Direct versus indirect effects of habitat fragmentation on community patterns in experimental landscapes

Habitat area and fragmentation are confounded in many ecological studies investigating fragmentation effects. We thus devised an innovative experiment founded on fractal neutral landscape models to disentangle the relative effects of habitat area and fragmentation on arthropod community patterns in red clover (Trifolium pratense). The conventional approach in experimental fragmentation studies is to adjust patch size and isolation to create different landscape patterns. We instead use fractal distributions to adjust the overall amount and fragmentation of habitat independently at the scale of the entire landscape, producing different patch properties. Although habitat area ultimately had a greater effect on arthropod abundance and diversity in this system, we found that fragmentation had a significant effect in clover landscapes with ≤40 % habitat. Landscapes at these lower habitat levels were dominated by edge cells, which had fewer arthropods and lower richness than interior cells. Fragmentation per se did not have a direct effect on local-scale diversity, however, as demonstrated by the lack of a broader landscape effect (in terms of total habitat area and fragmentation) on arthropods within habitat cells. Fragmentation-through the creation of edge habitat-thus had a strong indirect effect on morphospecies richness and abundance at the local scale. Although it has been suggested that fragmentation should be important at low habitat levels (≤20-30 %), we show that fragmentation per se is significant only at intermediate (40 %) levels of habitat, where edge effects were neither too great (as at lower levels of habitat) nor too weak (as at higher levels of habitat).