Dissimilarity analyses of late Pleistocene and Holocene vegetation and climate in eastern North America

J. W. Williams, B. N. Shuman, and T. Webb III

2001 Ecology 82:  3346-3362

Abstract —  Plant formations different from any extant today apparently were widespread in North America and Europe during the last deglaciation, produced by the independent biogeographic responses of plant taxa to climate change.  Dissimilarity analyses of modern and fossil pollen samples in eastern North America show that the unique plant associations centered around the Great Lakes at 14,000 calendar years before present (14 ka) with high dissimilarities from 17 to 12 ka.  The late-glacial fossil pollen assemblages are characterized by 1) high abundances of boreal conifers such as spruce and larch relative to their Holocene values, 2) high abundances of herbaceous types (sedge, sage, and ragweed), 3) high abundances of broadleaved deciduous types (ash, hornbeam, poplar, hazel, and willow), and 4) the low abundance or absence of pine, alder, and birch.  When the fossil pollen samples are assigned to biomes using the affinity score technique, the late Pleistocene pollen samples are assigned to mixed parkland, a biome that is not extant in North America today.  The fastest vegetational changes occurred between 13 and 11 ka, when the late Pleistocene vegetation reorganized into the Holocene biomes, which have persisted to today.  Simulations by the NCAR Community Climate Model, version 1, suggest that late-glacial climates were also unlike modern climates, featuring a ‘hyper-continental’ mixture of colder-than-present winters, warmer-than-present summers, and lower-than-present precipitation.  Dissimilarity analyses of the pollen data and CCM1 simulations for 21, 16, 14, 11, and 6 ka show that 1) the temporal and spatial distribution of high dissimilarities in the vegetation (relative to present) coincide with dissimilarities in simulated climate, 2) the timing and spatial distribution of changes in the vegetation and simulated climate also agree, and 3) the largest climatic and vegetational changes follow the peak period of dissimilarity from present.  Taken together, these three lines of evidence support the hypothesis that the no-analog plant associations were in equilibrium with orbital- and millennial-scale climate changes.  Non-climatic factors such as low atmospheric CO2 concentrations and the presence of now-extinct megafauna species may have increased the openness of the Pleistocene vegetation but by themselves cannot explain the observed mixture of boreal, temperate, and herbaceous taxa in the no-analog pollen assemblages, nor the prevalence of the no-analog pollen samples during the late glacial.

Figures
Figure 1:  Pollen sites used for each time interval.  The past extent of ice sheet extent and coastlines have been digitized for 6, 11, 14, 16, and 21 ka (Bartlein et al. 1998) from Dyke and Prest (1987) and Peltier (1994) and applied to the other 1,000 year intervals.  Areas covered by the Laurentide Ice Sheet are left blank.  All maps use an Albers equal-area projection (standard parallels at 33.33°N and 66.67°N, center at 70°N,100°W).  PDF (745 KB)
Figure 2, Top:  Squared-chord distances (SCD’s) between fossil pollen samples and their closest modern analogs.  Pollen samples with SCD’s>0.15 (dark gray) are considered to have no modern analog.  Bottom:  Vegetation change, measured by calculating, within each pollen record, the SCD between adjacent time intervals.  SCD’s greater than 0.15 indicate a floristic change comparable to travelling between modern vegetation formations (Overpeck et al. 1985).  For both sets of maps the dissimilarity calculations were performed for individual sites and interpolated onto a 50 km grid.  Blank areas indicate no data.  PDF (1524 KB)
Figure 3:  Representations of the late-Pleistocene and modern plant associations as biomes (top row) and as multi-taxon isopoll maps (Jacobson et al. 1987) (rows 2-4).  Biome maps were created using the affinity score technique (Prentice et al. 1996) and the biome definitions of Williams et al. (2000), with spruce parkland and mixed parkland added to represent late-Pleistocene plant associations.  Three taxa are mapped in each multi-taxon map, with a single isopoll shown for each plant taxon.  Primary colors (red, blue, cyan) indicate regions where only one of the three taxa is present in high.  Secondary colors (orange, purple, green) denote the associations between pairs of taxa; areas where all three taxa overlap are beige.  Differences in color between maps, such as those between the maps of the late-glacial and modern pollen assemblages, indicates a change in plant associations (Jacobson et al. 1987).  Second row:  spruce, sedge, and birch (at 20%, 5%, and 10%, respectively).  Third row:  ash, hornbeam, and elm at 5%, 3%, and 5%.  Fourth row:  larch, fir, and pine at 1%, 1%, and 20%.  Blank areas indicate no data.  PDF (697 KB) 
Figure 4:  Dissimilarity values for the fossil pollen data and CCM1 climate simulations.   The top two rows indicate the dissimilarity of the fossil pollen samples or climate simulations from their modern counterparts.  The bottom two rows show the magnitude of change in the vegetation or climate between adjacent time slices.  Because the intervals between time periods are larger than in Fig. 3, the SCD scale was broadened (row 3). For bottom two rows, the coastlines and ice sheets in each map are those of the older time period.  PDF (811 KB)
Figure 5: Scatter plots of the median dissimilarity from present for the CCM1 climate simulations for ENA, plotted against the median dissimilarity of ENA fossil pollen samples (filled circles and solid line) and the percent area of ENA covered by no-analog biomes.  All gridpoints assigned to the spruce parkland, mixed parkland, or conifer woodland biomes by the affinity score technique were considered to have no modern analog.  PDF (23 KB)
Figure 6: a) The median dissimilarity-from-present of ENA fossil pollen samples (solid line), and the median dissimilarity between fossil pollen sites for adjacent 1 ka intervals (dashed line), a measure of vegetational turnover (Jacobson et al. 1987, Overpeck et al. 1992).  The horizontal bar indicates the duration of diverse megafauna communities in North America (Mead and Meltzer 1984). b) The median dissimilarity-from-present of the CCM1 climate simulations for ENA (solid line) and the median dissimilarity between temporally adjacent climate simulations (dashed line). Pollen dissimilarities are measured as a squared chord distance; climate simulation dissimilarities are measured as a standardized Euclidean distance.  Both the pollen and CCM1 simulations show that the vegetational and climatic dissimilarity from present peak between 16 and 14 ka, before the period of maximal change between 13 and 11 ka.  c) Temporal evolution of some of the primary drivers of late-Quaternary climates, as used for the CCM1 paleoclimatic simulations:  northern hemisphere insolation (W/m2) for the summer (SJJA) and winter (SDJF) , land ice extent, and atmospheric CO2 concentrations.  Graph 6c is redrawn from Kutzbach et al. (1998).  PDF (347 KB)