A computerized clustering algorithm, DIVIDE, was used to gain insight into the relationships between physical site factors and existing forest communities in southwestern Virginia. A pair of dichotomous keys was produced that "predicted" the forest type most likely to occur in an area based on topography. Maps of predicted forest types, using the Trayis (1982) and SAF (1980) vegetation classification systems, were produced for the entire study area. Accuracy levels of 57 to 78 percent were obtained. There were no significant differences in classification accuracy between Trayis and SAF forest type predictions (P > 0.25). Herbaceous understory was sampled on the basis of cover, and mast production was estimated in each of the forest types. Forest types on sites had significantly greater amounts of forb and fern cover than those on drier sites. Production of grasses and leaves of woody plants was probably similarly affected, but differences were not significant. Estimates of acorn production were highest in old stands containing a high percentage of oaks. Oak stands on moist sites appeared to have higher estimated mast yields than those on dry sites, but differences were not significant. Based on the results of these analyses, forest types were rated tor deer suitability using compatibility matrices. Deer habitat suitability maps vere produced for the entire study area based on these matrices. Differences between overall suitability values for the Travis and SAF systems suggest that selection of an appropriate forest type classification system is important for wildlife managers. / M.S.

One hundred and twenty-six stands were sampled using a stratified random design in a 6880 hectare Appalachian Oak Forest in southwest Virginia. The objective was to determine the present distribution patterns of tree species and community types and to relate their occurrence to site factors. Of the 52 tree species recorded, the dominants in order of decreasing importance were Acer rubrum, Quercus rubra, Liriodendron tulipifera, Quercus prinus, Acer saccharum and Quercus coccinea. Most site factors of importance, including elevation, lithology, landform, slope position, steepness, exposure, and soil depth appeared to influence the distribution of individual species and community types through their effect on local moisture and temperature conditions. Different site factors best explained the distribution of various species in a multiple regression analysis, thus suggesting that these species are distributed independently of each other. Stand history and interaction between site factors appeared to confound the species-site relationships examined and may explain the large portion of variation unaccounted for by the regressions. Stands were placed in one of six community types with the aid of polar ordination. Three community types were dominated by oak and showed considerable intergradation of species and site preferences. The dominant oaks shifted from Quercus rubra to Quercus prinus to Quercus coccinea on progressively drier sites. Sugar maple-mixed hardwoods, Yellow-poplar, and Hemlock community types were more discrete when compared to the oak community types. Stands dominated by Acer saccharum or Tsuga canadensis occurred mostly in relatively cool, moist conditions at mid to upper elevations, while most stands dominated by Liriodendron tulipifera were in moist locations at lower elevations. Results indicate that shifts in the composition of forest stands when compared to presettlement forests may be attributed to selective logging practices and to the demise of American chestnut. / Master of Science

A decisionmaking framework was developed to assess atmospheric deposition impacts on regional softwood inventory in Virginia. This decision making framework consists of three segments: a forest inventory projection model, a timber production function, and a timber consumption model. The Timber Resource Inventory Model (TRIM) was used to project future forest inventory, given initial inventory data, yield information, and harvest request. The timber production function allows the estimation of the individual effects of input variables on stand growth and yield. The timber consumption model was linked with TRIM to simulate the interactions between timber removals and inventory levels. Algorithm analysis, sensitivity analysis, and an a priori analysis were used to examine the feasibility of TRIM for projecting atmospheric deposition impacts on inventory. Modification of growth and harvest decision variables in TRIM allows this impact estimation. Schumacher's yield model was modified to develop the timber production function according to goodness-of-fit, minimal collinearity, and biological rationale. Crown length was used as a surrogate of a. biological factor to reflect atmospheric deposition impacts on stand growth and yield. The small variance inflation factor allows the crown length elasticity to serve as a measure of the quantitative effects on the yield table. A system of predictor equations was added to the yield equation for simulating stand dynamics. A consumption function approach was used to develop the timber removals model. The BoxCox transformation, the stepwise regression procedure, and standard error were used to select the functional form, predictor variables, and estimates for the timber removals model. This removals model was linked with TRIM for simulating the interactions between removals and inventory levels for Forest Industry and Other Private. The existing forecasts of removals based on Forest Service projections were used for impact estimation for all ownerships. This decisionmaking framework was applied to the softwood inventory data in Virginia to demonstrate the impact estimation. Sensitivity analysis showed that the percentage reduction of inventory and removals is directly related to the crown length reduction. The larger the crown length reduction, the greater the percentage reduction of the inventory. The percentage reduction of yield tables due to the crown length reduction is slightly less than the overall percentage reduction of the inventory but is slightly greater than the overall percentage reduction of removals. The quantitative information on atmospheric deposition impacts on crown variables is a key to the impact estimation for inventory and removals. Also, this decision making framework can be used to measure some silvicultural practice effects on regional inventory. / Ph. D.

Master of Science

Two watersheds on the southern end of Havens State Wildlife Management Area, Roanoke County, Virginia, were sampled for vascular plant species richness. Two-hundred and forty-eight species were identified. Three methods of sampling for species richness in eastern forests were compared: timed-search meanders, belt transects, and plots. Meanders and transects located more species in the same amount of time as plots. Plot sampling encompassed only two-thirds of the richness known from the site. Species-area and species-effort relationships were described by exponential models (number of species = In (area + 1), and number of species = In (time + 1)). Models were used to predict numbers of species which might have been found had more area been sampled or had more time been spent searching. Species-area models yielded more conservative, and probably more accurate, predictions than species-time models. Predictions of species numbers were reasonable for areas as large as 60 ha, but were rather large for areas the size of Havens (2800 ha). Sufficiency of search effort was judged using species-area and species-effort curves. However, it was difficult to tell whether the curves approached horizontal or not. Therefore, this was not a good technique to judge sampling adequacy. Species composition observed by each of the three methods was different. Composition of species lists was 65% similar between meanders and transects, and only 51-58% similar between plots and other methods. These figures were within an expected range. A combination of two methods or repeated meanders was recommended in order to identify a higher proportion of the species present. Seasonal and observer differences, and the effect of learning and taxonomy on richness estimates were discussed. / Master of Science