Using multivariate clustering to characterize ecoregion borders

sets as the basis for visualizing ecoregion maps.1,2 Ecoregions are areas containing similar environmental conditions that are classified for particular purposes. For example, the US Department of Agriculture publishes a map of Plant Hardiness Zones, which divides the US into different ecoregions so that gardeners can select appropriate plants and shrubs for their particular area. The ecoregion is one of the most important concepts in managing and understanding landscape ecology.3,4 Unfortunately, ecologists have long struggled with exactly how and where to locate the dividing lines between ecoregions.5,6 Historically, the process of regionalization— drawing ecoregion borders—has been subjective: experts have attempted to integrate and weigh all of the environmental characteristics and draw the borders accordingly, often without being able to elucidate the logic behind their lines. This subjectivity leads to frequent revisions5–8 and disagreements over particular locations,9 and hampers widespread acceptance and use of such maps. Some experts, in fact, are unable to correctly identify actual ecoregion maps from synthetic maps simulated using a fractal technique.10 Part of the problem is the variable nature of the borders between ecoregions. Some borders are very sharp and distinct, and you can literally stand with your feet in two clearly different regions. Ecologists call such unequivocal and easyto-locate borders ecotones, because they represent sharp cuts. However, most borders are more like the M.C. Escher woodcut Sky and Water I (www.cs.rochester.edu/u/si/images/escher/birds_fish. gif), in which black birds slowly transform into white fish. Although the picture clearly contains two distinct creatures, it’s difficult to locate a line of demarcation between them. We define a new term, ecopause, to indicate the indistinct nature of such borders. But a border may actually change character along its length. For example, a border can begin in one geographic location as an ecotone and transform slowly along its length into an ecopause. Unfortunately, ecologists have had only simple lines with which to visualize these many types of borders. Locating ecoregion borders is a multivariate decision process that must consider a large geo-