Plant chemistry mediates interactions between plants and their environment. While intraspecific chemodiversity at the individual plant level is well-studied, the effects of chemodiversity at the plant community level on arthropod interactions need more attention. We conducted a field experiment to test how intraspecific chemodiversity affects plant-arthropod interactions. We manipulated plots of Tanacetum vulgare L., varying in chemotype richness and composition, and monitored four arthropod functional groups (herbivores, flower visitors, predators, and ants) over three seasons. We hypothesized that higher plot-level chemotype richness would enhance occurrence across all studied arthropod groups but have functional group-specific effects on abundance, resulting in reduced herbivore and ant abundance and increased flower visitor and predator abundance with increased chemotype richness. Using mixed models, we found that increasing plot-level chemotype richness had a limited effect on most arthropod group occurrences but led to significant changes in abundance. Increasing chemotype richness decreased herbivore abundance, while it increased flower visitor abundance. Predatory arthropods and ants remained largely unaffected. Furthermore, we found that the presence of specific plant chemotypes within a plot had variable effects for herbivores across years, as some chemotypes initially promoted higher abundances but were associated with lower abundances in the later years. We also found that the presence of specific chemotypes had a positive effect on the abundances of various flower visitor taxa. Predators and ants, on the other hand, showed weaker and more variable responses to specific chemotypes. Specifically, higher chemotype richness reduced the abundance of herbivores such as Uroleucon tanaceti (2022 and 2023) and Brachycaudus cardui (2023), while increasing visitation by flower visitors. These results highlight the role of plant chemodiversity in shaping insect communities and influencing ecosystem dynamics.

Data collection

From 2021 to 2023, we assessed herbivores, flower visitors, predators, and ants. Each year, herbivore communities were monitored by counting individual aphids on the plants from their first appearance until nearly none were found. Although the focus was on aphids, other herbivores were scarce. Therefore, the generalization to herbivores here reflects predominantly aphids. Counts were conducted 13 times from June to August 2021, eight times from May to August 2022, and three times from June to August 2023. Each count, executed within a day between 7 h and 19 h, involved a meticulous scan of each plant, which started from the tip of each stem and proceeded to the end, examining both sides of each leaf. When multiple aphid species were present on a plant, we recorded the total number of each. In cases of high aphid densities (>100), we estimated the total by counting aphids on 1 cm of stem or 1 cm² of leaf three times in different stems and leaves for each plant, then multiplying the average by the estimated lengths of stems (in cm) or the estimated areas of leaves (in cm²) colonized by the aphids. We used plot-level occupancy and cumulative plot-level number of aphids to analyze occurrence and abundance.

Flower visitors were assessed by counting the flower heads of individual plants during the flower-visiting recording. Flower visitors were counted three times in July 2021 and once in August 2021. Surveys lasted three minutes per plot and were conducted randomly from 9h to 15h under optimal weather conditions (no gusts of wind or precipitation). Flower visitors were identified at the order level. We used presence-absence data to analyze the effects of CR and chemotype presence on occurrence. However, as CR and individual chemotypes influence reproductive traits in our system (Ojeda‐Prieto et al., 2024), flower visitor abundance per flower head (visitation rate) assessed effects on abundance.

Predatory arthropods and ants on individual plants were recorded during herbivore counting for 2021, 2022, and the day after in 2023. Predators were classified into taxa/feeding guilds (Araneae, Coleoptera, Dermaptera, Diptera, Hemiptera, and Hymenoptera), specifying aphid parasitoids and parasitized aphids (counted as mummies found). Ant presence on plants, but not abundance, was noted. Predator and ant occurrences and parasitized aphid abundance were used for analyses for the three seasons. In 2023, predator and ant abundances were also analyzed.

Statistical analyses

Statistical analyses were conducted using R version 4.1.2 and RStudio 2023.09.1 (R Core Team, 2021). All statistics were performed at the plot level. The effects of chemotype presence and CR on the occurrence of herbivores, flower visitors, predators, and ants were assessed through binomial Generalized Linear Mixed-effect Models (GLMM) using the ‘lme4’ package (Bates et al., 2015). Linear Mixed-Effect Models (LMM) and Zero-Inflated Negative Binomial Models (ZINB), using the 'glmmTMB' package (Brooks et al., 2017), modeled the effect on abundances. P-values were estimated by type II Wald-Chi-Squared tests using the Anova() function (‘car’ package; Fox & Weisberg, 2019). Data visualization used academic-licensed Biorender® and 'ggplot2’ (Wickham, 2009), ‘grid’ (R Core Team, 2021), ‘gridExtra’ (Auguie et al., 2017), ‘ggpubr’ (Kassambara, 2020), and ’DHARMa’ (Hartig, 2022) packages.

Effects of chemotype richness and chemotype presence on arthropod taxa occurrence and abundance

Binomial GLMMs assessed the effect of CR and chemotype presence on occurrence. LMMs and ZINB tested the effects on abundance. Data were analyzed at the arthropod group and lower taxonomic levels (when indicated). Data were analyzed cumulatively for each year and for each independent sampling time.

References

Auguie, B., Antonov, A., & Auguie, M. B. (2017). gridExtra: Miscellaneous functions for “grid” graphics.

Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1–48.

Brooks, M. E., Kristensen, K., Van Benthem, J., Magnusson, A., Berg, C. W., Nielsen, A., Skaug, H. J., Maechler, M., & Bolker, B. M. (2017). glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. The R Journal, 9(2), 378–400.

Fox, J., & Weisberg, S. (2019). An {R} companion to applied regression (Third Edition). Thousand Oaks CA: Sage.

Hartig, F. (2022). DHARMa: Residual diagnostics for hierarchical (multi-level / mixed) regression model. https://cran.r-hub.io/web/packages/DHARMa/DHARMa.pdf

Kassambara, A. (2020). ggpubr: “ggplot2” based publication ready plots.

Ojeda-Prieto, L., Medina-van Berkum, P., Unsicker, S. B., Heinen, R., & Weisser, W. W. (2025). Intraspecific chemical variation of Tanacetum vulgare affects plant growth and reproductive traits in field plant communities. Plant biology (Stuttgart, Germany), 27(5), 785–801. https://doi.org/10.1111/plb.13646

R Core Team. (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing.

Wickham, H. (2009). ggplot2: Elegant graphics for data analysis. Springer New York. https://doi.org/10.1007/978-0-387-98141-3