1. Introduction

Herbivory is an important ecosystem function that mediates species coexistence and energy flow (Rosado-Sánchez et al., 2018; Muiruri et al., 2019). In forest ecosystems, herbivore feeding behavior and intensity have been shown to be affected by plant diversity and composition (Zhang et al., 2017; Grossman et al., 2019; Wang et al., 2023). Conversely, herbivores have been shown to alter plant survival and growth through feeding, which in turn affects the composition and dynamics of plant communities (Zvereva et al., 2012). Clarifying the interactions between plants and herbivores is critical for both forest pest management and our understanding of forest ecology and species coexistence (Cardinale et al., 2006; Wang et al., 2022; Li et al., 2023). However, elucidating the role of various factors that regulate these biodiversity-herbivore interactions has been a major challenge (Vehviläinen et al., 2007; Castagneyrol et al., 2013; Guyot et al., 2015; Kambach et al., 2016; Halliday et al., 2025).

Two competing hypotheses have been proposed to explain how biodiversity affects herbivory (Tahvanainen and Root, 1972; White and Whitham, 2000; Schuldt et al., 2010). The associational resistance (AR) hypothesis states that host plants in more diverse communities experience lower herbivory (Tahvanainen and Root, 1972; Root, 1973; Bommarco and Banks, 2003; Jactel et al., 2006). In contrast, the associational susceptibility (AS) hypothesis proposes that plant diversity may facilitate herbivore foraging or spillover, especially when preferred host resources become scarce (Otway et al., 2005; Barbosa et al., 2009), leading to greater overall damage in species-rich communities (White and Whitham, 2000). Evidence in support of either hypothesis has been mixed, especially for the vulnerable early-successional stage (Schuldt et al., 2014; Damien et al., 2016; Grossman et al., 2019).

Previous studies have indicated that several factors may determine whether associational resistance or associational susceptibility is conferred to trees in an ecosystem. Trees in more diverse forest ecosystems have been shown to suffer less herbivore damage (Setiawan et al., 2014; Damien et al., 2016; Guyot et al., 2016), although herbivore damage has been indirectly associated with neighborhood diversity, as higher diversity increases herbivore abundance and species richness (Scherber et al., 2010; Brezzi et al., 2017). Herbivore damage has been shown to be significantly influenced by species identity, i.e., plant species composition and functional traits (Sobek et al., 2009; Setiawan et al., 2014; Grossman et al., 2019). For example, herbivory levels have been found to be affected by leaf morphological characteristics (e.g., SLA, leaf toughness) (Poorter et al., 2004; Schuldt et al., 2012). In addition, studies have found that the feeding of herbivores increases as leaf toughness and C: N decrease (Poorter et al., 2004; Salgado-Luarte and Gianoli, 2012). Previous studies have also reported that herbivory is influenced by plant height apparency (Grossman et al., 2019). Plant apparency refers to the likelihood of a plant being detected by herbivores (Feeny, 1970; Endara and Coley, 2011), i.e., taller trees, which are more apparent than their neighboring plants, may experience higher levels of herbivore damage (Castagneyrol et al., 2013; Guyot et al., 2015; Damien et al., 2016). However, studies on these factors have generally produced conflicting results (Plath et al., 2011; Setiawan et al., 2014; Damien et al., 2016; Guyot et al., 2016). Furthermore, despite their acknowledged importance, no studies have yet examined how tree species identity, local neighborhood diversity, and focal-tree apparency together may affect herbivory in forests.

Most studies examining the diversity–herbivory relationship have focused on mature forest systems (Jactel et al., 2006; Schuldt et al., 2010; Zhang et al., 2023; Jia et al., 2024). However, early-successional stage forests are especially vulnerable to herbivory and play a crucial role in shaping forest structure (Barton and Hanley, 2013; Haase et al., 2015; Schuldt et al., 2015; Grossman et al., 2019). Some studies have reported lower herbivore damage in more diverse young plantations (Damien et al., 2016; Muiruri and Koricheva, 2017; Rosado-Sánchez et al., 2018), whereas others report increased damage or no effect (Haase et al., 2015; Schuldt et al., 2015; Zhang et al., 2017; Wang et al., 2019). In addition, many studies have reported that as trees grow larger, the negative effects of herbivory become more pronounced (Zvereva et al., 2012; Schuldt et al., 2015). However, although constant levels of herbivory can impact plant growth, the interactive effect of plant richness and herbivory on tree growth remains poorly understood (Zvereva et al., 2012).

In this study, our overall aim is to test whether biodiversity gives trees associational resistance or associational susceptibility to herbivores during the early stages of forest development. For this purpose, we tested three hypotheses: (1) herbivore damage in the early stages of forest development is mediated by tree species identity; (2) tree species diversity decreases herbivore damage; and (3) low neighboring tree diversity leads to high herbivore pressure, which inhibits tree growth. We used the Competition and Diversity Experiment-Biodiversity and Ecosystem Functioning (CADE-BEF) framework to test these hypotheses in southern China, as part of the global TreeDivNet network (https://treedivnet.ugent.be/experiments.html) (Shen et al., 2020, 2021).

2. Materials and methods 2.1. Study site

The CADE-BEF experiment (Fig. S1) was established in an abandoned field in January 2018 (23°30' N, 111°49' E) in Heerkou Town, near the Heishiding Forest Nature Reserve, Guangdong Province, China. The area experiences a subtropical humid monsoon climate, characterized by an average annual precipitation of 1740 mm and a mean annual temperature of 19.6 ℃ (Li et al., 2021). The wet season spans from April to September, while the dry season lasts from October to March. Monthly mean temperature ranges from 10.6 ℃ in January to 28.4 ℃ in July (Li et al., 2021; Luo et al., 2021). The experimental plantation is dominated by herbs and shrubs: Paspalum scrobiculatum, Sacciolepis indica, Digitaria radicosa, Echinochloa colona, Ageratum conyzoides, Phyllanthus urinaria, Lindernia crustacea, etc.

2.2. Experimental design

Eight native woody species were used in the CADE-BEF experiment, including Erythrophleum fordii (Fabaceae), Pinus massoniana (Pinaceae), Castanopsis fissa (Fagaceae), Castanopsis carlesii (Fagaceae), Schima superba (Theaceae), Elaeocarpus sylvestris (Elaeocarpaceae), Ilex rotunda (Aquifoliaceae) and Cinnamomum camphora (Lauraceae). All eight species are evergreen and possess high economic value. In addition, these species represent distinct functional groups with divergent ecological strategies. Specifically, P. massoniana and C. camphora are light-demanding species, whereas E. fordii, C. fissa, S. superba, C. carlesii, E. sylvestris, and I. rotunda are shade-tolerant species (Shen et al., 2021).

Tree saplings (1–2 years old) of selected species were planted in the CADE-BEF experiment from April 20 to May 10, 2018. Specifically, a total of 20,480 tree saplings were planted in eight blocks (32 m × 20 m). Each block had 40 plots (4 m × 4 m) (Fig. S2). Each plot contained 8 rows of 8 saplings planted 0.5 m apart (Fig. 1). Each plot within a block was randomly assigned one of 39 community compositions at planting: 8 monocultures, 18 two-species polycultures, 12 four-species polycultures, and 1 eight-species polyculture (Fig. 1). Monocultures, bicultures, and four-species polycultures were each replicated eight times (once per block), while the eight-species polyculture was replicated 16 times (Fig. S2). The individual positions within each plot and the plot positions within each block were randomized completely. Herbs and shrubs were removed biannually, during which time all upcoming vegetation between the planted saplings was cleaned (Shen et al., 2021).

2.3. Herbivory assessment

Herbivory assessments were conducted in August 2019 using standardized visual estimate methods, following established protocols widely applied in biodiversity–herbivory studies (Castagneyrol et al., 2013; Haase et al., 2015). In each plot, the central 36 tree individuals were monitored for herbivore damage, while the outermost row was excluded to reduce the strength of edge effects. Four saplings of each species were randomly selected within each plot (Fig. 1c). To ensure consistency in neighborhood diversity calculations, we excluded focal trees and neighbors that had died prior to the herbivory assessment. For broad-leaved saplings, two branches were randomly selected from both the upper and bottom parts of each sapling. Six leaves were sampled from each selected branch: three from the tip and three from the base. This sampling strategy ensured the inclusion of both young and old leaves in the assessment. In total, 24 leaves per sapling were used. For young pines, herbivory was assessed by inspecting all shoots within each yearly needle cohort and branch, with four branches selected per sapling (see broad-leaved saplings) (Haase et al., 2015).

We grouped leaf damage from two herbivore guilds (leaf chewers and skeletonizers) under chewing herbivory, as damage from leaf rollers was infrequent. Six herbivore damage percentage classes (0%, 1–5%, 6–25%, 26–50%, 51–75% and > 75%) were used to estimate leaf area removed by chewing herbivores for each leaf. Herbivore damage was averaged by each surveyed sapling and the median value of each class was used in statistical analyses (Castagneyrol et al., 2013; Schuldt et al., 2015). To improve the accuracy of the visual estimates, leaf damage assessment was conducted by the same individual with no less than 15 s per leaf used throughout the experimental investigation (Johnson et al., 2016).

2.4. Tree growth measurements

Sapling height and ground diameter were measured in November 2018 and November 2019. Height was recorded using a measuring pole, extending from the stem base to the apical meristem. Ground diameter, measured with a caliper to the nearest millimeter, was taken 5 cm above the ground. To ensure consistency, the measurement position was marked with white paint. The annual relative growth rate (RGR) of each sapling was calculated using the natural logarithm of the ratio of the 2019 size to the 2018 size: ln (size in year 2019)/(size in year 2018) (Paine et al., 2012; Schuldt et al., 2015).

2.5. Neighborhood diversity and tree apparency

As the saplings were still in the early stages of growth, the biotic influence from neighboring trees—such as shading, resource competition, and volatile cues—were expected to be strongest at the immediate scale. Therefore, we defined local neighborhood diversity based on the eight nearest neighbors of each focal tree (Skarbek et al., 2020): four situated at the main compass directions, 1.29 m away, and four positioned diagonally at a distance of 1.82 m (Fig. 1c). The neighborhood diversity (ND) of each sapling was assessed using the exponent of the Shannon diversity index, which accounts for the effective number of species weighted by their relative abundance (Shannon, 1948):

N D=-\sum\limits_{i=1}^N P_i\ ln\ P_i

P_i is the proportion of the number of individuals of each species out of the overall number of neighbor saplings, and N is the number of species.

Previous studies have suggested that the perception of the focal tree by herbivores will vary with the height of the focal tree and its neighbors (Castagneyrol et al., 2013). The manipulation of tree species diversity has implications for the vertical structure of plant communities. For example, in our experiments, Pinus massoniana, Elaeocarpus sylvestris and Ilex rotunda grow faster than Castanopsis carlesii, Schima superba and C. fissa (Shen et al., 2021). Studies have also shown that increased tree species diversity leads to greater variation in tree heights across experimental plots, which has a significant effect on the herbivore damage of the focal tree (Endara and Coley, 2011; Castagneyrol et al., 2013). The tree apparency (TA, one particular metric of plant apparency) is defined as the mean difference in height between a focal tree and its neighbors, calculated as follows (Castagneyrol et al., 2013; Grossman et al., 2019):

T A=\frac{1}{n} \times \sum\limits_{i=1}^n \frac{H F-H N_i}{d F, N_i}

HF is the total height of focal trees at the end of the growing season, HN is the total height of neighbors, dF, N_i is the distance between the focal tree and its neighbors, and n is the number of neighbors. TA > 0 indicates that the focal tree is taller (and therefore more apparent) than its neighbors, while TA < 0 indicates that the focal tree is less apparent.

2.6. Statistical analyses

The impact of sapling species identity on herbivore damage was analyzed using a one-way ANOVA followed by a Tukey post-hoc test. To assess the effects of tree neighborhood diversity and tree apparency on herbivore damage and tree growth, linear mixed-effect models were fitted using the lmer function from the "lme4" package (Bates et al., 2015), and p-values were calculated using the "lmerTest" package (Kuznetsova et al., 2017). Herbivore damage, relative height growth rate and relative ground diameter growth were used as response variables, respectively. In the herbivore damage model (for all species and each species), focal tree height, tree apparency, neighborhood diversity and two-way interactions between them were included as fixed explanatory variables, species identity and plots were treated as additive random factors in the model for all species, whereas only plots were treated as additive random factors in the models for individual species. All herbivore damage as the response variable were log-transformed for normality (Schuldt et al., 2015). All explanatory variables were standardized to have mean of zero and unit standard deviation. In the relative height and ground growth rate models, sapling initial height and ground diameter (measured in 2018), herbivore damage, and the interactions between them were included as fixed explanatory variables, and random effects were consistent with the herbivore damage model. The tested data were all normally distributed and no violations of homogeneity and normality of variance were found in the three models (Fig. S3). Models were compared using Akaike's Information Criterion (AIC) to identify the best-fit models (Zuur et al., 2009) (Table S1). All statistical analyses were conducted using R v.4.1.3 (R Core Team, 2022).

3. Results 3.1. Herbivory response to species identity, neighborhood diversity, tree height and apparency

The mean herbivore damage by leaf chewers across species was 2.6% ± 0.1 (standard error, SE), and the mean herbivore damage levels per species ranged from 0.1% ± 0.1–6.8% ± 0.3. Herbivore damage levels were highest in Castanopsis fissa, C. carlesii and Elaeocarpus sylvestris (6.8% ± 0.3, 4.8% ± 0.2 and 4.1% ± 0.2, respectively; Fig. 2). Herbivore damage levels were 'medium' for the other four species (Cinnamomum camphora, Schima superba, Erythrophleum fordii and Ilex rotunda) (3.2% ± 0.2, 2.2% ± 0.1, 1.1% ± 0.1 and 1.1% ± 0.1, respectively). The lowest mean herbivore damage levels were found on Pinus massoniana (0.1% ± 0.1) (Fig. 2).

For all species, species identity affected the mean herbivore damage by leaf chewers (P < 0.001). Fixed factors explained 5% of the total variance in the best-fit model of herbivore damage (Table 1). Herbivore damage was positively correlated with neighborhood diversity and tree apparency (P < 0.05, Fig. 3). Furthermore, focal tree height and tree apparency were also correlated. Specifically, herbivore damage generally decreased with focal tree height; however, decreases in herbivore damage were smaller as tree apparency increased.

The effects of focal tree height, neighborhood diversity, tree apparency and the interactions between them on the herbivore damage varied among species. For Erythrophleum fordii, herbivore damage was positively correlated with neighborhood diversity (P = 0.006) and its interactions with tree height (P = 0.032) (Table S2). Herbivore damage was positively correlated with tree height in Schima superba (P < 0.05) and in Cinnamomum camphora saplings (P = 0.003, Table S2). In contrast, herbivore damage was negatively correlated with tree height in Castanopsis carlesii and in Elaeocarpus sylvestris. Herbivore damage was also influenced by tree apparency in C. carlesii and in E. sylvestris (P < 0.05, Table S2).

3.2. Tree growth response to herbivory

The RGRs of sapling height across species were 0.53 ± 0.01SE, and the RGRs of sapling ground diameter across species were 0.58 ± 0.01SE. RGRs for both ground diameter and height were highest for Pinus massoniana, Elaeocarpus sylvestris and Ilex rotunda. RGRs were 'medium' for the other four species—Cinnamomum camphora, Castanopsis carlesii, Schima superba, and Castanopsis fissa (Fig. S4). RGRs were lowest for Erythrophleum fordii.

Relative growth rates of height and ground diameter were negatively correlated with initial sapling size. In addition, higher levels of herbivore damage increased the intensity of this negative effect (Fig. 4 and Table S3).

4. Discussion 4.1. Effects of species identity on herbivore damage

Herbivores are greatly influenced by their own nutritional needs and the suitability of host plants (Mattson, 1980). For example, studies have shown that herbivory levels can be affected by leaf morphological characteristics (e.g., SLA, leaf toughness) (Poorter et al., 2004; Schuldt et al., 2012) and that herbivory increases as leaf toughness and C: N decrease (Poorter et al., 2004; Salgado-Luarte and Gianoli, 2012). Our findings are consistent with these and other findings from biodiversity ecosystem forest experiments (Sobek et al., 2009; Setiawan et al., 2014; Grossman et al., 2019). We found that herbivore damage is significantly influenced by species identity, neighborhood diversity, and tree apparency. In addition, we found that herbivore damage affects tree growth.

Our finding that herbivore damage is significantly influenced by species identity supports our hypothesis that plant defenses and nutritional value mediate herbivory. At our site, herbivore damage levels were highest for Castanopsis fissa, C. carlesii and Elaeocarpus sylvestris. This level of herbivore damage may be due to large leaf size. Herbivore damage was also high for Erythrophleum fordii (nitrogen-fixing tree), which may be explained by their high nitrogen tissue concentrations (of nutritional value for herbivores) (Gubsch et al., 2011; Loranger et al., 2012; Luo et al., 2016; Meyer et al., 2017). Herbivore damage levels were the lowest for Pinus massoniana during the sapling phase, possibly because its neighbors are all taller evergreen broad-leaved trees that block the vision of herbivores and reduce feeding on P. massoniana (Dulaurent et al., 2012). An alternative explanation for this finding is that needle abscission may have contributed to lower than actual herbivory estimates for P. massoniana (Larsson and Tenow, 1980). However, conifers are less likely to attract herbivores, which may involve changes in volatiles emitted by the host trees (Vehviläinen et al., 2007; Sobek et al., 2009).

4.2. Effects of neighborhood diversity on herbivore damage

Previous studies have indicated that trees show less herbivore damage in more diverse forest ecosystems (Setiawan et al., 2014; Damien et al., 2016; Guyot et al., 2016). We found, however, that herbivore damage on the focal tree (e.g., Erythrophleum fordii) increased with neighborhood diversity. This finding is consistent with other biodiversity and ecosystem functioning tree planting experiments (Haase et al., 2015; Schuldt et al., 2015; Zhang et al., 2017). One explanation for these findings is that high neighbor diversity increases herbivore abundance and species richness, which could lead to more severe leaf damage by herbivores (Scherber et al., 2010; Brezzi et al., 2017). We also found weaker associations between herbivory and neighborhood diversity for other focal tree species. This may be because the focal tree and its neighbors were too small and distant at this early stage, which may have limited the interactions between them (Plath et al., 2011).

Previous research has suggested that associational susceptibility in forests is mostly observed in response to damage caused by generalist herbivores (Schuldt et al., 2010; Castagneyrol et al., 2014). Moreover, most leaf-chewing species are generalist herbivores (Giffard et al., 2012). Therefore, we hypothesize that generalist herbivores potentially contributed most to increased total herbivore damage in our early-successional ecosystem (Siemann et al., 1999; Schuldt et al., 2015). We will continue monitoring the herbivore community structure and collecting data in the future to further investigate this issue.

Our finding that herbivore damage increases with neighborhood diversity may be related to the spatial scale of our study. Previous studies have demonstrated that the impact of crop species richness on herbivore damage diminishes with increasing plot size (Bommarco and Banks, 2003). Other studies have reported that herbivore damage is more accurately predicted by community structure and diversity at smaller spatial scales than at larger ones (Grossman et al., 2019). Furthermore, in our current study, all tree diversity metrics were calculated based on individual counts, which may not adequately reflect the ecological influence of neighboring trees. Given the high asymmetry in tree size across different species—even during early plantation stages— it may be more ecologically meaningful to incorporate diversity metrics weighted by tree biomass or size (e.g., DBH). Larger individuals are likely to exert disproportionate effects on ecological processes such as light availability, volatile chemical emissions, and the deterrence or attraction of herbivores. Overall, future research should also consider multiple spatial scales, incorporate measures of tree biomass or size, and explore broader neighborhood effects to more comprehensively understand the relationship between tree diversity and herbivory.

4.3. Effects of tree apparency on herbivore damage

Plant apparency refers to the likelihood of a plant being detected by herbivores (Feeny, 1970; Endara and Coley, 2011). Accordingly, taller trees, which are more apparent than neighboring plants, may experience higher levels of herbivore damage (Castagneyrol et al., 2013; Guyot et al., 2015; Damien et al., 2016). Our study indicates that focal tree apparency (e.g., Castanopsis carlesii and Elaeocarpus sylvestris) is positively correlated with herbivore damage. This finding is consistent with previous studies that reported plant height apparency is positively correlated with leaf removal in certain species in the early forest succession stage in North America (Grossman et al., 2019).

Alternative explanations for our finding include the appropriate landing hypothesis, which proposes that herbivores' host selection may be influenced by plant odors: herbivores detect the plant's odors, then visually synthesize whether the plant's surface features are suitable for them, and finally decide whether to stay or leave. Consequently, faster-growing and taller saplings with similar initial size may be easier for chewing herbivores to find them (Lawton, 1983; Herms and Mattson, 1992; Finch and Collier, 2000). According to the growth-differentiation balance hypothesis, host plants cannot simultaneously allocate resources to both growth and defense (Herms and Mattson, 1992). Consequently, faster-growing plants may exhibit weaker defenses leading to damage by herbivores (Cornelissen et al., 2008). Additionally, taller saplings can provide more and larger palatable young leaves, which are more conducive to herbivore feeding (Lawton, 1983).

4.4. Effects of herbivore damage on tree growth

The RGRs for height and ground diameter in most tree species were correlated with their initial size and the interactions with herbivore damage. One explanation for this might be that smaller sapling species, starting from a smaller initial size, spend more time in the exponential growth phase, resulting in a higher average RGR (Turnbull et al., 2008). Many studies have reported that as trees grow larger, the negative effects of herbivory become more pronounced (Schuldt et al., 2015). We also found that herbivory influenced tree growth. This is also consistent with previous studies that showed long-term levels of herbivory (i.e., 4% and 8%) strongly affect tree growth and might influence tree performance (Zvereva et al., 2012). As the diversity of plant species escalates, the extent of herbivore damage also increases, potentially influencing the composition of leaf-associated fungal pathogens (Stout et al., 2006). This interaction may exacerbate the detrimental impacts of herbivory on plant growth. Furthermore, the roles of compensatory growth mechanisms and microbial interactions as potential modulators of growth warrant further investigation.

5. Conclusions

Our study shows that herbivore damage in young subtropical plantations is not only governed by tree diversity, but also by a three-way interplay of (ⅰ) the identity of focal tree species, (ⅱ) neighborhood diversity, and (ⅲ) focal tree apparency. This finding extends the associational-resistance/susceptibility theory by showing that apparency is a trait-based mechanism that can reverse the effects of diversity. Importantly, we show that early-stage damage already constrains growth in several species, highlighting a previously underappreciated pathway through which diversity can shape long-term stand dynamics. Nonetheless, we acknowledge that this connection was not consistently evident across all eight focal species. This highlights the need for long-term experiments to better disentangle the mechanisms underlying species-specific variation. Furthermore, future research should explicitly examine how changes in herbivore community structure interact with plant functional traits and multiple dimensions of plant diversity (including taxonomic, phylogenetic, and functional diversity) to shape herbivory patterns. The diversity-herbivore damage relationship warrants further investigation across multiple spatial scales and in several under-researched climatic zones.

Acknowledgments

We are grateful to the MAPPING team for coordinating the fieldwork necessary to establish this experiment. This work was financially supported by the National Natural Science Foundation of China (32330064), the Natural Science Foundation of Guangdong Province (2025A1515012138), the Scientific Research Foundation for High-level Talents of Anhui University of Science and Technology (2022yjrc07), the Open Project Program Foundation of Engineering Research Center of Underground Mine Construction, Ministry of Education (Anhui University of Science and Technology) (JYBGCZX2023104), the Anhui Key Laboratory of Mining Construction Engineering (Anhui University of Science and Technology), and the Opening Foundation of State Key Laboratory of Biocontrol (Sun Yat-sen University) (2024SKLBC-KF01).

CRediT authorship contribution statement

Zhi-Qiang Shen: Conceptualization, Methodology, Investigation, Writing – original draft, Funding acquisition. Xian-Hui Zhu: Investigation. Ming-Qiang Wang: Writing – review & editing. Ming Ni: Writing – original draft. Wen-Da Cheng: Writing – review & editing. Wei Lin: Data curation. Cheng-Jin Chu: Data curation, Writing – review & editing. You-Shi Wang: Supervision, Conceptualization, Writing – review & editing, Funding acquisition.

Data availability

The data presented in this study are included within the article. Additional datasets used for visualizations are available from the corresponding author upon reasonable request.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2025.09.008.