Edge detection in vegetation: Jornada revisited

We return to the problem of pattern detection with focus on locating vegetation edges. We assume transect sampling with contiguous quadrats and we postulate a vegetation model in which pattern confounds responses to random, serial, and possibly edge‐generating effects. Realizing that strong serial effects will enhance the appearance of vegetation continuity, we probe the residuals for evidence of edges. Two independently derived structures furnish the evidence: a deviations profile and an ordination map. The profile displays compositional changes along the transect as deviations from a theoretical continuum state. The ordination traces the transect as an irregular line in vegetation hyperspace. On this irregular line, the transect segments appear as apices and the compositional dissimilarities of the transect segments as inter‐apex distances. Sharp apices (in full dimension) isolated by long inter‐apex distances (also in full dimension) indicate possible edges. While the subtending angle to an apex reflects changes in population proportions, the criterion for measuring edge sharpness, the height of the deviations profile expresses quantitative effects, the criterion for measuring edge intensity. The deviations profile does not seem to have direct parallel in past edge detection. The inter‐apex distance has a parallel in the Ludwig‐Cornelius compositional distance. We describe the details of sampling and analysis, and present numerical results. We also include a short glossary in App. 5.