VARIATION IN LOCAL-SCALE EDGE EFFECTS - MECHANISMS AND LANDSCAPE CONTEXT

Ecological processes near habitat edges often differ from processes aw ay from edges. Yet, the generality of ''edge effects'' has been hotly debated because results vary tremendously. To understand the factors r esponsible for this variation, we described nest predation and cowbird distribution patterns in forest edge and forest core habitats on 36 r andomly selected plots in three states in the midwestern United States . We tested four hypotheses that may explain the variation and mechani sms responsible for edge effects among the 36 plots: (1) the landscape context, (2) the local predator community, (3) the local bird (host-p rey) community, and (4) the nest site microhabitat structure. We used artificial nests baited with quail and clay eggs to determine nest pre dation patterns and predators and used point count surveys to determin e cowbird and host abundance in forest edge and forest core habitats. Raccoons, opossums, canids, and birds accounted for most predation of artificial nests. Neither local host abundance nor mean nest concealme nt of artificial nests significantly influenced nest predation rates i n habitat edge or in habitat core. Nest predation was significantly gr eater in highly fragmented landscapes than in unfragmented landscapes and was significantly higher in edge habitats than in core habitats. H owever, detection of edge effects varied, depending upon landscape typ e. Higher predation rates in edge habitats were detected in highly and moderately fragmented landscapes, but not in unfragmented landscapes. Both mammalian and avian predator groups contributed to higher predat ion rates along edges in highly and moderately fragmented landscapes. Cowbird abundance was significantly related to host abundance, but the effect of hosts varied depending upon habitat type. In edge habitats, cowbird abundance was negatively associated with host abundance in al l three landscapes studied. By contrast, cowbird abundance was positiv ely associated with host abundance in core habitats. Once the effects of host abundance were removed, cowbird abundance in core habitat was greater in highly fragmented landscapes than in moderately and unfragm ented landscapes, but did not differ between the latter two. In edge h abitat, cowbird abundance did not differ between landscapes, but abund ance in edges tended to be highest in the highly fragmented landscape and lowest in the unfragmented landscape. Cowbird abundance did not va ry between edge and core habitat in any of the landscapes studied. We suggest that the first approximation to predicting the impact of agric ultural or permanently managed edges on forest songbird reproductive s uccess is to assess habitat characteristics at the landscape scale. Gi ven geographic location, local factors such as host abundance and pred ator composition should be assessed.