Quaternary Science Reviews 19 (2000) 489}508
Vegetation and environment in Eastern North America during the Last Glacial Maximum Stephen T. Jackson *, Robert S. Webb, Katharine H. Anderson, Jonathan T. Overpeck , Thompson Webb III, John W. Williams , Barbara C.S. Hansen Department of Botany, University of Wyoming, Laramie, WY 82071, USA Paleoclimatology Program, NOAA National Geophysical Data Center, 325 Broadway, Boulder, CO 80303, USA Institute for Arctic and Alpine Research, University of Colorado, Boulder CO 80309, USA Department of Geological Sciences, Brown University, Providence, RI 02912, USA The Pollen Connection, 402 South 6th Street, Stillwater, MI 55082, USA
Abstract Knowledge of the vegetation and environment of eastern North America during the Last Glacial Maximum (LGM) is important to understanding postglacial vegetational and biogeographic dynamics, assessing climate sensitivity, and constraining and evaluating earth-system models. Our understanding of LGM conditions in the region has been hampered by low site density, problems of data quality (particularly dating), and the possibility that LGM vegetation and climate lacked modern analogs. In order to generate improved reconstructions of LGM vegetation and environment, we assembled pollen and plant macrofossil data from 21 and 17 well-dated LGM sites, respectively. All sites have assemblages within the LGM timespan of 21,000$1500 calendar yr BP. Based on these data, we prepared maps of isopolls, macrofossil presence/absence, pollen-analogs, biomes, inferred mean January and July temperatures and mean annual precipitation for the LGM. Tundra and open Picea-dominated forest grew along the Laurentide ice sheet, with tundra predominantly in the west. In the east, Pinus-dominated vegetation (mainly P. banksiana with local P. resinosa and P. strobus) occurred extensively to 343N and possibly as far south as 303N. Picea glauca and a now-extinct species, P. critchxeldii, occurred locally. Picea-dominated forest grew in the continental interior, with temperate hardwoods (Quercus, Carya, Juglans, Liriodendron, Fagus, Ulmus) growing locally near the Lower Mississippi Valley at least as far north as 353N. Picea critchxeldii was the dominant species in this region. The Florida peninsula was occupied by open vegetation with warm-temperate species of Pinus. Eastern Texas was occupied by open vegetation with at least local Quercus and Picea. Extensive areas of peninsular Florida and the continental interior had vegetation unmatched by any modern pollen samples. The paleovegetational data indicate more extensive cooling in eastern North America at the LGM than simulated by either the NCAR CCM0 or CCM1 climate models. The occurrence of cool-temperate conifers and hardwoods as far north as 34-353N, however, indicates less severe cooling than some previous reconstructions. Paleoclimate inferences for the LGM are complicated by lowered atmospheric CO concentrations, which may be responsible for the open nature and dominance of conifers in LGM vegetation. 2000 Elsevier Science Ltd. All rights reserved.
1. Introduction The biogeography and climate of unglaciated eastern North America during the Last Glacial Maximum (LGM) have been the subject of long-standing interest
* Corresponding author. Tel.: 001-307-766-2819; fax: 001-307-7662851. E-mail address:
[email protected] (S.T. Jackson). Current address: Institute for the Study of Planet Earth, University of Arizona, Tucson, AZ 85721 USA. Current address: National Center for Ecological Analysis and Synthesis, Santa Barbara, CA 93101.
(Gray, 1884; Adams, 1902, 1905) and considerable disagreement (Deevey, 1949; Braun, 1950, 1955). The LGM is of particular interest to paleoclimatologists because orbital parameters were similar to today, but other boundary conditions and climatic forcings (e.g., CO , global ice-volume, sea level, sea-surface temperatures) were very di!erent (Webb and Kutzbach, 1998). Biogeography and vegetation during the LGM are of interest to biologists investigating the historical context of the biota (Brown and Lomolino, 1998), the antiquity and development of communities and ecosystems (Webb, 1988; Overpeck et al., 1992), and distances and rates involved in postglacial migrations (Cain et al., 1998; Clark et al.,
0277-3791/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 3 7 9 1 ( 9 9 ) 0 0 0 9 3 - 1
490
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
1998). LGM patterns in climate and vegetation are important to understand global climate sensitivity, assess hypothesized climate forcing mechanisms, and evaluate earth-system models (e.g., Atmospheric General Circulation Models (AGCMs), biogeochemical models, coupled ocean-atmosphere models). New paleoecological and paleoclimatic records have improved our knowledge of LGM vegetation and environment in eastern North America since the reviews of Peterson et al. (1979) and Wright (1981). In this study we present a new series of primary maps of pollen and plant macrofossil data, new maps of vegetation formations and biomes inferred from pollen data, and new paleoclimate maps based on application of climate response-surfaces to pollen data. Our aim is to summarize current knowledge, identify uncertainties, and draw attention to important questions and research priorities. Our compilation and synthesis should contribute to assessment of state-of-the-art AGCM simulations generated as part of the Paleoclimate Modeling Intercomparison Project (PMIP) (Joussaume and Taylor, 1995).
2. Data and Methods 2.1. Chronology of the Last Glacial Maximum Late Quaternary chronologies are often complicated by mismatches between C-based chronologies and calendar-year chronologies. These mismatches, caused by temporal variations in atmospheric concentrations of C, tend to increase with increasing age (Bard et al., 1990; Hughen et al., 1998). Previous syntheses have assigned the LGM an approximate age of 18,000 C yr BP, and have set bracketing dates of 500}1500 C yr. For example, in mapping macrofossil occurrences, Jackson et al. (1997) de"ned the LGM as 18,000$1500 C yr (i.e., 16,500}19,500 C yr). Calibrations based on Th/U dating of corals (Bard et al., 1990) indicate that this time span is 21,484#1716/!1956 calendar yr BP (i.e., 19,528 to 23,200 calendar yr BP). We therefore de"ned the LGM as 21,500$1500 calendar yr BP (16,925 to 19,330 C yr BP). 2.2. Site selection and age models Sedimentary records dating to the LGM in eastern North America are few and variable in quality; they include sequences from sediments of deep lakes (mostly in Florida) and shallow ponds or wetlands, and `snapshota records from alluvial sediments, loess, and buried soils. Age estimation of sediments in the southeastern US is confounded by low deposition rates, depositional and erosional hiatuses, and radiocarbon-dating uncertainty, which stems from carbonate problems, root intrusion, and low organic content of sediments (Webb and Webb,
1988; Jackson and Whitehead, 1993). The compilation of paleoenvironmental data sets involves tradeo!s between inclusiveness and data quality. In selecting sites for this study, we chose to minimize errors in age estimation and data quality, at the expense of site number and coverage (Table 1). We started with the 18 ka sites used in the COHMAP pollen database (Webb et al., 1993, 1998) and the macrofossil database of Jackson et al. (1997), and then added recently reported sites with apparent records in the LGM range. For sites with sedimentary sequences (i.e., lakes, ponds, and wetlands with at least two radiocarbon dates bracketing the LGM), we applied the criteria outlined by Webb and Webb (1988) in evaluating age-models. We calculated sediment accumulation rates for the intervals spanning the Last Glacial Maximum, and examined sediment descriptions. Sites with LGM sediment accumulation rates of (10 cm/1000 yr were eliminated from the data set (Webb and Webb, 1988). Records with accumulation rates of 10}25 cm/1000 yr were scrutinized for evidence of hiatuses. For `snapshota sites, at least one C date clearly associated with the pollen and/or macrofossil assemblages had to fall within the LGM interval. Application of these criteria resulted in the elimination of nine sites that were used in the COHMAP 18 ka pollen data set (Webb et al., 1993, 1998) and three (all COHMAP sites) that were used in the 18 ka macrofossil data set of Jackson et al. (1997). Five sites (Buckles Bog, Muscotah Marsh, Pigeon Marsh, Singletary Lake, White Pond) were rejected because sediment accumulation rates spanning the LGM period were (10 cm/1000 yr, suggesting hiatuses or poor temporal resolution (Webb and Webb, 1988). Camel Lake was eliminated because sediment accumulation rates are low (12 cm/1000 yr), the LGM interval is broadly bracketed by C dates (29,350 and 14,330 yr BP), and there are numerous lithological changes between those dates (including &40 cm of sandy clay) (Watts et al., 1992). Pollen-bearing sediments at Boney Spring, Ninepin 24, and Rayburn Salt Dome have no C dates within the LGM time interval as we de"ned it. Hiatuses may occur between bracketing dates at all three sites (they all have frequent lithological changes, including alternating organic and clastic beds), so reliable ages cannot be assigned to any of the pollen assemblages. Some assemblages may be within the LGM window, but it is not clear where. We also chose not to use pollen data from the Friesenhahn Cave Site (Hall and Valastro, 1995) owing to uncertainties in dating (four bone dates ranged from 14,020 to 19,600 C yr BP), and to unclear relationships between the stratigraphic positions of the pollen samples and the C dates. We retained the Nonconnah Creek Site in our data set, although we applied a di!erent age model than that adopted by Webb et al. (1993, 1998) and Jackson et al. (1997). The previous age-model interpolated between a C date of 17,195 yr BP and a presumed date of
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
491
Table 1 Pollen and macrofossil sites used in eastern North America LGM synthesis Site No. Site
Latitude
Longitude
Elev
COHMAP
Dating
Data
Reference
Anderson Pond, TN Anderson Run, OH
36302 39322 30
85330 83305 30
150 278
Y N
1, 1 S
P13,M7,S M1
3 4 5 6 7 8 9 10 11
Big Run Bog, WV Bob Black Pond, GA Browns Pond, VA Clear Pond, SC Conklin Quarry, IA Cupola Pond, MO Francis Lake, NJ Graham County Site, KS Jackson Pond, KY
39307 34319 30 38309 33329 41342 36329 40335 39320 37327
79335 84352 79337 78354 91333 91304 74331 99349 30 85343
980 285 620 10 218 244 189 675 212
N N N N Y Y Y N Y
2, 1; 5, 3, S 4, 2, S 1,
P6,M1,S P1,M8,S P4,M1 P26,S P2,M1 P4 P3 M1 P7,M5,S
12 13
Lake Tulane, FL Nonconnah Creek, TN
27325 35305
81318 89355
36 80
Y Y
S
P10,S P1,M1
14 15
Patschke Bog, TX Pinckneyville Creek Site 9, MS Pinckneyville Creek Site 19, MS Pinckneyville Creek Site 20, MS Powers Fort Swale, MO Princeton Sewer Plant Site, OH Rockyhock Bay, NC Quicksand Pond, GA Sheelar Lake, FL Tunica Bayou Site 23, LA Wolf Creek, MN
30313 31302 32
97304 91328 43
142 76
N N
6, 1 S
P3 M1
H.R. Delcourt (1978a, 1979) Burns (1958) and Goldthwait (1958) Larabee (1986) Jackson et al., unpublished Kneller and Peteet, 1993 Hussey (1993) Baker et al. (1986) Smith (1984) Cotter (1984) Wells and Stewart (1987) Wilkins (1985) and Wilkins et al. (1991) Grimm et al., 1993 P.A. Delcourt (1978b) and Delcourt et al. (1980) Camper (1991) Givens and Givens (1987)
31302 45
91328 46
79
N
S
P1,M1
Jackson and Givens (1994)
31302 54
91328 55
79
N
S
P1,M1
Jackson and Givens (1994)
36322 39316
90321 84327
91 180
N N
2, 1 S
P5 P1,M1
36306 34320 29319 30357 43 46307
76324 84352 82300 91331 10 94307
6 285 51 31 375
Y Y Y N Y
2, 6, 1, S 4,
P3,S P8,M11 P6,M5 P1,M1 P4,M3
Royall et al. (1991) Jackson and Miller, unpublished and Miller (1992) Whitehead, 1981 Watts, 1970 Watts and Stuiver, 1980 Jackson and Givens (1994) Birks, 1976
1 2
16 17 18 19 20 21 22 23 24
2 2, 2 1 1 1 1 4
1, 6 3 6 4
Denotes whether site was included in COHMAP database used in previous syntheses of Webb et al. (1993, 1998) and Williams et al. (1998, 1999). S"snapshot assemblage with at least one associated date within 1500 calendar years of LGM (21,500 yr BP); 1"closest bracketing date is within 1500 calendar years of LGM; 2"closest bracketing date is within 2000 calendar years of LGM; 3"closest bracketing date is within 2500 yr of LGM; 4"closest bracketing date is within 4000 calendar yr of LGM; 5"closest bracketing date is within 5000 calendar yr of LGM; 6"closest bracketing date is within 5500 calendar yr of LGM. Paired numbers respectively indicate upper and lower dates bracketing the LGM interval. P"pollen data; M"macrofossil data; S"Pinus-pollen size data. Numbers refer to number of LGM samples for each data type. In the case of Pinus-pollen size, one sample was used for each site. Watts (1970) published complete pollen and macrofossil sequences from a 5-cm diameter core from this site, with a single basal C date outside the LGM interval. In our mapping, we used data from two 10-cm diameter cores obtained by Jackson et al. (unpublished). These newer data corroborate the results of Watts (1970), but have more diverse macrofossil assemblages and more associated C dates (see Table 2). Site was used in Williams et al. (1998b) but not other COHMAP data syntheses. Age model modi"ed from that used by COHMAP (see text). Erroneously reported in Jackson et al. (1997).
&13,000 yr BP at the top of the pollen-bearing sediments (P.A. Delcourt, 1978a; Delcourt et al., 1980). The latter age estimate was based on pollen-stratigraphic correlations with distant sites (e.g., Anderson Pond, 350 km east) and on an assumption that deposition at the site continued uninterrupted until cessation of glacial meltwater in the adjacent Mississippi Valley (P.A. Delcourt, 1978b). In this paper, we treat the Nonconnah Creek site as a snapshot assemblage centered on the C of 17,195$505 yr BP. That date was obtained from `wood fragments, one shell of black walnut (Juglans nigra), and a cone of white spruce (Picea glauca), all collected in place from the surface of the
palate within the mastodon skull...; the position of the fossil material is stratigraphically correlated with the NC, TN-1-A-50 cm deptha (P.A. Delcourt, 1978a, p. 29). In our data set, we used only the pollen and macrofossil assemblages at 50 cm depth at NC, TN-1-A (P.A. Delcourt, 1978b; Delcourt et al., 1980), which are clearly associated with the C date. Other C dates at the site predate the LGM interval, and lithological evidence suggests hiatuses between those dates and the date we used. The data set used in our analyses consists of 24 sites, including 21 pollen sites and 17 macrofossil sites (Fig. 1; Table 1). Table 2 provides details of age-model
492
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
development for sites not used in the previous compilations of Webb et al. (1993) and Jackson et al. (1997). 2.3. Pollen data Fossil pollen data were obtained from the North American Pollen Database (NAPD), available at the World Data Center-A (WDC-A) for Paleoclimatology. (http://www.ngdc.noaa.gov/paleo/paleo.html). We calculated pollen percentages for each LGM assemblage based on a pollen sum consisting of 21 arboreal taxa
(Abies, Acer, Betula, Carya, Castanea, Celtis, Fagus, Fraxinus, Juglans, Larix, Liquidambar, Ostrya/Carpinus, Picea, Pinus, Platanus, Populus, Quercus, Tilia, Tsuga, Ulmus/Cupressaceae) and 8 non-arboreal taxa (Alnus, Ambrosia, Artemisia, Asteraceae, Chenopodiaceae/Amaranthaceae, Corylus, Cyperaceae, Poaceae). This sum included three pollen types (Larix, Ambrosia, Poaceae) that were not included in the analyses of Webb et al. (1993, 1998). Pollen percentages were interpolated simultaneously along spatial and temporal coordinates to a 100;100 km grid. We used a set of decision rules designed to avoid unwarranted extrapolation of the data in either space or time and to compensate for irregular temporal and/or spatial distribution of the fossil samples. First, we used initial search windows of 100 km and 600 years (centered on LGM at 21,500 calendar yr BP). If no fossil samples were found within the initial search, we used an iterative process that expanded the spatial search window dimensions at 50-km intervals to a maximum of 200 km and the temporal search window dimensions at 300-year intervals to a maximum of 1500 calendar years. A tri-cubic weighting of fossil samples within the spatial and temporal search window was used to calculate mean fossil pollen percentages for each grid node. If no fossil samples were found within the fully expanded search window, the grid node was identi"ed as having no data. 2.4. Macrofossil data
Fig. 1. Locations of LGM sites used in maps and analyses. Gray circles denotes sites with pollen data. Black triangles denotes sites with macrofossil data. Site numbers correspond to those in Table 1.
Macrofossil data were obtained from the North American Plant Macrofossil Database (NAPMD), available at the WDC-A for Paleoclimatology. We assessed presence/absence of macrofossils for each stratigraphic interval at each site (Jackson et al., 1997). If a species was represented by one or more occurrences at a site within a 3000-year time-window spanning the LGM (i.e., 20,000 to 23,000 calendar yr BP; 16,925 to 19,330 C yr BP), it was mapped as present at that site. If the taxon was absent from all LGM samples at the site, and was judged likely to have occurred in sediments of the type studied at the site, it was mapped as absent. For example, Quercus is
Table 2 Age-model information for sites not included in previous syntheses (Webb et al., 1993; Jackson et al., 1997). Additional details can be obtained from the NAPD web site Site
Number of C Dates
Age-Model Comments
Bob Black Pond (Core 1092/1) Bob Black Pond (Core 1093/1-2) Clear Pond Patschke Bog Powers Fort Swale Princeton Sewer Plant Site
1 4 9 4 8 1
Snapshot Linear interpolation Third-degree polynomial Linear interpolation Linear interpolation Snapshot, based on AMS-dated Salix twig (see Miller, 1992)
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
493
Table 3 Sites and sources used for Pinus pollen-size data Site
Estimated Age (C yr BP)
Data Source
Sample Size
Depth (cm)
Reference
Anderson Pond, TN Big Run Bog, WV Bob Black Pond, GA
18,915 16,702 18,230
D D R
100 U 50
675 228 143
Clear Pond, SC Jackson Pond, KY Lake Tulane, FL
18,180 18,279 17,652
R D R
30 U 50
932 455 922
Rockyhock Bay, NC
17,559
R
50
370
H.R. Delcourt (1978a, 1979) Larabee (1986) B.C.S. Hansen and S.T. Jackson (unpublished) Hussey (1993) Wilkins (1985) B.C.S. Hansen and S.T. Jackson (unpublished) Whitehead (1981)
R"raw data (provided directly by original analyst or in table in reference); D " digitized from percent-frequency histograms, converted to raw data by multiplying estimated percentages by number of grains measured. Number of pollen grains measured in sample. U indicates that sample size was not speci"ed in the original reference. In these cases, we assumed a sample size of 100 for purposes of calculation. Depth below sediment surface.
represented in macrofossil assemblages by relatively large and poorly dispersed organs (logs, acorns), so it was recorded as absent only for sites where large sediment volumes were examined (e.g., exposures of #uvial or lacustrine sediments or buried soils). Quercus was not mapped at sites where small-volume sediment cores were obtained. Other taxa treated in this manner include Fagus, Juglans, Liriodendron, Carpinus, Ulmus, Acer, and Carya. In contrast, Picea occurs in macrofossil assemblages as relatively small and well-dispersed organs (needles, twigs, seeds), and was mapped as absent at all sites where it did not occur. Other taxa treated in this way include Dryas integrifolia, Larix laricina, Abies sp., Pinus banksiana, P. resinosa, P. strobus, and Betula papyrifera. Species-level identi"cations can be made for Picea macrofossils based on cone morphology and/or needle anatomy (Weng, 1998; Jackson and Weng, 1999). Presence/absence of Picea mariana, P. glauca, and P. critchxeldii (Jackson and Weng, 1999) were mapped only at sites where large sediment volumes were studied (cones) and at sites where anatomical studies of needle anatomy have been conducted (needles). 2.5. Pine pollen diwerentiation The 13 species of Pinus in eastern North America today range from peninsular Florida to James Bay. Previous paleoecological syntheses have split these into northern and southern groups (e.g., Webb et al., 1993,1998). The northern group includes boreal (P. banksiana) and cooltemperate species (P. resinosa, P. strobus, P. rigida); the latter two species occur in the southern Appalachian highlands as far south as northern Georgia. The remaining nine species are warm-temperate and/or subtropical; none occur today farther north than 413N (Little, 1971).
Pollen of Pinus strobus can be di!erentiated morphologically from the other 12 species. P. strobus pollen occurs only in trace amounts at all LGM sites where it has been di!erentiated. Of the remaining 12 species (all Pinus Subgenus Pinus), pollen grains of two of the northern species, P. banksiana and P. resinosa, are substantially smaller (corpus breadth of 30}42 m) than those of the others (40}55 m) (Whitehead, 1964). Pollen-size studies of Pinus Subgenus Pinus have been conducted at several LGM sites (Table 3). These studies, together with macrofossil data, can be used to determine geographic distributions of the northern and southern Pinus species. We compiled pollen-size data from eight sites (Table 3) to identify the LGM distributions of northern and southern Pinus species. 2.6. Pollen Analogs and Vegetation Formation Maps We used squared-chord distance (SCD) to assess the similarity of each LGM pollen assemblage with each of 1744 modern pollen samples from eastern North America (obtained from the NAPD data set). We adopted the threshold value of 0.15 de"ned in previous studies (Overpeck et al., 1985, 1992) to identify fossil samples with no modern analogs. Ambrosia, Castanea, and Poaceae were eliminated from the pollen sums for the analog analyses to avoid e!ects of anthropogenic vegetation patterns of the past two centuries. We split northern and southern Pinus populations based on macrofossil and pollen-size data. If there were no modern samples with a SCD of 0.15 or less, the fossil sample was classi"ed as a no-analog sample (having no modern analog pollen spectra). Otherwise fossil samples were considered to have a modern analog and assigned to the modern vegetation formation with the greatest number of pollen samples having SCD
494
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
of (0.15 to the fossil sample. If there was a tie for the greatest number of analog samples, we assigned the vegetation formation with the lowest mean SCD. The LGM modern analog assignments were mapped to the 100;100 km grid using the spatial-temporal search-window approach used in mapping the fossil pollen data. As with the percentage data, if no fossil samples were found within the fully expanded search window, the grid node was identi"ed as having no data. For nodes with interpolated data, a hierarchy of decision rules were followed to determine which of the modern-analog vegetation types associated with the LGM samples within the search window should be assigned to the grid point. If a given modern-analog vegetation type within the search window was 10 times more frequent than any other vegetation type, the node was assigned that type. Otherwise, we calculated the tri-cubic weighted distance in both time and space between every fossil sample within the search window and the target node. We then calculated the normalized mean tri-cubic weighted distance for each vegetation type by summing the same number of samples for the most abundant vegetation types. In order to avoid bias from uneven distribution of fossil samples in time or space, each grid point was assigned the vegetation type with the smallest normalized mean tri-cubic weight. We also mapped the minimum SCD between each fossil sample and the modern samples to assess spatial patterns in strength of association. These data were gridded to the 100;100 km grid using the same approach described for the LGM fossil pollen data.
scores were gridded to the 100;100 km grid using the same approach described for the LGM modern analog maps. 2.8. Pollen-inferred paleoclimate We used response surfaces (Bartlein et al., 1986) and dissimilarity coe$cients (Overpeck et al., 1985) to infer paleoclimate variables (mean January and July temperature, mean annual precipitation) from the fossil pollen assemblages. Our methods follow those of Webb et al. (1993, 1998). We calculated squared chord distances between the interpolated pollen assemblage at each 100;100 km grid node and the `virtuala pollen assemblage de"ned by the "tted values for the 27 responsesurfaces corresponding to each of our pollen types. Inferred paleoclimate estimates for each point consisted of the climate values associated with the `virtuala assemblages most similar to the corresponding interpolated fossil assemblage. We used a squared chord-distance threshold of *0.4 to identify fossil assemblages with no suitable climatic analog based on comparison of 1744 modern pollen assemblages. More than 98% of the modern assemblages matched `virtuala assemblages with SCD of (0.4, and we observed a substantial increase in SCD at this point. Further methodological details are provided by Webb et al. (1993). The LGM pollen-inferred paleoclimate reconstructions were gridded to the 100;100 km using the same approach described for the LGM fossil pollen data.
2.7. Aznity scores and biome maps 3. Results We prepared biome maps for the LGM and modern vegetation by calculating the a$nity scores for each biome, using the methods described by Prentice et al. (1996) and Williams et al. (1998, 1999). We used the same pollen types and the same modern pollen samples as used in the analog analyses. In order to "lter out noise introduced by long-distance pollen transport and artifacts of pollen sums, we used thresholds of 5% for Pinus, 2.5% for Quercus, and 1% for all other taxa in calculating a$nity scores (Williams et al., 1998, 1999). In order to assess the strength of the biome assignments based on the a$nity scores, we calculated an index based on the highest a$nity score determined for the LGM pollen assemblage at each site divided by the maximum possible a$nity score for that site. The maximum possible a$nity score was the score that would have been attained by a site had all of the pollen types in the assemblage contributed to a single biome. The index we calculated is a measure of the proportion of the pollen assemblage that actually `voteda for the particular biome we assigned to the site. In general, the higher the index value, the greater the con"dence we place in the biome assignment. The LGM biome assignments and index
3.1. Isopoll and macrofossil maps Our isopoll maps (Fig. 2) show two striking features: an absence of strong latitudinal patterns for most taxa, and a longitudinal pattern in which Pinus dominates pollen assemblages along the Atlantic Coast and interior highlands and Picea dominates in the continental interior. Picea pollen percentages are highest in the lower Mississippi Valley and along the ice margin. Quercus, Acer, Alnus, Fraxinus, and Cupressaceae also have maxima in the interior, although Quercus and Cupressaceae are also abundant in peninsular Florida and the southern Atlantic Coastal Plain. Other temperate and boreal taxa (Abies, Ulmus, Betula, Ostrya/Carpinus, Corylus) are widely distributed but in trace quantities (Fig. 2). Cyperaceae pollen shows a strong latitudinal pattern, with highest percentages within 500 km of the ice margin, although it also has secondary maxima in Florida and east Texas (Fig. 2). Artemisia pollen occurs widely in trace amounts, and other herbaceous and graminoid taxa (Asteraceae, Chenopodiaceae/ Amaranthaceae, Poaceae, Ambrosia) have maxima
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
495
Fig. 2. Isopoll maps for selected pollen types in LGM assemblages.
in peninsular Florida and scattered occurrences elsewhere. Poaceae has a large maximum in eastern Texas (Fig. 2). Plant macrofossil sites are concentrated in the continental interior (Fig. 3). The macrofossil maps con"rm the
widespread occurrence of Picea, which ranged from the ice margin south to east-central Louisiana (Fig. 3). Pinus macrofossils are absent from sites in the interior lowlands and near the ice margin. Pinus banksiana macrofossils occurred in highlands of eastern Tennessee (305 m elevation)
496
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
Fig. 3. Presence/absence maps for selected plant macrofossil types in LGM assemblages. Closed circles denote presence; open circles denote absence. Presence/absence patterns of Pinus strobus and Betula papyrifera (not shown) are identical to that shown for Pinus resinosa. Patterns for Juglans nigra are identical to those for Carya. Patterns for Liriodendron tulipifera are identical to those for Fagus grandifolia. Patterns of Ulmus americana, Carpinus caroliniana, and Acer sp. are similar to that for Quercus sp.
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
and northwest Georgia (285 m), where Pinus resinosa, P. strobus, and Betula papyrifera also occurred (Fig. 3). Dryas integrifolia, a tundra/boreal woodland herb, occurred near the ice margin in Minnesota and Iowa, but not at ice-margin sites in Ohio. However, several bryophytes characteristic of tundra and open boreal woodlands occurred near the ice margin in Ohio (Miller, 1992). Tundra/woodland herb and bryophyte taxa also occurred at the Minnesota and Iowa sites (Birks, 1976; Baker et al., 1986). Occurrences of temperate deciduous trees (Quercus, Juglans, Carya, Fagus, Liriodendron, Ulmus, Carpinus) are restricted to the Lower Mississippi Valley (Nonconnah Creek and/or Tunica Hills), but sites where these taxa might be represented by macrofossils are lacking to the east (Fig. 3). Picea is represented by three species in eastern North America today. Two are represented in the LGM record (Fig. 3). Picea mariana occurred near the ice margin in Iowa, and P. glauca is recorded at one highland site in Georgia. Both of these species were probably more widespread; few sites where macrofossils could be identi"ed to species are represented. P. glauca has been recorded from a site in Kansas dating to 16,420 C yr BP (Jaumann, 1989) and an ice-margin site in SW Ohio dating to 19,600 C yr BP (C. Weng, N.G. Miller, S.T. Jackson, unpublished). Another, now-extinct species, P. critchxeldii (Jackson and Weng, 1999), is recorded at several LGM sites, both in the Lower Mississippi Valley and NW
497
Georgia (Fig. 3). This species, which grew in the Lower Mississippi Valley in association with temperate and cool-temperate hardwoods (e.g., Quercus, Juglans, Liriodendron, Carya, Carpinus, Fagus) and in Georgia with boreal and cool-temperate conifers (Picea glauca, Pinus banksiana, P. resinosa), was probably responsible for the Picea pollen maximum in the Lower Mississippi Valley (Fig. 2). The Pinus pollen-size plots (Fig. 4) indicate that the Pinus-dominated pollen assemblages throughout the Atlantic Coastal and interior highlands regions consisted predominantly of small Pinus pollen (i.e, P. banksiana and/or P. resinosa). Dominance of Pinus banksiana and/or P. resinosa north of 343N is corroborated by macrofossil records from E Tennessee and NW Georgia (Fig. 3). Temperate and/or subtropical pines, which produce larger pollen grains, occurred in the southern Florida peninsula (Fig. 4). Unfortunately, there are no pollen-size or macrofossil records between 283N and 343N, so the northern extent of the southern pines, the southern extent of the northern pines, and the degree of intermixing of the two groups cannot yet be determined. We note, however, that Pinus pollen from late-glacial sediments of Camel Lake in western Florida (620 cm depth, estimated age"13,000 C yr B.P.) have a bimodal size distribution; more than half the grains are within the size range of P. banksiana and P. resinosa (sample median"38.7 m; range"30.5}54.6 m)
Fig. 4. Box plots showing size distributions of Pinus Subgenus Pinus pollen grains from LGM sediments at seven sites. Midpoints of box plots represent sample medians; box boundaries represent the interquartile range (i.e., middle 50% of the distribution). Whiskers represent upper and lower adjacent values, and dots represent outside values (Cleveland, 1994).
498
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
(B.C.S. Hansen and S.T. Jackson, unpublished). This mixture may indicate that the boreal and cool-temperate Pinus species (P. banksiana and/or P. resinosa) occurred in the Gulf Coastal Plain at the LGM. Unfortunately, LGM sediments are apparently lacking at Camel Lake and have not yet been recovered from the Gulf Coastal Plain between peninsular Florida and the Lower Mississippi Valley.
3.2. Analog map Our LGM results (Fig. 5), based on a revised data set, represent an improvement over the previous LGM analog-based vegetation reconstruction for eastern North America of Overpeck et al. (1992). Along the southern margin of the Laurentide Ice Sheet, we map boreal forest in the east, with taiga limited to the far west (Fig. 5), consistent with the Overpeck et al. (1992) reconstruction. We also con"rm that mixed conifer/hardwood and boreal forest extended from the mid-Atlantic region across the center of eastern North America. The vegetation reconstruction for northern Florida is problematic. Pollen assemblages at Sheelar Lake are Pinus-dominated (Watts and Stuiver, 1980), but we do not know whether the species represented are boreal/cool-temperate, warm-temperate/subtropical, or a mixture. We did two analyses, classifying the LGM Pinus at Sheelar as all southern and then all northern. The southern analysis yielded closest analog-matches with southern mixed forest, and the northern analysis matched most closely with deciduous forest. Because we have no objective basis for di!erentiating between these categories, we map northern Florida as a mixture of the two types. LGM vegetation of northern Florida may have lacked modern vegetation analogs. Our LGM vegetation formation map includes a previously unmapped region of vegetation lacking modern analogs across the western sector of eastern North America and southern Florida (Fig. 5). The no-analog pollen assemblages in the Lower Mississippi Valley region were dominated by Picea, with an admixture of Quercus and non-arboreal taxa. The no-analog assemblages in southern Florida and Texas were dominated by a mixture of southern Pinus, Quercus, and non-arboreal taxa. Additional regions may have had no-analog vegetation. Although the Tunica Hills region is mapped as taiga in Fig. 5, two of the three LGM pollen assemblages lacked modern analogs, and the third, Pinckneyville Creek Site 19, had analog-matches with only two modern pollen assemblages. Conklin Quarry, the single pollen site from the Upper Mississippi Valley region mapped as taiga, had only one modern analog. Similarly, two sites in the area mapped as boreal forest south of the midcontinental ice sheet, Big Run Bog and the Princeton Site, each had only one modern analog (the same site for each), al-
though other sites in the region (e.g., Jackson Pond) had multiple analogs. Overall, our mapped reconstruction indicates extensive no-analog LGM vegetation in eastern North America, and probably underestimate the actual extent of no-analog vegetation, which we suspect spanned the entire Coastal Plain and Mississippi Valley and perhaps the entire study area. Overpeck et al. (1992) suggested that the Picea- and Pinus-dominated LGM assemblages were distinct from modern boreal forest assemblages, although the modern analog method was unable to resolve these di!erences with their data. Our revised dataset drew out some subtle di!erences between the composition of the LGM and the modern conifer-dominated forests. These di!erences are reinforced by consideration of the macrofossil data, which indicate that the forests of the Lower Mississippi Valley were dominated by a now-extinct species of Picea, growing with cool-temperate and temperate hardwoods (Fig. 3). The map of minimum SCD between each LGM fossil sample and modern samples showed the closest analog matches between modern and fossil assemblages were for southeastern mixed forest and deciduous forest in northern Florida and boreal forest along the northeastern margin of the ice sheet (Fig. 6). Analog matches for tundra and mixed forest in the Atlantic coastal region have moderate SCD values, and the no-analog regions have large SCD values when compared to modern pollen samples (Fig. 6). 3.3. Biome map Our modern pollen-inferred biome map, based on af"nity scores, is overall similar to those of Williams et al. (1998, 1999), and shows general correspondence to continental-scale biome patterns (Fig. 5). Our LGM biome map (Fig. 5) is similar to the LGM biome map produced using identical methods by Williams et al. (1998, 1999). The primary di!erences result from di!erences in the datasets used in the respective analyses (See Methods). Biome assignments are identical for all the sites shared by the datasets. The results show cool mixed forest, taiga, and steppe near the ice margin, and extensive cool mixed forest through the interior Southeast and lower Mississippi Valley (Fig. 5). Coniferous woodlands dominate along the Atlantic coast of the Carolinas and Florida. Eastern Texas is classi"ed as tundra, evidently because of the dominance of Poaceae, Cyperaceae, and Asteraceae pollen together with signi"cant representation of Pinus, Alnus, and Picea at Patschke Bog. Ratios of maximum a$nity scores to total possible a$nity scores show spatial patterns related to strength of the biome reconstructions. Ratios are lowest near the western edge of the ice margin, eastern Texas, and Florida, and highest in the continental interior and in the east (Fig. 6).
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
499
Fig. 5. Upper left: modern pollen sites and corresponding vegetation formations. Upper right: modern pollen sites and corresponding pollen-inferred biomes. Middle left: modern vegetation formations. Middle right: modern biomes. Lower left: Pollen-inferred LGM vegetation formations. Lower right: Polleninferred LGM biomes. Modern vegetation and biome maps adapted from KuK chler (1964) and Rowe (1972), and represent potential natural vegetation (sensu KuK chler, 1964).
500
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
Fig. 6. Maps of squared-chord distance values obtained in modern-analog analysis and ratio of maximum a$nity scores to total possible a$nity scores.
3.4. Pollen-inferred paleoclimate Eight of the LGM pollen samples were identi"ed as lacking modern analogs within the response surface analysis and not included in the reconstructions. These included two from Anderson Pond, all three from Wolf Creek, one from Powers Fort Swale, and samples from two of the three Tunica Hills sites (Pinckneyville Creek 20 and Tunica Bayou 23), leaving 102 samples from 18 sites. Except for the Anderson Pond samples, all of the samples that were eliminated were from regions of noanalog pollen assemblages (Figs. 5 and 7). Because we had no basis for assigning Pinus pollen at Sheelar Lake to either northern or southern groups, we assumed it was 50% of each type. The paleoclimate inferences did not di!er appreciably from analyses in which we estimated the proportion of each type based on relative distances from Sheelar to Lake Tulane (100% southern) and Bob Black Pond (100% northern). An analysis in which we assumed 100% southern Pinus at Sheelar yielded temperature estimates 2}43C higher than at Lake Tulane, which is ca. 300 km to the south.
Our new analyses (Fig. 7) suggest that eastern North America was colder during the LGM than indicated by previous reconstructions (e.g., Bartlein et al., 1998; T. Webb et al., 1993, 1998). LGM January temperatures inferred from our data set were 15}203C colder than modern and up to 103C colder than previous estimates. LGM July temperatures were up to 103C colder than modern and slightly cooler than previous results. Inferred LGM mean annual precipitation was as much as 40 mm less than modern and 10}20 mm less than previous results. However, our results show less precipitation immediately south of the Laurentide Ice Sheet than indicated by Webb et al. (1993, 1998). The substantial drop in inferred winter temperatures, together with the modest drop in inferred summer temperatures, can be attributed in part to in#uence of new data from the Lower Mississippi Valley indicating high abundances of Picea. The drier conditions inferred along the southern margin of the Laurentide ice sheet may result from the elimination of the no-analog Wolf Creek assemblages. The reliability of our LGM paleoclimate reconstructions probably varies regionally. In particular, inferred
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
501
Fig. 7. Maps of LGM inferred climate and climate anomalies (di!erence from modern climate). Gray shaded areas represent regions of no-analog pollen assemblages.
temperatures may be too cold in regions distant from the ice margin, especially the Lower Mississippi Valley, where pollen assemblages lack modern analogs. The response-surface paleoclimate reconstructions are dominated by an extinct species of Picea in the fossil pollen assemblages. Although we did not apply formal macrofossil-based constraints to our pollen-based inferences (e.g., Huntley, 1993, 1994), the macrofossil data (Fig. 3) suggest overestimation of LGM cooling. Presence/absence climate-response surfaces indicate that occurrence of such temperate and cool-temperate taxa as Quercus spp., Carya spp., Juglans nigra, Liriodendron tulipifera, Ulmus americana, Carpinus caroliniana, Fagus grandifolia, and Acer spp. at the Tunica Hills and Nonconnah Creek sites is unlikely at the inferred January temperatures of (!153C and July temperatures of (15}183C (J.T.
Overpeck, S.T. Jackson, K.H. Anderson, unpublished). We regard the pollen-based paleoclimate inferences for the regions along the ice-margin, the Atlantic Coastal Plain, and interior highlands as more accurate. Extant species of Picea are documented in these regions, and analog-matches with modern vegetation are reasonably close (Figs. 3 and 5, 6).
4. Discussion 4.1. Phytogeography of the Last Glacial Maximum in eastern North America The now-classic debate between Braun (1950, 1955) and Deevey (1949) centered on the magnitude of
502
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
southward displacement of boreal and temperate taxa during the Last Glacial Maximum and the antiquity of certain plant communities. Fossil evidence was meager before the 1960s, and Braun was able to accomodate the existing pollen and plant-macrofossil evidence within her model of modest to no biogeographic change south of the ice margin. Improved paleoecological methods and the discovery of new fossil sites led to apparent resolution of the debate in Deevey's favor. Whitehead (1965, 1973), Watts (1980, 1983), and Wright (1981) all inferred extensive forests of Picea and Pinus banksiana on the Atlantic and Gulf Coastal Plains and Piedmont. Delcourt and Delcourt (1981, 1993) mapped more complex patterns, with boreal forests extending south to 333N, and most of the Piedmont and Coastal Plain occupied by temperate deciduous forests of Quercus, Carya, and southern Pinus. The deciduous forests, however, were inferred from a single site (Goshen Springs, AL) where the sedimentary record indicates a hiatus spanning the LGM (P.A. Delcourt, 1978a; Watts, 1983; Webb, 1984). Delcourt and Delcourt (1981, 1993) mapped Picea forest in the Lower Mississippi Valley, adopting Braun's (1950) explanation that Picea glauca grew there owing to locally cool climate related to glacial meltwater. Our synthesis of available LGM data, which incorporates the most recent available data as well as a rigorous selection of sites aimed at maximizing data reliability and quality, is generally consistent with biogeographic and vegetational patterns inferred from paleoecological studies for the region north of 343N. Open boreal woodlands and/or tundra grew along the ice margin, with tundra particularly widespread in the continental interior (Upper Mississippi Valley). Vegetation of the interior and coastal regions south of the ice sheet was dominated by boreal or cool-temperate conifers. Boreal (Picea glauca, Pinus banksiana) and cool-temperate conifers (Pinus resinosa, P. strobus) grew at least as far south as 343N (NW Georgia), and may have extended as far south as the Gulf Coast of Florida. Vegetation of the Lower Mississippi Valley was dominated by a now-extinct species of Picea (P. critchxeldii), which coexisted with cool-temperate and temperate deciduous trees as far north as 353N (Nonconnah Creek). The only LGM records of southern Pinus species are from peninsular Florida (Fig. 4). Substantial uncertainties remain in our understanding of the biogeography of the Last Glacial Maximum. Large geographic gaps in the data network constrain interpretation. For example, the absence of well-dated LGM sites between 303N and 333N and east of 913W in the Atlantic and Gulf Coastal Plains prevents determination of the southernmost extent of boreal and cool-temperate conifers and the northernmost extent of warm-temperate Pinus spp. The geographic ranges of the extant and extinct Picea species need to be assessed by additional macrofossil studies. Did P. critchxeldii occur in the Lower Mississippi Valley north of 353N? How far south
did P. glauca and P. mariana occur? The Picea-dominated pollen assemblages in the continental interior are all from sites in or adjacent to the Lower Mississippi Valley. How far east and west of the Lower Mississippi Valley region did these Picea-dominated forests extend? Was the eastward transition to Pinus-dominated forests abrupt or gradual? The LGM ranges of many important boreal and cooltemperate conifers (e.g., Abies balsamea, Tsuga canadensis, Pinus strobus) and hardwoods (Betula papyrifera, B. allegheniensis, Acer saccharum, Tilia) are unknown. Sporadic macrofossil occurrences of P. strobus and B. papyrifera in LGM sediments of Bob Black Pond (NW Georgia) are associated with only trace amounts of these taxa in pollen assemblages (S.T. Jackson, unpublished). These and other taxa must have occurred in the Pinus banksiana/resinosa-dominated forests as di!use, local populations and hence were below pollen-percentage detection thresholds owing to dilution by Pinus Subgenus Pinus pollen. Similarly, LGM occurrences of most of the temperate hardwoods (Quercus, Carya, Liriodendron, Fagus, Juglans) and conifers (Pinus spp.) of eastern North America are undocumented except for the Lower Mississippi Valley sites (hardwoods) and peninsular Florida (conifers). Macrofossil occurrences of hardwood species are associated with low pollen percentages at the Lower Mississippi Valley sites, where pollen and macrofossil assemblages are dominated by Picea (Figs. 2 and 3). Picea evidently dominated forests in this region. We cannot yet determine whether the hardwood taxa were restricted to the Lower Mississippi Valley region or occurred more widely in scattered, small populations. Hardwoods may have been widespread east of the Mississippi Valley, but were below detection thresholds of the pollen-sensing system owing to abundant representation of northern Pinus. Additional pollen and macrofossil studies are needed to identify past occurrences of these and other taxa, which will be valuable both in constraining paleoclimatic inferences and in understanding the biogeography and postglacial migration dynamics of these taxa. Our data synthesis supports earlier conclusions that vegetation south of the LGM ice margin consisted of coniferous and mixed forests and woodlands dominated by boreal and cool-temperate conifers (Whitehead, 1965, 1973; Watts, 1980, 1983). The macrofossil data indicate substantial southward displacements of such taxa as Picea glauca, Pinus banksiana, P. resinosa, and Betula papyrifera. However, the existing paleoecological data tell us little or nothing about the extent of geographic displacement of most of the current southeastern #ora and fauna. We have no direct evidence from which to determine whether, for instance, the southern Appalachian endemic Magnolia fraseri (Little, 1970) occupied its present range, was restricted to a handful of scattered microhabitats within its present range, was
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
displaced 10 km south to the northern Piedmont, or was forced 10 km south into central Florida during the LGM and earlier glacial intervals. Geographic responses of plant species to climatic changes during the Late Quaternary have been individualistic, and complex patterns of change are well-documented for the Holocene (Davis, 1976; Webb, 1988; Jackson et al., 1997). We postulate that during the LGM and earlier glacial periods, the elements of the southeastern #ora displayed a broad spectrum of patterns, with some species occupying nearmodern ranges, others displaced far south of present ranges (to the Gulf Coast, Florida Peninsula, or even the West Indies), and many in intermediate positions. We observe that at least some temperate species grew locally at least as far north as 353N, only 500 km south of the ice margin. 4.2. Vegetation of eastern North America during the Last Glacial Maximum The analog and biomization techniques emphasize different aspects of fossil pollen data, and draw out di!erent features. The analog technique assumes that fossil pollen assemblages that match modern assemblages within a speci"ed range of SCD represent vegetation assignable to a modern vegetation type associated with the closest match. Fossil samples that do not match any modern assemblages within the speci"ed SCD are assumed to represent vegetation su$ciently di!erent in composition and/or structure as to represent a distinct, now-`extincta vegetation type. The analog technique emphasizes matches in terms of #oristic compositional similarity. The biomization approach, in contrast, converts pollen taxa to plant functional types (PFTs) (de"ned in terms of physiognomic and other structural features as well as assumed ecophysiological responses), and assigns biomes based on calculated a$nities of the PFTs comprising a pollen assemblage to existing biomes (Prentice et al., 1996; Williams et al., 1998, 1999). The biomes are broadly de"ned as assemblages of plant functional types characterized by physiognomic properties (vegetation structure, deciduousness, etc.) and climatic tolerances. Our analog results indicate extensive areas of vegetation lacking modern #oristic analogs at LGM. The noanalog regions are concentrated in the continental interior and southern Florida. The lack of modern analogs is reinforced, and perhaps explained in part, by the occurrence in much of the interior region of a now-extinct Picea species which was a major component of the vegetation (Jackson and Weng, 1999). Its association with temperate deciduous trees, and the absence of extant boreal taxa in pollen and macrofossil assemblages in much of the region, emphasize the peculiar character of the vegetation. The extensive no-analog regions indicate caution in paleoclimate inference and subjective, `narrativea mapping of vegetation.
503
The analog technique is useful in identifying regions of no-analog vegetation composition, but provides no further information on the nature of vegetation in those regions. The biomization approach is insensitive to absence of #oristic analogs, and hence can provide information on vegetation structure. The no-analog region in peninsular Florida is classi"ed as open conifer woodland, a biome that is now restricted to semiarid western North America (e.g., open woodlands dominated by Pinus and/or Juniperus spp.). The classi"cation of southern Florida and part of the Atlantic Coastal Plain emphasizes the open structure of the Pinus-dominated vegetation of the region at LGM. The no-analog region in the continental interior is classi"ed primarily as cool mixed forest. This is consistent with the coexistence of Picea critchxeldii with various temperate hardwood species (especially Quercus, Carya, Juglans, and Fagus) in the Lower Mississippi Valley. The dominance of Picea and the lack of modern analogs emphasize the di!erences between this vegetation and modern cool mixed forests of eastern North America, which are characterized by cooccurrence of Tsuga, Pinus and Picea with Acer, Betula, Fagus, and Fraxinus. The tundra biome inferred for Texas is clearly a mismatch and may represent a lack of resolving power of the biomization technique for assemblages dominated by taxa assignable to many PFTs and/or biomes (e.g., Cyperaceae, Poaceae, Asteraceae, Alnus). The strength of a$nity for that region is relatively weak (Fig. 6). Analog and biomization approaches can be viewed as complementary, each with corresponding advantages and liabilities. Analog techniques can #ag vegetation lacking modern analogs and provide taxonomic resolving power among #oristically di!erent but physiognomically similar vegetation types (e.g., conifer and mixed forest). Biome techniques can provide clues to the character of no-analog vegetation as well as yield independent corroboration of analog matches. We advocate their combined use in paleovegetational studies. The results of our pollen-based analog and biome analyses are generally consistent with the independent evidence from macrofossils. The macrofossils indicate boreal trees (Picea) along the ice sheet in the Ohio Valley, and open tundra or parkland in Iowa and Minnesota. Conifers, especially Pinus banksiana and Picea, dominate macrofossil assemblages in the interior highlands, consistent with the analog and biome assignments of coniferdominated forests. The Picea-dominated macrofossil assemblages of the Lower Mississippi Valley, together with the occurrence of cool-temperate hardwoods, support the biome assigment of this region to cool mixed forest. Macrofossil studies from the Atlantic Coastal Plain are few. However, LGM sediments at Clear Pond lack macrofossils of Pinus and other tree taxa (G. L. Jacobson, Jr., personal communication). Absence of arboreal macrofossils is consistent with our biome inference of open conifer
504
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
woodlands, and may indicate that trees of northern Pinus, which dominate the pollen assemblages, were scattered or even distant from the site. 4.3. Climate of eastern North America during the Last Glacial Maximum Comparison of pollen-inferred paleoclimates with paleoclimates simulated from AGCMs have indicated substantial data/model di!erences for the Last Glacial Maximum in eastern North America (COHMAP, 1988; Webb et al., 1993). Similar to previous data compilations (Webb et al., 1993, 1998), our LGM paleoclimate reconstructions indicate substantially cooler LGM conditions in eastern North America than simulated by either the CCM0 or CCM1 models (Kutzbach et al., 1993, 1998). The relatively warm conditions simulated by CCM0 most likely arose from unrealistically high LGM seasurface temperature prescriptions (Guilderson et al., 1994; Rosell-MeleH et al., 1998). In an analysis of CCM1, Bartlein et al. (1998) suggested that the anomalously warm simulated LGM temperatures in the southeastern United States can be explained in terms of spatial resolution and model-speci"c processes impacting circulation, advection, and local surface-energy and moisture-balance processes. However, our LGM temperature reconstructions for the region are similar to those reported by Webb et al. (1997) for a NASA-GISS LGM simulation using modern ocean heat transports. E!orts to deduce the LGM vegetation of eastern North America using LGM paleoclimate simulations (CCM0 and CCM1) coupled with pollen-climate response surfaces yield simulations of latitudinally aligned Picea and northern Pinus maxima south of the ice sheet and terminating north of 353N, with all areas to the south characterized by high simulated percentages of Quercus and southern Pinus (Webb and Bartlein, 1988; Webb et al., 1998). These LGM simulations are not consistent with the documented occurrences of Picea and northern Pinus as far south as 33}343N in the Coastal Plain and Piedmont, nor with the extensive observed Picea populations in the Lower Mississippi Valley between 353 and 313N. Although the latter case involves a now-extinct species, all extant Picea species in North America, Europe, and Asia are cool-temperate to boreal/montane, and imply temperatures cooler than those simulated for the region. Our pollen-inferred paleoclimate estimates di!er from those of Webb et al. (1993, 1998), who used similar methods. The di!erences derive entirely from the improvements and additions we made to the LGM pollen data set. Our data set indicates substantially lower January temperatures and moderately lower July temperatures in all parts of eastern North America except peninsular Florida. Precipitation estimates are similar for the eastern part of the study region, but our data set
shows more complex patterns in the continental interior, with low precipitation near the ice margin and high precipitation in and west of the Lower Mississippi Valley. Our January and July temperature maps show particularly low temperatures in the Lower Mississippi Valley, with January temperatures (!163C and July temperatures (143C. Our paleoclimate inferences may overestimate the extent of LGM cooling, particularly in the Lower Mississippi Valley and eastern interior highlands. In the former region, the Picea-dominated pollen assemblages dictate low inferred temperatures because comparably high Picea pollen percentages today are associated with low temperatures for July (9}153C) and January (\30}\103C) (see Figs. 17.8 and 17.9 in Webb et al., 1993). These high percentages occur in regions of boreal forest and taiga dominated by Picea glauca and P. mariana. Picea critchxeldii, which contributed all or most of the high Picea pollen percentages in the Lower Mississippi Valley during the LGM, was probably associated with warmer temperatures. The occurrence in the macrofossil assemblages of temperate and cool-temperate taxa (Quercus, Carya, Fagus, Liriodendron, Juglans, Carpinus, Ulmus) indicates July temperatures '15}183C and January temperatures '!153C (J. T. Overpeck, S. T. Jackson, K. H. Anderson, unpublished), suggesting temperature anomalies of ca. \5}\103C relative to modern conditions. These more modest anomalies are consistent with recent paleotemperature estimates from noble gases in LGM groundwater in central Texas (Stute et al., 1992), and late-glacial insect assemblages in northeastern Texas (Elias, 1997), and with northern Gulf of Mexico sea-surface temperature estimates from alkenone data (Jasper and Gagosian, 1989; Rosell-MeleH et al., 1998). Inferred LGM temperatures may also be too low in the eastern part of the study area. Because of the bimodal pattern of modern Pinus pollen percentages in eastern North America (Bartlein et al., 1986), high Pinus pollen percentages are associated with either dominance of boreal Pinus banksiana (with varying contribution of cool-temperate P. resinosa and P. strobus) or P. taeda and other warm-temperate species. Accordingly, high Pinus pollen percentages force either very warm or very cool temperatures. The occurrence of cool-temperate conifers (Pinus resinosa, P. strobus, Picea critchxeldii?) in LGM macrofossil assemblages at Bob Black Pond in NW Georgia may indicate warmer temperatures than implied by the pollen assemblages at that site, which are dominated by northern Pinus and Picea. Our methods for paleoclimate reconstruction also do not take account of e!ects of LGM atmospheric CO concentrations, which were approximately 65% (180 mol/mol) of pre-industrial concentrations (Barnola et al., 1987). Low CO concentrations at LGM would have had direct physiological e!ects on plants (Polley
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
505
et al., 1993, 1995; Johnson et al., 1993), and may have altered plant responses to climate in complex ways (Austin, 1992; Cowling and Sage, 1998; Saxe et al., 1998). In particular, decreased water-use e$ciency of plants owing to low CO during LGM may be respon sible for the low precipitation inferred in our climate reconstructions, and low CO may also account for the open nature of vegetation in much of the region (e.g., Cowling, 1999). E!ects of lowered CO on temperature reconstruc tions are more di$cult to assess. Dominance of conifers in forests and woodlands throughout the region could be attributable in part to lowered CO e!ects. Conifer leaves have a number of adaptations that increase carbon-use e$ciency, including evergreenness, longevity (up to 12 or more years), and mechanical strength (which reduces loss), and have several features (physiological cold-tolerance, needle clustering, prevention of low-temperature photoinhibition) that expand the circumstances under which net photosynthesis can occur (Smith and Brewer, 1994; Reich et al., 1995; Smith et al., 1997; Eckstein et al., 1999). These and related traits may have conferred competitive advantages on conifers relative to deciduous broadleaf trees during the LGM. Drought tolerance of conifers may have also provided advantages in terms of water-use e$ciency in a CO -limited environ ment. However, dominant conifers on most of the landscape of unglaciated eastern North America were boreal and cool-temperate species of Pinus and Picea; warmtemperate conifers were apparently restricted to Florida and perhaps the Gulf coast. These patterns cannot be explained without substantial temperature lowering, consistent with paleotemperature inferences from other data (Stute et al., 1992; Elias, 1997).
gion, thus making exact paleoclimatic inferences di$cult for a substantial part of eastern North America for the LGM. Reconstruction is further complicated by likely ecophysiological e!ects of lowered LGM CO concen trations on plants. The dissimilarity between LGM and modern vegetation composition (indicated by the extensive areas of pollen assemblages lacking modern analogs) is attributable to unique climatic conditions during the LGM, with lowered CO concentrations probably play ing an important role as well. Rates of vegetational and climatic change were relatively low during the LGM interval, peaking during the Lateglacial (Overpeck et al., 1991, Overpeck et al., 1992). Understanding of the biogeographic patterns, vegetational structure, climatic gradients, and environmental mechanisms of the peculiar `LGM worlda is critical to assessing climate and biotic sensitivity to various forcings. We need more high-quality data, which requires discovery of new sites as well as restudy of others. Application of response-surface models to plant macrofossil data may help re"ne paleoclimate inferences. We have little indication of the nature and roles of disturbance, particularly "re, in structuring LGM vegetation; the only stratigraphic charcoal records (Lake Tulane, FL and Clear Pond, SC) show low LGM charcoal in#ux (Watts and Hansen, 1988; Hussey, 1993). Charcoal records are needed from sites further inland. E!ects of lowered CO concentrations on vegetation physiognomy and composition need to be investigated. We foresee a greater role of modeling in assessing potential forcing mechanisms, both climatic (e.g., Webb et al., 1997) and ecological (e.g., Jolly and Haxeltine, 1997; Cowling, 1999).
5. Conclusions
This research was supported by the National Science Foundation (Climate Dynamics, Ecology, and Earth System History Programs) and the National Oceanic and Atmospheric Administration (Paleoclimatology Program). We thank George Jacobson and Don Whitehead for providing pollen-size data, Chengyu Weng, Jennifer Kearsley, and Darren (Paco) R. Van Sistine for technical assistance, William K. Smith and Robert S. Thompson for discussion, and Brian Huntley and William A. Watts for critical and helpful reviews.
Our data compilation and analyses indicate substantial biogeographic displacement and climatic cooling during the Last Glacial Maximum in eastern North America. The occurrence of Pinus banksiana, P. resinosa, and Picea glauca as far south as 343N implies a higher degree of LGM cooling than simulated by AGCMs. Similarly, the occurrence of extensive Picea-dominated forest in the continental interior as far south as 313N indicates considerable cooling relative to today. However, the macrofossil occurrences of cool-temperate hardwoods in the Lower Mississippi Valley as far north as 353N and cool-temperate conifers (Pinus resinosa, P. strobus) as far north as 343N in Georgia indicate that conditions were not as severe as implied by the pollen assemblages alone, which are dominated by pollen types (Picea, Pinus banksiana) that reach maxima today in boreal regions of eastern North America. The pollen assemblages lack modern analogs over much of the re-
Acknowledgements
References Adams, C.C., 1902. Postglacial origin and migrations of the life of the northeastern United States. Journal of Geography 1, 303}310, 352}357. Adams, C.C., 1905. The postglacial dispersal of the North American biota. Biological Bulletin 9, 53}71. Austin, M.P., 1992. Modeling the environmental niche of plants: implications for plant community response to elevated CO levels. Australian Journal of Botany 40, 615}630.
506
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
Baker, R.G., Rhodes II, R.S., Schwert, D.P., Ashworth, A.C., Frest, T.J., Hallberg, G.R., Janssens, J.A., 1986. A full-glacial biota from southeastern Iowa, USA. Journal of Quaternary Science 1, 91}107. Bard, E., Hamelin, B., Fairbanks, R.G., 1990. U-Th ages obtained by mass spectrometry in corals from Barbados: sea level during the last 130,000 years. Nature 346, 456}458. Barnola, J.M., Raynaud, D., Korotkevich, Y.S., Lorius, C., 1987. Vostok ice core provides 160,000 year-record of atmospheric CO . Nature 329, 408}414. Bartlein, P.J., Prentice, I.C., Webb, T.III, 1986. Climatic response surfaces from pollen data for some eastern North American taxa. Journal of Biogeography 13, 35}57. Bartlein, P.J., Anderson, K.H., Anderson, P.M., Edwards, M.E., Mock, C.J., Thompson, R.S., Webb, R.S., Webb III, T., Whitlock, C., 1998. Paleoclimate simulations for North America over the past 21,000 years: features of the simulated climate and comparisons with paleoenvironmental data. Quaternary Science Reviews 17, 535}548. Birks, H.J.B., 1976. Late-Wisconsinan vegetational history at Wolf Creek, central Minnesota. Ecological Monographs 46, 395}429. Braun, E.L., 1950. Deciduous Forests of Eastern North America. Macmillan, New York. Braun, E.L., 1955. The phytogeography of unglaciated eastern United States and its interpretation. Botanical Review 21, 297}375. Brown, J.H., Lomolino, M.V., 1998. `Biogeography (Second Edition)a. Sinauer, Sunderland. Burns, G.W., 1958. Wisconsin age forests in western Ohio. II. Vegetation and burial conditions. Ohio Journal of Science 58, 220}230. Cain, M.L., Damman, H., Muir, A., 1998. Seed dispersal and the Holocene migration of woodland herbs. Ecological Monographs 68, 325}347. Camper, H. A., 1991. `Pollen Analysis of Patschke Bog.a Unpublished M.S. thesis, Texas A&M University. Clark, J.S., Fastie, C., Hurtt, G., Jackson, S.T., Johnson, C., King, G.A., Lewis, M., Lynch, J., Pacala, S., Prentice, C., Schupp, E.W., Webb III, T., Wycko!, P., 1998. Reid's paradox of plant migration. BioScience 48, 13}24. Cleveland, W.S., 1994. `The Elements of Graphing Dataa. Revised Edition. Hobart Press, Summit, NJ. COHMAP Members, 1988. Climatic changes of the last 18,000 years: observations and model simulations. Science 241, 1043}1052. Cotter, J.F.P., 1984. The minimum age of the woodfordian deglaciation of Northeastern Pennsylvania and Northwestern New Jersey. Unpublished Ph.D. Thesis. Lehigh University, Bethlehem. Cowling, S. A., 1999. Simulated e!ects of low atmospheric CO on structure and composition of North American vegetation at the Last Glacial Maximum. Global Ecology and Biogeography Letters, in press. Cowling, S.A., Sage, R.F., 1998. Interactive e!ects of low atmospheric CO , and elevated temperature on growth, photosynthesis, and respiration in Phaesolus vulgaris. Plant, Cell and Environment 21, 427}435. Davis, M.B., 1976. Pleistocene biogeography of temperate deciduous forests. Geoscience and Man 13, 13}26. Deevey, E.S., 1949. Biogeography of the Pleistocene. Part I. Europe and North America. Geological Society of America Bulletin 60, 1315}1416. Delcourt, H.R., 1978a. Late quaternary vegetation history of the Eastern Highland Rim and adjacent Cumberland Plateau of Tennessee. Unpublished Ph.D. Thesis. University of Minnesota, Minneapolis. Delcourt, H.R., 1979. Late quaternary vegetation history of the eastern Highland Rim and adjacent Cumberland Plateau of Tennessee. Ecological Monographs 49, 255}280. Delcourt, P.A., 1978b. Quaternary vegetation history of the Gulf Coastal Plain. Unpublished Ph.D. Thesis. University of Minnesota. Delcourt, P.A., Delcourt, H.R., Brister, R.C., Lackey, L.E., 1980. Quaternary vegetation history of the Mississippi Embayment. Quaternary Research 13, 111}132.
Delcourt, P.A., Delcourt, H.R., 1981. Vegetation maps for eastern North America: 40,000 yr BP to the present. In: Romans, R.C. (Ed.), `Geobotany IIa. Plenum, New York, pp. 123}165. Delcourt, P.A., Delcourt, H.R., 1993. Paleoclimates, paleovegetation, and paleo#oras during the Late Quaternary. In: Flora of North America Editorial Committee (Eds.), Flora of North America. Vol. 1. Introduction. Oxford University Press, New York, pp. 71}94. Eckstein, R.L., Karlsson, P.S., Weih, M., 1999. Leaf life span and nutrient resorption as determinants of plant nutrient conservation in temperate-arctic regions. New Phytologist 143, 177}189. Elias, S.A., 1997. Paleoenvironmental signi"cance of late glacial insect fossil assemblages from the Aubrey Site, north-central Texas, USA. Studies in Quaternary Entomology * An Inordinate Fondness for Insects. Quaternary Proceedings Number 5, 105}111. Givens, C.R., Givens, F.M., 1987. Age and signi"cance of fossil white spruce (Picea glauca), Tunica Hills, Louisiana-Mississippi. Quaternary Research 27, 283}296. Goldthwait, R.P., 1958. Wisconsin age forests in western Ohio. I. Age and glacial events. Ohio Journal of Science 58, 209}219. Gray, A., 1884. Characteristics of the North American #ora. American Journal of Science 128, 323}340. Grimm, E.C., Jacobson Jr., G.L., Watts, W.A., Hansen, B.C.S., Maasch, K.A., 1993. A 50,000-year record of climate oscillations from Florida and its temporal correlation with the Heinrich events. Science 261, 198}200. Guilderson, T.P., Fairbanks, R.G., Rubenstone, J.L., 1994. Tropical temperature variations since 20,000 years ago: modulating interhemispheric climate changes. Science 263, 663}665. Hall, S.A., Valastro Jr., S., 1995. Grassland vegetation in the southern Great Plains during the Last Glacial Maximum. Quaternary Research 44, 237}245. Hughen, K.A., Overpeck, J.T., Lehman, S.J., Kashgarian, M., Southon, J., Peterson, L.C., Alley, R., Sigman, D.M., 1998. Deglacial changes in ocean circulation from an extended radiocarbon calibration. Nature 391, 65}68. Huntley, B., 1993. The use of climate response surfaces to reconstruct paleoclimate from Quaternary pollen and plant macrofossil data. Philosophical Transactions of the Royal Society of London 341, 215}224. Huntley, B., 1994. Late Devensian and Holocene palaeoecology and palaeoenvironments of the Morrone Birkwoods, Aberdeenshire, Scotland. Journal of Quaternary Science 9, 311}336. Hussey, T.C., 1993. A 20,000-year record of vegetation and climate at clear pond, northeastern South Carolina. Unpublished M.S. Thesis. University of Maine. Jackson, S.T., Givens, C.R., 1994. Late Wisconsinan vegetation and environment of the Tunica Hills region. Louisiana/Mississippi. Quaternary Research 41, 316}325. Jackson, S.T., Overpeck, J.T., Webb, T.III, Keattch, S.E., Anderson, K.H., 1997. Mapped plant-macrofossil and pollen records of Late Quaternary vegetation change in eastern North America. Quaternary Science Reviews 16, 1}70. Jackson, S. T., Weng, C., 1999. Late Quaternary extinction of a tree species in eastern North America. Proceedings of the National Academy of Sciences USA, 96: 13 847}13 852. Jackson, S.T., Whitehead, D.R., 1993. Pollen and plant macrofossils from Wisconsinan interstadial sediments in northeastern Georgia. Quaternary Research 39, 99}106. Jasper, J.P., Gagosian, R.B., 1989. Alkenone molecular stratigraphy in an oceanic environment a!ected by glacial freshwater events. Paleoceanography 4, 603}614. Jaumann, P.J., 1989. Evidence for Late Quaternary Boreal Environments in the Arkansas River Valley, South-central Kansas. Theoretical Aspects of Paleoecology and Climate Inferences. Unpublished M.Sc. Thesis. University of Kansas.
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508 Johnson, H.B., Polley, H.W., Mayeux, H.S., 1993. Increasing CO and plant-plant interactions: e!ects on natural vegetation. Vegetatio 104/105, 157}170. Jolly, D., Haxeltine, A., 1997. E!ect of low glacial atmospheric CO on tropical African montane vegetation. Science 276, 786}788. Joussaume, S., Taylor, K., 1995. Status of the palaeoclimate modeling intercomparison project (PMIP). In: Proceedings of the First International AMIP Scienti"c Conference. World Meteorology Organization, Geneva, pp. 425}430. Kneller, M., Peteet, D.M., 1993. Late-quaternary climate in the Ridge and Valley of Virginia, U.S.: changes in vegetation and depositional environment. Quaternary Science Reviews 12, 613}628. KuK chler, A.W., 1964. Potential Natural Vegetation of the Conterminous United States. American Geographical Society Special Publication 36, New York. Kutzbach, J.E., Guetter, P.J., Behling, P.J., Selin, R., 1993. Simulated climatic changes: results of the COHMAP climate-model experiments. In: Wright Jr., H.E., Kutzbach, J.E., Webb, T.III, Ruddiman, W.F., Street-Perrott, F.A., Bartlein, P.J. (Eds.), Global Climates Since the Last Glacial Maximum. University of Minnesota Press, Minneapolis, pp. 24}93. Kutzbach, J.E., Gallimore, R., Harrison, S., Behling, P., Selin, R., Laarif, F., 1998. Climate and biome simulations for the past 21,000 years. Quaternary Science Reviews 17, 473}506. Larabee, P.A., 1986. Late-quaternary vegetational and geomorphic history of the allegheny plateau at big run bog, Tucker county, West Virginia. Unpublished Master's Thesis. University of Tennessee. Little Jr., E.L., 1970. Endemic, disjunct and northern trees in the southern Appalachians. In: Holt, P.C. (Ed.), The Distributional History of the Biota of the Southern Appalachians. Part II: Flora. Virginia Poytechnic Institute and State University Research Division Monograph 2, Blacksburg, pp. 249}290. Little, E. L., Jr., 1971. Atlas of United States Trees, Vol. 1. Conifers and Important Hardwoods. United States Department of Agriculture, Forest Service Miscellaneous Publication Number 1146. Miller, N.G., 1992. A contribution toward a history of the Arctic moss #ora. Contributions of the University of Michigan Herbarium 18, 73}86. Overpeck, J.T., Webb III, T., Prentice, I.C., 1985. Quantitative interpretation of fossil pollen spectra: dissimilarity coe$cients and the method of modern analogs. Quaternary Research 23, 87}108. Overpeck, J.T., Webb, R.S., Webb, T. III., 1992. Mapping eastern North American vegetation changes of the past 18 ka: no-analogs and the future. Geology 20, 1071}1074. Peterson, G.M., Webb III, T., Kutzbach, J.E., van der Hammen, T., Wijmstra, T.A., Street, F.A., 1979. The continental record of environmental conditions at 18,000 yr BP: an initial evaluation. Quaternary Research 12, 47}82. Polley, H.W., Johnson, H.B., Marion, B.D., Mayeux, H.S., 1993. Increase in C3 plant water-use e$ciency and biomass over Glacil to present CO concentrations. Nature 361, 61}64. Polley, H.W., Johnson, H.B., Mayeux, H.S., 1995. Nitrogen and water requirements of C plants grown at glacial to present carbon dioxide concentrations. Functional Ecology 9, 86}96. Prentice, I.C., Guiot, J., Huntley, B., Jolly, D., Cheddadi, R., 1996. Reconstructing biomes from palaeoecological data: a general method and its application to European pollen data at 0 and 6 ka. Climate Dynamics 12, 185}194. Reich, P.B., Koike, T., Gower, S.T., Schoettle, A.W., 1995. Causes and consequences of variation in conifer leaf life-span. In: Smith, W.K., Hinckley, T.M. (Eds.), Ecophysiology of Coniferous Forests. Academic Press, San Diego, pp. 225}254. Rosell-MeleH , A., Bard, E., Emeis, K.-C., Farrimond, P., Grimalt, J., MuK ller, P.J., Schneider, R.R., 1998. Project takes a new look at past sea surface temperatures. EOS Transactions AGU 79 (33), 393}394.
507
Rowe, J.S., 1972. Forest Regions of Canada. Department of the Environment, Canadian Forestry Service Publication Number 1300, Ottawa. Royall, P.D., Delcourt, P.A., Delcourt, H.R., 1991. Late Quaternary paleoecology and paleoenvironments of the Central Mississippi Alluvial Valley. Geological Society of America Bulletin 103, 157}170. Saxe, H., Ellsworth, D.S., Heath, J., 1998. Tree and forest functioning in an enriched CO atmosphere. New Phytologist 139, 395}436. Smith Jr., E.N., 1984. Late-quaternary vegetational history at Cupola Pond, Ozark National Scenic Riverways, Southeastern Missouri. Unpublished Master's Thesis. University of Tennessee. Smith, W.K., Brewer, C.A., 1994. The adaptive importance of shoot and crown architecture in conifer trees. American Naturalist 143, 165}169. Smith, W.K., Vogelmann, T.C., DeLucia, E.H., Bell, D.T., Shepherd, K.A., 1997. Leaf form and photosynthesis. BioScience 47, 785}793. Stute, M., Schlosser, P., Clark, J.F., Broecker, W.S., 1992. Paleotemperatures in the southwestern United States derived from noble gases in ground water. Science 256, 1000}1002. Watts, W.A., 1970. The full-glacial vegetation of northwestern Georgia. Ecology 51, 17}33. Watts, W.A., 1980. The late Quaternary vegetation history of southeastern United States. Annual Review of Ecology and Systematics 11, 387}409. Watts, W.A., 1983. Vegetational history of the eastern United States 25,000 to 10,000 years ago. In: Porter, S.C. (Ed.), Late-quaternary environments of the United States Vol. 1. The Late Pleistocene. University of Minnesota Press, Minneapolis, pp. 294}310. Watts, W.A., Hansen, B.C.S., 1988. Environments of Florida in the Late Wisconsin and Holocene. In: Purdy, B. (Ed.), Wet Site Archaeology. Telford Press, Caldwell, NJ. Watts, W.A., Hansen, B.C.S., Grimm, E.C., 1992. Camel Lake: a 40,000 year record of vegetational and forest history from northwest Florida. Ecology 73, 1056}1066. Watts, W.A., Stuiver, M., 1980. Late Wisconsin climate of northern Florida and the origin of species-rich deciduous forest. Science 210, 325}327. Webb, R.S., Webb, T.III., 1988. Rates of sediment accumulation in pollen cores from small lakes and mires of eastern North America. Quaternary Research 30, 284}297. Webb, R.S., Rind, D.H., Lehman, S.J., Healy, R.J., Sigman, D., 1997. In#uence of ocean heat transport on the climate of the Last Glacial Maximum. Nature 385, 695}699. Webb, T.III., 1984. Discussion of: Late Quaternary vegetational dynamics and community stability reconsidered. Quaternary Research 22, 262. Webb, T.III., 1988. Eastern North America. In: Huntley, B., Webb, T.III. (Eds.), Vegetation History. Kluwer, Dordrecht, pp. 385}414. Webb, T.III., Bartlein, P.J., Harrison, S., Anderson, K.H., 1993. Vegetation, lake levels, and climate in eastern North America for the past 18,000 years. In: Wright Jr., H.E., Kutzbach, J.E., Webb, T.III., Ruddiman, W.F., Street-Perrott, F.A., Bartlein, P.J. (Eds.), Global Climates Since the Last Glacial Maximum. University of Minnesota Press, Minneapolis, pp. 415}467. Webb III, T., Anderson, K.H., Bartlein, P.J., Webb, R.S., 1998. Late Quaternary climate change in eastern North America: a comparison of pollen-derived estimates with climate model results. Quaternary Science Reviews 17, 587}606. Webb III, T., Bartlein, P.J., 1988. Late Quaternary climatic change in eastern North America: the role of modeling experiments and empirical studies. Bu!alo Society of Natural Sciences Bulletin 33, 3}13. Webb III, T., Kutzbach, J.E., 1998. An introduction to late quaternary climates: data syntheses and model experiments. Quaternary Science Reviews 17, 465}472.
508
S.T. Jackson et al. / Quaternary Science Reviews 19 (2000) 489}508
Wells, P.V., Stewart, J.D., 1987. Cordilleran-boreal taiga and fauna on the central Great Plains of North America, 14,000}18,000 years ago. American Midland Naturalist 118, 94}106. Weng, C., 1998. Species di!erentiation of Picea in the Quaternary fossil record. Unpublished Ph.D. Thesis University of Wyomug haramie. North American spruce (Picea) needles based on morphological and anatomical characteristics. Canadian Journal of Botany, in review. Whitehead, D.R., 1964. Fossil pine pollen and full-glacial vegetation in southeastern North Carolina. Ecology 45, 767}777. Whitehead, D.R., 1965. Palynology and Pleistocene phytogeography of unglaciated eastern North America. In: Wright Jr., H.E., Frey, D.G. (Eds.), The Quaternary of the United States. Yale University Press, New Haven, pp. 417}432. Whitehead, D.R., 1973. Late-Wisconsin vegetational changes in unglaciated eastern North America. Quaternary Research 3, 613}621. Whitehead, D.R., 1981. Late-Pleistocene vegetational changes in northeastern North Carolina. Ecological Monographs 51, 451}471.
Wilkins, G.R., 1985. Late-quaternary vegetational history at Jackson Pond Larue County, Kentucky. Unpublished Master's Thesis. University of Tennessee. Wilkins, G.R., Delcourt, P.A., Delcourt, H.R., Harrison, F.W., Turner, M.R., 1991. Paleoecology of central Kentucky since the last glacial maximum. Quaternary Research 36, 224}239. Williams, J.W., Summers, R.L., Webb, T.III., 1998. Applying plant functional types to construct biome maps from eastern North American pollen data: comparisons with model results. Quaternary Science Reviews 17, 607}628. Williams, J. W., Webb, T., III, Richard, P. J. H., Newby, P., 1999. Late Quaternary biomes of Canada and the eastern United States. Journal of Biogeography, in press. Wright Jr., H.E., 1981. Vegetation east of the Rocky Mountains 18,000 years ago. Quaternary Research 15, 113}125.
