Climate, Land use and Plant Richness Differently Shape Herbivory on Major Plant Functional Groups


 Interactions between plants and herbivorous invertebrates drive the nutritional quality of resources for higher trophic levels, nutrient cycling and plant-community structure. Thereby, shifts in functional composition of plant communities particularly impact ecosystem processes. However, the current understanding of herbivory is limited concerning climate, land use and plant richness, as comparative studies of different plant functional groups are lacking. This study was conducted on 81 plots covering large climatic and land-use gradients in Bavaria, Germany. We investigated foliar invertebrate herbivory rates (proportional leaf-area loss, following ‘herbivory’) in three major plant functional groups (legumes, non-leguminous forbs, grasses). As drivers we considered multi-annual mean temperature (range: 6.5–10.0 °C), local habitat type (forest, grassland, arable field, settlement), local plant richness (species and family level, ranges: 10–50 species, 5–25 families) and landscape diversity (0.2–3-km scale). Our results largely confirm higher herbivory on legumes than on forbs and grasses. However, herbivory in forests was similar across plant functional groups since herbivory on legumes was low, e.g. lower than on legumes in grasslands. We also observed differential responses of herbivory among plant functional groups in response to plant richness (family level only), but not to landscape diversity. Temperature did not affect overall herbivory, but in grasslands higher temperature decreased herbivory on legumes and increased on forbs and grasses. We conclude that climate, habitat type and family-level plant richness likely assert different effects on herbivory among plant functional groups. This emphasises the importance of functional groups for understanding community-level herbivory and ecosystem functioning.


Introduction 40
Herbivores feeding on living plant tissue make essential contributions to ecosystem functioning 41 (Biedermann et al. 2005). When becoming prey or through defecating, herbivores provide nutrients to other 42 trophic levels and release nutrients which otherwise would be retained in plant material for a longer time. 43 Despite its key role in food webs and nutrient cycles, herbivory generally causes negligible total biomass 44 loss (Schowalter 2016). Nonetheless, herbivory affects growth, phenology, productivity and competitive 45 Herbivory pressure strongly depends on the nutritious quality and palatability of plants (Loranger et al. 56 2012; Njovu et al. 2019), which varies substantially among major plant functional groupsnamely 57 legumes, non-leguminous forbs and grasses (Scherber et al. 2006). Legumes contain more nitrogen, e.g. 58 higher crude plant protein content and lower leaf C:N ratio, than forbs and grasses (Perez Corona et al. 59 1995;Leingärtner et al. 2014), whereas lignin content is higher in grasses (Perez Corona et al. 1995). 60 Consequently, high, medium and low foliar herbivory intensities are frequently observed on legumes, non-61 leguminous forbs and grasses, respectively (Scherber et al. 2006;Leingärtner et al. 2014). Furthermore, 62 plant functional groups encompass food plants of specialist herbivores since insect herbivores are often 63 7 processes (Roscher et al. 2004), sometimes additional groups are distinguished e.g. small and tall forbs 119 (Scherber et al. 2006, Fischer et al. 2019 or C3 and C4 grasses (Siemann et al 1998). 120 Per plant functional group, three plant individuals of three plant species were randomly selected for the 121 collection of three leaves (total of 27 plant individuals and 81 leaves per plot). Plant species assessed for 122 herbivory differed among plots, since not even a single plant species occurred across all plots (Table S1), 123 but represent herbivory of the local herbaceous vegetation. Leaves from the apical, middle and basal nodes 124 of each plant were pinched off, mounted in a notebook with transparent tape, pressed and dried for later 125 assessment. Both leaf position as selection criterion and digital assessment in the lab were chosen to 126 minimize observer bias and also to include leaves of different ontogenetic stages that may be 127 disproportionately affected by herbivory (Sand-Jensen et al. 1994). Mean leaf-area loss was determined 128 using the BioLeaf app (Machado et al. 2016), which automatically transforms colour images to binary 129 images (only black and white pixels) and calculates proportional mean leaf-area loss based on white parts 130 enclosed by black leaf area. Therefore, some prior image adjustments were needed: i) Nibbled leaf margins 131 were straightened or adjusted to restore the pre-damage leaf contour with a thin black line in order to capture 132 nibbled leaf margins as missing leaf area; and ii) overlapping leaf parts were separated with a thin white 133 line connecting the white space to the surroundings of the leaf to not falsely be assigned as missing leaf 134 area by the Bioleaf app. Images were adjusted using GIMP software (The GIMP Development Team 2017). 135

Measures of plant richness 136
Vegetation surveys were conducted between May and July 2019 (seven subplots on each plot, adding up to 137 10 m 2 total sampling area per site). Recorded plant species and families were summed up per plot to achieve 138 plant richness at species and family level. Ferns, horsetails and woody plants as part of the herb layer were 139 considered for total plant richness measures but not for herbivory assessment. A list of plant species and 140 families present on plots can be found in the supplement (Table S1). 141

Measures of local and landscape-scale land use 142
Local habitat-specific similarities among plots were captured by habitat type (forest, grassland, arable field 143 and settlement). 144 At landscape-scale, we calculated landscape diversity as Shannon Index from detailed land-cover maps 145 distinguishing six land-use categories: natural/semi-natural, forest, grassland, arable field, settlement and  were available in a plot or, exceptionally, also due to processing damage. To minimize bias through single 165 herbivory events and to assure sampling of at least two different plant species per plant functional group, 9 we excluded data from all plots of which we obtained proportional mean leaf-area loss data of ˂10 leaves 167 of each plant functional group prior to herbivory analysis. Taking also seven plots into account where no 168 herbivory sampling could be realized due to time constraints, this yielded 81 plots in 40 regions. 169 Herbivory data were analysed with beta regression to cope with continuous proportional data 170 (Yellareddygari et al. 2016; Douma and Weedon 2019). In preparation of beta regression, zeros were 171 replaced through a small value (0.00001; slightly lower than the smallest value) to allow for model 172 comparison with Akaike's information criterion (AIC), which is inappropriate for scaled data (Douma and 173 Weedon 2019). Proportional mean leaf-area loss data on legumes and forbs contained only a single zero 174 and data on grasses two zeros (2.5%). 175 As candidate predictors, we included plant functional group, multi-annual mean temperature as climatic 176 variable, land use at local (habitat type) and landscape-scale (landscape diversity; seven spatial scales in 177 separate models), and local plant richness (species and family level). Predictor values were z-transformed 178 prior to analysis, while the selected best models are presented with untransformed predictor variables. Ten 179 separate models were created, each of them containing plant functional group, multi-annual mean 180 temperature, one of the four land-use and plant-richness variables (at different spatial scales, if applicable) 181 and all interactions up to the three-way interaction term. Separate models were preferred over one model 182 containing all land-use and plant-richness variables to avoid over-parameterization. 183 The model including the three-way interaction of plant functional group, multi-annual mean temperature 184 and habitat type indicated a trend in grassland, which was further explored using a data subset of grassland 185 plots. An additional model containing multi-annual mean temperature, habitat type and their interaction 186 term was fitted to the subset with the rest of the analysis approach being equal to that of the other models. 187 A nested random term for 'plot' in 'region' (three plots per region) was included to account for plant 188 functional groups on the same plots and clustering of plots, which were located in closer vicinity than other 189 plots due to the nestedness of the study design (Redlich et al. 2021). This nested random term was retained 190 throughout the model selection process (Bolker et al. 2008). 191 The majority of maximum variance inflation factors were <4, which falls below the commonly applied 192 threshold for collinearity of variance inflation factor ˃10 (Chatterjee and Price 1991). The few cases in 193 which the variance inflation factor exceeded the threshold were in models containing interaction terms with 194 habitat type. Additionally, a correlation matrix of continuous predictor variables was calculated (Table S2) 195 and continuous predictors were plotted by habitat type (Fig. S2)  Effects of plant functional group, land-use and plant-richness on herbivory 213 We observed an overall proportional leaf-area loss due to invertebrate chewing across plant functional 214 groups of 1.35 ± 0.10% (mean ± se). Foliar herbivory on legumes was on average 2.3 times higher than on 215 forbs and 5.4 times higher than on grasses (Fig. 1a). Besides, plant functional group and habitat type 216 interactively affected herbivory (Fig. 1b). Herbivory on legumes was higher than herbivory on forbs except 217 in forests, where herbivory on legumes and forbs was similar. In forests, herbivory on legumes was also 218 lower than in grassland, and intermediate in settlements and agricultural fields. Herbivory on forbs and 219 grasses was similar across all habitat types. 220 Invertebrate herbivory did not depend on plant richness at species level (Fig. 2a). However, with increasing 221 total numbers of plant families, herbivory on legumes decreased while herbivory on forbs and grasses 222 increased (Fig. 2b). At the landscape-scale, invertebrate herbivory was similar across the covered landscape-223 diversity gradient ( Fig. 2; Table S3). 224 Interactive effects of plant functional group, climate and land-use (or plant richness) predictors on herbivory 225 Mean-annual temperature did not substantially affect overall herbivory and three-way interactions of plant 226 functional group, climate and single land-use or plant richness predictors were not supported by ∆AICc and 227 parsimony (Fig. 3, Table S3). Yet in grasslands, herbivory on legumes, forbs and grasses decreased, 228 increased and slightly increased with higher multi-annual mean temperature, respectively (Fig. 3, dashed  229 line: grassland subset, Table S4). 230 231

Discussion 232
In this study, we compared herbivory among three major plant functional groups in a wide range of typical 233 habitat types in temperate regions across large climatic gradients, and also taking into account land use at 234 local and landscape scale, and local plant richness. Herbivory differed between plant functional groups and 235 local habitat types as well as with plant richness at family level, but showed no significant response to plant 236 richness at species level, landscape diversity and multi-annual mean temperature at the studied gradients, 237 except for differential temperature effects among plant functional groups in grasslands (significant effect 238 in grassland subset). 239

How do land use and plant richness affect invertebrate herbivory on plant functional groups? 240
Although herbivory assessment was conducted on open herbaceous vegetation, the surrounding local 241 habitat types included in this study differ in multiple aspects that may impact invertebrate herbivores. 242 Among habitat types, we did not observe differences in herbivory on forbs and grasses but herbivory on 243 legumes was lower in forests than in grasslands and intermediate in settlements and arable fields. Higher 244 impact of herbivores on herbs in grasslands than forestsnot differentiating plant functional groupswas 245 also reported by Maron et al (2006). Our results, however, suggest that herbivory on legumes is more 246 sensitive to local land use (habitat type) than that on forbs and grasses, which emphasises the importance 247 of distinguishing between plant functional groups. One explanation could be that legumes are more prone 248 to specialist herbivory and that herbivore communities on herbaceous-vegetation patches in forests may 249 more frequently lack these herbivore species, since forests may constitute barriers to dispersal (Schmitt et 250 al. 2000) and since small, isolated patches face higher species extinction risk at simultaneously reduced 251 recolonization rate (Rösch et al. 2013). In analogy, highest herbivory intensities on legumes in grasslands 252 may result from larger habitat patches and fewer dispersal barriers. An even simpler explanation of the 253 observed herbivory pattern on legumes provides the 'habitat amount hypothesis' put forth by Fahrig (2013). 254 This hypothesis stresses the importance of the total amount of habitat area for species richness (across a 255 broad taxonomic range, including insects). Thus, more herbaceous vegetation in grasslands than forests 256 13 may have led to higher species richness of herbivores foraging on herbaceous plants and indirectly to higher 257 herbivory rates on legumes, under the assumption that higher herbivore species richness increases the 258 likelihood that herbivore species specialized on legumes are present. 259 While herbivory on legumes differed among habitat types, we observed similar herbivory among habitat 260 types on grasses and forbs. Herbivory on grasses was commonly low, thus when small habitat-type effects 261 occurred, they may have been rendered invisible. Forbs encompass species of several plant families, which 262 increases the likelihood that some of the forb plant families perceive herbivore damage. This suggests to 263 differentiate between more groups of forbs, e.g. distinguishing herbivory on plant-family level for forbs or 264 distinguishing small and tall forbs (see also Roscher et al. 2004). Alternatively, forbs and grasses may be 265 prone to more generalist herbivores, which have the potential to maintain herbivory even at low (generalist in legumes than forbs and grasses at higher temperature. More research will be needed to provide further 306 evidence on differential rates of herbivory among plant functional groups towards higher temperature and 307 to identify the major drivers. 308

Conclusion 309
We conclude that multi-annual mean temperature, plant richness at family level and land-use at local scale 310 (habitat type)but not at landscape scale (landscape diversity)assert differential effects on herbivory 311 among plant functional groups, with legume herbivory in grasslands being most affected. Herbivory on 312 legumes was higher in grassland than forest habitats, decreased with temperature in grasslands, and 313 decreased with family-level plant richness. Further research is needed to identify the drivers behind these 314 observations, whereby functional herbivore community compositionalong with plant functional groups 315 may provide valuable insights. Our study emphasises the importance of functional groups (of both plants 316 and herbivores) for understanding the response of community-level herbivory and ecosystem functioning. 317 318

Acknowledgements 319
We are grateful to the landowners, leaseholders, municipalities and the Bavarian State Forestry, who 320 facilitated this project. Special thanks go to Jan Weber and Kerstin Hikel for numerous hours spent on 321 herbivory assessment. We thank the Bavarian Office for Surveying and Geographic  Availability of data and material: We intend that data will be available from the Dryad Digital Repository. 336 Code availability: R Code is available from the corresponding author upon reasonable request.            Mean leaf-area loss to invertebrate herbivory per plot and plant functional group (legumes: pink circles, non leguminous forbs: green triangles: grasses: blue squares) is presented across plant richness at (a) species level and (b) family level, and (c) landscape diversity at 1-km spatial scale. Solid lines present predictions of best beta mixed models. Model selection was based on ∆AICc and parsimony. Mean leaf-area loss per plot and plant functional group (legumes: pink circles, non-leguminous forbs: green triangles: grasses: blue squares) is presented per habitat type across the multi-annual mean temperature gradient. Solid lines indicate predictions of the best beta mixed model based on the complete data set. Dashed lines show predictions based on the grassland subset. Model selection was done using ∆AICc and parsimony.