Delineation of Ecological Boundaries: Comparison of Approaches and Significance Tests

In this study, quantitative spatial methods such as cluster analysis with spatial constraints and edge detection algorithms are compared with respect to their abilities to delimit boundaries from two-dimensional sampled data. While cluster analysis with spatial constraints forms clusters among neighboring sites that are similar, edge detection algorithms delimit areas of high rate of change. To determine whether the delineated boundaries could have arisen by chance, boundary and superfluity statistics are used and their statistical significance is assessed by permutation tests. Advantages and limits of each approach are illustrated using data sets of tree densities from a second growth stand of northern deciduous forest in southern Quebec, Canada. It is found (1) that applying jointly these two types of approaches provides complementary information about the location and the intensity of the delineated boundaries; and (2) that the boundary and superfluity statistics are useful for assessing the statistical significance of boundaries