Habitat fragmentation is widely considered the most important near-term threat to biodiversity, and both ecological theory and experiments have played integral roles in understanding its consequences. Although past experiments and theory have laid a valuable foundation for understanding habitat fragmentation, several shortcomings have been realized. Two major shortcomings are that experiments are often manipulated at the patch scale (e.g., patch size, isolation) and that much work has neglected the importance of the “matrix” in which fragments are embedded. Yet we know that habitat loss and fragmentation is a landscape-scale process and that for many species, matrix effects can better explain species distribution than long-held tenets of fragment area and isolation. We conducted a landscape-scale experiment on a native insect herbivore (Chelinidea vittiger) that specializes on patchily distributed, prickly pear cactus (Opuntia humifusa), where both habitat loss and fragmentation were manipulated across 17 experimental plots (50 x 50m). During a pre-treatment year, we released 1700 individuals to standardize baseline population sizes and subsequently estimated population dynamics via mark recapture. During the second (treatment) year, habitat loss was manipulated (10-95% loss) and occurred either via random patch removal or aggregated removal (on the scale of observed pre-treatment dispersal distances). We also measured the structure of the surrounding matrix and estimated behavioral and demographic effects of the matrix with subsequent, individual-based experimental releases. Observed (meta)population dynamics were measured across 3-4 generations and contrasted to predictions from metapopulation theory.
Results/Conclusions
At the landscape scale, we found a threshold effect of habitat loss on the collapse of metapopulations, where habitat networks suffering >80% collapsed regardless of whether loss was random or aggregated. At lower levels of habitat destruction, aggregated loss resulted in greater population decline than random loss. Similar patterns occurred both in terms of adult abundance and the production of nymphs on cacti. Yet matrix quality explained population abundance as well or better than the whole-sale conversion of habitat. While observed metapopulation collapse could be partially predicted via metapopulation theory, the synergistic effects we observed were masked in metapopulation predictions. These results provide guidance for extending metapopulation theory to capture the synergistic effects of environmental change on the persistence of populations, highlight the need to scale such dynamics from patches to landscapes, and emphasize the critical role of the matrix on population viability.