A variety of modeling techniques have been used to determine the strength of association between suites of biotic and abiotic variables and species distributions. These associations can then be used to predict responses to environmental change, including climatic change. Bioclimatic models encompass a wide range of complexity. The simplest model is described as a "climate envelope." It is designed to describe static associations between a species' distribution and a single set of climatic variables (Grinnell, 1924, 1928). Modern statistical analyses and improved computer power have facilitated determination of complex suites of climatic and nonclimatic variables that correlate with the range boundaries for a given species (e.g., software such as BIOCLIM and GARP) (Stockwell and Noble, 1991). These models incorporate biological realism, such as local adaptation and differences in the nature of range limitations at different edges. Modern biogeographic models have demonstrated a high level of predictive power in cross-validation tests (Peterson and Cohoon, 1999).
In assessing the strengths of studies as indicators of response to climate change, it is helpful to consider where they lie along axes of time, space, and replication (numbers of populations, numbers of species, etc.). To assess changes in species distributions, data over large geographic areas are important, especially for areas that represent the boundaries of a species' range or migratory destination. To assess trends through time, frequent (yearly is ideal) observations over many decades are most informative. And to assess the generality of the result, good replication is necessary, with many populations/census sites per species to indicate distributional changes within species or many species per community to indicate community shifts.
Other reports in this collection