Biogeochemical Dynamics in Adjacent Mesoscale Eddies of Opposite Polarity

We examined the biogeochemical impact of paired mesoscale cyclones and anticyclones in spatial proximity (<200 km apart) in the North Pacific Subtropical Gyre. While previous studies have demonstrated that upwelling associated with the intensification of cyclonic eddies can supply nutrients supporting plankton productivity, we observed that steeper vertical gradients in inorganic nutrients increased nutrient fluxes due to diapycnal mixing during the mature stage of cyclonic eddies. The increased diapycnal nutrient supply was linked with expansion of eukaryotic phytoplankton biomass and intensification of the deep chlorophyll maximum (DCM) layer. This perturbation in the plankton community was associated with increased fluxes of biominerals (specifically particulate inorganic carbon and particulate silica) and isotopically enriched organic nitrogen in particles exported in the cyclone. The time‐integrated effects of thermocline vertical displacements on the lower euphotic zone were predictable deficits and surpluses of inorganic nutrients and dissolved oxygen, respectively. However, the stoichiometry of oxygen and inorganic nutrients differed from that predicted for production and consumption of phytoplankton biomass, requiring additional biological processes that decouple changes in oxygen and nutrient concentrations. The dynamics revealed by this study may be a common feature of oligotrophic ecosystems, where mesoscale biogeochemical perturbations are buffered by the DCM layer, which limits the ecological impact of eddies in the well‐lit, near‐surface ocean.


10.1029/2021GB007115
3 of 25 on SLA corr ), and assessment of time-varying eddy behavior (e.g., strengthening or weakening in SLA corr ). Eddy coordinates, age, and amplitude were tracked in time using a simple algorithm: For each daily map of SLA corr , the eddy center was defined as the minimum or maximum of SLA corr in a square with a side of 1.5° centered on the coordinates of the eddy center on the previous day. The amplitude of an eddy was defined as the value of SLA corr at the minimum or maximum. The age of each eddy was defined as the number of days since the time of first detection, which was subjectively assessed as the first day when a minimum or maximum detached from a larger area of positive or negative SLA corr . Similarly, the last day of eddy detection was subjectively defined as the day when a minimum or maximum of SLA corr could no longer be distinguished within a larger region with positive or negative SLA corr .
While we adopted a simplified tracking algorithm to describe eddy life history, eddies are also routinely cataloged and tracked within the altimetric Mesoscale Eddy Trajectories Atlas (META2.0) distributed by AVISO+ (Archiving, Validation and Interpretation of Satellite Ocean data; Schlax & Chelton, 2016). The four eddies sampled in this study were all reported in the eddy atlas, but the MESO-SCOPE cyclone was considered as a single feature together with a second cyclonic eddy during part of its lifetime. However, the trajectory of the surface drifters (described in Section 2.5) deployed during MESO-SCOPE showed cyclonic circulation around the SLA corr minimum defined as the eddy center in this study; hence it is likely that the two minima which were merged in the eddy atlas represented distinct features.

Water Sampling and Profiling Instruments
During transit, the hydrographic characteristics were measured using underway conductivity, temperature, and depth (CTD) profilers (underway CTD, Teledyne), while current speed and direction were measured using a hull-mounted acoustic doppler current profiler (ADCP, Workhorse 300 kHz, Teledyne). Upper ocean biogeochemical properties were characterized using a rosette system mounting 10 L Niskin® sampling bottles and profiling instruments including a CTD sensor (Sea-Bird 911 plus), a chlorophyll fluorometer (Seapoint SCF), a polarographic oxygen (O 2 ) sensor (SBE 43,, and a transmissometer (c-star, Sea-Bird). In situ chlorophyll fluorescence was calibrated using a linear regression against the concentration of chloropigments (comprised mostly of monovinyl and divinyl forms of chlorophyll a and b), measured fluorometrically. Similarly, the O 2 sensor was calibrated from a linear regression of dissolved O 2 concentrations determined at discrete depths measured using a Winkler titration system (see below). Beam attenuation measurements that exceeded three standard deviations from the mean (based on 20 m vertical bins) were removed in order to exclude observations presumably due to rare, large particles. Furthermore, to account for variations in sensor background values among deployments, the minimum value measured between 350 and 400 m was subtracted from each beam attenuation profile (this approach assumes that particle scattering is constant in the 350-400 m layer).
Light flux at discrete depths was calculated as the product of the daily-integrated downwelling cosine PAR irradiance above the sea surface measured with a shipboard sensor , and the fraction of downwelling PAR penetrating at depth measured using a free-falling optical profiler (Sea-Bird Hyperpro II). The average depth sampled for all vertical profiles was 161 ± 25 m (± standard deviation). Light penetration below the maximum depth was assessed by extrapolating the exponential decrease of light measured in the deepest 25 m of each vertical profile.

Eddy Transects
Physical and biogeochemical characterizations of each of the eddy couples were obtained by shipboard sampling along an approximately linear transect crossing both eddy centers (Figures 1a and 1b). An initial physical characterization was obtained while transiting using the ADCP and the underway CTD. Biogeochemical observations were obtained at a lower horizontal resolution along transects including five stations in 2016 and eleven stations in 2017, where samples were collected using the CTD rosette sampler. For these stations, hydrography and other profiling sensor observations were measured using the instruments attached to the CTD rosette.
During HL4 (in 2016) there was a failure of the conductivity sensor that prevented an accurate determination of salinity; hence, for this cruise only temperature observations from the underway CTD are presented. Since this problem only impacted the underway instrument, measurements of salinity and potential density associated with the biogeochemical observations were still available from CTD measurements collected when the ship was stationary.

High Vertical Resolution Sampling
In the four eddy centers, the layer surrounding the DCM was characterized using high vertical resolution sampling (approximately 5 m vertical intervals) at 15 discrete depths around the DCM (six depths above the DCM, eight depths below the DCM, in addition to the DCM). The DCM was identified during each CTD deployment as the depth of the fluorescence maximum (ranging from 96 to 151 m). This sampling was consistently done at 0700 local time to account for possible diel variability in biogeochemical properties. The objective of this high vertical Arrows depict the average current velocity and direction measured from the research vessel with an acoustic doppler current profiler in the 11-61 m depth layer; yellow stars depict the sampling stations along a transect where water samples were collected from the ship. Temperature contours along the transects are shown in (d) for HL4, and (e) for MESO-SCOPE. Data from 2016 include shipboard measurements as well as measurements collected with an underway profiling conductivity, temperature, and depth (CTD) system. Data from 2017 include only measurements collected with an underway CTD. resolution sampling was to characterize changes in water chemistry and plankton community composition above, within, and below the DCM.

MESO-SCOPE Lagrangian Sampling
During the 2017 expedition, we extended our observations through Lagrangian sampling of the eddy centers using surface velocity program (SVP, Pacific Gyre) drifters with a drogue centered at 15 m depth. Initially, eddy centers were identified using shipboard ADCP measurements of upper ocean currents and underway hydrographic characterization conducted while transiting across the eddies. Using this information, a SVP drifter was deployed in the predicted center of each eddy (lowest current velocity and deepest or shallowest isopycnal depth). These drifters advanced in arcs consistent with the geostrophic flow. A second SVP drifter was then deployed near the center of the drift arcs to decrease the distance from the eddy center. The research vessel followed these secondary drifters conducting high-temporal resolution Lagrangian sampling over 3-4 days. The distance between the drifter and the ship at the Lagrangian sampling stations averaged 1.1 ± 0.8 (standard deviation) km, with a maximum distance of 4.1 km.

Biogeochemical Sample Collection and Analyses
The analytical methods used for assessing upper ocean biogeochemical properties, including concentrations of dissolved O 2 , chlorophyll, dissolved and particulate nutrients, and fluxes of particulate matter derived from sediment trap collections followed protocols utilized by the HOT program (http://hahana.soest.hawaii.edu/hot/ methods/results.html). Briefly, concentrations of chlorophyll were determined from filtered seawater samples (25 mm diameter glass fiber filters; Whatman GF/F, nominal porosity 0.7 μm). Concentrations of chlorophyll a and phaeopigments were measured fluorometrically on a Turner Designs Model 10-AU fluorometer (Strickland & Parsons, 1972). Note that, in this environment, fluorescence from chlorophyll b is a major contributor to "phaeopigment" determinations (Karl, Bidigare, & Letelier, 2001). Seawater for subsequent determinations of particulate carbon (PC) and nitrogen (PN) was prefiltered through 202 μm Nitex® mesh screen, followed by positive pressure filtration onto precombusted 25 mm diameter glass fiber filters (Whatman GF/F). Filters were frozen at −20°C until analysis in the laboratory. Particulate carbon and PN concentrations were determined using an Exeter Analytical CE−440 CHN elemental analyzer (Exeter Analytical, UK) as described in Grabowski et al. (2019).
Samples for the determination of dissolved O 2 were collected into borosilicate flasks, fixed with manganous chloride and alkaline iodide, acidified, and dissolved O 2 concentrations were determined by titration with thiosulfate (Carpenter, 1965 ) concentrations were collected in acid-washed polyethylene bottles and stored frozen until analyzed using a SEAL Autoanalyzer III Foreman et al., 2016). For most samples where N + N concentrations were below 100 nM, concentrations were determined using a chemiluminescent method (Foreman et al., 2016). Iron (Fe) was sampled using a trace metal clean rosette and analyzed by inductively coupled plasma mass spectrometry after preconcentration by an automated SeaFAST system (Elemental Scientific; see Hawco et al. (2021), for full description).

Calculation of Isopycnal Anomalies and Nutrient Gradients
We calculated isopycnal concentration anomalies for inorganic nutrients and O 2 for both HL4 and MESO-SCOPE cruises to illustrate biogeochemical changes driven by past physical-biological processes in mesoscale eddies. These anomalies represent the concentrations of N + N, PO 4 3− , and O 2 measured along an isopycnal surface minus the cruise-averaged concentrations measured on the same isopycnal. Computing anomalies in this manner assumes similar initial concentrations of nutrients and O 2 for a given isopycnal surface, and that mesoscale physical-biological linkages drive variations in these properties.
Vertical gradients in inorganic nutrient concentrations were calculated as the slope of Model I linear regressions of nutrient concentration versus depth at depths between 10 m above the DCM and 90 m below the DCM from individual nutrient profiles collected during HL4 and MESO-SCOPE. We chose this portion of the water column because the vertical sampling resolution for nutrient concentrations was high enough to enable robust fitting of linear regressions; at shallower depths, N + N concentrations are very low and generally do not demonstrate 10.1029/2021GB007115 6 of 25 vertical structure. The depth of the top of the nutricline was defined as the intercept on the depth axis of the linear regression used to compute N + N vertical gradients, consistent with the approach adopted by Karl, Björkman, et al. (2001).
Isopycnal nutrient anomalies and vertical nutrient gradients were compared to near-monthly Eulerian observations conducted by the HOT program at Station ALOHA. These HOT program observations  were used to examine the relationships between SLA corr and the O 2 and inorganic nutrient anomalies along the 24.5 kg m −3 isopycnal surface (henceforth, we refer to an isopycnal surface based on its potential density anomaly, which is the potential density -1000 kg m −3 ). For these analyses, we eliminated potential seasonal variability in O 2 and inorganic nutrient concentrations by subtracting the monthly mean isopycnal concentrations of these properties. For consistency, when comparing these anomalies with those observed during the HL4 and MESO-SCOPE cruises, we also subtracted the HOT monthly mean concentration from the isopycnal anomalies calculated on the eddy transects.
The calculation of vertical nutrient gradients on HOT observations followed the same procedure described for HL4 and MESO-SCOPE. Vertical profiles where less than three measurements existed in the vertical layer around the DCM were discarded for these analyses.

Plankton Abundance and Biomass
Abundances of picocyanobacteria (Prochlorococcus and Synechococcus) and pigmented eukaryotes (pico-and nanoplankton), together with non-pigmented picoplankton (hereafter, heterotrophic bacteria), were determined by flow cytometry. For these analyses, seawater was collected in 2 mL cryovials, fixed with microscopy-grade paraformaldehyde (0.24% vol/vol final concentration), stored in the dark for 15 min, flash frozen in liquid nitrogen, and stored at −80°C. Plankton were enumerated with an Influx (Cytopeia) flow cytometer, with pigmented cells distinguished based on their scattering and fluorescence characteristics using 2 excitation lasers (457 and 488 nm). The abundance of non-pigmented picoplankton (herein defined as heterotrophic bacteria) was determined by staining cells with SYBR Green I DNA and subtracting the contributions from Prochlorococcus.
The carbon biomass of larger (4-100 μm) eukaryotic phytoplankton was estimated by automated imaging flow cytometry (Imaging FlowCytobot [IFCB], McLane) over the course of the Lagrangian sampling in 2017. In both eddy centers, daily discrete samples were collected every 4 hr (n = 6 at each depth) at 15 m and at the DCM from Niskin ® bottles mounted on the CTD rosette. Analyses were carried out within 2 hr of collection to ensure that cells were not deteriorating since samples were not preserved. The IFCB detects individual particles via laser-induced fluorescence and light scattering used to trigger image acquisition of all particles in successive ∼5 mL samples.
A training set was used to classify the images to the genus level based on morphological traits, as described in Dugenne et al. (2020). The output of the random forest classifier (Sosik & Olson, 2007) was manually corrected to provide accurate estimates of cell size distribution and concentration for individual genera. For this study, we selected phytoplankton genera, then computed carbon biomass based on a biovolume-to-C quota using class-specific regression coefficients (Menden-Deuer & Lessard, 2000). Standard errors associated with the intercept and slope of the biovolume-to-C regressions result in a ±20% uncertainty in larger phytoplankton biomass.

Quantification of Particle Export
During the MESO-SCOPE cruise, 12 free-drifting, surface-tethered, sediment trap arrays were deployed along a line connecting the eddy centers (at approximately 18 km spacing) to examine the downward flux of sinking particulate material. Each surface-tethered array included 12 individual particle interceptor trap collector tubes (Knauer et al., 1979). Traps were deployed at 150 m for 10-13 days, collecting sinking particles. The sediment traps were prepared and processed following HOT program methods . Trap samples were processed and analyzed for PC, PN, particulate phosphorus (PPO 4 3− ), PSi, particulate inorganic carbon (PIC; i.e., calcium carbonate), and stable nitrogen isotope composition in particles (δ 15 N-PN). In addition, sediment trap samples were used to quantify total mass of sinking particles. From the 12 individual collector tubes, 6 were processed for subsequent measurements of PC, PN, PIC, δ 15 N-PN; 3 were processed for PPO 4 3− ; and the remaining 3 traps were split and processed for PSi and total mass flux. Prior to filtration, trap contents were screened through a 335−μm Nitex mesh to remove any mesozooplankton (swimmers) that are not part of the passive flux of particles. Collection tubes for subsequent analyses of PC, PN, and PIC were filtered onto 25 mm diameter combusted glass fiber filters, while samples for PPO 4 3− analyses were filtered onto combusted, acid-washed glass fiber filters. Samples for PSi were filtered onto 47 mm diameter, 0.8 μm pore size polycarbonate filters. All filters were frozen until subsequent analyses. Filters for PC, PN, and δ 15 N-PN were analyzed on a high-temperature elemental analyzer (Carlo Erba NC 2500) coupled to a Finnigan MAT DeltaS mass spectrometer. Filters for total mass were analyzed as described in Grabowski et al. (2019). Filters for PIC determinations were placed in gas-tight vials, fumed with phosphoric acid, and the CO 2 evolved in the headspace quantified by infrared absorption (Grabowski et al., 2019). Particulate silica analyses followed the NaOH digestion method for biogenic silica described in Brzezinski and Nelson (1989). Particulate organic carbon (POC) flux was calculated as the difference between PC and PIC.

Transects Across Adjacent Eddies
The hydrographic and dynamic structure of the sampled eddies was consistent with expectations based on eddy polarity: cyclonic eddies were characterized by an uplifted thermocline and surface currents organized in a counterclockwise circulation, while anticyclones demonstrated deeper thermoclines and clockwise surface current circulation ( Figure 1). In both cruises, surface currents were strongest along the frontal boundaries separating the adjacent eddies. Both of the cyclones and anticyclones sampled as part of this study represented extreme mesoscale events relative to historical Eulerian observations of SLA corr at Station ALOHA ( Figure 2). SLA corr in the cyclone and anticyclone eddy centers averaged −18.4 and 23.8 cm during HL4 and −14.7 and 24.0 cm during MESO-SCOPE, respectively. In both cases, these SLA corr values were more than two standard deviations different from the mean SLA corr measured at Station ALOHA between 1993 and 2018 (Figures 2a and 2b).
The vertical position of the DCM also varied consistently with eddy polarity, with cyclones characterized by shallower DCM, but more intense chlorophyll fluorescence relative to anticyclones (Figures 2c and 2d). For both cruises, daily integrated incident PAR reaching both the DCMs and the 24.5 kg m −3 isopcynal surface was significantly greater in cyclones than in anticyclones (Wilcoxon rank sum test, p = 0.03; Table 1). However, differences in downwelling irradiance at the depth of the DCM were not significantly different between the two eddies sampled during the MESO-SCOPE cruise (Wilcoxon rank sum test p = 0.05).

Isopycnal Anomalies of O 2 and Inorganic Nutrients
Depth-dependent oscillations in the thermocline appear to underlie much of the variability associated with concentrations of inorganic nutrients and dissolved O 2 (Figures 2g and 2h). However, analyses of these properties along isopycnal surfaces revealed more subtle dynamics. For both cruises, robust anomalies for both O 2 and inorganic nutrients were observed along isopycnal surfaces located at depths surrounding the DCM. To assess the generality of mesoscale variability in biogeochemistry along isopycnal surfaces located near the depth of the DCM, we evaluated relationships between SLA corr and both O 2 and inorganic nutrient anomalies along the 24.5 kg m −3 surface using historical HOT program measurements ( Figure 4). These analyses revealed consistent patterns in both the HOT program observations and the eddy transects. In particular, the seasonally detrended O 2 concentration anomalies decreased with increasing SLA corr , while the de-seasoned concentration anomalies of both N + N and PO 4 3− increased with SLA corr (Figure 4). The consistency in concentrations of O 2 and inorganic nutrients to variations in SLA corr along this isopycnal surface suggests the patterns we observed are not specific to strong mesoscale eddies, but rather are more generally reflective of the full range of mesoscale motions observed in the region near Station ALOHA. Slopes of the isopycnal concentration anomalies versus SLA corr (Model II geometric mean regression) were −1.07 ± 0.07 mmol O 2 m −3 cm −1 , 0.102 ± 0.008 mmol N + N m −3 cm −1 and 0.00860 ± 0.00067 mmol PO 4 3− m −3 cm −1 . The resulting ratios of these slopes were −10.49 (mol O 2 : mol N), −124.4 (mol O 2 : mol P), and 11.86 (mol N: mol P). SLA corr explained less of the variability in the inorganic nutrient concentration anomalies than O 2 concentration anomalies (correlation coefficients values for O 2 , N + N, and PO 4 3− were −0.49, 0.21, and 0.23, respectively, all with p < 0.001), and departure from a linear relationship are particularly evident for N + N at low values of SLA corr (i.e., when SLA corr decreases below −10 cm, such as would occur with relatively strong cyclonic eddies). Potential Fe limitation of DCM plankton production was also investigated during the MESO-SCOPE cruise and results from that study are described elsewhere (Hawco et al., 2021). Briefly, dissolved Fe was depleted in the lower euphotic zone of the cyclone, averaging 73 nmol m −3 at a potential density of ∼25.16 kg m −3 , the same isopycnal where we measured the maximum cyclonic deficit of N + N and PO 4 3− . In this layer, the cyclone had 110 nmol m −3 less Fe than the anticyclone as Note. The depths of the DCM and of the isopycnal surface (m) are daily averages ± standard deviations from shipboard conductivity, temperature, and depth measurements. The isopycnal anomalies of O 2 and inorganic nutrients vary inversely due to primary production and mineralization, which produce O 2 and consume inorganic nutrients (or vice versa) in proportions that are relatively invariant across ocean basin scales (Anderson & Sarmiento, 1994;Redfield, 1958). To assess if the isopycnal variations of O 2 and inorganic nutrients in mesoscale eddies were consistent with the stoichiometry of the production and consumption of organic matter, we calculated the excess O 2 anomaly as follows: isopycnal anomaly of (O2 − O2 solubility) + 10.5 × isopycnal anomaly of (NO3 − + NO2 − ) where 10.5 represents the moles of O 2 produced when 1 mol of NO 3 − + NO 2 − is consumed for the synthesis of new organic matter, or the inverse for its consumption (Anderson, 1995;Johnson et al., 2010). For this analysis, we also subtracted O 2 solubility to avoid misinterpreting variations caused by changes in hydrography along the eddy transects. The excess O 2 anomaly was positive in both cyclone centers in a vertical layer located just above the DCM and approximately between 50 and 100 m (Figure 5a). In this layer, the excess O 2 anomaly reached values above 10 mmol m −3 in cyclones, while it was close to 0 or negative in both anticyclones ( Figure 5). The excess O 2 anomaly was predominantly negative below 100 m in all four eddy centers ( Figure 5), indicating that the O 2 concentration in these eddy centers was lower than expected assuming the water started with the average isopycnal concentrations measured along the transect and was later modified assuming fixed 10.5 O 2 :N (mol:mol) stoichiometry (Anderson, 1995).

Vertical Gradients and Mixing Estimates at the Top of the Nutricline
We calculated how mesoscale motions influenced vertical gradients in N + N and PO 4 3− based on the slopes of nutrient concentration versus depth relationships calculated from the individual nutrient profiles. Upward displacement of the nutricline associated with cyclonic eddies resulted in vertical gradients in N + N that were 1.8-9.2 times greater near the center of the cyclones than in the anticyclones (Figures 6a and 6b). Similarly, vertical gradients in PO 4 3− were 1.8-10.8 times greater near the center of the cyclones than in the anticyclones (Figures 6c and 6d). Changes in the gradients of N + N and PO 4 3− between cyclones and anticyclones were even larger when derived based on nutrient concentrations measured from high-resolution vertical sampling at depths around the DCM.
We also examined relationships between the vertical nutrient gradients and SLA corr based on the historical HOT program observations (Figures 6e  and 6f). These analyses revealed patterns consistent with the more limited observations during HL4 and MESO-SCOPE. Specifically, vertical gradients in concentrations of N + N and PO 4 3− decreased with increases in SLA corr (Figures 6e and 6f).

Mesoscale Variability in Plankton Biomass and Community Composition
Several consistent patterns in the near-surface waters and at the DCM emerged based on various proxy measurements of plankton biomass at the centers of both pairs of cyclones and anticyclones. In the near-surface waters, there were no apparent eddy-specific differences in concentrations of PC or beam attenuation, nor in chlorophyll a concentrations or fluorescence (Figure 7). In contrast, at the depth of the DCM, all of these proxy measurements of plankton biomass were elevated in the cyclonic eddies relative to the anticyclones. The concentration of PC was greater in the DCM of the cyclones relative to anticyclones by 1.3-and 1.4-fold, in the two expeditions (Figure 7a). Similarly, the concentration of chlorophyll a was greater in the DCM of the cyclones relative to anticyclones by1.2-and 1.3-fold (Figure 7c). This difference was even more pronounced for in situ chlorophyll fluorescence (Figure 7d), which includes contributions from pigments other than chlorophyll a and can be influenced by variation in fluorescence yield (Roesler et al., 2017). Fluorescence in the DCM was 1.3-and 1.8-fold greater in the cyclonic eddies than in the anticyclones (Figure 7d). Similarly, beam attenuation, an optical proxy for particle concentrations, was 1.4-and 1.5-fold greater at the depth of DCM in the cyclonic eddies relative to the anticyclones (Figure 7b) suggesting enhancement of plankton biomass rather than photoacclimation as drivers of these signals. We also evaluated the abundances of Prochlorococcus, Synechococcus, eukaryotic phytoplankton, and heterotrophic bacteria in the adjacent eddy pairs based on flow cytometric analyses. We found that Prochlorococcus abundances were greater in the waters above the DCM in both anticyclones, relative to the cyclones, and relative to the HOT program climatology (Figures 8a and 8e). In contrast, eukaryotic phytoplankton (restricted to pico-and nanoplankton) abundance was elevated in a broad layer near the DCM. Maximum abundance in this layer exceeded the 95th percentile of HOT observations, with abundances in these eddies >4,000 cells mL −1 (Figures 8c and 8g). Eukaryotic phytoplankton abundances in the near-surface waters were similar among the sampled eddies. Vertical distributions of Synechococcus were similar among the sampled eddies and similar to the HOT program observations of this genus (Figures 8b and 8f). Heterotrophic bacteria were more abundant in anticyclones than in cyclones in the near-surface and in waters below the DCM, but this pattern was reversed at depths just above the DCM.
During the MESO-SCOPE cruise, we also examined eddy-driven variability in the biomass of larger eukaryotic phytoplankton cells (4-100 μm in equivalent spherical diameter) imaged with an IFCB. In the near-surface waters, biomass estimates of most of the enumerated phytoplankton were similar between the eddies, with the exception of significantly greater biomass of Prymnesiophyceae in the anticyclone (Figure 9, Wilcoxon rank sum test, p < 0.05). Conversely, at the depth of the DCM, most phytoplankton classes demonstrated greater biomass in the cyclone than in the anticyclone (Figure 9). The dominant classes in the DCM of the cyclone were Prymnesiophyceae and Bacillariophyceae (specifically members of diatoms), whose biomass was 2.3 and 4.7 times greater than in the anticyclone, respectively (Figure 9b). Prymnesiophyceae genera, whose biomass was significantly larger in the DCM of the cyclone compared to the anticyclone (Wilcoxon rank sum test, p < 005), included Phaeocystis, Ophiaster, Umbellosphaera, Acanthoica, Gephyrocapsa, and Syracosphaera (in descending order of average biomass). The Bacillaryophceae genera enriched in the DCM of the cyclone were Pseudo-nitzschia, Chaetoceros, Cylindrotheca, and Navicula. Members of the Prymnesiophyceae and Bacillariophyceae are important controls on the formation of calcium carbonate and opal, respectively.

Mesoscale Variability in Particle Export
Drift tracks of the sediment trap arrays were consistent with geostrophic circulation of the two eddies sampled during MESO-SCOPE (Figures 10a and 10b). The resulting particle fluxes measured across the adjacent eddies revealed clear mesoscale structure in both the magnitude and stoichiometry of exported particulate matter. Most notably, despite similarities in POC, PN, and PPO 4 3− fluxes across the eddies, total mass, PSi, and PIC export were approximately 2−, 4−, and 2.9−fold greater, respectively, in the cyclone relative to the anticyclone (Figures 10c and 10d; Supporting Information Table S1). Moreover, total mass, PSi, and PIC fluxes were all inversely related to SLA corr (Figures 10c and 10d; Supporting Information Table S1). In contrast, fluxes of POC, PN, and  Table S1; Model II least-squares linear regressions, p > 0.05). In the cyclone, export of PIC accounted for 17% of the PC flux, with PIC comprising a lower fraction (∼10%) of PC flux in the anticyclone ( Table 2). The proportions of PIC and PSi to mass flux were inversely related to spatial changes in SLA corr , with PIC:mass flux ratios (g CaCO 3 :g mass) decreasing from 38% to 28%, and PSi: Mass flux ratios (g SiO 2 : g mass) varying from 5.8% to 2.1% across the cyclone to anticyclone transect ( Table 2). Because of the proportionally greater contribution of PIC and PSi to mass flux in the cyclonic eddy, POC to mass flux (g C: g mass) was positively related to SLA corr , with POC representing upwards of 34% of mass flux in the anticyclone and only ∼22% in the cyclone (Table 2). In addition to evaluating mesoscale spatial variation in fluxes of key bioelements, we also quantified changes in δ 15 N-PN. The δ 15 N-PN of exported particles was inversely related to changes in SLA corr across the dipole, with the cyclone exporting PN with a δ 15 N signature as great as ∼4.3‰, decreasing to as low as 3.0‰ in the anticyclone (Table 2). , defined as dN/dz and dP/dz, respectively, across adjacent eddies of opposite polarity for (a and c) HL4 and (b and d) MESO-SCOPE cruises. Panels e and f depict vertical gradients in N + N and PO 4 3− concentrations, respectively, versus SLA corr for the two eddy transects (blue and yellow symbols) and the Hawaii Ocean Time-series (HOT) program observations (gray symbols). Yellow squares in (a-d) and blue or yellow squares in (e-f) depict gradients estimated from the high vertical resolution profiles in the four eddy centers. Red lines in (a-d) and error bars in (e-f) depict ±1 standard deviation in the slopes. Red lines in (e-f) are linear regressions on HOT program nutrient gradients (Model II) with slopes of −0.0023 ± 0.0002 mmol m −4 cm −1 (r = 0.36, p = 5 × 10 −9 ) and −0.00018 ± 0.00001 mmol m −4 cm −1 (r = 0.35, p = 1 × 10 −8 ) for N + N and PO 4 3− , respectively. Estimates of vertical nutrient gradients were obtained by calculating the slope of the Model I linear regression of nutrient concentration versus depth in the vertical layer between 10 m above and 90 below the depth of the deep chlorophyll maximum.
The elevated contributions of PSi and PIC to particle export in the cyclonic eddy resulted in significant spatial differences in the elemental stoichiometry of particulate matter export across the mesoscale field. Elevated PSi flux in the cyclone coincided with decreasing POC:PSi, PN:PSi, and PPO 4 3− :PSi ratios (mol:mol) across the eddies (Table 2). Moreover, all three ratios were positively correlated to changes in SLA corr . Elevated fluxes of PIC within the cyclonic eddy resulted in decreasing PC:PIC, PN:PIC, and PPO 4 3− :PIC ratios across the cyclone to anticyclone transect, with all three ratios inversely related to spatial changes in SLA corr (Table 2). In contrast,  POC:PPO 4 3− and PN:PPO 4 3− ratios (mol:mol) did not vary with SLA corr , averaging 173 (±20) and 28 (±2.0), respectively. Intriguingly, although neither POC nor PN fluxes varied significantly with SLA corr , the ratio of POC:PN was inversely and significantly related to SLA corr (Table 2).
Once again, we leveraged HOT program observations to evaluate the generality of the mesoscale variability of particle fluxes across eddies. In this instance, we assessed the significance and strength of the Pearson correlations between HOT particle fluxes and SLA corr . Consistent with our sampling during MESO-SCOPE, we find a lack of correlation between SLA corr and sinking fluxes of PC, PN, and PPO 4 3− , but significant negative correlations in fluxes of both PSi and PIC and SLA corr (in the Supporting Information Table S1; p < 0.05). While the negative correlations of PSi and PIC are consistent with observations during MESO-SCOPE, we also found a significant positive correlation for HOT observations between δ 15 N and SLA corr (correlation coefficient of 0.14; p < 0.05), which was opposite in sign of the relationship observed based on the MESO-SCOPE sediment trap collections.

Mesoscale Impacts on the Nutricline and Lower Euphotic Zone
On two separate research cruises in the subtropical North Pacific Ocean, we sampled pairs of adjacent mesoscale eddies of opposite polarity in late spring and summer, providing an opportunity for comparative analyses of upper ocean biogeochemistry and plankton ecology associated with mesoscale features in close spatial proximity (<200 km). By coupling our sampling of adjacent eddies with a multi-decadal analysis of HOT program observations, we find robust differences between cyclones and anticyclones in concentrations of chlorophyll a, dissolved O 2 , inorganic nutrients, and particle load in the lower euphotic zone. However, differences in these properties were not apparent in the well-lit upper ocean waters.
The consistent biogeochemical responses to mesoscale motions observed in the lower euphotic zone reflect rapid plankton assimilation of nutrients as the nutricline undergoes mesoscale-driven oscillations across the exponentially varying light field. This variability in the lower euphotic zone is consistent with previous findings on the change in the vertical distributions of nutrients and biomass related to mesoscale activity at Station ALOHA (Barone et al., 2019;Church et al., 2009). The eddies sampled as part of this study bracketed the full dynamic   range of SLA corr variability observed at Station ALOHA and revealed that mesoscale motions can vertically perturb isopycnals in the lower euphotic zone by more than 100 m. These large vertical excursions alter the ambient light field on an isopycnal by more than an order of magnitude. Furthermore, we observed that light fluxes to the DCM were ∼2-fold greater in cyclones than in anticyclones, linking increases in available energy to increased diapycnal fluxes of nutrients. Such variations in light intensity play important roles in regulating variability in phytoplankton production in the light-limited DCM (Goldman, 1993;Goldman & McGillicuddy, 2003), with concomitant impacts on concentrations of nutrients and dissolved O 2 , a finding consistent with our analyses of historical HOT program observations.
In the current study, phytoplankton biomass remained elevated in the DCM of cyclonic eddies several weeks after the initial intensification phase (see eddy life cycles in Supplementary text S1). Based on these findings, we devised a simplified conceptual framework to describe temporal dynamics associated with cyclonic uplift of the nutricline and subsequent biological responses in the stratified ocean ( Figure 11). Prior to the intensification of the cyclonic eddy, the biomass maximum of eukaryotic phytoplankton coincides with the DCM, at a depth determined by the downward light flux (Figure 11a; Letelier et al., 2004). Mesoscale uplift results in upward displacement of the DCM community and the nutricline, increasing light to otherwise light-limited phytoplankton (Figure 11b). In response to this uplift, a transient phytoplankton community utilizes the additional light energy, increasing biomass and consuming the available nutrients. Net photosynthetic production by this community fuels export of organic matter from the euphotic zone resulting in negative nutrient and positive O 2 anomalies in the waters above the DCM constituting a "chemical wake" along the displaced isopycnal surfaces (Figure 11c). An important consequence of the resulting nutrient removal is a steepening of the nutricline and presumed intensification of diapycnal nutrient fluxes, sustaining elevated eukaryotic phytoplankton biomass within the DCM (Figure 11c). Our observations indicate that, following the transient phase of nutricline erosion, light flux reaching the DCM of a cyclone can be greater than the flux reaching the DCM of an anticyclone. This finding is consistent with the ecological dynamics of the DCM layer described by Cullen (2015), whereby an increase in diapycnal nutrient flux requires a greater supply of energy (i.e., light intensity) to be consumed. Increased light accelerates the growth of phytoplankton cells in the cyclonic DCM resulting in more efficient consumption of the diapycnal nutrient flux. In our study, we did not sample the cyclonic eddies in their intensification phase, so we do not have observations to describe the phytoplankton community responsible for the production of the chemical wake. However, the introduction of nutrient-enriched waters to the euphotic zone has been shown to promote growth of larger eukaryotic phytoplankton, which can export organic matter efficiently to the mesopelagic waters (McAndrew et al., 2007). This community would be expected to dominate phytoplankton biomass in the DCM during the transition between panels b and c in Figure 11 of our conceptual framework. As nutrients are removed, phytoplankton above the DCM, in the "chemical wake," would be increasingly dominated by smaller cells as picoplanktonic cyanobacteria (e.g., Prochlorococcus), which are more typical of oligotrophic environments and whose growth and removal are more tightly balanced (McAndrew et al., 2007).
Our framework highlights understudied biogeochemical dynamics associated with the mature phase of cyclonic eddies. In contrast to previous studies Sweeney et al., 2003), the novelty of our model is the focus on nutrient delivery to the lower euphotic zone attributable to diapycnal mixing in the mature stage of a cyclone rather than nutrient upwelling. Making the simplified assumption that eddy diffusivity does not vary between eddies of opposite polarity, diapycnal fluxes attributable to vertical nutrient gradients would be 9-fold and 11-fold greater for N + N and PO 4 3− , respectively, in the MESO-SCOPE cyclone relative to the anticyclone. Analyses of the results from a four-dimensional numerical model from Station ALOHA lend additional support to the idea that, during the mature phase of cyclonic eddies, erosion of the nutricline drives increased diapycnal nutrient fluxes supporting biomass production by eukaryotic phytoplankton near the DCM (Supplementary Text 2 and Supplementary Figure 3).
Our analyses also highlight that mesoscale dynamics can vertically decouple O 2 production from nutrient consumption. For example, during the MESO-SCOPE cruise, the 24.5 kg m −3 isopycnal was displaced upwards to a depth of 64 m in the center of the cyclonic eddy, which was 65 m shallower than the position of this isopycnal in the anticyclone. During the initial phase of eddy intensification, upward displacement of this isopycnal and associated increase of light in the cyclone would presumably stimulate net community production and subsequent export of this photosynthetically-derived organic matter. We see evidence of this as an excess of 23 mmol O 2 m −3 on the 24.5 kg m −3 isopycnal in the center of the cyclone relative to the average concentration of O 2 measured during the eddy survey on that isopycnal. Similarly, inorganic nutrient consumption associated with this cyclonic uplift resulted in isopycnal concentration anomalies of N + N and PO 4 3− of −0.21 mmol N m −3 and −0.043 mmol P m −3 , respectively. The resulting O 2 :N and O 2 :P molar stoichiometries (110:1 and 535:1, respectively) suggested excess net O 2 production with respect to the net consumption of inorganic nutrients. This phenomenon was widespread in both cyclone centers, where it was observed in a depth layer above the DCM approximately between 50 and 100 m. Conversely, below 100 m, all eddy centers were characterized by a deficit in O 2 with respect to the concentration expected based on inorganic nutrient concentrations. The observation of similar variability in eddies of opposite polarity is intriguing considering the isopycnal anomalies must sum to zero across each eddy transect.
Previous studies, including several based at Station ALOHA, have reported anomalous O 2 to nutrient stoichiometries similar to those described in this study. Emerson and Hayward (1995) found that O 2 :N stoichiometries deviated substantially from those expected based on Redfield mineralization of organic matter in the lower euphotic zone and upper mesopelagic waters. Leveraging autonomous profiling floats equipped with O 2 and NO 3 − sensors, Johnson et al. (2010) documented consistent supersaturation of O 2 in the well-lit euphotic zone waters and an apparent deficit of NO 3 − (relative to respiratory O 2 consumption) at depth. These authors found that the "overproduction" of O 2 relative to NO 3 − consumed in the upper euphotic zone could be balanced by NO 3 − consumption without corresponding O 2 production at depth (Johnson et al., 2010). A subsequent analysis of profiling float O 2 and NO 3 − data from the subtropical North Pacific confirmed anomalous O 2 and NO 3 − stoichiometry within discrete "regimes" that differentiate based on the vertical position of isopycnal surfaces (Ascani et al., 2013).
Mechanisms proposed to explain the anomalous ratios in the concentrations of O 2 and inorganic nutrients include remineralization of C-enriched dissolved organic matter (Abell et al., 2005;Emerson & Hayward, 1995;Fawcett et al., 2018), heterotrophic bacterial assimilation of NO 3 − (Fawcett et al., 2018), phytoplankton vertical migration (Letscher & Villareal, 2018), and nitrous oxide production accompanying N oxidation (e.g., ammonia or NO 2 − oxidation; Wilson et al., 2014). Our eddy-centric observations suggest different mechanisms may produce the observed anomalies in anticyclones and cyclones. In anticyclones, the subduction of near-surface water could provide high concentrations of C-rich organic matter to the lower euphotic zone, which would be later mineralized and result in lower O 2 concentrations relative to concentrations of inorganic nutrients that would be produced if the organic matter were mineralized in typical (i.e., Redfield) stoichiometry. Conversely, vertically-migrating phytoplankton might consume nutrients near the top of the nutricline and actively move upward to photosynthesize (Villareal et al., 1993), which could explain both the O 2 deficit below 100 m (where nutrients are taken up without O 2 production) and the O 2 surplus in the 50-100 m layer in the cyclonic eddies (where O 2 is produced without nutrient uptake).

Mesoscale Variability in Particle Fluxes
By deploying a suite of 12 individual sediment trap arrays across two adjacent eddies of opposite polarities, we obtained spatially-resolved information on particle export across a large mesoscale gradient (SLA corr for these deployments ranged between −12 and 25 cm). Although export of organic matter (POC, PN, and PPO 4 ) was similar across and within this mesoscale gradient, total mass flux varied more than 2-fold, coincident with approximately 4-fold changes in export of PSi and PIC. Export of total mass, PSi, and PIC were all greatest in the cyclone center (lowest SLA corr ) and decreased linearly with increases in SLA corr . Moreover, the δ 15 N signature of sinking PN points to different sources of N supporting export across these eddies. While these observations were collected in strong mesoscale eddies, they depict several patterns of particle flux variability broadly linked with different classes of mesoscale motions, as demonstrated by the analysis of historical observations from the HOT program. Consistent with measurements in MESO-SCOPE, the HOT particle fluxes of PC, PN, and PPO 4 were not significantly correlated with SLA corr , a finding previously reported by Barone et al. (2019), while a negative correlation was observed for PIC and PSi. Conversely, the increased δ 15 PN in the MESO-SCOPE cyclone is in the opposite direction of the HOT positive correlation of δ 15 PN and SLA corr .
We believe that the mesoscale change in particle export observed in our study is not directly linked to upwelling of nutrient-rich water, but rather reflects export linked to increased diapycnal nutrient fluxes. Two pieces of evidence support his view: (a) The correlations of particle fluxes with SLA corr point to a connection with the depth of the thermocline rather than with the vertical velocities that would be associated with eddy intensification or weakening; and (b) the cyclone sampled during MESO-SCOPE had its nearest intensification event 32 days before our sampling (in the Supporting Information S1). As a consequence, we attribute mesoscale variation in export detected in our measurements to the shallower and steeper nutricline associated with low SLA corr , increasing nutrient supply to the lower euphotic zone and supporting abundant eukaryotic phytoplankton, including diatoms and prymnesiophytes. Subsequent export of these phytoplankton directly impacts the export of biomineralized particles. In contrast, the deeper nutricline in anticyclonic eddies appears to favor the proliferation of small picoplanktonic cyanobacteria (i.e., Prochlorococcus; Barone et al., 2019). The MESO-SCOPE anticyclonic eddy also contained nearly twice the concentration of dissolved iron than is typically found at the DCM at Station ALOHA, which may have supported the increase in Prochlorococcus abundance near the DCM (Hawco et al., 2021).
One of the most striking observations from the MESO-SCOPE sediment trap deployments were the robust inverse relationships between SLA corr and PSi, PIC, and total mass flux. Enhanced PSi export has been previously reported for a wind-driven cyclone south of the Hawaiian archipelago, for a mode-water eddy in the Sargasso Sea, and for cyclones traversing Station ALOHA (Benitez-Nelson et al., 2007;Buesseler et al., 2008;Rii et al., 2008;Zhou et al., 2021); however, our observations provide evidence that cyclonic eddies may also promote PIC export. To our knowledge, these are the first observations linking export of PIC to mesoscale dynamics. The enhanced fluxes of PSi and PIC observed during MESO-SCOPE may provide some insights into the phytoplankton fueling particle export in the cyclone, specifically diatoms and prymnesiophytes, whose biomass was elevated in the DCM of the cyclone relative to the DCM of the anticyclone. Diatoms and prymnesiophytes are key phytoplankton taxa contributing to pools of PSi and PIC, respectively, but various zooplankton, including members of the Rhizaria, also have mineral skeletons composed of calcium carbonate (e.g., Foraminifera) or opal (e.g., Radiolaria). We could not quantify the abundance of large Rhizaria (>100 μm in diameter) with the methods employed in this study, so we cannot quantify contributions from these organisms to the observed variations in the export of PSi and PIC between eddies of different polarity.
A number of previous studies have identified elevated diatom and prynmnesiophyte biomass in the DCM of cyclonic eddies and mode water eddies (Benitez-Nelson et al., 2007;Bidigare et al., 2003;Buesseler et al., 2008;McGillicuddy et al., 2007;Rii et al., 2008). In a microscopy-based assessment of diatoms, Scharek et al. (1999) found low contributions to sinking particle export by members of the DCM diatom assemblage, but they noted how a substantial fraction of the diatom frustules were empty (lacking cytoplasm). Similarly, Benitez-Nelson et al. (2007) concluded that coupling between phytoplankton growth and microzooplankton grazing in Hawaiian lee cyclones resulted in export of empty diatom frustules, thereby increasing the PSi: POC ratio of the trap-derived material. Based on pigment and gene-based analyses, Li et al. (2013) concluded that diatoms and prymnesiophytes comprised approximately 2% and 20%, respectively, of total chlorophyll a in the lower euphotic zone at Station ALOHA. However, the proportion of diatom and prymnesiophyte genes exported, relative to the euphotic zone inventory, averaged 0.15% d −1 and 0.19% d −1 , compared to the 5% d −1 PSi export: inventory (Li et al., 2013). Based on high vertical resolution sampling of 234 Th deficits in a mature mode water eddy in the Sargasso Sea, Buesseler et al. (2008) concluded that particle export and remineralization occur in vertically distinct regions of the upper ocean, with the waters immediately below the DCM appearing to support active particle decomposition. These results appear consistent with the interpretation that a fraction of the biomass produced by diatoms and prymnesiphytes near the DCM is mineralized before reaching the sediment traps leaving empty frustules and calcified liths, which underlie the elevated PSi and PIC fluxes in the MESO-SCOPE cyclone.
In addition to clear relationships between mass, PSi, and PIC fluxes, the δ 15 N of the exported PN demonstrated a significant inverse relationship to SLA corr during MESO-SCOPE (i.e., the δ 15 N of PN was more enriched in the cyclone relative to the anticyclone). A number of previous studies have utilized a two-end member isotope model, leveraging differences in the δ 15 N signatures of deep water NO 3 − and N 2 , to estimate the proportion of PN export supported by NO 3 − assimilation and N 2 fixation at Station ALOHA (Böttjer et al., 2016;Casciotti et al., 2008;Dore et al., 2002;Karl et al., 1997). Assuming only two potential sources of N (NO 3 − assimilation and N 2 fixation) support PN export, these observations suggest that NO 3 − assimilation was relatively more important in the cyclone (where NO 3 − appeared to support upwards of 85% of PN export) while N 2 fixation was relatively more important in the anticyclone (where N 2 fixation was estimated to support upwards of 40% of PN export). However, observations from the HOT program suggest this is not a generalized pattern for these waters: despite increased steepness in the nutricline with decreases in SLA corr , the HOT observations indicate that the δ 15 N signature of sinking particles was positively related to variations in SLA corr , opposite in sign to the variations observed during MESO-SCOPE.
To compare the magnitude of particle export to the supply of nutrients from turbulent mixing, we computed diapycnal nutrient fluxes across the top of the nutricline for the MESO-SCOPE eddies. Doing so required assuming a value for eddy diffusivity. For this purpose, we used 1.1 × 10 −5 m 2 s −1 , a value determined based on a tracer release experiment conducted in the pycnocline of the North Atlantic (Ledwell et al., 1993), which is similar to the diffusivity of 10 −5 m 2 s −1 that Klymak et al. (2006) estimated from the rate of dissipation of kinetic energy in the open ocean far from the Hawaiian Ridge. A tracer release experiment in the North Atlantic targeted a layer near the DCM to estimate eddy diffusivity in relation to nutrient fluxes (Ledwell et al., 2008), deriving a value for eddy diffusivity of 3.5 × 10 −5 m 2 s −1 . However, the authors of this study acknowledge this value is likely elevated due to the passage of a series of storms through the sampling region (Ledwell et al., 2008); hence, we considered this diffusivity too high to be representative of the eddies sampled during the MESO-SCOPE cruise.
By assuming a diapycnal eddy diffusivity of 1.1 × 10 −5 m 2 s −1 , we estimate diapycnal fluxes in the MESO-SCOPE cyclone were 0.08 mmol N m −2 d −1 and 4.5 μmol P m −2 d −1 , equivalent to 20% and 34% of the measured PN and PPO 4 3− export fluxes, respectively. If this flux was consumed in a 20 m layer around the DCM, it would account for 10%-15% of the 14 C-based primary production (Hawco et al., 2021). In comparison, diapycnal fluxes in the anticyclone were 0.009 mmol N m −2 d −1 , and 0.5 μmol P m −2 d −1 , equivalent to 2% and 3% of the measured PN or PPO 4 3− export fluxes. While the magnitude of these fluxes shows that additional sources of nitrogen and phosphorus fuel the export of particles from the euphotic zone, it remains unclear why changes in diapycnal nutrient fluxes do not appear to drive changes in the export of particulate organic matter. One possible explanation is that the mesoscale variability of other unidentified sources of nitrogen and particularly phosphorus counterbalances changes in the supply of these elements through diapycnal mixing. Another possible explanation is that the new production sustained via the implied diapycnal nutrient fluxes is exported to depths deeper than our sediment trap collections through vertically migrating zooplankton. A prior study at Station ALOHA estimated that vertical fluxes mediated by diel vertical migrators were 38% and 79% of the sediment trap fluxes of N and P, respectively (Hannides et al., 2009). This mechanism could be more pronounced in cyclones, where the higher biomass of relatively large phytoplankton could be preferentially grazed by larger zooplankton species capable of diel vertical migration.
Another process that likely varied between the eddies was N 2 fixation. Abundances of Crocosphaera, a diazotrophic cyanobacterium, were elevated in the MESO-SCOPE anticyclone (Dugenne et al., 2020). Previous studies have documented increases in N 2 fixation and N 2 -fixing microorganism biomass associated with anticyclonic eddies (Davis & McGillicuddy, 2006;Fong et al., 2008;Wilson et al., 2017). Moreover, during the warm summer months (June-October), rates of N 2 fixation at Station ALOHA often peak during periods of positive SLA corr such as those associated with anticyclones (Böttjer et al., 2016;Church et al., 2009). The gradient in δ 15 N of PN export observed across the mesoscale eddies sampled in the current study further reinforce the apparent promotion of N 2 fixation in anticyclones during the summer in this region.
Despite the lack of variability in the export of particulate organic matter across eddies of opposite polarity, the isopycnal anomalies of O 2 and nutrients observed in the cyclone reflect the prior history of enhanced export associated with isopycnal uplift. During HL4 and MESO-SCOPE, 50-200 m depth integrated O 2 concentration anomalies in the cyclone centers were 2152 mmol O 2 m −2 and 772 mmol O 2 m −2 (taking into account changes in solubility), respectively, suggesting excess O 2 produced in the lower euphotic zone was not accompanied by concomitant consumption of O 2 via respiration of organic matter at the top of the mesopelagic zone. Moreover, suspended PC inventories during HL4 and MESO-SCOPE appear too low to account for the formation of O 2 isopycnal anomalies. We did not measure dissolved organic carbon, but concentrations of dissolved organic N and P were lower in the cyclone than in the anticyclones (data not shown). Such findings suggest that the organic matter produced during the formation of the O 2 and nutrient anomalies had already been exported to deeper waters. As a consequence, the biogeochemical signatures associated with nutrient consumption and O 2 production remained in the euphotic zone as a "chemical wake." Taken together, the isopycnal anomalies and sediment trap fluxes indicate that the MESO-SCOPE cruise likely sampled the cyclonic eddy at some point after the peak period of organic matter export. However, the enhanced diapycnal nutrient fluxes during the mature stage of the eddy kept sustaining a community of diatoms and prymnesiophytes in the DCM layer. Some of these cells underwent mineralization in the upper 150 m, exporting empty frustules and calcified liths that enhanced PSi and PIC fluxes.

Conclusions
By sampling adjacent eddies of opposite polarity we identified several impacts of mesoscale motions on plankton biology and upper ocean biogeochemistry in the subtropical North Pacific Ocean. Some key findings that emerged from our study are: 1. Phytoplankton growth following isopycnal uplift in cyclones steepened the vertical gradients in inorganic nutrients through the biological erosion of the top of the nutricline. This presumably increased the diapycnal flux of nutrients delivered to the lower euphotic zone and supported increased eukaryotic phytoplankton biomass during the mature phase of the eddies. Conversely, depression of isopycnals in anticyclones decreased nutrient concentration gradients in the lower euphotic zone, a dynamic that appeared to increase Prochlorococcus biomass above the DCM. 2. Mesoscale changes in inorganic nutrients and dissolved O 2 concentrations are vertically decoupled. For example, upward displacement of isopycnals associated with cyclonic eddies resulted in peak consumption of inorganic nutrients occurring deeper than the layer where O 2 production was maximal. In contrast, both cyclones and anticyclones appeared to have less O 2 than expected based on nutrient concentrations in the layer below the DCM. 3. Particle-associated biomineral fluxes of PSi and PIC were inversely related to SLA corr despite no differences in export of POC, PN, and PPO 4 3− . Elevated biomineral export at low SLA corr presumably reflects increased contributions from silicifying and calcifying plankton consistent with observations of elevated biomass of Bacillariophyceae and Prymnesiophytes in the cyclone DCM. Moreover, the δ 15 N of sinking particles revealed differences in N sources supporting production across adjacent eddies, with a larger contribution by N 2 fixation in the anticyclone.
Our study highlights the spatiotemporal complexity of ocean ecosystems, and the inherent challenges such complexity presents to Eulerian observations such as the HOT program. Across a region of the ocean stretching a few hundreds of kilometers, the presence of strong, mature, mesoscale eddies of opposite polarity resulted in striking differences in ocean biogeochemistry. We found eddy-dependent isopycnal displacements through the upper ocean light field alter the steepness of the nutricline, influence the biomass of eukaryotic phytoplankton (including members of the prymnesiophytes and diatoms) and Prochlorococcus, and impact the export of opal and carbonate. These observations expand our understanding of the biogeochemical impacts of mesoscale motions by describing the stoichiometry of the erosion of the nutricline in cyclonic eddies and by highlighting a series of cascading dynamics that result from this erosion.