Last modified 2nd October 1997


NORTH AMERICA; comprising the USA, Canada, Greenland and 'Beringia'.

+ North & Central America 18,000 14C years ago

+ North & Central America 8,000 14C years ago

+ North & Central America 5,000 14C years ago and present-potential


Return to QEN Home Page

List of References (separate document)

Key to the vegetation classification system used in the atlas


Distribution of sites used towards reconstructing the LGM palaeovegetation distribution for North America

18-20,000 14C years ago (last glacial maximum or LGM).

Sea level. LGM coastlines on the map follow the -150m bathymetric contour. Extensive lakes were present in the cordillera of the western USA, forming the Lake Bonneville system. The lake extent for 18,000 years ago is taken here from the maps shown by Benson & Thompson (1987).

Ice Extent. The extent of ice is taken from the standard published maps of Denton & Hughes (1981) and Dyke & Prest (1986, 1986). Although the broad patterns of ice extent at 18,000 radiocarbon years ago are known, some details are still disputed amongst glaciologists.

Eastern USA. There is a relative abundance of pollen, plant macrofossil, sedimentological and zoological evidence for the eastern USA (e.g. pollen and lake level evidence summarised by Webb et al. 1993), from just south of the ice sheet down to Florida. However, this has not been enough to prevent uncertainty and disagreement over the vegetation distribution at the LGM. The pollen, plant macrofossil and zoological evidence has been summarised and used to produce preliminary vegetation maps by Delcourt & Delcourt (1987, 1990), Woodcock & Wells (1990), Overpeck et al. (1993) and Tallis (1990) (see below).

Southern edge of the ice sheet. Various pollen and zoological fossil sites have been found close to the fringe of the LGM ice sheet (Ritchie 1982). A narrow zone of dry tundra (in the east, for example at a site in eastern Pennsylvania 60 km from the ice sheet; Watts 1989) and moister tundra (towards the west) is thought likely to have existed along the southern rim of the Laurentide ice sheet, and there is direct pollen evidence for its existence in the north-east (Webb et al. 1993). From pollen and molluscan evidence it seems to have given way to the south to a 'park tundra' (= forest tundra & open boreal woodland?) dominated by Picea. within about 100-150 km of the ice margin (Ritchie 1982). Fossils of tundra mammals (e.g. Musk ox Ovibos, Reindeer Rangifer) have only been found close to the LGM ice sheet margins, and even then in association with typical boreal-forest rodents (Tallis 1990), again suggesting the existence of a relatively narrow tundra belt sensu stricto. The LGM summer temperature depression relative to the present would have been around 12°C, according to an interpretation of the plant and molluscan fossils by Woodcock & Wells (1990).

Boreal forest or woodland in the eastern USA. Further south through the eastern USA, pollen-bearing cores yield evidence of a boreal-type forest or woodland. This included such elements as Picea and Pinus banksiana, with Pinus being dominant to the east of the Appalachians and Picea to the west (Tallis 1990). Strangely, deciduous forest elements (e.g. Quercus, Ulmus) seem to have occurred at low abundance throughout this 'boreal' forest region, possibly due to the peculiar climates of that time (Delcourt & Delcourt 1990), or some effect of lower CO2 levels (C. Loehle pers. commun. 1998).

In the Appalachian mountains, pollen evidence of temperature lowering from a site 300 km south of the ice front indicates that tundra would have been present at the higher altitudes (above about 800m) (Delcourt & Delcourt 1990, Woodcock &Wells 1990).

Relatively open forest east of the Appalachians? Conditions generally seem to have been drier than present, with a more open forest cover. Jack Pine (Pinus banksiana) seems to have been the predominant tree species east of the Appalachian Mountains - turning up for instance in central Maryland and Delaware - indicating fairly dry conditions and perhaps a tendency towards open-ness of the forest canopy (citation of work by Denny et al., in Wells 1992 p.579). In north-western Georgia, in the southern Appalacian foothills, (85 deg.W, 34 deg.N) Brook & Nickmann (1996) find pollen evidence that after a relatively mesic interval around 26,000 y.a., during the LGM-late glacial there was pine vegetation with a large proportion of dry, open meadows.

Summarizing the general picture of cold and drier conditions for the southern Appallacian foothills region (at the Anderson Pond site), P. Delcourt has suggested a mean annual precipitation of 500mm (in contrast to present totals of well over 1000mm), a January overall mean temperature of -20 deg.C and a July overall mean of 20 deg.C, during the full glacial-LGM period. Brook & Nickmann (1996) suggest that similar conditions would have previled at their site in the north-west Georgia foothills.

Relatively xerophytic full-glacial plant communities are recorded from the Atlantic Coastal Plain at White Pond (34N, 80W) in South Carolina and Singletary Lake in North Carolina (Wells 1992 p.624). It seems that at these sites a mosaic of open pine-dominated forest with some spruce and prairie grassland existed in these areas between 22,000 and 13,500 years ago (Watts 1980). In Maryland (38N, 75W) pollen evidence indicates pine-birch barrens or spruce parkland dominating in the interval between 30,000 and 13,000 years ago (Wells 1992 p.612). The relative abundance of chenopods (e.g. Atriplex) and compositae (Artemisia, Ambrosia) characteristic of the present steppe regions, and a mammal fauna including grazing herbivores (e.g. Bison) is interpreted by Tallis (1990) to suggest that the LGM vegetation of parts of the central and eastern USA was parkland or open woodland, not dissimilar to the present-day Eurasian forest-steppe (Tallis 1990).

There is abundant sedimentological evidence from the coastal plain of the eastern USA, showing arid conditions with blowouts and drifting sand during the approximate interval of the LGM (reviewed by Wells 1992). Along a broad band of the coastal plain, from Maryland south to Georgia, abundant blowout features (the 'Carolina Bays') and stabilised dunes stretch as much as 200 km inland from the present coastline, to the edge of the Appalachian foothills, and further inland round to the southern edge of the Appalachians (Well 1992). The age of these features is broadly constrained by the radiocarbon dating of underlying deposits that preceded dune activity, and the time at which the dunes finally became stabilised. It seems that dune activity began more than 22,000 14C years ago and had mostly ended by around 13,000 14C years ago (though with very slight sand movement in the lower river valleys extending into the early Holocene). The sheer density and extent of these aeolian features implies that in many areas at the LGM, the coastal plain of the eastern USA had a sparse open vegetation, with a mosaic of pine and spruce woodlands and more open herbaceous and xerophytic vegetation. Wells (1992 p.624) suggests on the basis of the observed dune features between North Carolina and central Florida, that during the glacial maximum the annual precipitation must have declined to around one fifth to one third (i.e. 200-450mm) of the present values of 1000 to 1400 mm, if temperature change is not accounted for (a larger change must be assumed if the lower LGM temperature is taken into account).

The generally rather dry nature of the glacial climate across the eastern USA may also be reflected by the occurrence of a 'blue clay' layer (probably wind-deposited) in the Carolina Bays. See this link on the blue clays

Considering all the evidence from various sources, we suggest on our map reconstruction that a 'temperate woodland' or 'open boreal' and not a closed forest vegetation would be most appropriate for most of the eastern USA at the LGM. This is in contrast with the map reconstructions presented by Webb et al. and Overpeck et al., based on their extensive reviews of pollen evidence, that in the eastern USA, forest was fairly closed in character throughout. Theirs seems a very surprising conclusion, and it perhaps calls into question the effectiveness of the computer pollen-based vegetation reconstruction method which they are using.

Was there a temperate deciduous forest belt? The position of the southern boundary of the 'boreal' forest in the eastern USA is still a matter of some uncertainty, as is the extent to which deciduous trees dominated the vegetation on the Gulf and Atlantic coastal plain . As mentioned above, pollen records show that at least some forest dominated by the 'boreal' conifer spruce (Picea) was present in the coastal plain of south Carolina at 34 N, and this tree was also present in west-central Georgia; possibly due to the existence of an extinct temperate-zone ecotype of spruce. However, sites at about 33N in north-central Louisiana and Alabama show strong dominance by deciduous forest species (Woodcock & Wells 1990), with the Alabama site having a xeric oak-hickory woodland suggesting drier-than-present conditions.

Wells (1992, p.625-630) reviews the literature of scattered glacial-age inland dunes along what is presently the Gulf Coast. He notes that there is considerable evidence of severe aridity in the area of the south-eastern Gulf Coast at the LGM. In south-eastern Louisiana, dune deflation features and the occurrence of soil calcite (caliche) nodules and layers, might suggest an annual rainfall of only around 400-600mm (if there were no temperature change) for the last glacial period, in contrast with the present annual rainfall of around 1450-1600mm (Wells 1992). This would seem to indicate conditions too dry for any continuous forest cover. However, the pollen evidence for forest cover seems to contradict this.

Paul & Hazel Delcourt offer the following comments on the LGM-full glacial environment at the time on the Gulf Coastal Plain;

The climate would have been " seasonally more dry than today, but with a more winter-wet and summer-dry pattern. This is consistent with the glacial-age formation of Deweyville river terraces with 6 to 8 times modern-day discharge producing them. Mean-annual temperature was still warm-temperate and perhaps 2 degrees C cooler on average annually. Summer-dry conditions favor drought tolerant species of oaks and hickories at the expense of southern pines (which today respond positively to moisture during their growing season). The impact of lowered sea level was critical in lowering groundwater tables only in the Florida karst area. In northernmost Florida and the rest of the Gulf coastal plain, layers of clastic, interbedded shale and siltstone served to perch the water table, decoupling the rolling sandy coastal plain from the potential drawdown by lowered sea level. The bottom line is the variety of full-glacial habitats permitted a mixture of both xeric and mesic forest communities to persist, more mixed-mesic deciduous foresets in fire protected sites along river valleys and stands of temperate oaks and southern pines across sandy interfluves. Areally-extensive forests are mapped as mixed oaks and pines. " (N.B., a 2 deg.C lowering of mean annual temperature seems very conservative by comparison with other estimates from across the USA and at lower latitudes also; even tropical estimates suggest a 5 deg.C lowering at around the LGM). On the basis of the pollen evidence, and from molluscan evidence, Woodcock & Wells (1990) drew a north-south boundary between boreal and temperate deciduous forest at about 33N for the LGM, apparently rather to the north of the boundary drawn by Delcourt and Delcourt (1990).

Another vegetation map has also been produced by Overpeck et al. (1992) for 18,000 14C years ago, based on numerical analysis of the pollen data through this region. The Overpeck et al. map shows what they call 'deciduous forest' covering the Florida peninsula and other south-eastern parts of the USA, with a 'mixed forest' area extending northwards to North Carolina. The southern edge of their 'boreal forest' zone seems hundreds of kilometres further north than this vegetation type when delineated by Delcourt and Delcourt, Woodcock and Wells and also the pollen core data reviewed in Watts & Stuiver (1980). It is difficult to assess to what extent this discrepancy between the Overpeck et al. map and the information from other sources is due to ambiguities in the definitions of vegetation types themselves, and to what extent it is due to errors in the methods used by those of either point of view. However, when faced with contrary interpretation from other sources, and after consultation with colleagues, we suggest that the Overpeck et al. maps are incorrect in certain key elements. It is worth noting that Overpeck et al. do say in their paper that broad-leaved trees were only a minor sub-component in the LGM vegetation of the south-eastern region (in which pines were apparently dominant), yet they still map it as 'deciduous forest'! There seems to be a problem with the vegetation definitions in their model, in this case at least.

Conclusion; deciduous forest barely existed in the southern USA at the Last Glacial Maximum. Contrary to many earlier interpretations, it now seems inappropriate to show any significant area of deciduous forest in the eastern US for the LGM. Current work by S. Jackson (Northern Arizona University), Eric Grimm (Illinois State Museum) and Bill Watts suggests that even in the south there was very little broad-leaved forest at the LGM, with perhaps only isolated pockets of deciduous forest that were surrounded by coniferous vegetation (M.B. Davis pers. comm. May 1993). These deciduous forest pockets at the LGM may have been confined to moister microsites, such as stream gullies (Ritchie 1982).

As mentioned above, spruce forest has been found to have been growing in Louisiana at the LGM. Tallis (1990) also maps a south-eastern pine forest extending from about 33N down to the Gulf coast and northern Florida, and suggests that it would generally have been fairly open in structure. In general it seems from the pollen profiles that the climate in the eastern USA was drier than most boreal forest in Canada or Siberia exists under at present (Jackson et al. 1997). Faced with the need to map the no-analogue south-eastern LGM vegetation according to some useful category or other, we have opted for 'main taiga', which it seems to resemble in terms of general composition and in not having a dense, closed forest cover in most areas. However, it was probably somewhat drier than the 'main taiga' which generally exists in the present-day world, and perhaps would not have had the same thick organic soil layer. This is evidently an example of the difficulty of pushing past no-analogue vegetation into categories which will be of some meaning to us in the present-day world.

Sparse, scrubby vegetation in Florida. In Florida, LGM conditions seem to have been cooler and much drier than at present. A 5.3°C cooler mean annual temperature in northern Florida is suggested on the basis of stable isotope ratios in groundwaters (Plummer 1993), and this generally agrees with the figures from molluscan evidence (Woodcock & Wells 1990). On the basis of various pollen-bearing cores and also indirect evidence from sand-dune geomorphology, many authors refer to a dry 'sand dune scrub' with a strong Pinus and small palm element and few broad-leaved trees existing at around the LGM (contrary to the interpretation by Overpeck et al. that this area be mappable as 'deciduous forest'). An open pine woodland, with a sub-component of deciduous trees, and with sandy steppe-like openings (Watts & Stuiver 1980) seems to have been present in northern Florida close to the LGM (18,500 years ago).

A hiatus occurs (18,500-14,000 14C years ago) at this site around the LGM, at a time of lowered groundwater levels (attributed to lowered sea level). For this period in which fossils are absent, one is forced to look only at sedimentological evidence. The occurrence of extensive dune activity across Florida during the last glacial (Wells 1992 p.626-628) suggests that there may have been hardly any vegetation on the Florida Peninsula at the LGM itself. Likewise on the Island of Bermuda, just to the south of Florida, there appears to have been extensive dune activity during the low sea level phase of the last glacial, indicating dry (and windy?) conditions and much less vegetation than at present (Herwitz 1992). It thus seems that even the 'sand dune scrub' diagnosis of the LGM vegetation of Florida may be on the optimistic side, with very few trees having been present at that time, and conditions perhaps more closely resembling a desert or semi-desert.

Central USA.

Steppe-tundra and woodland in what is now Prairie. In its 'natural' state, the Great Plains region would at present be covered by moist tallgrass prairie vegetation (a fairly mesic temperate steppe, with a tall continuous grassy herbaceous cover). This type of vegetation seems not to have existed in the central USA at the Last Glacial Maximum (Tallis 1990). Instead, some degree of coniferous woodland cover seems to have been present over much or most of the Great Plains region, especially towards the east. Boreal conifers seem to have extended westwards as far as the central Great Plains (Picea, Larix and some Pinus banksiana), under somewhat moister conditions than at present, west of about 90W. As far west as south-central Kansas and Nebraska, abundant macrofossils of Picea glauca and Pinus flexilis have been found, dating back to 14,000-18,000 14C years ago (Woodcock & Wells 1990). Sites in eastern Kansas, (39 N, 95W and 39N and 95W) show predominantly spruce pollen between about 24,000 and 15,000 years ago.

More-or-less wooded vegetation may have predominated in many places in the present-day Great Plains prairie region. However, in many places this may have been more like forest-steppe or open woodland than true closed forest. For example, south-central Nebraska has often been thought of as having been forested at the LGM but at Eustis (41N, 100W) and other sites in Nebraska, snail faunas in the LGM loess suggest an open woodland or scrub vegetation in a very cold climate with conditions slightly moister than at present, on a poorly developed loess soil, poor in organic matter (Rousseau & Kukla 1994), in contrast to the earlier conclusions of Wells & Stuart (cited by Rousseau & Kukla) who suggested on the basis of spruce charcoal and molluscan fossils that the central Great Plains were covered in a fairly dense taiga-like forest at around the time of the LGM. Possibly, both views may be true to some extent, although the balance of evidence would seem to favour an open wooded or parkland vegetation in most areas.

A fairly closed spruce (Picea) forest did occur at least as far west as central Kentucky (which is to the east of the present prairie belt boundary), in the uplands (Wilkins 1991). Ritchie (1982) cites the work of Grueger; in N.E. Kansas two spring-fed marsh sites have yielded pollen and macrofossils suggesting closed spruce forest at the LGM. However, the extent to which this forest was a product of the locally favourable conditions is not clear. Conditions were markedly cooler than today. Mean annual temperatures at about 35N in Kentucky seem to have been at least 7°C lower than today (Barry 1987), based on a range of botanical and zoological indicators.

Indeed, one must bear in mind the possibility that quite large parts of the Great Plains (the present tallgrass prairie) region were only sparsely wooded at the LGM, with a dense woody cover confined to stream valleys. Thus at a site in the Ozark Highlands of Missouri, where spruce and oak pollen predominates during the LGM, the presence of Equus (horse) and Paramylodon (ground sloth) fossils suggests that a parkland vegetation was locally predominant (Graham & Mead 1987). Elsewhere in the eastern Great Plains, browsers such as deer, longnose peccary and mastodon suggest a substantial wooded element in the vegetation (Graham & Mead 1987).

Steven Stokes (pers. comm. Nov. 1994) is sceptical that there could have been anything more than a very sparse or very localised woody cover over most of the Great Plains region, despite the abundance of spruce (Picea) in the pollen record. He points out the surprising absence of any evidence of iron mobilisation in the LGM soils from any sites in the region; in the present-day world even a very sparse coverage by Picea is enough to cause iron mobilisation. He suggests that apparent importance of spruce in the pollen record reflects nothing more than its ability as prolific pollen producer, together with long-distance transport and selective preservation of this relatively resilient pollen type.

The woody cover on the Great Plains generally seems to have thinned out westwards, with an increasing proportion of prairie herb species in the pollen records, and the increasing prevalence of herbivores associated with open ground and low or sparse vegetation (Tallis 1990). In Kansas (towards the west of the Great Plains region), conditions were generally cooler and moister than at present (Barnosky et al. 1987). As mentioned above, woody conifers extended into this area, but probably only as a relatively minor component in a mainly grassy vegetation.

In more southerly parts of the Great Plains region - in south-western Kansas, Oklahoma and Texas - pine (Pinus) and aspen (Populus tremuloides) woodland may have been more prevalent than the spruce woodland, which occurred further north (Woodcock & Wells 1990). However, the woody cover may have been overestimated due to selective concentration of tree pollen in poorly preserved samples. Hall & Velastro (1995) have concentrated on three well-preserved pollen sites to suggest that in the southwestern part of the Great Plains region in central Texas, the late-glacial (and probably the LGM) vegetation was probably a composite-rich grassland, with only local populations of pinyon pine and deciduous trees confined to stream gulleys. The open-ness of the woody cover is further supported by the prevalence of fossils of open-ground grazers (Equus, Camelops) in the western and south-western Great Plains for the LGM.

Very dry in the High Plains region.

In the High Plains, the drier western part of the present North American steppe region (which presently has a mainly short-grass prairie vegetation), LGM conditions seem to have been very arid (Wells 1992). This area, stretching approximately from southern Saskatchewan down to NW Texas, between about 50N to 32N., is mantled with evidence of major dune activity during the LGM period (Wells 1992 p.316). On the basis of the morphology and extent of the dunes of the Nebraska Sand Hills dating to the last glacial, between 22,000 and 12,500 14C years ago, Wells (1992 p.317) suggests an annual precipitation at the LGM of less than 50mm (not allowing for the much lower glacial temperatures), in contrast to the present 600mm. For most of the rest of the High Plains belt, Wells (1992 p.319) suggests that annual precipitation was well below 250mm (again, not allowing for temperature changes).

This view stands in contrast to previous work based on fossil pollen from the High Plains. At some sites of Last Glacial age, herbaceous pollen with a minor (<10%) arboreal element seems to predominate, whilst at others spruce pollen is relatively very abundant. High levels of spruce (Picea) pollen have been found from sites in South Dakota, and spruce macrofossils occur at a late-glacial site in the Nebraska sand hills area (40N, 99W). Other pollen sites towards the south of the High Plains region, such as in western Texas, also show a substantial element of Picea and Pinus pollen (e.g. at Llano Estacado). Wendorf (cited by Wells 1992, p.331) suggested that dense pine and spruce forest predominated at Llano Estacado. This seems rather unlikely since the area was dry enough to be undergoing deflation (wind erosion) at this time, and there is a striking absence of the podsolized soils that accompany such forests in the present-day world (Holliday 1987). A more likely explanation is that the conifer pollen is derived from long-distance transport and that selective preservation against decay (conifer pollen is generally more robust than angiosperm pollen) has disguised a mainly herbaceous plant community existing at that time (Hall & Valastro 1995). Sites elsewhere in the High Plains (in Wyoming and N.E. Colorado, see Wells 1992 p.323) suggest a predominantly herbaceous shortgrass prairie or semidesert vegetation, rich in Artemisia, during the last glacial.

Re-interpretation of the land-snail record of the southern High Plains in west Texas is also compatible with a dry environment (discussed by Hall & Valastro 1995). Animal fossils from within this region also tend to be those of grazing forms (Equus, Camelops, Bison antiqus, Mammuthus, Antilocapra) suggesting open grassland vegetation.

Considering the various sources of evidence, Wells (1992) suggests that the Great Plains region at around the LGM would have been largely barren over about 50 or 60% of its area, with other parts of it (especially at the peripheries) covered by semidesert or shortgrass prairie vegetation. Some vegetation may have been present in favourable localities even within the arid core of the central Great Plains region, as indicated by localised evidence of loess stabilisation at the surface (Wells 1992). It seems that the dust derived from this desert region is also the source of the extensive loess deposits that blanketed the eastern Great Plains during the last glacial period (Rousseau & Kukla 1994).

At the northern end of the High Plains, just east of the Rockies at the latitude of the Canadian border, conditions also seem to have been generally in an arid or semi-arid mode. The following is an excerpt from a note sent in by Alwynne Beaudoin (pers. comm. Jan. 1998), discussing the evidence for a semi-desert zone in the northern High Plains during the LGM through to the early Holocene (reproduced with permission);

" There are few pollen records extending back into the ca 13 kya BP interval (basically Goldeye Lake, Mitchell Lake and perhaps Chalmers Bog depending on one's interpretation of the dating). Mitchell Lake is probably the most comprehensive record. Carole Mandryk interprets the pollen assemblages as follows "Regional vegetation from c. 17 960 to 16 100 BP is interpreted as an extremely cold semi-arid Artemisia steppe, the vegetation c. 16 100 to 11 900 as an Artemisia-Betula shrubland, and the vegetation c. 11 900-10 200 BP as a Picea woodland, in an environment characterized by consistently arid and windy conditions" (J Paleolimnology 16: 37-57, 1996) ( this would appear to support the reconstruction of relatively arid LGM conditions, although semi-desert rather than extreme desert might be more approporiate for the QEN map; J.M.A.).

Western Cordillera

A complex picture. A great deal of data and interpretation have been published describing the vegetation of the western and south-western USA at around the LGM, due to the abundance of pollen and macrofossil sites (especially packrat middens). However, the sheer complexity of the landscape and climates makes it difficult to form any coherent picture. We have attempted here to summarise the salient features of LGM vegetation, whilst trying to avoid becoming distracted by local details.

Dry periglacial steppe-tundra in the Cordilleran north-west of the USA. Various pollen-bearing cores from the mainly mountainous Cordillera area seem to confirm the general picture of a dry, very cold and relatively sparsely vegetated landscape in the extreme north-western region, often with hiatuses in deposition at around the LGM itself (Barnosky et al. 1987).

In western Washington State, a subalpine parkland (under conditions much drier and colder than at present) was present in the valleys and on the coastal plain. In the Puget Trough of western Washington, the parkland of spruce and pine had a dry Artemisia tundra ground layer (Barnosky et al. 1987). Only on the coastal plain itself were conditions moist and mild enough to support a parkland in which mesophytic subalpine trees were abundant.

Further east across the cordillera, conditions were drier than in the west. North of about 42, cold dry, periglacial steppe-tundra (Artemisia etc.) seems to have prevailed across low elevations in eastern Washington State (from the Cascade Range eastward) and the Rocky Mountains (Barnosky et al. 1987, Tallis 1990). Upper elevations were presumably even more sparsely vegetated than the lowlands. Permafrost seems to have been continuous over large areas of Montana and Wyoming (Tallis 1990).

Moister-than-present conditions in the south-central cordillera. In the southern part of the western cordillera of the US, conditions seem to have been considerably cooler (7-8°C, with winter temperatures as much as 17-18°C cooler) than at present (Spaulding et al. 1983), with a vegetation zone lowering of around 1000m, but rather moister. Lakes covered considerable areas in Utah, Nevada and southern Oregon, due to higher rainfall and/or reduced evaporation. Elsewhere, south to about 30 N, lake levels seem to have been higher than present everywhere. The timing of the glacial lake phase maxima has been the subject of some disagreement in the literature. Thompson et al. (1983) suggest that although lake levels were in fact considerably higher than present at the LGM, the lake level and moisture maximum was was not reached until around 16,000 years ago. Some estimates of annual precipitation (based on plant fossil and lake level evidence) for the American south-west at the LGM have suggested that it was 25% or even 100 % higher than at present (Spaulding et al. 1983), even allowing for the temperature changes. However, due to revised dating of the timing of maximum lake phases (Thompson et al. 1995), and of the true nature of the regional vegetation, the upper estimates may prove to have been much too high.

On the basis of his review of the plant fossil evidence, Tallis (1990 p.208-211) suggests that the vegetation of the cordilleran south-west (between around 26 and 42 N) at the LGM can essentially be reduced to three components which formed a topographic mosaic, with the actual altitude of each zone varying somewhat according to latitude;

1) alpine tundra, above about 2500m.

2) subalpine parkland of open stands of spruce, pines and fir (1500-2500m), intermixed with present-day desert plants such as Artemisia and Atriplex. This is the zone now occupied by juniper-pinyon pine woodland, which had descended to lower altitudes. Local stands of mesic conifers (e.g. Pseudotsuga) seem to have been scattered in sheltered valleys amongst the hills, but only as a minor component in the vegetation. Delcourt & Delcourt (1990) report that there was Douglas fir-white fur-lumber pine forest in this altitudinal band in northern Arizona at the LGM, but it is not clear to what extent they are using the same evidence that Tallis regards as indicating only local stands of these conifers.

3) a lowland woodland zone, below about 1500m, of pygmy conifers (pinyon pines) and juniper, together with xerophytic shrubs and succulents such as prickly pear cactus. This was probably in fact not a continuous woodland but instead a mosaic between areas of semi-desert scrub and woodland. For example, Thompson & Mead (1982) summarise evidence from the area of the Great Basin (approx. 37-43 N) from packrat middens, which indicates that although patches of subalpine conifer woodland were present they did not in fact form a continuous corridor in the valleys. Instead the valleys probably contained a mosaic of dry steppe, meadow and conifer communities, and wooded or forested habitats in this region would probably have been discontinuous islands, albeit larger than today's areas of woodland (Thompson & Mead 1982). Indeed, O.K. Davis (pers. comm., Aug. 1995) suggests that at the LGM the present-day Great Basin desert areas would have had treeless vegetation throughout. He reaches this conclusion on the basis of his knowledge of pollen data from the area, suggesting that the packrat midden based data give a misleading conclusion. It thus seems possible that the LGM vegetation of the US south-west was much sparser than some (e.g. Brown 1978) have suggested on the basis of fossil and biogeographical evidence, and as compared to the 'caricature' of a relatively very moist and generally wooded palaeovegetation in the dry cordilleran zone that persists after reading most reviews and popular accounts of the evidence.

From a study of stomatal density and delta-13C of LGM-age leaves of Pinus flexilis, Van de Water et al. (1994) have suggested that water use efficiency in this tree species was 15% lower under LGM CO2 levels as compared with Holocene CO2 levels. The possible effect of this on vegetation cover is unclear.

The lowland woodland or woodland-mosaic zone did not occur extensively in the higher lands north of 37N, whilst south of 34N, evergreen oaks were important in the vegetation (Tallis 1990). At the lowest altitudes conditions were drier; below 500m juniper was often the only tree present and in the lower Colorado River valley there was a treeless semi-desert scrub. At present, the natural vegetation below 1500m throughout this south-west region is semi-desert scrub.

On the basis of his review of the palaeovegetation evidence, Tallis (1990) (p.210) has drawn a map of suggested LGM vegetation in the south-western US, showing large areas of the southern cordillera region covered with juniper-pinyon woodland. Further north, this graded into parkland vegetation (see above).

In the lowlands and hills near to the California coast, between about 37 and 32 N, macrofossil and pollen evidence indicates the local presence of mesic conifers (e.g. Sequoia), in a mosaic of chaparral vegetation, grassland and xeric woodlands (Tallis 1990, p.210-211). Barry (1987) presents floristically-based estimates of a 5°C cooling relative to the present, whilst Thompson et al. (1993) suggest a smaller temperature depression of 2-4°C based on a range of sources of information. Dry Artemisia -Juniperus semi-desert scrub covered much of the valley of central California. In the Sierra Nevada of California, ice seems to have been generally present above 2000m. In the uplands of south-eastern California, the familiar Pinyon-Juniper woodland was predominant at low-to-mid altitudes. Mesic montane conifers (e.g. Sequoiadendron) were locally present, at favourable sites within this woodland zone (Tallis 1990).

Moister than present in the southwestern semi-desert region, with open woodlands.

Moister-than-present conditions also seem to have prevailed further south of the Cordillera. In what are semi-deserts today, woodland vegetation may have covered lower and middle elevations at around the LGM (Thompson 1992, Thompson et al. 1993). In the Chihuahua Desert (which covers parts of W. Texas, S. New Mexico and N. Mexico), pine-juniper-oak woodland may have predominated at all altitudes. Likewise, in the Sonoran Desert region of southern Arizona, pine-juniper woodland predominated at the LGM. However, the picture is a complex one, and instead of woodland, Artemisia dry steppe was expanded in many northern and eastern parts (Thompson 1992).

Moister sagebrush vegetation in New Mexico. In New Mexico, vertebrate fossil evidence seems to confirm that areas which are now arid grasslands or grassy woodlands were occupied by cool, relatively dense sagebrush vegetation with elements from mixed coniferous forest (Harris 1993).

Grasslands and open woodlands widespread in Texas. The vegetation has been summarised in a table on page 117 of Nordt et al. (1994), drawing on various sources within the published literature. In central Texas, carbon isotope ratios in alluvial organic matter dating back to 15,000 years ago also suggest that late glacial conditions were generally cooler and moister than now, with 50-60% trees and a grassy vegetation with a higher percentage of C3 species in areas nowadays dominated by C4 grassland (Nordt et al. 1994).

It seems possible that various sorts of more-or-less wooded vegetation extended across much of Texas. Some areas in central Texas, dated to around 18,000 years ago (Nordt et al., citing Bryant & Holloway 1985), seem to have had deciduous forest vegetation in areas that became grassland/savanna throughout the Holocene, with July temperatures estimated on floristic evidence to have been around 5.5°C cooler than present (also backed up by noble gas ratios in groundwater in southern Texas). In north-west Texas (Nordt et al., citing Johnson 1987) however, the vegetation seems to have been a grassy parkland, with groups of deciduous trees, in areas that are now naturally shortgrass prairie. It seems that earlier suggestions that conifer woodland predominated at the southern edge of the High Plains were incorrect (see discussion above, in context of the High Plains region).

Moister and cooler than present conditions are also recorded for Edwards Plateau in west central Texas (near the Mexico border), based on a range of sources of fossil and other evidence (Blum et al. 1994). However, the vegetation here seems to have been broadly similar to the present savanna with mixed grasses.

Woodcock and Wells (1990) speculate that perhaps the deciduous forests recorded for the eastern parts of the Gulf Coast (?in Louisiana) might have extended around westwards to Texas, although there appears to be no direct evidence for this. In fact, pollen of xeric conifers occurs close to the Gulf coast in Texas (see map of pollen-bearing sites in Tallis 1990, p. 206). Thus the Gulf coast area might perhaps have had a xeric pine woodland belt.

Alaska and Beringia.

An ice cap originating in the Alaska Range of mountains extended down onto the lowlands, covering almost all of southern Alaska. Smaller ice caps were also present towards the north of Alaska. A broad tract of low-lying land - known as Beringia - was exposed by the lowered sea level, extending between the western coast of Alaska and the eastern tip of Siberia. Although there is not much doubt that the landscapes were almost completely treeless and generally drier than at present, the detailed character of the LGM vegetation remains somewhat controversial (Delcourt & Delcourt 1991, Anderson & Brubaker 1993).

The Beringian vegetation; what was it really like? Various pollen cores, taken from Alaska and the now submerged shelf areas, show that at around the LGM Beringia supported plant communities in which Artemisia, sedges and grasses (including the present short-grass prairie species Boutelona) were prominent. However, the detailed character of this vegetation - in terms of its ground cover, distribution and productivity - is a contentious issue. On one hand, Hopkins et al. (1982) suggest that there would generally have been a fairly lush vegetation, growing under warmer summer conditions than at present, and that this would have been necessary to sustain the grazing mammal fauna reported from the Alaska for the last glacial (Guthrie 1990). Bliss & Richards (1982) suggest a fairly continuous vegetation cover of moist sedge and shrub tundra in the hollows and a more open grassy steppe-tundra covering most of the area, on the gently rolling ridge tops. On the basis of the diverse and apparently abundant large mammal fauna of Beringia, Guthrie (1990) has suggested that this mosaic could have been of crucial importance in maintaining the species richness of grazers.

However, this general interpretation of the full-glacial Artemisia-rich plant communities as indicating a fairly lush and/or productive type of vegetation has been strongly challenged. Various pollen-bearing cores have been recovered from the submerged landmass, and these have been interpreted as showing varying degrees of barren-ness (e.g. see discussion in Barnosky et al. 1987, and references cited therein). For example, cores taken at various points across the submerged Beringia landbridge by Wendy Eisner (Ohio State University, Byrd Polar Research Center; reported in Geotimes, April 1993 p.10) are taken to indicate a general picture of dry, very sparse vegetation (?polar desert?) during the glacial period, switching to a moister climate with peat deposition at around the beginning of the Holocene. Measurements of the absolute pollen concentration of the plant species in LGM lake sediments from NW Alaska suggests that in fact the biomass was much lower than for present-day tundra, and closer to fell-field or polar desert (Ritchie & Cwynar 1982). This case is strengthened by the widespread occurrence (at low abundances) of cushion plants and other taxa indicative of sparse fell-field vegetation. Such moisture indicators as Sphagnum were virtually absent, and halophytes (suggesting arid soils) were relatively common. The general scarcity of organic debris, roots and buried turf layers in pond sediments and Beringian loess of LGM age also suggests a sparse vegetation with soils lacking in organic matter. The fossil insect assemblages are also characteristic of dry polar environments in the present-day world.

The widespread evidence of redistribution of sand and silt by aeolian processes apparently confirms the lack of a continuous vegetation cover in many places (Schweger et al. 1982). For example, layers of aeolian sand from the last glacial in the Primbloff islands, which would have been part of the Beringia landmass (Barnosky et al. 1987).

Guthrie's view that the diverse mammal communities of Beringia actually persisted through the LGM period has been challenged as being based on ambiguous and incorrect dating, with the fossils in fact dating from before and after the LGM interval. This particular issue still seems unresolved.

However, conditions were by no means uniformly so dry. Evidence of relatively moist tundra-like conditions in the Beringia-Alaska region comes from various pollen cores taken in Alaska and the now-submerged areas off the Seward Peninsula. Moisture-dependent graminoids and forbs, seem to have occurred in low-lying valleys and floodplains towards the western side of Alaska (Barnosky et al. 1987). At the sites off the Seward Peninsula, dating to around 16,000 14C y.a., the predominant vegetation seems to have been a rather dry, grassy tundra with abundant dwarf birch shrubs. Peat layers, with Sphagnum moss and dwarf shrub thickets locally, were present locally where pools had formed, and the ecosystem at these sites was much like a present-day moist tundra (Elias et al. 1996). But this is probably not the predominant picture of the area during the full glacial. Further east and further west (Elias et al. 1996), particularly on hillsides and ridge tops, a much sparser vegetation seems to have predominated, in the form of a xeric discontinuous Artemisia 'fell-field' tundra.

Winter temperatures seem to have been colder-than-present everywhere, but it has been suggested that summers in the northwest might actually have been warmer than now (evidence summarized in Barnosky et al. 1987).

In conclusion, the vegetation of the Beringian landbridge would no doubt have varied from one place to another, but we (the editors) are persuaded by various lines of evidence that there was an open, dry vegetation (certainly drier than most present-day tundra) in most places on the land bridge. As a preliminary hypothesis, one may adopt the view that the Beringian LGM vegetation was perhaps one-third by area a fairly dry tundra, and one third steppe-tundra, with the rest as polar desert.

The Beringian Atlas Project is focussing on broadscale reconstruction of the environments of Beringia over the last 17,000 years (however, their maps are not yet fully available online).

8,000 14C years ago (early Holocene).

Coastlines and ice extent. For lack of any better information, coastlines are assumed to be the same as they are for the present day. Large portions of the Laurentide ice sheet remained at 8,000 years ago due to thermal inertia; they were simply so large that they had not yet had time to melt in the warmer climates. The Laurentide ice extent is taken from a maps of Dyke & Prest (1986 & 1987). Small remnants of the Cordilleran ice cap ice may have still survived in the west, in the mountains close to the Pacific coast.

Eastern USA. Forested eastern USA. An abundance of pollen data (Webb et al. 1993) from this area shows that the eastern USA was already heavily forested by 8,000 years ago (Delcourt and Delcourt 1987). From the pollen percentages in these cores, it seems that many of the forest tree communities were of a somewhat no-analogue nature, but this should not prevent them being regarded as falling into the same broad biome categories as today. In the south-eastern USA, oaks (Quercus) generally seems to have been much more important as a forest component than today, comprising around 40% of the total pollen influx. Mapping on the basis of a formalised analysis of pollen data, Overpeck et al. (1992) also show a predominance of deciduous and mixed forest throughout the region by around 9,000 years ago (the nearest time slice to our 8,000 years ago one), with only a relatively small area of south-eastern pine forest present on the Florida peninsula.

Central and western USA.

Extension of prairie eastwards and northwards. Given the general evidence of drier-than-present conditions across the western and central USA (see below), at 8,000 years ago there was probably an eastwards and northwards (J.C. Ritchie pers. comm.) extension of the prairie belt at the expense of forests. Lake level evidence at various sites in the mid-west (at around 45-47N) indicates drier-than-present conditions, reaching a maximum dryness 7,700-4,000 years ago (Webb et al. 1993, Schwalb & Dean 1995). The strongest evidence for an expanded area of prairie is just to the west of the Great Lakes region, where there is an increased representation of prairie species in the numerous pollen cores obtained from this area (Webb et al. 1983, 1993), relative to the 'present-natural' situation recorded from several hundred years ago before extensive agriculture began in this area. Although it seems likely that there would in fact have been a mosaic of areas of forest (in river valleys and on moist soils) and prairie (on drier ground), it is convenient to talk simplistically in terms of a single prairie 'boundary'. In this sense, the boundary at 8,000 years ago stood about 4 Longitude further east than at present, suggesting a annual precipitation around 20% less than at present (Webb et al. 1983, 1993).

Episodes of drier conditions with dune activity in the western prairie region and in the South-western deserts. In the south-western USA, there are various signs that conditions were intermittently drier (and/or more windy) than at present, with widespread dune activity. This seems a contradiction of the more direct vegetational evidence from packrat middens.

There is extensive evidence of a period of early-to-mid Holocene dune mobility in the Southwest, occurring episodically between 8,000 and 5,000 years ago (Tcharkerian 1994). This is surprising because inland dunes are not currently active anywhere within the USA. Active dune fields seem to have been particularly widespread in the high plains of N.E. Colorado / S.E. Nebraska, indicating at most a sparse semi-desert vegetation cover (presently the natural vegetation would be short-grass prairie) between 9,500 and 4,800 years ago (Forman et al. 1992). Areas from which early-to-mid Holocene dune activity has been reported include N.E. Colorado, Nebraska, S.C. Wyoming, the Central Highplains and the Texas Panhandle, S. Arizona, S. Nevada, S. Oregon, E. Mojave desert and S. Mojave desert. To some extent the radiocarbon and thermoluminescence dates indicate distinct episodes of activity that were not necessarily contemporaneous between different areas, but the emerging picture may be that there was indeed a widespread and predominant pattern of aridity during large spans of Holocene time.

Equivocal signs from plant fossil evidence. The vegetation cover at times of dune activity remains uncertain; as Tchakerian (1994) points out, it may have been wind speed rather than lack of vegetation cover that was most important in allowing Holocene dune mobility. Certainly, it seems unlikely that there really were very extensive areas of absolute desert, since plant fossil evidence from close to these regions shows that there was a vegetation not so very different from today's semi-desert, and even rather moister in many places. Perhaps (S. Stokes pers. comm., February 1995) one must view the Holocene of the American south-west as consisting of conditions much like today's for most of the time, punctuated by short, intense episodes of drought and dune activity.

Generally during the Holocene, woody vegetation in the south-western uplands seems to have been less widespread than today, with less Juniper-Pine woodland, presumably due to drier conditions. From charcoal and plant-fossil evidence, sclerophyllous scrub vegetation such as chaparral seems to have been more widespread relative to woodland (relative to the present-natural) at 8,000 and 5,000 years ago (Delcourt & Delcourt in press). However, the woodland vegetation where it did occur was able to extend to higher altitudes. Most palaeoecological studies indicate that the montane tree lines in the Cordillera region were higher than today, presumably due to a warmer climate, between 9,000 and 5,000 years ago (Rochefort et al. 1994). Generally, the difference in tree-line altitude was about 100-200m. In the Chihuahua Desert of New Mexico, conditions inferred from packrat evidence also seem to have been drier than today's at 8,000 and 5,000 years ago.

In some areas of the Southwest, climates at 8,000 years ago generally seem to have been quite similar to the present. In Texas, climates even appear to have been slightly moister than at present (Nordt et al. 1994). Studying soil carbon isotopes, Cole & Monger (1994) have found evidence of an early Holocene switch from C4 to C3 species in the early Holocene (9,000-7,000 years ago). This occurred at a time when other evidence indicates a significantly drier climate relative to the glacial, that should have favoured C4 species during the Holocene. They attribute this anomalous shift to an increase in CO2 levels favouring the C3 species, although in fact one should point out that this does not seem to occur at the right time for the main late-glacial to Holocene CO2 shift.

Sorting out the confusion; south-western desert vegetation much the same as today's at 8,000 years ago and 5,000 years ago? There is such an abundance of Holocene plant fossil data from the south-western USA that attention in the literature seems to concentrate on local and quite minor differences rather than broad trends and similarities relative to the present. Thompson et al. (1993) have summarised general trends in effective moisture relative to the present, as they appear from the plant fossil record. They find that for both 9,000 years ago and 6,000 years ago conditions seem to have been somewhat moister than today. Of course, these time slices do not coincide exactly with the 8,000 years ago and 5,000 years ago slices used here, and it does seem that 5,000 years ago was associated more with dune-building episodes. Nevertheless, there seems to be little justification for drawing vegetation boundaries drastically different from those today on the basis of the dune evidence, when more direct evidence on the existence of similar-to-present plant communities is so abundant. As mentioned above, the early to mid Holocene dune-building recorded from the south-west may reflect only short-lived arid events in a fundamentally unstable climate, and might in any case be related more to stronger wind speeds than a complete loss of vegetation cover.

central Canada.

Canadian vegetation quite similar to the present. In deglaciated parts of Canada, the general vegetation cover by 8,000 years ago seems to have been quite similar to today's, although with less peatland scattered through it. The numerous pollen-bearing sites are summarised by Ritchie & Harrison (1993).

Retreating ice sheet. In the regions of Canada from which the ice sheet had most recently retreated, birch, alder and willow scrub would presumably have been the primary vegetation, followed later by forest-forming trees. Pollen sites dating from the earlier stages of deglaciation in the northern USA suggest that the landscape often remained treeless or only sparsely wooded for thousands of years after the ice had left a site, despite the climate apparently being suitable for forest vegetation (Ritchie 1982). Thus, some broad scale disequilibrium in vegetation structure can be expected to have been present around the retreating ice sheet areas that existed during the early Holocene. A fringe of open recolonising woodland vegetation is thus suggested on the map that we have presented here, as speculation.

In arctic north-eastern Canada, deglaciated areas already had a well developed covering of heath and shrub tundra by 8,000 years ago (Delcourt & Delcourt in press).

Spreading of peatland just getting underway by 8,000 years ago. Peat deposition in North America seems to have been an ongoing process during the Holocene. Peat deposition was at this time active further north than at present, presumably due to warmer temperatures. However, peat initiation and spread may have been a fairly late phenomenon in many areas according to Tallis (1990), due to relatively warm, dry conditions across much of North America, and also the late melting of parts of the Laurentide ice sheet in the north-west of Canada. Tallis presents a literature summary diagram (p.348), which shows peat initiation in central Canada starting around 5,000 14C years ago, Minnesota at 5,000 years ago, and W. Quebec & Hudson Bay lowlands at around 7,000 years ago, with only Newfoundland and Labrador at around 8,000 years ago. This general scenario of ongoing gradual spread of peatland area contrasts with the picture currently emerging for western Siberia and elsewhere, of simultaneous peat initiation over very large areas (C. Kreminetski, pers. comm., May 1994). It is difficult to know whether in North America, the published records of the age of basal peats (i.e. the age at which peat build-up started) are biased in some way towards sites where peat build-up began relatively late for this region in general .

Alaska and northern Canada. Warmer than present in Alaska and northern Canada. There is an abundance of fossil and other evidence from this region, allowing relatively detailed reconstruction of the spread of vegetation during the Holocene. In central and northern Alaska, conditions seem to have been slightly warmer and wetter than at present, whilst in the south-east of Alaska the climate was warmer and drier. The central Alaskan forests were established by 8,000 14C years ago (Anderson & Brubaker 1995), but in some parts of southern Alaska the arrival of spruce forest vegetation and coastal rainforest was delayed until after 5,000 years ago (remaining as shrub tundra with scattered trees), probably due to aridity (Anderson & Brubaker 1995, Elias 1995). Across Alaska, there was alder-birch shrub tundra in the west and spruce parkland in the east.

As in Siberia and Northern Europe, the climate between about 8,500 and 5,000 years ago in both Alaska and North-Western Canada seems to have been somewhat warmer than at present. Fossil evidence from various sites indicates that the treeline in north-west Canada was several tens of kilometres north of its present position, about 8,000 years ago (Tallis 1990, Edwards & Wolfe 1995), although it did not reach its furthest extent until after 7,500 years ago (Edwards & Wolfe 1995). Presumably, a similar pattern of northwards movement of vegetation zones might have been present across certain other deglaciated parts of Canada. In north-central Canada however, the tree line was apparently in about the same position at 8,000 years ago as it is today.

The North-western Cordilleran belt of Canada and the USA. Forested, but not as moist as at present. The western interior of Canada seems to have been forested by 8,000 years ago. The moist north-western conifer forests seem to have been somewhat more restricted than at present at 8,000 years ago (and also at 5,000 years ago), perhaps due to a slightly drier climate. Western red cedar (Thuja) - a moist climate species - was already present in the lowlands of SW British Columbia, but it was less abundant than now. Its place was taken by slightly more drought tolerant (but still mesophytic) Douglas fir, Pseudotsuga. (Delcourt & Delcourt. in press).

5,000 14C years ago.

Since the general picture at 5,000 years ago resembles that at 8,000 years ago fairly closely, much of the information is discussed in combination with that above.

Drier with more forest fires in eastern Canada? In eastern Canada, there was a switchover away from birch and more towards spruce forest during this period. Levels of combustion byproducts in Greenland ice cores suggest that there was a peak in fire fequency in the source areas of eastern Canada, between 6,000 and 3,000 y.a. (possibly beginning earlier) (Taylor et al. 1996). Taylor et al. suggest that this combustion peak may have been related a combination of warm, dry summer climates and also to the amount of combustible vegetation and species present in the successional forests and woodlands following the earlier retreat of the Laurentide ice sheet from the area.

Western USA. Warmer and drier than today. In the mountains of the western USA, treelines generally still seem to have been higher than today, by about 100-200m. A decline set in soon after 5,000 years ago and especially after 3,500 years ago. In the mediterranean climate zone in the west of this region, and in the maritime north-west, from the pollen record the climate seems to have been slightly warmer and drier than at present (Rochefort et al. 1994).

Central USA. Prairie vegetation extended. At 5,000 years ago, various pollen-bearing sites indicate that the prairie-to-forest boundary was still further north-east than its present/historical position, at about the same position as at 8,000 years ago (Bartlein et al. 1984). Lake level evidence from the midwest also supports the view that there was greater aridity, peaking between 7,700 and 4,000 years ago (Schwalb & Burns 1995)

South-western USA.

Dune activity.

There is extensive evidence of a period of early-to-mid Holocene dune mobility in the Southwest, extending up until about 5,000 years ago. However, plant fossil evidence from packrat middens suggests vegetation fairly similar to today's (see above for discussion of this apparent contradiction).

Drier, with sparser vegetation in West Central Texas.

In the Edwards Plateau region of west central Texas (close to the border with Mexico), conditions seem to have been slightly drier than present, based on a range of fossil and sedimentological data. The predominant vegetation appears to have been short grasses and scrub instead of the present vegetation of savanna with mixed grasses (Blum et al. 1994).

Higher treelines.

Treelines in the mountains of California seem to have been around 120-150m higher than today, between around 6,000 and 3,000 years ago (the dates varying according to the site) (Tallis 1990).


Treeline still further north. From the observed treeline changes, conditions in the far north still seem to have been warmer than today at 5,000 years ago, although changes in moisture regime and fire frequency may also have played a role. In north-western Canada, a range of plant fossil evidence indicates that the treeline was 50-60 km north of its present position between around 7,500 and 5,000 years ago (Tallis 1990). A recent map summary of available fossil evidence suggests that between 5,500 and 2,000 years ago the tree line in central Canada lay some 100 km further north than at present (Edwards and Wolfe 1995).

However, in Quebec the forest line may still have been about 500 km further south than at present, judging by the position a thousand years earlier at 6,000 years before present (Richard 1995), perhaps due to a cooler climate or to successional disequilibrium. In southern Alaska, forest was apparently still absent from some areas which became forested after 4,000 years ago. This may have been due to aridity (Elias 1995).

No agriculture yet in Great Lakes region. In the Great Lakes region of southern Ontario (and possibly elsewhere in the eastern temperate forest zone), the charcoal and pollen records indicate that white pine/oak temperate forests do not appear to have been disturbed by humans until farming communities of Amerindians (Iroquois) colonised the area in the 1300s A.D. (Clark & Royal 1995). Before about this time, cutting and burning of the temperate forest by humans may have been very minor .

Names and addresses of QEN participating experts (named in the text above) who have made direct contributions to this work on North America:

A. B. Beaudoin, Provincial Museum of Alberta, 12845-102nd Avenue, Edmonton, Alberta T5N 0M6, Canada.

P. Colinvaux, Department of Zoology, Ohio State University, 1735 Neil Avenue, Columbus, Ohio 43210, USA, and Smithsonian Tropical Research Institute, UNIT 0948, APO AA 34002-0948, USA.

M.B. Davis, Department of Ecology, Evolution and Behaviour, College of Biological Sciences, 109 Zoology, 318 Church Street S.E., Minneapolis, MN 55455, USA.

O.K. Davis, Department of Geosciences, University of Arizona, Tucson, Arizona, AZ 85721. e-mail;

H.R. Delcourt, Center for Quaternary Studies of the Southeastern United States, Department of Geological Sciences and Graduate Program in Ecology, University of Tennessee, Knoxville, Tennessee 37996, USA.

P.A. Delcourt, Center for Quaternary Studies of the Southeastern United States, Department of Geological Sciences and Graduate Program in Ecology, University of Tennessee, Knoxville, Tennessee 37996, USA.

A.S. Goudie, School of Geography, University of Oxford, 1 Mansfield Road, Oxford OX1 3TB, UK.

D.L. Peteet, NASA-GISS, New York, NY 10025, USA.

J.C. Ritchie, University of Toronto, Toronto, Canada.

M. Sarnthein, Christian-Albrechts University, Germany.

S. Stokes, School of Geography, University of Oxford, 1 Mansfield Road, Oxford OX1 3TB, UK.

T. Webb III, Department of Geological Sciences, Brown University, Rhode Island 02912, USA.

K. Whitlock, Department of Geography, University of Oregon, Eugene, Oregon 97403, USA.

S.C. Zoltai, Forestry Canada, Edmonton, Alberta, Canada.