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 processes are dominant at different spatial and temporal scales, the results obtained from such observations will be specific to the single scale 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, how rapidly, and under what circumstances the dominant processes change with scale of observation. Indeed, early work by Greig-Smith and associates (Greig-Smith 1952; Austin & Greig-Smith 1968; see Greig-Smith 1979; Kershaw & Looney 1985; Austin & Nicholls 1988) suggested that plant-plant interactions are typically important at small scales, but that the physical environment dominates at large scales. Using a gridded and mapped 6.6 ha portion of the Duke Forest on the North Carolina piedmont for a case study, we examined the importance of scale in vegetation studies by testing four hypotheses. First, we hypothesized that the correlation between vegetation composition and environment should increase with increasing grain (quadrat) size. Our results support this hypothesis. Second, we hypothesized that the environmental factors most highly correlated with species composition should be similar at all grain sizes within the 6.6-ha study area, and should be among the environmental factors strongly correlated with species composition over the much larger extent of the ca. 3500 ha Duke Forest. Our data are not consistent with either portion of this hypothesis. Third, we hypothesized that at the smaller grain sizes employed in this study (less-than-or-equal-to 256 m2), the composition of the tree canopy should contribute significantly to the vegetation pattern in the understory. Our results do not support this hypothesis. Finally, we predicted that with increased extent of sampling, the correlation between environment and vegetation should increase. Our data suggest the opposite may be true. This study confirms that results of vegetation analyses can depend greatly on the grain and extent of the samples employed. Whenever possible, sampling should include a variety of grain sizes and a carefully selected sample extent so as to ensure that the results obtained are robust. Application of the methods used here to a variety of vegetation types could lead to a better understanding of whether different ecological processes typically dominate at different spatial scales.