Trophic complexity alters the diversity–multifunctionality relationship in experimental grassland mesocosms

Abstract Plant diversity has a positive influence on the number of ecosystem functions maintained simultaneously by a community, or multifunctionality. While the presence of multiple trophic levels beyond plants, or trophic complexity, affects individual functions, the effect of trophic complexity on the diversity–multifunctionality relationship is less well known. To address this issue, we tested whether the independent or simultaneous manipulation of both plant diversity and trophic complexity impacted multifunctionality using a mesocosm experiment from Cedar Creek, Minnesota, USA. Our analyses revealed that neither plant diversity nor trophic complexity had significant effects on single functions, but trophic complexity altered the diversity–multifunctionality relationship in two key ways: It lowered the maximum strength of the diversity–multifunctionality effect, and it shifted the relationship between increasing diversity and multifunctionality from positive to negative at lower function thresholds. Our findings highlight the importance to account for interactions with higher trophic levels, as they can alter the biodiversity effect on multifunctionality.


| INTRODUC TI ON
Biodiversity has a positive effect on the magnitude of individual ecosystem functions and on the number of functions an ecosystem maintains simultaneously (multifunctionality) Hector & Bagchi, 2007;Hooper et al. 2005;Liang et al. 2016;Zavaleta et al. 2010). Empirical support for this diversity-multifunctionality relationship, across taxa and habitats, suggests that higher levels of biodiversity may be necessary to maintain ecosystem functioning than previously assumed based on single-function studies (Cardinale et al. 2012;Hooper et al. 2005;Lefcheck et al. 2015). Moreover, an increase in total biodiversity in an ecosystem often corresponds with an increase in trophic complexity, which can then alter ecosystem functioning (Haddad et al. 2009;Soliveres et al. 2016). However, the role of trophic complexity in influencing how biodiversity mediates multifunctionality is less well understood.
Nonproducer trophic levels (e.g., litter decomposers, herbivorous insects) can have positive or negative impacts on various ecosystem functions simultaneously (Dyer & Letourneau, 2003;Estes et al. 2011;Naeem et al. 1994Naeem et al. , 1995Schmitz, 2006;Strickland et al. 2013;Tiffin & Ross-Ibarra, 2014), leading to complex effects on multifunctionality. Depending on the functional role of the group and the ecosystem function considered, nonproducer trophic levels can either directly enhance or reduce the magnitude of function. For example, aboveground herbivores can decrease aboveground plant biomass, while aboveground predators nity to test the impacts of trophic complexity on the diversitymultifunctionality relationship. The study is a fully factorial design with 5 plant diversity treatments (1, 2, 4, 8, and 16 spp.) crossed with 4 trophic complexity treatments that represent four basic levels: a close to natural community (plants + litter fauna + aboveground mesofauna), two communities with half the functional groups absent (plants + litter fauna and plants + aboveground mesofauna), and a null community with plants alone. We note that species diversity within each level of trophic complexity could potentially affect ecosystem functioning and multifunctionality, but our experimental design cannot disentangle these effects.
Four ecosystem functions were measured: aboveground biomass, belowground root biomass, soil water retention, and biomass recovery after harvest in the following year.
We hypothesized that trophic complexity may impact multifunctionality and the shape of the jack-of-all-trades curve in two ways.
First, trophic complexity will affect the slope of the relationship between plant diversity and multifunctionality (i.e., the biodiversitymultifunctionality (BMF) effect) at different levels of ecosystem functioning, consequently resulting in vertical shifts in the jack-ofall-trades curve ( Figure 1). This would occur because trophic complexity could alter the magnitude of individual ecosystem functions either directly, through independent effects on functioning, or indirectly, through interspecific interactions that affect BEF curves.
Second, trophic complexity will horizontally shift in the jack-of-alltrades curve, especially at the inflection point where it crosses the x-axis ( Figure 1). This pattern could occur through the impact trophic complexity has on correlations between individual ecosystem functions in a community and the values of functions at which diversity F I G U R E 1 Conceptual framework illustrating hypothesized effects of trophic complexity on the biodiversity-multifunctionality effect (BMF) curve. The BMF is a measure of the slope of the relationship between diversity and multifunctionality (e.g., number of functions gained through the addition of species) whose value is dependent on the threshold (i.e., percent ecosystem function obtained) used in estimating multifunctionality. Positive effects of diversity correspond to positive BMF values, or a curve above zero and vice versa. The continuous black curve represents a hypothetical relationship between selected threshold value and the BMF for a plant community in the absence of trophic complexity (i.e., the plant only curve). The red and blue lines represent possible deviations from the plant-only curve with the addition of trophic complexity. The red curves represent trophic-induced changes to diversity effects on single ecosystem functions or BEF, which alter the flatness of the BMF curve. The blue curves represent trophicinduced changes to correlations between traits, which shifts the horizontal location of the BMF switch from positive to negative has a positive effect on multifunctionality. These potential impacts of trophic complexity therefore result in distinguishable, independent and nonmutually exclusive, shifts in the jack-of-all-trades curve (Heilpern et al. 2020). To test these hypotheses, we model the effects of trophic complexity on (i) individual functions, (ii) multifunctionality at multiple thresholds, and (iii) the biodiversity-multifunctionality (BMF) effect (the gain in number of ecosystem functions maintained above the given value with one additional species) across the range of possible thresholds.

| Experimental methods
Tall-grass prairie mesocosms were established adjacent to the bio-  (Table S1) were native perennial species used in previous experimental studies from the site (Seabloom et al. 2017;Tilman et al. 2001Tilman et al. , 2006. Pots with incomplete data on species identity were excluded from our analyses, resulting in a sample size of 94. The plant diversity treatments were crossed with trophic complexity treatments, such that pots included plants and aboveground meso- Following previous studies, these treatments were achieved by first applying a pesticide treatment on all the pots, which initially contained only sterilized local soil and litter, removing all fauna (Seabloom et al. 2017;Tilman et al. 2012). The pesticide used was esfenvalerate (DuPont™ Asana® XL), a natural pyrethrin insecticide, known to have no nontarget effects such as phytotoxicity or fertilization (DuPont, 2002;Mitchell, 2003). Treatments with aboveground fauna were then inoculated with identical sets of invertebrates that included all species obtained from sweep-netting adjacent vegetation. To control for biomass effects, equivalent sets of frozen, killed invertebrates were added to the treatments without aboveground fauna. Removal of aboveground fauna in appropriate treatments was ensured by monthly pesticide treatments, and a pesticide control was applied to the other plots by spraying equal volumes of water. Similarly, treatments with litter fauna were inoculated with active leaf litter each month (mesh bags filled with 40 g of leaf litter collected from the surrounding field placed on the ground near the mesocosms for a minimum of two weeks). Controls that included autoclaved leaf litter and litter fauna were simultaneously added to treatments with no litter fauna.
Appendix Figure S1 summarizes the counts and community compositions of these treatments.
The experiment was established in 1999 and seeded in spring 2000, and then run for a year. In July 2001, the experiment was ended, plants harvested, and aboveground biomass, root biomass, and soil water retention were measured. Further, in 2002, following natural recruitment, the biomass in each pot was measured to assess recovery.

| Ecosystem function measurements
Four ecosystem functions were analyzed. In 2001-after one year of the experiment-in each pot, we measured (i) total aboveground biomass, (ii) root biomass (i.e., the total belowground biomass), and (iii) water retention (quantified as the time taken for a fixed volume of water to flow into a collection flask at the bottom of the pot).
In 2002, following the harvesting of both aboveground and root biomass, we characterized a fourth ecosystem function: biomass recovery, which was characterized as the total recovered biomass in a pot after disturbance (the removal, in 2001, of the aboveground and root biomass). These four functions were chosen as they represent key properties of ecosystem function and have potential links to trophic complexity. Moreover, we also chose them for low correlations and thus independent contributions to multifunctionality.
Pairwise correlations between these show overall low correlations between the functions at the plot level (Appendix Figure S3), with aboveground biomass and water retention most highly correlated (r = 0.32).

| BEF curves
All analyses were performed using the statistical software R, version 3.4.4 (Team, 2009). The response of each ecosystem function to manipulated plant diversity and trophic complexity treatment was analyzed as a log-linear model using generalized linear mixed-effects models in the package lme4 (Bates et al. 2015). The likelihood of the full model: was compared using stepwise selection against the likelihoods of reduced models that did not include the interaction term of plant species richness with trophic complexity, and a simple model that did not include trophic complexity, using AIC values.

| Single ecosystem functions (BEF)
In generalized linear models of ecosystems functions, plant species diversity did not have a significant impact on any ecosystem functions measured (Figure 2). In simple models of ecosystem function against plant diversity alone (ecosystem function 1 ~ log(plant richness)), plant diversity did not have a significant effect on root biomass (R 2 = 0.0, p =.9), water retention (R 2 = 0.02, p =.13), or biomass recovery (R 2 = 0.0, p =.75), but we observed a significant positive saturating effect on aboveground biomass (log(plant richness) coef-ficient=−0.0017, R 2 = 0.12, p <.001). The total effects of species richness were low, and the average values of each function remained within a small range of values across different plant diversity treatments (Appendix Figure S2).
In the full model with plant diversity, trophic complexity and their interaction (ecosystem function 1 ~ log(plant richness)+trophic complexity + log(plant richness):trophic complexity), trophic complexity effects on single ecosystem functions were largely nonsignificant, except the litter mesofauna treatment for aboveground biomass (Figure 2). Further, when the effect of trophic complexity was removed using stepwise selection, the best-fit models for each ecosystem function did not include trophic complexity as a predictor; the simplest model with only plant diversity as the predictor had the lowest AIC.

| The jack-of-all-trades curve
The BMF increased and peaked at moderate thresholds, switching to a negative at high thresholds for all 4 treatments, following predic- Trophic complexity had an effect on both the height and location of the peak BMF (Figure 4). Across all thresholds, we found that treatments with at least one additional trophic level had consistently higher BMF values than the plant-only treatment. Among the three complexity levels, treatments with aboveground mesofauna had a higher peak BMF than the treatment with litter fauna as the only additional level. Moreover, the plant-only and plants + litter mesofauna treatments had no clear peak, while the other two treatments peaked F I G U R E 3 Number of functions above four different thresholds, indicated in the top right corner of each panel, against the number of species in the plot. Lines represent linear model fits for pooled data (black) and each treatment (colors). Legend shows color codes for treatments; only plants (NONE, green), plants and aboveground mesofauna (ABV, yellow), plants and litter mesofauna (LIT, brown), and plants and both aboveground and litter mesofauna (BOTH, gray). Actual data points for each plot represented as gray dots at similar intermediate thresholds (~ 35%-50%). We also found that treatments with one additional level of trophic complexity transitioned from positive to negative BMFs at higher thresholds ~ 70% -90% thresholds, than the plant-only treatment (shift ~ 50% -60%).
Although at high thresholds, differences among treatments with an additional trophic component were less detectable, each remained distinct from the treatment with plants alone.

| D ISCUSS I ON
We find that trophic complexity affects the relationship between biodiversity and ecosystem multifunctionality in two ways. First, the strength of the BMF effect is different across the spectrum of levels of ecosystem function, depending on trophic complexity; treatments with additional trophic levels beyond plants had higher strengths of BMF across thresholds (Figures 3, 4). Second, the shapes of the jack-of-all-trades curves were strikingly different, staying positive for higher thresholds in treatments with more than one trophic level ( Figure 4). Together, these findings are indicative of pervasive impacts of trophic complexity on the relationship between biodiversity and ecosystem multifunctionality.
Trophic complexity may alter BMF effects by changing the magnitude of biodiversity effects on individual ecosystem functions; this relationship has been observed in grassland communities similar to ours (Lefcheck et al. 2015;Soliveres et al. 2016). Interestingly, our analyses did not reveal significant impacts of trophic complexity on single ecosystem functions as modeled by BEF relationships (Figure 2, Appendix Figure S2), in contrast with studies on similar landscapes (Soliveres et al. 2016). However, functional groups within mesofauna that were not distinguished in this study (e.g., aboveground herbivores and predators grouped as aboveground mesofauna) could have opposing impacts on individual ecosystem functions. This could make it difficult to disentangle the positive and negative effects of these different groups on multifunctionality. Moreover, the effects of nonproducer trophic levels on plant communities and ecosystem functioning could be latent, delayed, or accruing over time and hence difficult to detect in short-term manipulations (Maguire et al. 2015;Root, 1996).
Although we did not observe significant impacts of trophic com- against treatments with plants alone, possibly leading to the observed significant effects. In addition to BEF mechanisms of complementarity and selection, plant diversity is observed to decrease herbivory damage in natural systems (Baraza et al. 2006;Hambäck et al. 2014). This indirect effect of herbivory on plant biomass could also potentially explain the amplified effect of plant diversity on multifunctionality that we observe in the presence of these trophic groups. Thus, it is possible that changes in BEF driven by trophic complexity have resulted in the observed shifts in the BMF curve ( Figure 4), although we find no evidence to support this.
Rather, given the impacts of trophic complexity on multifunctionality but the absence of detectable effects on single functions, our results suggest that trophic interactions could mediate BMF by altering trait correlations among plant species, a mechanism observed through numerical simulations (Heilpern et al. 2020). Although independent frameworks to assess identity effects of individual species and environmentally linked intraspecific trait variation in species for BEF and multifunctionality have been proposed (Laughlin, 2014;Meyer et al. 2018), the role of induced trait variation and shifts in function correlations through biotic mechanisms is less explored in F I G U R E 4 Effect of ecosystem-function threshold on the biodiversity-multifunctionality effect (the BMF). Each point represents the slope (i.e., strength) of the relationship between plant species richness and number of functions above the threshold when estimated using a linear regression model. Each BMF curve for each level of trophic complexity is plotted using a different color, as presented in the key (top right); only plants (PLANT ONLY, green), plants and aboveground mesofauna (ABV, yellow), plants and litter mesofauna (LIT, brown), and plants and both aboveground and litter mesofauna (BOTH, gray). The curves are smooth-spline interpolations. The dashed line represents the mean slope, and the beige polygon represents the bounds of the standard error of the slope in the pooled dataset the context of multifunctionality. Our experiment was not designed