Plant nutritional and structural diversity shape multitrophic arthropod communities and grassland productivity
Data files
Apr 11, 2026 version files 83.21 KB
-
JOE_Arthropod_richess_and_abundance.xlsx
16.12 KB
-
JOE_Liner_and_quadratic_model_comparison_new.xlsx
20.86 KB
-
JOE_Plant_traits_composition_and_ANPP.xlsx
31.03 KB
-
README.md
15.20 KB
Abstract
Arthropod communities, comprising diverse trophic groups such as herbivores, predators, and parasitoids, are intricately linked to plant traits that provide food and habitat. While it is well-established that changes in plant functional diversity (e.g., trait identity and diversity) can significantly alter arthropod diversity across trophic levels, the cascading effects on ecosystem functions remain less understood. Particularly, the role of multitrophic arthropod diversity in mediating the relationship between plant functional diversity and grassland productivity presents a critical knowledge gap in ecosystem ecology.
We employed a long-term plant removal experiment in the Inner Mongolian grassland to systematically investigate how variations in the community-weight mean and diversity of multiple plant traits influence the diversity (measured by taxon richness and abundance) of herbivores and their natural enemies. Furthermore, we explored how trophic interactions between herbivores and their natural enemies influence plant community productivity.
Our findings indicate that high diversity in plant nutritional traits (e.g., nitrogen, phosphorus, and sodium contents) negatively impacts plant productivity through both direct and indirect pathways. The adverse effect was mediated by an increase in the richness of sucking and chewing herbivores, which exploited high resource complementarity yet collectively suppressed plant productivity. In contrast, higher community-weighted means of plant structural traits (e.g., vegetative height and leaf lateral spread) were associated with greater plant productivity. This positive effect appears to arise from enhanced top-down control, whereby predators—particularly spiders—reduced both the richness and abundance of herbivores.
Synthesis. Our study reveals that herbivores and their natural enemies respond distinctly to the variation in the composition and diversity of plant nutritional and structural traits. We show that cross-trophic interactions—specifically, diversity within herbivore and predator guilds—constitute a primary pathway through which plant functional diversity influences grassland productivity. By disentangling the links between plant trait spectra, arthropod community structure, and ecosystem functioning, our findings provide key insights for biodiversity conservation and the design of ecosystem management strategies in grasslands.
Dataset DOI: 10.5061/dryad.dbrv15fgq
Description of the data and file structure
To explore how the functional diversity of plant nutritional and structural traits influences arthropod communities, and, subsequently, grassland ecosystem productivity, we conducted a comprehensive study using a long-term plant removal experiment in the Inner Mongolia grassland (Sasaki et al., 2019, Wu et al., 2015). Our unique experimental design allowed us to explore a broad range of variations in both the functional identity and diversity of plant nutritional and structural traits. Our study aimed to address two fundamental questions: First, how do variations in the community-weighted mean and diversity of plant nutritional and structural traits shape the diversity of herbivores and their natural enemies as measured by taxon richness and abundance? Second, to what extent do herbivore and natural enemy communities mediate the effects of plant functional traits on overall plant productivity? To address these questions, we formulated four hypotheses: (1) Herbivore diversity is primarily driven by both the community-weighted mean and diversity of plant nutritional traits (Fig. 1B) due to the strong bottom-up effects of increased plant nutrient availability and diversity on herbivore communities (Awmack and Leather, 2002, Behmer, 2009). (2) In contrast to herbivores, we hypothesize that the diversity of natural enemies is predominantly influenced by the community-weighted mean and diversity of plant structural traits (Fig. 1B). Previous research suggests that greater habitat space and habitat heterogeneity, resulting from diverse plant structures, can support more abundant and diverse predator populations, leading to stronger top-down control of herbivore communities (Letourneau et al., 2009, Prather et al., 2021, Langellotto and Denno, 2004). (3) The community-weight mean and diversity of plant nutritional traits affect plant productivity through both direct and indirect pathways (Fig. 1C; pathways a, b, c). The indirect pathway operates by modulating herbivore diversity, which consequently influencing plant consumption (Deraison et al., 2015, Borer et al., 2014, Zhang et al., 2019, Milcu et al., 2014). (4) The community mean and diversity of plant structural traits influence plant productivity through a cascade of direct and indirect effects (Fig. 1C; pathways d, e, f, g, c). The cascade operates by altering natural enemy diversity, which in turn affects herbivore diversity (Letourneau et al., 2009, Li et al., 2023, Prather et al., 2021, Schuldt et al., 2019, Valencia et al., 2015).
Files and variables
File: JOE_Arthropod_richess_and_abundance.xlsx: Richness and abundance of different arthropod guilds
Variables
Block: Description: The experimental block code in the Inner Mongolia Grassland Removal Experiment (IMGRE). The IMGRE consists of eight blocks (55 m × 85 m each), separated by 2 m walkways. For this study, we selected four out of the eight blocks (i.e., blocks 2, 4, 6, and 8) for our sampling efforts.
Treatment: Description: The abbreviation code for the experimental treatment in the Inner Mongolia Grassland Removal Experiment (IMGRE). PB, PR, PF and AB represent perennial bunchgrasses, perennial rhizomatous grasses, perennial forbs and annuals/biennials, respectively. “+” indicates the removal of the corresponding plant functional group and their combinations.
Plots: Description: The experimental plot code in each block of the Inner Mongolia Grassland Removal Experiment (IMGRE). Plots (6 m × 6 m each) within the blocks were separated by 1 m walkways.
Sucking/piercing herbivore richness (number of species / plot): Description: Number of sucking/piercing herbivore species sampled in an experimental plot.
Chewing herbivore richness (number of species / plot): Description: Number of chewing herbivore species sampled in an experimental plot.
Boring/mining herbivores richness (number of species / plot): Description: Number of boring/mining herbivores species sampled in an experimental plot.
Total herbivore richness (number of species / plot): Description: Number of total herbivore species sampled in an experimental plot.
Predator richness (number of species / plot): Description: Number of predator species sampled in an experimental plot.
Parasitoid richness (number of species / plot): Description: Number of parasitoid species sampled in an experimental plot.
Spider richness (number of species / plot): Description: Number of spider species sampled in an experimental plot.
Total enemy richness (number of species / plot): Description: Number of total enemy species sampled in an experimental plot.
Sucking/piercing herbivore abundance (number of individuals / plot): Description: Number of sucking/piercing herbivore individuals sampled in an experimental plot.
Chewing herbivore abundance (number of individuals / plot): Description: Number of chewing herbivore individuals sampled in an experimental plot.
Boring/mining herbivores abundance (number of individuals / plot): Description: Number of boring/mining herbivores individuals sampled in an experimental plot.
Total herbivore abundance (number of individuals / plot): Description: Number of total herbivore individuals sampled in an experimental plot.
Predator abundance (number of individuals / plot): Description: Number of predator individuals sampled in an experimental plot.
Parasitoid abundance (number of individuals / plot): Description: Number of parasitoid individuals sampled in an experimental plot.
Spider abundance (number of individuals / plot): Description: Number of spider individuals sampled in an experimental plot.
Spider abundance (number of individuals /plot): Description: Number of total enemy individuals sampled in an experimental plot.
File: JOE_Liner_and_quadratic_model_comparison_new.xlsx
Table S1 Model selection for the effects of CWM and Rao of plant nutrients traits, CWM and Rao of plant structural trait on plant productivity. In the liner mixed models (LMMs), each of plant variables were treated as fixed factors and block was treated as a random factor. Models with the first and second order polynomial were fitted. We compared the fit of the two models by using small sample size corrected Akaike information criterion (AICc) (Burnham and Anderson 2002). When the AICc difference (ΔAICc) was <2 between two significant functions, the simpler function was chosen, where ‘simplicity’ increased from linear to quadratic. Best fitted curves are in pink. AICc is Akaike’s information criterion; Marginal R2 (Rm2) represents model variation explained by fixed effects; Conditional R2 (Rc2) represents model variation explained by both fixed and random effects.
Table S2 Model selection for the effects of CWM and Rao of plant nutrients traits, CWM and Rao of plant structural trait on abundance and richness of herbivore and their natural enemies. In the liner mixed models (LMMs), each of plant variables were treated as fixed factors and block was treated as a random factor. Models with the first and second order polynomial were fitted. We compared the fit of the two models by using small sample size corrected Akaike information criterion (AICc) (Burnham and Anderson 2002). When the AICc difference (ΔAICc) was <2 between two significant functions, the simpler function was chosen, where ‘simplicity’ increased from linear to quadratic. Best-fitted curves are in pink. AICc is Akaike’s information criterion; Marginal R2 (Rm2) represents model variation explained by fixed effects; Conditional R2 (Rc2) represents model variation explained by both fixed and random effects.
Table S3 Model selection for the effects of the richness and abundance of sucking herbivore, chewing herbivore, and endophyte on plant productivity. In the liner mixed models (LMMs), each of arthropod variables were treated as fixed factors and block was treated as a random factor. Models with the first and second order polynomial were fitted. We compared the fit of the two models by using small sample size corrected Akaike information criterion (AICc) (Burnham and Anderson 2002). When the AICc difference (ΔAICc) was <2 between two significant functions, the simpler function was chosen, where ‘simplicity’ increased from linear to quadratic. Best fitted curves are in pink. AICc is Akaike’s information criterion; Marginal R2 (Rm2) represents model variation explained by fixed effects; Conditional R2 (Rc2) represents model variation explained by both fixed and random effects.
Variable (units)
Plant productivity (g/m2): Aboveground net primary production.
Herbivore abundance (number of individuals / plot): Number of total herbivore individuals in experimental plot.
Herbivore richness (number of species / plot): Number of total herbivore species in an experimental plot.
Enemy richness (number of species / plot): Number of total enemy species in an experimental plot.
Enemy abundance (number of individuals / plot): Number of total enemy individuals in an experimental plot.
CWM_leaf structure (biodiversity index/unitless): The first component of all the CWM of leaf traits, which explained 46.53% of the variance, was associated with plant structure traits.
CWM_leaf nutrient (biodiversity index/unitless): The second component of all the CWM of leaf traits, accounting for 32.31% of the variance, was linked to plant nutrient traits.
Rao_leaf structure (biodiversity index/unitless): The first component all the Rao of leaf traits, explaining 32.61% of the variance, was related to the diversity of plant structure traits.
Rao_leaf nutrient (biodiversity index/unitless): The second component all the Rao of leaf traits, explaining 30.05% of the variance, was related to the diversity of plant nutrient traits.
Sucking/piercing herbivore richness (number of species /m2): Number of sucking/piercing herbivore species sampled in an experimental plot.
Chewing herbivore richness (number of species / plot): Number of chewing herbivore species sampled in an experimental plot.
Boring/mining herbivores richness (number of species / plot): Number of boring/mining herbivores species sampled in an experimental plot.
Sucking/piercing herbivore abundance (number of individuals / plot): Number of sucking/piercing herbivore individuals sampled in an experimental plot.
Chewing herbivore abundance (number of individuals / plot): Number of chewing herbivore individuals sampled in an experimental plot.
Boring/mining herbivores abundance (number of individuals / plot): Number of boring/mining herbivores individuals sampled in an experimental plot.
File: JOE_Plant_traits_composition_and_ANPP.xlsx: Plant functional trait composition across different plant functional group removal treatments.
Block : The experimental block code in the Inner Mongolia Grassland Removal Experiment (IMGRE). The IMGRE consists of eight blocks (55 m × 85 m each), separated by 2 m walkways. For this study, we selected four out of the eight blocks (i.e., blocks 2, 4, 6, and 8) for our sampling efforts.
Treatment: The abbreviation code for the experimental treatment in the Inner Mongolia Grassland Removal Experiment (IMGRE). PB, PR, PF and AB represent perennial bunchgrasses, perennial rhizomatous grasses, perennial forbs and annuals/biennials, respectively. “+” indicates the removal of the corresponding plant functional group and their combinations.
Plots: The experimental plot code in each block of the Inner Mongolia Grassland Removal Experiment (IMGRE). Plots (6 m × 6 m each) within the blocks were separated by 1 m walkways.
Variables (units)
Plant species richness (number of plant species /m2): Number of plant species in plant communities.
ANPP(g/m2): Aboveground net primary production.
PR biomass(g/m2): Perennial rhizome grasses biomass.
PB biomass (g/m2): Perennial bunch grasses biomass.
PF biomass (g/m2): Perennial forbs biomass.
AB biomass (g/m2): Annuals and biennials biomass.
CWM_nitrogen (%): Community-weighted mean (CWM) of leaf nitrogen content.
CWM_lateral spread (cm): Community-weighted mean (CWM) of leaf lateral spread.
CWM_vegetative height (cm): Community-weighted mean (CWM) of vegetative height.
CWM_steam and leaf ratio (%): Community-weighted mean (CWM) of steam and leaf ratio.
CWM_phosphorus (%): Community-weighted mean (CWM) of leaf phosphorus content.
CWM_sodium (%): Community-weighted mean (CWM) of leaf sodium content.
CWM_water (%): Community-weighted mean (CWM) of leaf water content.
Rao_nitrogen (biodiversity index/unitless): Rao’s quadratic entropy (Rao’s Q) of leaf nitrogen content.
Rao_lateral spread (biodiversity index/unitless): Rao’s quadratic entropy (Rao’s Q) of leaf lateral spread.
Rao_vegetative height (biodiversity index/unitless): Rao’s quadratic entropy (Rao’s Q) of vegetative height.
Rao_steam and leaf ratio (biodiversity index/unitless): Rao’s quadratic entropy (Rao’s Q) of steam and leaf ratio.
Rao_phosphorus (biodiversity index/unitless): Rao’s quadratic entropy (Rao’s Q) of leaf phosphorus content.
Rao_sodium (biodiversity index/unitless): Rao’s quadratic entropy (Rao’s Q) of leaf sodium content.
Rao_water (biodiversity index/unitless): Rao’s quadratic entropy (Rao’s Q) of leaf water content.
CWM_leaf structure (biodiversity index/unitless): The first component of all the CWM of all plant leaf traits, which explained 46.53% of the variance, was associated with plant structure traits.
CWM_leaf nutrient (biodiversity index/unitless): The second component of all the CWM of all plant leaf traits, accounting for 32.31% of the variance, was linked to plant nutrient traits.
Rao_leaf structure (biodiversity index/unitless): The first component all the Rao of all plant leaf traits, explaining 32.61% of the variance, was related to the diversity of plant structure traits.
Rao_leaf nutrient (biodiversity index/unitless): The second component all the Rao of all plant leaf traits, explaining 30.05% of the variance, was related to the diversity of plant nutrient traits.
1. Study site
This study was conducted at the Inner Mongolia Grassland Ecosystem Research Station (IMGERS, 43°38’N, 116°42’El) of the Chinese Academy of Sciences, located in the Xilin River Basin of Inner Mongolia, China. The study site is characterized by a semiarid climate, with a mean annual precipitation of 346.1 mm and a mean annual temperature of 0.3°C. The soil at this location is classified as dark chestnut soil with relatively homogeneous physionutrient properties (Calcic Chernozems according to the IUSS Working Group WRB 2006). The plant community at our study site comprises approximately 20 species, with Stipa grandis and Leymus chinensis identified as the dominant species (Bai et al., 2004).
All necessary permits were obtained from IMGERS prior to carrying out the field work. To investigate the effects of plant functional diversity on ecosystem processes, we utilized the Inner Mongolia Grassland Removal Experiment (IMGRE), established in 2005 (Wu et al., 2015). The IMGRE consists of eight blocks (55 m × 85 m each), separated by 2 m walkways. For this study, we selected four out of the eight blocks (i.e., blocks 2, 4, 6, and 8) for our sampling efforts (Sasaki et al., 2019). The plots within the blocks were separated by 1 m walkways. The experimental design categorized all plant species into four plant functional groups (PFGs): perennial rhizome grasses (PR), perennial bunch grasses (PB), perennial forbs (PF), annuals and biennials (AB). The IMGRE employed a full combinatorial design, including the removal of all possible combinations of these four PFGs (Table S1). This resulted in 16 distinct PFG removal treatments, which were assigned to plots according to a randomized block design scheme, yielding a total of 64 experimental plots.
To minimize soil disturbance, individual target plants were removed by clipping aboveground parts and tillering at a soil depth of 0-3 cm. The removal treatments were applied annually in June from 2006 to 2009. Subsequently, PFG removals were halted from 2010 to 2018 to allow natural reassembly processes to reconstruct the plant communities. The original PFG removal treatments did not significantly alter the total plant biomass within the communities; however, they did have a substantial impact on the biomass of PR and PB across the various removal treatments (Fig. S2). Moreover, these removal treatments significantly influenced both the community-weighted mean and the diversity of plant nutritional and structural traits among the sixteen removal treatments (Fig. S3).
2. Plant productivity and plant trait measurements
To quantify ecosystem productivity, we measured the aboveground net primary production (ANPP) in August 2018. This timing coincides with the annual peak of aboveground biomass in these steppe communities, providing an accurate estimate of ANPP (Bai et al., 2004). ANPP was measured by clipping all the plants in one 1 m × 1 m quadrat within each plot at the soil surface. The green portions were separated from standing litter and sorted by species. All samples were oven-dried at 65°C for 48 h and then weighed.
To comprehensively assess plant functional traits, we focused on both nutritional and structural characteristics. Leaf nitrogen concentration (N, %), leaf phosphorus concentration (P, %), leaf sodium concentration (Na, %) and leaf water content (LWC, %), which are crucial determinants of plant quality for arthropods (Awmack and Leather, 2002). Vegetative height (cm), leaf lateral spread (cm), and stem and leaf ratio (%) were measured simultaneously. Such plant structural traits are key factors in determining habitat volume and heterogeneity for arthropods (Lawton, 1983, Biswas et al., 2016). Given the limited area of our experimental plots, we employed a non-destructive sampling method for plant traits (Sasaki et al., 2019). Specifically, we measured the nutritional and structural traits of 15 species at five locations, each at least 50 meters apart, in a nearby ungrazed area (Cornelissen et al., 2003). It is important to note that we did not assess traits for extremely rare species. While this approach may result in the omission of trait values from rare species, the traits of the 15 common species are representative for each plot when calculating functional diversity indices, as these species collectively account for over 95% of the total biomass within the communities (Sasaki et al., 2019). For each location at least nine fully expanded leaves from a minimum of three individuals per species were collected. Nutritional and structural traits were measured in August 2013 following standard protocols (Cornelissen et al., 2003).
3. Functional identity and diversity of plant traits
To assess the functional identity and diversity of plant traits, we calculated the community-weighted mean (CWM) trait values for each measured trait using the formula from Garnier et al. (2004).
Many studies have integrated multiple plant traits to calculate functional diversity indices at the community level (Garnier et al., 2004, Mokany et al., 2008, Xu et al., 2018, Zhang et al., 2019). However, this approach can obscure the distinction between plant nutritional and structural traits in the current study. Consequently, isolating the independent effects of these two trait categories on plant productivity and arthropod diversity becomes challenging. To address this issue, we quantified the functional diversity of plant communities for each trait using Rao’s quadratic entropy (Rao’s Q) as described by Botta-Dukát (2005).
Subsequently, we conducted principal component analyses (PCA) with Varimax rotation separately for community-weighted mean (CWM) and Rao’s quadratic entropy (Rao) of all measured plant traits. This approach ensures that the first and second PCA axes capture the maximum variation among the plant traits (Valencia et al., 2015). We utilized the PCA coordinates of the two main components with eigenvalues greater than 1 to replace most individual plant trait values in subsequent data analyses. By employing PCA with CWM and Rao, we were able to identify the functional composition and diversity of plant traits covarying across plant communities, while also focusing on independent variables in further analyses.
The PCA of CWM values revealed two main components that accounted for 78.84% of the total variance (Fig. S1A). The first component, which explained 46.53% of the variance, was associated with plant structural traits (hereafter referred to as CWM-plant structure traits) and showed positive correlations with the CWM of vegetative height and leaf lateral spread. A high value of the CWM-plant structure traits indicates that plant communities are dominated by species with tall vegetative height and broad leaf lateral spreads. The second component, accounting for 32.31% of the variance, was linked to plant nutrient traits (hereafter CWM-plant nutrient traits), positively correlating with the CWM of leaf N, P, and Na. A high value of the CWM-plant nutrient traits indicates that these communities are characterized by species with elevated leaf N, P, and Na concentrations.
Similarly, the PCA of Rao values identified two main components that accounted for 63.19% of the total variance (Fig. S1B). The first component, explaining 32.61% of the variance, was related to the diversity of plant structure traits (hereafter Rao-plant structure traits), incorporating the Rao values of vegetative height and leaf lateral spread. A high value of the Rao-plant structure traits indicates that the plant communities exhibit substantial variance in vegetative height and leaf lateral spread. The second component, accounting for 30.50% of the variance, was associated with the diversity of plant nutrient traits (hereafter Rao-plant nutrient traits), including the Rao values of leaf N, P, and Na concentrations. A high value of the Rao-plant nutrient trait suggests that these communities display significant variance in leaf N, P, and Na concentrations.
4. Arthropod sampling and identification
This study did not require ethical approval for arthropods sampling. Arthropods were collected from experimental plots using sweep net sampling (Ø32.0 cm) during the growing season (mid-August 2018). In each plot, arthropods were sampled with 50 sweeps using a muslin net between 10 am and 4 pm on rain-free days. The sweeping technique involved 180° arcs through the vegetation canopy, quickly reversing direction at the end of each arc (Doxon et al., 2011). The contents of the net were collected and stored in bottles containing ethyl acetate for later sorting. In the laboratory, arthropods were manually sorted into different orders and identified to the genus and species level using optical microscopy. Unidentified specimens were preserved in 75% ethanol and sent to taxonomists for morphospecies identification. In this study, we categorized all arthropods into distinct morphospecies. However, we opted to utilize species richness as our primary biodiversity metric instead of relying on higher taxonomic richness because species richness provides a more consistent and broadly applicable measure of biodiversity across various context. Based on observations and literature (Carmona et al., 2011), each morphospecies was classified into one of two trophic categories (herbivores and natural enemies; Table S3) and assigned to one of three feeding guilds (Carmona et al., 2011, Lu et al., 2022, Pratt et al., 2017). Herbivorous guilds included (1) sucking/piercing herbivores, (2) chewing herbivores, and (3) boring/mining herbivores , while enemy guilds comprised (1) parasitoids, (2) predators, and (3) spiders (Lu et al., 2022).