Mapping the world of the last 20,000 years

Mapping the world of the last 20,000 years

The palaeovegetation maps of the interlinked QEN web pages have been produced by Jonathan Adams (QEN and New Series maps) and Hugues Faure (QEN maps), with help and advice from a large number of colleagues. Whilst still subject to uncertainties due to sparse data coverage in many areas, these maps represent perhaps the nearest thing available to a 'consensus' picture, based on published review sources and advice from around 100 specialists in all branches of the study of Quaternary environments. However, as in any rapidly developing field, there are still considerable uncertainties and disagreements amongst Quaternarists concerning the actual global vegetation cover over the last 20,000 years. Some of the differences centre on the methodology to use for the mapping; should one use only precisely dated plant fossil data, or accept the mass of more loosely dated evidence and the relatively indirect information from sedimentological and palaeozoological sources? On one hand, if one uses only the 'best' (14C-dated plant fossil) data, one can be reasonably sure about certain limited areas, but much less sure about the extensive areas where such data are not available. On the other hand, if one accepts the more loosely dated and indirect data, it is possible to gain indications of global-scale environments, but at some risk of correlating patterns that actually occurred at different stages in time, and also at the risk of accepting some evidence that might eventually turn out to be a poor indicator of vegetation conditions.

Besides the QEN maps, alternative reconstructions for the Last Glacial and Holocene global vegetation have been published by various authors, and these vary in where they fall between the two alternative approaches. Frenzel et al. (1992) published various reconstructions of Northern Hemisphere land cover at the Last Glacial Maximum and during the Holocene, based on interpreting a very diverse range of data. Peng et al. (1994) and Peng (1994) have produced a set of Northern Hemisphere/global estimates based on the vegetation maps of Grichuk (1992) and the climate reconstruction maps of Frenzel (1992), which were based on a similar approach. In contrast, at the other end of the spectrum of methods, T.J. Crowley (1995 a,b) reconstruced LGM vegetation cover based on a 220-site pollen data base compiled by Webb et al. (1995)

These various maps are all part of the necessary process of arriving at more accurate depictions of the world of the past, and they have been an important influence on our reconstructions here. We have occassionally found limitations in these other reconstructions which we have tried to allow for in compiling the QEN maps. The recent formalized mapping study by Crowley (1995 a,b) for the Last Glacial Maximum arrives at a picture that in most respects resembles the vegetation distributions arrived at in the QEN maps and the Frenzel et al. maps. However, because Crowley's map is based only on LGM-dated plant fossil evidence it extrapolates over large distances because such 'ideal' data are relatively scarce. This has given rise to reconstructions of palaeovegetation that in some areas seem contradictory to a range of other sources of evidence (e.g. sedimentary and palaeozoological) which occur in abundance in the areas between the isolated plant fossil data points (e.g. Crowley finds a considerably reduced Australian desert area, which is contrary to widely accepted evidence from this region. He also suggests a belt of conifer forest across southern Siberia, which does not appear to be suggested by any pollen evidence within that region, and is also contradicted by a range of indicators suggesting arid conditions). The algorithm-based ecosystem category assignment and grid-based mapping system of the data points which Crowley has used, whilst providing an objective and easily reproducible methodology, might to have have rigidified the mapping process excessively in some instances (e.g. one finds closed broadleaved forest assigned for the LGM of the south-eastern USA even though those who published the evidence find reason for thinking that there were only isolated stands of deciduous trees in a more open and predominantly coniferous or scrub vegetation; e.g. Watts & Stuiver 1980. The suggested extension of a conifer belt accross southern Siberia might be another example of this). However, the fact that the different methodologies do arrive at global vegetation distributions that are in most other respects so similar is reassuring.

Various problems of methodology may have occurred in the LGM mapping effort by van Campo et al. (1993). In their case, a heavy reliance on the out-of-date CLIMAP (1976) land surface maps also seems to have resulted in some suprising choices of land ecosystem cover. For instance, in their map a major expansion of tropical forest to cover much of the Indian subcontinent is shown for the LGM, contrary to evidence for a regional aridity trend resulting from reduced monsoon intensity (e.g. Cullen 1981).

Various carefully compiled interdisciplinary maps have been published in the Northern Hemisphere atlas volume by Frenzel et al. (1992), based upon decades of experience by their authors. These represent an important starting point for our work, and we have added further data and comment to modify and improve these maps. In general, the maps show much greater aridity and desert expansion throughout the Northern Hemisphere for the Last Glacial Maximum, and data which have come forth since then tend to suggest even more aridity. Unfortunately, there has been some misinterpretation of the LGM map of Grichuk (1992) in the same volume. Grichuk's map appears to show relatively dense vegetation cover and forest in many parts of southern Europe and southern Siberia for the LGM. The Grichuk map has been accepted by at least some authors, for example by Peng et al. (1993), but consultation with Russian Quaternary scientists suggests that it does not a reliable picture, for there is abundant palynological and sedimentological evidence to show that conditions were far more arid, with woody cover almost absent across both Europe and Siberia. In fact, the origonal purpose of Grichuk's map was apparently to show broad biogeographic 'source areas' (containing possibly very small and scattered refugia) and not vegetation coverage as such (E. Zelikson pers. comm.).

A quite separate type of problem is evident with Prentice et al.'s (1993) set of global vegetation maps which were based on a GCM (General Circulation Model) simulation for the Last Glacial Maximum. Their reconstruction of vegetation, whilst showing many of the correct gradients in moisture distribution, has too much forest and dense vegetation, and not enough desert, when compared to the available paleoenvironmental evidence. For instance, large amounts of forest and wooded vegetation are shown extending across Europe, southern and central Siberia, and China, in areas where there is quite abundant plant fossil, animal fossil and sedimentological evidence for semi-desert and very dry steppe conditions at that time. The problem, it would appear, is the set of CLIMAP (1978) sea surface temperature data that the climate model was 'fixed' against. Evidence is now accumulating that in many parts of the tropics the sea surface temperatures were several degrees lower than the original CLIMAP data had suggested (e.g. Broecker 1995). If a model were to be run with substantially lower tropical sea surface temperatures, it would be likely to result in a substantially more arid climate in many parts of the world, agreeing more closely with the palaeoenvironmental evidence.

Different mapping philosophies.

Some of the differences amongst the various maps that have been published can be seen as the result of two different approaches to the difficult task of reconstructing global vegetation to the past. One approach, that emphasized for example by Prentice et al. (1993) and Crowley (1995a,b), emphasizes the features of objectivity and reproducibility that come with a strongly mechanical approach to the mapping; data and algorithms are input to a computer, and the resulting vegetation map is worked out by numerical processes. Generally, only data which can easily be quantified are used for the reconstructions of vegetation and climate. Non-plant fossil sources of information are avoided as they are seen as 'indirect' and 'imprecise'. This approach has the distinct advantage of avoiding any preconceived notions that may unconciously bias the conceptions of 'schools of thought'. A computer has no preconceptions other than the algorithms and data supplied to it, and it is thus a useful, independant source of hypotheses on the distribution of vegetation at a particular time in the past.

The approach that we have used in the QEN and New Series maps is a rather more traditional, less formalized approach, emphasizing the complex and interdisciplinary nature of the evidence. The mapping process we use is based upon many different sources of evidence (including the maps published by the numerical method described above!). In our view, data on the past land conditions are too sparse in many areas to be able to extrapolate only from the relatively small number of data points that a pollen database would use. For this reason we do not consider that it should be necessary to ignore the extensive evidence of sedimentological and geomorphological features such as sand dunes and buried soils, or the characteristic animal faunas of open ground or closed forests, when making decisions of how to reconstruct vegetation cover. We also consider that at present, a computer programme cannot fully substitute for the site knowledge and experience of fieldworkers in interpreting the plant fossil record of each region. Thus for example, in many mid-western pollen sites from the last glacial, spruce pollen is the dominant type present. Working from rules that seem reasonable elsewhere, a mechanical computer-based mapping process would tend to suggest spruce forest across most of the mid-western USA for the last glacial, whereas those who take cores and work on the region now consider that over much of the area this is only an illusion based upon selective transport and preservation of spruce pollen. The models and the mapping algorithms will certainly improve as time goes on, but they will always be relatively limited if they do not take into account the mass of additional data from sedimentology, zoology, and the basic interpretative understanding that field workers have for the complexities of each site and its local vegetation.

Our ecosystem reconstructions do emphasize 14C-dated evidence, whether palynological, sedimentological or zoological. However, the error bars on the age of each site often span a few thousand years, even with 14C. Lack of precision in dating is not as great a limitation as it might seem; often the indications are of a long and fairly uniform phase of arid conditions starting just after around 25,000 14C years ago, and ending just before the early stages of interglacial warming. Considering the lack of any indications of moist phases, and the relatively uniform depositional conditions, sedimentologists, geomorphologists and palynologists studying these sediments can reasonably hypothesise that an arid phase began sometime after 25,000 calendar years ago and continued up until around 13,000-14,000 calendar years ago. It now seems that from the latest evidence that different regions may have peaked in cold and aridity at slightly different times, just a few thousand years earlier or later than one another. Nevertheless, there is enough time constraint to say that at some stage during the Late Glacial time frame, if not necessarily the LGM 'sensu stricto', there was an overlapping period when very cold global conditions and severe aridity occurred across many different regions. This would have given a global vegetation distribution much like that which has been reconstructed in the QEN maps for the LGM/full glacial period, though perhaps slightly drier in some regions and slightly moister in others. New, more precisely dated evidence is tending to favour a slightly later global aridity maximum (17,000-15,000 14C y.a.) in many areas, even though the LGM was itself generally very arid on a global scale.

Whilst we do not claim that the QEN and New Series maps are anything more than a preliminary stage in the lengthy process of development of global palaeovegetation maps, we suggest that with their strongly inter-disciplinary approach they may represent the most up-to-date and accurate global scale ecosystem coverage so far achieved. By relying on direct consultation and cross-checking with regional experts from all across the world, as well as an extensive review of the current literature, we aim to bring the greatest possible amount of diverse expertise to bear upon the task. If you feel that you can contribute to the task, please send us your detailed comments and criticisms of the existing maps and data review, and we will make the necessary changes and add your name to the growing list of QEN participants!

More detailed description of the mapping process

Jonathan Adams, Environmental Sciences Division, Oak Ridge National Laboratory


"Who the heck is this Jonathan Adams, anyway?"