Insect community patterns in space: Statistical tools, simulations, and empirical data from southern Ontario coldwater springs

The purpose of this Ph.D. project was to explore spatial patterns of the species composition of insect communities theoretically using simulations as well as empirically based on an extensive field study. Additionally, the performance of current statistical tools for the analysis of multivariate spatially structured data is reviewed and suggestions for their improvement are made. The thesis is presented in three parts:;In part 1, empirical data from a 3-year field study of the insect communities in 30 coldwater springs in Southern Ontario, Canada, are presented. Using bi-monthly emergence sampling, a total of 86 insect species was collected from 9 taxonomic groups. A large fraction (55%) of these species had a low frequency and/or low abundances, and the emergence record of any given study year contained on average only about 50% of the total number of species recorded, possibly suggesting a high turnover in the species composition of the local insect communities. Canonical Correspondence Analysis (CCA) indicated that among an extensive set of environmental factors water temperature and discharge were most strongly correlated with community similarity. At the same time, the data showed a 'neighborhood effect' in that spatially adjacent sites frequently had highly similar insect communities even though sites had been selected as to maximize small-scale environmental heterogeneity. This suggests that other factors, such as meta-community dynamics, might play a decisive role in shaping the species composition in coldwater spring habitats.;Part II describes the architecture and implementation of the inter-site simulator, a Discrete Event Simulation environment for the simulation of spatially realistic population and community dynamics written in Python, a high-level, object-oriented scripting language. A practical application of the simulator to analyzing metapopulation persistence in patchy irregular landscapes is presented and the benefits of using Python as the computer language for implementing a simulation environment are discussed.;Part III is concerned with the description and refinement of statistical methods for the analysis of patterns of species distribution in space. First, PyDAS, the Python Data Analysis Servant, is introduced, a computer package that provides an extensible environment for the statistical analysis of spatially structured multivariate data. Then, two groups of methods commonly used for analyzing multivariate spatial data, Mantel-type permutation tests and spatially constrained ordinations, are introduced in detail and then evaluated using artificial data generated with the inter-site simulator presented in part II as well as the empirical data presented in part I. Several suggestions for improvements for both methods are made.;Finally, the effects of range restrictions and of long-term meta-community dynamics on local community similarity are explored in and Epilog, again using data generated with the inter-site simulator.