Biogeochemical feedbacks associated with the response of micronutrient recycling by zooplankton to climate change

Abstract Recycling by zooplankton has emerged as an important process driving the cycling of essential micronutrients in the upper ocean. Resupply of nutrients by upper ocean recycling is itself controlled by multiple biotic and abiotic factors. Although the response of these drivers to climate change will shape future recycling rates and their stoichiometry, their magnitude and variability are unaddressed by climate change projections, which means potentially important feedbacks on surface biogeochemistry are neglected. Here, we assess the impacts of climate change under the high emissions RCP8.5 scenario on the recycling of the essential micronutrients Fe, Zn, Cu, Co and Mn and quantify the regional control by zooplankton food quality, prey quantity, sea surface temperature and zooplankton biomass. A statistical assessment of our model results reveals that the variability in recycling fluxes across all micronutrients is mainly driven by the variability of zooplankton and prey biomass. In contrast, the variability in micronutrient recycling stoichiometry and its response to climate change are more complex and is regulated by zooplankton food quality and prey quantity. Regionally, the relative influence of each driver on recycling changes in our model by the end of the 21st century. Temperature becomes an important driving factor in the polar regions while the expansion of oligotrophic regions leads to the importance of food quality increase for low and mid‐latitudes. These responses lead to novel feedbacks that can amplify the response of surface ocean biogeochemistry and alter nutrient deficiency regimes.


| INTRODUC TI ON
Trace metals are essential micronutrients for marine phytoplankton and their scarcity in the environment can contribute to regulating primary productivity and ecosystem functioning in the surface ocean (Moore et al., 2013;Tagliabue et al., 2020), as well as being toxic at high concentrations (Brand et al., 1986;Debelius et al., 2011). A range of micronutrients have various essential roles in phytoplankton and zooplankton physiology and therefore affect the ocean cycles of major elements, such as carbon (C), nitrogen (N), silica (Si) and phosphorus (P; Morel et al., 2014). The relative amount of micronutrients in phytoplankton primary producers is known as the stoichiometry and is much more variable (3-to 20-fold) for micronutrients than for major elements, such as N and P (Twining & Baines, 2013). Zooplankton grazers | 4759 RICHON aNd TaGLIaBUE rely on the micronutrient content of their ingested prey to satisfy their nutritional requirements and after assimilation, any excess nutrients are directly excreted in the dissolved phase at the surface or egested as faecal pellets, which sink to depth (see Richardson, 2008;Steinberg & Landry, 2017).
Micronutrient concentrations in the upper ocean are generally very low (in the micro to picomolar range) and recycling by zooplankton has emerged as an important component of upper ocean micronutrient cycling in prior work. Recycling has been shown to play a crucial role for Fe (Boyd et al., 2015;Sarthou et al., 2008;Strzepek et al., 2005;Tagliabue et al., 2014) and under resource limitation, zooplankton recycling can sustain phytoplankton activity (see Laglera et al., 2017;Rafter et al., 2017). In this context, recycling of dissolved micronutrients by zooplankton increases their surface ocean residence time (Boyd et al., 2012;Hutchins & Bruland, 1994). The importance of micronutrient recycling by zooplankton stems from two main mechanisms. First, the mismatch between zooplankton and prey micronutrient stoichiometries can lead to high rates of recycling due to altered zooplankton assimilation efficiencies (Richon et al., 2020). Second, the reduced supply of micronutrients from external sources over much of the open ocean (Gaillardet et al., 2014;Mahowald et al., 2018) makes internal cycling a key driver of upper ocean inventories.
Nutrient recycling by zooplankton is influenced by many direct and indirect factors that are under both biotic and abiotic control.
Temperature directly impacts zooplankton distributions, their growth and a range of physiological rates (Richardson, 2008), as well as the underlying ocean circulation. Zooplankton biomass is a major driver of micronutrient recycling rates, as a greater abundance of zooplankton will lead to greater recycling fluxes. The amount of ingested prey affects recycling through its effect on nutrient assimilation, with greater overall prey ingestion leading to a reduced zooplankton assimilation efficiency and higher recycling rates (Richon et al., 2020;Xu & Wang, 2001). Finally, the absolute and relative micronutrient availability in the surface ocean influences phytoplankton uptake, which drives the phytoplankton micronutrient stoichiometry, which, in turn, impacts recycling by controlling food quality (Richon et al., 2020). Food quality was defined by Richon et al. (2020) as the dimensionless balance between the zooplankton stoichiometry and the stoichiometry of their ingested prey. When food quality is close to 1, zooplankton are feeding on stoichiometrically similar prey. If food quality deviates from 1, then prey are either richer or depleted in nutrients, which leads to excess or insufficient ingestion by zooplankton. A food quality factor close to 1 favours low recycling due to a high assimilation efficiency, while as food quality moves further away from 1, recycling increases as the assimilation efficiency drops (Richon et al., 2020). Despite its impacts on surface biogeochemistry, recycling is challenging to quantify in situ (e.g. Boyd et al., 2012). This means that the impacts of this range of biotic and abiotic drivers on zooplankton recycling and the biogeochemical and ecological consequences are difficult to identify (Mayzaud & Pakhomov, 2014). In this context, global ocean models can be useful as they represent various abiotic and biotic drivers of zooplankton recycling and their role in driving modelled recycling fluxes and stoichiometry can be assessed.
The consequences of climate change on ocean surface circulation, nutrient and ecosystem structure and function have been studied extensively (e.g. Bindoff et al., 2019;Kwiatkowski, Aumont, Bopp, & Ciais, 2018;Kwiatkowski et al., 2020;Lotze et al., 2019). Increased stratification due to climate change affects vertical nutrient supply and alters the strength of phytoplankton nutrient limitation (e.g. Tagliabue et al., 2020). These changes will also affect micronutrient uptake and phytoplankton stoichiometry while zooplankton recycling rates will respond to the integrated effect of the full suite of drivers. Temperature, plankton biomass and food quality may vary in different directions across different micronutrients and in distinct regions due to their unique cycling and distinct biological requirements. This means that the impacts of climate change on recycling fluxes and stoichiometry are poorly quantified and absent from climate change assessments of ecosystem change (Bindoff et al., 2019). Due to the key role of recycling for surface ocean biogeochemistry, important feedbacks that may operate between climate-driven changes in micronutrient recycling and resource availability for upper ocean ecosystems have been neglected.
In this article, we use a global coupled physical-biogeochemical model under a high greenhouse gas emissions scenario (RCP8.5) to assess the climate change impacts on surface micronutrient biogeochemistry. We focus on iron (Fe), cobalt (Co), copper (Cu), manganese (Mn) and zinc (Zn), included in an earth system model for the first time. Our results show that recycling fluxes decrease for all micronutrients in the low latitudes and increase in the mid-to high latitudes by the end of the 21st century. On the other hand, recycling stoichiometry (Micronutrient:C ratio in recycled material) varies depending on the metal. Then, we explore the changes in recycling drivers (zooplankton food quality, prey quantity, sea temperature and microzooplankton biomass) and use statistical models to identify how different driving factors influence micronutrient recycling in different ocean regions in the beginning and by the end of the century. Finally, we discuss how the changes in the different drivers may impact recycling and how this may feedback on surface micronutrient biogeochemistry.

| The NEMO-PISCES model and experimental approach
We use the global coupled physical-biogeochemical model NEMO/ PISCES , which represents nitrate, ammonium, phosphate, silicic acid and dissolved iron cycling, the full carbon and oxygen systems, two phytoplankton groups (nanophytoplankton and diatoms) and two zooplankton size classes (microzooplankton and mesozooplankton) and has been extensively used to study regional and global ocean biogeochemistry (e.g. Aumont et al., 2017;Gorgues et al., 2019;Kwiatkowski, Aumont, Bopp, & Ciais, 2018;Richon et al., 2017;Tagliabue & Resing, 2016). Recent developments of the PISCES model have included micronutrient cycling such as Cu (Richon & Tagliabue, 2019), Zn, Co (Tagliabue et al., 2018) and Mn to build a new version of the model called PISCES-BYONIC. The model is fully described and evaluated in supplement (Text S1). We present here the key information on the model.
In the standard version of PISCES-BYONIC, the phytoplankton macronutrient stoichiometry (C:N:P) is fixed, but it is variable for micronutrients (Fe, Co, Cu, Mn and Zn), chlorophyll and silica. The maximum micronutrient:C molar quotas are 80E-6 for Fe, 40 and 123 for Zn in nanophytoplankton and diatoms, respectively, 16E-6 for Cu, 1.2E-6 for Co and 8E-6 for Mn, which broadly reflects available observational constraints (e.g. Twining & Baines, 2013;Twining et al., 2015). The zooplankton molar micronutrient to carbon stoichiometry is fixed to 10E-6 for Fe, Zn and Cu, and to 0.16E-6 and 1E-6 for Co and Mn, respectively, following the more limited observational understanding (see Baines et al., 2016;Ratnarajah et al., 2014;Twining & Baines, 2013).
The impacts of climate change on micronutrient recycling and recycling stoichiometry were simulated using offline physical fields from the IPSL-CM5A climate model, as in previous work (Kwiatkowski, Aumont, Bopp, & Ciais, 2018;Tagliabue et al., 2020). We performed two simulations: a preindustrial control (PICONTROL) from 1801 to 2100 with atmospheric CO 2 concentrations fixed to the preindustrial value. Then, from 1851 to 2100, we performed a second simulation initialized from year 1851 of the PICONTROL, with CO 2 concentrations varying according to the historical pathway until 2005 and switching to the high emissions RCP8.5 scenario (Riahi et al., 2011) from 2006 to 2100. Previous studies using NEMO/PISCES under the RCP8.5 scenario showed a global increase in stratification, increased SST and decrease in surface macronutrients leading to changes in plankton distribution and stoichiometry Kwiatkowski, Aumont, Bopp, & Ciais, 2018).
For our simulations, we use constant external nutrient sources (hydrothermal vents, rivers and aerosols). Sedimentary sources of Co and Mn are O 2 dependent (Tagliabue et al., 2018). To assess our results, we define two periods of time: PRESENT (model results averaged over 1991PRESENT (model results averaged over -2000 and FUTURE (model results averaged over 2091-2100). Previous work with this model has shown that microzooplankton recycling accounts for most of micronutrient recycling fluxes (see Richon et al., 2020); therefore, we focus here on microzooplankton.

| Statistical methods to identify recycling drivers
We seek to analyse the biotic and abiotic drivers of recycling and recycling stoichiometry variance. Because of the intimate role of zooplankton recycling in the overall ocean biogeochemical cycling, we could not quantify the role of individual drivers using sensitivity tests. Instead, to identify the role of different factors influencing the variance in micronutrient recycling fluxes and stoichiometry, we used statistical models (Richon et al., 2020). The generalized linear models (GLM) given in Equations (1)  is higher than that of prey. Therefore, zooplankton may not fulfil its physiological demand for micronutrients through its diet. On the other hand, food quality <1 indicates that prey stoichiometry is higher than zooplankton demand and that zooplankton may ingest excess micronutrients.
Prey quantity is the amount of micronutrient available in the preys (phytoplankton and organic particles), it is defined as the sum of micronutrient concentration in each prey type weighted by microzooplankton preference for each prey (Equation 3): In Equation (3), P j i represents the concentration of micronutrient j in prey i and p i the preference of zooplankton for prey i, with i either nanophytoplankton, diatoms or organic particles.
These statistical models explore the roles of the biotic and abiotic parameters directly driving recycling changes. The choice of the explanatory variables in Equations (1) and (2) is based on current scientific understanding of each driver's impact on micronutrient recycling as well as the results from Richon et al. (2020; see also the references therein). Food quality (Schmidt & Hutchins, 1999), prey composition and biomass (Anderson & Harvey, 2019;Steinberg & Landry, 2017) and seawater temperature (Richardson, 2008) have all been shown to influence micronutrient recycling by zooplankton.
Zooplankton biomass is expected to directly influence micronutrient recycling since more zooplankton would lead to higher total recycling flux.
The ε term is the error term. A high error term may signify that some factors influencing micronutrient recycling were omitted in the GLMs. The omitted factors may be the indirect effects of physics (stratification and modification of currents) and changes in external micronutrient sources (e.g. sedimentary Fe, Mn and Co).
We use the R package relaimpo (https://cran.rproj ect.org/ web/packa ges/relai mpo/index.html) to calculate the percentage of (1) Micronutrient recycling flux = food quality + prey quantity Micronutrient recycling stoichiometry = food quality + prey quantity + sea temperature + zooplankton biomass + . (3) recycling and recycling stoichiometry variance explained by each statistical model in PRESENT and FUTURE periods. Relaimpo also allows us to calculate the fraction of recycling and recycling stoichiometry variance explained by each of the factors.

| RE SULTS
In this section, we use the results from our simulations to study the impacts of climate change on micronutrient recycling and recycling stoichiometry and use the statistical models to identify the main regional drivers of recycling.

| Recycling stoichiometry
The recycling stoichiometry, relative to C, differs across micronu-  Figure S4c,e). This may be due to the low maximum cellular quotas imposed for these micronutrients that restricts excess accumulation or the contribution of particulate matter to the zooplankton diet.

| Drivers of recycling flux
The GLM in Equation (1)

| Drivers of recycling stoichiometry
Unlike recycling fluxes, which are mostly driven by zooplankton and prey biomass variability, the drivers of micronutrient recycling stoichiometry are more complex, with different factors emerging depending on the nutrient and the region considered ( Figure 4). The (2)   The feedbacks from changes in micronutrient recycling stoichiometry on surface biogeochemistry are more complex since they are strongly influenced by zooplankton food quality in many regions (see Figure 4), which itself depends on the micronutrient stoichiometry of phytoplankton (Figure 2; Figure S4). The differential recycling of resources will change the micronutrient environment of phytoplankton and potentially modify nutrient limitation patterns or intensities in the future (Moore, 2016;Sterner, 2009). The resulting alteration of phytoplankton stoichiometry will feedback on zooplankton food quality ( Figure 5). It is important to note that the change in phytoplankton stoichiometry in response to changing recycling stoichiometry is complex and also depends on how far phytoplankton are from their assumed maximum quota and the specific micronutrient uptake kinetics (see Text S1; Table S3), which regulates the specific response of micronutrient uptake rates. Therefore, identifying the general sign of the recycling stoichiometry feedback is complex as it ultimately depends on how food quality deviates from 1 in response to the climate change alterations to recycling stoichiometry (since the zooplankton micronutrient assimilation efficiency increases as food quality gets closer to 1). For example, in the Equatorial Pacific and the Arctic, Zn food quality is projected to increase in response to climate change (see Text S2; Figure S6). The ensuing decline in the Zn recycling stoichiometry then acts to reduce the phytoplankton stoichiometry. This change in phytoplankton stoichiometry then exerts a negative feedback on the recycling stoichiometry that may contribute to lowering Zn food quality. On the other hand, food quality deviations away from 1 are common across many micronutrients ( Figures S5 and S6). The resulting increase in the recycling stoichiometry then exerts a positive feedback as the elevated phytoplankton stoichiometry that results drives food quality further still from 1.

| Quantifying the impact of food quality feedback on biogeochemistry
To further explore the nature of the feedbacks associated with the recycling stoichiometry and food quality, we repeated our climate change simulations with food quality factors for micronutrient recycling fixed at their preindustrial levels. In these simulations, the impacts of changes in food quality on micronutrient recycling are eliminated and the recycling stoichiometry only varies in response to changes in zooplankton biomass, prey quantity and temperature.
We focus here on the essential micronutrient Fe and examine the trends in the recycling stoichiometry (ratio of the Fe and C recycling rates) normalized to the first hundred years of the PI control simulation (1801-1900) both in the reference simulation and in the simulation with fixed food quality across different biogeochemical provinces ( Figure 6).
The recycling stoichiometry of Fe is lower throughout the 21st century when the food quality is fixed ( Figure 6). These results highlight how changes in the zooplankton food quality lead to high relative rates of iron recycling in response to climate change, amplifying climate change impacts on Fe recycling stoichiometry. The large differences between the two simulations in the Pacific regions (and to a lesser extent, in the Atlantic) confirm the major role for food quality in driving Fe recycling stoichiometry in those regions (see Figure 4).
Variations in Fe food quality in the Atlantic and Southern Ocean are more limited (Figure S6), leading to more muted food quality feedbacks in these regions (Figure 6b,d,f,g).
Both our model and available observations indicate that the low-latitude Pacific is iron limited. Therefore, recycling changes may trigger important changes in net primary production. Our results indicate that recycling stoichiometry feedbacks lead to greater relative rates of iron recycling due to alterations to the food quality of zooplankton. This enhanced surface retention of iron will contribute to lessening the extent of iron limitation of primary producers in the region, with implications for net primary production projections (e.g. Tagliabue et al., 2020). These amplifying feedbacks illustrated for Fe will likely also be important for Mn and Zn ( Figure 2), which also show large projected changes in food quality in the Pacific ( Figure S6).

| Constraining micronutrient recycling drivers
This article presents results from novel biogeochemical models of micronutrients (see Text S1 for the complete model description).
The dissolved micronutrient concentrations are well reproduced by our model in general (see Figures S1 and S2; Table S6), which raises confidence in our projections. However, micronutrient prey content and zooplankton micronutrient food quality are much more complex and available in situ estimates would provide important additional constraints on our model. The phytoplankton and zooplankton micronutrients stoichiometries we apply in our model, which also underlie the calculations of food quality and prey contents, are based on the available measurements from phytoplankton (Twining & Baines, 2013) and zooplankton (Baines et al., 2016;Ratnarajah et al., 2014). Evidence regarding the importance of multiple micronutrients for microbial activity is expanding (Browning et al., 2017(Browning et al., , 2021Peers et al., 2005;Saito et al., 2008;Wu et al., 2019). This highlights the need for a broader consideration of nutrient limitation, recycling and its sensitivity to climate change in Earth System Models.

| Implications for ocean biogeochemical cycles
Ocean biogeochemical cycles facilitate the ocean's regulatory role associated with a suite of ecosystem services and are largely governed by dissolved resource concentrations and lower trophic-level functioning (Bindoff et al., 2019). Modifications to environmental conditions through climate change will impact the drivers of ocean biogeochemical cycles, ranging from changes to nutrient availability to species biomass and physiology (e.g. Kwiatkowski et al., 2020;Laufkötter et al., 2015;Lotze et al., 2019;Tagliabue et al., 2020 Food quality emerged as an important driver of the projected changes in micronutrient recycling dynamics during our experiments. This term depends both on prey, which can vary, and zooplankton stoichiometry, which are currently fixed in our model. Similar to phytoplankton, it is possible that the micronutrient composition of zooplankton may change depending on the species, the size and life stage of the zooplankton, as well as environmental conditions (Baines et al., 2016). Any such variability in the zooplankton micronutrient stoichiometry would then become an additional factor governing the impact of climate change on micronutrient recycling.  and Kwiatkowski, Aumont, Bopp, & Ciais (2018) showed that accounting for variability in phytoplankton C:N:P quota under the climate change RCP8.5 scenario F I G U R E 6 Times series of the trends in iron recycling stoichiometry (normalised against the PICONTROL: 1801-1900 average) in the surface waters of different regions (see Figure 3 for region location). Black lines represent the reference simulation, red lines represent the simulation with fixed food quality [Colour figure can be viewed at wileyonlinelibrary. com] leads to similar declines in NPP as with a standard fixed stoichiometry version. However, food quality (here phytoplankton N and P content) also varies significantly with declines of 1%-6% globally, but as high as 20% in the Arctic. Because macronutrient and micronutrient recycling are decoupled in PISCES, assumptions regarding N and P stoichiometry do not affect micronutrient recycling. However, the changes in N and P food quality may lead to significant changes in N and P recycling, potentially leading to similar feedbacks than for micronutrients identified in this study.
Sea surface temperature changes are a fundamental component of the changes in ocean biogeochemistry in general because it underpins multiple direct and indirect biotic and abiotic processes (Doney, 2013). In our model, temperature directly affects growth, grazing and recycling rates via metabolic functions . However, our model does not account for any potential adaptation responses of the ecosystem to temperature change (Dam, 2012). These adaptive responses are uncertain and difficult to forecast at present but may be important and require further experimental and observational constraints.

| Implications for wider ecosystem functioning
The PISCES model represents the lower trophic levels of the planktonic ecosystem and neglects the interaction with zooplankton predators (fish, birds and mammals). However, the effects of temperature and biomass changes may impact zooplankton consumption by higher trophic-level predators in multiple ways. Shifts in the distribution of zooplankton predators due to shifts in thermal niches and invasion of new potential predators (Cheung et al., 2009) will combine with biomass changes that may modify the balance between zooplankton and predator biomass, with potential consequences for birds or fish stocks (Beaugrand, 2004;Bertram et al., 2001). Combined with the direct influence of temperature changes on the physiology and grazing rates of zooplankton predators, these impacts may further amplify the changes in zooplankton biomass simulated by our model and lead to additional feedbacks on ocean biogeochemistry.
Modifications to food quality and surface ocean stoichiometry may have consequences for upper trophic-level biomass and nutrient content, which may, in turn, affect human nutrition when commercially important fish are impacted (Hicks et al., 2019;Tagliabue et al., 2020). The broader consequences of changes in microzooplankton recycling we focus on here may have far-reaching consequences for ocean biogeochemical cycles and ecosystem functioning and require further experimental, field and modelling efforts in the future.

| CON CLUS ION
Recycling by microzooplankton occupies a central role in the biogeochemistry of the surface ocean and supports marine ecosystems. In this study, we examined how recycling of micronutrients responds to the change in climate associated with the high emissions RCP8.5 scenario. We find that the recycling of essential micronutrients is impacted by a suite of biotic and abiotic factors, which are all influenced by climate change. Our model experiments show that the net impact of climate change on surface micronutrient recycling results from complex interactions between different drivers, which may affect the rates of stoichiometry of recycling in different directions. As plankton biomass and physiology responds nonlinearly to climate change, the response of upper ocean biogeochemistry may conceal complex multifactorial interactions.
Previous studies examining the impact of climate change on ocean biogeochemical cycles and ecosystem functioning have tended to focus on the direct impacts for nutrient and plankton distributions Kwiatkowski et al., 2020;Laufkötter et al., 2015). However, ocean biogeochemical cycles and ecosystem functioning are determined by a network of complex interactions and feedbacks that include grazers. We find that the projected changes in micronutrient zooplankton recycling dynamics depend on drivers that vary across different micronutrients and by region. Moreover, these drivers often interact and exhibit feedbacks in response to climate change. Therefore, the response of micronutrient recycling by the end of the century is not straightforward. We find that modifications to the magnitude and stoichiometry of recycling lead to feedbacks on biogeochemical cycles. There is a positive feedback between climate change, nutrient levels, zooplankton biomass and gross recycling rates, but the nature of the feedback is much more complex for the stoichiometry of recycling which regulates resource deficiency. Therefore, it is important to understand and quantify the multifaceted drivers of micronutrient recycling by zooplankton and how these impact surface ocean biogeochemistry to reduce uncertainties in the projections of future ocean biogeochemistry.