Analytical developments in the use of resemblance measures in community ecology and applications to boreal forest Carabidae.

摘要

Understanding the factors influencing the distribution of species is one of the main goals of ecology. This thesis presents three contributions to better and more efficiently understand the factors defining the composition of ecological communities. First, I studied the impact of anthropogenic disturbances, habitat heterogeneity, and spatial autocorrelation on Carabidae in a mature boreal forest. I showed that carabids were influenced mainly by forest floor cover, soil drainage, and tree composition. Moderate levels of anthropogenic disturbance only mildly influenced the spatial distribution of the carabid assemblages. I concluded that, carabid diversity would be best conserved in boreal forests if a network of large forest patches were left after harvest. Second, I considered the difficulty of analysing multivariate data, the main challenge in analysing species communities. Canonical redundancy analysis (RDA) is a flexible approach to relating a species community to environmental constraints. Although flexibility flows from the fact that any resemblance measure can be used within this framework, there is little guidance for how to select from the large number of existing resemblance measures. Using communities simulated from 25 different species abundance distributions (SAD), I compared results from 16 different resemblance measures within the RDA framework. The results showed that, independent of SADs, all resemblance measures gave equivalent results whether the communities were recorded as abundance or presence-absence data. In light of these results, I proposed a new canonical ordination to make a consensus of RDAs across resemblance measures. In my simulations presence-absence data were directly derived from abundance data, and so I also evaluated if the information in presence-absence and abundance data gave equivalent result. I found that the data formats may be complementary. Lastly, in ecological applications, either abundance/cover or presence-absence data are collected when species communities are sampled. With the help of resemblance measures, I propose a new way to survey ecological communities that is intermediate between presence-absence and abundance data. This approach is more cost-effective than counting abundance yet more informative than recording presence-absence. Overall, this thesis contributes to understanding spatial distribution of carabids in boreal forests and provides new methods to analyse multivariate ecological data.