Last modified Tuesday 25th February 1997
INTRODUCTION TO THE REVIEW AND ATLAS OF PALAEOVEGETATION
VIEW NEW SET OF PALAEOVEGETATION MAPS (PRELIMINARY)Details of the vegetation categories used in the maps THE GLOBAL ATLAS OF PALAEOVEGETATION
Return to QEN Home Page
List of References (separate document)
A set of preliminary, broad-scale ecosystem map reconstructions is presented for the world at the Last Glacial Maximum (18,000 14C years ago) and the early Holocene (8,000 14C years ago), the mid Holocene (5,000 14C years ago) and for comparison 'present-potential' maps that may be regarded as approximating the late Holocene vegetation distribution as it would - or might - have been without agricultural modification. The maps were produced through consultation with an extensive network of experts and a range of literature and map sources, with the final decision in each case made by the editors. Accompanying each regional map is a general background text detailing the principal sources of evidence and the major uncertainties within this.
These maps are not intended as the 'last word' on the distribution of ecosystem types at these times - they are merely a rough attempt at appraisal of current knowledge and opinion. Nevertheless, the maps and the accompanying literature review should provide a valuable and readily accessible source of information on current opinion in the Quaternary community. It is also hoped that they will act as a forum for discussion on the distribution of palaeovegetation amongst those who are working in each region.
NEW! Discussion of the differences between various published palaeovegetation maps
The Radiocarbon age scale vs the 'real' (calibrated) years age scale.
Most information on the past 30,000 years or so is from sites or specimens that have been dated using radiocarbon (14C). However, the radiocarbon age scale that would be calculated from first principles (based on the decay rate of the 14C isotope) is not always reliable, because there have been fluctuations in the rate of production in 14C at the top of the atmosphere. Other dating methods (e.g. U/Th) can be used to attempt to check the 'true' age of specimens or sediment layers dated by 14C. The most convincing way to check the 14C age scale is through biological or sedimentological features which build up annual layers over long periods of time (e.g. tree rings, and annual layers of sediment building up on lake beds); counting back the annual layers will reveal the true number of years before the present, and comparing the 14C age of each tree ring or sediment layer will give an age scale for how 14C age can be converted into 'real' age. The most recently used apparent 'consensus' (e.g. Dahl & Nesje 1996) 14C-to-real age conversion scale is given below, but because it is possible that opinions on the appropriate age conversion will change as more data come in, the time slices of the maps are presently described according to a 14C age scale. The reader can use this preliminary age scale as a guide to the likely true age of each of the time slices and vegetation distributions given on the QEN web pages.
14C years ago=>Calibrated ('real') years ago
1,000 => 1,000
2,000 => 2,000
3,000 => 3,200
4,000 => 4,500
5,000 => 5,900
6,000 => 6,950
7,000 => 7,900
8,000 => 8,900
9,000 => 10,000
15,000 => 17,000
18,000 => 21,000
Over recent years there has been a great growth of interest in the global climate cycles which have occurred over the last 2.4 million years, during the Quaternary period. At least partly, this sense of interest reflects a hope amongst researchers that a better understanding of the Earth's past will allow improved prediction of human effects on the environment. There is also a great deal of purely academic curiosity amongst ecologists, archaeologists, anthropologists, climatologists and biogeochemists about the ways that the global environment has changed during the recent past, and the way in which each aspect of the global system has interacted with others.
To aid understanding of the how the Earth has changed since the peak of the last ice age, around 18,000 (radiocarbon) years ago, a set of maps is presented here depicting the world's vegetation/ecosystem cover on a region-by-region basis. These maps cannot be regarded as the definitive last word on the subject; instead they represent a necessary first step in the process of assembling data and opinion from the very large number of scientists who work on vegetation reconstruction. There are many within this field who will say that it is inappropriate to try to present vegetation maps for regions in which there is still a great sparsity of data, and no doubt there will be others who feel that their own well-founded views on palaeovegetation reconstruction have been ignored. However it is necessary to start somewhere, with a serious attempt at showing what the world might have looked like at specific times in the past. The editors would nevertheless urge anyone using these maps to take a glance at the literature review and description of methods presented in the following pages, to get some idea of the major uncertainties that remain.
As well as providing a public service for those outside the main Quaternary vegetation community, the QEN maps may prove useful to those within it. Being the first reasonably coherent interdisciplinary set of global biome maps for the LGM and early Holocene, they may act as a baseline for further work in vegetation reconstruction - even if only as a target to aim criticism at. There is still a tendency for scientists working on the Quaternary period to think on a purely local or sub-regional scale, when they should also be thinking in regional or global terms. This summary may help individuals to keep in touch with others in their field, even when they are studying palaeovegetation on the other side of the world. One may also hope that it will also help to bring about greater coherence in their ways of presenting information, by showing that with a little effort it is possible to make a genuine contribution to a global picture.
Methods and information sources used
The Quaternary Environments Network. This compilation of information summarises the efforts of J.M. Adams and H. Faure, in connection with an informal network of contributors which they have named the Quaternary Environments Network. The Network has operated for the past six years through a system of informal contact; experts around the world are asked for their opinions on the nature of palaeovegetation in particular regions for particular time slices, using whatever palaeoindicators they consider to be relevant and in return these experts are supplied by updated compilations of maps and literature reviews for parts of the world outside their own specialist region. Participants are named here with the aim of providing proper acknowledgement for their direct contributions to the process of gathering information (e.g. replying directly with advice, drawing maps, checking over the text, sending reprints etc.). Anyone who is displeased to have been named here can contact us, and we will immediately remove his/her name from the list.
A top-downwards approach; a culture shock for some ? There is undoubtedly a need to produce improved and up-to-date vegetation maps of the Quaternary; but how should one best go about it? In an ideal world, the editors of this network would have read every relevant paper, and they would understand its implications in just as much detail as the specialists who wrote that paper. However, this is of course impossible; when looking on a global scale it is always necessary to delegate the task to others to some extent, and this approach has been used extensively here. The alternative would be never to have attempted the mapping in the first place, and yet the need to compile an overall picture of the Quaternary world is a pressing one. It is necessary for someone to act as a catalyst for the putting together of current opinion from all the disciplines which relate (directly or indirectly) to Quaternary vegetation ecology.
At a time when a huge and expanding literature exists on the late Quaternary, the secondary use of data and delegation to input of expertise by others seems the only logical strategy for going about such a task. It is evident that this suggestion may be difficult to accept by many Quaternary scientists, because it runs against the cultural current of specialisation and direct citation of all primary sources that is dominant within late-20th century science.
A related issue is whether one should cite old and relatively outdated reviews when newer work is available. There is certainly a strong tradition within the sciences of citing important papers that are decades old and have long been superseded by other work. As QEN is a global-scale review, only the more recent and up-to-date research papers or reviews are included, unless an older piece of work contains some particularly important evidence that has been neglected by later authors. Our 'blindness' towards older work tends to rankle with many experts who see that their colleagues' and mentors' important origonal contributions are not even mentioned here. However, whilst one recognises the contributions made by everyone who has laid the foundations of the Quaternary sciences, the task here is to present a useful review of current information and opinion, and not to write a historical account of the subject.
The approach of the QEN review is to rely on openness and straightforwardness. In this review, the editors aim to acknowledge every single source of advice that they use, and updated versions of the literature review and maps are freely available to anyone who needs them through the World Wide Web address, or (by special request) in printed or diskette form.
Thus, throughout the process of putting together the maps and the accompanying review, the emphasis has been on a 'top-downwards' approach. Regional experts are consulted on the basis of their own published reviews and maps; these experts discuss the problems and uncertainties in the evidence with the editors, and perhaps recommend some more key papers to read. These correspondents are returned to later on, to check that what has been written in this review is factual and reasonable. Of course most of the real work, and the real thinking, is being done by those who are consulted by the editors and who publish the hundreds of papers on the Quaternary that appear each year. But a few speculative points based on recent discoveries are also added by the editors, as a provocation for further thought and response.
It is necessary that the editors emphasise once again the debt to everyone who has taken the time and interest to contribute data, ideas and opinions on palaeovegetation to the Network over the past several years; the task would of course have been impossible without their help. However, their contributions were made as tentative suggestions and not as definite conclusions, and it is inevitable that there will be disagreement amongst different individuals working within the field. The editors hope that they have been able to summarise the arguments fairly and without prejudice.
Types of information and their differing quality; the hierarchy of usefulness. The ideal way to understand the vegetation of the past would be to discover it fossilised en masse and in situ, in the way that certain Cenozoic and Palaeozoic plant communities have been found (Pires-O'Brien 1995). Unfortunately this does not often happen, but failing this ideal there is still a whole spectrum of less perfect sources which can give more or less useful information on what the vegetation in a particular area once looked like (Williams et al. 1993). There are no simple rules that can be applied everywhere towards interpreting the evidence, but in a general sense they can be said to form a hierarchy of usefulness.
The best indirect indicators of ancient vegetation structure and composition are pollen and plant 'macrofossils' (leaves, twigs etc.) which frequently turn up in lake deposits or in offshore sediments. When critically handled, pollen and plant macrofossils are a powerful tool in the mental process of vegetation reconstruction. Often, aided by the latest techniques of numerical reconstruction by computer analysis (bringing in the pollen production and transport characteristics of each plant species), one can gain a fair idea of the abundance of individual species of plants and their physical and ecological relations to one another within the ancient plant community (e.g. Prentice & Webb 1994). The most useful fossil evidence is that which turns up at many different sites, and is well dated, for example by radiocarbon dating. Unfortunately, the greatest problem with relying on plant fossil evidence is that it tends to be in short supply. Cores that yield fossil plant data back as far as the Last Glacial are few and far between, and where they do occur there is often a gap in deposition (known as a hiatus) at the critical period around the Last Glacial Maximum. Thus it is often necessary to extrapolate or interpolate; in space between different sites, and in time to before deposition in a particular lake or swamp began. Intuitively this seems likely to introduce considerable error, so it is important to admit whenever one is relying on it, as has been done here. However, from those well-dated cores where deposition does occur straight through from the glacial maximum one can see that there is usually (though not always) a marked similarity between the conditions at the LGM (approximately 18,000 radiocarbon years ago) and a couple of thousand years either side of it, except that the LGM itself tends to be more 'extreme' in terms of its difference from the present. Lack of precise dating is often a problem for fossil-bearing deposits, but more so for geomorphological features that lack organic matter (see below). Henc the case for an LGM or Holocene age must often rest on explaining the general context in which the evidence is found, and whether it seems to correlate with an overall pattern. If the locality did not in fact fit the broader pattern in its climate and ecology, then this assumption of synchroneity in the environmental or vegetation record may be incorrect.
The information that one can gain from plant fossils certainly has its limitations, as one can see from observing the processes of fossil preservation in the present-day world; for example leaves can be concentrated by stream action to give the impression of abundant trees when in fact there are only a few somewhere upstream. Sometimes, a particular resistant pollen type is concentrated by the way in which other pollen types decay more easily, giving a misleading picture of the composition of the palaeovegetation. A probable example of both these processes at work is the way in which spruce pollen, blown in from remote sources, seems to have accumulated in soils from the Great Plains and High Plains of the USA during the Last Glacial Maximum (conflicting with other sources of evidence from plant macrofossils and palaeopedology).
Sometimes, one can get an 'oasis effect'; a site which preserves large amounts of pollen or plant macrofossils (usually a swamp) tends to be the moistest location thereabouts, and the type of vegetation which grows there will tend to reflect the locally high moisture levels and deeper soils. Hence if one was to look only at the plant fossils one might sometimes get a misleading impression about what the rest of the landscape was like. Various possible examples of this 'oasis effect' are pointed out within the following text sections on regional histories; the contrast between pollen and dune evidence for the LGM of south-eastern Australia (Thom et al. 1994) is one instance, and other examples may include contradictory interpretations of the vegetation history of the Holocene steppe zone of Russia. Mistakes due to the 'oasis effect' are in fact more likely to occur on the very large scale. Very often the only plant fossil evidence is from moist sites (e.g. mountain areas receiving orographic rain, or coastal areas receiving moist sea winds) hundreds of kilometres apart, where lakes and swamps were in abundant supply. If conditions elsewhere were in fact too arid to preserve plant fossils, one would get a very misleading picture by simply extrapolating between the 'favoured' sites where preservation actually occurred. Thus it will always be necessary to use other sources of evidence (e.g. preserved ancient soils, dunes, animal fossils) that are not dependent on the same combination of conditions as plant fossil preservation, to back up or to test the conclusions based on plant fossils. Reconstruction based only on plant fossil evidence becomes invincible and unscientific if no such tests are accepted.
Below plant fossil evidence in the hierarchy of usefulness for vegetation reconstruction are other biological indicators such as animal fossils; animals in the present-day world often tend to live amongst vegetation of a particular structural form, or under climate conditions of a particular type (and hence, by an extra leap, to vegetation). However, on detailed study many animal species turn out to be more flexible in their present-day habitat requirements than ecologists had first realised, and it is always possible that they might have had changed preferences in the different world of the past. Nevertheless, this should not distract from the strong case which can be made using fossils from several species of vertebrates and invertebrates simultaneously.
Also potentially very useful are geomorphological/sedimentary indicators of the ways in which soils and sediments have moved in the past; if one finds evidence of actively moving sand dunes existing at a particular time in the past, then one can probably say for sure that there was only sparse vegetation cover at that site. In fact, the use of dunes as an indicator of past vegetation cover may not be quite as simple and reliable as has often been supposed; some studies in Australia indicate dune mobility with up to 35% vegetation cover if wind speed is high enough (Thomas & Shaw 1991). Thus, for the past one does not necessarily know what roles greater wind speed or lack of vegetation cover actually played in allowing increased dune mobility (there is independent evidence from oceanography and windbourne salt levels in ice cores that wind speeds over the oceans were up to 50% higher than present during the LGM). If a site was locally drier than the landscape in general (e.g. due to it being on a well-drained sandstone plateau) it is to be expected that dunes and desert-like conditions would form much more easily there (a sort of 'inverse oasis effect'). Thus one must also look for subtle clues of dune morphology and the spatial scatter and intensity of dune fields, and use this evidence in combination with other sources, before assigning large areas to desert conditions in the past. Very often, it is the sheer extent and depth of blown sand deposits which gives reason to suspect the past existence of a true desert and not merely sparse vegetation in a windy climate.
Other sedimentary indicators are more ambiguous. For example, the onset of stream sedimentation in a sequence may indicate a drying of climate resulting in loss of vegetation cover if one is in a forest or woodland zone. Under such conditions, the occasional storms which do occur are enough to sweep large quantities of material off the bare landscape of the headwaters, to deposit it downstream. On the other hand in a generally more arid zone, onset of sedimentation might mean a moistening of climate, because there was previously not enough surface runoff to cause sediment to be transported in any large amount. The interpretation must depend to some extent upon the present-day context, and upon the general direction of change indicated by other clues from the same region. Thus the case for a particular change having occurred in the past can rest upon a very tangled web of evidence.
Sedimentary clues to past vegetation do not necessarily have to relate to the surface vegetation cover in any way. Isotopic and dissolved gas temperature indicators (e.g. oxygen isotope ratios in cave calcite deposits, and the ratios of different noble gases in ground water) relate indirectly to the air temperature. This can give clues to the past climate conditions, and these data can be used to try to reconstruct vegetation given a general knowledge of the present-day relationships between vegetation and climate.
Generally less satisfactory amongst the sedimentological vegetation/climate indicators are those such as lake levels which merely indicate the overall direction in which climate changed - towards wetness or aridity - often without any real way of quantifying the magnitude of this change. Although models are nowadays being developed in order to quantify such changes, those who use them generally admit that they are very imperfect. Since vegetation is partly a function of climate, one can say that it would probably have become sparser or denser as the climate change occurred, but not by how much. Likewise, there are many things that can affect either the level of a lake or the local sedimentation conditions (including vegetation itself), and not just climate. Nevertheless, these qualitative indicators are often a useful means of either reinforcing or challenging a case constructed upon other sources of evidence.
Least direct, and probably least reliable, as an indicator of past vegetation are biogeographical clues based only on the present-day distribution of animals and plants. The argument in this case is that one can deduce what happened to these species during past glacial periods from the present patterns of distribution and diversity of species. However, it is perhaps too easy to dream up 'just so' stories about how things got to be the way they are now, and in any case one does not know if it was the last glacial period which was the crucial event in producing the observed present-day pattern. Despite these reservations, the QEN editors at least do still regard this source of evidence to be worth discussing as part of the overall case.
All these sources of evidence must be used if one is to stand a reasonable chance of building up a picture of what the world's vegetation actually looked like as far back as the last ice age. Pollen and plant macrofossil data are lacking from many areas when one looks back further than a few thousand years, and consequently the only recourse is to use whatever other evidence come to hand. These other sources must be used carefully, and in combination. The aim is to build up a reasonably well-founded picture of the past, which may or may not eventually turn out to be accurate, but which can be continually tested as new evidence comes to light.
General Circulation Models; their importance and limitations. There is another legitimate source of evidence that has been avoided in the present review. This is the approach based on GCM modelling, reconstructing the past world's climates from first principles and then predicting on this basis the vegetation that one might have found in it. This is partly because it is hoped that the picture gained from all the other sources of evidence can be used to 'test' the accuracy of GCMs, pointing out their potential flaws as predictors of other future or past climates. This aim is already foremost in the minds of many palynologists and geologists working in North America and in Europe, and numerically precise databases are being constructed with this specific purpose in mind (for example, the NOAA National Geophysical Data Center/World Data Center-A Palaeoenvironmental data sets). The QEN review does not aim to supplant these quantitative 'GCM-testing' databases, but it may at least help those who seek such data to keep in touch with the latest sources of information, and also remind them to exercise a necessary degree of caution in accepting such data on their own at face value.
However, there is also another reason that GCM data is not used here. This is the fact that in well-studied areas GCM-based predictions of palaeovegetation turn out to be extremely poor compared to a range of other more direct sources of data on climate and vegetation. There have certainly been some conspicuous successes in such areas as eastern North America and Europe in terms of 'predicting' Holocene climate and vegetation change on the basis of past global boundary conditions (Huntley 1992). However, for the Last Glacial Maximum some serious problems with this approach become apparent. For this older time slice, even the most sophisticated GCM-based predictions of global palaeovegetation are in broad scale error when tested against the diverse local sources of data. The most recent GCM-based predictions (e.g. Prentice et al. 1993) for the last glacial do in fact seem capable of producing a fairly realistic pattern of moisture gradients for the LGM, the problem is that the actual quantitative moisture level (at least in terms of predicted vegetation type) is evidently far too high in many areas if one compares this with the combined array of palaeoevidence. For instance, recent model predictions for the LGM suggest a belt of taiga across Europe and moist tundra or taiga across northern Eurasia for the LGM when there is a large amount of more direct evidence for very sparse steppe-tundra and polar desert. The predictive match generally seems so poor that in this review the decision has been made not to use quantitative GCM-based vegetation reconstructions for the LGM even in areas where palaeodata are sparse, preferring to make do with what there is from the fossil and sedimentary records. It now seems that the sea-surface temperature data which all previously published GCM reconstructions of the Last Glacial Maximum relied upon may be fundamentally in error (Broecker 1995), and in any case the GCMs do not incorporate any detailed assessment of the effects of airbourne desert dust and haze on full glacial climates (P.J. Valdes pers. comm.). Both these factors may help to explain why GCMs seem to overestimate the amount of rainfall available in the Last Glacial Maximum world, and give grounds for caution in accepting GCM palaeoclimate predictions uncritically.
Holocene GCM-based vegetation reconstructions might potentially be more reliable (judging by their recent results), but the most sophisticated models do not yet seem to have been used to reconstruct global-scale vegetation for the Holocene.
Although the use of global-scale climate models has been avoided by the editors, one must bear in mind that in extrapolating between isolated pieces of data, many Quaternary scientists must construct what is in effect a simple regional palaeoclimate model in their own head in order to reconstruct the pattern of environmental change which occurred. Thus for example isolated data points from across what is presently the south Asian monsoon belt are generally seen as reflecting an overall pattern of past climate across this region. This is based on the reasonable assumption that there would also have been a some sort of monsoon belt in the late Quaternary, and that this might have weakened or strengthened in unison across broad areas. It is in effect a very simple climatic model that most palaeoecologists intuitively construct in assuming (below) that in many areas, the present-areas of highest rainfall were likewise relatively moist at the LGM. If there are flaws in such ad hoc climatic 'models', then the resulting assumptions about past patterns of vegetation change may also be wrong. If the global-scale GCMs are also tested against similarly extrapolated assumptions, then one can hardly regard the regional data as a true test. Thus, in testing GCMs it is best to concentrate on individual items of 'hard' palaeodata (such as those partially but only poorly summarised in the text here), rather than the interpolations between them that have been represented in the biome maps. It is a case of different purposes in mind; the aim here is not primarily the testing of GCMs but rather to reconstruct the world's land vegetation for the past. In certain cases, the text will refer to the findings according to GCMs that (for instance) the appearance of large exposed shelf areas in South-east Asia would have produced a lower rainfall belt in the interior of this land mass. It is the quantitative predictions of the GCMs which seem intrinsically less reliable. Their qualitative predictions of general directions of rainfall gradients are more likely to be accurate. In this sense, making limited, explicitly stated and cautious reference to climate modelling data does not seem to be an inappropriate addition to the picture primarily based on palaeoevidence.
The present editors incline towards the view that in most, but not all, parts of the world the most 'extreme' (driest and coldest) climate during the last glacial occurred fairly simultaneously such that it can be mapped as a single time frame. This view is based upon considering the closeness of the independent dates for the most 'extreme' conditions which are available from very different places around the world, and upon the view that the global climate system was linked by atmospheric dust, aerosol, ocean circulation, ice albedo and greenhouse gas content into a system likely to behave as a relatively uniform whole. In time, with better dating information, this view might be proven to be incorrect.
The interdisciplinary approach. The interdisciplinary approach described above must often make use of subtle, complex human judgements that would thwart the ambitions of some Quaternarists to put their science on a par with physics and chemistry. There is at present a powerful movement within the palynological community towards a very rigorously defined approach towards palaeovegetation mapping (e.g. Prentice & Webb 1994). Only precisely dated pollen or macrofossil data are accepted as relevant to the task. Each pollen diagram is assigned to a biome category according to a numerical method, an interpolation algorithm is applied, and a palaeovegetation map appears on the computer screen.
This is a good idea, often useful (the present review has often made indirect use of numerically-based evidence, wherever it is to be found in the literature). But there is a danger that the numerical-algorithm method can be carried too far with its air of authoritative final judgement, flattening any differing opinions with such terms as 'rigour' and 'objectivity'. Scientists in fact have to live with a difficult, unhelpful world which often cheats them of the evidence they most need, and then throws in a few tricks to lead them off the trail. It is only when one begins to consider all sources of evidence together, and when one actually talks to the individuals who know each locality and its regional ecology, that one can see that the results of a one-sided numerical approach to mapping based only on plant fossils have often proven highly unsatisfactory. Unfortunately, the limitations of this approach are not properly admitted or even discussed in papers that have reconstructed vegetation using this method.
Part of the problem is that the exclusively plant fossil-based approach tends to focus on sites which are most favourable for plant fossil preservation, no matter how rare or unrepresentative these are of regional landscapes (the 'oasis effect', mentioned briefly above). Thus, amongst the published literature one may find woodland vegetation reconstructed by computer (see Huntley's work, cited below in the section on Europe) for the early Holocene in southern Russia, even though the Russian workers who study and know the area - and who took the cores, counted the pollen and published the results - say they find plenty of evidence for thinking that the vegetation was actually an open steppe with only local patches of woody vegetation in the stream valleys (with abundant zoological and sedimentological evidence to support this view). It is this localised woody vegetation that tends to show up disproportionately in the pollen record despite the fact that it is a small element in the landscape, because in a modern steppe landscape trees do tend to be confined to the same stream valleys and hollows where fossil preservation is occurring (Zelikson pers. comm.).
To complicate matters, the opposite effect can also occur; long-distance transport of pollen from abundant pollen producers and selective preservation of decay-resistant pollen types can give the impression that a plant species or genus was abundant at a site when in fact it was sparse or absent from the area. Long distance transport followed by selective preservation of spruce pollen initially gave palaeoecologists the impression that spruce woodland was much more widespread in the last glacial mid-western USA than was actually the case. Only careful, expert interpretation of the pollen record coupled with other independent forms of evidence (macrofossil, zoological, sedimentological) can show up such flaws in the plant fossil record.
The various published maps have been prepared by an exclusively plant fossil-based approach all seem to suffer to some extent from the problems that result from ignoring other sources of relevant data, and from ignoring the subtleties and peculiarities of individual sites. This is in itself not a great or a permanent problem, but there seems to have been a shyness amongst many who publish in the subject towards admitting the current limitations of their own numerically-based methods. Formalised, computer-based reconstructions will certainly improve in the future, but at the moment they do not seem sophisticated enough to trust uncritically. And no matter how sophisticated they become, they will always be relatively limited if they do not consider other sources of evidence external to the local pollen or plant macrofossil record. Rigorously dated plant fossil evidence is still in short supply, and in many areas of the world it will be lacking for years to come, if not forever. We must collectively ask ourselves whether we are prepared to wait that long before presenting more accurate palaeovegetation maps for discussion.
The form of the data
The present review is a summary and discussion of information and opinion on palaeovegetation distribution. Hence it should not be seen as a rival to databases (e.g. Prentice & Webb 1994) which attempt to summarise 'raw' (uninterpreted) data in a form which subsequently can be used for detailed analysis. Rather, the present review mainly takes and discusses information as interpreted within the literature, including that originally supplied in raw form by databases. Detailed information such as pollen percentages and the nature and frequency of radiocarbon dates is generally not referred to here. The reader will need to refer back to the original cited documents to find all these details. The raw information of dating and of sites are only discussed when they are seen as critical to the case of reconstructing a particular vegetation cover over a particular area, such as when there is a controversy in the literature concerning the palaeovegetation. Where the only information available from a particular area is regarded by the editors as being very sparse and potentially unreliable, such as through undated cores, this is indicated with the text. Otherwise, the reader can assume that most cores in the area are dated to a reasonable standard, through 14C or other methods.
To give the reader some idea of the distribution of data sources for the LGM (18,000 radiocarbon years ago), which is by far the most contentious of the three past time intervals mapped here, there are indicated the locations of main data sites or 'source regions' containing many data sites. The letters added to the maps indicate the general types of data used; P for plant fossil (pollen or macrofossil), S for sedimentary (sediment grade & chemistry, lake level, diatom etc.) & Z for animal fossil and anthropological evidence. The actual number of data points in each 'source area' is not indicated; the intention is simply to point out the main spatial gaps in the data and to guide the reader into the accompanying literature review in which such details can be discovered.
There is always an element of uncertainty involved in palaeoenvironmental reconstruction, and experts often disagree with one another. It has been necessary here to weigh up each case according to which author presents the most convincing set of evidence and arguments. The justification for presenting any particular vegetation reconstruction in the QEN maps is a complex and subtle process (which can be best understood by studying the main text of this database). The process is much more difficult to define than the tidy set of axioms and calculations in a model output, for example (although the parameters and axioms for a model run and subsequent vegetation reconstruction can often be said to be based on some fairly arbitrary assumptions).
The spatial resolution of the maps
It is evident from the sparsity of available palaeodata (see the main text below) that the effective resolution of the vegetation map reconstructions presented here is fairly low compared with the present-day vegetation maps one might find published in an atlas, for instance. To make allowance for topographic subtleties, the decision has been made not to constrain these maps to a uniform grid of data squares. Hence, one cannot in any simple way quantify their resolution. In fact, the working resolution varies greatly from one area to another, based on the density of sites from which useful data has been obtained. The data resolution of a very well studied area (e.g. late Holocene Europe) is certainly far greater than for an area in which only a few data points are available (e.g. the LGM Amazon). The reader should bear this in mind when studying the maps.
In the drawing-up of these maps, some attempt has been made to allow for topographical trends, with an approximate lowering of vegetation belts being assumed in the ways described from published literature sources and direct consultation with regional experts (see the main text below for the details). Thus for example, for the LGM, areas above 500m altitude (relative to present-day sea-level) are assumed to have been polar desert if they were north of 50N latitude, considering the evidence of severe temperature lowering at these latitudes. In general, where the landscape was a complex mosaic broken up topographically, no patch of uniform vegetation less than 50 km x 50km in extent is demarcated even if it occurs surrounded by a different vegetation type. Instead, the patch is 'absorbed' into the surrounding vegetation and in effect assumed not to exist. For the most part however, the palaeo-data are not even at a high enough density to allow such patches to be detected, so a uniform vegetation type is in any case assigned across the area.
Where a topographical mosaic extents over very large areas, such as in the Andean chain of highlands, a special ad hoc 'mosaic category' is created. For the Andes and Himalayas, a mosaic of montane desert and montane tundra is assumed for the areas beyond the foothills, with a particular percentage representation being assigned to each component of the mosaic (see main text below).
Other mosaic categories which could potentially be created are not used here. Thus for example, although a category for 'forest-savanna mosaic' is included in the vegetation scheme with subsequent mapping attempts in mind, there is enough doubt about its potential distribution, structure and carbon storage potential that the editors have for the present considered it better to map such areas as either forest or savanna. Swamp categories of vegetation, although they occur to some extent almost everywhere, are also not included in the maps here because they generally represent a sub-component scattered through the main vegetation type. The only exception is the very large uniform area of swampland close to the Hudson Bay in Canada.
Geographical boundaries in relation to climate
In all published palaeovegetation maps based on point sources of data, extrapolation or interpolation based on knowledge of present-day vegetation-to-climate relationships has been important. Each piece of data relating to climate or palaeovegetation provides a basis from which to deduce the climate/palaeovegetation for other adjacent areas. To some extent, the maps presented here are based by proxy on the vegetation-climate relationships that individual contributing authors have assumed, and on the pattern by which they feel climate would have varied across the region. Such factors are difficult to control for and to describe, for the authors who produced each individual palaeovegetation map often do not themselves state the assumptions which they are using. However, where such information is provided, it is included in the review below.
The boundary between tropical and temperate vegetation types is reconstructed on the basis of palaeotemperature estimates from various sources (see below). The distribution of the tropical-temperate boundary is assumed here to follow the minimum monthly temperature limits detailed in the vegetation scheme at the end of this paper. In fact, the precise position of the demarcation of 'tropical' from 'temperate' vegetation is fairly arbitrary if one looks at the problem objectively (see detailed discussion of this point in Adams 1993), but there is no doubt that there is a general gradient in vegetation ecology and composition as one travels away from the equator. It is always necessary to draw the tropical vs temperate line somewhere, and in the maps here the boundary for most tropical vegetation types is set at a coldest mean monthly value of 10°C. This follows the general correspondence between global present-day poleward vegetation boundaries and the present-day temperature isotherms (e.g. see the Times Atlas [Times 1992], which presents good general maps of both temperature and vegetation). For tropical rainforest vegetation, a somewhat higher temperature limit of 15.5°C (as the mean temperature of the coolest month of the year) seems to correspond well to the map boundaries drawn by various authors (Prentice et al. 1992, Prentice et al. 1993, also see discussion in Adams 1993). There is in fact no objective way of distinguishing a 'tropical rainforest' flora from a 'warm temperate forest' flora (Adams 1993); anyone who studies them objectively where they are contiguous in the field finds that the two simply merge into one another, both taxonomically and physiognomically. This is the case in the forest regions of south-east Asia, for example (Walter 1971). Where plant fossil evidence on forest composition is available, it is assigned here to a vegetation category according to the opinion of the individual authors who first published and interpreted the data, on the assumption that they have a fairly representative understanding of what does or does not constitute a 'tropical rainforest' flora. The formalised temperature limit is only used here as a substitute for direct floristic evidence, where it is lacking, and in order to interpolate between data points yielding plant fossil data.
For the boreal-to-temperate transition, a more generalised definition based on the prevalence of cold-climate conifers is used here as the basis for demarcating the boundary, and a similar floristic type of definition is used for tundra as distinguished from temperate steppe.
In drawing the palaeovegetation maps presented here, the general assumption is made that the overall pattern of isotherms remained approximately the same, but was lowered (this seems generally reasonable with the exception of certain areas close to ice sheets or where large shelf areas had appeared; all GCM reconstructions give generally similar qualitative spatial patterns to those existing at present, albeit shifted quantitatively), and that at around the outer boundaries of the tropics the temperature lowering at the LGM was 5.5°C (= 10°F.) Thus for the majority of vegetation types, the tropical-temperate boundary is moved equatorwards from the present 10°C (50 °F) isotherm for the coldest month, to what is presently the 15.5°C (60°F) isotherm.
Another important factor to include in extrapolating from scattered data points to produce a palaeovegetation map, is the likely pattern of variation in precipitation. If one finds evidence of a particular vegetation type having existed at particular site in the past, from this one can come up with a rough estimate of precipitation for that place. From this one can extrapolate the possible pattern of rainfall across a wider area, or if more than one data point is available interpolate between these. Extrapolation/interpolation of past rainfall patterns from proxy data points can be carried out using a numerical model (which may be coupled to a GCM), and although this brings with it an element of consistency this is in itself no guarantee that the model is correct.
Here, limitations on the availability of funding have necessitated a more ad hoc approach. The approximate relative precipitation patterns are generally assumed to have varied much as they do today (the present rainfall distribution being taken from the Times Atlas and other regional climatological maps), but to have been shifted in quantity. The amount by which rainfall shifted is dictated by the indications from regional sources of terrestrial palaeoevidence, except where there is evidence to the contrary. Assuming this, the boundaries of palaeovegetation types are extrapolated across from areas where relevant data have been found, assuming a broadly similar distribution of maxima and minima to those which occur today. In fact, there is good evidence that even during the LGM (see the main text of the database below), most areas that are relatively moist today tended to be relatively moist during the LGM (even if shifted substantially overall to drier and cooler conditions, they were still generally moister relative to their surroundings). Hence, extrapolation along these principles seems perfectly reasonable in most areas (though not all; see the database below for exceptions). Further case-by-case details of the extrapolations used in constructing the palaeovegetation maps can be found within the main body of the text, below.
No-analogue vegetation types
Mapping of past vegetation must have at least some reference to the present world concepts of vegetation types if the maps are to be meaningful. It is true that in many areas during the LGM, the vegetation combined species that normally grow well apart from one another in separate vegetation/ecosystem zones. The most striking example is the steppe-tundra (Tallis 1990) which brought together species of modern-day steppe and tundra vegetation respectively. To a lesser extent, the same is true of much of early and mid Holocene vegetation of the world. The possible reasons for why such unfamiliar combinations of species occurred in the past are many and varied. Shifting combinations of climatic parameters may have produced climates which suited the independent requirements of species that do not presently overlap in distribution. Changed atmospheric CO2 levels in the past might also have produced unfamiliar combinations of species. During the lower Holocene, disequilibrium in colonisation on poorly developed soils and 'sampling error' in population expansion from refugial populations may likewise have produced many of the no-analogue associations (Tallis 1990).
Whether these vegetation types have any sort of present-day analogue is sometimes a moot point (sometimes the past vegetation seems to resemble unfamiliar combinations of species which do occur in the present-day world, but are much more restricted in distribution). In any case, it is sometimes difficult to decide whether these no-present-analogue vegetation types were similar enough to any presently widespread vegetation type to be mapped and treated as such. In the case of the steppe-tundra of the LGM, the difference from present vegetation is suggested here as being too great to fit into any of the standard present-day vegetation categories. Instead, a separate category is created especially for it. In other instances, such as the Podocarpus-containing lowland tropical forests of the LGM (Colinvaux & Bush 1988), the need to produce at least some sort of working definition in present-day terms has overrided the differences from the present species composition and ecology which undoubtedly existed. In such cases, the no-analogue combinations of species are pointed out within the discussion text below.
Such 'lumping' of past vegetation into present-day categories is a necessary evil which may ultimately be avoidable as more and more detailed knowledge of relationships between species composition and ecological parameters such as carbon storage accumulates in the future. However for the present study the 'lumping' process is to some extent unavoidable.
Drawing the vegetation boundaries
In the real world outside of vegetation maps, most vegetation types either exist as mosaics or merge gradually into one another, so in many ways the decision as to how to divide up the world's vegetation for mapping purposes is quite arbitrary. Having said this, one has to draw the dividing lines somewhere and it is best to try to do this in a way that corresponds to pre-existing perceptions amongst ecologists on how the world's vegetation should be divided up.
It is probably true to say that the current situation in terms of presentation of palaeovegetation data in the Quaternary literature is one of anarchy. There are few real attempts being made amongst different authors to specify the exact character of vegetation in structural terms, even when the palaeo-data can allow a reasonable guess. Nor is there much attempt to use even the vaguest of categories which can be cross-compared between different regions, with local names such as 'fynbos' 'sage brush' or 'maquis' being used to the exclusion of more general ones such as 'evergreen scrub'.
In the absence of any generally agreed vegetation categories, and also with the necessary heavy reliance on proxy data (for example, erosion rates) as an indicator of palaeovegetation, it has been necessary to interpret the available data here in a fairly ad hoc way. In some cases, in the absence of satisfactory data, the positions of certain vegetation boundaries drawn for the past must be regarded as very arbitrary and simply a necessary provocation to further thought (when this is the case, it is made clear in the accompanying literature review, see below).
There is increasing fossil evidence (described at many points within this literature review) of past vegetation that was compositionally and perhaps structurally quite different from anything existing today. This presents problems for anyone attempting to provide vegetation maps for the past world, for there are no present-day analogues to describe them in terms of. In the case of the very unusual steppe-tundra vegetation of the LGM, a special no-analogue map category has been created. In other instances it has been necessary to 'slot' the past categories into what seem to be their closest present-day equivalents. Wherever such difficulties are encountered, they are discussed within the text.
Despite all the difficulties, one should still regard these mapped vegetation boundaries as having enough meaning to be worth presenting in a preliminary sense for broad-scale studies, as a significant improvement on previous efforts. It is hoped that by acting as a focus for further thought, and perhaps as a basis for a standardised classification system for palaeovegetation, these maps can be improved upon little-by-little over time.
Expressing the uncertainties in the vegetation map reconstructions
Because of the diverse and partly qualitative nature of the sources consulted, it is exceedingly difficult to express any simple measure of the amount of confidence to be placed in the patterns of vegetation distribution drawn onto the maps presented here. The reader must study the main literature review sections below to gain an idea of the detailed degree of uncertainty or confidence which has been placed in the data for each sub region by those who work in that area, and from this judge for his or her self whether the map reconstructions presented here are likely to be accurate. Unfortunately, the world is not reducible to a statistical confidence limit, at least not in the same way that one could produce a confidence limit for a set of laboratory carbon dated analyses. However, it is still possible to give an overall impression of how seriously one takes the inherent uncertainties in the information.
The Olson et al. scheme
Even in the present-day world, confusing, local definitions and names of palaeovegetation types often muddle the cross-comparison between different regions, so it is important to try to fit vegetation reconstructions into broader, more generally familiar categories. In compiling these maps, an effort has been made to conform as closely as possible to the well-known and widely cited scheme developed by Olson et al. (1983) that was used to present a database of present-day world vegetation cover.
The Olson et al. scheme may be useful, but it is far from perfect. When one examines the descriptions of vegetation categories given by Olson et al. in the accompanying literature for their map, it is obvious that their perceptions of what is or is not to be included within each vegetation type around the world are very loose ones (this looseness has been necessitated by the diverse nature of the published data that they were drawing on for compiling their maps; it often seems that no two sources ever use the same system for categorising vegetation). The palaeovegetation maps that presented here must in effect take this looseness even further, partly by necessity of the indirect nature of the evidence that one must rely upon in reconstructing past vegetation, and partly due to the vagueness with which descriptions of palaeovegetation cover are presented by different authors.
The following categories are used here for the accompanying vegetation maps, based broadly on the Olson et al. vegetation scheme, but heavily modified in certain respects. In particular some categories have been merged and others have been added, including some vegetation types for the past that have no close analogue in the present world. However, for the most part placed past vegetation types within present-day categories, because we do not believe their structural and floristic differences to have been great enough to warrant separate biome-level categories. A more detailed and complete rendering of the vegetation scheme that has been used as the 'ideal' for the mapping is given in the Appendix, at the end of this paper. Of course, this ideal cannot be followed closely even for mapping the present-day world, because the necessary information on precise vegetation structure and cover are simply not available. It is simply a hypothetical goal, to try to aim for if not necessarily to achieve.
Although time is of course a continuum, it has been necessary here to concentrate the reconstructed scenarios around particular slices of time that seem particularly significant in relation to the processes taking place in the global system. The perception of what is or is not a particularly 'significant' time slice depends on what aspects of the global system one is interested by, and in this respect the choice has been biased by an interest in the relationship between vegetation ecology and carbon storage on land. Another criterion has been the relative 'stability' of vegetation at each time slice, avoiding trying to make anything of the confusing and blurred picture from times such as the late glacial/earliest Holocene in which the world seems to have been a mosaic of landscapes and vegetation, all at various stages of rapid change. At such times, a slight inaccuracy in the dating or a gap in the data could make a vast amount of difference.
The last glacial maximum. This earlier time-slice, the last glacial maximum (LGM), is placed at around 18,000 years ago in radiocarbon years. Note that due to changes in radiocarbon abundance at different times in recent Earth history (due to changes in the carbon cycle or in cosmic ray flux) the ages given by radiocarbon dates may be wrong. It is now thought that 18,000 years ago in radiocarbon terms corresponds to about 20,000-21,000 years ago in 'real' years (Bard et al. 1990).
The LGM is generally defined as the stage during the last glacial cycle at which the greatest mass of ice was present on Earth, showing up in ocean carbonates as a peak of 18O. It is also thought of as being the time at which at which other components of the ocean-atmosphere system were at their most 'glacial' (e.g. lowest global temperatures, lowest atmospheric CO2 concentration, and apparently greatest aridity in many continental regions) (Crowley & North 1991).
In fact, there are numerous signs that not all attributes and processes reached their peak of 'glaciality' (in the sense of maximum cold, maximum ice extent, and maximum difference in water balance relative to the present) at the same time during the last general glacial phase; example Colinvaux et al. (1989) suggest that the lowest temperatures and maximum glacier extensions in tropical uplands may have occurred several thousand years before those at higher latitudes. The most recent evidence is suggesting that in fact the LGM in terms of the 'ice maximum' was not the maximum in terms of global cold and aridity. This extreme in terms of climate may have coincided with Heinrich Events (surges of icebergs) that affected the whole of the planet's climate, and that what has been thought of as single glacial maximum may actually be two fairly similar intense cold episodes, separated by a short-lived but slightly milder episode that corresponds to the maximum land ice extent itself. If this interpretation is correct, the global cold and aridity maximum described here may need to be shifted slightly in time to lie both just before and just after the LGM sensu stricto. The 'LGM' as described here should be seen as the 'globally most arid and cold conditions to have occurred simultaneously during the last 20,000 radiocarbon years' and not in terms of 'greatest global ice extent'.
In cases where there is no proper dating evidence available, it is too easy to assign all ambiguously dated 'cold' and 'arid' events in the palaeoenvironmental record to the time around LGM, simply because well-dated evidence from elsewhere shows cold or aridity at that time. Having said this, in a situation where one has to scramble for tenuous clues as to the true nature of palaeoenvironments at around the LGM, one can also intuitively understand that this 'circular' search for relatively poor quality evidence to reinforce a picture based on good quality evidence might lead to the best overall predictions in the longer run. For example, if one has a whole series of undated lake cores from across a broad region, each showing a single marked arid event, it seems reasonable to suggest in a tentative way that this arid event occurred synchronously throughout the region. If a neighbouring region has dated information that also shows a single well-marked arid event, a reasonable preliminary hypothesis (in the absence of any other information) is that the arid events in both adjacent regions occurred simultaneously. Carbon dating of evidence from around the world suggests that in most areas (though by no means all!) this approach would in fact be correct in identifying the conditions that occurred simultaneously at approximately the LGM*.
It is important to bear in mind that in some areas for which there is continuous well-dated evidence, climates only a few thousand radiocarbon years before or after 18,000 14C years ago were very different from the LGM itself, often being much moister (and as mentioned above, the 'true' LGM in terms of land ice extent may have been a brief milder phase between two closely placed and more extreme cold events). The ambiguous use of the word 'glacial' in the literature to include periods thousands of years before and after the LGM often confuses readers, who tend to assume that this refers to the LGM itself. This is not the place to resolve fundamental issues of stratigraphy, but authors should perhaps take note of this problem. They could perhaps take care to specify when they are talking about the last glacial period 'before', 'after' or 'at around' the LGM.
*In some cases this method could confuse the Younger Dryas with the LGM, but then again the vegetation conditions of the two periods seem to have been generally similar in terms of direction of vegetation change, but with the LGM being markedly more extreme in most areas.
In this review, wherever undated or ambiguously dated evidence is cited, it is pointed out as such and treated cautiously, and presented as only possibly relevant to the overall picture. Certainly, in some areas where other evidence is lacking it has been allowed it to influence the vegetation map reconstructions presented (this fact is explicitly stated wherever it is used). Of course, when relying on review sources or expert overviews one cannot always be sure what quality and types of evidence individual experts are willing to accept as convincing. For the most part it is only possible to report on their final recommendations or conclusions about palaeoenvironmental conditions, and not all the detailed contortions of the decision-making process which allowed these experts to arrive at their conclusions.
The early Holocene, 8,000 radiocarbon years ago. This second time slice is intended to represent the world whilst it was reaching the final stages of deglaciation (some remnants of the North American ice sheets still remained but they were melting fast; Denton & Hughes 1982), as vegetation and soils were approaching a relatively steady state after the rapid changes that had taken place during the previous millennia. In terms of 'real' years, the 8,000 radiocarbon year timeslice was probably nearer 9,000 years ago, though there are still considerable uncertainities.
There are some indications that a sudden drought episode set in northern and eastern Africa (and possibly elsewhere) either just before or just after 8,000 14C years ago (Gasse & van Campo 1994), but this time slice is intended to represent the moist conditions immediately prior to this particular climate switch. Although the broad climatic background was much more similar to that of today, relative to the LGM, climates in most areas do seem to have been noticeably different from those of the present. There is less ambiguity in the dating for this time, and data are generally much more abundant, so there is greater confidence that the biome reconstructions presented here are reliable.
In many areas of the world, the indications are that the 8,000 14C years ago vegetation was already quite similar to today's, at least in the sense of broad biome structure (e.g. temperate deciduous forest as opposed to closed boreal forest) although not necessarily in terms of detailed community composition (e.g. the relative abundance of particular tree species). Wherever published review sources agree in emphasising that the vegetation was broadly similar to that existing today, little emphasis in our review is placed on the nature of the evidence used to justify this conclusion, although it is given some cursory mention. Instead, the emphasis is on the nature and validity of those sources of data apparently indicating a significant difference from the present-natural vegetation. For the many regions where there are no indications of conditions being significantly different (on the available evidence, which is often not of ideal quality), we have merely filled in 'present-natural' vegetation boundaries as the most reasonable surrogate. Knowing that many parts of the world were not so very different from today, it seems better for the sake of global-scale studies to make a reasonable but conservative guess rather than leaving swathes of the land surface unclassified (e.g. it would be unrealistic to make a global biogeochemical study that did not include the whole land surface; one has to put some values into the sums). It may eventually turn out that significant differences were present, in which case the mapped vegetation boundaries will need to be altered accordingly. For the time being, it is important that the user of the maps consults the accompanying literature review for each region to know just how much weight to give to the reconstructions within each region.
Although agriculture was certainly present in several parts of the world by 8,000 years ago, it generally seems that it was not a significant modifier of vegetation on anything more than a localised scale (e.g. see Tallis 1990, and discussions in the text below). Thus, on these maps there is no attempt made to mark out any areas of croplands or other land cleared for agriculture or settlement.
The mid-Holocene, 5,000 years ago. By around 5,000 14C years ago, it seems that climate and potential vegetation patterns were significantly closer to those of the late 20th century than those that existed 8,000 14C years ago. Hence, this time interval provides a state relatively similar to the 'present potential', the difference being that it actually existed and is not a hypothetical state in the way that the present-potential is.
There certainly would have been important differences from the vegetation patterns that would potentially exist at present (e.g. the Sahara was much more densely vegetated than now at 5,000 14C years ago), and there has been an attempt made here to map those differences that appear in the literature and have been pointed out by contributors to this database. However, in many areas, it seems that the world's vegetation at 5,000 years ago was very similar to the 'present natural' vegetation that is mapped in atlases and ecology textbooks (although some slight differences are usually apparent). This is hardly surprising in some ways, for the idea of what was the natural state of the world's present vegetation is sometimes inferred from pollen records of whatever vegetation seems to have existed before agriculture caused major landscape modifications. However, even apart from this circularity it does seem that over much (but not all) of the Earth's surface, climates around 5,000 years ago were very similar to those of today (e.g. Crowley & North 1991, Williams et al. 1993), and the vegetation that existed was very similar in character to that which now survives in regions not subject to intense agricultural activity, and in remnant pockets within otherwise mainly agricultural areas.
By around this time, farming populations or cultures were spreading rapidly outwards across various regions, probably from independent centres of innovation (Tallis 1990). Although the vegetation of the world had undoubtedly been influenced in some important ways by preagricultural human activities (such as burning of semi-arid vegetation by hunter-gatherers) since well before the LGM, clearance for agriculture seems likely to have been of relatively localised importance in terms of overall vegetation ecology and structure at 5,000 years ago. There certainly seem to have been some 'false dawns' for extensive agricultural impact seen in the environmental record; for example, the elm decline observed around 5,000 years ago in north-western Europe (and earlier in southern and central Europe) was initially ascribed to a sudden and widespread increase in animal herding, but it now seems far more likely that disease and/or climate was largely to blame (Rackham 1980, Huntley & Birks 1983).
There were some exceptions to the generalisation if one focuses in on specific areas; the first irrigation schemes in Mesopotamia began at around 5,000 years ago. In southern Greece, some large areas of cultivation were probably present by 5,000 years ago, but even here the main phase of forest loss and soil erosion began about 1,000 years ago later (Tallis 1990). However, concentrating on general patterns across broad areas of the Earth's surface rather than just taking the few isolated exceptions, it seems that the overall impact of agriculturists on vegetation at 5,000 years ago was still fairly light. Tallis (1990) concludes that pollen diagrams from around the world do not tend to show a significant deflection attributable to agriculture until around 4,000 years ago, even in such agricultural 'cradles' as the middle east. This view seems borne out by my own survey (the text below) of the literature and of the opinions of palynologists. In many regions of the world, the first signs of extensive (rather than simply sporadic) modification of landscapes by herders and farmers appear between around 4,500 and 3,000 years ago, and these signs have increased towards the present.
The climatic shift that appears to have caused the 'elm decline' in Europe was quite possibly part of a much more extensive episode of cold or aridity that occurred around this time. Recent evidence from the Greenland ice cores shows a sudden fall in atmospheric methane at around 5,000 years ago (Chappellaz et al. 1993), probably reflecting some temporary decrease in biological activity in tropical or boreal latitudes. In the context of this event, the maps here should be regarded as representing the system immediately before the onset of the cold and/or drought event.
The 'present-potential' vegetation. At the present, large areas of the world have an extremely modified and fragmented semi-natural vegetation cover (e.g. Olson et al. 1983, Milanova et al. 1994) that makes it difficult to know what things could have been like if agriculture had not arisen, and if human populations had not exploded as they have done. Over large areas of Europe, there is nothing but cropland with barely a tree in sight. Over much of India and Bangladesh, the vegetation is so influenced by crop-growing, grazing and fire that it is completely a matter of speculation as to what it would otherwise be like (Milanova et al. 1994). Likewise at the fringes of the rainforest zone of Africa and Amazonia, one sees domestic cattle grazing and regular fires set to provide them with fodder. Yet everywhere in atlases and textbooks one sees maps of the 'present potential'; the vegetation that was 'meant' to be here but which does not in fact exist.
What is one to make of this? Can we really know what the present vegetation would be like if human history had been different? The answer is, both 'yes' and 'no'. In many areas, the vegetation seems almost untouched by human influence, or at least we see no obvious intervention occurring. In many other areas, there are enough remaining fragments of what appears to be a sort of semi-natural vegetation to allow plant geographers to suggest what the vegetation would become like if humans were to suddenly be removed from the face of the Earth. However, there are also many places that are obviously too heavily grazed, burnt, trampled and ploughed for anyone to know how far things would change without these influences. Sometimes the only way that plant geographers can support their opinions is to refer to pollen evidence from before the time of agriculture, making a tangled web of the real past and the imaginary present.
What is the point of presenting such 'present-potential' maps here if one knows that they represent a non-existent world? Their benefit comes in the need for inter comparisons. At a basic level, the widespread familiarity of 'present potential' maps - for all their faults - allows any reader familiar with seeing such maps to use them as an anchor point for comparison with the world of the past. There are also other pragmatic reasons for using them. Very often within the Quaternary literature, individual authors refer to the boundary between one vegetation type and another having shifted by a certain amount under Holocene or LGM conditions. To know where the starting point is, one has to connect up the pieces of remnant vegetation, then allow for where the authors think the natural vegetation boundary would have lain, and from this draw a shifted boundary for the past. In other areas, the necessary early or mid Holocene vegetation evidence is completely lacking, so that the best one can do in the reconstruction (other than either leaving the area completely blank or assigning a completely arbitrary vegetation cover) is to throw backwards in time a present-potential vegetation map and hope that it is not too inaccurate, whilst making some alterations to allow for broad climatic shifts that seem to have been present. This is what the editor has been forced to do in some places here by the inherent gaps in the data.
Hence, for the sake of comparison, and honesty, the editor has presented here a vegetation map of the 'present-potential' world, representing a compilation of the Olson et al. present-actual maps and various regional present-potential vegetation maps (e.g. the now famous vegetation maps of P.E. Preston-James in the Times Atlas; 1992). There has been no attempt to indicate which areas are subject to most intense human activity through anthropogenic burning, grazing, forestry, agriculture etc. Milanova et al. (1994) have produced a comprehensive set of maps indicating the approximate intensity of modification by humans, and by consulting these the reader can make up his or her mind as to how much to trust the true accuracy of the 'present-potential' maps presented here. Generally, the greater the human modification at present then the less trustworthy the 'present-potential' reconstruction actually is. Thus, whilst one can be almost certain that the position of the forest - tundra boundary in Canada is not primarily a result of human activity (at least in the position it was in the 1950s, and before greenhouse warming begins to act too strongly), one should strongly suspect that the forest-savanna boundary in central Africa is very much a product of the agricultural human populations that live throughout this zone.
Some other useful web pages
NOAA's National Geophysical Data Center, palaeoclimatology page. Lots of links to useful data sources on the past.
Association of Canadian Palynologists.
Tropical Geomorphology Newsletter. Discussion of geomorphological processes in the low-latitudes. Such information is useful as a palaeoenvironmental indicator.
Suggestions and comments from readers
Any further additions, suggestions or criticisms of the maps, or of the interpretations presented here, should be directed to J.M. Adams or H. Faure. The editors hope that they have cited the opinions of individual contributors as they had intended, whilst taking responsibility for any misunderstandings which might arise from these citations.
Index to the regions
This text is divided up on a regional basis, roughly corresponding to traditional notions of 'the continents', although the detailed choice of boundaries for each region is by necessity fairly arbitrary. The order of the regional treatment is as follows;
Europe eastwards to the Urals, and also Asia Minor
Northern Eurasia Mostly Russia east of the Urals
Southern and Eastern Eurasia The middle east, and from the central Asian desert southwards and eastwards to Malaysia/Indonesia
Africa, including Madagascar, Arabia and the Levant
Australasia, including Australia, New Guinea and New Zealand
North America comprising the USA, Canada, Greenland and 'Beringia', with Mexico and the Caribbean
South and Central Americawith Mexico and the Caribbean
VIEW NEW SET OF PALAEOVEGETATION MAPS (PRELIMINARY)
Jump to the List of References (separate document)
As was made clear in the introductory section, this data summary has only been made possible by the collaboration of a large number of experts from around the world, in contributing information and opinion to the Quaternary Environments Network. The foresight and open-mindedness of all those who have helped us is greatly appreciated. We hope that we have been able to do their work justice. Anyone who feels that we have misquoted or misunderstood his or her work should contact us and we will act upon the problem immediately.
The task of compiling this information was greatly aided by travel grants from the British Council and from the Laboratoire de Géologie du Quaternaire. Also of considerable importance has been the generosity of Professor A.G. Goudie, Head of the School of Geography of the University of Oxford in making its facilities available to J.M. Adams during the long cold, arid phase in which this work went unfunded. The organizational skills of Lilliane Faure have also been essential in making possible the continuance of the project through such difficult times.