2.3) Data retrieval
To characterise each of the 139 relevant studies, we compiled information for six main categories: publication data, geographic and land-use data, herbivore data, plant data, trait data and sampling design (Figure 1, Appendix A, B). All data was compiled using information provided in the papers, except for climate information which was obtained by overlaying study locations with ESRI World Terrestrial Ecosystems, Temperature Regime data (Esri, USGS, TNC) (Sayre et al., 2020). Land use at the study area were extracted from information provided within the publications, although collated into four categories: pastoral, natural area/minimal use, protected area and research facility. These features were used as descriptive variables to quantitatively summarise plant trait-herbivore research in grasslands and to examine trends in a) the effect of vertebrate and invertebrate herbivores on plant traits and b) the response of vertebrate and invertebrate herbivore forage selection to plant traits. In scenario a) plant traits are the response, while in scenario b) plant traits are the effect and herbivore forage selection is the response. As some studies included multiple herbivory treatments or multiple sites, there were a total of 172 records from which we extracted trait and herbivore responses.
Where trait information had been collected across multiple sites or species, all trait responses are recorded. Plant traits were grouped a priori into broad categories; morphological, biochemical, physiological and phenological. A list of all traits and the broad category they fall into can be found in Appendix C. The response of plant traits to herbivory or the response of herbivore forage selection to the plant traits was extracted from each article and recorded within the ordinal scale; negative, positive and not significant. For example, a study which found leaf nitrogen to be lower in grazed plants compared to ungrazed plants, would be given a negative score. Alternatively, if herbivore selection increased with leaf nitrogen, this would be given a positive score. For traits with more than 5 responses recorded across all studies, we calculated the proportion of trait responses which were negative, positive or not significant.
As our study is focussed on reviewing and examining coarse differences between vertebrate and invertebrate focussed herbivore research, we did not to collect information on effect size. We do however present the proportion and total number of studies which found positive, negative or non-significant results. As such this study only accounts for statistically significant results and not overall practical significance.