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

Diversity within trophic levels influences the number of ecosystem functions maintained simultaneously by a community, or multifunctionality. Depending on threshold cutoffs applied to measuring these functions, the diversity-multifunctionality relationship changes from positive until intermediate thresholds to negative effect at high function thresholds. Although the presence of multiple trophic levels or trophic complexity affects levels of functions, its effect on the diversity-multifunctionality relationship has not been experimentally tested. We simultaneously manipulated plant diversity and trophic complexity in a multifactorial tall-grass prairie mesocosm experiment at Cedar Creek, Minnesota, USA and measured multiple ecosystem functions. Trophic complexity altered the diversity-multifunctionality relationship in two key ways: it lowered the maximum strength of the diversity-multifunctionality effect and it resulted in a switch from positive to negative relationship between increasing diversity and multifunctionality at lower function thresholds. Our findings suggest that global declines in trophic complexity will exacerbate the reduction in ecosystem multifunctionality as a result of widespread declines in biodiversity.


Introduction 32
Increasing biodiversity within a trophic level, both along experimental and naturally occurring 33 gradients, is associated with a positive and saturating increase in the magnitude of ecosystem functions 34 when considered individually and often a monotonic increase in the number of functions maintained 35 simultaneously 1-5 . Empirical support for this diversity-multifunctionality relationship, across taxa and 36 habitats, suggests that higher levels of biodiversity may be necessary to maintain ecosystem 37 functioning than previously assumed based on single-function studies 2,6,7 . 38 39 Multifunctionality of ecosystems is sensitive to several factors, prominently: (1) the levels of 40 biodiversity within trophic levels 4,7 and (2) a desired "threshold" magnitude of each ecosystem 41 function scaled from zero to 100% (above this level, a community is considered to maintain that 42 particular function and contribute to multifunctionality) 7-9 . When low thresholds are considered (e.g., 43 maintining at least 10% of every function measured) diversity generally increases the number of 44 functions maintained above that threshold; a positive diversity-multifunction relationship. But this 45 effect of diversity on multifunctionality, or slope of this relationship between number of ecosystem 46 functions maintained and diversity, is expected to increase with the selected threshold only until 47 moderate threshold values, then decreases and switches to a negative diversity-multifunction 48 relationship at high thresholds. This pattern of changes in the diversity-multifunctionality relationship 49 is referred to as the "jack-of-all-trades" effect 8 (Fig 1). In effect, community average values for 50 functions are expected to be low in diverse communities compared to high functioning monocultures, 51 allowing them to maintain most functions, but only at the average value contributed by each species. 52 53 Most diversity-function studies and tests of this pattern have focussed on a single trophic level, most 54 often the plant or producer community, but changes in diversity of non-producer trophic levels can also 55 have significant impacts on several ecosystem functions 10-16 . Plant diversity can affect diversity and 56 community composition of both aboveground and litter fauna which by removing biomass and through 57 decomposition, can then affect ecosystem functioning. However, non-producer trophic levels can either 58 enhance the effect of plant diversity on an ecosystem function 17 (e.g. nutrient uptake) or reduce it 18 59 (e.g. primary productivity). The number of trophic levels, or trophic complexity, can thus potentially 60 impact the diversity-multifunctionality relationship, but this effect is poorly understood. Current 61 studies do not allow for a direct comparison of ecosystems that vary in trophic complexity because 62 such explorations would require simultaneously manipulating plant diversity as well as the number of 63 non-producer trophic groups. 64 65 Here, we explore the effects of diversity and trophic complexity on ecosystem multifunctionality using 66 an experimental approach. In the absence of any effect by non-prodcuer trophic groups, we hypothesise 67 that plant diversity would increase multifunctionality until moderate thresholds and decrease 68 multifunctionality at high thresholds, resulting in a jack-of-all-trades relationship. We hypothesise that 69 trophic complexity could have impacts on multifunctionality at two different levels; (1) altered 70 diversity-function relationships for single functions -the number and identity of non-producer trophic 71 levels affects each ecosystem function distinctly, either strengthening or dampening the effect of plant 72 diversity which aggregates to strengthened or dampened multifunctionality effects and (2) non-73 producer trophic level affect plant traits at the species level altering correlations between measured 74 functions. If trait trade-offs are reduced in the presence of multiple trophic levels, communities would 75 maintain multiple functions for higher than moderate thresholds and vice versa. This shift the threshold 76 at which the diversity-multifunctionality effect switches from positive to negative (Heilpern et al. in 77 review). Each of these effects can be measured on the jack-off-all-trades curve, or the relationship 78 between the strength (slope) of the diversity-multifunction effect (DME) and the selected threshold for 79 measuring multifunctionality. Based on the hypotheses, we expected two corresponding effects of 80 trophic complexity to this curve: (1) a change in the magnitude of the DME across thresholds, resulting 81 in either a taller or flatter curve (2) a change in the threshold at which DME shifts from positive to 82 negative effect measured along the horizontal location on the x-axis. Figure 1 provides a conceptual 83 framework for these predicted outcomes. We note that these predicted responses to trophic complexity 84 are not mutually exclusive; the curve flatness, shift location, both, or neither may respond to 85 differences in trophic complexity. 86

87
To test this framework, we simultaneously manipulated tall-grass prairie plant diversity and trophic 88 complexity in 94 tall-grass prairie mesocosms at Cedar Creek, Minnesota, USA. In these mesocosms, 89 we varied plant diversity from 1 to 16 species, following a standard, stratified log 2 randomised design. 90 We simultaneously varied trophic complexity following a factorial design resulting in 4 trophic 91 treatments; above-ground insect-dominated mesofaunal communities only (INS), below-ground litter 92 mesofaunal communities only (LIT), both aboveground and litter mesofauna (BOTH) or no non-93 producer trophic levels (NONE). For comparison, we pooled all the data (POOLED). 94 95 We measured four ecosystem functions; aboveground biomass, belowground biomass, soil water 96 retention and biomass recovery after harvest. We calculated the mean values of these functions across 97 the different plant diversity and trophic complexity treatments. We standardised the values of each of 98 the functions between 0 and 100 for the entire dataset and calculated a combined multifunctionality 99 metric, the number of functions maintained above a given threshold for each community across this range of thresholds following the standard approach 4,8 . We then analysed the diversity-101 multifunctionality effect (the DME) as the slope of the linear fit of the number of functions maintained 102 above the threshold against plant diversity. Finally, we tested whether this relationship was sensitive to 103 trophic complexity. 104 105

106
For every function measured, communities on average, independent of diversity and trophic complexity 107 treatments, had low values; most communities failed to maintain functions at high values (Fig 2). The 108 average values of each function remained within a small range of values across different plant diversity 109 treatments ( Fig S1). Trophic treatment did not alter the curves substantially for any single function, 110 although there were some differences at high threshold values for water retention and biomass recovery 111 (Fig 2) and at high diversity treatments for water retention and aboveground biomass (Fig S1). 112 113 Ecosystem multifunctionality showed the predicted pattern to changes in plant biodiversity, with 114 positive effects at low thresholds and negative effects at high thresholds (Fig. 3). The plant diversity-115 multifunctionality relationship for the pooled dataset was positive for a threshold of 25% (slope=0.03, 116 p=0.36) and 50% (slope=0.007, p=0.81) but was negative for 75% (slope=-0.14, p=0.4) and 90% 117 thresholds (slope=-0.11, p=0.36) (Fig 3). At the 90% threshold, most plots had no functions above 118 cutoff, making a biodiversity or trophic effect less discernible, whereas at 25% threshold, most 119 communities mauntained functions even without the effect of plant diversity or trophic structure. The 120 biodiversity effect on multifunctionality was most observable at moderate thresholds. Similarly, the 121 effect of trophic complexity was significant only at 75% (-0.12, p<0.05) and marginally significant at 122 90% (-0.07, p=0.06) while at a 25% threshold (-0.05, p=0.59) and 50% threshold (-0.14, p=0.16), these 123 differences were not detectable. 124

125
The diversity-multifunctionality effect (DME) is sensitive to trophic complexity. In the pooled dataset, 126 where all treatments were taken together, the DME increased and peaked at moderate thresholds, 127 switching to negative effect at high thresholds, following predictions of the jack-of-all-trades effect 128 (Fig 4). Trophic complexity had an effect on both the height and location of the peak DME (Fig 4). 129 Comparing the curves of the four different treatments, we found that the addition of either the above- In agreement with recent literature 7,8 , our results show that biodiversity is necessary for maintaining 141 multiple functions from low to moderate function threshold values in grassland communities but at 142 higher thresholds, increases in diversity are inversely associated with multifunctionality (i.e., the jack-143 of-all-trades effect) (Figs 3-4). 144 145 We advance understanding of biodiversity and multifunctionality, however, by also showing that this 146 jack-of-all-trades and master-of-none pattern is sensitive to trophic complexity. Although the bottom-147 up effect of plant diversity on herbivore diversity, and potentially ecosystem functions, has been 148 observed in long-term experimental data 17,19-21 , our study tests the top-down (aboveground fauna) and 149 donor-controlled (litter fauna) effects of trophic complexity on multifunctionality. The simultaneous presence of aboveground and litter fauna trophic levels, the condition closest to natural systems where 151 trophic complexity is common, showed a switch to negative diversity-multifunctionality effects 152 (DMEs) at a very low threshold and maintained persistent negative DMEs throughout most thresholds 153 values (Fig. 4). Any reduction in trophic complexity altered the diversity-multifunctionality 154 relationship. Thus, both the location and height of peak DMEs are affected by trophic complexity, as 155 suggested in our conceptual framework (Fig. 1). 156

157
The nature and direction of the trophic impact on DME depends on the trophic component considered. 158 When examined across thresholds, both aboveground and litter fauna amplified the DME at low 159 thresholds but at high thresholds, only aboveground fauna amplified the effect. This suggests that the 160 presence of aboveground arthropods increases the contribution of biodiversity in providing ecosystem 161 multifunctionality at moderate thresholds. This could be because plant diversity is known to decrease 162 associational effects to herbivory damage 22, 23 leading to an additional advantage of diversity in the 163 presence of aboveground fauna. 164

165
In contrast to what we observed for multifunctionality, our analyses did not reveal impacts of trophic 166 complexity on any single ecosystem function when aggregated across plant diversity treatments (Figs. 167 2-3). Further we did not observe any impacts of the trophic complexity treatments on average 168 ecosystem function value at any level of plant diversity (Fig S1). Although the treatment with plants 169 alone performed better at water retention along all thresholds (Fig. 2) and intermediate diversities (Fig  170   S1), this observation alone cannot explain the differences between the treatments. 171

172
Despite the lack of significant effects on single ecosystem functions, we observe trophic complexity 173 effects at the multifunctionality level. Our results also show that the presence of multiple trophic 174 components decreases the magnitude of the DME across all thresholds, consistent with results from a long-term experiment from the same region demonstrating that non-producer trophic components 176 obscure the diversity-productivity relationship due to a loss in complementarity effects with the 177 addition of herbivores 17 . As the plants we used were taxonomically and functionally quite distinct, 178 identity effects 24 , or unique contributions of individual species to the ecosystem functions considered, 179 are also likely to be responsible for the nature of the biodiversity-multifunctionality relationship we 180 To assess multifunctionality across diversity treatments, measurements of four ecosystem functions -227 aboveground biomass, belowground biomass, water retention and biomass recovery after harvest -were 228 chosen and analysed using published methodology of the threshold approach 18 . To this end, the 229 maximum value for each ecosystem function was calculated as the mean of the five highest function 230 values in the entire experiment. Each ecosystem function in a pot was then standardised between this 231 maximum and the minimum value in the experiment. For every 5% between 0 and 100, each pot was 232 scored for the number of functions maintained above that threshold. For each threshold, the slope of 233 linear model between the number of functions maintained above threshold and the manipulated plant 234 diversity in the community was defined as the diversity effect on multifunctionality (DME). DME was 235 analysed for the pooled dataset as well as the dataset split into the four trophic complexity treatments. 236 The magnitude of the peak and the point at which the curve of biodiversity effect vs. threshold crosses 237 the x-axis for each of the treatments were examined in comparison with the pooled data for reasons 238 described in the introduction. 239 240