Preliminary Evaluation of Methods for Classifying Forest Site Productivity Based on Species Composition in Western North Carolina

The species indicator approach to forest site classification was evaluated for 210 relatively undisturbed plots established by the USDA Forest Service Forest Inventory and Analysis unit (FIA) in western North Carolina. Plots were classified by low, medium, and high levels of productivity based on IO-year individual tree basal area increment data standardized for initial stocking. Chi-square analysis of contingency tables indicated that productivity classes were not independent (P < 0.05) of the frequencies of occurrence for 4 of 27 common tree species. Multiple logistic regression of a binary variable formed by the high productivity class compared to the combined low and medium classes resulted in a model consisting of elevation and seven significant (P < 0.05) species that produced a classification accuracy of 85 percent; a similar model based on the low productivity class resulted in classification accuracy of 70 percent. A multinomial logistic regression model indicated that elevation and six species were significantly (P < 0.05) associated with the three productivity classes, but overall classification accuracy dropped to 61 percent, mainly due to the poor predictability of low productivity classes. Chestnut oak (Quercus prinus) and serviceberry (Amelanchier spp.) were the most consistent indicator species. Results of this exploratory study suggest that using indicator species for site classification shows promise in hardwood stands by avoiding problems associated Lvith conventional methods based on s i te index. INTRODUCTION Forest productivity evaluation based on indicator species -where the presence of certain vegetative species is associated with the rate of tree growth on forestland -has received relatively little attention in the United States (Daubenmire 1976). Indicator species integrate the complex array of forest environmental components important for tree growth and their presence can used as a phytometer to conveniently assess productivity (Kimmins 1987). Vegetative composition is the basis of the habitat type approach to site classification in much of the arid western United States (Daubenmire 1976), but elsewhere other methods are generally used to evaluate forest productivity (Carmean 1975). Site index, the most widely used method of evaluating forest productivity, is also based on the phytometer premise (Carmean 1975, Spun: and Barnes 1984), but requires the acceptance of a number of underlying assumptions that are typically unknown for the subject stand (Beck and Trousdel 1973). Using site index is problematic in hardwood stands, which tend to be many-aged and consist of mixed species. Determining the age and height of sample trees in these stands is laborious and prone to measurement error; in addi t ion , s i t e index relationships are typically based on simple guidecurve relationships that may be inaccurate (Carmean 1975). Adopting indicator species for si te classification could prove particularly useful in growth and yield equations that use tree lists to drive the models, such as the forest vegetation simulator (Teck et al. 1996), because plot inventory data could also be used for s i te qual i ty determinat ion. Several problems with using indicator species for site quality assessment soon become apparent. Among the most important of these is the paucity of quantitative information on the physiological requirements of different tree species. A considerable body of qualitative information exists on vegetation and environmental associations in the southern Appalachian Mountains, and occurrence of species has been frequently used to classify vegetative associations on forest sites (Whittaker 1966, Mowbray and Oosting 1968, Golden 1974, Callaway et al. 1989). Lacking, however, are quanti tat ive relationships -based on the presence (or absence) of multiple tree species -that could be used to assess s i te productivi ty for management purposes. Additional site classification problems include how to determine measurement units of forest productivity and the number of categories to use in ’ Research Forester, USDA Forest Service, Southern Forest Experiment Station, Bent Creek Experimental Forest, Asheville, NC 28806 ’ Project Leader, USDA Forest Service, Southern Forest Experiment Station, Bent Creek Experimental Forest,