SCALE DEPENDENCE OF VEGETATION-ENVIRONMENT CORRELATIONS - A CASE-STUDY OF A NORTH-CAROLINA PIEDMONT WOODLAND

Vegetation and its correlation with environment has been traditionally studied at a single scale of observation. If different ecological pro cesses are dominant at different spatial and temporal scales, the resu lts obtained from such observations will be specific to the single sca le of observation employed and will lack generality. Consequently, it is important to assess whether the processes that determine community structure and function are similar at different scales, or whether, ho w rapidly, and under what circumstances the dominant processes change with scale of observation. Indeed, early work by Greig-Smith and assoc iates (Greig-Smith 1952; Austin & Greig-Smith 1968; see Greig-Smith 19 79; Kershaw & Looney 1985; Austin & Nicholls 1988) suggested that plan t-plant interactions are typically important at small scales, but that the physical environment dominates at large scales. Using a gridded a nd mapped 6.6 ha portion of the Duke Forest on the North Carolina pied mont for a case study, we examined the importance of scale in vegetati on studies by testing four hypotheses. First, we hypothesized that the correlation between vegetation composition and environment should inc rease with increasing grain (quadrat) size. Our results support this h ypothesis. Second, we hypothesized that the environmental factors most highly correlated with species composition should be similar at all g rain sizes within the 6.6-ha study area, and should be among the envir onmental factors strongly correlated with species composition over the much larger extent of the ca. 3500 ha Duke Forest. Our data are not c onsistent with either portion of this hypothesis. Third, we hypothesiz ed that at the smaller grain sizes employed in this study (less-than-o r-equal-to 256 m2), the composition of the tree canopy should contribu te significantly to the vegetation pattern in the understory. Our resu lts do not support this hypothesis. Finally, we predicted that with in creased extent of sampling, the correlation between environment and ve getation should increase. Our data suggest the opposite may be true. T his study confirms that results of vegetation analyses can depend grea tly on the grain and extent of the samples employed. Whenever possible , sampling should include a variety of grain sizes and a carefully sel ected sample extent so as to ensure that the results obtained are robu st. Application of the methods used here to a variety of vegetation ty pes could lead to a better understanding of whether different ecologic al processes typically dominate at different spatial scales.