Holocene variation in spatial scales of vegetation pattern in the upper Great Lakes

While continental-scale patterns of vegetation change during the Holocene clearly record the influence of climatic change, the factors governing change at the landscape scale are less clearly defined. In order to characterize the scales of processes determining vegetation patterns during the Holocene, we analyzed a network of 52 pollen sites in the upper Great Lakes region. Pollen percentage data for three dominant tree genera (pine, Pinus; oak, Quercus; birch, Betula) were interpolated from samples bracketing four target years (500, 2500, 4500, and 6500 yr BP). Smoothed isopoll maps of taxon abundance for each target year show broad trends in pollen abundance that correspond to climatic gradients. Residuals, representing the deviation of each pollen datum from the smoothed valued, indicate the amount of spatial variation in pollen abundance independent of that already modeled as a broad gradient. The three genera differ in the magnitude and pattern of residual variation. Oak residuals are relatively small in magnitude, while pine and birch residuals are relatively large and show greater local variability in sign and magnitude. This indicates local variability in tree abundance, as pollen of all three taxa is readily dispersed by wind. Spatial correlograms, which summarize the strength of spatial autocorrelation as a function of distance between pairs of sites for a given taxon, were calculated separately for each target year and allow the quantification of the dominant scale of variability of each taxon. Oak correlograms corroborate the mapped data in indicating the dominance of region-wide trends. In contrast the birch and pine correlograms indicate that factors operating at scales of 150 to 300 km are as important as region-wide trends in governing pollen abundances. The structure of the correlograms for birch and pine pollen changes through time, with birch showing a more patchy spatial pattern in the mid-Holocene (4500 and 6500 yr BP) as compared to the late-Holocene (500 and 2500 yr BP). Pine, in contrast, shows a more strongly autocorrelated pattern in the mid-Holocene. Our results suggest that substrate, an environmental constraint on vegetation at scales of tens to hundreds of kilometers, has been important in governing the spatial distribution of birch and pine in the upper Great Lakes region. The changing distribution of birch and pine is attributed, in part, to changes in the relative abundance of ecologically dissimilar species within these genera. Further, these observations suggest that spatial scales of tree abundances are dynamic and that constraints imposed by substrate vary in importance in response to long-term climatic variation.