Species identity and the functioning of ecosystems: the role of detritivore traits and trophic interactions in connecting of multiple ecosystem responses

dissimilarity of detritivore functional traits was not associated with dissimilarity of ecosystem functioning. These results serve as an important reminder that consumers influence ecosystem processes via multiple energy channels and that food web interactions set important context for consumer-mediated effects on multiple ecosystem functions. Given that species are being lost, gained and redistributed at unprecedented rates, we can anticipate that changes in species identity will have additional ecosystem consequences beyond those predicted by species’ primary functional role.


Introduction
A major goal in ecological research is to understand the consequences of changes in biological diversity for ecosystem functioning (Cardinale et al. 2012, Tilman et al. 2014, Isbell et al. 2017).Therefore, many scientists have conducted experiments where they manipulated species richness and measured the productivity or resource uptake of focal communities (Hooper et al. 2005, Balvanera et al. 2006, Cardinale et al. 2006).Cumulatively, these experiments have demonstrated that less diverse communities consistently produce less biomass, and typically support reduced ecosystem process rates (Cardinale et al. 2011).
Notably, previous conclusions about the influence of diversity change on ecosystem functioning have been based predominately on plant communities, despite the fact that consumers feeding at higher trophic levels may be relatively more vulnerable to environmental change (Voigt et al. 2003, Hines et al. 2015a).Moreover, in many cases, species losses from ecosystems can be balanced by species gains (Vellend et al. 2013, Dornelas et al. 2014), which has led to skepticism about whether conclusions from classic biodiversity experiments adequately describe ecosystem responses to realistic changes in biological communities (but see Eisenhauer et al. 2016, Wardle 2016).That is, communities often experience high compositional turn-over in species identity (β-diversity) in response to natural and anthropogenic disturbances in environmental conditions, even when there is no change in local species richness (Dornelas et al. 2014).Therefore, it is important to anticipate how changes in species identity will influence the functioning of ecosystems.
One basis for considering species contribution to ecosystems is to identify functional groups of species that share similarities in traits, or groups of traits, that determine the way the group regulates an ecosystem process (Laureto et al. 2015).Such traits can be quantified on axes that describe morphological, physiological, behavioral and life history characteristics (Ellers et al. 2018).Emphasis is often placed on 'effect' traits related to resource use, consumption-based processes (Pla et al. 2012), and to a lesser extent, how species activity may influence the structure of their habitat (Lawton 1994, Jones andLawton 1995).For example, invertebrate detritivores drive decomposition via direct consumption of detritus, by influencing microbial dynamics, and by physical shredding of litter (Zimmer et al. 2004).In this context, functional group is considered to be synonymous with trophic level.Theory suggests that for a given trophic/functional group, dissimilarity in traits indicates differences in proficiency with respect to a focal ecosystem process (Chapin et al. 1997, Louca et al. 2018).
Although discrete functional groups have a superficial appeal, it has long been recognized that these groupings mask more complex influences of species on multiple processes that sustain an ecosystem (Hines et al. 2015b, Risch et al. 2018).It is tempting to suggest that trait dissimilarity may also be used to predict dissimilarity in multiple ecosystem functions (Heemsbergen et al. 2004).However, hierarchical (Holzwarth et al. 2015) or nonlinear relationships between trait dissimilarity and dissimilarity of single functions (Cadotte 2017) may inhibit detection of multivariate traitfunction relationships.Accordingly, it has been suggested that measuring more ecosystem functions (ecosystem multifunctionality), rather than traits, could be used to document species' unique contributions to ecosystems, and could be an informative way to quantify ecosystem responses to biodiversity change (Byrnes et al. 2014).
Unfortunately, measures of ecosystem multifunctionality have not been standardized across study systems or trophic groups, and careful consideration of methods is needed to interpret results (Bradford et al. 2014, Manning et al. 2018, Giling et al. 2019).For example, in a research synthesis describing the consequences of changes in biodiversity on ecosystem multifunctionality of four distinct trophic groups, Lefcheck et al. (2015) found no evidence that changes in detritivore diversity influence ecosystem multifunctionality.Yet, in the seven independent experimental manipulations of detritivore diversity that fit the criteria for their analysis, typically fewer than three functions were reported (mean = 2.5, mode = 2, n = 17) (Lefcheck et al. 2015).Additionally, all of the reported functions describe responses related to decomposition or detritivore growth, which would be more typical of consumer-resource dynamics (Lefcheck et al. 2015).Given that detritivores are notorious omnivores, quantifying their influence on either brown (i.e.soil-and decompositionbased processes) or green (primary producer-based processes) functions in isolation may insufficiently assess their influence on ecosystem functioning (Hines andGessner 2012, Mulder et al. 2013).Experiments reporting a broader array of ecosystem functions that are more clearly linked to food web dynamics and realistic species changes are needed to assess the multifunctional role of detritivores in ecosystems (Hines et al. 2015b, Mancinelli andMulder 2015).
Here we use model communities, composed of common salt marsh plant and invertebrate species, to test how changes in detritivore communities influence ecosystem multifunctionality.We quantify six ecosystem functions that serve as proxies for the suite of ecological processes that broadly represent ecosystem functioning (Hector andBagchi 2007, Gamfeldt et al. 2008).To assess how well classic functionalgroup assignments capture species' main roles in ecosystems (i.e.how effects of different detritivores extended beyond effects on decomposition), we also test whether effects of detritivores on ecosystem functions are dependent upon food web compartment (belowground, aboveground or combinations of aboveground and belowground functions), or the number of functions considered (1-6).We expect that all detritivores will affect decomposition and subsequent nutrient mineralization, but that the multifunctional role of detritivores will be revealed by examining a broader range of ecosystem functions associated with food web interactions (Fig. 1).Specifically, 1) by consuming and mineralizing leaf litter, detritivore activity may subsidize nutrients available for plant uptake, enhance plant quality for herbivores, and indirectly support herbivore population growth.Additionally, Figure 1.Conceptual diagram and paired table showing feeding interactions and functional traits of four numerically dominant salt marsh detritivores: Littoraria irrorata, Melampus bidentatus, Orchestia grillus, Littorophiloscia vittata.Feeding interactions are shown as arrows pointing from consumer to resource.Species functional traits related to food sources (detritus, microbes, live plants) and activity (body size, ecosystem engineering and vulnerability to spider predation) characterize functional dissimilarity among four numerically dominant salt marsh detritivore species, while unique combinations of traits can be used to characterize species identity.Throughout the text species are referred to by their genus name.
2) by serving as alternative prey, detritivores can maintain predator populations, and enhance predation on herbivores.However, leaf litter structure provides a component of habitat complexity that is the preferred habitat of spiders, in part, because it provides refuge from cannibalism (Langellotto and Denno 2004).Therefore, 3) by modifying habitat structure (i.e.ecosystem engineering) detritivores can indirectly influence populations of predators and their herbivore prey.Depending on the vulnerability of detritivores to predation, and the strength of their influence on decomposition, some predicted influences on the aboveground community can differentially promote predator populations (1), subsidize herbivore populations (2) or potentially destabilize predatorherbivore dynamics (3).Therefore, the extent to which each effect attenuates or cascades to influence all measured ecosystem functions likely depends on the unique combinations of traits of the detritivores (Fig. 1).

Study system
Along the Atlantic coast of North America, the single cordgrass species, Spartina alterniflora, can form expansive monospecific stands (Valiela et al. 1976, Mendelssohn 1979, Mendelssohn et al. 1981, Bradley and Morris 1991).In the decomposition based 'brown' food web compartment, there are four common detritivorous species: Littorophiloscia vittata, Orchestia grillus, Melampus bidendatus, Littoraria irrorata (Fig. 1 for description of traits, abundances and trophic interactions).The aboveground 'green' food web compartment is numerically dominated by the phloem-feeding herbivore Prokelisia dolus (Delphacidae: Hemiptera) and the generalist spider predator Pardosa littoralis (Araneae: Lycosidae) (Denno et al. 2002).Prokelisia herbivores are sensitive to changes in host plant nitrogen, and can reach densities as high as 50 000 individuals m −2 in patches of high-quality (high % N) S. alterniflora plants (Hines et al. 2006).Pardosa wolf spiders have been reported to consume Prokelisia (Denno et al. 2002), as well as detritivorous prey, such as isopods and amphipods (Morse 1997, Hines andGessner 2012) (Fig. 1).Pardosa aggregate in areas rich in Spartina leaf litter because the structural complexity of the litter provides refuge from cannibalism, allowing for enhanced prey suppression (Langellotto and Denno 2004).

Experimental design
To determine the influence of changes in detritivore species on multiple ecosystem functions, we established five detritivore treatments in microcosms.Species densities were chosen to fall within the range of naturally variable field densities, and also to establish communities with comparable detritivore biomasses (Fig. 1): no detritivore, 50 Littorophiloscia vittata, 50 Orchestia grillus, 50 Melampus bidentatus or 5 Littoraria irrorata) (Graça et al. 2000, Zimmer et al. 2002, 2004).These microcosm densities correspond to 52 Pardosa m −2 , 130 Littoraria m −2 and 1315 individuals m −2 for each the species stocked at 50 individuals per microcosm.To tease apart the influences of detritivores that are realized via nutrient/decomposition pathways from influences on ecosystem functioning that are a consequence of their role as prey (Hines andGessner 2012, Hines et al. 2015a), the detritivore treatments were crossed with two levels of spider predation (4 Pardosa littoralis present or absent).The full experimental was a randomized complete block, 5 × 2 factorial design experiment, which consisted of 10 treatments that were replicated in each of nine spatial blocks, resulting in 90 microcosms.The experiment was located in a common garden habitat, at the Smithsonian Environmental Research Center in Edgewater, MD.
Each microcosm consisted of a sand-filled pot (22 cm diameter × 21 cm deep) with five transplants of small Spartina plants, 25 g S. alterniflora leaf litter and 30 adult Prokelisia herbivores (25 females and 5 males, corresponding to 825 individuals m −2 ).Potted plants were enclosed in clear plastic tube cages (21.5 cm diameter × 30 cm tall), capped with a mesh top, and pots were placed in plastic pans (30 cm diameter × 10 cm deep) filled with water.Treatments were established in a sequential fashion that approximated the phenology of field interactions.Detritivore species, which remain in meadow habitats year-round, were added on 7 June 2007 and were allowed to settle and feed before herbivores emerged.Prokelisia herbivores, which overwinter as nymphs in leaf litter and emerge as adults by the end of June, were added to microcosms on 25-27 June 2007 and allowed to settle for four days, before four juvenile Pardosa spiders were added to half of the microcosms.
The experiment was terminated on day 60, when visual counts indicated that the second generation adult planthoppers began a slight natural decline (Denno and Roderick 2003).Because detritivores and spiders were often hidden in the leaf litter, it was not possible to obtain a visual assessment of their densities throughout the course of the experiment.Therefore, final invertebrate densities were obtained by collecting all animals from microcosms using a vacuum sampler fitted with mesh bags.Arthropods were killed using ethyl acetate, transferred into plastic bags, before they were counted in the laboratory.
To assess plant growth and litter mass loss, S. alterniflora litter, root and shoot biomass was harvested, washed and sieved through 1.7 mm sieves, freeze-dried and weighed to the nearest 0.01 g.Because damage imposed by phloem feeders can be difficult to detect, we did not distinguish between loss of plant biomass caused by detritivore grazing and phloem feeding herbivores.Instead, herbivore abundance was used as a proxy for herbivory by strict herbivores, which is in line with rationale for energy flux to herbivores (Barnes et al. 2018, Gauzens et al. 2019), but could overlook the potential for non-linear density dependent feeding rates (Hines et al. 2016)  herbivore populations during the course of the experiment, and is a proxy related to secondary production of herbivores.Pardosa predators molted to adults but, due to longer generation times, they did not reproduce during the time course of the experiment.Nonetheless, for consistency, we also consider persistence of predators as a measure of predation potential.Soil inorganic soil nitrogen (NH 4 + and NO 3

Statistical analysis
We used a mixed model ANOVA (lme Pinheiro et al. 2014) to assess the influence two fixed factors (i.e.detritivore identity and predator presence) on six individual functions as well as ecosystem multifunctionality.For all analyses, the microcosm was considered as the experimental unit in which response variables were reported and spatial block was included as a random factor.To calculate average ecosystem multifunctionality, we used the averaging approach (Byrnes et al. 2014), and included the average of all possible reported response variables in each microcosm (shoot biomass, root biomass, litter decomposition rate k, nutrient concentration [NO 3 − ], herbivore abundance, predator abundance).Predator abundance was not included in the calculation when predators were absent.Because standardized nutrient concentrations of NH 4 + and NO 3 − were similar, and neither molecule was sensitive to treatments in this experiment, we chose to report NO 3 − in the average ecosystem multifunctionality metric and text.
We used another mixed model analysis to consider how two additional factors (i.e.food web compartment, and number of functions included in the average multifunctionality metric) influenced the average multifunctionality responses in the original 2 × 5 factorial experimental.In this analysis, we calculated average multifunctionality for each microcosm using all viable combinations of 1, 2, 3, 4, 5 and 6 standardized functions that filled the criteria of being from aboveground compartment, belowground compartment or mixed compartment multifunctionality metric that included a combination of at least one aboveground and one belowground function.The experimental design was necessarily unbalanced as some combinations of covariate calculations were not possible (e.g.mixed aboveground-belowground compartment with only one function, or aboveground compartment with three functions when predators were absent).For this analysis, we were interested in how the number of functions and/or the focal compartment influenced variation in multi-functionality results, rather than the specific combination of functions included in calculations for each microcosm.Therefore, we included the identity of function combinations in the model as a random factor.Spatial block was also included in the model as a random factor.Following significant (p < 0.05) mixed model results, we used pairwise comparisons with Tukey's adjustment for multiple comparisons to identify significant treatment differences (lsmeans: Lenth 2018).All data met assumptions of homogeneity of variance and normality of residuals.
To test whether dissimilarity in species functional traits were associated with dissimilarity in multiple ecosystem functions, we constructed distance matrices both for species traits and for the ecosystem functions performed by model communities with and without predators (Gower 1971).Subsequently, we used a partial mantel test to examine correlations between the three distance matrices (species traits, functioning of communities with or without predators) (Vegan: Legendre andLegendre 1998, Oksanen et al. 2019).

Individual functions
The most rapid decomposition occurred when Melampus was present (Fig. 2A).The other two belowground functions, nitrate availability (Fig. 2B) and root production (Fig. 2C) were not influenced by detritivore species either alone (open circles), or when detritivores and predators were present in combination (filled circles).Consequently, only one-of-three belowground ecosystem response variables was sensitive to turnover in detritivore identity (Fig. 2A-C, Table 1).
Aboveground ecosystem functioning was comparatively more sensitive to changes in detritivores, and observed effects were partly moderated by predator-detritivore interactions (Fig. 2D-F).In the absence of detritivores, Pardosa spiders survived (Fig. 2D).Prokelisia populations were suppressed to low densities in detritivore-free communities when spiders were present compared to when spiders were absent (Fig. 2E).These differences in herbivory (i.e.herbivore abundance) attenuated, leaving plant shoot biomass unaffected by the presence or absence of predators in control treatments (Fig. 2F).Microcosms with Littorophiloscia also supported predation (i.e.predator survival; Fig. 2D), but predators did not suppress herbivores as strongly when this source of alternative prey was available (filled circles; Fig. 2E).Orchestia was less effective as a source of alternative prey (Fig. 2D-E).That is, there were no differences in predator or herbivore survival when Orchestia was present, compared to when detritivores were absent (Fig. 2D-E).Communities where Melampus was present had enhanced Spartina shoot biomass compared to when Orchestia or Littoraria were present, but no significant effects on predators, or herbivores (Fig. 2F, Table 1).Microcosms with Littoraria were associated with the reductions in all three aboveground ecosystem functions, namely reduced predator survival (Fig. 2D), lower herbivore abundance in the presence and absence of predators (Fig. 2E), and less plant shoot biomass (compared to Melampus Fig. 2F).

Ecosystem multifunctionality
Ecosystem multifunctionality (EMF) was sensitive to changes in detritivore identity and the presence of predators (Fig. 3, Table 2).Trait dissimilarity, however, was not associated with dissimilarity in ecosystem functions (partial Mantel test between traits and ecosystem functions with predators: 16000706, 2021, 10, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/oik.08333by Universitaetsbibliothek Leipzig, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License r = 0.44, p = 0.12, between traits and ecosystem functions without predators: r = −0.11,p = 0.63).We observed both the highest and the lowest values of EMF in communities with detritivores that were not susceptible to predation by spiders (Melampus and Littoraria, Fig. 3).Intermediate values of EMF were recorded in communities with detritivores that were susceptible to predation (Littorophiloscia and Orchestia) (Fig. 3).Presence of predators typically reduced EMF, presumably by reducing herbivory (i.e.herbivore abundance) (Fig. 3).

Aboveground-belowground compartments
The effects of detritivore identity on EMF were predominately realized via effects in the aboveground compartment,

L i t t o r o p h i l o s c i a O r c h e s t i a M e l a m p u s L i t t o r a r i a
C o n t r o l

L i t t o r o p h i l o s c i a O r c h e s t i a M e l a m p u s L i t t o r a r i a
C o n t r o l  and comparatively few effects were observed for belowground functions or by including combinations of functions from both aboveground and belowground compartments.Moreover, and variability in EMF dampened with increasing number of functions averaged (Fig. 4, Table 2).

Detritivore density
Littorophilosica populations maintained near initial densities in absence of spider predation (1157 ± 392 individuals m −2 ; mean ± SE, n = 9), and were limited when Pardosa predators were present (424 ± 278 individuals m −2 ; mean ± SE, n = 9).Orchestia populations maintained lower population densities overall, and were only weakly suppressed by the presence of spiders (524 ± 36 individuals m −2 ; mean ± SE, n = 9) compared to the spider-free control (601 ± 52 individuals m −2 ; mean ± SE, n = 9).All snails survived but none reproduced during the experiment, which resulted in final densities that were consistent with the initial densities stocked in the microcosms.

Discussion
Our results show that changes in detritivore identity influence multiple aspects of ecosystem functioning.Our model communities exemplify three typical pathways by which detritivores' influences on EMF can arise.First, prolific detritivory by Melampus enhanced decomposition, and conceivably litter mineralization released nutrient resources that could be available for rapid plant uptake and growth.We did not, however, observe changes in the standing stocks of nutrients in the presence of Melampus.This could have been because we chose to assess all six ecosystem functions at the same time, rather than reporting maximum possible influence of detritivores on all ecosystem functions that respond at different rates.That is, nitrate could have been rapidly mobilized within hours or days, whereas effects on predator population growth could be realized only after a full year.Furthermore, the slight trends toward increased plant root and shoot biomass in the presence of Melampus were not significant when quantified as individual functions, but they contributed to increased multifunctionality when considered in an aggregated metric (EMF) together with enhanced decomposition.This suggests that multifunctionality may be useful for capturing cumulative effects on ecosystem functioning despite temporal variation of effects for individual functions.Second, detritivorous Littorophiloscia served as alternative prey and moderated predator-herbivore interactions.Yet, the observed positive effects on predation and herbivory (measured as predator and herbivore abundance) did not coincide with enhanced Table 2. Mixed model ANOVA results show the main and interactive effects of detritivore identity, predator addition and number of functions included in ecosystem multifunctionality (EMF) calculated using aboveground functions, belowground functions or mixtures of aboveground and belowground functions ( average ecosystem multifunctionality, demonstrating limitation in the utility of the multifunctionality metric.Third, as strong omnivorous grazers Littoraria reduced ecosystem multifunctionality by directly consuming multiple resources (i.e.shoots and litter), and we observed reduced abundances of the consumer communities that rely on those resources (i.e.herbivores and predators).Considered together, these results corroborate findings that many detritivores are influential omnivorous species (Hines and Gessner 2012, Mulder et al. 2013, Wolkovich et al. 2014), and suggest that multi-trophic data is needed to draw generalities regarding the influence of detritivores on ecosystem functioning.
Although our results confirm a number of previously reported feeding interactions among the species in our model communities, they also extend several species' ranges of influence across compartments (aboveground-belowground), and in doing so they refute some prior claims of high functional redundancy among detritivores.This finding is consistent with the central claim made by proponents of ecosystem multifunctionality (Byrnes et al. 2014).Namely, expanding the number of functions measured will increase the likelihood of quantifying a unique effect of any given species (Isbell et al. 2011, Meyer et al. 2018).In our analysis, however, averaging an increasing number of functions clearly dampened variation and altered the magnitude of multifunctionality responses, because there are several possible combinations of one or two functions, but only one possible combination that includes all six measured functions in average EMF.This numerical artifact of the average multifunctionality metric has been critiqued openly for obfuscating results (Bradford et al. 2014).Whether it is an attribute, or an artifact, is debatable, and such disputes of multi-function combinations in this response variable echo critiques of multi-species combinations as independent variables in classic biodiversity ecosystem function (BEF) experiments (Huston 1997).For decades, controversy surrounded whether BEF relationships were a real consequence of biodiversity change or whether they were an artifact of classic BEF experimental design that did not account for 'sampling effects' (i.e. a greater chance of including a high functioning species in more diverse mixtures) (Huston 1997, Hooper et al. 2005, Eisenhauer et al. 2016, Wardle 2016).To resolve that debate, post hoc statistical methods were used to determine if a particular species was driving effects (i.e.'sampling effects'), or if synergistic effects stemmed species interactions (i.e.complementarity or facilitation) (Loreau 1998, Cardinale et al. 2007).Employing somewhat parallel rationale, Byrnes et al. (2014) suggested using threshold values to test if any particular function is driving results, and Manning et al. (2018) suggested that better grouping and naming of functions would be useful to explain which functions would interactively influence ecosystem multifunctionality.Both suggestions focus on number and strength of component processes rather than how they connect and feedback to influence one another.Interpretation of our results, however, hinge upon consideration of trophic interactions, suggesting that a food web, or more systems-based perspective, will provide a useful complement to multifunctionality metrics (Hines 2019).
Indeed, our analysis is part of a convergence of biodiversity-ecosystem function (BEF) research and trophic ecology (Hines et al. 2015b, Wang andBrose 2017).Our results show that detritivores' influences operate across trophic levels, and that the magnitude and the direction of observed effects depend on the presence or absence of predators.Therefore, our findings contribute to a classic debate in trophic ecology about the occurrence of trophic cascades.For many years, scientists argued about whether, and when, predators indirectly benefited plants by consuming herbivores (Schmitz et al. 2000, Shurin et al. 2002, Borer et al. 2005).The outcome of this debate has served as major impetus for predator conservation in agricultural (Tscharntke et al. 2007), and wild ecosystems (Ripple et al. 2014, Stier et al. 2016).Similar to our study, this research has expanding evidence for the ecosystem consequences of direct and indirect species interactions to a wider variety of ecosystems and facets of biodiversity (e.g.species identity and trait-based effects) than some classic BEF research that predominately focused on plant species richness in grasslands and forests.Without overlooking the importance of predators, our results also support the primacy of detritivores as species that can influence multiple ecosystem functions by simultaneously influencing green and brown energetic pathways.An important implication is that species interaction patterns may be more informative than trophic level for establishing conservation priorities.The interpretation that interaction patterns explain important variation in our ecosystem multifunctionality data, is consistent with that of Lefcheck et al. (2015) who suggest that omnivory and multi-channel feeding underpin their finding that herbivores had a stronger diversity-multifunctionality relationships than plants or dead organic matter.Yet, Lefcheck et al. (2015) found remarkably few suitable studies reporting ecosystem consequences of changes in detritivore diversity, highlighting the importance of original empirical studies like this one.Notably, experiments documenting connections among aboveground and belowground sub-systems have historically reported multiple ecosystem responses (Wardle 2002), but these studies did not specifically quantify ecosystem multifunctionality as reported here.
In our experiment, we test the consequences of change in species identity rather than change in species richness.Our results from single-detritivore species do not allow us to evaluate whether co-existing detritivore species would compete with or facilitate each other, and hence whether changes in detritivore-species richness would modify the influences on multiple ecosystem functions that we report.This reflects a recent initiative to emphasize that multiple types of diversity change confront ecosystems, including changes in species traits, dominance and identity (Hillebrand et al. 2018).Despite the prevalence of species turnover in ecosystems around the world (Dornelas et al. 2014, Blowes et al. 2019), the ecosystem consequences of turnover within a given species richness remain an overlooked component of many BEF experiments (Eisenhauer et al. 2016).In the present study, we kept detritivore-species diversity low, which may maximize our ability to detect species-specific influences on ecosystem functioning.
Notably, environmental change often drives changes in biodiversity, whether it be changes in species richness, or species identity.However, we did not impose an environmental change treatment to select for our detritivore communities.Those drivers would likely alter the magnitude of species influence on any given function as well as the strength of feedbacks between functions (Hines andGessner 2012, Hines et al. 2015a).In our analysis, differences among species were realized as impacts on auxiliary ecosystem functions in addition to species' primary functional role.This finding is likely to hold true for spatial and temporal turnover of many organisms in ecosystems around the world.However, we were not able to attribute species trait dissimilarity to dissimilarity in ecosystem functioning, highlighting the challenges in predicting how diversity change will influence ecosystem functioning.Moreover, we conducted our experiment in closed microcosms which prevent us from assessing the influence of dispersal (i.e.immigration and emigration) on species interactions.Nonetheless, our model communities capture realistic feeding rates and local population dynamics (i.e.invertebrate reproduction and mortality) during the main growing season for several of the major players in Atlantic coastal salt marshes, allowing us to document some baseline influences of detritivore species on multiple ecosystem functions.Notably, individual ecosystem functions need not respond simultaneously, and less is known about timing of belowground interactions compared to aboveground phenology (Eisenhauer et al. 2018).If duration of root 16000706, 2021, 10, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/oik.08333by Universitaetsbibliothek Leipzig, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License production was prolonged or delayed compared to shoot production, examination of an extended belowground growing season may reveal hidden ecosystem functioning that was not detected in this experiment (Abramoff andFinzi 2014, Blume-Werry et al. 2015).
In conclusion, salt marsh detritivores and predators are highly sensitive to environmental variation (Hines et al. 2015a), and we quantify three important ways changes in detritivore species assemblages can influence multivariate ecosystem responses.Our results show that changes in detritivore species composition are likely associated with multiple functional pathways by which detritivores influence aboveground food web interactions.This is especially important when considered in light of mismatches in aboveground and belowground biodiversity patterns (Cameron et al. 2019).That is, although species in the aboveground food web may coexist across environmental gradients in soil conditions, the strength and consistency of their interactions, and their extended influence on multiple ecosystem functions may be contingent upon the identity of the dominant detritivore species.If multifunctionality metrics are used to assess whole ecosystem responses to changes in consumer community composition, then evaluation of food web interactions provide relevant mechanistic underpinnings needed to evaluate responses.

Figure 3 .
Figure 3. Direct and predator mediated effects of changes in detritivores on average ecosystem multifunctionality based on standardized values of six variable that represent key ecosystem functions above and below the ground: plant shoot biomass, herbivore abundance, predator abundance, plant root biomass, decomposition, nutrient concentration (NO 3 − ).

Figure 4 .
Figure4.The effects of food web compartment (aboveground, belowground or mixed) and number of ecosystem functions (one to six) on the relationship between detritivore identity and ecosystem multifunctionality when predators are present (black circles) and absent (white circles).
or non-consumptive effects of predators(Peckarsky et al. 2008).That is, herbivory was considered as ability of the ecosystem to support two generations of 16000706, 2021, 10, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/oik.08333by Universitaetsbibliothek Leipzig, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

Table 1 .
Mixed model ANOVA results show the main and interactive effects of the detritivore identity and predator-addition treatments on ecosystem multifunctionality (EMF) derived from the six individual functions (herbivore abundance, shoot biomass, predator abundance, litter biomass, root biomass and NO 3 − concentrations) measured in common garden microcosms after 60 days exposure to treatments.Significant p-values (p < 0.05) appear in bold for clarity., 2021, 10, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/oik.08333by Universitaetsbibliothek Leipzig, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License *df show degrees of freedom (numerator, denominator) for all response variables except predator abundance which is 4, 32.16000706 Table 1 for compartment designations of individual functions).Significant p-values (p < 0.05) appear in bold for clarity., 2021, 10, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/oik.08333by Universitaetsbibliothek Leipzig, Wiley Online Library on [05/04/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License *Numerator degrees of freedom.Denominator degrees of freedom approximated with Saitterthwaite's method.16000706