Identifying the key drivers of changes in species biomass, abundance, and distribution across trophic levels in terrestrial ecosystems represents a fundamental challenge in ecology. The shape of trophic pyramids—reflecting the relative biomass or abundance of various trophic groups, such as herbivores, omnivores, predators, and parasitoids—is anticipated to vary with both the quantity and nutritional quality of plants. However, a comprehensive understanding of how these factors influence the shape of trophic pyramids in grassland ecosystems remains elusive. In this study, we utilize replicated monocultures of 15 herbaceous species in the Inner Mongolian grassland to investigate the effects of plant biomass, macronutrients (including protein, carbohydrate, and phosphorus), and micronutrients (such as sodium) on the structure of arthropod trophic pyramids and the abundances within different trophic levels. Our results demonstrated that plant biomass, leaf protein-to-carbohydrate ratio, and sodium content collectively contributed to a top-heavy structure in arthropod trophic pyramids, characterized by a relatively higher proportions of predators and parasitoids. Specifically, plant biomass enhanced this top-heaviness both directly, by increasing the abundance of predators and parasitoids, and indirectly, by increasing chewing and sucking herbivores, which in turn bolstered the abundance of predators and parasitoids.  Furthermore, leaf sodium and the protein-to-carbohydrate ratio positively influenced parasitoid abundance through increasing the abundance of sucking herbivores and endophytes. Given that climate change and human activities, such as nitrogen fertilization and saline water irrigation, are altering plant biomass, nutrient composition, and sodium availability globally, our findings suggest that these changes will have significant cascading effects on arthropod trophic structures and overall ecosystem functioning.

Study site

The monoculture experiment was established in June 2014 at the Inner Mongolia Grassland Ecosystem Research Station (IMGERS, 116°42′E, 43°38′N) of the Chinese Academy of Sciences, located in the Xilin River Basin of Inner Mongolia, China (Bai et al. 2004). The mean annual temperature of the study area is 0.3°C, with the lowest and highest monthly mean temperature occurring in January and July, respectively.  The mean annual precipitation is 346 mm, with more than 80% of the annual rainfall falling in the growing season from April to September. The plant community is dominated by Leymus chinensis (perennial rhizome grass) and Stipa grandis (perennial bunchgrass), both of which are widely distributed in the Eurasian steppe (Bai et al. 2004).  The soil texture of the study area is loamy sand (Calcic Chernozem according to the ISSS Working Group RB, 1998).

To remove plants and seed prior to experiment establishment, a 40 × 40 m area was bulldozed and manually ploughed, removing 10 cm of top soil. The area was then divided into four blocks, with each block containing 20 plots (each 1.2 × 1.2 m), and the entire area was fenced. Within each block, the 20 most common native species in the regional plant community were established in monocultures within randomly assigned plots.  Seeds were sourced locally, being collected by hand in 2013. Of the initial 20 plant species, only 15 established within plots (Appendix S1: Table S1; 60 plots).

Plant biomass and leaf nutrients measurements

In mid-August 2018 and 2019, plant biomass was sampled by clipping all plants in a 50 × 50 cm quadrat at the soil surface within each plot. All plants were oven-dried at 65°C for 48 h, and weighed to measure dry plant biomass (g m^-2) of each plot (Bai et al. 2004). Leaf concentrations of non-structural carbohydrates, protein, phosphorus, and sodium were measured from five leaves each from three individual plants/ plot, as these are important determinants of plant nutritional quality for arthropods (Awmack and Leather 2002, Joern et al. 2012, Kaspari 2021). Leaves were dried for 24 h at 65°C, ground using a ball mill prior to chemical analysis. Leaf phosphorus and sodium concentrations were measured using an elemental analyzer (VarioEL Element Analyzer; Hanau, Germany), protein content was measured in the Bradford assay, and non-structural carbohydrates (used as carbohydrates in the leaf protein-to-carbohydrate ratio) were measured using the phenol–sulphuric acid method (Clissold et al. 2006). Leaf nutrient concentrations were averaged within a plot, and the plot was used as the unit of replication prior to data analysis. Monocultures of different plant species covered a wide range of plant biomass, macro- and micronutrient variations.

Arthropod sampling and identification

Arthropods were collected from the monoculture plots using a sweep net (diameter 32.0 cm) in mid-August 2018 and 2019. Sampling took place between 10:00 am and 4:00 pm on rain- free days. We conducted 50 sweeps per plot (25 sweeps per sampling period, with two sampling periods in August). Each sweep involved a 180° arc through the vegetation canopy, followed by a quick turn and reverse to capture vegetation-dwelling arthropods.  We varied the sweep direction to ensure comprehensive coverage. Directions A and B consisted of 8 sweeps each, as the 32.0 cm-diameter net effectively covered the plot width. Direction C involved 9 sweeps to account for the longer diagonal distance.   The contents of the sweep net were put in bottles containing ethyl acetate until sorting.  In the laboratory, all arthropod individuals were identified to species by optical microscopy, or to morphospecies identified to genus or family. Based on personal observations and a literature review (Carmona et al. 2011, Lu et al. 2022), each morphospecies was assigned to one of four trophic categories: herbivores, omnivores, predators, and parasitoids.  Herbivores were further assigned to one of the three feeding guilds: sucking/piercing herbivore, chewing herbivore, and endophytes (Carmona et al. 2011, Lu et al. 2022), based on whether specimens had chewing mouthparts, sucking mouthparts or fed within tissues, respectively. However, sweep net sampling may be less effective for capturing ground‑active omnivores and predators, such as ants and beetles, as well as endophytic insects that inhabit plant tissues.  We further quantified abundance of herbivores, omnivores, predators, and parasitoids across the monoculture plots of different plant species.

Community trophic mean

We calculated the community trophic mean (CTM) of arthropod trophic positions for each monoculture plot (Ricotta and Moretti 2011, Welti et al. 2020).

We assigned a trophic level value of 2 to herbivores and of 2.5 to omnivores. Omnivores, such as ants, occupy a trophic position that is intermediate between herbivores and predators. This classification is supported by N isotopic analyses across a diverse range of grasslands, which indicate an average trophic position of 2.5 for omnivores (Welti et al. 2020). Predators and parasitoids were both assigned a trophic level of 3. While predatory arthropods may serve as hosts for parasitoid species (Fei et al. 2023), this interaction could suggest a higher the trophic level for parasitoids as a group. However, most parasitoids are specialized, primarily targeting herbivorous insects as their hosts (Kraaijeveld et al. 1998, Frago and Zytynska 2023).  In contrast, predators exhibit a more polyphagous feeding, strategy, preying on a wider variety of organisms, including herbivores, other predators, and parasitoids (Hurd and Eisenberg 1990, Rosenheim et al. 1995, Rosenheim 1998, Colfer and Rosenheim 2001). Therefore, it is not justified to assign a higher trophic level to parasitoids than to predators. CTM of trophic level simplifies biological complexity and is a straightforward indicator of community trophic structure (Welti et al. 2020). A high CTM of trophic level indicates a top-heavy trophic pyramid (e.g., relatively higher proportions of predators and parasitoids) while a low CTM of trophic level indicates a bottom-heavy trophic pyramid (e.g., relatively higher proportions of herbivores and omnivores; Welti et al. 2020). Because the high mobility and feeding activity of chewing insects, such as grasshoppers, across different monoculture plots may influence CTM estimates, we also recalculated the CTM after excluding these insects.