Bruce R. Gervais, Department of Geography, 2555 East San Ramon Ave M/S SB69, California State University, Fresno, CA 93740, USA (E-mail email@example.com; tel. +559-278-2678; fax +559-278-7268).
1A sediment core recovered from Poteryanny Zub Lake on the Kola Peninsula of Russia (68°48′91″ N, 35°19′32″ E) provides fossil pollen and stomate evidence for Pinus sylvestris treeline development and movement. The site today lies in birch forest tundra, approximately 25 km north of the pine treeline and 90 km south of the Murman Coast.
2The record begins approximately 10 050 14C years before present (bp). The post-glacial tundra, typified by various herbaceous taxa and Artemisia, was followed by Betula-dominated woodland forest with Polypodiaceae-Gramineae pollen types. Such vegetation assemblages are not analogous to any in the western Kola Peninsula today.
3Stomate evidence demonstrates that Pinus sylvestris immigrated to the study site as early as 8150 14C bp. Pine forest probably expanded northwards from sparse or small disjunct populations, which established c. 1000 14C bp before the expansion evident in the pollen and wood macrofossil records (7200 14C bp). Betula and other shrub and herb taxa declined at this time, possibly as a result of shading by the pine canopy.
4The presence of fossil Pinus sylvestris stomates and wood, together with AP : NAP ratios, indicates that the pine treeline was some 100 km north of its present position by the end of the middle Holocene warming (7000–6000 14C bp). This places it near the Barents Sea coast, currently tundra vegetation. An approximate 2 °C increase in regional summer temperatures would be necessary to move the treeline to the present-day coastline.
5From 6000 14C bp the pine treeline gradually retreated southward to its present modern position. Birch woodland has grown at the site since c. 3000 14C bp.
The Holocene history of boreal forest development and treeline history in northern Fennoscandia has long been a topic of biogeographical and ecological interest. Holocene treeline establishment and changes have been reconstructed primarily by means of pollen and wood macrofossil analysis (Hyvärinen 1975, 1976; Eronen & Hyvärinen 1981; Birks et al. 1996; Seppä 1996; Mäkelä 1997). Pollen percentage and influx data have provided evidence of major latitudinal and altitudinal shifts of the northern treeline in response to global climate controls, but it is difficult to infer confidently the local presence of forest trees from pollen percentage data (MacDonald 1993). Radiocarbon-dated tree subfossil wood remains have also been widely used to infer the timing of past changes in the position of northern forests. Recent subfossil wood and other macrofossils from northern Scandinavia provide evidence that the post-glacial northward movement of forest trees may have occurred up to several millennia earlier than inferred in pollen studies (Kullman 1995a, 1995b, 1996, 1998a). It appears that, in some instances, the first appearance of a forest tree species may be indicated by preserved wood, cone or seed remains but be undetected in pollen records (e.g. Kullman 1998b). A significant drawback to subfossil wood analysis, however, is the difficulty of finding full continuous records that represent both the earliest and latest presence of a given taxon. Smaller plant remains, such as needles, seeds or lignified stomates, which are often found in lacustrine sediments, are becoming more frequently used in treeline reconstructions, and offer considerable promise by providing continuous evidence of the local presence of treeline conifers (e.g. Hansen et al. 1996; Clayden et al. 1997).
Despite the numerous studies of Holocene treeline development and change in northern Fennoscandia, until recently there was little such information available for the adjacent Kola Peninsula of Russia. This study addresses post-glacial development and changes in the conifer treeline position on the Kola Peninsula – defined here by the northern limits of Pinus sylvestris L. (Scots pine) – by presenting new pollen and pine stomate evidence, which can be augmented with previously reported subfossil pine stump evidence (MacDonald et al. 2000a). The problem of identifying the first local presence of pine is addressed by extending analyses of the pollen and wood data to include pine stomates. Stomates from conifer needles generally provide continuous and relatively reliable evidence of the local presence of taxa at the northern and alpine treeline, including the pine treeline on the Kola Peninsula (Hansen 1995; Clayden et al. 1996, 1997; Hansen et al. 1996; David 1997; Gervais & MacDonald 2001).
Information regarding treeline development and movement on the Kola Peninsula is important to understanding long-term vegetation dynamics in adjacent Fennoscandia. Hyvärinen (1975) reported pine forests occurring in the Varanger region of northern Finland as early as 8500 14C years before present (bp). Although hard water may have affected the dates from samples of bulk organic lake sediments, later research validated these findings (Eronen & Hyvärinen 1981; Seppä 1996; Mäkelä 1997). It is, however, unclear how these far northern regions came to be among the first in Fennoscandia to be forested with pine (Hyvärinen 1975; Eronen & Hyvärinen 1981; Seppä 1996). If the Kola Peninsula is a potential eastern route by which Pinus sylvestris entered northern Fennoscandia and subsequently expanded southward (Seppä 1996), then pine immigration should be registered here prior to the northern coastal region of Fennoscandia. Previous palaeoecological studies from the Kola Peninsula have been at low temporal resolution and/or lack independent chronological control (Kremenetski & Patyk-Kara 1997; Kremenetski et al. 1997). Here we present evidence from the treeline on the western Kola Peninsula that allows us to address the nature and timing of post-glacial vegetation change and the northward advance of Pinus sylvestris.
The Kola Peninsula is bounded by the Barents Sea to the north and the White Sea to the south. Its geology consists of Precambrian crystalline igneous rocks of the Baltic Shield, overlain by a mantle of Late-Quaternary glacial till, glaciofluvial deposits and peat. The regional climate is heavily influenced by the relatively warm Murman Coast current, a branch of the North Atlantic current. Mean temperatures range from −9 °C to −13 °C in January and + 9°C to + 14°C in July. Mean annual precipitation is in the order of 500–700 mm and occurs primarily in summer and autumn (Murmansk Atlas 1971; Arctic Atlas 1985). The latitudinal coniferous treeline in this region is defined by the northern limit of Pinus sylvestris, which occurs between the northern tundra and the boreal forest to the south (Murmansk Atlas 1971). The other important conifer in the region is Picea abies[L.] (Karst) (Norway spruce), which currently occurs in the southern, central and north-west Kola Peninsula.
The coring site, informally named Poteryanny Zub Lake (Lost Tooth Lake), is located at 68°48′91′′ N, 35°19′32′′ E, some 25 km north of the northern pine limit and approximately 90 km south of the Murman Coast (Fig. 1). This lake was chosen for its small size, limited fluvial input and its location just north of the modern treeline, which makes it ideal for detecting previous latitudinal displacement of the pine treeline. Poteryanny Zub Lake is 131 m a.s.l., has an area of c. 3.2 ha and has a maximum depth of 5.5 m. The core site was in the deeper southern lobe of the kidney-shaped lake. The northern lobe is significantly shallower and supports emergent aquatic plants. Eroding peat banks are exposed along portions of the shore, and the lake has one small inlet and one outlet stream.
Botanical nomenclature follows Tutin et al. (1993). The nearest Pinus sylvestris stands grow approximately 25 km south from the study site. Surrounding vegetation is described according to published vegetation maps (Murmansk Atlas 1971; Arctic Atlas 1985) and field observations. The surrounding field-layer vegetation is dominated by Betula spp., including Betula pubescens ssp. tortuosa, Betula nana and Betula pendula, interspersed with patches of Juniperus communis. Salix spp. grow at the moistest sites on the lake shore, whereas Empetrum nigrum and Vaccinium myrtillus are common at dry sites. Vaccinium vitis-idaea and Ledum palustre are less abundant. Sphagnum spp. cover the ground layer in the moist depressions. Other local taxa include Calluna vulgaris, Cornus suecica, Epilobium angustifolium, Equisetum spp., Eriophorum vaginatum, Gymnocarpium dyropteris, Rubus saxatilis, Rubus chaemamorus, Trollius europaeus and Vaccinium uliginosum. Ground lichens, primarily Cladonia, are ubiquitous.
SEDIMENT SAMPLING AND PROCESSING
Lake coring was conducted in April 1998 from the surface ice over the deepest portion of the lake. Four duplicate 2.5 m sediment cores were taken with a modified Livingstone piston corer (Wright et al. 1984). The cores were extruded in the field, wrapped in plastic and aluminium foil and secured in plywood containers for transport to the laboratory. The cores were stored at c. 4 °C until subsampled. Subsamples of 1 cm3 were taken every 2 cm from 1 to 82 cm and 164 to 250 cm of the core in order to detect potential high-frequency early- and late-Holocene climatic events. Otherwise, the core was subsampled at 4-cm intervals for pollen, stomate and loss-on-ignition (LOI) analysis (Dean 1974). Subsamples were taken from the interior of the core to minimize the chance of contamination. Laboratory processing of the sediment was performed at the University of California at Los Angeles (UCLA) and followed Faegri & Iversen (1964). Two calibrated Lycopodium clavatum tablets (c. 12 544 spores per tablet) were added to each subsample to calculate pollen influx values (Stockmarr 1971; Maher 1972; Davis et al. 1973). Naturally occurring Lycopodium clavatum were distinguished from the L. clavatum marker spores on the basis of the dark colouration and crumpled condition of the latter. Only terrestrial pollen types were included in the pollen sum that was used to calculate pollen percentages, including indeterminate grains (corroded, folded and concealed) and unknown grains. Extra-regional types, such as Corylus, Juglans and Ulmus, and terrestrial spores were not included in the cumulative pollen sum. Processed sediment residue was stained and mounted in silicone oil on microscope slides. Minimum counts of 500 terrestrial pollen grains per subsample were made at 400× magnification. Pollen nomenclature is based on Moore et al. (1991), and identification was based on pollen keys, reference photomicrographs (Moore et al. 1991; Rielle 1998), and the UCLA pollen reference collection.
Stomate content was determined while counting pollen. In samples where stomates were not encountered during routine pollen counts, a minimum of 10% of the sediment was scanned (2500 Lycopodium markers) for stomates at 100× magnification. Conifer stomates were identified using Hansen's (1995) key and the UCLA stomate reference collection. Occasionally stomates thought to belong to unidentified non-coniferous taxa were found, but these were easily differentiated from coniferous stomates and were not included in the analysis.
RADIOCARBON DATES AND CHRONOLOGY
Following subsampling, the core was split in half and 1-cm-wide cross-sections were taken at regular intervals and rinsed with distilled water through a 0.6-mm screen to isolate plant macrofossils for radiocarbon dating. Accelerator mass spectrometry (AMS) was used to determine radiocarbon ages for these macrofossils. For intervals where they were scarce sediment material was used for radiocarbon control (Table 1). AMS preparation was performed at the University of Colorado at Boulder INSTAAR Laboratory for AMS Radiocarbon Preparation and Research. Sedimentation rates for the core were estimated by fitting a third-order polynomial to the radiocarbon ages. Pollen and spore influx rates for major taxa were also calculated.
Table 1. List of radiocarbon dates from Poteryanny Zub Lake
14C age bp±1 SD
1800 ± 60
4630 ± 50
5180 ± 50
5720 ± 50
8860 ± 65
8890 ± 55
9500 ± 65
9410 ± 70
Approximately 80 pollen and spore types were identified, of which 25 dominant types were included in the pollen diagram. A complete list of pollen and spore data may be obtained from the authors. A stratigraphically constrained incremental sum of squares cluster analysis (program CONISS) of pollen percentages was performed to establish local pollen assemblage zones. Only terrestrial pollen taxa with values of 2% or greater were included in this analysis.
Pollen percentages from the transect of lake surface samples on the Kola Peninsula (Gervais & MacDonald 2001) were ordinated by principal components analysis (PCA), using the statistical program CANOCO (Jongman et al. 1995). The rates and directions of vegetation change through time in the long-core record were estimated and compared with the modern pollen and spore spectra using the same PCA analysis at 500-14C year intervals. Samples within 100 years of each 500-year interval were averaged for the analysis, and only terrestrial taxa having two or more consecutive values of 2% or greater were included (i.e. Juniperus communis, Betula, Picea, Pinus, Salix, Asteraceae, Cyperaceae, Gramineae, Lycopodiaceae, Polygonaceae, Polypodiaceae and Ericales). Although the finer ecological details provided by rarer pollen types are lost, we restricted our analysis largely to broad family groupings in order to obtain a generalized, taxonomically non-biased pollen signal. Ordination of surface samples formed three major vegetation types (delineated in the PCA by ellipses) of tundra, birch woodland and pine forest vegetation types (Gervais & MacDonald 2001). The long-core time series samples were ordinated passively with these surface sample data in order to compare the fossil vegetation assemblages with modern ones.
Ratios of local arboreal pollen (AP) (Pinus, Picea, Betula, Alnus and Sorbus) and all non-arboreal pollen and spores (NAP) were calculated from the pollen percentages at each sample interval by dividing the sum of AP by the sum of NAP. In a previous study (Gervais & MacDonald 2001), 31 lakes along a c. 200-km latitudinal transect, spanning from the northern Murman Coast south to the interior spruce-pine boreal forest, were sampled and analysed for their modern pollen spectra. An AP : NAP ratio was regressed against geographical distance from treeline in order to derive a statistical model to estimate the modern treeline position using the pollen spectra. Applying this linear regression equation, Y = −12.491X + 84.963 (R2 = 0.74, P < 0.000) (where X is the pollen ratio and Y is the distance in kilometres along the study transect), to the long-core record allowed us to estimate treeline location relative to the study site for any given period of time during the record. Pollen ratio and treeline displacement estimations were graphed and fitted with a smoothing 11-term weighted moving average. We did not separate Betula pollen grains into tree and non-tree types. Inclusion of shrub-form birch (e.g. B. nana) in the arboreal pollen ratio is likely to contribute some error to the model, particularly in the early, non-analogue portion of the core, when birch woodland dominated the site, but the high R2 value and significance of the model gives us confidence in its skill.
CORE STRATIGRAPHY AND RADIOCARBON AGES
The sediment from the Poteryanny Zub Lake core is largely composed of brownish-green lacustrine gyttja. Macroscopic plant fragments are found in portions of the core but are never abundant. The basal sediment from approximately 250–240 cm is composed of a mixture of sand and gravel (Fig. 2). This sediment grades abruptly to fine silty grey clay with low organic content from 240 cm to approximately 230 cm. The transition from fine silty clay to gyttja is gradual, beginning roughly at 225 cm. Above 220 cm organic gyttja occurs for the remainder of the core.
Eight AMS-radiocarbon ages were obtained from plant fragments or bulk sediment (Fig. 2, Table 1). The youngest age (1800 ± 60 14C bp) at 28 cm depth was measured from a woody twig, while the oldest date (9500 ± 65 14C bp) occurs at 220 cm depth and was obtained from an unidentified bark fragment. The two samples of aquatic moss produced radiocarbon ages younger than indicated by the wood or bulk sediment (Fig. 2). The moss sample at 230 cm thus produced a date younger than the bark sample found at 220 cm (Table 1). Terrestrial macrofossils are normally considered more reliable than aquatic moss when analysed for radiocarbon age estimates because of hard-water effects (MacDonald et al. 1991). However, the hard-water effect is not an important factor at our site because of the lack of carbonate rocks in the region. Terrestrial plant macrofossils could have been reworked from deposits in or outside the lake. Rather than arbitrarily omit inverted dates to permit linear interpolation of interval ages, we included all radiocarbon dates in the analysis and fitted a third-order polynomial to them to estimate the age of specific intervals and to calculate sediment and pollen accumulation rates: Y = 0.00042X3 − 0.283X2+ 84.6645X, where X is the sample depth and Y is the corresponding age of the sample. The fit has an R2 value of 0.98 (Fig. 2). The extrapolated basal age of the core is 10 050 year bp. The gradually changing sedimentation rate from relatively fast to slow suggests that a gradual decrease in sediment focusing is occurring at the lake (Lehman 1975).
The pollen percentage diagram was divided into six informal assemblage zones using the stratigraphic clustering program CONISS. Major changes in pollen and spore percentages (Fig. 3) and influx rates (Fig. 4b,c) are outlined here.
Zone 1A: 10 050–9760 14C bp (250–236 cm)
The base of Zone 1A is typified by herbaceous and shrub species, particularly Artemisia (≤ 10%) Gramineae (≤ 15%), Cyperaceae (≤ 17%), Rumex acetosella type (≤ 21%), Salix (≤ 35%) and Betula (≤ 30%). Organic content in this zone is very low, less than 5%. As much as 15% of the pollen is crushed and unidentifiable. Cumulative pollen accumulation rates for this zone, calculated from the sum of terrestrial and aquatic pollen and spore influx, quickly increase, beginning at c. 500 grains cm−2 year−1 and rising to almost 2000 grains cm−2 year−1.
Zone 1B: 9760–9070 14C bp (236–204 cm)
The base of Zone 1B is marked by rapidly increasing values of Betula, changing from 30% to 85%, and a concurrent decline in Salix, Artemisia, Cyperaceae and Gramineae. A spike of Ericales occurs, reaching as high as 15%, and Gramineae, which forms an important constituent of the herbaceous canopy, reaches values of 35%. Lycopodium clavatum and ferns (Polypodiaceae) begin to increase, each reaching 5%, while Pediastrum, an aquatic algae, displays a pronounced spike in this zone of 35%. Organic content of the sediments increases from 5% to 45% and pollen accumulation rates from approximately 2000 grains cm−2 year−1 to 6000 grains cm−2 year−1.
Zone 1C: 9070–8230 14C bp (204–168 cm)
This zone is marked by a decline in Ericales. Betula reaches sustained maximum values of almost 90%. Most herbaceous taxa, while still present, are no longer a large component of the pollen spectrum. Gramineae, however, maintains values of 8%, and Polypodiaceae reaches its highest values (25%). Lycopodium clavatum reaches its maximum value of 11%. Maximum values of sediment organic content occur, with LOI reaching as high as 60%. Pollen accumulation rates fluctuate around 8500 grains cm−2 year−1 and reach their highest value of c. 12 000 grains cm−2 year−1.
Zone 1D: 8230–7150 14C bp (168–130 cm)
Pine pollen percentages begin at 10% at the base of this zone and reach as high as 30%. Betula pollen remains high, ranging from 80% to 90%, but experiences a gradual decline. Lycopodium clavatum also declines. Gramineae and Polypodiaceae sustain values of 8% and 15%, respectively. Alnus pollen values rise for the first time, and Juniperus is also present, with values for some samples reaching 5%. Organic content of the sediment remains between 45% and 50%. Pollen accumulation rates begin to decline and range from c. 8000 to c. 4000 grains cm−2 year−1. Pine stomates, although not in every sample, appear regularly, beginning at 8150 14C bp near the base of the zone.
Zone 2A: 7150–2430 14C bp (130–32 cm)
Pine is the most important taxon, contributing between 45% and 60% of pollen for much of this period. Betula decreases to approximately 40%. Alnus pollen remains at 3% and Ericales increases to sustained values of 5%. Pediastrum also increases at this time, as does Sphagnum, and Juniperus values are consistently depressed (< 3%). Organic content of the sediment remains between 45% and 50%. The pollen accumulation rate reaches a peak of almost 8000 grains cm−2 year−1 and gradually declines to values near 2000 grains cm−2 year−1. Pine stomata occur at all intervals but end abruptly at the end of this zone at 2700 year bp. Pine stump macrofossils found in the vicinity (MacDonald et al. 2000a) also date from this period.
Zone 2B: 2430 14C bp to present (32–0 cm)
Pinus and Betula percentages remain constant at 30% and 55%, respectively, to the present day. Juniperus increases to 10%. Ericales remains at 5%, Gramineae gradually increases to almost 10%, and Corylus values rise to 5%. Organic content of the sediment ranges between 45% and 50% from the base of this zone to present. Pollen accumulation rates range from 2000 grains cm−2 year−1 to c. 500 grains cm−2 year−1 at the top of the core.
PCA-based comparison of fossil and modern pollen samples indicates that substantial changes in vegetation composition have occurred during the period under investigation (Fig. 5). For the modern surface samples (Gervais & MacDonald 2001), Polypodiaceae, Gramineae, Ericales, Cyperaceae and Juniperus have large negative loadings on the first axis, while Pinus and Picea have the highest first-axis loadings. The first principal component, axis 1, accounts for 62% of the variance in the pollen spectra. The second axis has high positive loadings by herbaceous and shrub taxa (Polypodiaceae, Gramineae, Ericales) and a high negative loading by Betula. Picea and Pinus load close to the origin on the second axis. The second axis accounts for 37% of the variance in the pollen spectra.
Fossil pollen sample ordinations plotted over time show a rapid progression from high to low second-axis scores between 10 000 and 9000 14C bp, while first-axis scores change little. From 9500 to 8500 14C bp first- and second-axes scores remain low. The fossil pollen assemblages from the early period between 10 000 and 7500 14C bp plot away from any modern pollen samples from the Kola Peninsula and should be considered to represent a vegetation that has no large regional analogue on the Kola Peninsula today. Between 8000 and 7000 14C bp first axis scores rise rapidly and following c. 7300 14C bp the vegetation becomes analogous to modern birch woodland. Between 6000 and 3500 14C bp first-axis scores reach their maximum, indicating the dominance of pine-forest vegetation. From 4000 to 1500 14C bp the first- and second-axis scores are reduced, indicating the southward recession of the pine treeline and increasing importance of Betula at the site as birch woodland is re-established.
Rates of floristic composition change can be approximated by the changes in distance in ordination space over time. Rapid change occurs in the early Holocene, with the immigration of herbaceous taxa and the Betula forest between 10 000 and 9000 14C bp. Rapid change also occurs between 8000 and 7000 14C bp during the period of rapid pine forest establishment and between 3500 and 2000 14C bp during the period of pine treeline retreat.
Treeline movement was calculated using AP : NAP ratios. The pollen spectra for the period prior to c. 7300 14C bp do not have an analogue to modern pollen samples and therefore may provide misleading results. However, from 7300 to 6000 14C bp, the treeline advanced c. 100 km north of the lake study site, placing it near the present Murman Coast (Fig. 6). Stomate evidence from a site near the Murman Coast suggests that pine had indeed reached that region at this time (Snyder et al. 2000). Starting at c. 6500 14C bp the treeline retreated southward until c. 2500 14C bp. From 2500 14C bp to present, the treeline has remained relatively stationary, with a slight northward shift occurring in the last 1300 14C bp.
GENERAL VEGETATION RECONSTRUCTION
Vegetation development at the study site is altogether typical of similar sites to the west in northern Fennoscandia (e.g. Hyvärinen 1975; Prentice 1982; Seppä 1996; Barnekow 1999). Earliest vegetation assemblages were unlike any large associations encountered in the landscape today, while inferred vegetation of the past 7000 years is largely similar to modern vegetation communities in the region. The non-analogue vegetation (10 050–7200 14C bp) represents a relatively complex early Holocene successional sequence of vegetation change dominated by herbaceous taxa and Betula spp. The last few decades of the widespread Younger Dryas (YD) cooling interval are likely to occur in the pollen sequence at the very base of the core (Zone 1A), with rapidly decreasing values of Artemisia and rapidly increasing values of Salix, marking the close of the YD (Hyvärinen 1975; Prentice 1982; Björck et al. 1996; Seppä 1996).
Early Holocene vegetation consisted of a relatively rich herbaceous flora, which was rapidly replaced by dense birch woodland (Betula spp.) with a characteristic understorey of lycopods (Lycopodium spp.) and pteridophytes (Polypodiaceae) and herb angiosperms. Stomate and pollen evidence indicates that the immigration and population expansion of pine at the study site was slow between 8150 and 7200 14C bp and rapidly accelerated following 7200 14C bp. After 3500 14C bp pine forest retreated south of the study site, to be replaced by birch-forest tundra up to the present day. Corylus does not now grow on the Kola Peninsula, and its high values following 3500 14C bp probably indicate more extra-regional pollen deposition as a result of the more open vegetation following the southward retreat of pine rather than the local presence of this taxon. Accordingly, Corylus pollen deposition today is greatest in the northernmost (i.e. open tundra) sites (Gervais & MacDonald 2001).
pinus sylvestris arrival and early holocene vegetation response
It is difficult to determine from pollen percentages when a tree taxon arrives at a given study site. On the Kola Peninsula, pine pollen values in modern lake sediments never drop below 22% of the terrestrial sum, even in the northernmost tundra regions (Gervais & MacDonald 2001). In the early Holocene, however, stomate data for Poteryanny Zub Lake indicate that pollen percentage values as low as 4% coincide with the local presence of pine at 8150 14C bp (Fig. 3). This discrepancy between low pollen percentages and stomate evidence of pine presence is to be expected as early Holocene pine forests on the Kola were not analogous to contemporary ones and are therefore not comparable in this way.
Pine pollen influx rates (grains cm−2 year−1) of 500 or greater have been used to infer local presence of Pinus sylvestris in northern Fennoscandia (Hyvärinen 1975). However, pollen influx values are influenced by sediment focusing, fluvial input of pollen and radiocarbon control, making this 500-count threshold potentially unreliable (Davis et al. 1973, 1984; Hyvärinen 1976). First and last pine values above this threshold (8150 and 2600 14C bp, respectively) coincide remarkably well with the first and last occurrences of stomates at this site (8150 and 2700 14C bp, respectively). We consider the combination of fossil stomates and pollen influx data to be conclusive evidence that pine was present at our site by 8150 14C bp.
Pollen influx evidence suggests that sites located very near the Barents coastline supported the earliest pine forests in northern Fennoscandia (Hyvärinen 1975; Eronen & Hyvärinen 1981; Seppä 1996; Snyder et al. 2000) around 8500 14C bp, roughly contemporary with its appearance on the northern Kola Peninsula. It cannot therefore be concluded that the Kola Peninsula provided a corridor for early Holocene pine migration to the northern Fennoscandian coastal regions.
The first significant and continuous rise in pine pollen percentage values at our study site occurs at 7200 14C bp, c. 1000 14C years after the first stomate occurrence. This demonstrates that individuals or small populations of pine, difficult to detect using pollen or macrofossil data, established in the early Holocene well before forest density increased enough to add significant amounts of pine pollen to the spectrum (Fig. 3). Small disjunct tree populations establishing far in advance of the main range limits of northward-migrating forests typify many other range expansions in northern Europe and North America (Bennett 1985, 1988; Peteet 1991; Kullman 1995b, 1998a, 1998c; Mäkelä 1997; Barnekow 1999). Coniferous tree species, including Pinus sylvestris, Larix sibirica and Picea abies, have gone entirely undetected in pollen records for part or all their Holocene history in northern Scandinavia (Hafsten 1992; Kullman 1995a, 1996, 1998a, 1998b). Pine seeds, winged and evolved for long-distance transport, probably dispersed early and more or less synchronously throughout the northern Fennoscandian region (Birks 1993; Kullman 1998c). Until climatic amelioration, these early propagules would probably have been restricted in terms of population size and geographical extent, to specific microhabitats, making them quite difficult to detect by whatever means. Subsequent population expansion from these outlying stands was probably the result of differences in the timing of climatically controlled regional population expansion rather than the timing of first arrival.
Early Holocene winter insolation values were lower than present in northern Fennoscandia (Kutzbach 1987; MacDonald et al. 2000b). Climatic factors were probably more amenable to pine forest expansion when the treeline was located along the northern coastal fringe rather than when it became located further inland to the south. Although Pinus sylvestris does well under continental climates with sparse snow cover (Kullman 1993), relatively severe winter temperatures with reduced snow cover have been inferred for the time in question (Ruddiman & McIntyre 1981; Selsing & Wishman 1984; Huntley 1988; Kutzbach & Gallimore 1988; Kullman 1998b) and could have resulted in pine seedling root desiccation or meristem tissue damage, preventing successful pine establishment at inland sites (Grace & Norton 1990; Kullman 1991; Gervais & MacDonald 2000; MacDonald et al. 2000a). In southern Scandinavia today, Pinus sylvestris is able to grow at a higher elevation along the coast where the climate is strongly oceanic, while at sites to the north-east its altitudinal range limit is relatively reduced by a cooler, more continental climate (Eronen 1979; Selsing & Wishman 1984; Eronen & Huttunen 1987). Maximum sea surface temperatures in the North Atlantic region occurred at 8500–8000 14C bp (Koçet al. 1993). It is plausible that early Holocene ocean-atmosphere conditions significantly ameliorated winter temperatures at sites located very near the coastline, allowing for limited pine forests to be established as soon as seeds arrived, while conditions farther inland were still largely unfavourable for their expansion.
The early stages of pine immigration coincide with the period of maximal birch forest establishment. Competitive displacement of birch by pine at this time has been reported from sites throughout northern Fennoscandia (e.g. Hyvärinen 1975, 1976; Mäkeläet al. 1994; Mäkelä 1997). At our site, Betula appears to respond to early pine immigration by slightly decreased influx rates following the first pine stomate at 8150 14C bp and more significant declines later as pine influx values rise at c. 7200 14C bp (Fig. 4b). As it appears that pine displaced birch on the landscape, it is possible that the rise of pine was slowed by competition from birch. Pines, for example, do not establish well beneath closed birch canopies. However, it is also possible that climatic change between 8150 and 7200 14C bp caused both the increase in pine and decrease in birch. Appearance of the first pine stomate coincides with decreases in herb-type pollen and spore counts. A protracted decrease in herb-type pollen and spores occurs when pine influx values rise sharply from 7200 14C bp (Fig. 4a,b), remaining depressed until 2700 14C bp, the date of the last pine stomate. The decline in abundance of terrestrial herbs is likely to reflect shading out of these light-demanding plants by the pine canopy.
MID-HOLOCENE PINE TREELINE MAXIMUM EXTENT AND RETREAT
Our AP : NAP ratio appears to be a relatively reliable approximation of treeline position. PCA analysis, pollen percentage and pine stomate data, pine pollen influx values and AP : NAP data indicate that dominant vegetation at the study site changed from birch woodland to pine forest at approximately 7000 14C bp and reverted back to birch woodland beginning approximately 2700 14C bp. AP : NAP ratio estimates of maximum pine forest northward advancement (between 7000 and 5800 14C bp, Fig. 6) roughly correspond in timing with maximum pine pollen influx values (Fig. 4b) and with maximum subfossil pine wood frequencies (MacDonald et al. 2000a). Between 7000 and 6000 14C bp, pine forest appears to have moved as far north as the Murman Coast, 100 km north of the present treeline position. Pine stomate and pollen data indicate that pine were locally present along the Murman Coast from c. 7500 to c. 6500 14C bp (Snyder et al. 2000).
The pine treeline in northern Fennoscandia and on the Kola Peninsula coincides with the mean 10 °C July isotherm. Presently the July isotherm along the northern Barents Sea coast of the Kola Peninsula is 8 °C (Murmansk Atlas 1971). Northward movement of the pine treeline to regions near the Barents coast thus represents an approximate 2 °C warming during the middle Holocene thermal maximum, similar in magnitude to other estimates in the region (Kullman 1995b; Barnekow 1999). The southward retreat of the pine treeline between c. 6200–2500 14C bp appears to parallel the gradual reduction of summer solar insolation at this latitude, so that from 3000 14C bp the treeline has been several kilometres south of our study site.
We thank Gennady Matishov and colleagues at the Murmansk Marine Biological Institute for their logistical help. Financial support for this project was provided by the NSF Palaeoclimatology of Arctic Lakes and Estuaries (PALE) grants (ATM-9600126/9632926). This is PARCS contribution number 188.