Proceedings of the Royal Society B: Biological Sciences
You have accessResearch articles

Functional and phylogenetic relationships link predators to plant diversity via trophic and non-trophic pathways

Jing-Ting Chen

Jing-Ting Chen

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

College of Biological Sciences, University of Chinese Academy of Sciences, Beijing, People's Republic of China

Contribution: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing

Google Scholar

Find this author on PubMed

,
Ming-Qiang Wang

Ming-Qiang Wang

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

Key Laboratory of Mountain Ecological Restoration and Bioresource Utilization & Ecological Restoration Biodiversity Conservation Key Laboratory of Sichuan Province, Chengdu Institute of Biology, Chengdu, People's Republic of China

[email protected]

Contribution: Conceptualization, Data curation, Formal analysis, Methodology, Supervision, Writing – original draft

Google Scholar

Find this author on PubMed

,
Yi Li

Yi Li

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Beijing, People's Republic of China

Contribution: Methodology

Google Scholar

Find this author on PubMed

,
Douglas Chesters

Douglas Chesters

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

Contribution: Methodology

Google Scholar

Find this author on PubMed

,
Arong Luo

Arong Luo

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

Contribution: Methodology

Google Scholar

Find this author on PubMed

,
Wei Zhang

Wei Zhang

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

State Key Laboratory of Biocatalysis and Enzyme Engineering of China, School of Life Sciences, Hubei University, Wuhan, People's Republic of China

Contribution: Data curation, Methodology

Google Scholar

Find this author on PubMed

,
Peng-Fei Guo

Peng-Fei Guo

School of Pharmacy, Guizhou University of Traditional Chinese Medicine, Guiyang, People's Republic of China

Google Scholar

Find this author on PubMed

,
Shi-Kun Guo

Shi-Kun Guo

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

College of Biological Sciences, University of Chinese Academy of Sciences, Beijing, People's Republic of China

Google Scholar

Find this author on PubMed

,
Qing-Song Zhou

Qing-Song Zhou

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

Contribution: Methodology

Google Scholar

Find this author on PubMed

,
Ke-Ping Ma

Ke-Ping Ma

State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Beijing, People's Republic of China

Contribution: Methodology

Google Scholar

Find this author on PubMed

,
Goddert von Oheimb

Goddert von Oheimb

Technische Universität Dresden, Institute of General Ecology and Environmental Protection, Pienner Straße 7, 01737 Tharandt, Germany

Google Scholar

Find this author on PubMed

,
Matthias Kunz

Matthias Kunz

Technische Universität Dresden, Institute of General Ecology and Environmental Protection, Pienner Straße 7, 01737 Tharandt, Germany

Contribution: Data curation

Google Scholar

Find this author on PubMed

,
Nai-Li Zhang

Nai-Li Zhang

The Key Laboratory for Silviculture and Conservation of Ministry of Education, Beijing Forestry University, Beijing, People's Republic of China

Google Scholar

Find this author on PubMed

,
Xiao-Juan Liu

Xiao-Juan Liu

State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Beijing, People's Republic of China

Contribution: Methodology

Google Scholar

Find this author on PubMed

,
Helge Bruelheide

Helge Bruelheide

Institute of Biology/Geobotany and Botanical Garden, Martin Luther University Halle-Wittenberg, Halle (Saale), Germany

German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Leipzig, Germany

Contribution: Methodology

Google Scholar

Find this author on PubMed

,
Andreas Schuldt

Andreas Schuldt

Forest Nature Conservation, Georg-August-University, Goettingen, Germany

Contribution: Conceptualization, Methodology, Supervision, Writing – review & editing

Google Scholar

Find this author on PubMed

and
Chao-Dong Zhu

Chao-Dong Zhu

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

College of Biological Sciences, University of Chinese Academy of Sciences, Beijing, People's Republic of China

State Key Laboratory of Integrated Pest Management, Institute of Zoology, Chinese Academy of Sciences, Beijing, People's Republic of China

[email protected]

Contribution: Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing – review & editing

Google Scholar

Find this author on PubMed

    Abstract

    Human-induced biodiversity loss negatively affects ecosystem function, but the interactive effects of biodiversity change across trophic levels remain insufficiently understood. We sampled arboreal spiders and lepidopteran larvae across seasons in 2 years in a subtropical tree diversity experiment, and then disentangled the links between tree diversity and arthropod predator diversity by deconstructing the pathways among multiple components of diversity (taxonomic, phylogenetic and functional) with structural equation models. We found that herbivores were major mediators of plant species richness effects on abundance, species richness, functional and phylogenetic diversity of predators, while phylogenetic, functional and structural diversity of trees were also important mediators of this process. However, the strength and direction differed between functional, structural and phylogenetic diversity effects, indicating different underlying mechanisms for predator community assembly. Abundance and multiple diversity components of predators were consistently affected by tree functional diversity, indicating that the variation in structure and environment caused by plant functional composition might play key roles in predator community assembly. Our study highlights the importance of an integrated approach based on multiple biodiversity components in understanding the consequences of biodiversity loss in multitrophic communities.

    1. Introduction

    Human activities and environmental changes are threatening global biodiversity, which plays a key role in the provisioning of ecosystem functioning and services [1]. Previous studies have investigated the impacts of increasing biodiversity loss on ecosystem structure, stability and functioning [2], such as nutrient cycling [3], multi-functionality [4] and decomposition [5]. These studies provided evidence that plant diversity contributes to important changes in ecosystem functioning. Further studies have shown that promoting plant diversity also modifies the biodiversity and ecosystem functions of consumers [6,7], and that plant diversity positively affects both the diversity and trophic interactions at higher trophic levels [811]. Declines in the abundance of arthropods and other consumers have also recently been corroborated [12], because biodiversity loss cascading through higher trophic levels may have repercussions on the functioning of ecosystems [13,14].

    Importantly, effects of plant biodiversity loss have been shown to be most pronounced for herbivores, whereas higher trophic levels showed weaker responses [8,9]. The ‘enemies hypothesis’ has been proposed to explain the relationship between plant diversity and higher trophic levels [15], and posits that predators are more abundant and diverse in speciose plant communities because more stable and variable prey resources are provided by more heterogeneous environments [15,16]. This hypothesis is commonly used to explain the diversity and community assembly mechanisms of generalist predators, rather than specialist predators. While some previous studies have in fact shown that arthropod predator abundance, species richness and their food web stability can be positively affected by plant species richness [8,17], others have found only weak or even negative effects (e.g. [16,18]). This makes it difficult to derive generalizable conclusions on how higher trophic level interactions and their top-down effects on ecosystems might change with plant diversity loss.

    To date, most studies have focused on species richness as a measure of biodiversity, although functional and phylogenetic diversity may better capture the multifaceted concept of biodiversity and provide mechanistic insights into biodiversity relationships across trophic levels [10,11,13,16]. For example, several studies have suggested that the functional diversity of predators plays a key role in prey suppression and has important effects on ecosystem functioning [19,20]; functional diversity captures species' morphological, physiological or behavioural features across communities and therefore allows insights into interactions of organisms with their environment [21]. Phylogenetic diversity reflects evolutionary processes and characters of species and may therefore help to understand the mechanisms driving patterns of distribution and co-occurrence among species [22,23], which is difficult to recognize based only on taxonomic diversity. Previous studies have found significant effects of plant diversity on higher trophic levels with either functional trait- or phylogeny-based approaches [24,25]. Considering both measures at the same time may allow for more comprehensive insights into the effects of biotic filters on community assembly and species coexistence. Recent studies have found that phylogenetic and functional diversity can reflect the high redundancy among members of species-rich communities, which is key to ecosystem stability [26]. Moreover, measures of phylogenetic and functional trait diversity often provide better fits in explaining species distribution and coexistence patterns [22,23,27]. Insights into these additional facets of diversity at higher trophic levels, and their associations with changes in plant diversity, might therefore help to give more mechanistic information in trophic interactions, but such studies are woefully underrepresented in biodiversity–ecosystem functioning researches [16].

    Here, we analyse how taxonomic, phylogenetic and functional diversity of arboreal spiders (species-rich assemblages of a key generalist predator taxon) are directly (assumed to be non-trophic from plants) or indirectly (via herbivore preys) affected by changes in multiple components of plant diversity. We test the effects of different components of tree diversity (taxonomic species richness, phylogenetic, functional and structural diversity) on the diversity components (taxonomic species richness, phylogenetic and functional diversity) of arboreal spiders. In addition, we explore the mediating role of herbivore species richness and abundance to spider communities. We expect that plant species loss affects taxonomic, phylogenetic and functional diversity of spider communities largely via changes in plant phylogenetic and functional diversity. Specifically, we hypothesized that (a) spider abundance and species richness are driven mainly by plant functional diversity, as it represents greater niche opportunities for prey and spiders [28,29]; (b) spider phylogenetic and functional diversity is determined by phylogenetic diversity and functional composition of trees, which are reflective of plant trait similarity and is highly associated with potential prey of spiders (e.g. herbivores). We expect that the phylogenetic indices (especially mean pairwise distance, MPD and mean nearest taxon distance, MNTD) show different effects on spider communities compared to effects of tree species richness, with MPD and MNTD capturing ancient and recent splits in the phylogeny [30], respectively. By contrast, Faith's phylogenetic diversity (PD) shows similar patterns to tree species richness, because of their highly correlated relationship. Moreover, we expected that (c) indirect effects of tree diversity acting via the modification of prey (herbivore) communities play a more important role than direct, potentially non-trophic effects of plants (e.g. habitat structure and microclimate).

    2. Material and methods

    (a) Study sites

    The study was conducted at the ‘BEF-China’ experiment sites, located in Jiangxi Province, southeast of China (29°08′-29°11′N, 117°90′-117°93′E), which is currently the largest tree diversity experiment worldwide [31]. The climate of the study area is subtropical, with a mean annual temperature of 16.7°C and mean annual precipitation of 1821 mm [32]. In this study, we focus on 64 very intensively studied, randomly distributed plots with a tree species-richness gradient (32 plots at each site) from monocultures to 24-species mixtures (16 monocultures, and eight, four, two, one and one mixtures of two, four, eight, 16 and 24 species, respectively, on each study site). We excluded 11 plots because of high mortality of trees, therefore, the data from 53 plots were used for further analyses (see the electronic supplementary material for details). Our sampling completely covered all tree species compositions and all species richness levels of all very intensively studied plots of both sites (electronic supplementary material, table S2).

    (b) Sampling and species identification

    We conducted field surveys for arboreal spiders and lepidopteran herbivores at 53 plots, for a total of 25 440 trees at the BEF-China study sites. Samples were collected six times in both 2017 and 2018 (April, June and September in each year). A beating technique was used to allow for a direct assessment of arthropods at the level of individual trees [10,33]. We knocked down the arthropods from trees onto a beating sheet (1.5 m × 1.5 m) and collected all the fallen arthropods. We sampled all trees from the first row of each plot for a total of 80 living trees in each plot. As spiders and lepidopteran herbivores were the most abundant and diverse functional groups collected, they may be treated as a representative community of predators and herbivores. Lepidopteran larvae were identified by molecular species delimitation (see [10]). For spider species identification, we combined morphological identification and molecular species delimitation. We further constructed a phylogenetic tree of spiders sampled following the method described by Chesters [34]. We used a set of functional traits related to spider survival and physical activity. The electronic supplementary material provides more details on data collection and phylogenetic analysis.

    (c) Plant functional diversity, structural diversity and environmental covariates

    We calculated plant functional diversity by using a range of plant leaf traits which have been proven to influence the diversity and community structure of arthropods (particularly for herbivores) by nutritional condition [9,35]. In addition, to test the potential effects of plant structural diversity on spiders, we used plant vertical structural diversity (VD) to represent the heterogeneity of tree stratification at the plot level. Moreover, four environmental covariates, plot means of elevation, slope, eastness and northness were used in our analysis. Details are provided in the electronic supplementary material.

    (d) Statistical analysis

    All analyses were performed using R v. 4.0.5 (http://www.R-project.org) with ade4, iNEXT, lavaan, MuMIn, picante, FD and vegan. We pooled spiders from the six sampling periods per plot. We included four functional indices of spiders (functional richness, FRic; functional evenness, FEve; functional divergence, FDiv and functional diversity, Rao's Q), and three components of phylogenetic diversity indices (Faith's PD, MPD, MNTD) of both tree and spider communities in our analyses. To exclude the potential effects of species richness on functional and phylogenetic metrics, we also calculated the standardized effect sizes (SES) of functional (FRic.ses, FEve.ses and FDiv.ses) and phylogenetic (PD.ses, MPD.ses and MNTD.ses) diversity with null model. The selection and complementarity effects were calculated using R package partitionBEFsp [36]. The correlation among all predictors and the variance inflation factors (VIF < 3) of all models were tested to make sure that the analyses were not affected by multicollinearity (electronic supplementary material, figure S4).

    (i) Spider abundance and biodiversity

    We assessed the sampling completeness of spiders by calculating the sample coverage with the cumulative number of sampling tree individuals per plot using the R package iNEXT (electronic supplementary material, figure S3).

    To assess whether plant diversity effects can be detected even at higher (not directly associated) trophic levels, we first used linear models with spider data aggregated at the plot level to analyse the correlations between tree diversity and environmental covariables with spider abundance, species richness, functional diversity and phylogenetic diversity. As biotic predictors, tree species richness, tree vertical structural diversity, tree functional diversity based on Rao's Q of the 10 major traits, and tree phylogenetic MNTD were used in the linear models. Because of the high correlation between tree species richness and tree MPD (Pearson's r = 0.78, p < 0.001), we calculated alternative models where we replaced tree species richness with tree phylogenetic MPD, and then selected the best fitting model according to the Akaike information criterion (AICc, corrected for small sample size; [37]). As for abiotic predictors, we used study site, elevation, northness, eastness and slope. We included the interactions between site and tree species richness, tree functional diversity, tree VD and tree phylogenetic MNTD in our models. We used model simplification by excluding predictor variables in an automated stepwise backwards selection based on the corrected AICc values, to obtain the models with the smallest sample sizes with lowest AICc values. Model residuals were checked for normality and homogeneity of variances. Tree species richness, abundance, richness, PD and FRic of spider were log transformed in all analyses to improve normality and homogeneity of the model residuals. All continuous predictors were standardized (mean = 0, standard deviation = 1) prior to the analyses to enable direct comparisons of model estimates.

    (ii) Path models

    To assess the extent to which the effects across trophic levels on spiders are direct (from plants to predators) or indirect (mediated by herbivores), path models were fitted with the R package lavaan, into which herbivore abundance and species richness were added. Tree species richness, tree MPD, tree functional diversity, tree VD, herbivore abundance, herbivore species richness and spider abundance were considered as potential predictors in our models. The abundance and species richness of herbivores and spiders were log-transformed to improve normality. Separate models were fitted for spider species richness, phylogenetic diversity (PD, MPD and MNTD) and functional diversity (FRic, FEve, FDiv and FD Rao's Q).

    For our initial model (electronic supplementary material, figure S14), we assumed that tree species richness as an experimental treatment affects tree phylogenetic diversity, functional diversity and structural diversity. Furthermore, we assumed that all plant-based predictors directly and indirectly (via effects on abundance and phylogenetic diversity) influence herbivore and predator diversity. We expected predator abundance to influence predator species richness and the other diversity metrics (as viable populations of more species can be supported by more individuals; see [38]). Recent studies have indicated that increasing plant diversity creates more resources for more arthropod individuals, and increasing species richness as a consequence of sampling rare species [9,10]. In alternative models, we tested for significant residual covariances between the spider abundance and all of the spider diversity metrics (instead of direct pathways connecting spider abundance and diversity), as abundance and diversity might not be completely independent and might have reciprocal effects on each other (e.g. [38]). We sequentially removed non-significant pathways to reduce the AICc values and enhance the robustness of the models. Although the sample size in our study was at the lower boundary for SEM analysis, we consider it sufficient to provide a reliable potential relationship between the factors [39,40].

    To compare the direct (potentially non-trophic) and indirect effects (via herbivores) of the different components of tree diversity on the different components of spider diversity, absolute effect sizes were calculated based on standardized path coefficients connecting plant diversity components and herbivores with spider richness. Then, summing up all absolute effect sizes for each individual predictor, allows comparing the role the predictor has for direct and indirect paths in the model [9].

    3. Results

    In total, 6962 spiders were collected across the six sampling periods, including 332 morphotypes; 2104 COI sequences of spiders were grouped into 258, 259, 326, 258 and 325 MOTU by Mothur, BlastClust, ABGD, jMOTU and PTP, respectively. We chose the results of jMOTU for further analyses as it gave the most consistent results when compared with the other four methods (electronic supplementary material, table S3). The most species-rich families in the experimental sites were Araneidae (55), Salticidae (48), Theridiidae (40) and Thomisidae (33); the most abundant family was Araneidae (2719 individuals of 55 MOTU) (electronic supplementary material, table S4). One hundred and thirty-two barcodes of MOTU were identified to family, 93 to genus and 73 to species using SAP analysis (electronic supplementary material, figure S2). The sample coverage per plot come close to an asymptote, indicating that our sampling data have a high degree of completeness and are comparable across study plots for further analysis (electronic supplementary material, figure S3).

    The net biodiversity effect along the tree species richness gradient increased markedly for spider richness and slightly for spider abundance. The increasing biodiversity effects were mainly brought about by increasing complementarity effects, while the all selection effects were consistently negative (electronic supplementary material, figure S5).

    Spider abundance and species richness significantly increased with tree functional diversity (figure 1a, table 1). Moreover, spider abundance and richness were not related to other environmental covariates (elevation, slope, northness and eastness). The results for spider PD were similar due to the strong correlation with species richness (Pearson's r = 0.953, p < 0.001; figure 1b, table 1). Spider MNTD was significantly negatively related to tree functional diversity (figure 1c, table 1). The FRic of spiders was positively and significantly related to tree functional diversity (figure 1d, table 1). The MNTD, FDiv and FD (Rao's Q) were positively and significantly related to tree MNTD (figure 2, table 1). The correlations between tree MNTD and spider species richness, MNTD, FRic, FEve and FD (Rao's Q) were found to differ significantly between the two study sites (table 1). The results with tree species richness were consistent with the results for models using tree MPD (except for spider FD (Rao's Q); table 1, electronic supplementary material, table S6).

    Figure 1.

    Figure 1. Relationship between tree functional diversity and (a) abundance, and (b) PD, and (c) MNTD and (d) FRic of spiders. The significant relationship is shown by a regression line (with 95% confidence bands).

    Figure 2.

    Figure 2. Relationship between mean nearest taxon distance (MNTD) of tree and (a) MNTD, (b) functional divergence (FDiv) and (c) FD (Rao's Q) of spiders. The significant relationship is shown by regression line (with 95% confidence bands).

    Table 1. Summary results of linear models for abundance, species richness, phylogenetic diversity index (PD, MPD and MNTD) and functional diversity (FRic, FEve, FDiv and Rao's Q) of spider communities with tree species richness. Depicted are the standardized parameter estimates (with standard error, t and p values) of variables remaining in the minimal model. The results of the models after backward selection are shown, and predictors indicated by ‘—’ were not included in the final model.

    Est s.e. t p Est s.e. t p
    abundance species richness
    (intercept) 4.81 0.11 43.77 <0.001 3.522 0.070 50.24 <0.001
    tree FD (Rao's Q) 0.30 0.09 3.39 <0.01 0.167 0.057 2.92 <0.01
    tree MNTD −0.24 0.13 −1.91 0.06 −0.139 0.082 −1.70 0.096
    site B −0.31 0.17 −1.89 0.06 −0.163 0.105 −1.55 0.129
    tree MNTD: site B 0.32 0.17 1.94 0.06 0.231 0.106 2.18 <0.05
    AICc 102.25 54.60
    PD MPD
    (intercept) 1.95 0.06 30.50 <0.001 0.493 0.007 66.59 <0.001
    tree FD (Rao's Q) 0.17 0.05 3.20 <0.01
    tree MNTD −0.13 0.07 −1.79 0.08
    site B −0.15 0.10 −1.55 0.13 −0.024 0.012 −1.91 0.062
    tree MNTD: site B 0.21 0.10 2.14 < 0.05
    tree elevation −0.013 0.006 −2.08 <0.05
    AICc 44.98 −198.73
    MNTD FRic
    (intercept) 0.28 0.01 37.09 <0.001 −11.451 0.368 −31.15 <0.001
    tree FD (Rao's Q) −0.02 0.01 −2.77 <0.01 0.882 0.300 2.94 <0.01
    tree MNTD 0.01 0.01 2.11 <0.05 −0.752 0.430 −1.75 0.087
    site B 0.03 0.01 2.12 <0.05 −0.388 0.553 −0.70 0.486
    tree elevation 0.02 0.01 2.98 <0.01
    tree MNTD: site B 1.255 0.558 2.25 < 0.05
    AICc −193.66 230.22
    FEve FDiv
    (intercept) 0.65 0.01 68.52 <0.001 0.855 0.009 94.28 <0.001
    tree MNTD 0.02 0.01 1.83 0.07 0.015 0.007 2.11 <0.05
    tree VD 0.00 0.01 −0.30 0.77 −0.015 0.007 −2.19 <0.05
    tree B 0.00 0.02 −0.16 0.87 0.011 0.014 0.79 0.434
    tree MNTD: site B −0.02 0.01 −1.61 0.12
    tree VD: site B 0.08 0.02 3.54 <0.001
    tree FD (Rao's Q) −0.011 0.008 −1.42 0.163
    tree FD (Rao's Q): site B 0.026 0.016 1.65 0.105
    AICc −158.63 −163.98
    FD (Rao's Q)
    (intercept) 0.039 0.001 43.76 <0.001
    tree MNTD 0.002 0.001 2.07 <0.05
    AICc −378.63

    When replacing the estimated functional and phylogenetic diversity with their standardized effect size values in the linear models, we found that spider functional and phylogenetic diversity were still related to multiple diversity components of trees. Spider MPD.ses was significantly negatively related to tree FD (Rao's Q). FRic.ses and FDiv.ses of spiders were positively related to tree MNTD, and MNTD.ses and FDiv.ses of spiders were negatively related to tree vertical structural diversity. As expected from our null model approach and in contrast to PD, PD.ses was not correlated with any of the tree diversity indices, indicating that spider PD changed in a predictable way with increasing spider species richness.

    Path analyses revealed that the tree species richness had no direct effects on species richness, phylogenetic diversity (expect for MPD) or functional diversity of spiders, but they were strongly mediated by MPD, FD and VD of trees, herbivore abundance and richness, and spider abundance (figure 3, electronic supplementary material, figure S7). The effects of tree FD and VD on spider abundance were direct, while the effects of tree MPD were indirect (via herbivores; except for MPD and FD Rao's Q). Spider FDiv was mainly affected by tree VD, while spider FD (Rao's Q) was directly promoted by tree MPD (figure 3c,d). In most cases, spider abundance was strongly promoted by herbivore species richness, which was mediated by herbivore abundance. The effects of tree diversity on spiders mainly operated by affecting spider abundance, which in turn strongly promoted spider species richness, PD and FRic (figure 3a, electronic supplementary material, figure S7a, figure S7c), while decreasing spider MNTD and FEve (electronic supplementary material, figure S7b, figure S7d). The reciprocal effects play a key role between MPD, FDiv and FD (Rao's Q) of spider and spider abundance (figure 3b–d). Directional effects of spider abundance on spider diversity received stronger support (lower AICc values, except for spider MPD, FDiv and FD Rao's Q) than alternative models assuming reciprocal effects between the two variables (using a covariance term instead of a direct path; AICc values for directional effects versus covariance effects were 882.6397 versus 887.9045 for species richness; 892.8918 versus 907.6858 for PD; 971.2691 versus 977.2375 for MNTD; 955.3165 versus 965.4696 for FRic; 961.4451 versus 972.5263 for FEve). The results of path models were similar when herbivore species richness was replaced by herbivore PD (electronic supplementary material, figure S9; except for spider MNTD and FD Rao's Q), however, with increasing the relative low direct effect of tree FD on herbivore PD. In the models with herbivore MPD, functional diversity and phylogenetic diversity of trees still showed both direct (potentially non-trophic) and indirect (via herbivores) connections with spider abundance in all cases (electronic supplementary material, figure S10), which in turn regulated spider diversity (except for spider FDiv and FD Rao's Q; see electronic supplementary material, figure S10a–f). Moreover, in none of the models with herbivore MPD (electronic supplementary material, figure S10) and in contrast to those with herbivore richness (figure 3, electronic supplementary material, figure S7) and herbivore PD (electronic supplementary material, figure S8), the herbivore diversity metric was not directly affected by herbivore abundance. When herbivore species richness was replaced with herbivore MNTD (electronic supplementary material, figure S10), abundance and diversity of spider were mainly affected by herbivore abundance directly, rather than via herbivore MNTD.

    Figure 3.

    Figure 3. Path model of effects of tree species richness, tree phylogenetic diversity, tree vertical structural diversity, tree functional diversity, herbivore abundance and herbivore diversity on abundance, (a) species richness (χ2 = 17.428, d.f. = 16, p = 0.358), (b) MPD (χ2 = 16.943, d.f. = 14, p = 0.500), (c) FDiv (χ2 = 20.219, d.f. = 16, p = 0.211), (d) FD Rao's Q (χ2 = 14.333, d.f. = 15, p = 0.699) of spiders. Blue arrows (with standardized path coefficients) indicate significantly positive effects, red arrows show significantly negative effects and grey dashed arrows indicate a non-significant effect or covariance. Arrow width was scaled by the standardized path coefficients.

    Moreover, the results of summed absolute effect sizes for the predictors indicate the functional diversity had more significant direct effects (potentially non-trophic) on most diversity components of spiders (except for spider MPD, FDiv and FD Rao's Q) than the other tree diversity indices. Nevertheless, the effects of tree richness and phylogenetic diversity on nearly all diversity components were indirect (via herbivore; except for spider MPD, FDiv and FD Rao's Q). In addition, the indirect absolute effect size of tree species richness via herbivore was stronger than the direct effect (figure 4).

    Figure 4.

    Figure 4. Direct and indirect of plant diversity on predator diversity. The summed effects of tree species richness (tree SR), phylogenetic diversity (tree MPD), vertical structural diversity (tree VD) and functional diversity (tree FD) on spider species richness, phylogenetic diversity (PD, MPD and MNTD) and functional diversity (FRic, FEve, FDiv and FD Rao's Q), obtained from the path models in figure 3 and electronic supplementary material, figure S7, are shown in bars. Effect sizes are calculated as absolute values. Effects are either direct or indirect (darker hues, left bar of each diversity component, effect of plant diversity on spider abundance and diversity through direct paths) via herbivore abundance and diversity (lighter hues, right bar of each diversity component, effect of plant diversity on herbivore abundance and diversity, which in turn affected spider abundance and diversity). Absolute effect size was calculated follow the method in Schuldt et al. [9].

    4. Discussion

    Our study clarifies linkages between the diversity of tree communities and arthropod predators. It shows the relative importance of direct (probably largely non-trophic from plants to predators) and indirect effects (trophic relationships via herbivores) from plants to predators. It suggested that these effects on spiders are primarily driven by herbivore abundance and diversity, while tree diversity components (phylogenetic diversity, functional diversity and vertical structural diversity) had weak effects on predator diversity. These findings provide extensive insights into the potential mechanisms structuring the relationships between tree diversity and predators in forest ecosystems.

    (a) Direct and indirect structuring of predator communities by tree diversity components

    Our results indicate that the relationships between tree species richness and spiders were to a larger part indirect and mediated by herbivore abundance and richness. More tree species in a forest plot might provide more prey for predators via the accommodation of more diverse food resources to primary consumers [41]. Effects of tree species richness occurring via herbivores accounted for a large part of the indirect effects of tree species richness on spiders. However, when broadening the scope by including tree functional, phylogenetic and structural diversity, our results showed additional effects of these diversity components on spider abundance, species richness, phylogenetic and functional diversity.

    The different effect directions of functional, phylogenetic and structural diversity have important ecological implications in that plant phylogenetic diversity, functional and structural diversity emerge as key predictors of the diversity components at the higher trophic levels, although the underlying mechanisms may differ to some extent. Direct effects of both functional diversity and structural diversity might indicate an important role of how tree diversity modifies the physical structure (which might influence spiders via e.g. availability of shelter, possibilities for web attachment, modification of microclimate [42]). Moreover, functional diversity also potentially influences the volatile chemical landscape of plant communities. Tree phylogenetic diversity might be linked more strongly to trophic relationships via its structuring role of herbivore communities [11,43] as important prey organisms, although some of these effects are certainly also reflected in the indirect effects of tree functional diversity by filtering herbivores and other potential prey for spiders via their dependence on leaf traits such as leaf palatability or nutrients [11,44,45].

    Moreover, a higher functional diversity of plant communities not only promoted the coexistence of more spider species, but may also indicate a potentially stronger prey control signal caused by broader resource use within functionally more diverse spider assemblages (as higher FRic reflects higher volume of trait space occupied by an assemblage; [46]). It should be noted that tree functional diversity had more consistent effects than tree phylogenetic diversity on spider communities, as most of the significant effects of phylogenetic diversity were represented in only one of the study sites. These results might at least partly be caused by the non-overlapping tree species pool between the two sites, which have different tree species identities that are not directly captured by phylogenetic diversity and point to the importance of differentiating among more distinct components of biodiversity. The absolute values of indirect effect size of tree MPD are similar to tree FD (direct effect) for spiders (except for spider MPD, FDiv and FD), suggesting that tree phylogenetic diversity can also play key roles in spider community assembly when more trophic levels were considered (e.g. herbivore in our case). Moreover, this pattern cannot be predicted only by simple regression modelling (as shown by the non-significant effects of tree MPD on spiders in our results of linear models; table 1).

    The direct effects of tree structural diversity are further indications of non-trophic tree diversity effects on spiders that might act either directly via vegetation structure or via microclimatic conditions that are altered by tree structural diversity [47,48]. The negative relationships between tree vertical structural diversity and spider abundance and functional diversity indicate that environments with highly heterogeneous conditions can hamper locating and capturing prey [49] and may promote specific functional types of predators adapted to these conditions. Our findings corroborate the results from a previous study on predators in our experiment at an earlier stage [9], where structural diversity effects were even stronger (which could be due to such environmental effects playing a particularly important role during earlier developmental forest stages [50], or to a different set of predatory taxa considered).

    The additional path models involving the three herbivore phylogenetic indices (PD, MPD and MNTD) helped us to further tease apart the direct and indirect pathways among trees, herbivores and spiders. The results indicated that the effects of trees on spiders via herbivores are general effects of tree diversity that can be observed irrespective of whether modifying effects of herbivores are quantified via herbivore species richness or phylogenetic diversity. The effects of herbivore PD were similar to herbivore species richness effects on spiders in the pathway models, indicating that herbivore species richness captured a large part of the effect of PD because of higher correlation between herbivore PD and species richness (Pearson's r = 0.94, p < 0.001). The relative lower direct effects of tree FD on herbivore PD in the path models highlights the importance of the phylogenetic associations between herbivores and their hosts, indicating phylogenetic conserved adaptations to nutritive and defensive characteristics [11]. Tree phylogenetic MPD had consistent effects on spiders in pathway models, irrespective of which herbivore diversity components were considered. Furthermore, there were no additional effects of herbivore phylogenetic diversity on spiders in the SEM models when herbivore species richness was replaced by herbivore phylogenetic diversity, suggesting that spider diversity cannot be simply explained by herbivore phylogenetic diversity. As spiders are generalist predators, their diet breadth is weakly affected by the phylogenetic dependencies of prey, which therefore should have little effects on spider diversity [51]. Because of a current lack of functional trait records for herbivores, we were not able to study the potential effects of herbivore functional diversity on spider diversity. However, some key traits of herbivores (such as resistance traits) have shown significant phylogenetic signal in previous studies [11,52], indicating that phylogenetic diversity of herbivores may be useful in predicting the functional trait space of herbivore assemblages. Nevertheless, the additional potential effects of herbivore functional diversity may not be fully captured by phylogenetic diversity. This remains an avenue for further research, in addressing how such relationships are affected by functional diversity of herbivore communities and relevant key functional traits. Moreover, it will also allow us to more directly understand the effects of trees on spiders via linking the associations between plants and herbivores in functional dependencies.

    (b) The role of tree functional and phylogenetic diversity components in driving predator community structure

    The results of net biodiversity effects, and its two components, i.e. additive complementarity and selection effects, allowed us to clarify the relationships between biodiversity and ecosystem functions. Negative selection effects in the context of spider richness mean that some tree species with relatively high monoculture spider richness had a lower spider richness in mixtures, or vice versa, some tree species with relatively low monoculture spider richness had higher richness in mixtures. Overall, this means that more diverse tree mixtures dampened the effect of particular tree species on spider richness. However, differences in tree compositions led to strong and positive complementarity effects, outweighing the negative selection effects. In other words, the effects caused by the combination of different tree species were more important than that of the particular tree species.

    The different components of tree diversity had distinct effects on spider diversity components, which warrants a closer inspection of these relationships in order to better understand driving forces of tree diversity effects. By contrast to the generally positive effect of tree functional diversity on spiders, spider MNTD was negatively related to tree functional diversity, indicating that more closely related spider assemblages increased with tree functional diversity. A potential explanation could be increasing competition among closely related spiders due to limited prey resources caused by lower plant functional diversity, which would promote separate niche utilization and in turn accelerate the functional differentiation of spider species [20]. By contrast, plots with high functional heterogeneity (high Rao's Q) of plants might provide more heterogeneous physical structures, reducing cannibalism and allowing even closely related (low MNTD) spiders to coexist [53].

    Tree phylogenetic MPD showed no direct impacts on spider abundance or any component of spider diversity (except for spider MPD). Nevertheless, tree phylogenetic MPD showed a relatively important role in the SEM models via pathways including herbivores. As already mentioned above, this pattern might be driven by the fact that biotic niches and therefore trophic relationships are determined by plant phylogenetic diversity and can shape species composition of organisms (see also [25]).

    Phylogenetic diversity effects might also be in part explained by functional traits and diversity. In our study, the positive effects of tree phylogenetic MNTD on spider MNTD, FDiv and FD (Rao's Q) indicates that spider species were phylogenetically and functionally similar and characterized by trait values closer to the centre of the overall functional trait space in communities with more closely related tree species [54]. Although spider communities are often thought to show weak associations with specific tree species, some studies have indicated that even such generalist predators might be structured by tree species identity, possibly by effects of structural traits or prey spectrum (e.g. [55]). This pattern was independent of changes in spider species richness, as we found that the results were consistent when the observed FDiv was replaced by FDiv.ses of spiders. Moreover, the negative effects of tree vertical stratification on MNTD.ses could be explained by the heterogeneity of vertical stratification providing a finer niche occupation that would allow more closely related spiders to co-occur (low MNTD.ses). In turn, the co-occurrence of more closely related (and therefore presumably functionally more similar) spiders might reduce the overall functional difference among dominant spiders (low FDiv).

    By contrast to observed PD, there were no significant relationships between standardized PD of spiders and any of the tree diversity indices, indicating that spider PD changed in a predictable way with increasing spider species richness. This points to the fact that the different diversity components we considered were not completely independent from each other, and especially phylogenetic diversity was correlated with species richness. Our results therefore also illustrate how species richness relationships can be explained by variability in functional and/or phylogenetic community characteristics, with the differences in their effect sizes being indicative of the specific mechanisms that potentially underlie these relationships.

    Interestingly, the MPD and FEve of spiders were not related to tree diversity, irrespective of taxonomic, phylogenetic and functional diversity of the tree communities. Lack of effect for MPD means that deeper in the phylogeny, spider composition is unaffected by tree diversity and the potential competition effects suggested for MNTD. This is in line with the expectation that competition is most important for terminal branches, between very closely related congeners, whereas resources seem to be available and competition is less important for different spider families [56]. The same seems to be the case for FEve, possibly because spider composition is driven more strongly by family and genera composition rather than composition within genera. However, our results differ from a previous study in older stands of natural forests of the study region, where epigeic spider FEve decreased with increasing MPD of the woody plant communities [57]. The different patterns might indicate that there are different coexistence mechanisms between epigeic and arboreal spiders, or that patterns might change over the course of forest succession.

    The effects of plant diversity on the diversity (species richness, PD and FRic) of spiders were indirectly mediated by spider abundances. This indicates that mechanisms related to the more-individuals hypothesis (e.g. [9,10,38]) play an important role in structuring plant–consumer relationships at our study site (see also [9,10]). Our results for species richness represent a biologically meaningful pattern among trees, herbivores and spiders because we used a standardized sampling strategy. We tried to standardize the beaten branch volume per tree when sampling in the field, so that species richness and abundance represent habitat volume-specific diversity, rather than being directly biased methodologically by abundance. Thus, our results support a biological pattern that tree diversity or herbivores promote spider abundance per tree (volume), which in turn affects spider richness. This result is in line with recent studies at our experimental sites [9,10], highlighting the importance of a more comprehensive assessment of the interactions between biodiversity change and abundance declines across trophic levels in times of declines in consumer abundances and biomass caused by human-induced environmental change [58]. However, it is notable that spider MPD, FDiv and FD (Rao's Q) were not directly linked to spider abundance in the path models. The latter is likely due to an increase in the number of common spider species in species-rich families decreasing the mean dissimilarity to the nearest relative, which was also suggested by a previous study on herbivores at our study sites [10]. The (indirect) covariation between the three metrics (MPD, FDiv and FD Rao's Q) and spider abundances might indicate more complex, reciprocal interactions between the two, rather than a unidirectional effect of abundance [54].

    It was unexpected that spider MPD, FDiv and FD (Rao's Q) were less affected by herbivores, meaning that overall phylogenetic splits and deviance of functional dissimilarity of spider communities were not significantly determined by herbivore richness. This might be because for generalized predators phylogenetic structure is usually not correlated with the number of species of prey [59].

    We are aware that the collection of spiders from three seasons might cause potential variation of the results because of the seasonal effects, but we were unable to analyse such seasonal effects due to the limited sample size in the individual seasons. Hence, our results represent the general pattern of the associations for the entire year among the trees, herbivores, and spiders [11]. There is no doubt that seasonal variation is a further perspective in exploring the pattern and mechanism on trophic cascade of multi-trophic levels [60]. Admittedly there is a potential limitation in our study, as only lepidopteran larvae were considered as prey in our analysis. However, Lepidoptera are one of the most diverse assemblages in these subtropical forests [61] providing very abundant food resources for predators. Hence, our study provides key evidence on the potential effects of abundant prey as well as important herbivores on spiders. Nevertheless, it will be an interesting option for future studies to expand the scope and include additional potential prey groups, as most spiders are generalist predators preying on a wider range of arthropods (e.g. Hemiptera, Diptera). Moreover, it would be valuable to take tree mortality into account in future studies, as tree survival is dynamic under natural conditions.

    5. Conclusion

    Overall, our study provides new insight into the relationships among multiple components of both predator diversity and tree diversity, far beyond the range of components considered in previous studies in forest ecosystems. Our findings highlight that incorporating a broader range of diversity indicators (from taxonomic to functional and phylogenetic) at the level of both the plant community and predator community can help to develop a more comprehensive understanding of how changes in biodiversity at different trophic levels will affect ecosystem structure and functions. Moreover, our results provide further evidence for the importance of monitoring consumers abundance and biomass (as their species richness can be mainly mediated by abundance). Importantly, while cascading tree diversity effects via prey/feeding relationships can play an important role in driving predator community diversity, direct effects especially of tree functional and structural diversity on predators underline that at same time mechanisms underlying biodiversity effects extend well beyond trophic relationships and require further study in future biodiversity–ecosystem functioning research. Altogether, these findings call for management that sustains natural enemy populations in forest ecosystems by increasing plant functional and phylogenetic diversity to sustain provisioning of key ecosystem services such as pest control.

    Data accessibility

    Data are available on the BEF-China project database at https://data.botanik.uni-halle.de/bef-china/datasets/656. Sequence data can be accessed on Genbank https://www.ncbi.nlm.nih.gov/genbank/ (accession numbers: OP816739-OP816940; OP816503-OP816512). Statistical and sequence data can be accessed on Science Data Bank (https://www.scidb.cn/s/QjUZf2; https://www.scidb.cn/s/iayiMj).

    The data are provided in the electronic supplementary material [62].

    Authors' contributions

    J.-T.C.: conceptualization, data curation, formal analysis, investigation, methodology, project administration, writing—original draft, writing—review and editing; M.-Q.W.: conceptualization, data curation, formal analysis, methodology, supervision, writing—original draft; Y.L.: methodology; D.C.: methodology; A.L.: methodology; W.Z.: data curation, methodology; P.-F.G.: methodology; S.-K.G.: methodology; Q.-S.Z.: methodology; K.-P.M.: methodology; G.v.O.: data curation; M.K.: data curation; N.-L.Z.: methodology; X.-J.L.: methodology; H.B.: methodology; A.S.: conceptualization, methodology, supervision, writing---review and editing; C.-D.Z.: conceptualization, funding acquisition, methodology, project administration, supervision, writing—review and editing.

    All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

    Conflict of interest declaration

    None of the authors has a conflict of interest.

    Funding

    This project was supported mainly by the Strategic Priority Research Program of the Chinese Academy of Science (grant no. XDB310304), partly by the National Science Fund for Distinguished Young Scholars (grant no. 31625024), the National Science & Technology Fundamental Resources Investigation Program of China (grant no. 2018FY100401) and a grant (2008DP173354) from the Key Laboratory of the Zoological Systematics and Evolution of the Chinese Academy of Sciences. M.-Q.W. is also funded by the National Natural Science Foundation, China (grant no. 32100343). J.-T.C. was supported by the International Research Training Group TreeDì jointly funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – 319936945/GRK2324 and the University of Chinese Academy of Sciences (UCAS).

    Acknowledgements

    We are grateful for support from the BEF-China consortium (especially, those from Bo Yang and Shan Li). We thank all local workers/assistants, who participated in our field work. We thank Zhongwei Deng from Hubei University for help in taxonomic identification of spiders. We are also indebted to Robert Kallal for providing the backbone phylogenetic tree of spiders. We thank Wenzel Kröber for the assessment of tree traits. We thank the support of Zhejiang Qianjiangyuan Forest Biodiversity National Observation and Research Station.

    Footnotes

    Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.6350555.