Stratification, nitrogen fixation, and cyanobacterial bloom stage regulate the planktonic food web structure

Abstract Changes in the complexity of planktonic food webs may be expected in future aquatic systems due to increases in sea surface temperature and an enhanced stratification of the water column. Under these conditions, the growth of unpalatable, filamentous, N2‐fixing cyanobacterial blooms, and their effect on planktonic food webs will become increasingly important. The planktonic food web structure in aquatic ecosystems at times of filamentous cyanobacterial blooms is currently unresolved, with discordant lines of evidence suggesting that herbivores dominate the mesozooplankton or that mesozooplankton organisms are mainly carnivorous. Here, we use a set of proxies derived from amino acid nitrogen stable isotopes from two mesozooplankton size fractions to identify changes in the nitrogen source and the planktonic food web structure across different microplankton communities. A transition from herbivory to carnivory in mesozooplankton between more eutrophic, near‐coastal sites and more oligotrophic, offshore sites was accompanied by an increasing diversity of microplankton communities with aging filamentous cyanobacterial blooms. Our analyses of 124 biotic and abiotic variables using multivariate statistics confirmed salinity as a major driver for the biomass distribution of non‐N2‐fixing microplankton species such as dinoflagellates. However, we provide strong evidence that stratification, N2 fixation, and the stage of the cyanobacterial blooms regulated much of the microplankton diversity and the mean trophic position and size of the metabolic nitrogen pool in mesozooplankton. Our empirical, macroscale data set consistently unifies contrasting results of the dominant feeding mode in mesozooplankton during blooms of unpalatable, filamentous, N2‐fixing cyanobacteria by identifying the at times important role of heterotrophic microbial food webs. Thus, carnivory, rather than herbivory, dominates in mesozooplankton during aging and decaying cyanobacterial blooms with hitherto uncharacterized consequences for the biogeochemical functions of mesozooplankton.


| INTRODUC TI ON
Approximately 93% of the excess heat energy trapped since the 1970s has been absorbed into the oceans, leading to a variety of changes in ocean conditions (Jewett & Romanou, 2017 and references therein). The most rapid warming of the mean sea surface temperature (>0.9°C) is observed in land-locked or semienclosed seas such as the Baltic Sea, North Sea, Black Sea, Japan Sea/East Sea, and East China Sea and over the Newfoundland-Labrador Shelf (Belkin, 2009). The large heat absorption alters the fundamental physical properties of the ocean by increasing the stratification of the water column (Jewett & Romanou, 2017 and references therein).
This process indirectly impacts chemical and biological properties such as changing nitrogen and carbon biogeochemical cycles that can result in increasing standing stocks of filamentous, N 2 -fixing cyanobacteria such as Trichodesmium (Gattuso et al., 2015;Jewett & Romanou, 2017;Paerl & Huisman, 2008;Roy et al., 2011). So far, we lack a consistent understanding of how stratification and blooms of unpalatable, N 2 -fixing, filamentous cyanobacteria affect the food web structure in aquatic ecosystems (Steinberg & Landry, 2017).
The number of trophic levels between autotrophs and mesozooplankton is a critical determinant to estimate the transfer of net primary production into the biological carbon pump and into higher trophic levels, such as fish, according to empirical data and ecosystem models (Jewett & Romanou, 2017;Peck et al., 2018;Steinberg & Landry, 2017). Direct grazing on filamentous, N 2 -fixing cyanobacteria by mesozooplankton organisms only sporadically takes place by some specialists (Engstroem et al., 2000;Loick-Wilde et al., 2012;O'Neil, 1999). The transfer of nutrients from unpalatable, filamentous, N 2 -fixing cyanobacterial genera such as Trichodesmium, Nodularia, or Aphanizomenon into mesozooplankton is thought to be mainly indirect through grazing on a microbial food web (summarized by Motwani, Duberg, Svedén, & Gorokhova, 2018;Wannicke, Korth, Liskow, & Voss, 2013). In these microbial food webs, auto-, mixo-, and heterotrophic microorganisms incorporate and grow on ammonium, amino acids (AA), or other dissolved organic nitrogen (DON) forms (Stal, Staal, & Villbrandt, 1999) that can be exuded in large quantities from filamentous cyanobacteria during N 2 fixation (Adam et al., 2016;Mulholland, Bernhardt, Heil, Bronk, & O'Neil, 2006;Ploug, 2008;Stal et al., 2003). The majority of studies suggest that mesozooplankton are herbivorous when filamentous, N 2 -fixing cyanobacteria dominate a microplankton community because the animals rely on co-occurring autotrophic phytoplankton species (Hannides, Popp, Landry, & Graham, 2009;McClelland, Holl, & Montoya, 2003;Mompeán, Bode, Gier, & McCarthy, 2016). All of these trophic studies used nitrogen stable isotope ratios in AAs to identify the mean trophic position (TP) of mixed field mesozooplankton samples without experimental manipulations (Chikaraishi et al., 2009;McClelland & Montoya, 2002;Steffan et al., 2015). Only recently, a study showed that mesozooplankton at times can also be carnivorous during a filamentous, N 2 -fixing cyanobacterial bloom according to TP estimates . Different from other studies this bloom was in a very late, decayed stage but with surprisingly fast amino acid turnover during N 2 fixation by the remaining intact cells .
The TP approach is based on the stable nitrogen isotope composition (δ 15 N) of the AAs phenylalanine (Phe) and glutamic acid (Glu) but sometimes also on phenylalanine (Phe) and alanine (Ala, reviewed by Ohkouchi et al., 2017). The δ 15 N of Phe is determined during its synthesis by autotrophs from an inorganic nitrogen source, and its isotopy remains virtually unaltered across trophic levels because no carbon-nitrogen atomic bonds are cleaved during the metabolism of essential Phe in heterotrophs (Chikaraishi et al., 2009;McClelland & Montoya, 2002;Steffan et al., 2015). The δ 15 N-Phe in heterotrophs such as mesozooplankton is therefore a proxy for the dominant inorganic nitrogen source (e.g., nitrate or N 2 ) in an aquatic food web McClelland & Montoya, 2002;McMahon, McCarthy, Sherwood, Larsen, & Guilderson, 2015;Sherwood, Lehmann, Schubert, Scott, & McCarthy, 2011). Glu and Ala belong to a group of AAs called "trophic AAs." Their isotopic signatures undergo large isotopic fractionation during the incorporation of a diet by a consumer (Montoya, Carpenter, & Capone, 2002;Steffan et al., 2015). In combination, Phe and Glu are the pair most often used for the stepless determination of the mean TP (TP Glu/Phe ) of a mixed field plankton sample without experimental manipulations (Chikaraishi et al., 2009;McClelland & Montoya, 2002;Ohkouchi et al., 2017), while the mean TP based on the δ 15 N-Ala and δ 15 N-Phe (TP Ala/Phe ) was recently suggested to better resolve the protist level (Décima, Landry, Bradley, & Fogel, 2017).
The Baltic Sea is one of the largest brackish ecosystems in the world and overall not a nutrient-poor sea; rather, it experiences over fertilization (Conley, 2012;Voss et al., 2011). However, in midsummer every year, its dissolved nitrate and phosphorus pools are largely depleted, and during this time, massive blooms of unpalatable, N 2 -fixing, filamentous cyanobacteria, namely, Nodularia and Aphanizomenon, occur (Kahru, Elmgren, Di Lorenzo, & Savchuk, 2018;Karlson et al., 2015;Wasmund, 1997). In addition, the Baltic Sea faces warming already today (Belkin, 2009;Kahru, Elmgren, & Savchuk, 2016;Suikkanen et al., 2013). This scenario leads to an increased summer surface temperature and stratification, which are among the primary variables that regulate variations in the intensity and occurrence of Nodularia and Aphanizomenon blooms in the offshore waters of the central Baltic Sea (Kahru & Elmgren, 2014;Kononen et al., 1996;Vahtera et al., 2007;Wasmund, 1997). Despite large structural (e.g., average water depths of only 58 m) and functional (e.g., low rate of exchange with North Atlantic waters and predominantly brackish conditions) differences between the Baltic Sea compared to marine coastal and offshore ecosystems, these regular and extensive cyanobacterial blooms offer the opportunity to develop a more mechanistic understanding of the effect of stratification and unpalatable, N 2 -fixing, filamentous cyanobacterial blooms on the number of trophic levels between autotrophs and mesozooplankton (Reusch et al., 2018).
In this study, we empirically identified how the trophic structure in planktonic food webs changed along with different abiotic and biotic factors across the Baltic Sea in summer. We determined the impact of N 2 fixation and the trophic structure of planktonic food webs according to two AA nitrogen stable isotope-based biogeochemical proxies (δ 15 N-Phe and TP Glu/Phe ) from two mesozooplankton size fraction samples from stations across the Baltic Sea. The proxies were connected to the results from Eglite et al. (2018) and Loick-Wilde et al. (2018) and to a large set of environmental variables, for example, mixed layer depth, nutrient concentrations, N 2 fixation rates (in parts from Loick-Wilde et al., 2018), microplankton cell carbon concentrations and biodiversity, and the sizes of the bulk and metabolic nitrogen pools in mesozooplankton. This process allowed us to conceptualize how the mean TP Glu/Phe of mesozooplankton can change along with the stratification, N 2 fixation, and filamentous cyanobacterial bloom stage in an aquatic ecosystem.

| MATERIAL S AND ME THODS
Samples were collected on a transect including a total of 59 hydrographical stations (Stas.) across the Baltic Sea in July and August 2015 during cruise M117 on board the RV Meteor. Fourteen stations were sampled for biogeochemical variables ( Figure 1). Hydrographic data and water samples for chlorophyll a (Chl. a) and nutrient determinations were obtained by deployment of a Seabird SBE-911 plus CTD equipped with oxygen and fluorescence sensors and mounted on a rosette sampler with thirteen 5-L GO-FLO bottles (Hydro-Bios, Kiel, Germany). Nutrient samples were analyzed directly on board according to Rhode and Nehring (1979) and Hansen and Koroleff (2007).
Additional samples of particulate organic matter (POM) were taken from the Chl. a maximum by filtering 0.5-1.0 L of seawater through precombusted Whatman GF/F filters (0.7 μm pore size, 25 mm in diameter) for elemental (total carbon and total nitrogen) analyses. All filters were stored at −20°C after shock-freezing in liquid nitrogen (−196°C). Additionally, 200-500 mL samples of surface water (from depths of 1 m, 5 m, 10 m, 15 m, and 20 m) were filtered for Chl. a concentrations on glass-fiber filters (Whatman GF/F) that were shockfrozen in liquid N 2 and stored at −80°C. The 96% ethanol extracts were used for fluorometric analysis according to the guidelines of HELCOM (2017). Nano-and microplankton samples (summarized as microplankton) were collected from an integrated depth between 0 and 10 m at 12 stations across the Baltic Sea, including two upwelling sites (Table 1). At Stas. TF12, TF360, TF109, TF113, TF213, TF259, and TF271, additional microplankton samples from a 20 m depth were taken. All microplankton samples were fixed with acid Lugol's solution, and at least 500 counting units were counted in sedimentation chambers under the inverted microscope as described in the manual of HELCOM (2017). Microplankton organisms were assigned to genera (species, if possible) and size classes, for which the specific volume was estimated as shown by Olenina et al. (2006). From these biovolume estimates, the carbon content of the taxa (in the following cell carbon biomass) was calculated after Menden-Deuer and Lessard (2000), as recommended by HELCOM (2017). Microplankton included autotrophic, mixotrophic, and heterotrophic species. N 2 fixation rates into particulate organic nitrogen were measured according to Montoya, Voss, Kaehler, and Capone (1996)  Zooplankton were collected by vertical tows through surface waters using a UNESCO WP-2 net (0.25 m 2 mouth opening) fitted with a 100 μm mesh size at 14 stations (Table 2). Once on deck, the mesozooplankton were concentrated in a light trap for 3 h before further processing to avoid sample contamination with phytoplankton cells. Then, the samples were divided into 100-300 μm and >300 μm size groups by sieving the samples through a 300 μm nylon mesh and stored in the manner described for the aforementioned filters. Plankton samples were then analyzed in the laboratory for total organic nitrogen (TN) and carbon (TC), and AA composition, as well as for bulk and compound-specific isotope ratios.  . Individual concentrations and nitrogen stable isotopes of 11 AAs included the so-called "source" AAs glycine -Gly-, lysine -Lys-, and phenylalanine -Phe-; the "trophic" AAs alanine -Ala-, aspartic acid -Asp-, glutamic acid -Glu-, isoleucine -Ile-, leucine -Leu-, proline -Pro-, and valine -Val-; and the "metabolic" AA threonine -Thr-, categorized by Germain, Koch, Harvey, and McCarthy (2013), McClelland and Montoya (2002), and Chikaraishi et al. (2009) according to the sensitivity of each AA to trophic enrichment in 15 N. It should be noted that Asp and Glu also include the amide forms asparagine and glutamine, respectively, with the N isotopic signature coming only from the α-amino-N from both compounds as the amide N is lost during hydrolysis. Details of the elemental analysis-isotope ratio mass spectrometry (EA-IRMS), gas chromatography-mass spectrometry (GC-MS), and gas chromatography-combustion-isotope ratio mass spectrometry (GC-C-IRMS) analyses can be found in Eglite et al. (2018) and in the Supporting Material S1.

| Proxies derived from amino acid analyses
The δ 15 N values of Phe from the two mesozooplankton size fractions were used as time-integrating proxies for the dominant inorganic nitrogen source sustaining the different planktonic food webs across the Baltic Sea McClelland & Montoya, 2002;McMahon et al., 2015;Sherwood et al., 2011). The proportion of Phe from N 2 fixation in mesozooplankton can be estimated according to a modified two-source mixing model after Montoya et al. (2002). For this model, it is assumed that the δ 15 N-Phe value of the mesozooplankton is a mixture of the δ 15 N TA B L E 1 Hydrographic characteristics, nutrient and Chl. a concentrations, and microplankton biodiversity indices (H': Shannon-Wiener's diversity; J 0 : Pielou's evenness; and d: Margalef's species richness) found in the surface mixed layer (ML) in July/August 2015. Dashes stand for no data. Chl. a from fluorometrical data from discrete measurements are given in mg m −3 and from the water column sensor in relative units, and both variables were significantly correlated (Supporting Figure S1)   The justification for this is that subsurface nitrate has the same δ 15 N value of 3.5 ‰ in both ecosystems (Casciotti, Trull, Glover, & Davies, 2008;Korth, Deutsch, Frey, Moros, & Voss, 2014). The TDF represents the trophic discrimination factor between Glu and Phe (7.6 ± 1.2‰) at each trophic shift between a consumer and its diet, and λ represents the first trophic level (=1) of the food web.  (Table 2) and coupled ΣV and TP values (Ohkouchi et al., 2017) were found (Supporting Figure S3). Therefore, we proceeded with the TP Glu/Phe estimates (in short TP) for both mesozooplankton size fractions.
The sum of nitrogen in AAs was used as a measure of the metabolic nitrogen pool size in mesozooplankton according to Mayzaud and Conover (1988). The metabolic nitrogen pool is a critical measure to estimate the availability of AAs as precursors for other macromolecules, such as nucleotides or fatty acids, or as an energy pool in mesozooplankton (Mayzaud & Conover, 1988

| Estimation of mixed layer depth
For the mixed layer depth (MLD) estimation, the density anom- Sea. Therefore, a slightly modified difference threshold method was applied. The mean σ T of the upper layer was calculated stepwise down to depth z d , and the difference Δσ T (z d ) to σ T (z d ) was derived: Starting at the surface, the MLD was reached when Δσ T (z d ) exceeded a threshold of 0.1 g kg −1 + the standard deviation of σ T in the range 0 to z d (mixed layer). The results were verified by visual inspection of the density profiles.

| Statistical analyses
A cluster analysis was used to identify cell carbon-specific communities of microplankton taxa based on samples from the upper 10 m from 12 stations across the Baltic Sea, including two upwelling areas. A similarity (SIMPER) analysis additionally identified which taxa shaped the respective communities. Cluster analyses including the available microplankton data from 20 m depth have also been tested but did not show significant differences to the more consistent data set from the upper 10 m (Supporting Figure   S2).
cyanobacteria (Clarke & Warwick, 2001 and TP 300 ) and the proxies for the dominant inorganic nitrogen source (δ 15 N-Phe 100 and δ 15 N-Phe 300 ) of the two mesozooplankton size fractions, which were of special interest for this study.
All biodiversity indices, community similarities, and the PCA were analyzed using PRIMER-6 Software (Primer-E Ltd., UK).
The data sets of N 2 fixation rates from three upwelling stations

| Microplankton communities across the Baltic Sea
The cluster analysis based on the microplankton cell carbon concentrations in the upper 10 m (Supporting Data SD1_1) identified four different microplankton clusters across the Baltic Sea in summer (Supporting Figure S2)

| Nitrogen source and trophic position proxies in mesozooplankton
The According to equation 1, essential Phe that received its nitrogen from N 2 fixation (diazotroph Phe) contributed up to 84% and up to 65% of the Phe pool in the small and large mesozooplankton size fraction samples, respectively, in the central Baltic Sea ( Figure 3B).
The TPs in the zooplankton communities ranged from 1.8 to 2.9 and covered the range from predominantly herbivorous to predominantly carnivorous communities across the Baltic Sea with distinct regional patterns (Figure 4).

| Statistical analyses to identify the environmental controls on the food web structure
A PCA was used to identify the environmental controls on the planktonic food web structure in the Baltic Sea during blooms of unpalatable, N 2 -fixing, filamentous cyanobacteria. A total of four PCs were necessary to cover all abiotic variables from While there is substantial variation explained by PC 3 and PC 4 (together explaining 23.9% of the variability), here we focused on PC 1 and PC 2 (together explaining 48.5% of the variability) to identify environmental controls specifically on the planktonic food web structure and its nitrogen supply across the Baltic Sea ( Figure 5).
The reason for this is that already PC 2 comprised both the mean trophic position and inorganic nitrogen source proxies for the two

| Stratification, nitrogen sources, and microplankton biomass and diversity
Salinity determined the distribution of many non-N 2 -fixing microplankton species with the most profound impact on dinoflagellate biomass. However, stratification regulated much of the availability of new nitrogen sources, the biomass of phytoplankton key species such as N 2 -fixing Nodularia and non-N 2 -fixing unicellular diversity in microbial food webs (Hewson et al., 2009;Loick-Wilde et al., 2017;Sheridan, Steinberg, & Kling, 2002). Accordingly, higher microplankton diversities were found at the central Baltic Sea stations compared to that in most of the near-coastal stations that largely lacked N 2 -fixing, filamentous cyanobacteria, even when salinity was higher at the near-coastal stations (Table 1).
Interestingly, microplankton diversity further increased from the Arkona toward the Bornholm and southern and eastern Gotland Basins (Table 1) namely of Nodularia, in the Baltic Sea has earlier been observed (Kanoshina, Lips, & Leppänen, 2003;Wasmund, 1997). However, in contrast to many diatoms, senescent unpalatable filamentous cyanobacteria are positively buoyant as long as their gas vacuole is intact (Sellner, 1997), which results in the degradation of their biomass components, such as AAs in surface waters, namely, in the highly decayed Nodularia bloom in the eastern Gotland Basin  in the Baltic Sea and from which the diazotroph nitrogen can be directly incorporated by mesozooplankton species (Motwani et al., 2018;Stal et al., 2003). Three lines of evidence support the idea that generally, this trophic relationship was also responsible for the incorporation of diazotroph nitrogen into mesozooplankton during our study. First, the group of Nostocales, namely, N 2 -fixing, unpalatable, filamentous Nodularia and Aphanizomenon, were responsible for the high N 2 fixation rates in the microplankton community in rather, spatial and zooplankton size-specific differences occurred.

| Planktonic food web structure across the Baltic Sea in summer
In the Arkona Basin (Stas. TF113 and TF109) and Bornholm Basin (Sta. TF213), herbivorous TPs in animals from the smaller mesozooplankton-sized fractions indicate that smaller animals must have intensively grazed directly on co-occurring autotrophs such as unicellular cyanobacteria during the younger N 2 -fixing cyanobacterial blooms. In contrast, TPs of 2.9 and 2.5 for small mesozooplankton in the southern (Sta. TF259) and eastern Gotland Basins (Sta. TF271), respectively, indicate that smaller animals must have predominantly preyed on mixo-and heterotrophic microorganisms in the more decayed cyanobacterial blooms (Figure 2). Additionally, larger animals showed a clear increase in their mean trophic position with aging cyanobacterial blooms according to a predominantly omnivorous feeding behavior in the Arkona Basin (Sta. TF113) and toward a predominantly carnivorous feeding behavior in the eastern Gotland Basin. In the southern and eastern Gotland Basins, carnivorous animals from both size fractions must have intensively preyed on diverse microbial biocenoses of bacteria, unicellular cyanobacteria, and mixo-and heterotrophic flagellates that had developed in association with the positively buoyant, aging or decaying Nodularia colonies ( Figure 6).
A common theme from bulk carbon and nitrogen pools in mesozooplankton was that larger animals rather than smaller animals in general were better able to keep body stoichiometry constant across the salinity gradient. The underlying causes for the variation in the C:N ratios with salinity in the smaller animals are potentially manifold (Steinberg & Saba, 2008) and may include changes in the mesozooplankton community structure (Walve & Larsson, 1999), changes in the lipid pools (Gismervik, 1997), or changes in ammonium excretion and growth efficiency (Checkley & Entzeroth, 1985;Koski, 1999;Walve & Larsson, 1999). In contrast, the metabolic nitrogen pool in mesozooplankton did not change with salinity but primarily with the biomass of N 2 -fixing Nostocales outside the predominantly decayed bloom in the eastern Gotland Basin. An increase in the metabolic nitrogen pools of mesozooplankton due to the intensive incorporation of diazotroph amino acids from N. spumigena at constant C:N ratios has previously been observed in field mesozooplankton from the central Baltic Sea (Loick-Wilde et al., 2012). This supports the idea that the amino acid nitrogen pool is a sensitive measure to quantify the impact of unpalatable, filamentous, N 2 -fixing cyanobacterial blooms on mesozooplankton physiology. In the decayed bloom, the size of the metabolic nitrogen pools in carnivorous mesozooplankton was more similar to the small pools in the near-coastal herbivorous animals. Thus, we can only speculate that differences in the turnover times of AAs associated with food quality were responsible for the relatively small metabolic nitrogen pools in mesozooplankton, which deserves further investigation.
Our data support the prevailing view that direct grazing of mesozooplankton on palatable phytoplankton species that co-occur with the unpalatable, filamentous, N 2 -fixing species dominates in a highly stratified, nitrate-depleted water column. However, in association with the increasing decay of blooms of unpalatable N 2 -fixing cyanobacteria, carnivory became the dominant feeding mode in mesozooplankton, which stresses the at times important role of heterotrophic microbial food webs for the planktonic food web structure. In the next step, empirical data of the mean trophic position of mesozooplankton and the dominant new nitrogen source for biological production will be used to calibrate and validate current biogeochemical models, including end-to-end models (e.g., physics to fish to human sectors, Peck et al., 2018).
Finally, an increase in sea surface stratification linked to sea surface warming, which tends to slow the nutrient supply to the surface, is projected for future oceans (Gattuso et al., 2015;Roy et al., 2011). Under these conditions, unpalatable N 2 -fixing Trichodesmium are especially favored, and their large blooms are expected to further expand under future global warming scenarios because enhanced N 2 fixation was found to persist under high CO 2 irrespective of phosphorus limitation Walworth et al., 2018). Analogously to Nodularia blooms in the Baltic Sea (Wasmund, Nausch, & Voss, 2012), blooms of Trichodesmium in other marine systems develop seasonally in association with mixing of DIN-depleted but phosphorus-rich upwelling waters into warm, stratified oceanic waters that contain seed populations of Trichodesmium (Deutsch, Sarmiento, Sigman, Gruber, & Dunne, 2007;Hegde et al., 2008;Hood, Coles, & Capone, 2004). Further, tropical and subtropical Trichodesmium blooms also decay in surface waters, which is namely triggered by the deepening of the mixed layer depth (Devassy, Bhattathiri, & Qasim, 1979;Hood et al., 2004) as found for Nodularia blooms (Kanoshina et al., 2003;Wasmund, 1997). It is a testable hypothesis that the observed changes in the planktonic food web structure in the Baltic Sea also take place in decay- ing Trichodesmium blooms in other marine systems, which would imply that a carnivorous feeding behavior in mesozooplankton can become more common under future global warming scenarios.

ACK N OWLED G EM ENTS
The authors wish to thank the captain and crew of the RV Meteor and the cruise leaders G. Nausch and O. Wurl for their invaluable F I G U R E 6 Biocenoses associated with the highly decayed Nodularia bloom in the mixed layer of the eastern Gotland Basin, including (a) unidentified bacteria and (b) unicellular cyanobacteria.

(a) (b)
assistance at sea. Susanne Busch is highly acknowledged for the microplankton counts and pictures. Fruitful discussions with C. Schrum and T. Neumann helped to consider the modeling perspective. This project was funded by the German Research Foundation (DFG) grant LO 1820/ 4-1 to NLW.