Adaptation and extinction in experimentally fragmented landscapes

Competition and disturbance are potent ecological forces that shape evolutionary trajectories. These forces typically work in opposition: when disturbance is infrequent, densities are high and competition is intense. In contrast, frequent disturbance creates a low-density environment in which competition is weak and good dispersal essential. We exploited recent advances in genomic research to quantify the response to selection by these powerful ecological forces at the phenotypic and molecular genetic level in experimental landscapes. We grew the annual plant Arabidopsis thaliana in discrete patches embedded in a hostile matrix and varied the number and size of patches and the intensity of disturbance, by creating both static and dynamic landscapes. In static landscapes all patches were undisturbed, whereas in dynamic landscapes all patches were destroyed in each generation, forcing seeds to disperse to new locations. We measured the resulting changes in phenotypic, genetic, and genotypic diversity after five generations of selection. Simulations revealed that the observed loss of genetic diversity dwarfed that expected under drift, with dramatic diversity loss, particularly from dynamic landscapes. In line with ecological theory, static landscapes favored good competitors; however, competitive ability was linked to growth rate and not, as expected, to seed mass. In dynamic landscapes, there was strong selection for increased dispersal ability in the form of increased inflorescence height and reduced seed mass. The most competitive genotypes were almost eliminated from highly disturbed landscapes, raising concern over the impact of increased levels of human-induced disturbance in natural landscapes.