Data from: Edge-mediated compositional and functional decay of tree assemblages in Amazonian forest islands after 26 years of isolation

1.Islands formed upstream of mega hydroelectric dams are excellent experimental landscapes to assess the impacts of habitat fragmentation on biodiversity. We examined the effects of plot-, patch- and landscape-scale variables on the patterns of floristic diversity across 34 forest islands that had experienced 26 years of isolation since the creation of the 4,437 km2 Balbina Hydroelectric Reservoir of central Brazilian Amazonia. In addition, three undisturbed continuous forest sites in neighbouring mainland areas were also sampled across a comparable elevational gradient. 2.We identified all live trees ≥10 cm DBH at species level within a total of 87 quarter-hectare forest plots and conducted a comprehensive compilation of functional attributes of each tree species. We then examined species-area relationships (SARs) and the additional effects of patch and landscape scale metrics on patterns of tree assemblage heterogeneity, both in terms of taxonomic and functional diversity. 3.Despite a clearly positive SAR, edge-mediated forest disturbance was the single most important driver of species composition and abundance within islands. Our results suggest that non-random floristic transitions within island plots followed a predictable pattern, with different life-history traits either penalizing or rewarding local persistence of different functional groups. Distance to edges mediated the probability of tree mortality induced by windfalls and episodic surface fires, clearly resulting in faster species turnover and unidirectional changes in guild structure within small islands where light-wooded fast-growing pioneers largely replaced heavy-wooded species of the old-growth flora. 3.Synthesis ─ Following a simultaneous 26-year post-isolation history, we disentangle the effects of habitat loss and insularization on tree assemblages within a large set of Amazonian 'true' forest islands, of variable sizes, sharing a uniform open-water matrix. Area effects are expressed via a response to edge effects, with trees in smaller islands being more vulnerable to edge-related surface fires and windthrows. Additionally, forest edge effects can be a powerful driver of non-random floristic transitions across islands within the Balbina archipelago via a process of rapid pioneer proliferation, drastically affecting both the taxonomic and functional composition of insular tree communities. Finally, our results indicate that detrimental effects of forest fragmentation induced by hydroelectric dams are considerably stronger than those of forest patches embedded within a terrestrial vegetation matrix.tree species dataData were collected in 2012 in 87 one-quarter hectare forest plots within the Balbina Hydroeletric Dam. The first column indicates the scientific name of each species recorded on at least one forest plot. The first line indicates forest plot identification. See the 'metrics data' for information of the island name according to each forest plot identification. The matrix contains the presence/absence data of each tree species within each of those 87 forest plots.Data_sps.xlsmetrics dataPlot, patch and landscape metrics of each surveyed forest plot within the Balbina Hydroeletric Dam. Abbreviated headlines are as follows: 'local' = island name; 'plot_ID' = plot number identification; 'area' = island area in hectares; 'slope' = the difference between maximum and minimum plot elevations based on the Shuttle Radar Topography Mission; 'windstorm' = the angular difference (0 - 90°) between the main axis of each rectangular plot and the median angular direction of prevailing strong winds recorded prior to and during convective windstorms; 'edge_mean' = the mean linear distance between the five midpoints of every 50-m section of each plot and the nearest points along the island perimeter; 'fire' = the understorey burn or fire severity measured according to a composite ordinal score (0–3); 'prox' = modified McGarigal et al. (2012) proximity index by considering both the aggregate area and distance to any land mass within 500 m of each island; 'dist_mainland' = the shortest linear distance from each island to the nearest mainlandData_metrics.xl