The predictability of phytophagous insect communities: host specialists as habitat specialists.

The difficulties specialized phytophagous insects face in finding habitats with an appropriate host should constrain their dispersal. Within the concept of metacommunities, this leads to the prediction that host-plant specialists should sort into local assemblages according to the local environmental conditions, i.e. habitat conditions, whereas assemblages of host-plant generalists should depend also on regional processes. Our study aimed at ranking the importance of local environmental factors and species composition of the vegetation for predicting the species composition of phytophagous moth assemblages with either a narrow or a broad host range. Our database consists of 351,506 specimens representing 820 species of nocturnal Macrolepidoptera sampled between 1980 and 2006 using light traps in 96 strict forest reserves in southern Germany. Species were grouped as specialists or generalists according to the food plants of the larvae; specialists use host plants belonging to one genus. We used predictive canonical correspondence and co-correspondence analyses to rank the importance of local environmental factors, the species composition of the vegetation and the role of host plants for predicting the species composition of host-plant specialists and generalists. The cross-validatory fit for predicting the species composition of phytophagous moths was higher for host-plant specialists than for host-plant generalists using environmental factors as well as the composition of the vegetation. As expected for host-plant specialists, the species composition of the vegetation was a better predictor of the composition of these assemblages than the environmental variables. But surprisingly, this difference for specialized insects was not due to the occurrence of their host plants. Overall, our study supports the idea that owing to evolutionary constraints in finding a host, host-plant specialists and host-plant generalists follow two different models of metacommunities: the species-sorting and the mass-effect model.