The interactions between adjacent trophic levels are essential for ecosystem functioning and stability. Grazing by domestic herbivores is an essential interaction in grasslands, but little information is available on the nature of relationship between plant and insect diversity under grazing by large herbivores.
We examined the effects of large herbivores on the relationship between plant and insect diversity with five grazing treatments (control, cattle, goat, sheep and a mixture of the three grazing types) across three plant diversity levels (low: 4–5 species, intermediate: 8–9 species and high: 15–17 species) in a meadow steppe.
We found that the grazing treatments did not significantly affect plant species richness, but reduced plant biomass, plant height and cover. Grazers affected variation in plant height differently at different plant diversity levels, and this variation increased at the low plant diversity level and decreased at the high plant diversity level after grazing. A similar pattern was observed for insect species richness: grazing had a positive impact at the low plant diversity level, but had a negative impact at the high plant diversity level.
In the absence of grazing, insect species richness was positively associated with plant species richness, but it decreased with increasing plant diversity in the grazing treatments. This was attributed to strong responses of insect species richness to plant height heterogeneity under grazing by large herbivores, implying that plant structural heterogeneity is more important than plant diversity in influencing insect diversity in grazed grasslands.
Synthesis and applications. Grazing by large herbivores may reverse the positive relationship between plant diversity and insect diversity by modifying plant structural heterogeneity. Therefore, the spatial heterogeneity of vegetation structure should be given more attention in future work on plant–insect interactions. This study further highlights the importance of using large herbivore grazing in management actions, not only to maintain diversity but also to mediate trophic interactions in grasslands.
One of the most dramatic consequences of global change is the rapid loss of biodiversity in many ecosystems (Sala et al. 2000). This loss of biodiversity will have profound impacts on ecosystem functioning and processes such as primary productivity and interactions between trophic levels (Hooper et al. 2005; Burns, Collins & Smith 2009). The relationship between species diversity and the interaction among trophic levels is important because of the recent interest in using food web theory to predict and understand community patterns, as well as proving essential for biodiversity conservation planning (Gamfeldt, Hillebrand & Jonsson 2009; Srivastava & Bell 2009). Recent studies reporting the effects of species diversity on trophic interactions further indicate a growing interest on this topic (Thébault & Loreau 2005; Schmitz 2007; Unsicker et al. 2010; Wang et al. 2010a).
According to the bottom-up effects hypothesis, the abundance or diversity of a lower trophic level controls that of an adjacent, higher trophic level (Hunter & Price 1992). Plant diversity has long been recognized as an important determinant of the abundance and diversity of herbivorous insects (Root 1973). Increases in the number of plant species in natural ecosystems consistently results in higher insect species richness and abundance (Knops et al. 1999; Haddad et al. 2001).
However, within grassland ecosystems, plants and insects often coexist with large herbivores, potentially confounding the ecological relationship between plant and insect diversity. Large herbivores play important roles in shaping plant community structure through selective foraging and trampling (McNaughton et al. 1989; Burns, Collins & Smith 2009). Studies have shown that grazing by large herbivores may affect plant diversity in complex ways, including increasing plant species diversity by directly consuming competitively dominant plant species and accelerating nutrient cycling through their excretions (Smith et al. 2000), decreasing plant species diversity by over-foraging preferred species (Wardle et al. 2001; Howe, Brown & Zorn-Arnold 2002), or have no effects on plant species diversity (Adler et al. 2005). Outcomes of grazing depend on factors such as the type of large herbivores, timing, intensity, and frequency of grazing, and soil and climatic factors (Olff & Ritchie 1998; Bakker et al. 2006).
Large herbivores can affect species diversity of other trophic levels directly or indirectly (Murray, Frank & Gehring 2010; Veen et al. 2010). Indirect effects are particularly common in insects (Kruess & Tscharntke 2002; Joern 2005; Davidson et al. 2010). Leaf-feeding insects can be depressed by large herbivores because they compete for the same plants (Bailey & Whitham 2003). Similarly, vertebrate herbivores reduce the abundance of caterpillars and grasshoppers (Huntzinger, Karban & Cushman 2008) and beetles (Dennis et al. 2008) because of limited food supplies and defensive structures of plants induced by grazing. In contrast, grazing by large herbivores may also facilitate insect diversity (Cagnolo, Molina & Valladares 2002). For example, bison grazing has been shown to result in higher grasshopper diversity by increasing plant species richness (Joern 2005).
Most of the studies reporting the effects of large herbivores on biodiversity focus on the diversity of a single trophic level, either plants or insects, while the influence of large herbivore grazing on the relationship between plant and insect diversity has rarely been examined. One of the major complications in understanding this linkage is that the effects of grazing on plant and insect diversity are not independent. Thus, whether the effects of grazing on the relationship between plant and insect diversity operate via changes in plant diversity or via altered insect diversity still remains to be tested (Joern 2005; Dennis et al. 2008). In addition, changes in plant structure and plant productivity caused by grazing are likely to play important roles in affecting insect diversity (Johnson & Matchett 2001) because vegetation clearly acts as the physical habitat and food resources for most insect species. Therefore, plant structure–induced and/or plant productivity–induced changes in insect diversity resulting from grazing may determine the responses of relationship between plant and insect diversity to grazing. However, to the best of our knowledge, no previous studies have empirically examined the links in the relationship between plant diversity and insect diversity to grazing by large herbivores.
We conducted an experiment in the native grassland of eastern Eurasian steppes to determine the effects of large herbivore grazing on the relationship between plant and insect diversity. Specifically, we addressed the following questions: (i) how does large herbivore grazing affect vegetation characteristics such as plant diversity, plant structure, plant biomass, plant structural heterogeneity and the resulting insect diversity pattern? (ii) Does insect diversity increase along the gradient of increasing plant diversity in the absence of large herbivores? And (iii) does grazing by different large herbivores alter the assumed positive relationship between plant diversity and insect diversity?
Materials and methods
The study site is located at the Grassland Ecological Research Station of Northeast Normal University, Jilin Province, China (44°45′N, 123°45′E). The site has a semi-arid, continental climate with mean annual temperature ranges from 4·6 to 6·4 °C, and annual precipitation ranges from 280 to 400 mm. More than half of the precipitation is received during the growing season, especially between June and August. Annual potential evapotranspiration is approximately three times as much as annual precipitation. Soils are mixed saline and alkaline (pH 8·5–10).
The study site lies in the eastern region of the Eurasian Steppe Zone. The main vegetation type is meadow steppe dominated by two grasses Leymus chinensis Tzvel. and Stipa baicalensis Roshev. (Wang & Ba 2008). Other species include grasses such as Phragmites australis Trin., Calamagrostis epigejos Roth. and Chloris virgata Swartz; legumes such as Lespedeza davurica Schindler and Midicago ruthenica C. W. Chang; and forbs such as Kalimeris integrifllia Turcz., Potentilla flagellaris Willd., Artemisia scoparia Waldstem et Kitailael and Carex duriuscula C. A. M.
Experimental design and animal management
A relatively flat area within the study site with homogenous soil conditions was enclosed in 2006 to create nine blocks. Fifteen 5 × 5 m quadrats were randomly placed along each of four parallel transects in each block. Percentage cover and number of individuals of each species within these quadrats were recorded in August 2006. The importance value (I.V. = relative cover + relative density + relative frequency, Odum 1971) of each species in each block was calculated. Species with an importance value >1 was used to evaluate and classify the nine blocks (Table S1, Supporting Information). The nine blocks were classified into three plant diversity levels: low (4–5 species with an importance value >1), intermediate (8–9 species) and high (15–17 species), resulting in three replicates (blocks) for each level of diversity.
Five plots were established in each block enclosed with barbwire and randomly assigned to the following grazing treatments: (i) grazing by cattle (a hybrid of native and yellow breed), (ii) grazing by goats (Liaoning Cashmere breed), (iii) grazing by sheep (small-tail Han breed), (iv) grazing by cattle, goats and sheep and (v) no grazing (control; Fig. S1, Supporting Information). The distance between neighbouring plots was about 18 m.
A moderate grazing intensity with approximate 60% of above-ground biomass removal was used in all grazing treatments. This was achieved using two cattle (221 ± 5·5 kg) at the stocking rate of 7·14 sheep ha−1, eight goats (34 ± 1·6 kg) at the stocking rate of 7·08 sheep ha−1 or eight sheep (33 ± 1·6 kg) at the stocking rate of 7·21 sheep ha−1. The mixed-grazing treatment was comprised of two cattle, eight goats and eight sheep at the stocking rate of 7·15 sheep ha−1. Plot size was 0·05 ha for all grazing treatments except the mixed-grazing treatment, which was 0·15 ha, so that comparable stocking rate can be achieved. Grazing started from the second week of July in 2007 and 2008, respectively, to ensure there was sufficient new growth for grazing. Grazing occurred twice a day, from 06:00 to 08:00 h and from 16:00 to 18:00 h, and was terminated when about 60% of the available forage had been removed. Large herbivore grazing was delayed immediately after a significant precipitation event (>8 mm).
Vegetation attributes, such as species richness, plant height and cover were measured along two, 20-m cater-corner transects in each plot. Ten 0·25 × 0·25 m quadrats were evenly placed along each transect. All plant species within the quadrat were identified, and the percentage cover of each species was visually estimated; plant height was measured to the nearest centimetre using a ruled rod (Joern 2005). Structural heterogeneity was estimated as the coefficient of variation (CV) of plant height in each quadrat. Above-ground biomass was measured by clipping standing plant materials to 1 cm above ground level using shears from five randomly located 0·5 × 0·5 m quadrats per plot. Plant material was oven-dried for 48 h at 80 °C to obtain dry weights. Vegetation measurements were conducted at the beginning (July), middle (August and September) and end (October) of the growing season in 2008.
Insect sampling and identification
We followed the standard sweep net survey method (40 cm in diameter) to estimate insect species richness and abundance (Evans, Rogers & Opfermann 1983; Haddad et al. 2001; Joern 2005; Schaffers et al. 2008). Insects were collected four times from July to October 2008. Insect specimens were collected only under favourable conditions (sunny days with minimal cloud cover and calm or no wind) from 09:00 to 15:00 h. Insects were sampled along two, 2-m wide parallel transects in each plot. These transects were at least 1 m away from the plot boundary to minimize edge effects. Each net was swept vigorously side to side while walking steadily along these transects. Each sample consisted of 30 sweeps, and two samples were carried out in each plot to ensure that those samples were representative. The contents of the sweep nets were preserved in bottles containing ethyl acetate. All grazing plots were sampled on the same days in random order in each sampling date. All individuals were identified to species but only adult insects were counted. Specimens that could not be identified to species were separated into recognizable taxonomic units. We collected 37 962 individual insects belonging to 57 families and 9 orders: Orthoptera, Hemiptera, Coleoptera, Diptera, Hymenoptera, Homoptera, Lepidoptera, Mantodea and Neuroptera (Table S2, Supporting Information). Approximately 84% of specimens were identified to species with the remaining being identified to genus (7%), family (6%) or order (3%).
Insect abundance was quantified as the total number of adult individuals. They were further placed into one of the four feeding guilds: herbivores, predators, detritivores or parasitoids (Perner et al. 2005). A small number of morphospecies (<1%) could not be sufficiently identified for guild placement and were thus omitted from analysis.
Subsamples of plant biomass, plant height, plant cover and variation in plant height were pooled according to plots. The cumulative species richness of plant and insect and the accumulative abundance of insect (both total abundance and the abundance of each trophic guild) throughout sampling periods in given year were analysed.
All data were assessed for normality using sas (Version 6.12; SAS Institute Inc. 1989). Two-way analyses of variance (anova) with plant diversity levels, herbivores treatments and their interaction as fixed effects were used. Tukey's multiple comparison was used for post hoc analysis of significant differences among factors. Data were further analysed using one-way anova within each plant diversity level or grazing treatment if the interaction between the two was significant. Significant level was set at P ≤0·05.
Plant species richness averaged 11, 13 and 16 per twenty 0·25 × 0·25 m quadrats in the low, intermediate and high diversity levels, respectively (data not shown). Please note that this was different from the pre-treatment plant richness because different sampling methods were used (see descriptions in 'Materials and methods' for details).
Plant cover was significantly affected by the interaction between plant diversity and large herbivores (Table 1). At low and intermediate plant diversity levels, plant cover was significantly reduced by all grazing treatments, with the exception of the sheep treatment that was not significantly different from any treatments (Fig. 1a). At high plant diversity level, plant cover was markedly reduced by all grazing treatments apart from the mixed grazing that was not significantly different from the controls but higher than that of all other grazing treatments (Fig. 1a). In the no-grazing controls, plant cover at high plant diversity was greater than that of intermediate diversity (P =0·022), a similar pattern was also observed in the mixed-grazing treatment (P =0·029). Plant cover was not significantly different among plant diversity levels within the goat-grazing treatment (P =0·539). Within the cattle-grazing treatment, plant cover was lower at intermediate plant diversity than that of low plant diversity (P =0·012), but there was no significant difference between low and high plant diversity levels. Plant cover within the sheep-grazing treatment decreased gradually from low to high plant diversity level (P =0·02).
Table 1. Two-way anova table of the effects of large herbivore grazing and plant diversity on plant species richness, plant cover, plant height, variation in plant height and plant biomass
Plant species richness
Variation in plant height
Plant diversity (PD)
PD × G
Plant height significantly decreased with increasing plant diversity and was markedly reduced by all grazed treatments compared with the controls (Tables 1 and 2).
Table 2. Plant height and plant biomass in different grazing treatments (control, cattle, goats, sheep and mixed grazing) at low, intermediate and high plant diversity levels
Plant diversity levels
Plant diversity effects
Values are means ± SE. Significant main effects of grazing by large herbivores are indicated by small letters (a–c), and significant main effects of plant diversity are indicated by capital letters (A, AB and B). Values with different letters are significantly different within each grazing treatments or each plant diversity levels (P ≤0·05).
Plant height (cm)
45·9 ± 0·5
42·4 ± 1. 5
39·6 ± 0·8
35·4 ± 1·4
39·9 ± 1. 5
40·6 ± 1·0A
42·7 ± 2·2
31·2 ± 1·1
37·8 ± 5·28
37·14 ± 6·14
41·7 ± 2. 9
35·9 ± 1·2AB
36·9 ± 1·6
27·9 ± 1·6
27·1 ± 2·91
30·7 ± 3·3
32·5 ± 1·9
31·0 ± 1·3B
41·5 ± 1·5a
34·3 ± 2·3b
33·5 ± 2·1b
34·4 ± 1·64b
35·6 ± 1·4b
Plant biomass (g sampling quadrat−1)
29·3 ± 2·9
19·8 ± 1·3
24·3 ± 1·0
24·4 ± 0·5
22·5 ± 1·0
24·1 ± 1·0B
33·7 ± 1·2
28·0 ± 2·2
25·0 ± 1·5
22·3 ± 2·6
28·2 ± 1·8
27·4 ± 1·5A
38·4 ± 2·8
26·3 ± 0·7
24·3 ± 1·9
26·4 ± 0·9
27·8 ± 3·1
28·7 ± 1·5A
33·8 ± 1·7a
24·7 ± 1·5b
24·5 ± 0·8b
24·4 ± 1·0b
26·2 ± 1·4b
Variation in plant height
The interaction between grazing and plant diversity had a significant effect on the variation in plant height (Table 1). At low plant diversity level, variation in plant height was significantly higher in cattle- and sheep-grazing treatments than that in the controls, but variation in plant height was dramatically higher in the mixed-grazing treatment than that in no-grazing controls at the intermediate plant diversity level (Fig. 1b). At the high plant diversity level, all grazed treatments significantly reduced the variation in plant height compared with ungrazed treatment. In no-grazing controls or goat-grazing treatment, variation in height was not significantly different among plant diversity levels (Fig. 1b). Variation in plant height was significantly reduced by cattle and sheep grazing at the intermediate and high plant diversity levels, compared with the low plant diversity level (P =0·007 for cattle, P =0·039 for sheep). Variation in plant height was significantly greater at the intermediate plant diversity level than that of high plant diversity level (P =0·017), but neither was significantly different from low plant diversity level in the mixed-grazing treatment (Fig. 1b).
Plant biomass was significantly higher at intermediate and high diversity levels than low diversity level and was dramatically reduced by all grazing treatments compared with the controls (Tables 1 and 2).
Insect species richness
Insect species richness was affected by the interaction between plant diversity and grazing (Table 3). At the low plant diversity level, insect species richness was enhanced by all grazing treatments except goat-grazing treatment in which insect species richness did not significantly differ from other treatments (Fig. 2a). At the intermediate plant diversity level, mixed grazing resulted in greater insect species richness than that of other grazing treatments, but none was significantly different from the controls (Fig. 2a). At the high plant diversity level, insect species richness was reduced by cattle and goat grazing but not sheep or mixed grazing (Fig. 2a). Insect species richness was higher at intermediate and high plant diversity levels than that of low plant diversity level in the no-grazing controls (P =0·032), but the trend was reversed within the sheep-grazing treatment (P =0·032). Insect species diversity decreased from low to high plant diversity levels within cattle-grazing treatment (P =0·026), but was not significantly different among plant diversity levels within the goat-grazing treatment (P =0·574). Within the mixed-grazing treatment, insect species richness was higher at the intermediate plant diversity level than that at the high plant diversity level (P =0·015).
Table 3. Two-way anova table of the interactive effects between plant diversity and large herbivore grazing on insect species richness, insect abundance and abundance of four insect guilds, including herbivores, predators, detritivores and parasitoids
Insect species richness
Plant diversity (PD)
PD × G
Insect abundance was significantly affected by the interaction between grazing and plant diversity (Table 3). Insect abundance significantly increased in grazed plots (Fig. 2b). In the low plant diversity plots, insect abundance was significantly higher in all grazing treatments than that in the controls, except for the goat-grazing treatment, and abundance was highest in sheep- and mixed-grazing treatments (Fig. 2b). At the intermediate plant diversity level, insect abundance was dramatically higher in goat and mixed-grazing treatments than that in the controls, cattle- and sheep-grazing treatments (Fig. 2b). At the high plant diversity level, insect abundance was highest in the mixed-grazing treatment, was lowest in the controls and cattle-grazing treatments and was intermediate in goat- and sheep-grazing treatments (Fig. 2b). Within the controls, cattle- or goat-grazing treatment, insect abundance was significantly higher at intermediate and high plant diversity levels than that of low plant diversity level (P =0·005 for controls; P =0·031 for cattle; P <0·0001 for goat). Within the sheep-grazing treatment, insect abundance was markedly higher at the high plant diversity level than that of low and intermediate plant diversity levels (P =0·019). Within the mixed-grazing treatment, insect abundance significantly increased along the gradient of increasing plant diversity (P =0·001).
Herbivorous insects were the most abundant feeding guild (95·7%) followed by predators (3·8%), parasitoids (0·3%) and detritivores (0·2%; Figs 2c and 3). The interaction between plant diversity level and grazing significantly affected herbivore insect abundance, but not the abundance of predator, detritivore or parasitoid insects (Table 3). In the low plant diversity plots, herbivore insect abundance was higher in cattle-, sheep- and mixed-grazing treatments than that in the controls, and it was higher in sheep- and mixed-grazing treatments than that in the goat-grazing treatment (Fig. 2c). In the intermediate plant diversity plots, herbivore insect abundance was higher in the goat- and mixed-grazing treatments. Herbivore insect abundance was highest in the mixed-grazing treatment, lowest in the controls and intermediate in the cattle, goat- and sheep-grazing treatments at the high plant diversity level (Fig. 2c).
The abundance of predators, detritivores and parasitoids was not significantly different among plant diversity levels, but was markedly affected by grazing (Table 3). Predator abundance was significantly higher in the mixed-grazing treatment than other grazing treatments; none of the grazing treatments was significantly different from the controls (Fig. 3a). Detritivore abundance was highest in the mixed-grazing treatment, lowest in the controls and sheep-grazing treatments and intermediate in the cattle- and goat-grazing treatments (Fig. 3b). Parasitoid abundance was significantly higher in the controls, goat- and mixed-grazing treatments than that in the cattle-grazing treatment (Fig. 3c).
The positive relationship between plant and insect diversity has been reported in other experimental manipulations and observational studies in a number of ecosystems (Knops et al. 1999; Haddad et al. 2001). Our study also confirmed this relationship in the absence of large herbivores. However, we found that the positive relationship between plant diversity and insect diversity was reversed by large herbivore grazing (Fig. 4b). Almost all previous studies on the relationship of plant diversity and insect diversity were conducted without natural disturbance (Koricheva et al. 2000; Haddad et al. 2009). Our study indicated that the relationship between plant and insect diversity in natural grazing grasslands is more complex than previously thought.
We found that large herbivore grazing did not affect plant species richness, but significantly altered insect species richness, and the responses of insect to grazing varied with plant diversity levels (Fig. 2a). Therefore, grazing by large herbivores probably affects insect diversity by modifying other vegetation characteristics such as community structure and productivity. The ‘resource productivity’ hypothesis predicts that more productive plant community supports more abundant insect species because of ample food supplies (Rosenzweig & Abramsky 1993; Perner et al. 2005). In our study, large herbivore grazing significantly reduced plant biomass at all plant diversity levels (Table 2), but the reduction in plant biomass did not consistently affect insect species richness. Insect species richness was reduced at the high plant diversity level but enhanced at the low plant diversity level (Fig. 2a). It is commonly understood that insect diversity responds to other aspects of the plant community and that taller vegetation supports more insect species (Treweek, Watt & Hambler 1997; Pöyry et al. 2006). Our results show that the effect of plant height on insect species richness was inconsistent (Table 2; Fig. 2a). Therefore, changes in plant productivity and plant height cannot explain the variation in insect diversity at all plant diversity levels, which differs from former studies, where differences in insect diversity were simply attributed to differences in plant productivity and plant architecture (Hartley, Gardener & Mitchell 2003; Woodcock et al. 2009).
Plant structural heterogeneity may affect species diversity at higher trophic levels (Agrawal, Lau & Hamback 2006; Beals 2006; Wang et al. 2010b). Our results show that grazing increased heterogeneity in plant height (measured as the coefficient in variation) at the low plant diversity level and decreased it at the high plant diversity level (Fig. 1b). Grazing-induced changes in plant height heterogeneity matched those in insect species richness at all levels of plant diversity (Fig. 4a,b). Increased structural heterogeneity by grazing offers a greater range of microsites for oviposition and hibernation for insects (Lawton 1983) and provides refuge from natural enemies (Joern 2002). Furthermore, heterogeneous vegetation structure may facilitate thermoregulation of insects, which is important for many insect activities, especially digestion and resource acquisition (Yang & Joern 1994). Increased structural heterogeneity leads to more suitable habitat structure and an increased number of potential food plant species for insects (Joern 2005). Therefore, more heterogeneous structure increases the availability of both food resources and suitability of habitat, supporting a variety of insect species.
Our results demonstrate that plant structural heterogeneity may be more important than other vegetation characteristics in predicting insect assemblages. But this result does not deny the importance of plant diversity, because insect diversity was positively related to plant diversity when structural heterogeneity was not changed because of the lack of grazing (Fig. 4b). Insect diversity responded strongly to the change in plant heterogeneous structure induced by grazing in our study. It has been suggested that plant structural heterogeneity induced by disturbance may encompass other decisive factors, such as plant species richness and environmental conditions (Randlkofer et al. 2010). Results from this study are consistent with emerging paradigms that emphasize the importance of increasing disturbance and heterogeneity in promoting biodiversity (Fuhlendorf & Engle 2001; Joern 2005).
The reversed relationship between plant diversity and insect diversity under grazing is attributed to the strong response of insect species richness to structural heterogeneity of plant community induced by large herbivores (Fig. 4a,b). The effects of grazing on vegetation vary with time and space (Olff & Ritchie 1998). Large herbivores may cause changes in plant species richness and composition of plant community over long time-scales (Collins & Smith 2006), influencing insect diversity at time-scales longer than our study (Jonas & Joern 2007). Therefore, the underlying mechanisms regulating the relationship between plant diversity and insect diversity under grazing may vary at different temporal and spatial scales. Further studies focusing on these underlying mechanisms will further reveal dynamic relationship between plants and insects.
The magnitude of changes in insect abundance induced by grazing was enhanced along the gradient of increasing plant diversity (Fig. 2b). Insect abundance was positively correlated with plant species richness (Haddad et al. 2001), which was verified by our study. Insect abundance was higher in grazed than ungrazed plots, possibly due to more available oviposition sites (Fisher 1994), more new growth to attract insects (Gao et al. 2008) and changes in microclimate favouring hatching of insects (Loftin et al. 2000) following grazing. Furthermore, grazing induced a change in trophic composition with increased herbivore abundance and decreased the numbers of predators when a single large herbivore species was present and clearly impoverished the abundance of herbivore, predator and detritivore when a mixture of large herbivores species was present (Figs 2c and 3). The causes of changes in trophic composition under grazing are unclear as we were unable to separate the complex response to large herbivores at each trophic level in this study. These possibilities and alternative explanations merit further exploration and have implications for the management of grasslands using large herbivore grazing.
In conclusion, our results show that the relationships among plant, insect and large herbivore grazing in grassland ecosystems are more complex than previously thought. Insect diversity was positively related to plant diversity in the absence of large herbivores, but large herbivore grazing reversed this positive relationship. The changes were attributed to the strong responses of insect diversity to plant structural heterogeneity that was directly influenced by grazing. Grazing by large herbivores has the potential to alter the delicate balance of neighbouring trophic relationship, influencing processes and functions of grassland ecosystems.
We greatly thank Oswald Schmitz for useful comments and editing on an earlier revision of the manuscript, Bingzhong Ren and Wentao Gao for the identification of insect species, Yuanyuan Peng, Shengping Li and Jushan Liu for their assistance in the field and laboratory. This project was supported by the National Natural Science Foundation of China (No. 31070294, 31072070, 31100331), NECT-11-0612 and the State Agricultural Commonweal Project (201003019).