Search for: All records

Despite global biodiversity losses, trends at local and regional scales are context dependent. Recent studies have been criticized for lacking baselines preceding human impacts, and few such studies have addressed the landscape scale. Our aim was to quantify temporal trends in landscape‐scale tree diversity during an unambiguous period of massively increased anthropogenic disturbance and to test the hypothesis that land use can increase landscape‐scale diversity via increased environmental heterogeneity.

Eastern USA.

1620–2008.

Trees.

We combined data from archival land surveys and modern‐day forest inventories in the north‐eastern USA to quantify tree genus diversity at the scale of towns (“landscapes”). We modelled change in diversity over time as a function of the proportion of the landscape historically converted to agriculture, historical temperature increases and nitrogen deposition, and other abiotic and spatial variables. We also tested for scale‐dependent changes in beta diversity.

Overall, tree genus diversity (Shannon and Simpson indices) changed minimally over time on average, but the magnitude of change increased with the maximum historical percentage of the town in agriculture. Other predictor variables had minimal influence. Beta diversity increased over time for nearby pairs of towns and decreased over time for more distant towns.

Forests have regrown on much former agricultural land, and our results support the hypothesis that increased landscape‐scale environmental heterogeneity, attributable to land use, increased tree diversity. Where agricultural land use was uncommon, declines in diversity might be attributable to effects of logging and fire suppression. Even the strongest driver of biodiversity loss at local and global scales (human land use) can lead to increases in biodiversity at the landscape scale, in addition to scale dependence of biotic differentiation versus homogenization.

Both ecological drift and environmental heterogeneity can produce high beta diversity among communities, but only the effect of drift is expected to be enhanced in communities of small size. Few studies have explicitly tested the influence of community size on patterns of beta diversity. Here we applied a series of analyses aimed at testing the influence of drift versus environmental heterogeneity on beta diversity among tree communities on islands of variable size.

Thousand Island Lake, Zhejiang Province, China.

We used data on mapped tree communities and environmental conditions for 20 small islands (<1 ha) and nine large islands (>1 ha) created via the construction of a hydroelectric dam in 1959. Beta diversity was calculated using abundance‐based multiple‐site dissimilarity based on the Bray–Curtis index. On the basis of the hypothesis of ecological drift among small islands, we tested for higher beta diversity among small than large islands using: (a) raw data (b) controlling for the number of individual sampled on a given island, and (c) controlling for the contiguous sampling area and thus for intra‐island environmental heterogeneity. We also tested the prediction that the relationship between species composition and environmental variables should be weaker on small islands using canonical correspondence analyses.

Using raw data and controlling for the number of individuals, community dissimilarity was significantly greater among small islands than among large islands. However, when controlling for contiguous sampling area this difference disappeared. Contrary to the prediction based on ecological drift, the strength of overall composition–environment relationships was not significantly weaker for small islands in any of the analyses, and environmental heterogeneity increased faster with area among small islands than among large islands.

Despite a result using raw data that was consistent with the hypothesis of ecological drift, our full set of results clearly indicated the high beta diversity among small islands was more likely due to environmental heterogeneity rather than ecological drift. This result points to a clear need to control for sampling area among habitats of different size when testing for statistical signatures of drift.