Abstract
In large-scale gradient studies, selection of the best research sites is critical but time-consuming and costly. Multivariate methods can be used to quickly identify suitable sites from existing data bases. Based on a study of acid deposition in the Great Lakes region (the Michigan Gradient Study), we illustrate the use of multivariate methods in screening potential research sites for similarity. Sites were examined using cluster analysis, principal coordinates analysis, and correspondence analysis. The graphical displays generated by the multivariate methods were used to identify similar sites across the gradient. A list of 31 potential sites was reduced to 5 similar research sites and several alternative sites. The results of the multivariate methods compared well with more traditional methods of research site selection but allowed for multiple comparisons of many potential sites using a variety of data from existing data bases. By eliminating sites that are unacceptable with respect to available data, the multivariate methods reduce the number of sites that require field visitation prior to final site verification. This process represents a substantial savings in time and effort when dealing with a long list of potential research sites.