*Correspondence and present address: S. A. Finkelstein, Department of Geography, University of Toronto, 100 St George Street, Toronto, Ontario, Canada, M5S 3G3 (tel. +1 416 978 3375; fax +1 416 946 3886; e-mail firstname.lastname@example.org).
1Species-level pollen identifications within the genera Acer, Fraxinus and Juglans have the potential to increase the detail available from regional- and landscape-scale studies of palaeo-forest dynamics.
2Data from the North American Pollen Database (NAPD) and the North American Modern Pollen Database (NAMPD) for sites in eastern North America for which species-level identifications have been recorded enabled us to consider pollen taxonomy at high resolution.
3Species within each of Acer, Fraxinus and Juglans have important differences in habitat, functional type and responses to climatic change. Analysis of the modern distribution of these taxa and their pollen rains confirms that species-level pollen identifications provide detailed ecological information, but the lack of distinction to the species level in many fossil and modern pollen studies renders palaeoenvironmental reconstructions incomplete.
4Within each of three selected high-resolution sites, ordinations indicate that individual species of Acer, Fraxinus and Juglans follow different trajectories through the Holocene, showing that analysing only generic categories results in the loss of ecologically valuable information. For example, pollen of Acer rubrum increased in abundance in response to canopy openings in a southern Ontario forest around 500 years ago, while that of A. saccharum declined. Similarly, Juglans nigra pollen percentages increased while J. cinerea decreased at this time.
5Regional-scale comparisons of pollen percentage curves indicate that, despite the uncertainties associated with the low temporal resolution and the chronologies, it is reasonable to conclude that individual species of Acer, Fraxinus and Juglans responded synchronously to palaeoclimatic changes within each region.
6Taking analyses to the specific level shows that different species followed different tracks in their post-glacial history, a fact previously blurred by less resolved taxonomy. This information is meaningful in biogeographical terms, providing much more specific evidence of how the selected trees behaved in the past.
Documenting and understanding major vegetation changes, especially as caused by human activity, is an important goal of ecology. For example, in North America, comparisons of present-day forests with those prior to European settlement have identified important changes (e.g. Bormann et al. 1970; McIntosh 1972; Siccama 1974; Foster 1992; Brisson & Bouchard 2003), although pollen diagrams suggest that at least some of the changes documented in these studies began pre-settlement (Gajewski 1987). Many pollen diagrams, however, show tree taxa only to genus level, limiting the precision of any comparison with vegetation surveys.
Analyses of pollen assemblages in lake sediment cores illustrate the large-scale migration of the major tree taxa of eastern North America during since the last ice age (e.g. Davis 1981b; Williams et al. 2004). Description below the level of genus is, however, generally restricted to genera that are monospecific in eastern North America, such as Fagus (Bennett 1988), Tsuga (Davis 1981b), Abies (Webb 1987) and Castanea (Davis 1981b). Similarly, few local or regional studies have plotted trees at other than the genus level (but see Maenza-Gmelch 1997; Davis et al. 1998). In many other cases, however, some tree pollen has been identified to subgeneric level, although the information was not presented or interpreted. These data have accumulated in publicly available databases, but have not previously been exploited.
Long-term datasets supplied by pollen records are increasingly being used to test ecological hypotheses, and our study considers the potential for evaluating species interactions for Acer, Fraxinus and Juglans, all of which can be subdivided palynologically to species or subgeneric level. Such species-level discriminations have the potential to improve our interpretation of pollen diagrams at different timescales.
When synthesizing large numbers of pollen records across a region, it is important to determine whether the changes identified in the pollen diagrams are synchronous, as this permits the separation of climate impacts from other factors (Gajewski 1987, 1993). Many factors can influence the vegetation in the area surrounding the lake from which a pollen diagram is prepared, e.g. fires and the nature of the substrate, as well as biological factors, such as herbivory or epidemics. It is assumed that cores showing a synchronous change in one or several pollen taxa indicate a climatic impact, as climatic changes are large-scale and should affect all of the sites and many of the taxa in any region. Other effects, such as fires, are more local. A fast-moving epidemic or introduction of an exotic taxon may also appear synchronous, given the resolution of the pollen record, but should affect only the taxon involved, whereas a climatic change affects most or all of the plants in the region.
The North American Pollen Database (NAPD, Grimm 2000) and the newly released North American Modern Pollen Database (NAMPD, Whitmore et al. 2005) are large repositories of pollen data that have been made available, in part, for the purpose of enabling regional- or continental-scale palaeovegetation reconstructions. We use analyses at high taxonomic resolution to access untapped information in these archived pollen records. The specific objectives of our study are:
1to confirm that reliable pollen identifications are possible within Acer, Fraxinus and Juglans;
2to show the extent to which individual Acer, Fraxinus and Juglans species in eastern North America are found in different climatic envelopes and ecological niches, and have distinct modern pollen distributions;
3to document the extent to which species-level distinctions have been made by palynologists for pollen types within Acer, Fraxinus and Juglans;
4to use pollen records with high temporal and taxonomic resolution to determine how individual species within Acer, Fraxinus and Juglans have responded differently to environmental changes during the Holocene;
5to evaluate the potential for individual species within the genera Acer, Fraxinus and Juglans to provide palaeoclimatic information by determining if the abundance of these pollen types changes synchronously within a region through the Holocene.
Materials and methods
We examined and photographed reference pollen slides from our collection to verify the taxonomic characters used to separate species within the genera of interest. Reference material was prepared using standard methods (Fægri & Iversen 1989) from herbarium specimens. Unless specified otherwise, plant nomenclature follows Gleason & Cronquist (1963) throughout.
Maps of the modern pollen rain of Acer, Fraxinus and Juglans from eastern North America (Whitmore et al. 2005; http://www.lpc.uottawa.ca/data/index.html) were compared with tree distribution maps (Thompson et al. 1998). We then plotted the relative abundance of the pollen as a function of mean July temperature and total annual precipitation to illustrate the temperature and precipitation requirements of each species. The maps showing species ranges and modern pollen representation include data for all of North America (4634 samples); the climate response surfaces only include data points from east of 110° W and south of 55° N (2274 samples) to facilitate a comparison among the different species.
For regional comparisons of pollen curves, we focused on three well-studied regions with high densities of pollen records: southern Ontario, southern Québec and the western Great Lakes region. Within each region, we selected well-dated sites with high temporal resolution between samples (Fig. 1, Table 1). For each region, we used principal components analyses (PCAs) on standardized pollen percentages to analyse the tree pollen data from the site with the highest mean temporal resolution (Table 1).
Table 1. Sites from which pollen records were used in this study. Sites with an asterisk were used in the principal components analysis. Resolution is the mean number of years between pollen samples
We then created stratigraphic diagrams showing fluctuations through the Holocene of normalized per cent pollen abundance within each of the three study regions. The pollen sum includes all upland trees, shrubs and herbs. We superimposed a lowess smoothing function (span = 0.2, iterations = 1) on the pollen curves to emphasize long-term trends. All dates are expressed as calibrated (calendar) years before the present (Stuiver et al. 1998).
It is possible to separate pollen grains within the genera Acer and Juglans reliably, but palynological identification of species is more problematic in Fraxinus. We compared pollen grains of the eight native Acer species found in our study regions (Table 2). A. spicatum is the most readily identified as it is the only tricolporate Acer grain in the region. Tricolpate types are divided into striate (A. pensylvanicum, A. rubrum), reticulate (A. nigrum, A. saccharum) and rugulate (A. negundo, A. saccharinum), and then to individual species on the basis of the patterns made by the exine structure and sculpturing elements (Helmich 1963; Richard 1970; Adams & Morton 1972–1979; McAndrews et al. 1973). The pollen of A. barbatum Michx. is not found in the pollen databases nor in most pollen treatments, but it is similar to A. saccharum and future taxonomic analyses may place them in the same species (Burns & Honkala 1990). Acer barbatum pollen are included in A. saccharum-type although Helmich (1963) indicates that its reticulum is coarser than that of pollen grains of A. saccharum or A. nigrum.
Table 2. List of species of Acer, Fraxinus and Juglans in eastern North America (Gleason & Cronquist 1963). Pollen characteristics are taken from Cushing (1963), Richard (1970) and McAndrews et al. (1973). Note: Acer pseudoplatanus L. and A. platanoides L. are omitted because they are introduced. A. barbatum Michx., F. caroliniana Mill. and F. profunda (Bush) Bush (= F. tomentosa Michx. f. in Gleason & Cronquist 1963) are omitted because they are not found in pollen keys nor in the NAPD, but note that A. barbatum has A. saccharum-type pollen and F. caroliniana and F. tomentosa have tetracolpate pollen grains
Distinguishing pollen characters
Acer negundo L.
Tricolpate, more finely rugulate, no colpus membrane
A. nigrum Michx. f.
Tricolpate, reticulate, colpi do not meet at poles
A. pensylvanicum L.
Tricolpate, striate, striae in ‘digital’ pattern
A. rubrum L.
Tricolpate, striate, striae parallel
A. saccharinum L.
Tricolpate, coarsely rugulate, with colpus membrane
A. saccharum Marsh.
Tricolpate, reticulate, colpi meet at poles
A. spicatum Lam.
Fraxinus americana L.
Tetracolpate, finer reticulum, thick exine
F. nigra Marsh.
Tricolpate or rarely tetracolpate, colpi do not meet equator at right angles, coarser reticulum than F. americana-type
F. pennsylvanica Marsh.
Tetracolpate, coarser reticulum, larger and thinner exine
F. quadrangulata Michx.
Tricolpate, as in F. nigra
Juglans cinerea L.
Periporate, 7–11 pores
J. nigra L.
Periporate, 12–18 pores
Eastern North American Fraxinus pollen is divided into two types, referred to throughout our study as tricolpate (F. nigra, F. quadrangulata) and tetracolpate (F. americana, F. pennsylvanica) (Cushing 1963; Adams & Morton 1972–1979; McAndrews et al. 1973). Richard (1970) further divided F. americana from F. pennsylvanica on the basis of a tighter reticulum and slightly thicker exine in F. americana (Table 2). He notes, however, that secure distinctions between these two types depend on the simultaneous viewing of pollen grains of each type, and we confirmed this by studying reference material. Therefore, we limit our discussion to tri- and tetracolpate Fraxinus pollen types. The two south-eastern Fraxinus species, F. caroliniana P. Mill. and F. profunda (Bush) Bush (F. tomentosa in Gleason & Cronquist 1963) also have tetracolpate pollen grains but these are not found in the NAPD nor are they described in pollen keys. In rare cases, F. nigra-type pollen can be found in a tetracolpate form but it may be distinguished from the F. americana-type grains on the basis of a coarser reticulum and colpi that do not cross the equator at right angles (Cushing 1963).
There are two species of Juglans in eastern North America: J. cinerea and J. nigra (Gleason & Cronquist 1963). The distinction between pollen grains of these two species is determined by the number of pores; J. nigra has 12–18 pores, whereas J. cinerea has 7–11 (McAndrews et al. 1973). Cushing (1963) indicates that 10–11 pores should be used to divide the two Juglans species. Pore size may also be used as a distinguishing character (Adams & Morton 1972–1979).
ecological differences within genera
There are ecological differences between species within eastern North American Acer, Fraxinus and Juglans. Differences can be seen in range maps, ecological affinities and response surfaces. Furthermore, the distribution of the modern pollen in sediment samples reflects some of these between-species differences.
Eastern North American Acer species differ in terms of functional type, stature, shade tolerance, regeneration requirements and other ecophysiological characters. A. saccharum is the most shade-tolerant (Burns & Honkala 1990), capable of long suppression in the understorey, and has high water and nutrient requirements (Sipe & Bazzaz 1994). A. rubrum is characterized by architectural and physiological plasticity, making it well suited to gap or understorey environments, and a good competitor on exposed microsites (Sipe & Bazzaz 1994). It can be found at drier sites than other Acer species and its distribution is bimodal, with populations found in wet, dry, shaded or open habitats (Curtis 1959; Warren et al. 2004). Acer negundo is similarly widely distributed; A. saccharinum is moderately shade-tolerant and is the Acer species associated with the wettest conditions (Curtis 1959). A. spicatum has a more boreal and high-elevation distribution (Curtis 1959) and is of subcanopy stature, whereas A. pensylvanicum is a shade-tolerant shrub or subcanopy tree (Burns & Honkala 1990). Acer nigrum is more common in the western part of the study region where a warmer and drier climate prevails (Burns & Honkala 1990).
Acer pollen is rarely found outside of the range of the tree, except in the cases of A. saccharum and Acer undifferentiated (Fig. 2). Incidences of A. saccharum pollen outside the range of the tree may represent long-distance transport but could also reflect the presence of scattered populations of A. saccharum across the American south-east or stands of A. barbatum, which has A. saccharum-type pollen. In all cases, within the range of the trees, there are regions where the corresponding pollen is lacking, indicating that palynological distinctions have not always been made for those types. In the central to south-eastern United States, there are few incidences of undifferentiated Acer pollen (Fig. 2), and higher counts of specific Acer species, reflecting the higher taxonomic resolution of studies from that region.
The paucity of species-level distinctions of Acer pollen types across most of eastern North America makes it difficult to see the range of environmental conditions associated with particular species, based on pollen data (Fig. 3). Nevertheless, two of the less common Acer species show different tolerances. A. pensylvanicum pollen is found in moist, cool habitats, and A. negundo pollen has important representation in warmer areas. Because pollen counts are lacking from large parts of the ranges of the species, these response surfaces must be interpreted with caution.
Pollen morphotypes in Fraxinus do not correspond to the habitat associations within the genus: F. nigra and F. pennsylvanica are species of wet ground or swamp environments whereas F. quadrangulata and F. americana are associated with mesic or moist habitats (Gleason & Cronquist 1963). F. nigra in particular is associated with a cooler, humid climate (Burns & Honkala 1990).
Tetracolpate Fraxinus pollen grains (F. americana, F. pennsylvanica) are found throughout the range of the two trees producing them, although the western range limit of the pollen is slightly to the east of the tree limit (Fig. 2). Tricolpate Fraxinus pollen (F. nigra, F. quadrangulata) is recorded at sites across all of eastern North America, including the south-eastern United States, where Thompson et al. (1998) report that the trees absent. Gleason & Cronquist (1963), however, indicate that F. quadrangulata can be found from Oklahoma to Alabama, explaining the presence of tricolpate Fraxinus pollen in that region. Response surfaces (Fig. 3) do not show a clear difference between the two Fraxinus pollen types, which is expected given the broad climatic ranges associated with each one, and the lack of correspondence between pollen morphotypes and habitat associations.
Range maps indicate a more northern distribution for J. cinerea relative to J. nigra (Fig. 2). Ecological studies corroborate that J. nigra is less winter-hardy (Burns & Honkala 1990) and is more commonly associated with dry savanna or prairie environments (Curtis 1959). The modern pollen distribution reflects these species differences. Relatively abundant amounts of Juglans pollen are only found within the range of the plant. Pollen of J. cinerea is more abundant in New England and southern Québec, whereas pollen of J. nigra is only rarely found in these regions (Fig. 2). Response surfaces confirm that pollen of J. cinerea is more abundant in cooler and wetter areas (Fig. 3).
species level distinctions in the database
For Acer, Fraxinus and Juglans distinctions have been made in some cases within the NAPD, but these data have not been considered analytically to their full potential. There are substantial numbers of sites with counts of species-level types within Acer, Fraxinus and Juglans in the NAPD (Table 3). Despite the fact that many palynologists divide these taxa into individual species, these distinctions are frequently neither plotted nor analysed (Ritchie 1995). In part, this is due to the low numbers of grains typically recorded for these taxa, but also due to the need for efficiency in plotting and a perception that pollen data are best used for genus- or family-level palaeovegetation reconstructions. However, because these identifications are frequently submitted to the NAPD and archived, there is potential to obtain ecological information from these sites, particularly when multiple sites are considered together. Table 3 also shows that the ‘undifferentiated’ categories contain large numbers, indicating that many palynological studies are not conducted at high taxonomic resolution.
Table 3. Counts of pollen grains of Acer, Fraxinus and Juglans species in the North American Pollen Database (NAPD) and the North American Modern Pollen Database (NAMPD). Species counts include ‘types’, and ‘cf’ (e.g. A. negundo-type included within A. negundo). The ‘undifferentiated’ category in the table includes the NAPD category ‘undiff.’ as well as the category genus name only (e.g. Acer undifferentiated =Acer undiff. +Acer)
No. of grains
No. of sites
No. of grains
No. of sites
Not in database
Not in database
responses to holocene environmental changes
Within single sites, species of Acer, Fraxinus and Juglans responded differently to environmental changes through the Holocene. For the site with the highest temporal resolution (Table 1) within each of our three focal regions, PCAs (Fig. 4) show that congeneric species do not necessarily follow the same patterns through time, thus confounding the interpretation that we might make if genera only are considered.
At Crawford Lake (McAndrews & Boyko-Diakonow 1989), A. saccharum and A. rubrum show almost opposite trends on PCA axis 1, as do J. cinerea and J. nigra, indicating that the abundances of pollen of individual species within a genus have varied in a dissimilar fashion over the past 1700 years (Fig. 4a). By contrast, the two Fraxinus types show similar loadings on axis 1. When genera only are considered, Acer is closely correlated with Fagus, as is A. saccharum when species distinction is available (Fig. 4a). Juglans is apparently primarily composed of J. cinerea and is well correlated with Ulmus (Fig. 4b). When pollen counts are lumped into generic groupings for Acer and Juglans, associations of A. rubrum and J. nigra with Betula and Populus in the upper right quadrat of the ordination diagram can no longer be detected (Fig. 4a,b). The ordination shows the changes in species composition that occurred in the forest surrounding Crawford Lake beginning about 500 years ago, coincident with the onset of the Little Ice Age as well as increased human activity in the surrounding area (Clark & Royall 1995). These disturbances resulted in canopy openings that decreased the competitive advantage of slow-growing, shade-tolerant A. saccharum and lowered its pollen percentages. Acer rubrum is able rapidly to exploit canopy openings and other disturbances, and this is reflected in the rapid increases in its pollen percentages at this time. Pollen of J. nigra increased at this time of disturbance relative to that of J. cinerea, which can be explained on the basis of the association between J. nigra and more open forest canopies (Curtis 1959; Szeicz & MacDonald 1991).
Five Acer species were recorded in the 5500-year record at Tourbière Lanoraie St Jean (Comtois 1982). In the ordination showing all species, the first axis separates A. saccharinum and A. pensylvanicum, with positive loadings on axis 1, from A. rubrum, A. spicatum and A. saccharum, with negative loadings (Fig. 4c). The three Fraxinus species recorded at this site are again poorly differentiated on axes 1 and 2 (Fig. 4c). When counts for all Acer species are grouped, Acer remains correlated with Quercus and Ulmus (Fig. 4d), as were A. saccharum and A. spicatum in the species-level ordination (Fig. 4c), but evidence for the correlation of A. pensylvanicum and A. saccharinum with Pinus strobus at the positive end of axis 1 is lost. At South Rhody Peatland (Booth et al. 2004), A. saccharum and A. rubrum, and F. americana and F. nigra, are differentiated on axis 1 only weakly (Fig. 4e). When the Acer and Fraxinus species are grouped into generic categories (Fig. 4f), the position of Acer suggests an averaging effect: neither Acer species reflects closely the position of the Acer genus, which is intermediate in location on the ordination plot with respect to the two species. Similarly, the genus Fraxinus is correlated with Cupressaceae (Fig. 4f), indicating PCA axis 1 values between those of F. americana and F. nigra in the species-level ordination (Fig. 4e).
synchrony of changes
Within any region, pollen of Acer, Fraxinus and Juglans species show synchronous changes, when accounting for uncertainties in chronology and resolution. We compared the species pollen curves in the three regions (Figs 5–7) to determine if the broad-scale, regional variations in the abundance of these pollen types are synchronous, which would suggest an overall climatic influence on these taxa. However, given the chronological constraints of the datasets available, it is unlikely that exact temporal correlations will exist in pollen curves for the species of interest. Thus, we can only ask whether the curves suggest that similar changes occurred at the regional scale.
The onset of the Tsuga pollen decline is plotted for sites across Ontario (Fig. 5a). This decline presumably occurred synchronously in response to the rapid spread of a pathogen (Davis 1981a; Fuller 1998), yet at the 14 sites considered, the chronologies supplied in the NAPD suggest that its date falls within a 1100-year window. It is unclear if these differences are due to a time-transgressive decline or due to insufficient dating control, but the lack of synchrony does suggest that many profiles in the NAPD are inadequately dated, and/or have poor temporal resolution between samples.
Nevertheless, the pollen curves for a single species across a region show some overall similarities, and fine-scale variations are seen in sites with higher temporal resolution. In southern Ontario, A. saccharum increased in abundance around 8000–7000 bp and reached maximum abundance between 5000 and 2000 bp. It declines over the most recent 1500 years, and most steeply since the Little Ice Age (Fig. 5a). Similarly, in southern Québec, A. saccharum curves show two peaks around 5000 and 2500 bp (Fig. 5c). Nearby lakes show similar, finer scale details: in Ontario, Lac Bastien, Found and Second Lakes show maxima in A. saccharum pollen around 1000 years ago, whereas the peak at the more south-eastern Nutt, Graham, High, Tonawa and Cranberry Lakes occurs 4000 years ago (Fig. 5a). At the southernmost sites in Ontario, Van Nostrand Lake, Porqui Pond, Wylde Bog and Wylde Lake, A. saccharum peaks even earlier, around 5000 years ago (Fig. 5a). In the western Great Lakes region (Fig. 5e), A. saccharum peaks at about 7000 bp in the easternmost lakes (Wood, Gass, Radtke Lakes and Lake Mendota), and around 3000 years ago in the northernmost lakes situated on the Upper Peninsula of Michigan (Spirit, Kitchner, Canyon, Lost, Camp 11 and Mud Lakes). The profiles for the easternmost Michigan sites (Ryerse, Wolverine Lakes and South Rhody Peatland) show similar profiles, with rises in A. saccharum pollen particularly since 2000 bp.
The curves for A. rubrum pollen are more difficult to interpret because of the overall scarcity of pollen counts of this type (Fig. 5b,d,f). The Québec sites contain the highest counts of A. rubrum (Fig. 5d) and illustrate the dissimilarity in the pollen curves for these two Acer species. At Lac à Sam and Lac Castor, for example, A. rubrum peaks around 4200 bp, while A. saccharum peaks 300 years later (Fig. 5c). Mont Shefford and Lac Geai, located further south, have a similar peak in A. rubrum 4000 years ago, but a later peak in A. saccharum, occurring around 3000 years ago. At several sites across the three regions, a very recent increase in A. rubrum occurs (e.g. Second, Van Nostrand Lakes, Lac Romer, Mont Shefford, Spirit, Kitchner, Mud, Gass and Radtke Lakes). The short, very high-resolution dataset from Crawford Lake shows a pronounced increase in A. rubrum 500 years ago, and a corresponding decline in A. saccharum (Fig. 5a,b). Finally, at 28 of the 32 sites with data for both A. saccharum and A. rubrum, A. rubrum pollen appears in the records following A. saccharum, indicating dissimilar arrival and migration histories for these taxa in the early Holocene (Fig. 5).
The pollen curves for tri- and tetracolpate Fraxinus (Fig. 6) are more complex than those of Acer spp. In Ontario, three classes of tricolpate Fraxinus curves can be distinguished. At Lac Bastien and Found Lake, tricolpate Fraxinus decline from an early maximum; at Graham, High, Tonawa Lakes and Wylde Bog, the curve rises to a mid Holocene maximum, and at some lakes (e.g. Nutt Lake), few changes are observed (Fig. 6a). The data from Québec show some synchrony in Fraxinus pollen curves. Tricolpate pollen types peak between 4000 and 2000 years ago at Lac Geai, St Calixte and Lac Manitou (Fig. 6c), whereas a synchronous peak occurs earlier in tetracolpate Fraxinus types at Lac à Sam, Lac Castor, St-Gabriel and Lac Romer, around 4500 years ago (Fig. 6d). In the western Great Lakes region, tricolpate Fraxinus pollen has decreased since the early Holocene at many of the sites, whereas tetracolpate grains have oscillated through time (Fig. 6e,f). Tricolpate Fraxinus pollen appears earlier in the records than the tetracolpate Fraxinus types in 24 of 34 profiles shown here (Fig. 6).
The curves for Juglans cinerea are broadly similar among the regions (Fig. 7). In Ontario and Québec, where the pollen is more abundant, maxima are well defined between 6000 and 4500 bp at Cranberry Lake, Barry Lake, Van Nostrand Lake, Wylde Bog and Wylde Lake (Fig. 7a) and earlier at the Québec sites Saint-Gabriel, Lac Romer and Lac des Atocas, where an abrupt rise occurs around 7000 years ago (Fig. 7c). Counts are low for J. nigra across the whole study region (Fig. 7b,d,f), limiting the interpretability of those pollen curves. A pronounced rise in J. nigra can, however, be seen in the very high-resolution record from Crawford Lake, coincident with a decline in J. cinerea 500 years ago (Fig. 7a,b). In 24 of 27 sites shown here, pollen of J. cinerea appears earlier in the records than that of J. nigra, suggesting dissimilar migration chronologies for the two species (Fig. 7).
Despite a lack of a comprehensive pollen flora for North America, analysts are making use of regional and taxonomically specialized texts to make distinctions to the species level for pollen types within North American Acer, Fraxinus and Juglans. Fraxinus pollen types remain difficult to separate to the species level, but with further close morphological study, advances may be possible.
Analyses from the NAPD show that, in some cases, distinctions have been made within these genera and submitted to the database. However, the large numbers of undifferentiated Acer, Fraxinus and Juglans pollen grains also in the database indicate that high taxonomic resolution is not always routine in palynological analysis. There are three main reasons why identifications to the species level may be missing from the NAPD. First, the analyst may not have identified these taxa due to the requirements of the study or their confidence in the pollen identification. Second, the data may not have been entered into the database, as was the case for many data originating from the Cooperative Holocene Mapping Project (COHMAP) database, in which data files were taxonomically limited due to computer memory constraints existing at that time. Third, pollen may not be identified or analysed at the species level because of a perception that species-level identifications will not improve upon reconstructions of past environments; our analysis, however, of pollen data from the NAPD shows that higher taxonomic resolution does lead to a more detailed characterization of palaeoenvironmental change.
The low counts for some of the pollen types considered here precluded thorough synoptic analyses for eastern North America. However, for some of the taxa for which species-level pollen counts were more numerous, we showed several examples of small- and broad-scale synchrony in the dynamics of tree species through time, suggesting the possibility of species-specific responses to climatic change. By looking in detail at pollen data from a small number of well-dated sites in ordinations, we also illustrated the ways in which the dynamics of individual species within a genus differ through time.
Our review of ecological differences between congeneric species and dissimilarities in environmental conditions associated with each one indicates that palynological studies made at high taxonomic resolution can provide a richer view of palaeoenvironmental change. For genera comprising several species, especially when important autecological differences exist between the species, as within the genus Acer, considering the dynamics of the genus as a whole to reconstruct palaeoenvironments provides an incomplete picture of past environments. Plotting the changing abundance of individual species within Acer, Fraxinus and Juglans through the Holocene shows that congeneric species follow different patterns through time. These results lend further support to the idea that low taxonomic resolution can be a source of error in matching fossil pollen data with modern analogues (Sawada et al. 2004; Williams et al. 2004). Using higher taxonomic resolution is a promising way to increase our knowledge of past environments. One major limiting factor, however, remains poor chronological control. Profiles with more dates and higher sampling resolution are needed for the quantitative assessment of species-level interactions through the Holocene.
Tracking species-level interactions through time by pollen analysis provides the potential for addressing key questions in forest ecology. For example, A. rubrum has recently been increasing across forests of eastern North America; hypotheses to explain this increase include fire suppression, elevated deer browsing on competitors, disease or elevated concentrations of atmospheric CO2 (Abrams 1998). At the same time, A. saccharum has been declining (Bauce & Allen 1991). The decline in A. saccharum also remains poorly explained, variously ascribed to acid rain, climatic changes, disease or ‘natural stand regulatory processes’ (reviewed in Minorsky 2003). The pollen records presented here, tracking the late Holocene dynamics of A. rubrum and A. saccharum, show that at several sites A. saccharum has been declining for the past 2000 years, and A. rubrum increasing since the time of the Little Ice Age. These longer-term processes must be considered alongside the other hypotheses of factors affecting these taxa. Palaeodata suggest that millennial- and centennial-scale climate changes are important in the dynamics of these forest trees.
At Crawford Lake, the rise in A. rubrum and J. nigra, and declines in A. saccharum and J. cinerea at the time of the Little Ice Age (McAndrews & Boyko-Diakonow 1989), suggest that these taxa might be responding to lower temperatures or to a wetter climate, or these changes may be a reflection of opportunistic capitalization on openings in the forest canopy. Canopy opening may be due to a long-term decline of Fagus (Gajewski 1987), a deep-shade-casting canopy dominant; A. rubrum is a better competitor than A. saccharum in gap conditions (Sipe & Bazzaz 1995). J. nigra also increased at this time while J. cinerea declined at Crawford Lake, even though J. nigra is the more southern species and presumably better adapted to warmer conditions. There have been few studies on the differences in regeneration requirements within Juglans species, but these pollen data suggest that J. nigra and J. cinerea may have different responses to canopy openings. In this case, pollen analysis of the dynamics of congeneric species can be used to distinguish ecophysiological mechanisms for observed changes from climatic forcing.
The interpretation of some of the pollen data presented here relies on pollen counts that are low. Small counts can be explained by low pollen production, poor dispersal or preservation, or long-distance transport. Therefore, rare types are often eliminated from the analysis and discussion of palynological data. It has been shown, however, that rare pollen taxa contain important information. For example, Gajewski et al. (1993) showed that very low values of pollen could trace the arrival of Pinus banksiana, which is found only in particular regions of the landscape at the treeline in northern Québec. In a study of the Holocene expansion of Fagus into Ontario, Bennett (1988) argued that low pollen counts are interpretable. He compared the rate of increases in Fagus pollen through time with an exponential population growth model, and showed that even when Fagus pollen is rare (< 0.05% of total pollen), it increases exponentially at the same rate as later in the records when Fagus constitutes 5% of total pollen. This suggests that even the expansion of small local populations can be sensed in the pollen record (see also Birks 1981).
We have shown that that pollen records contain information that has been insufficiently interpreted. High temporal resolution, many radiocarbon dates and high taxonomic resolution will be key elements of the next generation of pollen studies, and will allow palynological datasets to be used to their full potential.
The work was funded by a Natural Sciences and Engineering Research Council of Canada (NSERC) grant to K.G. and by an NSERC Postdoctoral Fellowship to S.A.F. We acknowledge contributors to the pollen databases, as well as R.K. Booth, for providing pollen data, J. Whitmore for data preparation and M. Sawada for discussions. We thank Eric C. Grimm for helpful comments on an earlier version of the paper.