2486 POLLEN ANALYSIS, PRINCIPLES

sites in the Swiss Alps. Palaeogeography, Palaeoclimatology, Wasylikowa, K. (2005). Palaeoecology of Lake Zeribar, Iran,
Palaeoecology 159, 251–259. in the Pleniglacial, Lateglacial and Holocene, reconstructed
van Zeist, W., and Bottema, S. (1977). Palynological investigations from plant macrofossils. The Holocene 15, 720–735.
in western Iran. Palaeohistoria 19, 19–85. Wasylikowa K, Witkowski A, Walanus A, et al. (2006)
von Grafenstein, U., Eicher, U., Erlenkeuser, H., et al. (2000). Isotope Palaeolimnology of Lake Zeribar and its climatic implications.
signature of the Younger Dryas and two minor oscillations at Quaternary Research, in press.
Gerzensee (Switzerland): Palaeoclimatic and palaeolimnologic Wick, L. (2000). Vegetational response to climatic changes recorded
interpretations based on bulk and biogenic carbonates. in Swiss Late Glacial lake sediments. Palaeogeography,
Palaeogeography, Palaeoclimatology, Palaeoecology 159, Palaeoclimatology, Palaeoecology 159, 231–250.
215–229.

Plate Tectonics see Glacial Landforms: Glacitectonic Structures and Landforms; Glaciation, Causes:
Tectonic Uplift-Continental Configurations; Glaciations: Late Quaternary in North America

Pliocene Environments see Paleoclimate Reconstruction: Pliocene Environments

POLLEN ANALYSIS, PRINCIPLES

H Seppä, University of Helsinki, Helsinki, Finland concentration of sediments is usually high, in lake
ª 2007 Elsevier B.V. All rights reserved. sediment often of a magnitude of 103105 grains
cm3, and pollen analysis can often be carried out
from sediment samples smaller than 1.0 cm3 in
volume. The sedimentary pollen records are therefore
Introduction unique in their stratigraphic completeness and detail.
Pollen analysis, the study of fossil pollen and spores, This is one of the major advantages of pollen analysis,
is the principal technique for long-term (102105 yr) namely, a methodological flexibility permitting
vegetation reconstructions. Pollen-stratigraphical adjustment of the technique according to the time
results have greatly shaped the development of ideas dimensions relevant for the study. Consequently, the
on the nature of vegetation and climate during the applications of pollen analysis can vary from investi-
Quaternary glacial–interglacial cycles. They have gations of the vegetation dynamics of the continents
provided unique insights into long-term ecological during the glacial–interglacial cycles to studies of
patterns and processes such as dispersal dynamics annual to subdecadal patterns of the secondary
and distribution changes of tree species, the ancestry, plant succession after a local forest disturbance.
development, and continuity of modern plant com- The spread of airborne pollen by wind also results
munities and biomes, and the sensitivity and response in methodological complexities, some of which con-
mechanisms of plants and vegetation to climate stitute the most important limitations and pitfalls of
changes. They have also shed light on the origin and the method. The crux of the problem is that many
spread of early agriculture and the role of humans in different patterns of plant abundance can yield the
influencing vegetation. same flux densities of pollen types (Prentice, 1988).
Plants liberate enormous quantities of pollen into The reason for this is that the pollen dispersal mode
the air, the production of pollen grains in one forest from a singular source such as a single tree is typically
hectare rising into the billions (Pohl, 1937). In the leptokurtic; most of the pollen is deposited within
atmosphere, pollen is efficiently transported and inte- 101103 m from the source, but a small proportion
grated by wind. Pollen is therefore widely and evenly of pollen is transported in the atmosphere over
spread, and fossil pollen records are available from longer, theoretically indefinite distances. It is there-
various sedimentary environments such as seas, lakes, fore difficult for a paleoecologist to know whether
ponds, peatlands, soils, and humus layers. The pollen small proportions of given taxa in sediment reflect a
 POLLEN ANALYSIS, PRINCIPLES 2487

few individuals of a tree species that formerly grew in the male flowers of plants and delivered during flow-
the vicinity of the coring site or whether they came ering. The purpose of the pollen grain is to germinate
from a larger population that grew far away from the on a stigma and provide the sperm nucleus to the
site (Davis, 2000). The problem of precisely defining ovalus. For this purpose, pollen is carried by wind,
the pollen source area and quantitatively reconstruct- insects, or other animals from the anthers to the sta-
ing past plant assemblages is crucial, not only men of the female flower. Wind-pollinated plants,
because of the risk of erroneous interpretations of such as most of the conifers, many other angiosperm
the nature of past vegetation and environments, but trees, grasses, and sedges usually produce more pol-
also because it has been the obstacle for utilizing the len than animal-pollinated plants. The estimated
full potential of pollen analysis as a central, routinely annual pollen production rates of common, wind-
used tool in ecological research. pollinated plants reflect the enormous dimensions of
Here, the general features of pollen analysis as a pollen production: one shoot of hemp can produce
method in Quaternary science are reviewed. The over 500 million and a shoot of Pinus contorta can
focus is particularly on the theoretical basis of pollen produce 8–9 million pollen grains per year; a male
analysis as a technique for vegetation reconstruction. plant of Rumex acetosa produces 400 million pollen
Some examples of the application of the technique grains, one 10-yr-old branch system of birch, spruce,
are also included. These are selected to reflect the or oak can produce more than 100 million grains,
variable time and space scales that characterize the and pine over 350 million pollen grains in a year
use of pollen analysis, and also to highlight the pro- (Pohl, 1937). The pollen productivity of the animal-
blems and pitfalls that limit the use of the technique. pollinated plants is, in general, much lower, but the
production rates of some animal-pollinated plants,
such as lime, willows, and heather, are comparable
Origin of Pollen and Spores to that of wind-pollinated plants.
Pollen grains are unicellular, microscopic (15–100 mm)
organs that contain the male gametophyte generation Sedimentation and Preservation
of plants. Spores are, in principle, the corresponding
Only a fraction of pollen grains falls on the stamen of
organs of the pteridophytes (ferns, mosses, and allies)
the maternal part of the flower and fertilizes the
but differ morphologically and functionally from pol-
ovule. A great majority of pollen falls on land or
len (Fig. 1). Pollen grains are produced in anthers of
water and decays, unless pollen is deposited in an
anoxic sedimentary environment where it will be
preserved. During sedimentation, the soft-bodied
cell contents of pollen grains are lost, and the fossil
pollen grains consist of hard and chemically resistant
exines. In addition to the sedimentary environment,
preservation of pollen grains depends on the thick-
ness and sporopollenin content of the exine, but in
practice palynologists usually assume that all pollen
grains are equally preserved in lake sediments or peat
deposits, at least if they are deposited directly from
the atmosphere on to the sediment.
Pollen is recovered from sediments with various
types of sediment corers. The subsamples are treated
in the laboratory by a series of physical and chemical
treatments. The standard procedure includes treat-
ment of the sample with hydrochloric acid, potas-
sium hydroxide, hydrofluoric acid, and a mixture of
sulfuric acid and acetic anhydride (Fægri and Iversen,
Figure 1 Different kind of pollen and spore types: (A) Vaccinium 1991; Moore et al., 1991; Bennett and Willis, 2003).
vitis-idaea (cowberry), a tetrad pollen type; (B) Potamogeton Hydrofluoric acid is a strong acid that dissolves the
natans (broad-leaved pondweed), an inaperturate pollen grain; silicates, while sulfuric acid and acetic anhydride
(C) Tussilago farfara (coltsfoot), an echinate trizonocolporate pol-
form strong acetylate that removes cellulose and
len type; (D) Cystopteris fragilis (fragile fern), a monocolpate
spore; (E) Corylus avellana (hazel), a trizonoporate pollen type; most of the other organic remains from the subsam-
(F) Silene dioica (red campion), a polypantoporate pollen grain ple. The residue is then stained with saffranin and
([Link] suspended in silicone oil or glycerol for slide
2488 POLLEN ANALYSIS, PRINCIPLES

preparation. Pollen grains are normally counted with (Prentice, 1988). This pattern of pollen dispersal
500 magnification. from a source explains why plants near a lake are
more strongly represented in the pollen input of a
basin than those growing progressively further from
Pollen Dispersal and Spatial the basin, and implies that some form of distance-
Representation weighting must be applied to plant abundances while
modeling or comparing modern pollen assemblages
Understanding pollen dispersal processes in the air is with modern plant abundances (Prentice, 1988;
a prerequisite for a precise and explicit definition of Sugita, 1994).
the study area in pollen-based paleoecological A central concept in the Prentice–Sugita model is
research. Pollen analysts have long compared surface the ‘relevant source area of pollen’ (RSAP). RSAP
samples with the surrounding vegetation in order to refers to the size of the area around a given study
assess how pollen percentages correspond with the site, which controls the individualistic features in the
vegetation. Such comparisons show that small hol- pollen assemblages of the site in relation to the uni-
lows (small openings within forests) receive great form regional background pollen. The vegetation pat-
quantities of pollen from the vegetation around the terns within the RSAPs are therefore reflected in the
sampling site, whereas centers of larger basins, for lake-to-lake differences in pollen input in areas of the
example, lakes, receive predominantly wind-trans- same regional vegetation (Sugita, 1994; Davis, 2000).
ported and well-mixed pollen from the air. The RSAP for a site can be estimated by calculating r2
Consequently, the source area of pollen records or maximum likelihood function score as a measure of
from lakes is the regional vegetation, whereas small goodness-of-fit between the pollen assemblage and the
hollows have a smaller source area and reflect more distance-weighted plant abundance around each study
the local vegetation around the site (Jacobson and site (Broström et al., 2004). The distance beyond
Bradshaw, 1981). This simple model has been used which r2 or the likelihood score no longer improves
as a general rule of interpretation of pollen-stratigra- defines the radius of the RSAP for that particular site
phical records from sampling sites of different sizes. (Fig. 2A). The simulation outputs using the Prentice–
Such a qualitative and intuitive definition does not, Sugita model in an idealistic patchy vegetation indi-
however, permit specification of the spatial scale of cate that the RSAP of sites with a radius of 2 m is
the study with the precision that would be compar- about 50–100 m from the edge of the site, 300–400 m
able to that of modern ecologists. A challenge for for sites with a radius of 50 m (small lakes), and 600–
palynologists is to more quantitatively and precisely 800 m for sites with a radius of 250 m (medium-size
define the pollen source areas, and, hence, to gener- lakes) (Fig. 2B; Sugita, 1994).
ate more spatially precise vegetation reconstructions. As with all models, the Prentice–Sugita model con-
Comprehensive efforts to modeling the pollen dis- tains many idealistic assumptions regarding pollen
persal and defining the source areas have been carried transportation and deposition mechanisms, local
out since the 1980s, and the resulting model is and regional vegetation patterns, and basin config-
usually termed the ‘Prentice–Sugita model’ after its uration. It does not, for example, account for stream-
key developers (Prentice, 1988; Sugita, 1993, 1994). borne pollen, a major source of pollen in many lakes.
The principal purpose of the model is to quantify the It is thus unclear how reliably the model can be
pollen input from the vegetation surrounding the applied to define the source area of fossil pollen
sampling site, either to the center of a basin or on assemblages in real-world conditions. Thus far, the
the entire surface of a basin. According to the model, most important implication of the Prentice–Sugita
the input depends on the pollen productivity of the model may have been that palynologists have been
plant species, the mean plant abundance at a given forced to take a fresh look at site selection in palyno-
distance from the center of the basin, and a pollen logical research. Studies designed to investigate
deposition function. The pollen deposition function, stand-scale forest disturbance or local prehistoric
based on Sutton’s physical equation for a ground land use must use small hollows or lakes with a
source, defines the proportion of pollen remaining radius of less than 50 m, whereas palynologists inter-
airborne at a given distance from the plants and is ested in regional-scale pollen signals (e.g., in climate
individual to all pollen types because of their indivi- reconstructions) should focus on lakes with a radius
dual deposition velocities. A characteristic feature of 250 m or more (Fig. 3). In treeless environments
common to the shapes of all the deposition functions such as the Arctic and Antarctic regions where the
is their leptokurtic nature: a majority of pollen will use of pollen analysis is often futile, owing to the
be deposited near the source, but a proportion will be strong influence of long-distance pollen carried
transported over theoretically indefinite distances from the forested regions, future studies must focus
 POLLEN ANALYSIS, PRINCIPLES 2489

Open region
38,200
Likelihood function score

38,000

37,800

37,600
400
37,400

37,200
(A) 0 200 400 600 800 1,000 1,200 1,400 1,600

R = 2m R = 50 m
1.0 1.0

0.8 0.8
Pollen proportion

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
R = 250 m R = 750 m
1.0 1.0

0.8 0.8
Pollen proportion

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Sites Sites
(B) Sugar maple Hemlock Yellow birch

Figure 2 The definition of the relevant source area of pollen (RSAP) in an open region. The correlation between the modern pollen
assemblage from a moss polster and plant abundance is assessed by likelihood function scores, calculated over distance from the moss
polster. The changing scores indicate the changes in the goodness-of-fit of the linear model between pollen loading and plant
abundances. The likelihood function score first decreases as the goodness-of-fit improves but reaches an asymptote at about 400 m,
indicating that the correlation between pollen and the surrounding vegetation does not improve beyond this distance; the RSAP of this
specific sampling site therefore has a radius of 400 m (Broström et al., 2004).

on small hollows or small lakes, because only at such the pollen records do not directly reflect plant abun-
sites can detailed vegetation reconstructions be dances because some taxa are overrepresented and
achieved for an area with a radius of a few kilometers others underrepresented in pollen samples. A poten-
around the pollen site. tial way to transform the fossil pollen percentage
values to correspond more precisely with real relative
plant abundances is to use correction factors devel-
Plant–Pollen Representation
oped individually for each pollen type. In order to
Since the first comparisons of pollen records and obtain such factors, it is necessary to study the
modern plant abundances, it has been known that pollen–plant relationships by comparing modern
2490 POLLEN ANALYSIS, PRINCIPLES

Figure 3 (A) A lake of about 20 ha, a typical sampling site for pollen analysis focusing on reconstruction of Postglacial regional
vegetation patterns in the boreal zone of Europe. The dominant tree species in the surrounding forest are pine and spruce (Photo by
Charlotte Sweeney). (B) A representative small hollow site in temperate woodland in Eriksberg, Blekinge, Sweden. The sediment core
was obtained from the center of the small overgrown pond. Photo: Gina Hannon. Sugita, S. (1994). Journal of Ecology 82, 881–897.

plant abundances and modern pollen values within Table 1 Two examples of the calculation of the extended
R-value (ERV model 1) slope parameter (ai) and y-intercept para-
the forests. Foundations for this work were laid in the meter (Zi) estimates, based on data shown in Figure 4
temperate woodlands in North America and Europe
in the 1960s. Davis (1963) defined an index, termed Pinus Acer
R-value, for describing the ratio between pollen per- 20 m 70 m 120 m 20 m 70 m 120 m
centages and vegetation percentages, and used the
R-value to convert the original pollen sums to reflect ai FC 1.22 1.32 1.38 0.04 0.03 0.03
the vegetation percentages. In contrast, Andersen ai EA 0.37 0.51 0.58 0.03 0.02 0.03
zi FC 15.99 15.19 14.86 4.65 4.77 4.74
(1970) compared absolute pollen accumulation zi EA 42.58 40.45 39.31 2.27 2.32 2.33
values analyzed from moss polsters with the vegeta-
tion percentages of corresponding tree species in the Jackson ST and Kearsley JB (1998) Quantitative representation
same woodland. Significantly, he showed that mod- of local forest composition in forest-floor pollen assemblages.
Journal of Ecology 86: 474–490.
els describing pollen input cannot only consist of
multiplying terms that reflect the relative productiv-
ity differences, but must also include additive terms fk is the site factor ensuring that the right-hand side
to account for the background pollen component of the equation will total 100%. The site factor can
originating outside the vegetation area used for cali- be expressed as
bration purposes. The relationship between pollen ! !
percentage values and the vegetation percentage can X X
therefore be depicted as a linear regression curve in fk ¼ 1= aj njk 1– xj
j j
which the slope indicates the pollen productivity of
the given species and the y-axis intercept shows the Of the two examples, Pinus represents a taxon
amount of background component of the same spe- with prolific pollen productivity and efficient pollen
cies transported from outside the study area (Fig. 4 dispersal; it has therefore high i and zi values. Acer
and Table 1). is a taxon with low pollen productivity and weak
The ERV model presents the relationship between dispersal capability; it has low i and zi values.
the pollen percentage and the surrounding vegetation Pinus shows increasing slope and decreasing y-inter-
as cept with an increasing vegetation sampling area.
Two different data sets are shown, obtained from
pik ¼ ai nik fk þ zi
Fish Creek (FC) and the Eastern Adirondacks (EA)
where pik is the pollen percentage of the taxon i at in northern New York, USA (Jackson and Kearsley,
site k and nik is the abundance of the taxon in the 1998).
vegetation.  is a coefficient describing the pollen An improvement combining features of both the
production of taxon i, and zi is the background com- R-value method and Andersen’s linear regression
ponent of pollen coming from outside the study area. method is an extended R-value (ERV) method,
 POLLEN ANALYSIS, PRINCIPLES 2491

Pinus which is based on the percentage values but takes

100 20 m 70 m into account the background pollen input (Prentice
80 and Parsons, 1981; Prentice, 1988). In addition, the
ERV method includes a site factor that ensures that
60 the corrected percentages at each site total 100%.
40 Due to the inclusion of the site factor, ERVs also
account for the Fagerlind effect, a nonlinearity typi-
20
cal of pollen percentage data. The linear regression
methods assume a linear relationship between the
% pollen

0
pollen values and plant abundance. However, the
100 120 m 120 m DV
percentage pollen data are not linear, because they
80 must total 100%. Therefore, an increase in the pollen
type of one taxon must be accompanied by a decrease
60
of another, even if the actual abundances of the other
40 taxa remain constant. In general, the ERVs give a
markedly better fit to pollen percentage–plant abun-
20
dance data than the classical linear regression
0 models.
0 20 40 60 80 100 0 20 40 60 80 100 Although valuable and rational in principle, the
%BA generation of reliable correction factors is a compli-
cated process, and their use is never simple or
Acer straightforward. Nevertheless, when the representa-
100 20 m 70 m tion differences between the main plant taxa are sig-
nificant, the corrected values undoubtedly often
80
reflect the original plant abundances more closely
60 than the uncorrected values. Yet, it is important to
remember that the linear models for pollen input
40
assume constant background pollen and constant
20 pollen productivity for any given taxon, whenever it
appears (Prentice, 1988; Jackson and Kearsley,
% pollen

0
1998), implying that the use of correction factors is
100 120 m 120 m DV constrained in space and time to the overall vegeta-
80 tion type for which they were originally generated.
The influence of new climate regimes on pollen pro-
60 ductivity is poorly understood, but it is clear that a
40 change in an overall vegetation structure leads to a
change in background pollen input. Vegetation
20
changes during the Holocene have been of such mag-
0 nitude that the background pollen input has
0 20 40 60 80 100 0 20 40 60 80 100 undoubtedly changed, and the linear relationship
%BA obtained from modern samples is therefore not a
Figure 4 Scatter diagrams showing the relationship between precise or reliable estimate for correcting the fossil
Pinus and Acer pollen percentages and their basal area pollen values on millennial or longer timescales. This
(BA) percentages in the surrounding vegetation (Jackson and may be the one reason for the fact that, despite the
Kearsley, 1998). Pollen percentages for individual sites are con-
development of correction factors and methods for
stant, but three different vegetation sampling areas have been
surveyed: 20 m ¼ 20-m vegetation sampling radius, 70 m ¼ 70-m estimating RSAPs, few pollen-based efforts have been
vegetation sampling radius, 120 ¼ 120-m vegetation sampling carried out to quantify past plant abundances in the
radius. Results from two different data sets are shown, from form of biomass or canopy coverage over an expli-
Fish Creek (open circles) and the Eastern Adirondacks (closed citly defined area.
circles) in northern New York, USA (Jackson and Kearsley,
A potential way to avoid the problems related to
1998). Scatter plots such as these provide the basis for estimat-
ing the pollen productivity of a plant species; species with high percentage data and to estimate plant abundance in
productivity have a steep slope and those with efficient dispersal the surrounding landscape independently for each
a high y-axis intercept. species is to calculate the direct pollen accumulation
2492 POLLEN ANALYSIS, PRINCIPLES

rate (PAR) values individually for each species. Despite these caveats, in some regions the PAR
Instead of relative percentage data, PAR is a calcula- technique has been successfully applied to estimate
tion of the accumulation of pollen at the sediment past tree population densities around the study lakes.
surface (per cm2 or other space unit) per year (Davis Especially valuable have been the applications of
and Deevey, 1964). This value is individual and inde- PAR values in the ecotones of the forested and tree-
pendent for each pollen and spore type and therefore less biomes, such as boreal forest and the arctic tun-
avoids the problems arising from the interdependent dra, where the advances or retreats of the main tree
nature of the relative records. PAR values are usually taxa are recorded as major changes in PAR values,
calculated for the most important tree species, the while less change is apparent in percentage values. In
aim being for PAR values to provide precise records suitable circumstances, it is possible to use modern
of past tree population sizes, measured either as surface samples or pollen trap results for generating
canopy coverage or biomass. However, methodolo- threshold PAR values that define the absence, pre-
gical investigations have shown that the PARs can sence, or dominance of key tree species and apply
vary by a factor of 3 up to 10, not only within lakes these values to the fossil PAR records for providing
but also between lakes located in the same area and semiquantitative estimates of past population densi-
surrounded by similar vegetation. The predominant ties in the study region (Fig. 5). To go further in the
reason for the within-basin variability of PAR is the direction of quantification will require a survey of
spatially and temporally uneven pattern of matrix modern population sizes of key tree species in the
sedimentation and pollen deposition in the lake RSAPs of modern sites used for calibration of the
basin (Davis et al., 1984). PAR values.

Lake Akuvaara Lake Tsuolbmajavri Lake Bruvatnet Lake Toskaljavri

within pine forest 30 km from 50 km from the 80 km from the
the pine treeline pine treeline pine treeline
pollen cm–2yr–1 pollen cm–2yr–1 pollen cm–2yr–1 pollen cm–2yr–1
0 500 1,000 1,500 2,000 0 500 1,000 1,500 2,000 0 500 1,000 1,500 2,000 500 1,000 1,500 2,000
0

0
2,000

2,000
4,000

4,000
cal yr BP
cal yr BP

6,000
6,000

8,000
8,000

10,000
10,000

Pine present

Pine dominant

Pine present

Pine dominant

Pine present

Pine dominant

Pine present

Pine dominant

Figure 5 Pollen accumulation rate (PAR) values provide a means of estimating past plant abundance independent of pollen percentage
values. Here, four Holocene PAR records of Pinus are presented from the Fennoscandian tree-line zone and compared with the modern
threshold PAR values delimiting the absence, presence, and dominance of Pinus in the lake surroundings. The Holocene records indicate the
invasion of Pinus into the region over 9 cal kyr BP and subsequent population expansion at about 8.3 cal kyr BP after which Pinus occurred
sparsely up to Lake Toskaljavri, at present about 80 km from the Pinus distribution limit. The retreat of the Pinus tree line started after 6 cal kyr
BP and accelerated at about 4 cal kyr BP. No significant late Holocene population reduction can be observed at Lake Akuvaara, where Pinus
has been the dominant tree species since its invasion. Modern threshold PAR values are based on the sediment core top values
from different vegetation zones in the same region (Seppä and Hicks, 2006). Data from Hyvärinen (1975) and Seppä and Hicks (2006).
Jackson, S. T. and Kearsley, J. B. (1998) Journal of Ecology 86, 474–490.
 POLLEN ANALYSIS, PRINCIPLES 2493

Applications of Pollen Analysis P < 0.0001 P < 0.0279

Local-Scale Studies 14 Cerealia absent

12 n = 611
As indicated by the pollen-representation models, the
10 Cerealia present

Pollen (%)
RSAPs for pollen records obtained from small forest n = 326
8
hollows or moss polster are about 50–200 m. Such
records reflect forest history on a stand scale, the 6

same spatial scale on which most modern forest ecol- 4

ogists operate, and the scale for which the classical 2
forest succession models describing the predictable 0
sequence of plant groups on annual to decadal time- (A) Fagus Picea
scales were developed. Consequently, paleoecologists
14
working on local-scale pollen studies can share the
concepts, research problems, techniques, and 12

perspectives of forest ecology. This provides oppor- 10

Pollen (%)
tunities for close collaboration with modern ecolo- 8
gists, whose understanding of local dispersal, 6
establishment, and other time-related ecological pro-
4
cesses emerges from observations of contemporary
processes and may thus underestimate the underlying 2

critical historical factors (Bradshaw, 1988). 0

c.
a

a
us

us

us

us

p
AP
Another major advantage of focusing on local- (B)

nu
ul

ce

ro
de
in

ag

yl

rc
et

N
th
Pi

Al
or
*P

ue

er
*B

*F

An
scale studies is that it is often possible to relate

**
*C
**

*Q

th
**

**

O
reliably past changes in species composition and

**
structure to external factors causing the changes. Figure 6 Stand-scale pollen records reflect the contrasting inva-
Independent evidence for the external factors can be sion patterns of Picea abies and Fagus sylvatica in northern
Europe. (A) Pollen percentage values (mean þ SD) with a total
obtained, for example, from analyses of sediment terrestrial pollen sum for P. abies and F. sylvatica in small hollow
charcoal, tree rings, historical archives, and climate samples with and without cereal grain pollen. The percentages of
reconstructions based on techniques independent of F. sylvatica are higher when cereal pollen are recorded in the same
vegetation dynamics. Comparisons of these records samples, while P. abies does not indicate similar relationship. (B)
The mean forest stand composition in southern Scandinavia during
with local-scale pollen records often demonstrate the
the invasion of P. abies and F. sylvatica. The percentage values
importance of such local processes as fire, human (mean þ SD) were adjusted by stand-scale correction factors that
disturbance, and storms in vegetation dynamics. allow for differential pollen representation. The light brown bars
Immigration or expansion of a species that is simul- indicate sites invaded by P. abies and the brownish green bars
taneous with a local disturbance such as fire has sites invaded by F. sylvatica. P. abies favored forest stands where
Betula, Pinus, or F. sylvatica were abundant, whereas F. sylvatica
likely taken place as a response to the altered growth was established in sites with Quercus, Alnus, Corylus, and other
conditions and interspecific competition. For exam- deciduous trees. All sites are totally or mostly ‘closed canopy’ sites
ple, the local establishment of beech at its distribu- with RSAP roughly comprising the surrounding 20–50 m (Bradshaw
tion limit in northern Europe seems to have been and Lindbladh, 2005).
greatly facilitated by human activities, especially dis-
turbance by fire, and this association with historical
Regional-Scale Studies
disturbances is reflected in the species’ modern pat-
chy distribution pattern (Bradshaw and Lindbladh, Conventionally, pollen records have been produced
2005) (Fig. 6). On the other hand, stand-scale inva- from small- to medium-sized lakes and mires and
sions during periods with no evidence for distur- reflect the vegetation of the surrounding hundreds
bance, such as that of hemlock in its western range of square kilometers. Pollen records on this scale
limit in North America, may be related to rapid or have had a major influence on our understanding of
gradual climate changes and associated changes in vegetation history and concepts of population, com-
local growing conditions (Parshall, 2002) (Fig. 7). munity, and biome dynamics during the Quaternary
These examples reflect the individualistic response period. Regional-scale pollen records have shown
patterns and invasion histories of tree species and dramatic vegetation changes during glacial–intergla-
demonstrate the great relevance of local-scale, pol- cial cycles in regions adjacent to ice sheets and have
len-analytical investigations for basic and applied provided unique insights into the nature of vegetation
forest ecology. during the previous interglacials. They have also shed
2494 POLLEN ANALYSIS, PRINCIPLES

Percentage of Tsuga pollen

0 10 0 10 20 0 10 20 0 10 20 0 10 20
0
st Expansion
Expansion
st
500 Expansion Expansion
p Colonization

1,000 st
Age (yr BP)

p
1,500 p Colonization Colonization
Regional
arrival of p
2,000 hemlock
st Colonization

2,500

3,000
Hemlock Drummond Cutacross Owen Diamond
Lake
(A) (B) (C) (D) (E)
Forest hollows
Figure 7 A combined use of stand-scale and regional-scale records can provide a detailed picture of past population dynamics, as
exemplified by the investigation of hemlock population history in Wisconsin, USA. Hemlock pollen percentage values from a lake
(A, Hemlock lake) show the regional arrival of the species at about 2 cal kyr BP. Four stand-scale records (B–E) from small hollows
indicate that after its spread, hemlock occurred in the region as scattered populations until a rapid expansion at about 500 cal yr BP. The
expansion has been a regionally synchronous event, suggesting a climatic change toward moister and cooler conditions. Records of
stomata (st) analyzed from the pollen samples corroborate the pollen evidence (Parshall, 2002).

light on significant vegetation zone shifts during the expansion occurred in response to a change of exter-
interglacials, such as the early to mid-Holocene nal factors. In many cases, it may be this expansion
expansion of the savanna vegetation into the modern that is reflected in the visible rise of the pollen curve,
Saharan desert in Africa, the northward advance of not the immigration. Motivated by these observa-
the boreal conifer forest zone of the Northern tions, some palynologists have also adopted a fresh
Hemisphere, and the major deforestations caused by approach to interpreting pollen data. A comparative
humans in many ecosystems during the late examination of combined and mapped pollen records
Holocene. from Beringia, a region that had been earlier inter-
By combining regional records over large areas, it preted as being tundra with no forests or even trees
has been possible to reconstruct the location of the present, showed consistent patterns indicating a
refugias of the main tree species during the last gla- presence of tree species during the Last Glacial
ciation and to map their subsequent spread during Maximum and subsequent expansions of the popula-
the interglacial (Huntley and Birks, 1983; Bennett, tions (Brubaker et al., 2004). Overall, the detection
1997). According to conventional interpretation, a of a presence of taxon from pollen data, especially in
Postglacial immigration of a new plant species to a a sparsely vegetated region, remains obscure, and the
region is reflected in a distinct rise of the percentage most reliable results can be achieved by an integra-
values of the corresponding pollen type. Many recent tion of pollen analysis with analyses of stomatal and
studies, however, have challenged this view by show- plant macrofossil from the same sediment cores
ing that the regional-scale pollen records may regu- (Birks and Birks, 2001).
larly provide underestimates for the date of
immigration of the new species into a given area. Continental-Scale Reconstructions
The occurrence of macrofossils and stomata of tree The continental-scale applications of pollen analysis
species have provided firm evidence for the presence usually involve pollen-stratigraphical analyses of
of tree species hundreds or thousands of years before long sediment cores obtained from ocean bottoms
any significant rise in the taxon’s pollen curve. This or from continental sites where sediment has
suggests that the species has occurred sparsely, prob- accumulated undisturbed over tens of thousand of
ably under conditions limiting both population years, for example, in crater lakes or lakes located
growth and pollen productivity, until a population in tectonic depressions. Long continental pollen
 POLLEN ANALYSIS, PRINCIPLES 2495

records have been obtained from Columbia, the pollen records and to examine whether the recently
western United States, France, Greece, and Italy realized abrupt climatic events and cycles typical of
(Hooghiemstra et al., 1993; Whitlock and Bartlein, the glacial periods influenced the vegetation of the
1997; Allen et al., 1999; Tzedakis et al., 2004). These continents adjacent to the Atlantic. Long continental
sediment cores span tens or even hundreds of thou- pollen records can also be used for quantitative cli-
sands of years and can cover full glacial–interglacial mate reconstructions, since on the timescale repre-
cycles. In addition to providing evidence of the sented by these records, vegetation patterns were
nature of vegetation during the previous interglacials predominantly controlled by the climate regimes.
and interstadials, the records reflect the pace of vege- Possibilities for precise reconstructions are limited
tation dynamics in response to the Milankovitch- by the often low level of analog between modern
scale climate dynamics, demonstrating alternations pollen assemblages and those representing the glacial
of forests and treeless vegetation types consistent periods. In general, however, the results indicate
with variations in solar insolation and global ice- trends that are comparable to those obtained by
volume. independent techniques such as the stable isotope
A difficulty with long pollen records is that their records of ice cores.
chronological control is, in general, weakly con-
strained, often being obtained by correlating distinct
events and periods of the pollen diagrams with pre- Conclusions and Future Directions
cisely dated, long records such as ice cores. More The main function of pollen analysis in Quaternary-
reliable correlations can be obtained by analyzing stratigraphical investigations has been to identify
continental and marine proxies from the same deep- chronostratigraphic periods, to correlate sediment
sea sequences or by applying tephrochronological sequences, and to provide general reconstructions of
techniques to generate more precise chronologies vegetation and climate. Conventional pollen dia-
for the continental cores (Fig. 8). Improved chrono- grams are still valuable when produced in new
logical precision has prompted palynologists to inves- regions or from key sites where long, uninterrupted
tigate the timing and duration of the interglacials in pollen records can be obtained, but in regions where

e
a lora
xa tax g ulif
axa ria n ta a n Li
oni
c
er t osibe ane e itae e ta
xa eae
Depth
nth
ic nkt n e r d i terr cacea m pos p p min us
(m) Be Pla Pio Eu Me Eri Co Ste Gra Pin
17 MIS 7a 17

18 MIS 7c 18

19 MIS 7e 19

20 20

21 21

MIS 9a
22 22

23 23
MIS 9c

24 24
MIS 9e

25 25

5 4 3 2 1 0 0 20 0 20 0 0 20 0 20 40 0 20 40 0 20 40 0 20 40 60 80
δ18σ %

Figure 8 A long, summary pollen diagram from marine sediment core MD01-2443, off the Portuguese coast (Tzedakis et al., 2004).
The corresponding marine isotope stages MIS 9e to MIS 7a and the benthic and planktonic foraminera isotope records are indicated on
the left. The estimated interval covered by the sequence is 190–350 kyr BP.
2496 POLLEN ANALYSIS, PRINCIPLES

the palynological studies have a long history, conven- Bradshaw, R. H. W. (1988). Spatially-precise studies of forest
tional pollen analysis is being replaced by carefully dynamics. In Vegetation History (B. Huntley and T. Webb,
designed studies that address specific questions and III, Eds.), pp. 725–751. Kluwer, Dordrecht.
Bradshaw, R. H. W., and Lindbladh, M. (2005). Regional spread
concentrate on specific time periods. This is particu- and stand-scale establishment of Fagus sylvatica and Picea
larly so in the application of pollen analysis in abies in Scandinavia. Ecology 86, 1679–1686.
Quaternary paleoclimatology, where the recent dis- Broström, A., Sugita, S., Gaillard, M.-J., and Pilesjö, P. (2004).
coveries of major Quaternary climate changes, such Estimating the spatial scale of pollen dispersal in the
cultural landscape of southern Sweden. The Holocene 15,
as the Heinrich Events during the last glacial, the
252–262.
abrupt warmings and coolings during the last glacial Brubaker, L. B., Anderson, P. M., Edwards, M. E., and
termination, and the short-lived cold period at about Lozhkin, A. V. (2004). Beringia as a glacial refugium for
8.2 ka, highlight the need to focus on high-resolution boreal trees and shrubs: New perspectives from mapped
pollen-stratigraphical data from optimally located data. Journal of Biogeography 32, 833–848.
Davis, M. B. (1963). On the theory of pollen analysis. American
sites in order to investigate the occurrence of these Journal of Science 261, 897–912.
climate events in the continental regions. Davis, M. B. (2000). Palynology after Y2K – Understanding the
Yet there is another major trend in the pollen ana- source area of pollen in sediments. Annual Review of Earth and
lysis, where the aim is to reconstruct past vegetation in Planetary Science 28, 1–18.
Davis, M. B., and Deevey, E. (1964). Pollen accumulation rates:
more precise terms. The fundamental objective is to
Estimates from late-glacial sediment of Rogers Lake. Science
provide quantitative vegetation reconstructions that 145, 1293–1295.
incorporate such central ecological attributes as flor- Davis, M. B., Moeller, R. E., and Ford, J. (1984). Sediment focus-
istic composition, species abundance, and the spatial ing and pollen influx. In Lake Sediments and Environmental
distribution of plants in an explicitly defined context History (E. Y. Haworth and J. W. G. Lund, Eds.), pp. 261–293.
University of Minnesota Press, Minneapolis.
of time and space, and assess the role of internal and Fægri, K., and Iversen, J. (1991). Textbook of Pollen Analysis.
external factors as drivers of change. For this purpose, Wiley, Chichester.
we must improve our understanding of modern rela- Hooghiemstra, H., Melice, J. L., Berger, A., and Shackleton, N. J.
tionships between vegetation patterns and pollen (1993). Frequency spectra and palaeoclimatic variability of the
high-precision 40–1450 ka Funza I pollen record (Eastern
input, including the critical processes of pollen pro-
Cordillera, Columbia). Quaternary Science Reviews 12, 141–
ductivity and dispersal from plants to the deposition 156.
site. Despite having made significant progress, pollen Huntley, B., and Birks, H. J. B. (1983). An Atlas of the Past and
analysts are still far from reaching these ambitious Present Pollen Maps for Europe: 0–13000 Years Ago.
goals, but remain motivated by the breakthrough Cambridge University Press, Cambridge.
Hyvärinen, H. (1975). Absolute and relative pollen diagrams from
potential of the method in truly quantitative vegeta- northernmost Fennoscandia. Fennia 142, 1–23.
tion reconstructions. Jackson, S. T., and Kearsley, J. B. (1998). Quantitative representa-
tion of local forest composition in forest-floor pollen assem-
blages. Journal of Ecology 86, 474–490.
See also: Paleobotany: Overview. Pollen Methods and Jacobson, G., and Bradshaw, R. H. W. (1981). The selection of
Studies: Use of Pollen as Climate Proxies; Numerical sites for paleovegetational studies. Quaternary Research 16,
Analysis Methods; Surface Samples and Trapping; Stand- 80–96.
Scale Palynology; BIOME Model of Vegetation Moore, P. D., Webb, J. A., and Collinson, M. E. (1991). Pollen
Reconstruction; POLLSCAPE Model. Analysis. Blackwell, Oxford.
Parshall, T. (2002). Late Holocene stand-scale invasion by hem-
lock (Tsuga canadensis) at its western range limit. Ecology 83,
1386–1398.
References Pohl, F. (1937). Die Pollenerzeugung der Windblütler. Beiheftezum
Botanischen Centralblatt 56, 365–470.
Allen, J. R. M., Brandt, U., Brauer, A., et al. (1999). Rapid envir- Prentice, I. C. (1988). Records of vegetation in time and space: The
onmental changes in southern Europe during the last glacial principles of pollen analysis. In Vegetation History (B. Huntley
period. Nature 400, 740–743. and T. Webb, III, Eds.), pp. 17–42. Kluwer, Dordrecht.
Andersen, S. T. (1970). The relative pollen productivity and pollen Prentice, I. C., and Parsons, R. W. (1981). Maximum likelihood
representation of north European trees, and correction factors linear calibration of pollen spectra in terms of forest
for tree pollen spectra. Danmarks Geologiske Undersøgelse, II composition. Biometrics 39, 1051–1057.
96, 1–99. Seppä, H., and Hicks, S. (2006). Intergration of modern and past
Bennett, K. D. (1997). Evolution and Ecology: The Pace of Life. pollen accumulation rate (PAR) records across the arctic tree-
Cambridge University Press, Cambridge. line: A method for more precise vegetation reconstructions.
Bennett, K. D., and Willis, K. (2003). Pollen analysis. In Tracking Quaternary Science Reviews (in press).
Environmental Change Using Lake Sediments (J. P. Smol, H. J. B. Sugita, S. (1993). A model of pollen source area for an entire lake
Birks and W. M. Last, Eds.), pp. 5–32. Kluwer, Dordrecht. surface. Quaternary Research 39, 239–244.
Birks, H. H., and Birks, H. J. B. (2001). Future uses of pollen Sugita, S. (1994). Pollen representation of vegetation in
analysis must include plant macrofossil analysis. Journal of Quaternary sediments: Theory and method in pacthy vegeta-
Biogeography 27, 31–35. tion. Journal of Ecology 82, 881–897.
 POLLEN METHODS AND STUDIES/Use of Pollen as Climate Proxies 2497

Tzedakis, P. C., Roucoux, K. H., de Abreu, L., and Shackleton, N. J. Relevant Website
(2004). The duration of forest stages in southern Europe and
interglacial climate variability. Science 306, 2231–2235. [Link] –
Whitlock, C., and Bartlein, P. J. (1997). Vegetation and climate Uppsala Universitet: Institutionen för Geovetenskaper:
change in northwest America during the past 125 kyr. Nature Paleobiologi.
388, 57–61.

POLLEN METHODS AND STUDIES

Contents
Use of Pollen as Climate Proxies
Reconstructing Past Biodiversity Development
Numerical Analysis Methods
Databases and their Application
Surface Samples and Trapping
Stand-Scale Palynology
Changing Plant Distributions
BIOME Model of Vegetation Reconstruction
POLLSCAPE Model
Archaeological Applications

Use of Pollen as Climate The increasing concern over the current and future
changes in climate has led to greater interest in the
Proxies use of pollen studies to quantify past climate changes,
S Brewer, J Guiot and D Barboni, Université Paul notably for periods of large variations in climate such
Cézanne, Aix-en-Provence, France as the transition between Glacial and Interglacial
ª 2007 Elsevier B.V. All rights reserved. periods. The results obtained by the transformation
of pollen data into climate equally form an ideal
source of data for testing the output of global circu-
lation models during the past (paleo-GCMs), a pro-
Introduction cess that allows an assessment of these models for
Climate strongly influences the geographical distribu- future predictions under different conditions (see
tion of plants and of vegetation types because not all Data-Model Comparisons).
plants have evolved the same tolerances for extreme Past variations of the climate have influenced the
temperatures and the same temperature and moisture distribution of terrestrial vegetation, and the changes
requirements for optimal photosynthesis (Woodward, in these distributions may be reconstructed by study-
1987). Past variations of the climate will therefore ing the record of fossil pollen recorded in lake sedi-
have influenced the distribution of terrestrial vegeta- ments and peat bogs (see Pollen Analysis, Principles).
tion, and the changes in these distributions may be The underlying principle of all attempts to recon-
reconstructed by studying the record of fossil pollen struct climate from pollen data is that of uniformitar-
recorded in lake sediments and peat bogs (see Pollen ianism, that is, the information required for
Analysis, Principles) However, the relation between understanding past conditions may be obtained
the pollen composition of lake sediments (pollen from the relationship between the modern distribu-
diversity and relative abundance) and the species com- tion of the taxa concerned and the climate (Birks and
position of the surrounding vegetation is not straight- Birks, 1980). The distribution of each taxon is con-
forward because pollen production, dispersion, and trolled by a set of climatic limits, and combinations
preservation vary from one species to the other. In of taxa indicate a particular set of conditions, with a
order to reconstruct past climates from pollen data, greater or lesser accuracy. By relating the modern
it is therefore necessary to directly relate modern pol- distribution of taxa and climate, and subsequently
len data to climatic values, and then apply this relation applying this relationship to the observed assemblage
to fossil pollen samples. of fossil pollen taxa, it is then possible to reconstruct