EDGE-DETECTION ALGORITHMS FOR 2-DIMENSIONAL ECOLOGICAL DATA

The widely advocated statement that ecotone movement may be useful in studies of the impacts of global warming implies not only that ecotones exist but that they can be delineated spatially. With sampled field data, the accuracy of the detected ecotone is related to the data type and its spatiotemporal resolution. In the present study, I introduce two edge detection algorithms for regularly (lattice-wombling) and irregularly (triangulation-wombling) two-dimensional sampled data. I investigate the reliability of these algorithms in detecting potential ecotones using simulated vegetation data that follow the individualistic, continuum-gradient, and community-type patterns. Ecotones were defined quantitatively as long narrow regions of high rates of change. Under this definition, significant ecotones were found mostly in the community-type patterns using either of the edge detection algorithms (lattice-wombling or triangulation-wombling) and a systematic or random sampling design, respectively.