In this study, we performed ocean simulations based on the offline version of the coupled physical/biogeochemical model NEMO-PISCES. NEMO (Nucleus for European Modelling of the Ocean) version 4.2 is a model of global ocean circulation comprised of three major components: the ocean dynamical code OPA , the sea-ice model SI3 , and the marine biogeochemical model PISCES (Pelagic Interaction Scheme for Carbon and Ecosystem Studies, , Fig. a).

The ocean dynamics simulated by NEMO is used as forcing to the PISCES model. PISCES simulates marine biological productivity, plankton dynamics and biogeochemical fluxes. The standard version (PISCES-STD) includes 24 prognostic variables with five nutrients (i.e. nitrate, silicate, phosphate, ammonium and iron) and four plankton compartments: two phytoplankton groups (diatoms and nanophytoplankton) and two zooplankton size-classes: microzooplankton and mesozooplankton. PISCES-STD integrates a detailed representation of the biogeochemical cycles of carbon, dissolved and particulate organic matter (with two size classes: sPOC for the carbon content of small organic particles (1–100 µm) and bPOC for the carbon content of big organic particles (100–5000 µm), total alkalinity and dissolved oxygen ). In PISCES-STD, phytoplankton growth is constrained by light availability, temperature, and nutrients (N, P, Fe, and Si) concentrations. Phytoplankton and small organic particles are consumed by both zooplankton groups and mesozooplankton additionally feed on microzooplankton and large particles. PISCES-STD considers mesozooplankton as a single PFT, where the flux-feeding mode is implicitly accounted for in addition to the explicit representation of suspension feeding : mesozooplankton are parametrized as a single population with a proportion of flux-feeders that is calculated as the ratio of flux-feeding to total mesozooplankton grazing and has a Holling type II functional response.

In this study, we chose to represent three PFTs for the mesozooplankton compartment, with distinct feeding strategies to differentiate active organisms from passive ones, while also differentiating suspension feeders from flux-feeders (Fig. ). Compared to PISCES-STD, we explicitly modelled a flux-feeding mesozooplankton compartment and further separated the suspension feeding mesozooplankton into two separate compartments: active cruise-feeders and passive ambush-feeders. Cruisers (also called cruise-feeders, CF) account for both cruise feeders sensus stricto and feeding-current feeders, though we do not explicitly distinguish between the two of them in our study, as their diets are assumed identical here . From this point on, we refer to them as cruisers–organisms that actively swim or generate feeding currents to encounter prey and mates, similar to calanoid copepods . This active behaviour increases predation risk but also enhances the likelihood of encountering prey items .

Ambushers (AF) are organisms that adopt a sit-and-wait strategy . They wait motionless for motile prey items to pass within their reach or they capture their prey directly colliding with them . Despite having a lower feeding efficiency and lower probabilities of finding mates, this strategy has the advantage of a much lower mortality rate (up to an order of magnitude; ) as well as lower metabolic expenses . In our study, we do not distinguish between active ambushers that capture their preys by active attacks , like Oithonid copepods and chaetognaths, and passive ambushers that passively capture their prey, like ctenophores or foraminifera.

Flux-feeders (FF) are predominantly passive organisms, such as pteropods (but they could also represent active feeders like copepods of the Temora and Oncaea genera), that feed on rapidly sinking organic particles . They inhabit the interface between the euphotic zone and deeper waters, acting as “gatekeepers” of the mesopelagic zone by regulating carbon transfer in the water column . This feeding strategy also contributes to lower mortality rates and higher growth efficiency.

In the new configuration developed in this study, called FOREFF (for FORaging EFFort), the three main feeding strategies of mesozooplankton are considered, each of them being represented by one PFT. Their dynamics follows Eq. (): ∂MX∂t=(1-σunass)×eMX×GMX×fMX(T)(1-Δ(O2))MX-rMXfMX(T)MXKm+MX+3Δ(O2)MX1-mMXfMX(T)(1-Δ(O2))MX∑XMX

In this equation, MX represents the mesozooplankton biomass of one of the three newly modelled feeding groups X (AF, CF, and FF) based on a Michaelis–Menten parameterization with no switching and a threshold, to avoid extinction of mesozooplankton at very low food concentration . The first right-hand term represents growth, where σunass is the non-assimilated fraction of ingested food, eMX is the growth efficiency, GMX represents the ingested matter by mesozooplankton, fMX(T) is the temperature dependence and Δ(O2) is an oxygen factor. A full description of the equations for GMX is provided in the Appendix . The second term represents mesozooplankton metabolic losses due to basal respiration and swimming activity, at a rate rMX and where Km is a half-saturation constant. The last term represents mortality by density-dependent processes such as predation and diseases, with the quadratic mortality coefficient mMX. Here we choose a formulation of quadratic mortality corresponding to predation by a generalist predator: the predation pressure on one group depends on the total mesozooplankton biomass. Consequently, the more advantageous a strategy is in a given region, the more it tends to outcompete and exclude alternative strategies. A full description of the parameters and their values is given in Table . Both suspension feeders feed indiscriminately on small living organisms and particulate marine snow, similar to the standard representation of mesozooplankton in . Only flux-feeders feed exclusively on particles, due to their feeding mode. All three terms have the same temperature dependence with a Q10 set to 2.14 and as we assume that mesozooplankton are unable to cope with anoxic waters, the growth rate and quadratic mortality are reduced and the metabolic losses are enhanced in oxygen depleted regions (Δ(O2), ).

In addition to the explicit representation of these three PFTs (Fig. ), the FOREFF configuration implements a non-dimensional foraging effort p for active organisms (i.e. cruisers). The foraging effort p varies between 0–1 and represents an optimization of the fitness via the fraction of time spent foraging. The parameter is adapted from and implemented in Eq. () (see Appendix  for more details). It is based on the assumption that ambushers have an invariant foraging effort due to their passive behaviour, while cruisers may modify their swimming activity in response to prey abundance to reduce the cost and risk of searching for prey items and optimize their fitness . The foraging effort of cruisers varies in response to prey density (see Fig.  for the theoretical curve) in order to maximize their fitness, balancing food intake, predation risk, and the metabolic cost of searching for food . Thus, at high prey densities, cruisers reduce their foraging effort to lower both predation risk and metabolic expenditure, while at intermediate prey densities, the foraging effort reaches its maximum value of 1. At low prey densities, the foraging effort decreases, implying that cruisers no longer swim or swim very little but do not have access to food, so they eventually die. Moreover, the foraging effort is set to zero when the prey concentration falls below a minimum threshold concentration Rmin (see Eq. ), as in . In our case, this threshold is 1.56 mmolm-3, which corresponds to the minimum prey concentration at which the energetic gain from foraging offsets the maintenance costs of cruisers.

To represent these feeding strategies and incorporate the foraging effort, model parameters (Table ) are adjusted as follows to reflect trade-offs between growth, reproduction, and survival .

Since cruisers swim continuously to encounter prey items, they face a higher predation risk than ambushers . Consequently, their quadratic mortality parameter mMCF is set three times higher than the one of ambushers (0.015 (µmolCL-1)-1d-1, , see Table ). The metabolic losses parameter rMCF of cruisers is also set higher (0.03 d-1), to account for the increased energetic expenses due to active feeding (Eq. ) introduced from . Additionally, we differentiate the maximum grazing rates for cruisers gmMCF and ambushers gmMAF, based on the data analysis from . Higher maximum grazing rates are assigned to cruisers than ambushers (0.8 and 0.2 d-1, respectively).

Five sensitivity experiments were conducted to investigate the relative impact of feeding strategies and the effect of foraging effort on ocean biogeochemistry and ecosystem functioning. A visual representation of the various configurations and sensitivity experiments (FOREFF, NO_FOREFF, LGE and KILL_XX experiments) can be found in the Appendix .

The first experiment (i.e. NO_FOREFF) is carried out to investigate the impact of foraging effort. NO_FOREFF is the same as FOREFF except that cruisers have a foraging effort set to a constant value of one.

The second experiment (i.e. Low Growth Efficiency, LGE) is less similar to the basic FOREFF model as it is based on a different set of hypotheses and does not include a variable foraging effort for cruisers. In LGE, parameters are adapted to differentiate active and passive feeding strategies through their efficiency at capturing prey items (see Table ). This experiment is set to study a different way to represent the trade-offs between the metabolic cost associated to swimming, grazing capacity and mortality from predation. In LGE, the feeding efficiency of cruisers is estimated to be three to ten times higher than the feeding efficiency of ambushers. Thus, their half-saturation constant for grazing (Kg) is set to 10 µmolCL-1 whereas that of ambushers is set to 30 µmolCL-1. Furthermore, active feeding is thought to increase metabolic losses by 15 %–25 % . We represent this process by decreasing the maximum growth efficiency eMCF of cruisers from 0.4 to 0.34 . The quadratic mortality mMCF for cruisers is set to be four times higher than for ambushers (0.02 (µmolCL-1)-1d-1, following ) to reflect higher predation risks inherent to cruise-feeding. Finally, we assume that the basal metabolism is the same for all groups, meaning that their respiration parameter rMX is identical.

The last three experiments (i.e. KILL_AF, KILL_CF and KILL_FF) include the foraging effort for cruisers, have similar parameters to FOREFF (see Table ) and are designed to eliminate one PFT, respectively ambushers, cruisers and flux-feeders, by setting their maximum grazing rate (or flux-feeding rate) to zero. This way, we are able to get more insights about the relative impact of each group on ecosystem dynamics and their contribution to the carbon cycle.

To characterize and compare the biogeography of the two suspension-feeding groups (cruisers and ambushers, ) across experiments, a dominance index is defined based on their biomass MX (Eq. ). This index is calculated at each time step and on every vertical level, then averaged over the year and the top 150 m. Positive values close to 1 indicate a dominance toward cruisers, negative values close to -1 signify a dominance toward ambushers, and values around 0 suggest a co-dominance of the two groups. 2Index of dominance =MCF-MAFMCF+MAF

To evaluate how mesozooplankton feeding strategies impact biogeochemical fluxes, we focused on carbon export. We investigated carbon export at 150 and 1000 m (respectively C150 and C1000), and calculated the efficiency of carbon transfer from 150 to 1000 m (Eq. ), which indicates how efficiently sinking organic matter is exported to the deep ocean. 3Carbon Transfer Efficiency =100×C1000C150

Simulations were run offline for 20 years using the coupled model NEMO-PISCES. The configuration and circulation are the same as in . We used the ORCA-2 global configuration of NEMO, which has a spatial horizontal resolution of 2° that increases to 0.5° latitudinal resolution at the equator. Along the vertical dimension, it has 31 vertical levels, with a thickness increasing from 10 m at the surface to 500 m at 5000 m. Nitrate, phosphate, silicate, and oxygen are initialized from the climatology of the World Ocean Atlas 2009 , DIC and alkalinity from GLODAP-v1 and iron and DOC are initialized from an existing quasi-steady state simulation .

The reference FOREFF simulation was evaluated against in situ data. To do so, we used the Biomass Distribution Models (BDM)-ensemble developed by in that estimates monthly fields of mesozooplankton biomass for the global epipelagic ocean. Data from the monthly climatology from MAREDAT (MARine Ecosystem DATa, ) re-gridded on the ORCA2 grid and integrated over the top 200 m was used to train the BDMs pipeline. Monthly satellite data from the Ocean Colour Climate Change Initiative project (OC-CCI, ) were re-gridded on the ORCA2 grid, and are used to evaluate the surface fields of chlorophyll a concentration. The modelled fields of mesozooplankton biomass were annually averaged and integrated over 200 m. For surface chlorophyll, a mask corresponding to the seasonal lack of data is applied to the modelled data.

Field-based estimates of global biomass are lacking for the three mesozooplankton PFTs . Therefore we have evaluated the quality of our PFT-specific fields against observations in a more indirect fashion. To do so, we used the global distribution maps of copepod functional groups published by . defined eleven functional groups (FGs) based on five species-level functional traits (i.e. body size, trophic group, feeding mode, myelination and spawning mode) and modelled the distribution of these groups across the global surface ocean based on field occurrences and species distribution models. The maps of estimate where the environment is most suitable for the copepod functional groups to be present or not (i.e. habitat suitability indices). They do not aim to represent actual biomass patterns, but they are useful to compare the biogeography of copepod PFTs based on in situ observations. Here, we focused on the copepod functional groups that best correspond to the suspension feeders (cruisers and ambushers) we modelled. The following functional groups (FG) of were used to evaluate the biogeography of our cruisers: FG1 (small, myelinated cruise-feeding herbivores), FG5 (medium size, current/cruise-feeding carnivores) and FG6 (large myelinated current-feeding herbivores). For ambushers, we considered: FG4 (small, amyelinated ambush-feeding carnivores), FG8 (small, amyelinated ambush/current-feeding carnivores) and FG10 (large, amyelinated ambush-feeding omnivores). For both cruisers and ambushers, we pooled together and summed the habitat suitability indices of their corresponding copepod functional groups and then calculated the dominance index following Eq. (). This way, we obtained a map of the dominance index that is comparable to the one based on our model projections for the global surface ocean.

Our modelled fields of mean annual total mesozooplankton biomass concentration and surface chlorophyll concentration are in line with observations (Fig. ). The Pearson correlation coefficient between observed and modelled mesozooplankton biomass concentration is equal to 0.49 (see Table ). Regions of high mesozooplankton biomass concentrations are correctly simulated although biomass is slightly overestimated compared to observations (Fig. a and c), which is indicated by a positive bias (+0.05 mmolm-3, Table ). The modelled mean annual mesozooplankton biomass concentration is coherent with previous studies, where higher concentrations are found in the subpolar regions such as in the Northern Atlantic and Pacific oceans .

Our model also reproduces the regions of high phytoplankton biomass (Pearson correlation coefficient =0.24) although it overestimates the concentration of surface chlorophyll as evidenced by a positive bias (+0.11 gChlm-3, Table ), especially in the Southern Ocean (Fig. b).

In addition to explicitly modelling cruisers, ambushers, and flux-feeders, the main novelty of our study is to model the foraging effort of cruisers which represents the effort invested into searching for prey items as a function of their availability. As active behaviours account for a higher predation risk, this foraging effort is also an asset to increase their overall fitness, while avoiding predators .

Figure a illustrates the foraging effort of cruisers averaged over the top 150 m. Consistent with Fig. , the foraging is zero in highly oligotrophic regions. This suggests that cruisers would decrease or cease their foraging and eventually die in the least productive regions, such as the subtropical gyres, due to insufficient prey availability to meet their metabolic needs. The impact on cruisers concentration has been shown on Fig. a, where ambushers completely outclass cruisers in regions of low productivity (see also Fig. a). The foraging effort peaks around one in regions with intermediate productivity to maximize ingestion and declines in areas of high prey concentrations. In very productive regions, the decrease in foraging effort suggests that cruisers are able to save energy and reduce their predation risk while benefiting from abundant prey concentrations.

The seasonal variation of the foraging effort over the year is presented through a map of its temporal variance (Fig. b). The highest seasonal variations are found at high latitudes and decrease towards the equator. High latitudes are characterized by strong seasonal variations in the prey concentration (for instance in the Southern Ocean, prey concentration varies from 3.33 mmolCm-3 during the seasonal bloom to 0.22 mmolm-3 in Austral winter), which leads consequently to important variations in the foraging effort of cruisers. During winter, when phytoplankton and microzooplankton concentrations are very low, the foraging effort decreases to zero. Conversely, during the favourable season, the foraging effort remains close to 1, except during the spring bloom when prey concentration may locally become sufficiently high to trigger its down regulation (Fig. ). Seasonal variations are smaller yet still important at the edges of the subtropical gyres. These variations are caused by their seasonal spatial contraction and expansion which leads prey abundance to fluctuate around the minimum prey concentration required to sustain a non-zero foraging effort (i.e. Rmin, see Eq. ). At the center of the subtropical gyres, prey concentration remains below Rmin all year long, resulting in a consistently null foraging effort and no seasonal variability. In the highly productive regions of the low latitudes, such as the eastern boundary upwelling systems, prey abundance remain always high all year long. Consequently, the foraging effort displays very modest variations in these productive regions, as evidenced in Fig. .

To investigate the seasonality of suspension feeders (cruisers and ambushers), we focus on the Southern Ocean (south of 60° S), where strong variations are observed, in accordance with the temporal variance of the foraging effort (Fig. b). This region, with its highly seasonal environment, is examined to explore the relationship between the biomass of suspension feeders and the foraging effort (Fig. b). Cruisers consistently dominate over ambushers in these regions throughout the year (Fig. a). Overall, we find that if one group of suspension feeders dominates in a region, it dominates all year long. Therefore, our model does not predict alternation between suspension feeders (see Appendix, Fig. ). In the Southern Ocean, both cruisers and their foraging effort exhibit a similar seasonal pattern (Fig. , blue and orange dots) peaking in March at 0.41 mmolCm-3 and 0.4 respectively. The favourable season is also characterized by a very large spatial variability of both the biomass and the foraging effort. During this season, prey concentration in the Southern Ocean increases (to reach a maximum of 3.3 mmolCm-3, with a minimum of 0.22 mmolCm-3 in Austral winter). This resulting larger pool of prey items combined with an enhanced foraging effort boosts the concentration of cruisers. The large spatial variability underscores the contrast between HNLC regions and highly productive areas near Antarctica, as well as downstream of islands and plateaus. Following the summer period, cruisers' biomass steadily declines, until it reaches its minimum in November (0.065 mmolCm-3, i.e. 84 % lower than its maximum summer value). Yet, cruisers remain four times more abundant than ambushers (0.016 mmolCm-3). Ambushers biomass remains low throughout the year (0.023 mmolCm-3) and presents weak seasonality (Fig. , in red). A small peak of ambushers (0.03 mmolm-3) occurs after the seasonal bloom, in April. Similar temporal patterns are observed at high Northern latitudes (>60° N, not shown here), with a peak of foraging effort and of cruisers' concentration in August, followed by a peak of ambushers later in October. At low latitudes (between 0–30° N/S, not shown here), seasonal variations are much lower (less than 20 % for ambushers) as these regions are mainly characterized by low productivity in the gyres throughout the year, except in Eastern Boundary Upwelling systems. Ambushers largely dominate over cruisers such that no successive dominance is observed in these regions as well (see Appendix, Fig. ).

In this section, we first compare the reference simulation FOREFF with two experiments with three PFTs: (i) NO_FOREFF, which is similar to FOREFF but with a constant foraging effort for cruisers (set to its maximum value of 1) and (ii) LGE (Low Growth Efficiency, for cruisers), in which both suspension feeders have the same maximum grazing rate, but cruisers are assigned a lower half-saturation constant for grazing and reduced growth efficiency. Next, we examine the impact of removing one of the feeding groups in FOREFF, hence reducing the representation of mesozooplankton from three to two PFTs.

Distinguishing three mesozooplankton feeding groups impacts the amount of carbon export at depth (Table ). In the reference FOREFF configuration, total carbon export is 5.01 GtCyr-1 at 150 m and 1.69 GtCyr-1 at 1000 m. The spatial pattern of the amount of carbon exported at depth is similar to the one obtained by , with highest export in productive regions and at high latitudes (Fig. a). These global carbon export values are within the range of recent independent studies , but they are lower than values found in previous PISCES-based model studies (6.9 GtCyr-1 at 150 m depth by , or 7.71 GtCyr-1 at 100 m depth by ). In our model, the global carbon transfer efficiency, defined as the carbon flux at 1000 m relative to the flux at 150 m, reaches a proportion of about 33.7 % (Table ).

In our reference FOREFF simulation, carbon transfer efficiency is maximum (up to 50 %) in the productive areas at high latitudes, intermediate (around 30 %–35 %) in regions of moderate productivity at mid and low latitudes, and minimum (less than 30 %) in highly oligotrophic regions (center of the gyres) and in the Eastern Boundary Upwelling Systems (less than 10 %). The latter corresponds to those regions where flux-feeders thrive (see flux-feeders concentration averaged between 150–1000 m in the Appendix Fig. a), as they are able to efficiently feed on abundant sinking particles, hence lowering the carbon export at depth (Fig. b).

Variations in carbon export across different model experiments are controlled by two main factors. First, the production of organic particles in the upper ocean, which partly depends on the relative contributions of the two suspension feeding modes. Suspension feeders influence production both directly, through differences in grazing intensity and mortality losses, and indirectly, by modulating microzooplankton biomass and primary productivity. Second, the fate of sinking organic particles, and thus the transfer efficiency, is affected by flux-feeders. According to our experiments, an increase in grazing by suspension feeding mesozooplankton leads to a higher export at 150 m (Table ). This result is expected since suspension feeding mesozooplankton are the main source of large organic particles through both fecal pellet production and mortality. Furthermore, cruisers appear to be more efficient at sustaining export than ambushers. This is demonstrated by the experiments NO_FOREFF and KILL_CF, both of which result in a significant reduction in export at 150 m of -6.41 % and -11 %, respectively (Table ). Spatially, the most substantial decreases in export in NO_FOREFF and KILL_CF occur at high latitudes, where a sharp decline in export (-48.8 % and -60 %, respectively) aligns with a significant reduction in cruisers abundance (Fig. c for NO_FOREFF and Fig. i in the Appendix for KILL_CF). In contrast, the variation of carbon export is much smaller when ambushers are eliminated (KILL_AF). Spatially, this corresponds to moderate increases in export in productive regions, balanced by moderate decreases in less productive areas (see Appendix, Fig. g).

In all experiments, absolute changes in average transfer efficiency remain relatively modest globally (less than 5.5 %), except when flux-feeders are eliminated. In the latter case, average transfer efficiency is strongly increased from 33.66 % to 45.96 % (KILL_FF, Table  and Fig. f). The KILL_FF experiment demonstrates the critical role played by flux-feeding on the fate of the particulate organic matter sinking down in the mesopelagic domain. Spatially, the impact of flux-feeders is maximum in productive regions such as upwelling systems and the high latitudes (Fig. f), where their abundance in the mesopelagic domain is high thanks to a higher concentration of organic particles exported from the upper ocean (see Appendix, Fig. f). In the other experiments, such as NO_FOREFF and KILL_CF, a decrease in flux-feeders concentration at depth generally leads to an increase in the transfer efficiency and vice versa (see Appendix, Figs.  and ).

Theoretical modelling studies on feeding strategies of zooplankton have shown various biogeographies for these organisms . For instance, the model proposed by focuses on encounter rates between zooplankton and their prey, which are controlled by trade-offs between motility, body size, and predation risk. Their model predicts a stronger competition among suspension feeders at high latitudes, where ambushers tend to dominate over cruisers. It is also the case in our Low Growth Efficiency (LGE) experiment, where ambushers dominate in productive regions and exclude cruisers (Fig. c). Similarly, and showed how trade-offs, specifically the net energy gain vs. predation , allow to predict the biogeography of these organisms. They suggest that more passive suspension feeders, such as ambushers, would dominate in regions characterized by higher prey levels, higher turbulence and higher predation risk. Conversely, cruisers would perform better at intermediate or lower food levels as well as at lower levels of turbulence . This leads to a pattern similar to our LGE experiment as well. These modelling studies suggest that ambushers would dominate on a global scale, particularly at high latitudes and in areas with high prey densities. This overall dominance of ambushers is a consistent finding across all our experiments (Fig. , top row, and Table ). However, the preference for a passive ambushing strategy in regions with higher prey concentrations is simulated only in our LGE experiment (Fig. c).

To our knowledge, the study by is the only modelling work that predicts a biogeography similar to that simulated in our experiments FOREFF and NO_FOREFF. This study examines the distribution of passive and active organisms alongside their body size, demonstrating that small passive-feeding organisms tend to dominate in low-productivity environments, whereas large active-feeding organisms prevail in more productive systems. Consistent with our findings in FOREFF and NO_FOREFF, the study also shows that active feeding strategies are entirely eliminated under low food availability. However, a key distinction from our work and the work of is their explicit representation of the mesozooplankton size distribution, which plays a crucial role in shaping the predicted biogeography. Among small organisms, passive feeding consistently emerges as the dominant strategy, explaining its prevalence in low-productivity environments. By contrast, large organisms, which thrive in highly productive systems, are hypothesized to be exclusively active feeders due to their high sinking speeds, rendering a passive strategy unviable.

Our predicted biogeography (Fig. ) can also be compared to recent observational studies, particularly the global surface distribution of copepod functional groups established by . This study integrates species-level occurrence observations, species distribution modelling and a species-level functional trait data, providing an empirical biogeographical perspective on suspension feeders. In , ambushers and cruisers are classified into three distinct functional groups each (see Sect. ). They observe a spatial pattern of active and passive organisms similar to that predicted in our FOREFF experiment. Their co-dominance index (Fig. ) suggests that, as in FOREFF, active organisms dominate over passive ones at high latitudes and in highly productive regions such as the Eastern Boundary Upwelling Systems. However, the observed biogeography exhibits greater co-dominance than our model predictions, as indicated by co-dominance index values closer to zero (Fig. ). Notably, in the oligotrophic subtropical gyres, active feeders are not completely absent in contrast with what we found in FOREFF.

However, caution is required when comparing the observation-based biogeography from to our modelling results. Their study is based on presence data and habitat suitability indices estimated from species distribution models, and hence it does not consider biomass which is the property simulated by our model. The fact that their approach is based on presence and habitat suitability rather than biomass, may introduce bias and underestimate the relative proportions of the different groups.

Cruise-feeding and ambush-feeding mesozooplankton display distinct spatial biomass distribution but similar vertical profiles. Both groups are more concentrated in surface layers, whereas flux-feeders are found in the deeper part of the euphotic zone and prevail in the mesopelagic and bathypelagic domains. In the euphotic zone, ambushers are found everywhere and dominate over cruisers except in productive regions and at high latitudes, as mentioned above (see Sect. ). The different experiments showed a strong sensitivity of the biogeography of the suspension feeding groups to the assumptions made on the trade-offs between the energy obtained from feeding and invested into competing functions such as growth, reproduction and survival .

In the reference configuration of our study (i.e. FOREFF), adaptive behaviour is incorporated using the theoretical framework proposed by . Comparing this reference setup to the NO_FOREFF sensitivity experiment provides insights into the effects of variable adaptive foraging effort. In productive regions with high prey concentrations, such as low-latitude upwelling systems and high latitudes, reduced foraging effort explains the dominance of cruisers over ambushers (Figs. a and a), as minimizing predation losses becomes more critical than maximizing energy gain. When food levels are low (<1.5 mmolCm-3), cruisers stop feeding because saving energy becomes critical. In the subtropical gyres, cruise feeders are outcompeted by ambushers and even completely eliminated when foraging effort is kept to its maximum value, because food availability is never sufficient to sustain their metabolic needs. At high latitudes, reduced metabolic expenditure resulting from ceasing foraging enables cruisers to better endure the winter and, hence, maintains a sufficient population to outcompete ambushers when preys becomes abundant in the spring (Fig. ). This is evident from the sharp decline in their abundance predicted in the NO_FOREFF experiment (Fig. d). The ability of mesozooplankton to adjust their foraging effort thus plays a crucial role in their success in seasonally productive regions, such as high latitudes and low-latitude upwelling systems. However, even in regions where active feeders dominate, ambushers are never entirely excluded (see Appendix, Fig. b). Thus, ambush feeding remains a viable predation strategy across all regions of the upper ocean, unlike active feeding modes, as already shown in previous studies .

In the LGE experiment, we assumed different hypotheses than in the reference experiment FOREFF. While keeping the same maximum grazing rate for both suspension feeding groups (cruisers and ambushers), we assigned to cruisers a lower half-saturation constant for grazing to reflect their superior foraging efficiency and a lower gross growth efficiency to represent their higher metabolic needs relative to ambushers . Under these assumptions and the parameters values prescribed in the LGE experiment, cruisers are outcompeted by ambushers in productive regions and high latitudes (see Table  and Appendix, Figs. c and c). This outcome is driven by the fact that cruisers experience a predation mortality rate four times higher than that of ambushers, requiring them to assimilate at least four times more food to remain competitive. Yet, with a lower gross growth efficiency and a half-saturation constant only three time lower, such assimilation level remains unachievable.

Our various configurations implement a common set of trade-offs related to feeding modes in different manners: active organisms are more efficient foragers and reproducers but experience a greater predation risk and higher metabolic losses. So far, there are still too little experimental data enabling us to quantitatively constrain these trade-offs accurately. Furthermore, previous theoretical and laboratory studies provide a broad range of possible parameter values for representing mesozooplankton feeding modes, adding to the challenge of accurately constraining these dynamics . This challenge is reflected through important variations in our experiments, such as the spatial and temporal repartition of the suspension feeding groups and the impact on carbon export. This strong sensitivity of zooplankton and its role in plankton ecosystem dynamics and carbon cycle has been previously evidenced .

Two parameters were particularly important in our modelling experiments: the maximum grazing rate of suspension feeders (cruisers and ambushers) gMSF and the quadratic mortality rate mMX. A higher maximum grazing rate for cruisers was required to reproduce a biogeography where they dominate in high-latitude and highly productive regions, but are outcompeted in low-productivity areas. A similar biogeography was found by who made the same assumption. When similar maximum grazing rates are prescribed, cruisers are generally outcompeted except at low food and turbulence levels as found in previous modelling studies . The assumed excess in predation risk due to an active feeding mode, i.e. the value of mMX, is also a key parameter. When maximum grazing rates are identical for both cruisers and ambushers, a high excess risk leads to an exclusion of the former in highly productive regions and at high latitudes (as in our LGE experiment), whereas a weak excess risk results in a domination by cruisers everywhere in the surface ocean.

We find flux-feeding mesozooplankton to be more abundant below the euphotic zone where they outcompete the suspension feeding modes (Fig. , ). This is not surprising since the main source of energy for mesozooplankton in the mesopelagic layer is the flux of sinking organic particles, making flux-feeding the most advantageous mesozooplankton feeding strategy. In particular, they prefer large, rapidly sinking particles, as the particle flux constrains the probability of feeding and flux-feeders would clear large particles more efficiently than smaller ones, that sink more slowly . In the interior of the ocean, the abundance and vertical distribution of flux-feeders closely align with the flux of particles. Their abundance peaks in highly productive areas and declines with depth, mirroring the particle flux, which is itself influenced by flux-feeders. The depth at which flux-feeders become dominant depends on the euphotic depth and therefore, on surface productivity (the euphotic depth being shallower in productive zones and deeper in oligotrophic regions, ). As a result, in the top 150 m (see Appendix, Fig. , right panels), the vertically integrated biomass of flux-feeders is comparable to that of suspension feeders in the highly productive regions where the euphotic depth is shallow and the flux of particles elevated .

showed that suspension feeders do not significantly affect carbon export at depth due to insufficient clearance rates. In contrast, flux-feeders play a major role in regulating deep-sea carbon export, influencing both the vertical attenuation and the overall magnitude of particle flux . In our study, we show that carbon export and transfer efficiency are strongly influenced by flux-feeders. This is particularly true in highly productive areas such as the Eastern Boundary Upwelling Systems, where flux-feeders are very abundant (see Appendix, Fig. , right panels) and where the carbon efficiency is minimal (Fig. b). Our different experiments show that an increase in flux-feeders abundance decreases carbon efficiency and vice versa, which is especially evident in the experiment where flux-feeders are removed (KILL_FF). This experiment simulates the highest carbon transfer efficiency values (Fig. f) due to the absence of flux-feeders' grazing on particles. It thus highlights the key role that these organisms play in the water column, in particular in highly productive regions: they decrease the efficiency of carbon export, increase the remineralization of particles in the upper mesopelagic zone and thus favour productivity in the upper ocean.

Another key process affecting the fate of organic particles in the ocean interior is their degradation by heterotrophic bacteria . recently showed that flux-feeders have a greater influence on flux attenuation than bacteria in the upper mesopelagic zone, as bacterial degradation accounts for only 7 %–29 % of flux attenuation. We also compare the remineralization of particulate organic carbon by bacterial activity and by flux-feeder grazing (not shown here; see Appendix, Fig. ). Our results indicate that, on a global scale and between 150–1000 m, bacterial activity has a greater impact than flux-feeder grazing (see Appendix, Fig. a and b). However, regionally, the dominance index between flux-feeders and bacterial activity reveals a stronger influence of flux-feeders in coastal and highly productive regions (see Appendix, Fig. c), with a tendency towards co-dominance in regions of intermediate productivity and at high latitudes. This highlights that, in areas where flux-feeders are abundant within the 150–1000 m depth range, their activity surpasses bacterial activity, underlining the key role flux-feeders play in the carbon cycle.

As in any theoretical modelling exercise, our results strongly rely on our hypotheses and parameter choices. Even though this study was designed to investigate the impact of mesozooplankton functional diversity on ecosystem dynamics and carbon fluxes through their feeding modes, the mesozooplankton compartment was only expanded to three coarse feeding strategies. However, observations reveal greater diversity in these feeding modes. Among cruise-feeding zooplankton, some organisms generate feeding currents such as the copepod Temora longicornis, to either filter prey items from the current or capture them when entrapped . Others swim actively, such as the copepod Centropages hamatus and employ raptorial strategies upon detecting prey items using chemotactic, rheotactic, or visual cues, which influence both detection efficiency and dietary preferences. Similarly, within ambushers, distinguished between passive ambush feeders that encounter and capture prey items passively such as Oithona nana or Acartia tonsa copepods and active ambush feeders that actively attack their prey such as ciliates of the Mesodinium genus . This wide diversity in the foraging techniques and detection modes controls the feeding efficiency and the types of prey that are ingested by mesozooplankton, a diversity that is only crudely represented in our modelling framework. Thus, our model and experiments underestimate the diversity of feeding strategies by considering only three main groups.

This large diversity in the feeding strategies and their success finds its source in the foraging but also in the defence trade-offs which reflect the fundamental dilemma between eating and being eaten . Here, we use a simple representation of these trade-offs between gains (ingestion) and losses (metabolic costs and predatory risk) from a specific set of hypotheses solely based on the foraging activity. Yet, other factors modulate the behavioural strategy of the zooplanktonic organisms. For instance, mate seeking leads to a systematic higher mortality by predation in males, particularly among ambushers, such as Oithona copepods . Additionally, zooplankton can mitigate their mortality risk through defensive adaptations such as chemical signalling and morphological changes to reduce ingestion likelihood, as well as avoidance and escape behaviours – like seeking spatial refuges, performing diurnal vertical migrations or forming swarms for euphausiids – to decrease predators' chances of successful encounters and captures . For instance, diurnal vertical migration performed by active feeders such as Calanoid copepods, is an efficient defence strategy and may reduce lateral transport that may contribute to their success in very productive upwelling systems .

In our study, we assumed that mesozooplankton are not able to switch from one feeding strategy to another. In other words, organisms are assigned an obligate feeding strategy which cannot change depending on the abiotic or biotic conditions. Yet, some zooplankton taxa, such as Acarcia tonsa copepods , can switch from one to another predation mode, for instance as a function of prey abundance (as it is the case for A. tonsa) and type . Integrating this flexibility in our model might change the dynamics of the zooplankton community, notably inducing more co-dominance of the predation modes since there would be a very rapid switch in occurrence of the feeding mode, whereas, in our approach, a change in the dominance by a feeding mode is only achieved through a change in the relative abundance of specialized functional groups. Representing this ability to switch and having more co-dominance would also lead to predicting seasonal alternations among suspension feeders, which is not the case here, as both groups have the same prey preferences. Thus, in our modelling framework, variations in the relative abundance of different prey types cannot induce such alternations. Yet, introducing a generalist feeding group capable of dynamically adjusting its feeding strategy would require a detailed understanding of the trade-offs between generalist and specialist feeding strategies. To our knowledge, these trade-offs have not been quantified so far. While some modelling studies did explore the dynamics of adaptive feeding strategies, we are unaware of any that integrate both obligate and facultative feeding strategies .

Observations and laboratory experiments also suggest that ambushers tend to consume larger, more motile preys such as microzooplankton or dinoflagellates, while cruisers prefer smaller less motile preys . In our modelling experiments, prey preferences are not modified from the standard configuration of the PISCES model (PISCES-STD), which includes only one PFT for mesozooplankton. Consequently, we assume that both suspension feeding groups share the same diet. Allowing for differences in prey preferences is another perspective that could lead to a different spatial and temporal distribution of cruisers and ambushers. It could also modify the structure and composition of the prey community through a trophic trait cascade . However, this approach would require accurate representation of the dynamics of both microzooplankton and phytoplankton, in particular a description of their motility capacities .

Finally, body size is a master trait that impacts metabolic losses, ingestion rates, diet and predatory losses and that also interacts with feeding strategies. Additionally, it plays a crucial role in shaping diurnal vertical migration patterns and fecal pellet size, as larger body size is associated with deeper migrations and increased fecal pellet size. Both factors significantly impact carbon export, particularly its effectiveness in sequestering carbon in the ocean .

Body size also governs sinking speed, which, according to Stokes' Law, increases with the square of the organism's equivalent spherical diameter. For ambushers, the length of repositioning jumps scales approximately with body length, meaning jump frequency should scale with their body length . As a consequence, ambush feeding mode becomes increasingly risky and energetically more costly for larger organisms, making it less advantageous. showed that above a size of about 1 mm, ambush feeding is systematically outcompeted by active feeding modes, as they would lose too much energy in swimming to maintain their position in the water column. Observations also suggest that ambushers are rare among large copepods, while active organisms are found among smaller and larger copepods ( and see Appendix, Fig. ). Additionally, the study by also considered organism size as a major trait. The biogeography they predicted for active organisms varies depending on their body size whereas more similar spatial distributions are obtained for passive organisms, no matter their size. Furthermore, in the trait dataset used by , the majority of large copepod species corresponds to active organisms such as cruisers, while for the smallest copepods species, there is no clear dominance between passive and active organisms (see Appendix, Fig. ). This is correlated with Fig. 5 of , where the largest copepods are found at high latitudes, and would thus correspond to their cruisers (Fig. 5b and k of ), while the smallest ones are found at lower latitudes and would correspond to ambushers (Fig. 5b and i of ).

Therefore, dividing suspension feeders into at least two size classes and including DVM and/or fecal pellet production in our model could provide a more realistic representation of pelagic ecosystems, with larger cruisers contributing more to carbon export through deeper migrations and larger fecal pellets.