Future uses of pollen analysis must include plant macrofossils
Before the development of pollen analysis following the seminal paper by von Post (1916), plant macrofossils were used to investigate Quaternary vegetational history. Basic aspects of Quaternary vegetational and climatic changes were established, including interglacials, the full-glacial, and the late-glacial (e.g. by Clement Reid, Hartz & Milthers, Jessen, etc.). However, quantitative pollen analysis was more immediately appealing and successful, particularly in the hands of Johs. Iversen, and the use of plant macrofossils faded during the 1950s.
Pollen analysis proved to be excellent for reconstructing floristic and vegetational history, and hence past climate changes, at a variety of spatial and temporal scales. Here we will restrict our discussion to the full-glacial, late-glacial, and Holocene time periods and to temperate, arctic, and alpine areas. In this present age of computer models and the need for quantitative palaeovegetational and palaeoclimatic reconstructions, we ask if pollen analysis alone can provide useful, adequate, and reliable data in this context.
The principles of pollen analysis assume that there is an interpretable relationship between the pollen deposited in the sediment and the vegetation that produced it. There are many limitations to pollen analysis that have been long known, and are most recently discussed by Ritchie (1995). Pollen identification has limited taxonomic resolution, and many ecologically important taxa have indistinguishable pollen grains. The composition of the pollen rain is exceedingly complex, comprising different proportions of taxa ranging from species that produce small amounts of locally dispersed pollen to other species that produce huge quantities of wind-dispersed pollen. Its source area is also complex, and although generalised models have been developed, every site probably differs from every other, both in space and in time. In temperate forest situations, pollen analysis functions adequately at producing evidence of major palaeoecological change. However, in treeless situations (arctic and alpine), local pollen production is usually low. The distant pollen rain increases in importance, and becomes over-emphasised in floristic and vegetational reconstructions. Misleading interpretations can easily be made. The problem is particularly acute in tree-line situations, where non-local tree pollen can mask the local pollen production and make interpretations of tree-line changes exceedingly difficult. The problem is not adequately addressed by the use of pollen influx (grains cm2 yr−1) due to complications and uncertainties arising from sediment dating, sediment redistribution in a lake by focusing, bioturbation, or mass movements, and difficulties in estimating pollen concentration. Pollen analysis can also give misleading interpretations in arctic and alpine areas, and in treeless full- and late-glacial environments. As the climate changes occurring around the last glacial termination are of intense interest at present, it is very important to be sure that correct palaeoclimatic reconstructions are used in the spatial and temporal framework of quantitative climate data that is needed for the analysis of causes and processes of the deglaciation. Interpretations from pollen analysis have often been biased by its inadequacies in such situations, and palaeoclimatic reconstructions and their use in models should be evaluated before being used uncritically.
Plant macrofossils can enhance the information from pollen analysis in treeless situations. They can often be identified with greater taxonomic precision than pollen, and they tend to be much more locally distributed from their source. Therefore, they reflect the local flora and vegetation, and hence palaeoclimatic conditions. Macrofossils can often be frequent of taxa that produce very little pollen or have poorly preserved pollen, and of taxa that do not produce pollen, particularly ecologically restricted moss taxa and aquatic Charophyceae. However, macrofossils are not usually produced in such abundance as pollen, and larger volumes of sediment are required for their study. Because of their variable local production and dispersal, their representation is very difficult to quantify beyond general terms. Nevertheless, their representation is linked closely to the vegetation that produced them, leading to reconstructions at an ecological or site scale and to the possibility of interpretations of vegetation change and dynamics. In addition, terrestrial plant material is ideal for AMS 14C dating, overcoming problems with bulk sediment dating such as imprecision because of the amount of sediment needed from a core, old carbon effects, contamination by roots, etc.
We give some examples of how plant macrofossils have modified and corrected interpretations made from pollen analysis alone within our present context.
Norwegian late glacial
Following the Danish example, late-glacial pollen diagrams from Norway have been interpreted as pioneer vegetation developing into more or less open birch forest in the interstadial (Allerød), which was then destroyed completely or partially during the cold Younger Dryas interval ( Birks, 1994a). A synthesis delimited vegetation regions and the ecotones between them, and hence reconstructed regional climate variation during the late glacial ( Birks, 1994a). Macrofossil analyses from several of the pollen sites revealed surprising discrepancies in this pattern, particularly in the Allerød. Where birch forest was thought to be best developed in southern Norway ( Paus, 1989), macrofossils showed only the presence of dwarf birch (Betula nana). Tree birches only arrived locally at the beginning of the Holocene ( van Dinter & Birks, 1996). At the coast where the late-glacial birch forest was thought to be open ( Mangerud, 1970; Paus, 1990), macrofossils suggested that the predominant vegetation was Empetrum heath and snowbeds dominated by Salix herbacea and that birch trees were absent ( H. H. Birks, 1993). In central Norway pollen analyses also suggested that tree birch was locally present, especially inland ( Larsen et al., 1984 ; Kristiansen et al., 1988 ), but the macrofossil pattern was similar, with snowbeds at the coast at Kråkenes ( Jonsgard & Birks, 1995) and Betula nana in more continental areas inland (H. H. Birks & M. van Dinter, unpublished). The macrofossil results provided evidence of the local vegetation and environment, showing that the interpretations from pollen analysis had been seriously confused by long-distance transported tree-Betula pollen from the south ( van Dinter & Birks, 1996). Therefore, the late-glacial climatic interpretations (e.g. Birks, 1994a) had to be drastically revised.
Full-glacial vegetation of Beringia
Full-glacial pollen assemblages from Beringia are dominated by Gramineae, Cyperaceae, and Artemisia, and have been reconstructed as a productive grassland, ‘steppe-tundra’, or ‘mammoth steppe’ which could support large populations of grazing mammals. However, studies involving finer pollen-taxonomic resolution and pollen influx measurements suggested another interpretation of the pollen results as a more arctic, less productive tundra vegetation ( Cwynar, 1982). The controversy became intense and could not be resolved satisfactorily because of the lack of taxonomic differentiation in the main pollen types, the lack of modern analogues for the pollen assemblages, and the lack of information on the source area and scale of representation of the pollen rain (e.g. Guthrie, 1990; Colinvaux, 1996; Ritchie, 1995). Two plant macrofossil studies have disentangled the controversy remarkably easily. Using macrofossil and Coleoptera analyses from sediment cores from the full- and late-glacial terrestrial Bering Land Bridge, now under the Bering Sea, Elias et al. (1996, 1997) showed that low-lying regions were occupied by mesic tundra with shrubs and boggy pools rather than grass steppe. In the second study, direct vegetation evidence was provided by macrofossil analysis of a 18,000 year-old fossil loess soil buried and preserved in situ by tephra on the Seward Peninsula bordering the Bering Land Bridge ( Wolf & Birks, 2000). The vegetation here was open sedge-, herb-, and moss-dominated tundra typical of dry soils in the arctic today, and had no resemblance to ‘steppe-tundra’ or ‘mammoth steppe’. We have to conclude that the arguments about the existence of a no-analogue ‘steppe-tundra’ biome have been based on the composition of the long-distance-transported pollen rain that dominates the pollen spectra because of the very low pollen production by the local vegetation.
Comparison of modern and fossil pollen assemblages has often revealed ‘no-analogue’ fossil assemblages (e.g. Anderson et al., 1989 ; Overpeck et al., 1985 ; Whitlock et al., 1993 ). The ‘no-analogue’ pollen assemblages are presumed to have originated from ‘no-analogue’ vegetation, and even a ‘no-analogue’ environment and climate. In many instances, contemporary plant macrofossil assemblages yield perfectly good analogues with modern vegetation types, implying that the ‘no-analogue’ pollen assemblages are little to do with the vegetation of the time but have resulted from the prominence of far-distance pollen components such as Artemisia, Betula, and Pinus in situations of low local pollen production.
Deciduous trees in the late-glacial of Minnesota
Late-glacial pollen diagrams from the mid-west USA are dominated by Picea pollen, but consistently contain smaller amounts of pollen of thermophilous deciduous trees (e.g. Quercus, Ulmus, Ostrya-Carpinus, Acer, Corylus) (e.g. Wright, 1971). It has been debated but not decided if this pollen is long-distance transported from south of the late-glacial Picea-Larix forest, or whether the thermophilous trees were locally present in small quantities. Macrofossil studies (e.g. Wright & Watts, 1969; Baker, 1965; Birks, 1976) failed to find any remains of these species, although the fact that they have been recorded in early Holocene sediments in Iowa indicates that they can be preserved and identified from sediments ( Baker et al., 1996 ). The macrofossil evidence for local late-glacial vegetation in Minnesota is consistent with a boreal type of spruce forest indicating a rather cool climate that would be unsuitable for thermophilous tree taxa. It is likely that the relatively low pollen production of Picea ( Watts, 1979) has allowed the representation of thermophilous pollen types transported from more southern regions. Macrofossil studies are needed in likely refugia of thermophilous trees to test this hypothesis.
Pollen analysis is a blunt instrument for detecting changes in tree lines and limits because pollen is easily transported from below or above and the record becomes blurred. The local representation of macrofossils sharpens the pollen-analytical tool, and a combination of the techniques can lead to rather precise evaluations of tree-line positions. A good example is the study by Jackson (1989) in the Adirondack Mountains, eastern USA. Macrofossils were able to pinpoint the local occurrence of the tree taxa, and in addition, differentiate within a genus (e.g. Betula, Pinus), and show how different species occurred at different times. In Norway, the tree line is frequently formed by Betula pubescens or B. tortuosa, whose pollen is difficult to distinguish from B. nana. Unpublished macrofossil and pollen studies by H. J. B. Birks, H. H. Birks, S. M. Peglar, and N. H. Bigelow have demonstrated that tree birch extended above the present tree-line from the earliest Holocene in southern Norway, but only after c. 8500 14C BP in northern Norway. The tree-line descended to its present altitude at 1000 14C yr bp (southern Norway) and around 3000 14C bp (northern Norway). The two regions seem to have different climatic histories during the Holocene. These local site-changes are not detectable at all in the pollen diagrams. In north Sweden, Barnekow (1999) elegantly demonstrated a similar tree-line pattern to northern Norway using macrofossils in an altitudinal transect of sites.
The spread of trees after deglaciation, earliest colonization, and successional forest development can be traced using macrofossils. Watts (1979) demonstrated the expansion of shrub birch before tree birch, and the coexistence of Picea and Pinus in the early Holocene of eastern N. America, developments that could not be interpreted from the pollen record alone. In southern Norway, the latitudinal movements of tree species are being traced during the Holocene using macrofossils to ascertain their local presence along major climatic transects (H. J. B. Birks, N. H. Bigelow, unpublished). The input of macrofossils to tree-line studies clearly needs to be explored further (e.g. Wick & Tinner, 1997).
The hypothesis that plants survived the glacial periods on nunataks and coastal refugia in Norway sparked much heated debate ( H. J. B. Birks, 1993). Pollen spectra from full-glacial sediments were unable to provide definitive evidence either way. However, with their higher taxonomic precision and local representation, plant macrofossils demonstrated that many of the plants in question were present beyond the limits of the ice-sheet during the last glacial period, and, as they occurred in different places at different times, they probably had good dispersal abilities and could follow its fluctuating margins, colonising disturbed unstable ground ( Birks, 1994b). Although direct evidence from the nunataks themselves is lacking, it is unnecessary to propose these as the sole glacial refugia because the plants were present elsewhere. Most of them are arctic and mountain plants of open habitats that are restricted today by habitat loss due to forest and soil development during the Holocene.
These are a few examples where macrofossils have provided more precise and accurate interpretations than could be made from pollen analysis alone. With the current interest in reconstructing past biomes and climates, it is essential that palaeovegetational and palaeoclimatic estimates are quantitative and as precise as possible. Palaeoclimatic estimates are needed for hindcasting validation of general climate models, so that they can be tuned to give climate forecasts as realistically as possible. They are also needed for modelling past global circulation in attempts to understand global climate changes associated particularly with the last glaciation, its termination, and the Holocene. The provision of such palaeoclimatic data depends on the accuracy of reconstructions from individual sites. We have outlined some cases where pollen-analytical data are seriously misleading and plant macrofossils resolved the situation. In addition, other multi-proxy evidence at individual sites (e.g. Birks et al., 2000 ) can provide and independently validate reconstructions but we have no space to discuss multi-proxy studies here. If a synthetic edifice is to be constructed from palaeoclimatic or environmental data, it is essential that the basic data be of high quality from individual sites, as these are the building bricks on which the edifice will entirely depend.
It seems that plant macrofossil studies are due for a revival in order to refine and correct palaeoenvironmental interpretations made from pollen analysis alone. Large regional reconstructions of past vegetation and climate are being made using international pollen data-bases. It is essential to be aware of interpretive traps in the use of pollen data alone, especially in areas and time-periods of low local pollen production, before making maps of past biomes and climates. Macrofossil analysis is not a difficult technique, but it requires expertise in botanical taxonomy and morphology, together with experience of individual species requirements and vegetation ecology. This is sadly lacking in young botanists today, because it is rarely provided by Universities in this time of running down of ‘unfashionable’ classical botany, herbaria, and teaching of plant taxonomy, systematics, morphology, and field-based natural history and vegetation ecology.