Abstract
Characterising plant-herbivore interactions is important to understanding the processes that influence community structure and ecosystem functioning. Traditional methods used to identify plant-herbivore interactions are being superseded by non-destructive molecular approaches that can infer interactions with greater resolution and accuracy from environmental DNA (e.g. faeces and regurgitate). However, few studies have compared the success of using different sample types and whether they provide similar or contrasting information about species' diet. Here we compared the success of DNA amplification and host plant species identification using restriction fragment length polymorphism (RFLP) applied to faecal and regurgitate samples collected from alpine grasshoppers Paprides nitidus Hutton during a grassland community mesocosm experiment. We found that DNA amplification success was 23% and 86% higher for faecal than regurgitate samples from female and male grasshoppers, respectively. In contrast, successful host plant identification using RFLP was 9% higher for regurgitate than faecal samples. The mean number of host plant species identified per sample (1.40) did not differ between sample types or grasshopper sexes. Of the 136 paired faecal-regurgitate samples, just 41% and 74% produced exactly or partially matching host plant identifications, respectively, indicating that different sample types provided complementary information about herbivore diet. Some plant species were more likely to be identified from faecal samples than expected by chance, and we found that this identification bias skewed towards plant species with higher investment in leaf tissue. We conclude that multiple sample types may be required to fully characterise an invertebrate herbivore species' diet.