seeded, or irrigated were not sampled. Plots were established 

 within portions of stands that appeared to be relatively uniform in 

 topography and vegetation structure. Within an area, one to five 

 plots were chosen from different topographic positions and where 

 judgement indicated a marked change in vegetation composition. 



Plot selection focused on contemporary stands of vegetation 

 without reference to successional relationships among stands. No 

 attempt was made to solely sample remnants of presettlement vegeta- 

 tion. 



The data were recorded on Natural Heritage Program Community 

 Survey (Figure 1) and Reconnaissance Soil Characterization (Figure 

 2) forms for each plot, and Site Survey Forms (Figure 3) for each 

 site. Instructions for completing these three forms are provided 

 in DeVelice (1991). The Community Survey Forms are basically the 

 same as the general plot data and ocular plant species data forms 

 used by the USDA Forest Service (USDA 1987) . Complete lists and 

 canopy cover estimates of vascular plant species were recorded 

 within each 375 m^ circular study plot. Site information such as 

 altitude, slope, aspect, parent material, landform, and erosion 

 characteristics were also recorded for each plot. A 20-inch deep 

 reconnaissance soil pit was excavated at each site and information 

 was collected to determine soil subgroup and available water 

 holding capacity. 



Data Analysis 



Analysis focused on using a combination of classification, to 

 determine community types, and ordination (gradient analyses), to 

 describe general patterns of communities in relation to environ- 

 mental factors. Classification was accomplished using two-way 

 indicator species analysis (TWINSPAN; Hill 1979a) in the CEP MS-DOS 

 computer package (Mohler 1987). Ordination was achieved using the 

 detrended correspondence analysis (DCA) algorithm in the CANOCO 

 computer package (Ter Braak 1988) . The input data were species 

 cover variables recorded in each plot. Both TWINSPAN and DCA are 

 based on the same mathematical strategy (i.e., reciprocal ave- 

 raging; Hill 1979a, b) and thus offer direct comparisons between the 

 results of ordination and classification. 



All default options in the TWINSPAN algorithm were used except 

 that pseudospecies cut levels were set at 0, 5, and 20 percent 

 cover. Also, all default options were used in running the ordina- 

 tions except that rare species were downweighted. First, the 

 entire data matrix of 52 stands and 268 species were analyzed. 

 Second, to reduce the amount of variation being considered, which 

 is substantial in the whole matrix, the species list was thinned to 

 the 55 most characteristic species. 



