TOC measured on the transect and deepwater TOC). Boxes show quartiles, and whiskers show 95% confidence intervals. Far right shaded column in both plots indicates previous gravitational POC flux ranges in the CCE region at nonfrontal locations. Horizontal dashed lines show average previous gravitational flux in the CCE determined by sediment trap (purple) and ²³⁴Th (blue). In Right, results from nonfrontal cycles 4 and 3 (offshore/left and coastal/right, respectively) are shown for comparison.

We also estimated sinking particle flux from two high-resolution vertical sections across the front (Fig. 2) using a steady-state ²³⁸U−²³⁴Th deficiency equation (Fig. 3B). Although this approach has substantial uncertainty due to the complex 3D structure of E-Front and the short residence time of water in the feature, this simple export proxy produced results similar to the sediment traps, showing elevated export relative to typical measurements made from ²³⁴Th in the CCE. Averaging the steady-state export estimates for the two transects, 104 mg C⋅m⁻²⋅d⁻¹ (84 to 124, mean and 95% C.I.) for transect 1 and 185 (135 to 232) mg C⋅m⁻²⋅d⁻¹ for transect 2, these ²³⁴Th-based measurements for the 30-km-wide E-Front are approximately twofold higher than the typical export flux determined by ²³⁴Th in the CCE (80 mg C⋅m⁻²⋅d⁻¹). Similarly, average export from sediment trap deployments in the frontal region (305 mg C⋅m⁻²⋅d⁻¹; cycles 1, 2, and 5) was ∼2.5× higher than the typical nonfrontal value of 121 mg C⋅m⁻²⋅d⁻¹, as well as greater than the trap fluxes on the offshore or coastal sides of the front (133 and 150 mg C⋅m⁻²⋅d⁻¹, respectively).

Subduction of Organic Matter.

The two high-resolution vertical sections across the front showed contrasting spatial patterns of ²³⁴Th. Our initial transect (4−5 August) showed a subsurface deficiency maximum (30 m to 70 m depth) at the front (Fig. 2A), which could result from subduction of high-deficiency surface waters or from localized subsurface particle production and sinking. Although the latter scenario is known to occur in offshore oligotrophic regions of the CCE, we found no evidence of subsurface PP maxima in any of our five Lagrangian cycles or in a PP model parameterized from cruise data, which showed most productivity in the upper 25 m of the water column. The measured ²³⁴Th section thus suggests along-isopycnal subduction of high-deficiency water from the coastal side of the front. This along-isopycnal ²³⁴Th feature was further associated with enhanced subsurface concentrations of POC (Fig. 2C) and TOC (Fig. S3), consistent with POC subduction at the front. In contrast, similar measurements 2 wk later did not show evidence of frontal subduction (Fig. 2B) but rather a shoaling of the ²³⁴Th isocline between isopycnals 1,025 and 1,025.5 kg⋅m⁻³ caused by mild upwelling. Such differences are an expected feature of dynamic fronts, in which meanders produce alternating regions of subduction and upwelling. Such dynamics (combined with the long half-life of ²³⁴Th) also likely explain the larger-scale patterns of surface ²³⁴Th during our transects (Fig. 1). Downwelling conditions lead to convergent frontal features that set stark boundaries between coastal (high deficiency) and offshore (low deficiency) water. Upwelling conditions lead to divergent features, with high gravitational flux at the front creating a wider high-deficiency signature.

Fig. S3.

TOC concentrations measured along (A and C) transect 1 and (B and D) transect 2. C and D show ΔTOC, which is the difference between measured TOC and deep-water TOC. Note the high subsurface ΔTOC concentrations near the base of the euphotic zone (white line) along transect 1 near the core of the front (15 km to 30 km along transect), which coincide with the subduction feature from Fig. 2A.

To further investigate and quantify the export flux from subduction, we used a dynamically consistent, data-assimilating ROMS model (29, 30). Consistent with observational interpretations, modeled vertical velocities confirmed relatively strong subduction over most of E-Front during transect 1 and weak upwelling during transect 2 (Fig. 4). Using measured POC and ΔTOC (the difference between E-Front TOC and typical deep-water TOC) concentrations at the base of the euphotic zone and modeled vertical velocities, we calculated areally averaged POC subduction rates of 475 (270 to 687, 95% C.I.) and −25 (−82 to 31) mg C⋅m⁻²⋅d⁻¹ for transects 1 and 2, respectively, and ΔTOC subduction rates of 379 (195 to 549) and −17 (−60 to 28) mg C⋅m⁻²⋅d⁻¹. The fate of the subducted POC is unknown, but it coincided with high subsurface $\mathrm{N}{\mathrm{H}}_{\mathrm{4}}^{+}$ concentrations (>1 μmol⋅L⁻¹) on transect 1, suggesting rapid remineralization beneath the euphotic zone. Importantly, this advective transport, averaging 225 (POC) or 181 (ΔTOC) mg C⋅m⁻²⋅d⁻¹, must be added to the gravitational flux (305 mg C⋅m⁻²⋅d⁻¹) to determine total export enhancement at E-Front, because the subducted bulk suspended material is not measured by sediment traps or ²³⁴Th:²³⁸U deficiency.

Fig. 4.

Vertical velocity sections at 34°36’N from the data assimilating ROMS model. Each section is a 6-d average centered at the midpoint sampling time of cross-frontal (A) transect 1 or (B) transect 2. Black lines are isopycnals. Black circles are ²³⁴Th sampling locations. Red shades are positive (upward) velocities, and blue shades are negative (downward) velocities. Note the strongly downward velocities along the core of the front in A and the weak upwelling velocities in B.

Using a chlorophyll−light diagnostic model parameterized with measured ¹⁴C PP from the Lagrangian cycles, we calculated PP across the transects to determine the fraction of PP exported as POC out of the base of the euphotic zone. Because typical formulations of the e ratio (export/PP) do not account for advective flux, we define the e_POC ratio as the sum of advective and sinking POC export divided by ¹⁴C PP. The calculated e_POC ratio of 67% for transect 1 exceeded the average f ratio at the front (52%) determined using $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ and $\mathrm{N}{\mathrm{H}}_{\mathrm{4}}^{+}$ concentrations and an ecosystem model parameterized for the CCE. For transect 2, the e_POC ratio of 11% was lower than the f ratio (45%).

Mechanisms of Enhanced Gravitational Flux.

To elucidate the mechanisms causing enhanced gravitational export at the front, we investigated phytoplankton and zooplankton dynamics during each Lagrangian experiment. Cycle 1 (at the front) had higher vertically integrated ¹⁴C PP (1,451 ± 140 mg C m⁻²⋅d⁻¹) than coastal or offshore cycles (Fig. 3A). Previous nonfrontal experiments exhibited decreased export efficiency (e ratio) when PP was high (19, 24). In contrast, at the front (cycle 1), the plankton community maintained a high e ratio (30%) despite enhanced phytoplankton production.

We evaluated two nonexclusive mechanisms that could cause an increase in export efficiency at the front: mesozooplankton fecal pellet production and increased silicification by diatoms. Mesozooplankton fecal pellets are major contributors to sinking flux in the CCE (19). Cycle 1 showed substantially enhanced grazing (and mesozooplankton biomass) relative to the other cycles, with grazing rate estimates (6.3 ± 2.9 mg Chl a m⁻²⋅d⁻¹) on the high side of previous measurements in the CCE (Fig. S2D). Similarly, mesozooplankton herbivory was enhanced at E-Front in transect samples (Fig. 5 E and F), and we measured high concentrations of phaeopigments and high phaeopigment:Chl a ratios (> 4, highest for cycles 1 and 5) in the sediment traps, indicative of high fecal pellet flux produced by herbivorous zooplankton.

Fig. 5.

Biological and chemical sections during (A, C, and E) transect 1 and (B, D, and F) transect 2. (A and B) Si_ex (= [H₄SiO₄] – [$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$] × R_Si:NO3). Negative values are indicative of Fe stress. (C and D) Variable fluorescence (Fv/Fm). Low values are indicative of Fe stress. (E and F) Mesozooplankton biomass and grazing rates from vertical net tows. Gray lines show surface salinity (gradient region indicates location of the surface expression of E-Front).

Fe limitation has been shown to drive enhanced export in the CCE by increasing silicification by diatoms that continue to take up Si despite reduced organic matter production (31). The increased cellular Si:N ratios amplify the ballasting effect of diatoms, leading to higher sinking rates of aggregates and fecal pellets that contain diatoms. Decoupled Si and N uptake also leads to low dissolved Si(OH)₄:$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ ratios in the water column. Negative values of Si excess [Si_ex = [Si(OH)₄] − [ $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$] × R_Si:NO3, where R_Si:NO3 is the Si(OH)₄:NO₃ ratio of upwelled water, equal to 1 mol:mol for the CCE (32)] are diagnostic of Fe limitation in the CCE and are evident in samples from the E-Front transects (Fig. 5 A and B) and cycle 1 (Fig. S2B). Conversely, Si_ex values were high (and grazing low) for cycle 2, the only near-front cycle that did not show enhanced export. Variable fluorescence (indicative of the photosynthetic status of phytoplankton and correlated with Fe availability) was low at the front (Fig. 5 C and D), consistent with Fe limitation. The ratio of nitrate:dissolved Fe (micromolars of $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$:nanomolars of dFe) is another diagnostic feature of Fe limitation in the CCE, with ratios in excess of 5 indicating significant potential for Fe limitation of diatoms (31–33). Consistent with our interpretation of Fe limitation in the main axis of the front, we found $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$:dFe values of 10 to 26 in the upper 50 m during cycle 1 and consistently >5 on the coastal edge of the E-Front transects. A 3-d deckboard Fe-enrichment experiment during cycle 1 also showed increased $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ drawdown, ¹⁵$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ uptake, and biomass production in Fe-amended bottles relative to control incubations (Fig. S2C). Thus, increased efficiency of carbon export at E-Front appears to be linked to Fe limitation of diatoms, which leads to increased silicification and rapid sinking of the heavily ballasted fecal pellets produced by zooplankton grazing, although other mechanisms may contribute as well.

Export at Fronts.

Although our study focused on a single mesoscale feature, the two mechanisms that drove high gravitational flux at E-Front are likely common for eastern boundary upwelling systems (EBUS), where fronts are typically associated with elevated phytoplankton and zooplankton biomass (26, 34). For example, Fe limitation was shown to increase export in a gradient region between cyclonic and anticyclonic eddies close to Point Conception (31), and significantly elevated mesozooplankton biomass and organic aggregate abundance (28) were demonstrated at a front ∼300 km southeast of E-Front.

Front frequency in the CCE, measured by autonomous in situ gliders along California Cooperative Oceanic Fisheries Investigations (CalCOFI) lines 80 (near our study site) and 90 (extending offshore from the Southern California Bight), indicates that 8% of CCE water is within 15 km of a density front (34). If we assume that the greater than twofold enhancement of gravitational flux at E-Front is generally representative of the region, then over 14% of the total sinking particle flux in the region occurs near mesoscale fronts. Additionally, our mean estimate of ∼225 mg C m⁻²⋅d⁻¹ for advective POC export in subducted water parcels at E-Front suggests that total particle export at CCE fronts may exceed 25% of regional gravitational flux. Given the global importance of EBUS in total oceanic production and the similarity in their physical dynamics (35, 36), we can surmise that EBUS fronts are likely globally important loci for carbon transport into the ocean’s interior.

Although the present results clearly indicate the importance of mesoscale fronts to carbon sequestration in a contemporary coastal ocean system, they likely underestimate the role of fronts in a warmer, future climate. Within the CCE, a decadal-scale trend of increasing frontal frequency has been linked to long-term increases of upwelling favorable winds (25, 37). Continued strengthening of land−sea temperature differences, combined with increased stratification in other regions, will likely further increase the importance of particle export at mesoscale fronts to the global carbon cycle.

Materials and Methods

Cruise Overview.

Our sampling scheme involved three distinct aspects: (i) 3D mapping of the large-scale physical structure of the front with a towed SeaSoar instrument; these surveys (referred to as SeaSoar surveys) were combined with surface mapping of ²³⁴Th deficiency and other biogeochemical properties in the region; (ii) Nearly synoptic 50-km transects across the frontal feature while measuring biogeochemical and ecological properties at 6 to 8 depths at 10 to 13 stations across the front; and (iii) Lagrangian experiments (referred to as “cycles”) during which we followed an in situ array drogued at 15 m depth, on which we attached bottles for experimental incubations including H¹⁴$\mathrm{C}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ uptake (38). The array provided a moving frame of reference for a suite of other measurements including mesozooplankton biomass and grazing rates, ²³⁴Th:²³⁸U disequilibrium measurements, nutrients, POM, and biological standing stocks. An identically drogued sediment trap array was deployed simultaneously on each cycle.

The front was initially located using satellite sea surface temperature (SST) and sea surface height (SSH), an autonomous Spray glider, and a free-fall Moving Vessel Profiler (28), then the region was mapped (SeaSoar Survey 1) from 30 July to 2 August. This initial survey was immediately followed by cross-frontal transect 1 (4 August, 1515 hours to 5 August, 1115 hours). After transect 1, we conducted the five cycles in different locations relative to the front. Cycles 1 to 5 lasted from 6 to 9 August, 10 to 12 August, 13 to 15 August, 16 to 18 August, and 18 to 20 August, respectively. After completing these experiments, cross-frontal transect 2 was conducted in waters near the location of transect 1 from 20 August, 1645 hours to 21 August, 0745 hours. The region was mapped again (SeaSoar Survey 2) from 21 to 25 August.

Export Measurements.

Particle-interceptor trap (PIT)-style sediment traps (19, 39) with 8 to 12 cylindrical, 70-mm-diameter tubes with baffles on top were deployed at a depth near the base of the euphotic zone as determined from fluorescence profiles (60 m for cycles 1 and 3, 70 m on cycles 2 and 5, and 100 m on cycle 4). Tubes were deployed with a formaldehyde brine for a period of ∼2.25 d. After recovery, overlying material was removed by suction, and samples were split for C and N analyses by CHN analyzer, C:²³⁴Th ratio analysis, and pigment measurements (Chl a and phaeopigments) by the acidification method. The ²³⁴Th concentrations were measured using standard small volume methods (40), including a ²³⁰Th tracer spike, filtration and beta counting at sea on a RISO beta counter, background beta counts >6 mo after the cruise, gravimetric addition of ²²⁹Th, and quantification of the ^229:230Th ratio by inductively coupled plasma (ICP) MS at the Woods Hole Oceanographic Institution Analytical Facility. The ²³⁸U−²³⁴Th deficiency was calculated after estimating ²³⁸U activity from salinity (41), assuming steady state, and vertically integrating.

Biological and Chemical Measurements.

Samples were collected by Niskin bottle for measurements of biological and chemical standing stocks and rates. Chl a was measured by fluorometer with acidification. POC was measured with a CHN analyzer. TOC was measured on a Shimadzu analyzer. Nutrients [$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$, $\mathrm{N}{\mathrm{H}}_{\mathrm{4}}^{+}$, $\mathrm{P}{\mathrm{O}}_{\mathrm{4}}^{\mathrm{3-}}$, Si(OH)₄] were measured by autoanalyzer. Phytoplankton variable fluorescence was measured using the Advanced Laser Fluorometer (42). PP was measured by uptake of H¹⁴$\mathrm{C}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ in triplicate 250-mL bottles incubated in situ for 24 h. Mesozooplankton were collected by vertical (on front transects) or oblique (during semi-Lagrangian cycles) net tows with a 0.71-m diameter, 202-μm mesh bongo net. Mesozooplankton grazing rates were determined from gut fluorescence measurements using gut turnover times calculated from a temperature-dependent equation (43).

Physical Model and POC Subduction.

To measure the passive transport of organic carbon by subduction, we first used kriging to compute gridded fields of POC and ΔTOC (where ΔTOC is the difference between measured shallow TOC and average deep TOC concentrations) along the two transects. We then calculated flux using the equations J = [POC] × w or J = [ΔTOC] × w, where J is flux (milligrams C per square meter per day) and w is the vertical velocity (meters per day) derived from a dynamically consistent data-assimilative model. Data assimilation was conducted within the ROMS with a 9-km grid resolution using a four-dimensional variational approach that repeatedly adjusted initial and boundary conditions to minimize the mismatch between the model and physical measurements (e.g., temperature, salinity) measured on our cruise (30).

SI Materials and Methods

Cruise Overview.

The August 2012 Process cruise of the CCE LTER program on the R/V Melville consisted of three distinct components: two mesoscale towed SeaSoar surveys to determine the physical structure of the frontal region, two semisynoptic transects across the front with conductivity temperature depth (CTD) casts to a depth of 300 m (with samples only drawn from the upper 100 m), and five quasi-Lagrangian experiments (hereafter “cycles”) of 1- to 3-d duration conducted in the general vicinity of the front (Fig. 1A). During the SeaSoar surveys (which covered a region of ∼1° × 1° and lasted ∼4 d), we sampled near surface from the ship’s uncontaminated flow-through system for dissolved nutrients, Chl a, and ²³⁴Th, with a horizontal resolution of ∼14 km across the front and ∼30 km along the front (Fig. 1 B and C). On the two frontal transects, we traversed a total of 45 to 50 km with 13 or 10 CTD casts over the course of 20 h or 16 h. CTD profiles extended to 300 m depth, although seawater samples (nutrients, POC, TOC, ²³⁴Th, Fv/Fm) were taken only from the upper 100 m (six to eight depths). Quasi-Lagrangian experiments (cycles) were initiated with the deployment of a sediment trap array (19). An experimental array with mesh bags was used for in situ incubation of ¹⁴C PP at eight depths (38). Each array included a satellite-equipped float and a holey-sock drogue centered at a depth of 15 m. Water-column concentrations of Chl a and nutrients were measured daily at eight depths in the euphotic zone, and ²³⁴Th concentrations in the upper 200 m were measured once or twice per cycle at the drifter location.

Plankton Community Measurements.

Samples for Chl a (282 mL) were filtered on Whatman glass fiber (GF/F) filters, extracted in 90% (vol/vol) acetone for 24 h and analyzed fluorometrically before and after acidification (44). Fifty-milliliter samples for nutrient analysis (nitrate, ammonium, and silicic acid) were collected from Niskin bottles, immediately gravity-filtered through a 0.1-μm-pore-size cartridge filter, and frozen for analysis at the University of California, Santa Barbara Analytical Facility. One-liter samples were filtered through precombusted GF/F filters for POC concentration and analyzed at the Scripps Institution of Oceanography Analytical Facility. Forty-milliliter acidified (pH < 2) samples were collected from Niskin bottles on the transects and measured for TOC concentration by high-temperature combustion (Shimadzu TOC-V analyzer). The ¹⁴C PP was measured using quadruplicate 250-mL samples (including a dark incubation) incubated in situ on the Lagrangian array at eight depths on each day of every cycle. Polycarbonate bottles were filled from the Niskin bottle with silicon tubing, spiked with H¹⁴$\mathrm{C}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ and incubated for 24 h. Samples were then filtered onto a GF/F filter and acidified with HCl to remove inorganic ¹⁴C, and beta decays were measured on a Beckman−Coulter scintillation counter following addition of scintillation mixture. Phytoplankton F_v/F_m was measured (both underway and on discrete samples drawn from Niskin bottles) using the Advanced Laser Fluorometer (42, 45). Mesozooplankton were collected from either vertical (front transects) or oblique (Lagrangian cycles) bongo tows with a 0.71-m-diameter, 202-μm mesh net with filtering cod end. Samples were split with two fractions strained through a series of nested filters (5, 2, 1, 0.5, and 0.2 mm) and frozen in liquid nitrogen for either dry weight or gut pigment measurements. Grazing rates were determined from mesozooplankton pigment concentration (Chl a + phaeopigments) using gut turnover times calculated from a temperature-dependent equation (43): K (per minute) = 0.0124 e^{0.0765T (°C)}.

Sample Collection for Dissolved Fe.

Trace metal clean samples were collected using Teflon-coated 5-L Niskin-X bottles (Ocean Test Equipment) mounted on a powder-coated rosette, equipped with a CTD and auto-fire module (Seabird Electronics), suspended from a coated metal cable (Space-Lay Wire Rope). Bottles were tripped at preprogrammed depths during the up-cast, while moving upward at minimum winch speed. Immediately following retrieval, Niskin-X bottles were transferred into a Class 100 trace metal clean van and filtered in-line using acid-washed Teflon tubing and an acid-washed Acropak-200 (0.2 µm) capsule filter pressurized by filtered air. Filtered samples for dissolved iron (dFe) analysis were placed in 250 mL of acid-cleaned low-density polyethylene bottles, acidified to pH 1.8 (Optima HCl), and stored until analysis in the laboratory.

Dissolved Iron Analysis.

DFe samples were analyzed via flow injection analysis using sodium sulfite for reduction of dFe (33, 46). This method has been shown to be sensitive and provide accurate results with respect to Sampling and Analysis of Fe (SAFe, D2) and the GEOTRACES program (GS) consensus samples. Values obtained for D2 (0.91 ± 0.02 nmol⋅L⁻¹, n = 10), and GS (0.57 ± 0.02 nmol⋅L⁻¹, n = 8) compare well to the most recent consensus values. Deckboard grow-out experiments were conducted to test for Fe limitation in 4-L acid-cleaned polycarbonate bottles. Two controls and two treatments with 5 nmol⋅L⁻¹ DFe added were incubated for 3 d. NO₃⁻ and Chl a were measured daily, and, each day, a 500-mL subsample was drawn, spiked with ¹⁵$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ (600 nmol⋅L⁻¹, final concentration), and incubated for 5 h to 6 h to calculate ¹⁵$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ uptake rates.

PP Model.

Because PP was not measured on the frontal transects, we parameterized a simple diagnostic model to predict PP from Chl a and photosynthetically active radiation (PAR). Because we were interested not in instantaneous growth rates (which are dependent on surface irradiance) but rather in the average PP expected for any sample on the transect, we normalized to PAR. The model was formulated as

$$\text{PP}=\text{chl}\times P_{max}\times \left[\left(1-\text{exp}\left(\raisebox{1ex}{$\%\text{PAR}$}\!\left/ \!\raisebox{-1ex}{$I_k$}\right.\right)\right)\right],$$

where P_max is the maximum chl-specific production rate of the phytoplankton community [(milligrams C per cubic meter per day) /(milligrams Chl a per cubic meter)], %PAR is the percent surface irradiance, and I_k is a light saturation parameter. P_max and I_k were determined from a least squares minimization of the model−data misfit for ¹⁴C PP measurements made on the five cycles after log transformation to normalize the ¹⁴C PP distribution. Because the data were log-transformed, all samples with measured ¹⁴C PP = 0 were excluded from the analysis, leaving a total of 71 samples that were used for model parameterization. Optimized parameters were P_max = 45.7 mgC⋅mg Chl a⁻¹⋅d⁻¹ and I_k = 9.5. Residual plots of model−data residuals vs. $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ showed no consistent model overestimation at low $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$_, suggesting that the addition of nutrient limitation would not improve the model fit.

Because $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ uptake was not measured on the cruise, we determined the f ratio using a phytoplankton model based on North Pacific Ecosystem Model for Understanding Regional Oceanography (NEMURO) (47), parameterized using in situ CCE rate measurements (48). In this model, phytoplankton nutrient limitation is modeled using the equation

$$f\left(N\right)=\frac{{\text{NO}}_3}{{\text{NO}}_3+K_{\text{NO}3}}\text{exp}\left(-\psi \times {\text{NH}}_4\right)+\frac{{\text{NH}}_4}{{\text{NH}}_4+K_{\text{NH}4}},$$

where K_NO3 and K_NH4 are half-saturation constants for $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ and $\mathrm{N}{\mathrm{H}}_{\mathrm{4}}^{+}$, respectively, and ψ is an ammonium inhibition parameter for $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ uptake. This equation can be rearranged to calculate an f ratio, regardless of light or other limiting factors,

$$f\hspace{0.17em}\text{ratio}=\frac{\frac{{\text{NO}}_3}{{\text{NO}}_3+K_{\text{NO}3}}\text{exp}\left(-\psi \times {\text{NH}}_4\right)}{\frac{{\text{NO}}_3}{{\text{NO}}_3+K_{\text{NO}3}}\text{exp}\left(-\psi \times {\text{NH}}_4\right)+\frac{{\text{NH}}_4}{{\text{NH}}_4+K_{\text{NH}4}}}.$$

Because diatom N uptake was likely substantially reduced by Fe stress in the frontal region (Results and Discussion), we use the Li et al. (48) parameterizations for nondiatoms of K_NO3 = 1.0 μmol⋅L⁻¹, K_NH4 = 0.1 μmol⋅L⁻¹, and ψ = 1.5 μmol⋅L⁻¹. We multiplied these f ratios by PP estimates across the transect to determine nitrate uptake rates, and then divided vertically integrated nitrate uptake rates by vertically integrated PP to determine average f ratios across the transects.

Sediment Trap.

VERTEX-style sediment traps (4 to 12 tubes per crosspiece) were deployed near the base of the euphotic zone (60 m on cycles 1 and 3, 70 m on cycles 2 and 5) and at 100 m (19, 39). Tubes were filled with a formaldehyde-preserved salt brine before deployment [0.1 μm filtered seawater, 50 g⋅L⁻¹ NaCl, 4% (wt/vol) formaldehyde]. Deployment duration varied from 30 h to 78 h. After recovery, lower-salinity water above the brine interface was removed by gentle suction, and the remaining material was filtered through a 200-μm filter, which was sorted microscopically to remove swimming mesozooplankton. Samples were then split on a rotary splitter and filtered through either GF/F (for C/N analyses) or quartz (QMA) filters (for C:²³⁴Th ratios). Samples for C and N were analyzed at the Scripps Institution of Oceanography Analytical Facility.

Thorium-234 Measurements.

Thorium-234 is a radioactive isotope (half-life = 24.1 d) produced from the decay of ²³⁸U. Whereas ²³⁸U is conserved in the surface ocean (covarying with salinity), ²³⁴Th is scavenged onto particles. Thus, measurements of ²³⁸U−²³⁴Th disequilibrium serve as a proxy for the removal of sinking particles from the surface ocean (49, 50). Total ²³⁴Th concentrations were measured from 4-L samples taken either from the ship’s flow-through system (SeaSoar Surveys) or from Niskin bottles (cycles and front transects) using established small volume methods (40, 51). After collection, samples were acidified with HNO₃ and 1 mL of ²³⁰Th was added as a yield tracer. Bottles were then shaken and stored for 4 h to 9 h. Samples were then basified to a pH of 8 to 9 with NH₃OH, and KMnO₄ and MnCl₂ were added. After >9 h of coprecipitation of Th with manganese oxide, samples were vacuum-filtered onto a QMA filter, mounted on a RISO sample holder, and counted on a RISO low-level beta multicounter at sea. After >6 mo, samples were recounted for background. Yield analyses were then performed by dissolving manganese oxide from filters in 10% (vol/vol) H₂O₂/8M HNO₃, gravimetrically adding ²³⁰Th, and determining the ratio of ²³⁰Th:²²⁹Th by mass spectrometry at the Woods Hole Oceanographic Institution Analytical Facility. During the Lagrangian cycles, ²³⁴Th flux was calculated assuming a steady-state equation of ${}^{234}\text{T}\text{h}\hspace{0.17em}\text{export}=\left(A_{238\text{U}}-A_{234\text{Th}}\right){\lambda }_{234\text{Th}}$, where A_238U and A_234Th are the vertically integrated activities of ²³⁸U and ²³⁴Th, respectively, and λ_234Th is the decay constant of ²³⁴Th (52). The ²³⁸U was determined from salinity using the equation ²³⁸U (dpm per liter) = 0.0704 × salinity (53). To determine carbon flux from ²³⁴Th export, we multiplied by the C:²³⁴Th ratio of particles collected in the sediment traps.

Depth of the Euphotic Zone.

Because most CTD casts on the cruise were conducted at night, we used beam transmission to estimate the depth of the euphotic zone. Measured light extinction coefficients from daytime CTD casts were used to establish a predictive relationship with simultaneously measured beam transmission. This relationship was subsequently used to calculate vertically varying light extinction coefficients and depth of the euphotic zone (0.1% light level) for each nighttime CTD cast.

Data-Assimilating ROMS Model.

To calculate vertical velocities at the front for estimating advective POC export, we used a dynamically consistent data assimilation approach (29, 30). In situ CTD data (temperature, salinity) and satellite-derived sea surface height and temperature were assimilated into a ∼9-km grid resolution model domain from 30°N to 40°N and 115°W to 131°W with 42 vertical layers. Data assimilation was conducted within the ROMS using a four-dimensional variational approach that adjusts initial conditions and surface forcing to minimize differences between model and in situ and remote measurements. The model estimated ocean states for 30 d spanning the length of the cruise. Model vertical velocities were extracted for times, locations, and depths corresponding to our in situ transects and cycles. With ∼9-km grid resolution, the ROMS model results do not capture some of the fine-scale variability and ageostrophic secondary circulation that operate at subgrid spatial scales (54). This submesoscale circulation may be nonnegligible. However, even without resolving submesoscale flows, the model agreed with tracer-derived evidence of subduction on transect 1 and upwelling on transect 2 and provided reasonable mean estimates of upwelling and downwelling velocities in different regions of E-Front throughout our occupation of the region.

Vertical Advective Organic Carbon Flux.

To assess the flux of organic carbon transported out of (or into) the euphotic zone by subduction (or upwelling), we used separate approaches for POC and ΔTOC concentrations, where ΔTOC is the difference between measured TOC concentrations on the transect and the expected TOC concentrations in water masses upwelled in the CCE (i.e., refractory TOC). Because deep-water TOC is likely variable depending on the source water for upwelling, we computed ΔTOC by subtracting TOC concentrations representative of deeper source water from concentration in surface water of similar density, as determined from a time series of samples taken in the study region. We then calculated advective carbon flux using the equation J = [POC] × w or J = [ΔTOC] × w, where J is flux (milligrams C per square meter per day) and w is the vertical velocity (meters per day). These equations thus assume that POC is predominantly produced in the euphotic zone of the CCE (i.e., there is negligible refractory POC upwelled into the CCE), but that TOC includes a large refractory component. POC and ΔTOC advective flux can then be considered net advective flux estimates, which are considerably less than gross organic carbon flux (calculated by multiplying vertical velocities by the total TOC concentration). [POC] and [ΔTOC] at the depth of the euphotic zone for each profile on the frontal transect were determined from linear extrapolation of the two nearest POC or ΔTOC measurements. Vertical velocity (w) at the same locations and times was determined from daily averages of vertical velocities extracted from the ROMS model (Fig. 4). Uncertainty in w was determined from the variability in daily averaged vertical velocities over the 6-d window centered at the sampling point.

Thorium-234 Export Across the Frontal Transect.

We determined ²³⁴Th flux out of the euphotic zone along the frontal transect using a one-dimensional steady-state approximation without advection (52),

$${\text{Export}}_{\text{eup}}=\hspace{0.17em}\underset{z=0}{\overset{eup}{{\displaystyle \int }}}\left(A_{\text{U}238}-A_{\text{Th}234}\right){\lambda }_{\text{Th}234}dz,$$

where A_U238 is the activity of ²³⁸U, A_Th234 is the activity of ²³⁴Th, λ_Th234 is the decay constant of ²³⁴Th, eup is the depth of the euphotic zone, and z is depth. The comparison of sediment trap flux to steady-state estimated export on our Lagrangian cycles (Fig. S4A) suggests that this provided a reasonable estimate of vertical ²³⁴Th flux. However, the steady-state approximation led to a large overestimate for cycle 3 (likely tied to local upwelling), consistent with previous findings (55) that ²³⁴Th advective processes need to be considered in the ²³⁴Th budget for regions with high kinetic energy. We thus consider the one-dimensional steady-state approach to be a useful proxy, but caution that it can overestimate or underestimate flux at a particular location.

Fig. S4.

Sediment trap ²³⁴Th results. (A) Agreement between ²³⁴Th flux into sediment traps (x axis) and simultaneous estimates of ²³⁴Th flux (at the same depth horizon) made using a one-dimensional steady-state ²³⁴Th equation without upwelling. Gray symbols are measurements made in nonfrontal regions on previous CCE LTER cruises. Blue diamonds are shallow sediment traps. Red squares are 100-m sediment traps. Error bars are ± SD of replicate trap tubes. (B) Type I linear regression of C:²³⁴Th ratio of sinking particles collected in the sediment trap (y axis) on the average POC concentration in the water column above the sediment trap: C:²³⁴Th = 54 + 0.30 × POC, r² = 0.53, P < 0.05.

Because measurements of the C:²³⁴Th ratio were not logistically possible at each station along the front transects for converting ²³⁴Th export to carbon equivalents, we used a regression of C:²³⁴Th ratio (micrograms C per dpm) from sediment trap measurements on the Lagrangian cycles against the average POC concentration (milligrams C per cubic meter) in the water column above the sampling horizon (C:²³⁴Th = 54 + 0.30 × POC, Fig. S4B; regression details in SI Materials and Methods, Statistical Analyses).

Statistical Analyses.

Uncertainties for in situ measurements (e.g., POC, ²³⁴Th counting statistics, ^229:230Th molar ratio for yield analyses) were assumed to be normally distributed about the mean; hence, uncertainties for individual measurements were determined by SE propagation methods. When duplicate or triplicate measurements were made (e.g., for sediment trap measurements), uncertainty was derived from the SE of the replicated measurements. Unless otherwise stated, all reported measurements are means ± SEs.

Spatial interpolations of measured variables (e.g., ²³⁴Th and POC) for front transects and mesoscale mapping were determined objectively using kriging (56–58). Two-dimensional empirical semivariograms were computed for each dataset and used to calculate best fits to an anisotrophic semivariogram model calculated from a spherical model. Ordinary kriging techniques were then used to compute gridded property estimates for ²³⁴Th, ²³⁸U−²³⁴Th deficiency, and POC and TOC concentration (Figs. 1 and 2 and Fig. S3).

To estimate the C:²³⁴Th ratio of sinking particles on frontal transects when cruise sampling did not allow deployment of sediment traps or in situ pumps, we used a relationship between C:²³⁴Th ratios measured by sediment traps during our Lagrangian cycles and the average POC concentration in the water column above the sediment traps. Because our goal was prediction of the C:²³⁴Th ratio from the average POC concentration, we used a Type I linear regression. Confidence bounds on the functional relationship and on the specific locations at which we needed to predict the C:²³⁴Th ratio were determined using the functions “fit” and “predint” from the Matlab Statistics toolbox. The uncertainty in predicting C:²³⁴Th value at a specific point is greater than the uncertainty in the functional relationship plotted in Fig. S4B.

Calculation of vertical carbon fluxes at the euphotic zone depth along the front transect (e.g., in Fig. 3B) depended on several different types of measurements (e.g., in situ ²³⁴Th or POC measurements, model-derived average vertical velocities, C:²³⁴Th ratios estimated from linear regression with average POC concentration). Because the uncertainty estimates could not be assumed to be normal, we used Monte Carlo techniques (sampling with replacement, 1,000 iterations) to determine probability distributions for each derived flux estimate, and used these distributions to calculate mean, median, 95% confidence intervals, and quartiles.

SI Results and Discussion

Front Hydrography.

E-Front separated a cyclonic eddy (CE, on the coastal side of the front) from an anticyclonic eddy (AE) farther offshore (59, 60). E-Front separated dense, salty water in the CE from the fresher water of the AE. The core of the front was most evident in plots of sea surface salinity, with the center of the front at a salinity of ∼33.35 (Fig. S1). Density gradients across the frontal region (although substantial) were weaker than salinity gradients because a cold-water tongue descended along the front core (Fig. S1 B and D). This low-temperature feature may represent the influence of a cold-water filament originating in the north. The vertical structure of E-Front (Fig. S1 E and F) exhibited sloping isopycnals, with the 1,024.7 kg⋅m⁻³ isopycnal outcropping near the core of the front and descending to a depth of ∼50 m to 60 m farther offshore. Along these sloping isopycnals, our first transect through E-Front showed salty, relatively warm water extending fairly deep into the front along the 1,025 kg⋅m⁻³ isopycnal, whereas the second transect (16 d later) showed this salty water restricted to shallower depths as it was displaced deeper in the water column by colder, fresher water representative of the deep water beneath the AE.

E-Front had a distinct T-S signal that we used to diagnose the relative locations of each of our Lagrangian cycles with respect to the front (Fig. S2A). Cycle 1, on the coastal edge of the front, had both frontal and coastal influences. Cycles 2 and 5 both showed substantial influence of offshore water from the AE, but with an additional influence of the front water. For cycle 2, this frontal influence was evident in colder surface temperatures than in the AE, whereas, for cycle 5, the influence was seen in more saline sea surface waters than in the AE. It thus seems likely that these two cycles were positioned on the offshore edge of the front, where the front had weaker surface expression, but sloping isopycnals likely had a substantial influence on the subsurface water column. Surface T-S plots for cycles 3 and 4 indicated that these cycles were in the CE (coastal) and AE (offshore), respectively.

Export Measurement Accuracy.

Sediment trap collection efficiency was determined by comparing ²³⁴Th flux into sediment traps to ²³⁴Th export rates from water-column measurements and simple ²³⁸U−²³⁴Th deficiency models (61). We previously deployed these sediment trap arrays on cruises in the CCE and the Costa Rica Dome (19, 24, 62) during Lagrangian cycles in water parcels that were selected for their mesoscale homogeneity (i.e., absence of fronts) where the steady-state ²³⁴Th model is more likely valid. These prior measurements have consistently shown relatively close agreement between sediment traps and ²³⁴Th, with the exception of one cycle during October 2008 (Fig. S4A; excluding cycle 4 of P0810, r² was 0.52, P < 0.01). The anomalous cycle occurred near Point Conception shortly after a shift to strong winds out of the north, which likely created strong upwelling and introduction of ²³⁴Th to the surface waters. We thus suspect that, for that experiment, the ²³⁴Th measurements underestimated true flux, rather than the sediment traps overestimating export. If we exclude that cycle and consider only the other CCE data, the median ratio of sediment trap:steady-state derived ²³⁴Th flux is 1.04. A similar analysis for the data from the Costa Rica Dome resulted in a sediment trap:steady-state ratio of 1.22. This agreement between independent methods in two separate regions suggests that our sediment trap arrays were relatively accurate collectors of sinking particles.

We can similarly compare the steady-state ²³⁴Th model to the ²³⁴Th collected by sediment traps during the five cycles of CCE-P1208 (Fig. S4A). Results showed relatively good agreement for four of the five cycles, but showed substantial overestimation of sinking flux by the steady-state ²³⁴Th method for cycle 3; this was likely due to downwelling (average vertical velocity derived from the ROMS model for the upper 100 m was −2.3 m⋅d⁻¹ at cycle 3 drifter locations and between −0.8 and 0.8 m⋅d⁻¹ for other cycles), which leads to net flux of ²³⁴Th out of the upper layer that is not accounted for in the standard steady-state ²³⁸U−²³⁴Th disequilibrium model. This result highlights the importance of including vertical transport in ²³⁴Th models in the vicinity of E-Front. However, we caution against using instantaneous vertical velocities in steady-state thorium models that include advection, because vertical velocities experienced by the water parcel throughout the 24.1-d half-life of ²³⁴Th affect the ²³⁴Th concentrations measured.

Plankton Ecosystem Structure.

One of our primary goals was to determine whether potentially enhanced export at the front resulted from in situ plankton processes or simply from the advection of high-particle-load water from other regions. Toward this end, we assessed several properties of the phytoplankton and zooplankton communities associated with biomass, activity, and physiological status. Detailed planktonic ecosystem measurements were focused on the Lagrangian cycles when additional time on station allowed for more measurements, but were also made, to a lesser extent, on our frontal transects.

The e ratio (ratio of export/¹⁴C PP) gives an estimate of ecosystem export efficiency. Sediment trap e ratios were 30% for cycle 1 (the most conclusively front-related cycle) and 16% and 33% for cycles 2 and 5, respectively (the two cycles that were likely on the offshore edge of the front). These export efficiencies contrast with e ratios of 20% for cycle 3 (coastal) and 67% for cycle 4 (offshore). It thus seems that the enhanced export measured at the front (and particularly on cycle 1) is due not to a change in efficiency but rather to increased PP at the front. Prior export measurements in the CCE have shown a strong negative correlation between ¹⁴C PP and export efficiency (19, 22); e ratios > 10% are common when ¹⁴C PP < 700 mg C⋅m⁻²⋅d⁻¹ and often exceed 20%. However, at higher productivity, export efficiencies decrease substantially. Above ¹⁴C PP of 1,000 mg C⋅m⁻²⋅d⁻¹, export efficiency seldom exceeded 10% (figure 12 of ref. 19). Near E-Front, this relationship no longer appears to hold, and export efficiency remains relatively high even at high primary productivity rates. E-Front was thus anomalous not in its high production rates but rather in the fact that export efficiency did not decrease at high productivity levels.

In cold CCE water parcels, diatoms are the dominant ballasted phytoplankton taxa. Evidence of diatom Fe stress can be seen in decreased variable fluorescence, decreased dissolved Si:$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$, and increased $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$:dFe, all of which were found in the E-Front region. Si_ex (a conservative tracer equal to [H₄SiO₄] – [$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$] × R_Si:NO3, where R_Si:NO3 is the H₄SiO₄:$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ ratio of water upwelled in the CCE, which is approximately equal to 1) was slightly less than −1 μmol⋅L⁻¹ in surface waters on cycle 1 (the most front-influenced cycle) and decreased to less than −2 μmol⋅L⁻¹ at 20 m depth before increasing to ∼−1 μmol⋅L⁻¹ at the base of the euphotic zone (Fig. S2B). During cycle 1, we also conducted 3-d deckboard growout experiments (two controls and two with +5 nmol⋅L⁻¹ dissolved Fe). Results showed increased Chl a, decreased $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$, and enhanced rates of ¹⁵$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ uptake in the +Fe treatments, confirming Fe limitation (Fig. S2C). Cycle 3 (coastal) also showed evidence of subsurface Fe limitation, with surface Si_ex approximately equal to 0, but decreasing with depth in the euphotic zone. Similar evidence for Fe limitation can be seen in cross-frontal nutrient sections from the front transects. Si_ex was less than 0 in surface waters at the core of the front and on the front’s coastal edge, and reached values close to −3 μmol⋅L⁻¹ within the subsurface core of the front (Fig. 5 A and B). Patterns of elevated near-surface $\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$:dFe ratios were generally correlated with low Si:$\mathrm{N}{\mathrm{O}}_{\mathrm{3}}^{\mathrm{-}}$ ratios in both cycles and transects. Diatom Fe limitation at the front is also supported by variable fluorescence (Fig. 5 C and D), which shows strikingly low values (Fv/Fm < 0.2) in surface waters through much of the front region. Enhanced silicification by Fe-stressed diatoms thus likely led to increased particle ballasting, as found in another gradient region in the CCE (31).

In the CCE, gravitational flux variability has also been linked to mesozooplankton fecal pellet production (19, 22). Using paired day−night bongo tows and the gut pigment method, we found that grazing was substantially increased on cycle 1 (front) where high sediment trap flux was measured (Fig. S2D). Grazing rates on cycle 1 were also on the high end of measurements made on previous Lagrangian cycles in the CCE (19, 38). High mesozooplankton grazing rates at the front were further supported by measurements made during our transects that showed increased mesozooplankton biomass and grazing in the vicinity of E-Front (Fig. 5 E and F). Thus, diatom Fe stress led to the production of dense phytoplankton in surface waters, and mesozooplankton repackaged these dense, small particles into large, dense, rapidly sinking fecal pellets that largely escaped euphotic zone remineralization.