Rapid photoacclimation during episodic deep mixing augments the biological carbon pump.

Episodic deep mixing events are one component of the biological carbon pump that physically transports organic carbon into the mesopelagic. Episodic deep mixing also disrupts summertime thermal stratification thereby changing the light field and nutrient concentrations available for phytoplankton growth. Phytoplankton survival and growth below the mixed layer following restratification depends on how rapidly cells can employ a variety of photoacclimation processes in response to the environmental changes. To compare the relative timescales of summertime episodic deep mixing events with the timescales of phytoplankton photoacclimation processes, we first analyzed autonomous float data to survey the frequency and magnitude of deep mixing events in the western North Atlantic Ocean. Next, we simulated a sustained deep mixing event in the laboratory and measured rates of acclimation processes ranging from light harvesting to growth in a model diatom and green alga. In both algae increases in chlorophyll (Chl) were coupled to growth, but growth of the green alga lagged the diatom by about a day. In float profiles, significant increases in Chl and phytoplankton carbon (C phyto) were detected below the mixed layer following episodic deep mixing events. These events pose a previously unrecognized source of new production below the mixed layer that can significantly boost the amount of carbon available for export to the deep ocean.

Light availability is a predominant factor limiting primary production across large regions of the world's oceans (Behrenfeld et al. 2016). Phytoplankton use a collection of processes to optimize growth depending on light availability (Lewis et al. 1984; Falkowski and LaRoche 1991; Fisher and Halsey 2016). The timescales under which these photoacclimation processes operate vary from milliseconds to days. For example, non‐photochemical quenching (NPQ) occurs within seconds of an increase in light intensity (Lohr and Wilhelm 1999; Milligan et al. 2012) and is vital for phytoplankton survival during periodic exposure to high light to prevent photoinhibition and damage to the photosynthetic reaction centers (Milligan et al. 2012). Slower‐acting acclimation processes are observed over timescales of days and generally require energy investments for anabolic activities such as pigment synthesis and growth (Laws and Bannister 1980; MacIntyre et al. 2002; Ross and Geider 2009).

Physical stratification of the water column is a critical factor impacting phytoplankton acclimation and accumulation. The Disturbance‐Recovery Hypothesis attributes phytoplankton accumulation following water column restratification to decreased grazing pressure brought about by deep mixing (Behrenfeld 2010). The Critical Turbulence Hypothesis posits that phytoplankton accumulate as turbulent motion decreases during water column stabilization (Huisman 1999; Franks 2015; Smith et al. 2015). In both cases, during prolonged periods of thermal stratification phytoplankton acclimate to high light intensities and low nutrient concentrations by decreasing chlorophyll (Chl) content to match downstream carbon fixation rates (Laws and Bannister 1980; Geider et al. 1997). Deep wind‐driven mixing can disrupt thermal stratification in the upper water column (Rumyantseva et al. 2015), altering the availability of light and nutrients within the mixed layer (Findlay et al. 2006; Rumyantseva et al. 2015). These episodic deep mixing events can last from hours to days during summer months and abruptly shift the phytoplankton growth environment to low light and nutrient replete conditions (Rumyantseva et al. 2015).

The biological carbon pump describes a collection of processes that move particles from the surface mixed layer into the deep ocean (Boyd et al. 2019). These collective processes operate across timescales ranging from days to months (Bishop et al. 1986; Dall'Olmo et al. 2016) and over spatial scales ranging from meters to thousands of kilometers (Levy et al. 2013; Omand et al. 2015). In addition to particle aggregation and sinking (Buesseler et al. 2007), particle injection pumps (PIPs) include the animal‐mediated, eddy‐subduction, large scale subduction, and mixed layer pumps (Boyd et al. 2019). The mixed layer pump includes seasonal shoaling and episodic deep mixing, which trap a large fraction of phytoplankton biomass below the mixed layer following restratification (Gardner et al. 1995; Kemp and Villareal 2013; Lacour et al. 2019). The contribution of the mixed layer pump to carbon export remains poorly understood because of the challenges associated with measuring the fates of carbon in the sub‐mixed layer. However, optical profiling (BGC Argo) floats now allow more intense scrutiny of the long‐established concepts of particle export, leading to fundamental changes in our understanding of biological carbon sequestration (Dall'Olmo et al. 2016; Boyd et al. 2019; Briggs et al. 2020).

At times during the spring through fall seasons, and especially in tropical and subtropical regions, light penetrates deeper than the mixed layer, leading to the potential for cells that were recently detrained from the mixed layer to grow, provided they can adjust their physiology fast enough to harvest energy under low light conditions before the cells are eaten by grazers, lysed by viruses, or sink out of the euphotic zone (Wiggert et al. 2000). The ability for phytoplankton to accumulate at depth, following a mixing event, depends on how rapidly the sum of photoacclimation processes (from light harvesting through carbon metabolism) integrate to affect growth in prolonged low light conditions. Recent analyses of a single deep mixing event in the North Atlantic showed that post‐mixing plankton communities that were kept in low light increased Chl within about 3 d (Morison et al. 2020), but whether those increases in Chl were coupled with growth was less clear (Graff and Behrenfeld 2018).

Few studies have investigated photoacclimation during sudden and sustained decreases in light intensity, such as those that occur during deep mixing events (Ross and Geider 2009; Graff and Behrenfeld 2018). Most research on photoacclimation has examined phytoplankton physiology under balanced growth (Laws and Bannister 1980; Fisher and Halsey 2016) or in response to rapid increases in light to understand mechanisms involved in photoprotection of photosynthetic reaction centers (Lavaud et al. 2004; Zhu and Green 2010). On timescales ranging from seconds to hours, diatoms decrease NPQ and alter carbon allocation mechanisms to sustain growth under decreasing light and even after exposure to darkness (Lohr and Wilhelm 1999; Bailleul et al. 2010; Poll et al. 2019). Green algae also decrease NPQ capacity, and their growth rate can slow drastically following a shift from high light to low light conditions (LaRoche et al. 1991). The time required to reach a new acclimated growth state depends on the magnitude of environmental change (Cullen and Lewis 1988) and potential differences in phytoplankton adaptive growth strategies across taxonomy (Litchman et al. 2012).

We first evaluated the frequency, duration, and intensity of episodic summertime mixing events in the North Atlantic Ocean. These events provided the basis for a laboratory culturing experiment that simulated phytoplankton light and nutrient environments before and after a deep mixing event, during which cells became trapped in a dim sub‐mixed layer. This experiment allowed us to study the timescales of photoacclimation processes in different phytoplankton. We found physiological trade‐offs provided a distinct advantage in the rate of photoacclimation in the model diatom, Thalassiosira pseudonana, compared to the model green alga, Dunaliella tertiolecta, but in both algae Chl increased in parallel with growth. Evaluation of these responses in the context of episodic deep mixing events revealed new production during and following mixing that provide a significant and unrecognized source for carbon export following mixed layer restratification.

Methods

Deep mixing events detected using in situ float retrievals

Deep mixing events were defined according to three criteria (Fig. S1): (1) the event must have occurred during the summer “stratified season” (mid‐April to mid‐September), (2) the event must have mixed below the average stratified mixed layer depth (MLD) determined for each float (Table 1), and (3) the change in mixing depth from profile x and x‐1 was less than 20% of the maximal mixing depth to confirm continued mixing (indicating that the MLD was still deepening or moderately shoaling, Fig. S1). This approach means that sustained mixing events (> 2 d in length) were only detected when profiles were made over consecutive days. The median light level in the MLD (I MLD; see calculation below) from the day before each deep mixing event was compared to I MLD during the deepest mixing for each event to characterize the changes in the light field.

Table 1

Float identifications, location ranges, and durations of deployments in the western North Atlantic Ocean. Data collected from optical sensors were used to assess deep mixing events in the region. Average MLD is the average mixed layer depth (m) detected by each float during the stratified season. Average change in temperature is given for the upper 100 m of the water column in each profile (°C). Data found at: ftp://misclab.umeoce.maine.edu/floats/ and https://seabass.gsfc.nasa.gov/experiment/NAAMES. Figure 1a gives the float tracks in the NAAMES region.

Float ID	Latitude minimum	Latitude maximum	Longitude minimum	Longitude maximum	Deployment duration	Daytime profiles	Nighttime profiles	Average MLD	Average change in temperature
572	53.6355	44.9197	−35.9373	−44.496	18 November 2015–2018 August 2017	173	59	28	1.3
573	52.5134	46.262	−25.7142	−40.034	16 November 2015–2113 April 2017	198	76	27	0.9
574	54.2803	52.3049	−40.1799	−47.5699	15 November 2015–8 March 2017	172	100	23	1.5
648	59.3369	54.8588	−39.0513	−55.4132	18 May 2016–2028 June 2018	95	137	27	0.8
846	53.7795	51.6096	−29.6168	−39.9949	17 September 2017–5 August 2018	62	4	29	2.4
847	54.0937	51.8813	−36.2085	−40.6644	13 September 2017–9 December 2018	152	11	33	2.6
848	54.6303	51.7053	−31.9608	−41.3701	12 September 2017–1 July 2018	165	8	33	1.8
849	51.6101	43.4794	−37.3061	−45.5249	10 September 2017–2 July 2018	166	8	28	1.7
850	47.7712	41.5737	−39.1929	−45.7988	8 September 2017–1 July 2018	168	11	25	0.8
851	46.2062	36.6883	−39.6306	−44.2569	6 September 2017–1 July 2018	170	17	25	1.7
852	44.4255	39.3833	−40.4172	−46.568	5 September 2017–2029 June 2018	170	11	27	1.2

Fig 1

(a) Tracks of the 11 floats in the North Atlantic Ocean during the stratified season. Each point represents a single profile during the lifetime of the floats. Points are colored according to float identification (Table 1). Dashed line depicts the boundary between subarctic and subtropical regions (Della Penna and Gaube 2019). (b) Location, duration, and intensity of 33 sustained deep mixing events (lasting at least 3 d) identified using optical sensors on the 11 autonomous profiling floats. Size of each point designates duration (in days) of the deep mixing event and color is scaled to show the magnitude of decrease in the average light level within the mixed layer from the pre‐mixing stratified state to the deepest mixed state.

Float‐derived properties of the water column

Water column structure, underwater light field, and phytoplankton carbon were evaluated in the North Atlantic Ocean using data collected from 11 Seabird Navis BGCi floats deployed during three cruises (November 2015, May 2016, September 2017) as part of the North Atlantic Aerosols and Marine Ecosystem Study (Fig. 1a, NAAMES; Behrenfeld et al. 2019). The floats recorded daily to weekly depth profiles (Data found at: ftp://misclab.umeoce.maine.edu/floats/ and https://seabass.gsfc.nasa.gov/experiment/NAAMES). A 0.03 density anomaly threshold (based on salinity and temperature) was used to define the MLD (Brainerd and Gregg 1995; McDougall and Barker 2011).

The median light level in the MLD (I MLD) was calculated according to Morel and Berthon (1989) (Eq. 1, where the diffuse attenuation coefficient [K d, m−1] was calculated from float‐measured Chl and daily photosynthetically available radiation [PAR] extracted from the MODIS‐Aqua satellite database [NASA Goddard Space Flight Center, Ocean Ecology Laboratory, Ocean Biology Processing Group]). Satellite‐retrieved PAR was used rather than PAR measured by the float because many float profiles were collected during the night or PAR was not available (Table 1).

Depth‐resolved changes in Chl and phytoplankton carbon (C phyto) throughout the stratified season were assessed using float‐retrieved fluorescence and backscatter, respectively. Chl was corrected for daytime NPQ according to Xing et al. (2012). Backscatter was converted to C phyto following the methods of Graff et al. (2016) and then integrated C phyto (∫C phyto) was calculated between the first post‐mixing MLD (1–3 d depending on profiling frequency) and the base of the euphotic zone (Z 0.1%, the depth at which measured PAR was less than 0.1% of incident PAR [Morel and Berthon 1989]). We call this region of the vertical water column the dim sub‐mixed layer. Changes in ∫C phyto following mixing were determined using the same water volume (the thickness of the dim sub‐mixed layer did not vary). Changes in ∫C phyto (ΔC phyto) in the dim sub‐mixed layer were calculated using the difference in ∫C phyto between profiles collected post‐mixing and the mixing profile immediately prior to restratification. New carbon production in the dim sub‐mixed layer was calculated by summing ΔC phyto in profiles made over 9 d following a mixing event. Carbon was only summed when the consecutive profile showed that the water column remained stratified. Events for which post‐mixing Z 0.1% was shallower than MLD were removed from the analysis because no dim sub‐mixed layer was present.

Lab‐simulated deep mixing events

T. pseudonana (Hustedt) Halse et Heimdal CCMP 1355 and D. tertiolecta CCMP 1320, two model phytoplankton of the diatoms and green algae, respectively, and with extensive physiological records, were grown separately to steady state growth in triplicate 300 mL N‐limited continuous cultures maintained at 18°C with 180 μE m−2 s−1 in f/2 + Si medium (Guillard 1975) with 100 μM NaNO3. The specific growth rate (μ, d−1) of the continuous culture was set by the rate of inflow of media (D) via a peristaltic pump according to Eq. 2, where V is the volume of culture. Cultures were maintained at 0.4 d−1 and bubbled continuously with sterile air to ensure even dispersal of cells and non‐limiting CO2 and O2.

Cultures were deemed to be in steady state growth following at least 10 cell divisions and cell densities that varied less than 5% over a 3‐d period. Cell counts were performed using a Multisizer 3 Coulter Counter with a 70 μM aperture (Beckman Coulter; Miami, FL). Once in steady state, 100 mL of each N‐limited continuous culture was diluted into 400 mL fresh f/2 + Si media with 100 μM NaNO3 to mimic the infusion of nutrients from below the mixed layer and placed at a 10 μmol photon m−2 s−1 growth irradiance with constant mixing to simulate the deep mixing event and dim sub‐mixed layer post‐mixing. Constant light conditions were used so that the bulk physiology of cultures would be independent of the time‐of‐day (Halsey et al. 2013). Use of fluctuating light conditions, however, can reveal different energy allocation strategies for photoprotection (Derks et al. 2015; Jallet et al. 2016; Andersson et al. 2019). Change in light intensity over the course of the simulated mixing event was modeled using Chl‐based K d. The maximum decrease in light intensity during the incubation was 15% or 1.5 μmol photon m−2 s−1. The modeled change in light intensity during the experiment was slightly greater than changes in light intensity during growth of both species in batch cultures grown with the same starting light intensity and cell densities as in the simulated deep mixing events and to final cell densities that were slightly higher than observed in the simulated deep mixing experiments.

Measurements of physiology were done immediately prior to dilution and for up to 7 d following the transition to the simulated deep mixing. For each sampling time‐point, Chl was measured in triplicate by filtering 5 mL of culture onto 25 mm glass fiber filters (Whatman GF/F) and extracting for 24 h at −20°C in 90% acetone. Absorptivity was measured using a UV‐Vis spectrophotometer (Shimadzu, Kyoto, Japan) and calculated using the equations for each taxa from Ritchie (2006). Samples of T. pseudonana were collected for pigment analysis by filtering 5 mL of culture onto a pre‐combusted GF/F and flash frozen prior to analysis by high‐performance liquid chromatography (HPLC) using a 250 mm C18 column, a Waters separation module, and a photodiode detector array. Calibration was done with Chl from Anacystis nidulans while mixed pigment standards were from DHI in Denmark (Wright 1991; Latasa et al. 1996; Jeffery et al. 1997; Bidigare et al. 2005). The large and rapid decrease in NPQ capacity in T. pseudonana prompted us to quantify its pigment profile.

Phytoplankton carbon and nitrogen was determined using 2, 3, and 4 mL culture samples filtered onto pre‐combusted GF/F filters and frozen until analysis using an Exeter Analytical A1 elemental analyzer (Coventry, England). Five milliliters of culture filtrate was re‐filtered onto a GF/F filter, frozen, and analyzed in the same way. Carbon and nitrogen measured in the filtrate was subtracted from the filtered samples to account for any dissolved organic carbon and nitrogen in the media.

Photosynthetic efficiency of photosystem II (F /F ) was measured using a custom‐built Fast Repetition Rate Fluorometer (FRRf) (Kolber et al. 1998). One hundred microliters of sample was diluted in triplicate into 3 mL F/2 + Si media and dark acclimated for 5 min to allow for NPQ relaxation prior to excitation and measurement. The relative number of available reaction centers per Chl (nPSII/Chl) was calculated from the functional cross section of photosystem II ( and the minimum fluorescence (F 0).

NPQ was measured using pulse amplitude modulation fluorometry (PhytoPAM, Walz, Effeltrich, Germany). 3.5 mL of culture was incubated at 18°C and dark acclimated for 10 min. Following dark acclimation, the light level was increased in 20 increments (ranging from 0 to 600 μmol photons m−2 s−1) for 5 min periods at each light level for measurement. NPQ was calculated as the difference between the maximum dark fluorescence (F ) at the beginning of measurement and the maximum fluorescence at each incubated light level () (Eq. 3). NPQ was reported for the growth irradiance during the pre‐ and post‐mixing environments.

Carbon uptake and gross carbon growth efficiency

Short term carbon incorporation was determined using 14C‐labeled bicarbonate incubations and used to estimate gross carbon production (Fisher and Halsey 2016). For the initial time point, 5 mL of culture was diluted into 8 mL of F/2 + Si medium and spiked with 5 μCi NaH14CO3. Post‐transition, 13 mL samples were collected and not diluted, and then spiked with 7.5 μCi NaH14CO3. Samples were transferred in 1 mL aliquots into 12 seven‐ml scintillation vials and incubated for 20 min at 10 different light levels ranging from 0–500 μmol photon m−2 s−1 to generate a photosynthesis‐irradiance curve modeled using the Jassby and Platt (1976) equation. The rate of photosynthesis at the growth irradiance (P IG) was determined from the resulting modeled curve.

Carbon uptake over 24 h to measure net carbon production was done using 5 mL of culture spiked with 1 μCi NaH14CO3 and incubated at the growth irradiance measured in chemostat and batch culture. A 3 mL culture sample spiked with 1 μCi NaH14CO3 and incubated in the dark was used to determine a background. After incubation for 24 h, bulk samples were then aliquoted into 1 mL samples for measurement. All samples were acidified with 10% HCl and degassed overnight. The total activity of NaH14CO3 added to each sample was measured using 50 μL of sample, 50 μL phenethylamine, and 900 μL water. Gross carbon growth efficiency (GGE) was calculated as the ratio of net carbon production (24 h 14C‐uptake rate) to gross carbon production (20 min 14C‐uptake rate).

ANOVA was conducted to determine the statistical significance of differences in properties between species and within measurements of each species across the time course of the simulated mixing event. Post hoc multiple comparisons were conducted to determine what days were contributing to the significant differences and the p‐values were corrected using the Bonferroni correction. Correlation tests were performed to determine physical parameters associated with new carbon production post‐mixing. Photosynthesis irradiance curves were analyzed for the Chl‐ and carbon‐normalized light limited slope of the curve (α b and α C, respectively) and the Chl‐ and carbon‐normalized light saturated rate of photosynthesis (P b max and P C max, respectively) using Sigma Plot version 14.

Results

Mixing events observed using in situ float retrievals

Evaluation of the 11 floats deployed in the western North Atlantic Ocean from November 2015 to August 2018 revealed 340 episodic deep mixing days associated with 93 separate deep mixing events (Table S1). Thirty‐three deep mixing events were three or more days in duration (Fig. 1). The most sustained deep mixing event lasted for 26 d, and the average duration of the 33 deep mixing events was 5.2 ± 0.6 (SE) days. The change in I MLD during mixing ranged from 5 to 186 μmol photons m−2 s−1 (Fig. 1b). This survey in the western North Atlantic Ocean revealed that deep mixing events were of significant magnitude and duration to induce phytoplankton to change their acclimation state from a high‐light and possibly nutrient‐limited physiology to a low‐light, nutrient‐replete physiology.

Laboratory‐based simulated deep mixing event

We next simulated a deep mixing event in the lab to observe how the timescales of phytoplankton acclimation processes vary between phytoplankton taxa and how these timescales of acclimation compare to the duration of the mixing event. The simulated deep mixing event was designed using a change of about 170 μmol photons m−2 s−1, which was at the high end of the changes observed in situ (Fig. 1b). We considered that a smaller change in light availability would give similar, but more muted physiological responses. The experiment began with cells fully acclimated to a high light, nutrient limited environment and then the cells were shifted to a low light, nutrient replete environment for up to 7 d to mimic the sustained deep mixing events observed in situ.

Acclimation of light harvesting parameters

T. pseudonana responded rapidly to the low light environment of the simulated deep mixing event by rapidly altering light harvesting processes. Photosynthetic efficiency (F V/F M) in the diatom remained high throughout the experiment, beginning at 0.49 on day 0 (during high light acclimated growth) and reaching 0.59 during deep mixing (Table 2). T. pseudonana achieved this high photosynthetic efficiency by holding σPSII constant (Fig. 2a), but rapidly increasing Chl content fivefold such that it reached its maximum level of 0.39 pg Chl cell−1 by the third day (Fig. 2c). The increase in Chl was accompanied by a decrease in nPSII/Chl from 4.2 to 0.5 by the third day and stayed relatively stable until day 5 when Chl content drops as cells reached stationary phase (Fig. 2a). Chl increased at the expense of photoprotective pigments, specifically the xanthophyll carotenoid diatoxanthin, which decreased 100%, within 24 h (Table 3). Loss of photoprotective pigments was accompanied by an immediate 92% decrease in NPQ capacity expressed at the growth irradiance during the first 24 h, but NPQ stabilized at 8% of its high light acclimated capacity for the remainder of the simulated deep mixing event (Fig. 2d).

Table 2

Photosynthetic parameters from 14C‐based photosynthesis irradiance curves and FRRf measurements following the transition from pre‐mixing (day 0) to post‐mixing (days 1–5 in Thalassiosira pseudonana and days 1–7 in Dunaliella tertiolecta). R 2 represents the fit of the modeled photosynthesis irradiance curve.

Day	D. tertiolecta									T. pseudonana
	α b	P b max	P b IG	α c	P c max	P c IG	E k	R 2	F V/F M	C : N	α b	P b max	P b IG	α C	P C max	P c IG	E k	R 2	F V/F M	C : N
0	2.0 (1.2)	247 (99)	191 (23)	0.03 (0.02)	3.6 (1.7)	2.8 (1.6)	142 (36)	0.95 (0.003)	0.52 (0.006)	13.8	3.0 (1.8)	242 (211)	216 (31)	0.05 (0.03)	4.3 (3.4)	3.6 (0.5)	294 (243)	0.95 (0.006)	0.48 (0.008)	8.6
1	4.3 (0.5)	498 (178)	41 (71)	0.16 (0.05)	23 (15)	1.6 (0.5)	123 (56)	0.83 (0.12)	0.59 (0.004)	5.1	5.4 (1.6)	453 (15)	51 (48)	0.29 (0.06)	25 (3.4)	2.7 (0.5)	93 (31)	0.94 (0.021)	0.58 (0.009)	3.0
2	3.2 (0.7)	327 (11)	31 (32)	0.22 (0.04)	22 (2.1)	2.1 (0.3)	108 (27)	0.91 (0.06)	0.57 (0.003)	5.3	5.9 (1.2)	431 (4.1)	55 (43)	0.42 (0.05)	31 (2.6)	3.9 (0.4)	76 (15)	0.95 (0.009)	0.59 (0.003)	3.7
3	2.5 (0.04)	300 (57)	24 (23)	0.15 (0.05)	20 (9.8)	1.4 (0.5)	123 (25)	0.95 (0.02)	0.57 (0.003)	7.5	5.1 (0.3)	369 (70)	48 (40)	0.47 (0.04)	34 (6)	4.4 (0.3)	73 (19)	0.93 (0.025)	0.58 (0.006)	3.6
4	2.7 (0.8)	258 (11)	26 (15)	0.16 (0.03)	16 (2.6)	1.5 (0.3)	105 (36)	0.98 (0.01)	0.57 (0.005)	6.8	4.6 (1.5)	302.2 (4.55)	42 (28)	0.48 (0.2)	30 (3.1)	4.4 (1.8)	75 (26)	0.94 (0.043)	0.57 (0.004)	4.0
5	2.0 (0.2)	231 (14)	19 (15)	0.14 (0.01)	17 (4.0)	1.4 (0.1)	116.0 (19)	0.97 (0.01)	0.58 (0.003)	6.6	7.9 (3.4)	502 (155)	73 (43)	0.85 (0.5)	54 (23)	7.9 (4.1)	68 (9)	0.97 (0.003)	0.57 (0.007)	4.2
6	2.0 (0.4)	223 (0.4)	20 (20)	0.17 (0.01)	19 (3.2)	1.6 (0.1)	113 (27)	0.93 (0.06)	0.58 (0.005)	6.0	–	–	–	–	–	–	–	–	–	–
7	2.9 (1.2)	209 (125)	27 (14)	0.25 (0.1)	17 (1.1)	2.3 (1.0	85 (34)	0.97 (0.01)	0.58 (0.007)	8.1	–	–	–	–	–	–	–	–	–	–

Fig 2

Light harvesting parameters in Thalassiosira pseudonana (blue squares and lines) and Dunaliella tertiolecta (black triangles and lines) pre‐mixing (day 0) and following transition to the post‐mixing environment (days 1–7). Functional cross section of photosystem II (σPSII, dashed line), minimum fluorescence (F o, dotted line) and relative number of reaction centers per Chl (solid line) in T. pseudonana (a) and D. tertiolecta (b). Chl per cell normalized to the highest value for each species (c). Non‐photochemical quenching capacity (d). T. pseudonana data shown only through day 5 because cells began to enter stationary phase.

Table 3

Percent change in primary and accessory pigment concentration (ng cell−1) from day 0 (stratified, pre‐mixing environment) to day 1 in Thalassiosira pseudonana (post‐mixing environment).

Pigment	Δ [pigment]D1–D0
Diatoxanthin	−100
Chlorophyllide	−36
Diadinoxanthin	28
Divinyl chlorophyll a	32
Violaxanthin	62
β‐Carotene	73
Chlorophyll C	76
Fucoxanthin	77
Chlorophyll a	77
Monovinyl chlorophyll a	78

Light harvesting processes in D. tertiolecta required 6 d to acclimate to the decrease in light availability. Photosynthetic efficiency remained between 0.52 and 0.57 throughout the experiment (Table 2). nPSII/Chl remained at 4.2 mol PSII pg Chl −1 for 1 d following the simulated deep mixing, but dropped by 50% on day 2, and then required two more days to stabilize at a value of 0.7 (Fig. 2b). σPSII decreased by 32% by the fourth day (Fig. 2b), but Chl slowly increased 6.25‐fold over the entirety of the simulated mixing event (Fig. 2c). The NPQ capacity of D. tertiolecta decreased by 46% over the first 24 h, and then stayed constant for the next 6 d (Fig. 2d). Thus, in D. tertiolecta the trade‐off of minimizing photoprotection in favor of increasing light harvesting capacity was far slower and subtler in magnitude compared to T. pseudonana.

Acclimation of growth processes

T. pseudonana exhibited little delay in growth following the sudden exposure to low light. Cell carbon content increased immediately from 6.6 to 7.2 pg C cell−1 by day 1 (Fig. 3a). The decrease in cell carbon from 7.3 to 5.8 pg C cell−1 on day 4 and the decrease on day 6 were the result of cell division associated with population doublings reflected in cell density measurements (Fig. 3a, 3c ). Prior to the simulated deep mixing, T. pseudonana was growing at a light‐saturated, nutrient‐limited specific growth rate of 0.4 d−1. Within 2 d following the simulated deep mixing, T. pseudonana recovered to a mean light‐limited growth rate of 0.3 d−1 (Fig. 3b). These rapid growth responses were mediated by the remarkably high GGE exhibited by T. pseudonana. GGE was 0.75 on day 0, decreased to 0.25 on day 1, and recovered to an average GGE of 0.60 over the remaining 5 d of the experiment (Fig. 3b).

Fig 3

Growth processes in Thalassiosira pseudonana (blue squares and lines) and Dunaliella tertiolecta (black triangles and lines) pre‐mixing (day 0) and following transition to post‐mixing environment (days 1–7). Cellular carbon content (solid lines) and cellular volume (dashed lines) (a). Growth rate determined from changes in particulate carbon (T. pseudonana data shown only through day 5 because cells began entering stationary phase) (b). Normalized cell density (solid line) and gross carbon growth efficiency (GGE; dashed line) for T. pseudonana (c) and D. tertiolecta (d).

Photosynthesis‐irradiance relationships were used to investigate how the changes in the capacity for light harvesting reactions balanced with downstream changes in carbon metabolism throughout the experiment. P b max in T. pseudonana increased nearly twofold over the first 24 h, from 242 to 453 μmol C mg Chl−1 h−1 and remained higher than the rate on day 0 throughout the simulated mixing (Table 2). Within 1 d α b increased from 3.0 to 5.4 μmol C mg Chl−1 h−1 (μmol photons m−2 s−1)−1 and subsequently maintained a high value between 5 and 8.8 μmol C mg Chl−1 h−1 (μmol photons m−2 s−1)−1. Nevertheless, the resultant light‐saturation index (E k), which began at 300 μmol photons m−2 s−1, stabilized at around 80 μmol photons m−2 s−1 within 1 d (Table 2). Photosynthesis irradiance parameters behaved similarly whether normalized to Chl or carbon content (Table 2). When normalized to carbon, and using appropriate unit conversions, the rate of carbon fixation at the growth irradiance (P C Ig) overestimated specific growth rate by about threefold.

Growth responses were comparatively sluggish in D. tertiolecta following exposure to low light. By day 1, D. tertiolecta decreased cell carbon content by 52% (from 46 to 21 pg C cell−1), but this decrease was associated with cell division in 38% of the population (Fig. 3d). Consistent with cell division, the average cell volume decreased from 211 to 153 μm3 between day 0 and day 1 (Fig. 3a). D. tertiolecta maintained a steady‐state specific growth rate of 0.4 d−1 during high light acclimation. Upon mixing, growth halted and then recovered within 3 d to a growth rate of 0.38 d−1 but only averaged a growth rate of 0.1 d−1 for the remainder of the mixing event (Fig. 3b). During this period, cell carbon reached 38 pg C cell−1 by day 7. GGE also dropped from 0.67 on day 0 to 0.20 for days 1 through 7 (Fig. 3d).

Photosynthesis irradiance parameters in D. tertiolecta were responsive to the change in light and nutrient environment caused by the simulated mixing event. From day 0 to day 1 α b increased from 2.0 to 4.3 μmol C mg Chl−1 h−1 (μmol photons m−2 s−1)−1 but decreased back to 2.0 μmol C mg Chl−1 h−1 (μmol photons m−2 s−1)−1 in the following days (Table 2). Similarly, P b max increased from 246 on day 0 to 498 μmol C mg Chl−1 h−1 on day 1, and then declined back to 231 μmol C mg Chl−1 h−1 by day 5. The same trends emerged when the photosynthesis irradiance parameters were normalized to carbon content (Table 2), and P C Ig was about twice the specific growth rate. The parallel behaviors in α and P max yielded a relatively constant E k that shifted only slightly from 140 μmol photons m−2 s−1 during high light acclimated growth to about 100 μmol photons m−2 s−1 throughout the simulated deep mixing experiment (Table 2).

In situ new production following deep mixing events

Our lab‐simulated deep mixing events suggested that increases in cell Chl content below the mixed layer following restratification could be associated with growth and new production. To investigate this idea, we returned to the autonomous float profiles to characterize Chl and carbon dynamics following the deep mixing events.

Profiles during the 93 deep mixing events (Table S1) showed decreases in surface Chl caused by deepening of the mixed layer and dilution of the standing stock (example profiles given in Fig. 4b,e,h, gray lines; Figs. S2–S6). Following deep mixing, Chl increased in the dim sub‐mixed layer, with one event showing an increase of 2.3 mg m−3 within 3 d post‐mixing in the subtropical region (Fig. 4h). This Chl increase was associated with a C phyto concentration that increased 1.6‐fold during the same period (Fig. 4i). C phyto concentrations sometimes decreased within 3 d post‐mixing but then increased by the next profile collection (i.e., Fig. 4c). C phyto concentrations increased in the dim sub‐mixed layer within 9 d post‐mixing in 68 of the 93 mixing events detected (Table S1). Temperature‐salinity plots showed little change in these properties in the water masses studied, thus, changes in Chl and C phyto following restratification were attributed to biological activity and not to encroaching water masses with different characteristics (Fig. 4a,d,g, Figs. S2–S6).

Fig 4

Mixing‐induced new production in the dim sub‐mixed layer. Temperature vs. salinity (a, d, g), chlorophyll concentration (b, e, h), and C phyto concentration (c, f, i) with depth determined from profiling floats (profiles shown for before mixing, blue lines; during mixing, gray lines; and up to three profiles after mixing; orange, red, and dark red lines, from floats representing the subarctic region (float 574; a, b, c), boundary (float 572; d, e, f), and subtropical region (float 850; g, h, i). Integrated carbon was determined from C phyto concentrations in the dim sub‐mixed layer (shaded in c, f, i) during mixing and post mixing to determine new production caused by deep mixing events (Table 5).

Table 5

Statistical analyses for new carbon production in the dim sub‐mixed layer detected by 11 floats in the North Atlantic Ocean. Region was split into SA, subarctic; B, boundary; ST, subtropical and season was divided into SP, March to May; MS, June to July; LS, August to October. One‐way ANOVA were performed with follow up pairwise comparisons using a Bonferroni correction for significant ANOVAs (significant result of ANOVA or post hoc tests [p‐value < 0.05] indicated by asterisks).

Variables		Mean			ANOVA		Post hoc test
Categorical	Dependent	SA	B	ST	F value	P value	SA vs. B	SA vs. ST	B vs. ST
Region	MLDMIXING (m)*	39 (2.8)	52 (6.7)	39 (1.9)	3.0	0.051	–	–	–
	MLDPOST‐MIXING (m)†	21 (1.1)	21 (1.7)	20 (1.2)	0.34	0.710	–	–	–
	Z 0.1% (m)	35 (1.3)	40 (2.1)	69 (6.5)	26	<0.001*	1.000	<0.001*	<0.001*
	Thickness of DSML (m)‡	12 (1.8)	19 (2.3)	48 (6.4)	28	<0.001*	0.470	<0.001*	<0.001*
	Mixing duration (d)	4.6 (0.7)	4.4 (0.9)	3.8 (0.7)	0.2	0.819	–	–	–
	Days post mixing (d)§	3.4 (0.3)	4.4 (0.4)	3.7 (0.4)	2.1	0.126	–	–	–
	C phyto production (mg C m−2 d−1)	160 (37)	41 (12)	55 (19)	5.0	0.008*	0.014*	0.055*	1.000
	Percent change C phyto ¶	59 (11)	24 (7.0)	40 (8.6)	0.7	0.481	–	–	–

MLDMIXING is the depth of the mixed layer during the profile immediately prior to restratification.

MLDPOST‐MIXING is the depth of the mixed layer during the first profile following restratification.

Dim sub‐mixed layer (DSML) was defined as the water layer between the MLDPOST‐MIXING and Z 0.1%.

Days post mixing is the number of days between restratification and the final profile used to calculate ΔC phyto.

Determined from ∫C phyto up to 9 d post mixing minus ∫C phyto from the profile immediately prior to restratification divided by ∫C phyto from the profile immediately prior to restratification.

Depth integrated C phyto (∫C phyto) in the dim sub‐mixed layer immediately prior to restratification and up to 9 d post‐mixing (see gray shaded areas in Fig. 4) were used to determine changes in ∫C phyto (ΔC phyto) induced by deep mixing events. In total, 40 of 93 deep mixing events resulted in ΔC phyto values greater than 50 mg C m−2 up to 9 d post‐mixing (Table S1). ΔC phyto averaged across all deep mixing events for each float ranged from 69 to 631 mg C m−2 (Table 4). There were no significant differences in ΔC phyto determined within 1–3 or 4–9 d post‐mixing (p = 0.25, Table 5). New carbon production induced by all deep mixing events detected by each float ranged from an average of 21.1 to 294 mg C m−2 d−1 (Table 4). Float 574 detected 893 mg C m−2 d−1, which was the highest rate of new C production associated with a single mixing event (Fig. 4f, Table S1). Mean mixing‐induced production values were significantly higher in the subarctic region than in the boundary or subtropical regions (p < 0.001, Table 5, Figs. S2–S6, Table S1). However, no differences were observed in new carbon production determined for different time‐periods during the stratified season (e.g., spring, early summer, or late summer) in any region (two‐way ANOVA interaction p = 0.8, Table 5). Thus, mixing‐induced production is unlikely to derive from the progressive seasonal stratification change associated with the seasonal mixed layer pump, thereby increasing confidence in their attribution to post‐mixing photoacclimation responses.

Table 4

New carbon (ΔC phyto) and new carbon production in the dim sub‐mixed layer induced by deep mixing and averaged across all events for each float. The ranges of ΔC phyto and new carbon production across all events detected by each float are also given. Data for each mixing event detected by all floats are given in Table S1.

Float	Subregion*	Number of mixing events	Average ΔC phyto 1–3 d post mixing (mg C m−2)†	Average ΔC phyto 4–9 d post mixing (mg C m−2)‡	Average ΔC phyto up to 9 d post mixing (mg C m−2)	Range of ΔC phyto (mg C m−2)	Daily new carbon production post mixing (mg C m−2 d−1)	Range of new C production post mixing (mg C m−2 d−1)
574	Subarctic	14	558	291	445	−27 – 1386	294	−6 – 893
648	Subarctic	11	275	332	237	0.4 – 833	70	0.1 – 320
846	Subarctic	6	166	212	105	−52 – 388	86	−11– 388
847	Subarctic	5	129	120	139	−218 – 428	46	−36 – 104
848	Subarctic	7	332	57	245	21 – 757	104	11 – 378
572	Boundary	14	192	231	153	−119 – 932	26	−40 – 197
573	Boundary	8	213	308	186	−99 – 819	50	−75 – 185
849	Boundary	8	172	196	215	−10 – 613	45	−10 – 102
850	Subtropical	6	334	565	631	−0.2 – 1632	106	−0.1 – 272
851	Subtropical	8	206	101	189	60 – 359	98	10 – 359
852	Subtropical	12	121	54	69	−65 – 374	21	−32 – 62

Subregion defined according to Della Penna and Gaube (2019).

ΔC phyto (mg C m−2) in the dim sub‐mixed layer 1–3 d post‐mixing depending on profiling frequency.

ΔC phyto (mg C m−2) in the dim sub‐mixed layer 4–9 d post‐mixing depending on profiling frequency and when the water column remained stable (no additional mixing occurred).

Discussion

The deep mixing events identified in the western North Atlantic using autonomous profiling floats are almost certainly only a fraction of the total number of episodic deep mixing events that occurred during the stratified season in that region during the 3 years the floats were deployed (Koeve et al. 2002; Waniek 2003). Nevertheless, this data set shows that episodic deep mixing events occur with significant frequency and expose phytoplankton to large changes in the light environment. In this study, we sought to characterize the timescales of photoacclimation in different phytoplankton relative to the magnitude and duration of deep mixing events to determine whether phytoplankton have the capacity to contribute new production during and immediately following deep mixing events.

The contribution of episodic deep mixing events to new production and export is currently not well known, due to their transient nature and the lack of knowledge or attention given to phytoplankton responses below the MLD (Dall'Olmo et al. 2016; Llort et al. 2018). Deep mixing events during the stratified season that were at least 3 d in duration caused light availability to decrease up to 80% in the mixed layer (Fig. 1b) and likely also increased nutrient availability, especially in the late summer. These environmental changes present an opportunity for phytoplankton to grow if they can quickly acclimate to the low light conditions. Below, we first discuss the differences in photoacclimation strategies measured in a diatom and a green alga. These growth behaviors are next discussed in the context of photoacclimation responses to episodic deep mixing events observed using profiling floats and the potential for new production below the mixed layer to augment the biological carbon pump.

Photoacclimation strategies

In nearly all measures of physiology, T. pseudonana responded to the simulated deep mixing event nearly twice as fast as D. tertiolecta. Differences in photoprotective pigment compositions between the two species (e.g., diatoxanthin is present in T. pseudonana, but not in D. tertiolecta) may partially mediate the photoacclimation response‐times (Lohr and Wilhelm 1999; Nymark et al. 2013). The loss of the photoprotective pigment, diatoxanthin, in T. pseudonana occurred concomitant with a dramatic reduction in NPQ capacity (to 8% of its observed maximum) and significant increases in light harvesting pigments. In another diatom, Phaeodactylum tricornutum, the light harvesting protein LHCX1, functions as a constitutive, low‐level source of NPQ and a possible structural component of the fucoxanthin Chl protein supercomplex (Bailleul et al. 2010). The 77% increase in fucoxanthin in T. pseudonana, was likely derived from β‐carotene (Kuczynska et al. 2015), which also increased 73%, and supports the idea that this species relies on the dual activities of the fucoxanthin Chl protein supercomplex for NPQ and funneling light energy to reaction centers (Zhu and Green 2010).

T. pseudonana maintained a high GGE (0.65 compared to 0.25 in the green algae) during simulated deep mixing, indicating the diatom's ability to rapidly reallocate energy and carbon into shorter, more efficient pathways compared to the green alga. For example, across a wide range of growth rates diatoms rely more heavily on non‐carbon respiratory pathways (e.g., Mehler reaction, terminal oxidase activity) in the light than in the dark (Fisher and Halsey 2016). Conservation of carbon storage reserves during the light period allows their use in energy generation during sudden light deprivation, a strategy likely shared by other diatoms (Nymark et al. 2013). In this study the diatom recovered its GGE to > 0.65 within 2 d of the shift from 180 to 10 μmol photons m−2 s−1, suggesting that T. pseudonana employs carbon pathways that permit high conservation of energy and carbon (Heydarizadeh et al. 2017; Wagner et al. 2017). Diatoms appear to use phosphoenolpyruvate and pyruvate as “hubs” in central metabolism that are then gated for carbon flow into respiratory or biosynthetic pathways (Heydarizadeh et al. 2017; Wagner et al. 2017). Furthermore, the peroxisomal glyoxylate cycle appears to play a key role in conserving carbon in diatoms through salvage pathways (Davis et al. 2017) and bypass of decarboxylation steps in the tricarboxylic acid cycle. The low carbon demand in silica frustule synthesis (Wagner et al. 2017) is another reason why diatoms may outpace other algae during photoacclimation.

Despite their markedly different GGEs during the simulated deep mixing event, the photosynthetic parameters, α and P max determined from short‐term 14C‐uptake measurements shifted similarly in both algae. The immediate twofold increases in α b and P b max and decreased C : N following transition to the mixing condition reflected relief from nitrogen limitation (Table 2) and a shift to unbalanced growth caused by the simultaneous decrease in light availability. These changes caused carbon to be redirected away from pathways causing rapid turnover of newly fixed carbon and into pathways leading to longer‐term storage products (Wilhelm et al. 2006; Jakob et al. 2007), a behavior previously described in algae growing at different steady state growth rates (Behrenfeld et al. 2004; Fisher and Halsey 2016).

The parallel behaviors in α b and P b max help explain how changes in carbon metabolism are coping strategies used by both algae. The experiment began with cells in steady state nutrient‐limited, light saturated growth. Under these conditions the turnover rate of newly fixed carbon was relatively fast, resulting in low α b and P b max values (Halsey et al. 2010). The shift to low light and nutrient repletion caused the algae to synthesize a greater proportion of carbon storage reserves (longer term polysaccharides) from carbon skeletal structures causing a slow carbon turnover rate and high α b and P b max values (Wilhelm et al. 2006; Halsey et al. 2013). Over the next few days as cells remained N‐replete and in low light, they progressively shifted carbon metabolism toward more rapid carbon turnover causing parallel decreases in α b and P b max that reached values previously observed under light limited, N‐replete growth (Fisher and Halsey 2016). The two to threefold higher P C Ig values compared to specific growth rates for both species reflect the high demand for storage reserves compared to low molecular weight carbon end products throughout the experiment (Halsey et al. 2013). Thus, covariation of α b and P b max and α C and P C max observed in the field over the course of days reflects cell strategies that direct new carbon into storage or catabolism (respiration) in response to transient changes in light and nutrient availability (i.e., episodic deep mixing events). In contrast, observed seasonal variations in α b and P b max are caused by growth‐rate dependent changes in carbon metabolism (Halsey et al. 2010).

Diatoms are known for their efficient growth under low light and rapid responses to increasing light availability (Geider et al. 1986; Fisher and Halsey 2016; Poll et al. 2019). In our experiments, photoacclimation of the diatom T. pseudonana outpaced the green alga, D. tertiolecta, following the shift from high to low light. Several physiological properties facilitated the speed of the diatom response, including efficient replacement of photoprotective pigments with Chl, parallel adjustments in light harvesting and carbon metabolism, and reallocation of energy into carbon‐conserving pathways (Ross and Geider 2009; Heydarizadeh et al. 2017). Maintenance of a constant Rubisco content in the diatom across the light and nutrient environments of our experiments likely also facilitated rapid carbon fixation upon relief from N limitation (Li and Campbell 2017). These behaviors can boost the potential yield of new production in the dim sub‐mixed layer following mixing‐restratification events.

New production in the dim sub‐mixed layer

Evidence that Chl accumulation in the dim sub‐mixed layer following a mixing‐restratification event is coupled to biomass accumulation and is not simply a pigment response suggests that episodic deep mixing events can augment the biological carbon pump. Algae trapped below the mixed layer during seasonal shoaling are recognized as contributing to carbon export (Lacour et al. 2019), but little attention has been given to new carbon production in the dim sub‐mixed layer immediately following restratification (Graff and Behrenfeld 2018). Investigating whether phytoplankton grow following a sudden decline in light required us to link the timescales of physical changes in the environment (mixing events) and timescales associated with phytoplankton physiology. Upregulation of Chl was coupled to growth in both species, but the diatom recovered growth within 2 d, while the green alga lagged behind by a day (Fig. 3b). Regardless of the speed with which phytoplankton photoacclimate, increases in Chl occurred in parallel with shifts in other metabolic processes that direct harvested light energy into growth. In the float profiles following many of the deep mixing events, pulses of Chl were observed within these same timescales (i.e., 2–6 d; Fig. 4, Figs. S2–S6) and were coupled with increases in C phyto (Fig. 4). Decreases in C phyto in the dim sub‐mixed layer (Fig. 4c, S2–S6) were occasionally observed following deep mixing and are likely due to sinking, grazing or lysis.

New phytoplankton production in the dim sub‐mixed layer following deep mixing is likely underpinned by decreased grazing pressure (Graff and Behrenfeld 2018; Morison et al. 2020). The initial consequences of deep mixing are (1) that phytoplankton standing stocks are diluted and separate the grazers from their phytoplankton prey, as posited in the Disturbance‐Recoupling Hypothesis that explains phytoplankton bloom events (Behrenfeld and Boss 2014) and (2) an immediate drop in phytoplankton growth rate (Fig. 3b). As phytoplankton acclimate to the new low light environment, their growth rate accelerates, and ∫C phyto at least temporarily outpaces losses to predation (Fig. 4c). Importantly, this acceleration in growth rate (not the growth rate itself) facilitates phytoplankton accumulation, as was observed by the positive ΔC phyto values in the dim sub‐mixed layer following mixing events (Fig. 4c,f,i and Table 4, Figs. S2‐S6). In a sense, these small scale episodic deep mixing events provide a temporally‐condensed view of the seasonal bloom cycles explained by Behrenfeld and Boss (2014).

Recent studies suggest that the PIP, which includes physical movement of carbon into the sub‐mixed layer by episodic deep mixing, could account for as much carbon sequestration as the gravitational pump (Boyd et al. 2019). Quantifying the PIP carbon flux pathways relied heavily on autonomous profilers (Omand et al. 2015; Dall'Olmo et al. 2016), but these studies did not investigate the impact of photoacclimation processes on the standing stock of carbon below the mixed layer. Across all floats and regions, we show that episodic deep mixing induced an average of 86 ± 23 mg C m−2 d−1 in the post mixing dim sub‐mixed layer (average daily new carbon production post mixing, Tables 4 and 5). While this value is ca. 10% of new production in the mixed layer during roughly the same period (840 ± 110 mg C m−2 d−1, Table S2, Fox et al. 2020), the relative contribution of individual events varied significantly, ranging up to 110% of new production in the mixed layer (Tables S1 and S2). The variation in the magnitude of mixing induced production shown in this study highlights this process as an important additional component to known PIPs. Importantly, there was no relationship between mixing‐induced new production and time of year within the mid‐April to mid‐September stratified season, which suggests that episodic deep mixing events and rapid phytoplankton photoacclimation combine to enhance export by the mixed layer pump throughout the annual cycle. Thus, while the mixed layer pump is generally viewed as being primarily associated with spring seasonal restratification (Dall'Olmo et al. 2016; Boyd et al. 2019; Lacour et al. 2019), episodic deep mixing and photoacclimation during the stratified summer months should be recognized as an additional mechanism of particle injection to the mesopelagic zone.

New carbon production induced by deep‐mixing was observed throughout the North Atlantic region and was significantly greater in the subarctic region (160 mg C m−2 d−1) compared to the boundary (41 mg C m−2 d−1) and subtropical region (55 mg C m−2 d−1). No parameter describing the light environment over time or space was found to explain this regional difference. For example, despite the greater light availability afforded by the deeper Z 0.1% and thicker dim sub‐mixed layer in the subtropical compared to the subarctic regions, significantly more carbon accumulated below the mixed layer in the subarctic region (Table 5). Furthermore, the relationships between these parameters and all three regions were nonlinear, suggesting that the parameters controlling mixing‐induced new production vary regionally or temporally. Alternatively, the controlling parameter was not considered in our analysis. We hypothesize that the combined influences of the phytoplankton community composition (see below) and interactions with grazers underpin the greater amount of new carbon production detected in the subarctic region. We suggest that future research that explores the influence of multiple variables would be especially useful for quantifying mixing‐induced new production and its impact on the overall biological carbon pump (Boyd et al. 2019).

The effect of climate change on the frequency and intensity of storms over the global oceans remains unclear, but reportedly trends toward increased storm events in high latitude regions, especially in the southern hemisphere (Fischer‐Bruns et al. 2005; Chang et al. 2012; Ardyna et al. 2014). Thus, deep mixing‐induced new production in the dim sub‐mixed layer should be recognized as providing a substantial boost to the biological carbon pump.

The growth responses of the diatom and green algae were used to coarsely evaluate the potential influence of community composition on new production in the dim sub‐mixed layer following mixing events. Both diatoms and green algae were present in the North Atlantic in varying proportions during the NAAMES campaign (Bolaños et al. 2020). Our results show that diatoms could dominate cell abundances from three to 9 d post mixing when they make up at least a quarter of the population immediately post mixing (Fig. S7). This prediction ignores the role of grazers which can quickly establish top‐down control following a disturbance event (Graff and Behrenfeld 2018; Morison et al. 2020) and may have prey‐specificities that will impact the phytoplankton composition. Nevertheless, diatoms are often associated with high carbon flux out of the euphotic zone, attributable to their higher sinking rates made possible by the ballast effect of their biomineral frustules and their affinity for aggregation (Klaas and Archer 2002; Moigne et al. 2015; Durkin et al. 2016). Thus, the physiological advantages of diatoms that we measured in the lab could result in the prevalence of diatoms in the dim sub‐mixed layer and greater carbon export following mixing events.

Our study shows that episodic deep mixing events boost the estimated total carbon pool available for export. Phytoplankton growth in the dim sub‐mixed layer can rapidly replenish the carbon pool diluted by deep mixing events that occur in the months following spring stratification, thus increasing the system's f‐ratio. How the additional impacts of community composition and predation impact these growth responses merit further study. The timescales under which photoacclimation processes operate underlie the potential for accumulation of phytoplankton following environmental disturbance. Remote detection of Chl and ΔC phyto from floats facilitate highly resolved temporal and spatial observations of phytoplankton behaviors and were studied here in the context of the dynamic physical environment. This integrated approach is helping to shed light on the range of adaptive strategies used by diverse phytoplankton species to response to environmental changes and how these strategies impact marine carbon cycling.

Data Availability Statement

The data presented are available from the SeaBASS repository at (https://seabass.gsfc.nasa.gov/experiment/NAAMES) and the University of Maine (ftp://misclab.umeoce.maine.edu/floats/).

Conflict of Interest

None declared.

Fig. S1 Conceptual diagram to describe mixing event criteria. The first criterion is that the mixing must have occurred during the summer stratified season (March 15–October 15). The second criterion is that the mixed layer depth (MLD, solid line) must have penetrated the average stratified MLD for the float during the stratified season (30 m in this example). The third criterion is that after a deep mixing profile (D0), the subsequent day's MLD must have continued to deepen or only shallowly shoaled, which is determined by the change in the MLD being less than 20% the maximum MLD (e.g., 20% of 60 m = 12 m in the figure) . In this example, D0 is the start of the mixing event because the MLD is deeper than 30 m. D1 is included in the mixing event because D0–D1 < 12 m. However, D1–D2 > 12 m and is thus not included. This mixing event was 2 d in duration.

Fig. S2. Mixing‐induced new production in the dim sub‐mixed layer. Temperature vs. salinity (a), chlorophyll (b), and C phyto (c) depth profiles determined from profiling floats (profiles are shown for before mixing, during mixing, and up to three profiles after mixing—blue, gray, orange, red, and dark red solid lines, respectively) from float 573. Heavy dashed line (‐ ‐ ‐) is the mixing depth and dotted line (• • •) is the euphotic depth, Z 0.1%. Carbon in the region from MLDpost to Z 0.1%post (the dim sub‐mixed layer, shaded region) was summed both during mixing and post mixing to determine new production caused by deep mixing events (Table 5).

Fig. S3. Mixing‐induced new production in the dim sub‐mixed layer. Temperature vs. salinity (a), chlorophyll (b), and C phyto (c) depth profiles determined from profiling floats (profiles are shown for before mixing, during mixing, and up to three profiles after mixing—blue, gray, orange, red, and dark red solid lines, respectively) from float 648. Heavy dashed line (‐ ‐ ‐) is the mixing depth and dotted line (• • •) is the euphotic depth, Z 0.1%. Carbon in the region from MLDpost to Z 0.1%post (the dim sub‐mixed layer, shaded region) was summed both during mixing and post mixing to determine new production caused by deep mixing events (Table 5).

Fig. S4. Mixing‐induced new production in the dim sub‐mixed layer. Temperature vs. salinity (a), chlorophyll (b), and C phyto (c) depth profiles determined from profiling floats (profiles are shown for before mixing, during mixing, and up to three profiles after mixing—blue, gray, orange, red, and dark red solid lines, respectively) from float 850. Heavy dashed line (‐ ‐ ‐) is the mixing depth and dotted line (• • •) is the euphotic depth, Z 0.1%. Carbon in the region from MLDpost to Z 0.1%post (the dim sub‐mixed layer, shaded region) was summed both during mixing and post mixing to determine new production caused by deep mixing events (Table 5).

Fig. S5. Mixing‐induced decrease in new production in the dim sub‐mixed layer. Temperature vs. salinity (a), chlorophyll (b), and C phyto (c) depth profiles determined from profiling floats (profiles are shown for before mixing, during mixing, and up to three profiles after mixing—blue, gray, orange, red, and dark red solid lines, respectively) from float 572. Heavy dashed line (‐ ‐ ‐) is the mixing depth and dotted line (• • •) is the euphotic depth, Z 0.1%. Carbon in the region from MLDpost to Z 0.1%post (dim sub‐mixed layer, shaded region) was summed both during mixing and post mixing to determine new production caused by deep mixing events (Table 5).

Fig. S6. Mixing‐ induced decrease in new production in the dim sub‐mixed layer. Temperature vs. salinity (a), chlorophyll (b), and C phyto (c) depth profiles determined from profiling floats (profiles are shown for before mixing, during mixing, and up to three profiles after mixing—blue, gray, orange, red, and dark red solid lines, respectively) from float 852. Heavy dashed line (‐ ‐ ‐) is the mixing depth and dotted line (• • •) is the euphotic depth, Z 0.1%. Carbon in the region from MLDpost to Z 0.1%post (dim sub‐mixed layer, shaded region) was summed both during mixing and post mixing to determine new production caused by deep mixing events (Table 5).

Fig. S7. Effect of different algal community compositions present immediately post mixing on the community composition three (light shades) and 9 d (dark shades) later. Cell densities are shown for diatom (D, blue) to green algal (G, orange) compositions of 0.5 : 1, 0.75 : 1, 1 : 1, 3 : 1, and 1 : 3. Data were calculated using initial post‐mixing concentrations of 103 cells L−1 and growth rates of 0.3 d−1 for the diatom and 0.2 d−1 for the green alga (taken from Fig. 3b), and growth initiated 2 d post mixing for the diatom and 3 d post mixing for the green alga.

Table S1. All mixing events included in new carbon production analysis.

Table S2. Modeled depth integrated net primary production (NPP, mg C m−2 d−1) for the euphotic zone (Z 0.1% NPP), mixed layer depth (MLD NPP), and dim sub‐mixed layer for the subarctic and subtropical regions in the North Atlantic Ocean during the stratified season from Fox et al. 2020. Standard error of the average shown in parentheses. aData for individual months were averaged to give values for the stratified season.

Click here for additional data file.