Extending community ecology to landscapes

A goal of landscape ecology is to infer processes or constraints that generate spatial pattern in communities and ecosystems . The rich tradition of plant community ecology is now being extended to address spatial pattern in vegetation over large spatial extents. The challenge ill this is that vegetation pattern oil landscapes is fine-grained, which presents sampling problems for large study areas. Further spatial autocorrelation in ecological data, coupled with strong patterns of correlation among environmental factors (such as the gradient complexes governed by elevation) make it difficult to make it clear inferences about the agents patterning landscape-scale vegetation. Here we review the methods of plant community ecology as extended to landscapes and illustrate the challenges with a case study from Sequoia-Kings Canyon National Park in California's southern Sierra Nevada. We outline an iterative approach to such studies, with threee stages. The first stage is a pilot study to characterize the spatial scaling of environmental factors presumed to be important to vegetation: this stage can often be conducted virtually, using digital terrain data. The second stage is iterative and consists of building a preliminary explanatory model using a combination of ordination, classification, and Mantel tests: all analyses based on the same ecological distance or dissimilarity matrices. This preliminary model is then attacked to find its uncertain or sensitive parts and these parametric conditions are mapped into geographic space to identify candidate sites for follow up field studies in the third stage. This approach ensures that the most uncertain aspects of the preliminary model are refined in an efficient manner. As the approach proceeds toward a richer understanding of species environment relationships and vegetation pattern, a need emerges for new kinds of field studies and novel extensions to existing statistical analyses. We discuss possible extensions of these as a natural consequence of this iterative process of model construction and revision.