Rapid responses of high-mountain vegetation to early Holocene environmental changes in the Swiss Alps

Authors


Willy Tinner (fax + 41 31 631 49 42; e-mail willy.tinner@ips.unibe.ch).

Summary

  • 1The Early Holocene sediment of a lake at tree line (Gouillé Rion, 2343 m a.s.l.) in the Swiss Central Alps was sampled for plant macrofossils. Thin (0.5 cm) slices, representing time intervals of c. 50 years each from 11 800 to 7800 cal. year bp, were analysed and the data compared with independent palaeoclimatic proxies to study vegetational responses to environmental change.
  • 2Alpine plant communities (e.g. with Salix herbacea) were established at 11 600–11 500 cal. year bp, when oxygen-isotope records showed that temperatures increased by c. 3–4 °C within decades. Larix decidua trees reached the site at c. 11 350 cal. year bp, probably in response to further warming by 1–2 °C. Forests dominated by L. decidua persisted until 9600 cal. year bp, when Pinus cembra became more important.
  • 3The dominance of Larix decidua for two millennia is explained by dry summer conditions, and possibly low winter temperatures, which favoured it over the late-successional Pinus cembra. Environmental conditions were a result of variations in the earth's orbit, leading to a maximum of summer and a minimum of winter solar radiation. Other heliophilous and drought-adapted species, such as Dryas octopetala and Juniperus nana, could persist in the open L. decidua forests, but were out-competed when the shade-tolerant P. cembra expanded.
  • 4The relative importance of Larix decidua decreased during periods of diminished solar radiation at 11 100, 10 100 and 9400 cal. year bp. Stable concentrations of L. decidua indicate that these percentage oscillations were caused by temporary increases of Pinus cembra, Dryas octopetala and Juniperus nana that can be explained by increases in moisture and/or decreases in summer temperature.
  • 5The final collapse of Larix decidua at 8400 cal. year bp was possibly related to abrupt climatic cooling as a consequence of a large meltwater input to the North Atlantic. Similarly, the temporary exclusion of Pinus cembra from tree line at 10 600–10 200 cal. year bp may be related to slowing down of thermohaline circulation at 10 700–10 300 cal. year bp.
  • 6Our results show that tree line vegetation was in dynamic equilibrium with climate, even during periods of extraordinarily rapid climatic change. They also imply that forecasted global warming may trigger rapid upslope movements of the tree line of up to 800 m within a few decades or centuries at most, probably inducing large-scale displacements of plant species as well as irrecoverable biodiversity losses.

Introduction

Climatic conditions above the alpine tree line are so harsh that they impede the growth of trees, allowing cold-adapted plants to form species-rich communities in meadows and snowbeds, and on stony ground. Many alpine species have evolved locally since the late Tertiary and can thus be considered as the Alps’ primary diversity (Theurillat et al. 1998). Because of the fragmented and restricted habitats at the top of high mountains, alpine vegetation is one of the types most endangered by the predicted global warming of 1.4–5.8 °C (IPCC 2001). Extinctions may occur where mountains are not sufficiently high to maintain snow cover under warmer conditions (Theurillat et al. 1998), and landslides and rockfalls triggered by thawing of permafrost (Williams 1995; Cohen 1997) may also cause instability. One possibility for assessing ecosystem responses to climatic change outside the range of recent observation is to consider palaeoclimatic and palaeoecological records. For instance, at the Late Glacial – Holocene transition 11 500 years ago, it has been estimated that temperature in the Alps rose by 3–4 °C in only 48 years (von Grafenstein et al. 2000; Schwander et al. 2000), and it is likely that Holocene temperatures oscillated by about 1–2 °C within decades to centuries (von Grafenstein et al. 1998; Heiri et al. 2003). While many recent palaeoecological studies in the Alps have focused on climate reconstruction (e.g. Bortenschlager 1972; Zoller 1977; Burga 1991; Wick & Tinner 1997; Haas et al. 1998), the requirement for independent proxies of environmental change and high temporal resolution (Tinner & Ammann 2001) had lead to relatively few studies addressing past ecological relationships between organisms and their environments. Most past studies have a resolution of 100–300 years, which is sufficient to indicate centennial-scale Holocene climatic trends but not biosphere responses to rapid environmental change. We have therefore retrieved new lacustrine sediment cores from Gouillé Rion, a small lake at the present tree line, and increased the temporal resolution of a previous palaeoecological study (Tinner et al. 1996) from 285 to 51 years for the period 11 800–7800 cal. year bp. To reconstruct climate-sun-vegetation linkages during periods of rapid ecosystem reorganization, we compare independent palaeoclimatic data and proxies of solar radiation with our palaeovegetational records. We consider the nature and speed of vegetational dynamics in response to environmental changes such as variations in temperature, solar radiation and moisture availability, as well as the chronology and structure of previously detected tree line oscillations.

Study area and methods

The climate of the study area is of the central Alpine dry type, with annual precipitation of about 500–700 mm at low altitudes and 900–1100 mm at 2300 m a.s.l., where annual mean temperature and July mean temperature are about 1 °C and 9 °C, respectively. The study site Gouillé Rion (46°09′25″ N, 07°21′45″ E) is a small lake (0.5 ha) situated in the Swiss Central Alps at 2343 m a.s.l. The upper limit of closed forest (timberline) currently occurs below the site (at about 2100 m) and is dominated by Larix decidua, although single well-developed Pinus cembra trees > 5 m grow at 2320 m, and small individuals of ≤ 2 m height reach about 2350–2400 m. Thus the lake is situated at about the modern tree line (for further details on the geographical position and environmental conditions of Gouillé Rion see Tinner et al. 1996; Wick & Tinner 1997; Tinner & Theurillat 2003).

Parallel sediment cores (GR 6 and GR 7) were taken from the central part of the lake at 7.60 m water depth, using a Livingstone piston corer with a tube diameter of 8 cm. The cores were divided into 1-m sections, corresponding to depths of 0–100 cm, 100–200 cm and 200–300 cm for GR 6 and 38–138 cm and 138–238 cm for GR 7. After correlation with lithostratigraphical markers, the cores were subsampled in 0.5-cm slices (volume = 22 cm3). Contiguous macrofossil samples were taken at 224–189 cm in GR 7 vs. 194–168.5 cm in GR 6, and at 177–132.5 cm in GR 7 vs. 138–116 cm in GR 6. Overlapping subsampling was made to check the biostratigraphical variation in the two different cores. Subsamples were wet sieved through 0.2-mm mesh. Plant macrofossils were identified with reference material from the Institute of Plant Sciences of the University of Bern and keys by Schoch et al. (1988) and Tobolski (1992). The macrofossil diagrams show numbers (e.g. needles, seed, fruits, leaves) and areas (e.g. bark, wood, charcoal) per 22 cm3, or percentages. Aquatic plants, Cyperaceae and area measurements were excluded from the macrofossil sum, and only taxa that could be unambiguously attributed to trees (indicating forests), shrubs (shrublands) or creeping shrubs and upland herbs (alpine meadows, debris vegetation, snowbeds) were considered for percentage calculation (Tinner & Theurillat 2003). Adjacent samples were amalgamated where macrofossil concentrations were insufficient for percentage calculation (basal depths 210–204 cm). Throughout profiles GR 6/7, groups of six neighbouring samples were amalgamated to allow comparison with previously published low-resolution results from the same site (core GR 2, 10 m distant; Tinner et al. 1996; Wick & Tinner 1997). The diagrams were subdivided into local macrofossil assemblage zones (LMAZ) by the method of optimal partitioning (Birks & Gordon 1985) as implemented in the program ZONE, version 1.2, written by Steve Juggins. To determine the number of statistically significant zones, we used the program BSTICK (Bennett 1996). Only macrofossil types contributing to the macrofossil sum defined above were included in the zonation and in detrended correspondence analyses (DCA; Kovach 1995). Macrofossil analysis was undertaken by P. Kaltenrieder.

Terrestrial plant macrofossils from both GR cores were dated by accelerator mass spectrometry (AMS)-techniques at the Sub Atomic Physics Department, Utrecht University.

Results

chronology

The 14C ages are presented in Table 1. In comparison with neighbouring dates and the radiocarbon-dated biostratigraphy of Gouillé Rion GR 2 (Tinner et al. 1996), the dates UtC 7713 and UtC 7699 appear too young and are rejected. The age-depth curve for Gouillé Rion was formed by linear interpolation of the calibrated 14C dates (Fig. 1). The sediments below 209 cm depth were dated by lithostratigraphical comparison with the radiocarbon-dated record GR 2 (Tinner et al. 1996). Half centimetre sample resolution translates to a mean temporal sample resolution of 51 ± 24 years for the period 11 700–7700 cal. year bp (210.5–172 cm).

Table 1.  Radiocarbon dating of the sediments of Gouillé Rion (GR 6/7)
CoreLaboratory numberMaterialDepth (cm)Conventional radiocarbon date bpCal. year bp 95% limits*Cal. year bp in diagrams
  • *CALIB 4.3, INTCAL 98 (Stuiver et al. 1998).

  • Outside diagram limits of this study.

  • Rejected dates.

  • L = leaves, N = needles, NU = nut, P = periderm, SH = short shoots, W = wood.

GR7UtC 7705Conif. P   163–163.5  6030 ± 50  6729–7002 6 830
GR7UtC 7702Indet. W   173–173.5  7100 ± 60  7762–8105 7 901
GR6UtC 7716P. cembra W173.5–174  6900 ± 70  7591–7924 7 947
GR6UtC 7715P. cembra NU178.5–179  7630 ± 60  8346–8540 8 407
GR6UtC 7714Decid. P183.5–184  8080 ± 60  8777–9247 9 011
GR6UtC 7713P. cembra SH, N, conif. W188.5–189  7400 ± 50  8039–8343 9 275
GR7UtC 7701P. cembra W, P192.5–193  8450 ± 70  9284–9546 9 486
GR7UtC 7700Larix MB, W199.5–200  9080 ± 7010 154–10 40110 222
GR7UtC 7699Dryas L, Larix N203.5–204  8870 ± 90  9565–10 21710 945
GR7UtC 7698Larix N, S. herbacea L, indet W   206–20910 100 ± 8011 259–12 29311 669
Figure 1.

Radiocarbon dating of the sediments from the cores GR 6/7 of Gouillé Rion. Vertical bars give the sampling depth of terrestrial macrofossils. Horizontal bars show the 95% confidence interval of calibrated ages. Rejected dates indicated by X (see also Table 1) are not included in the linear depth/age model.

taphonomic processes: indications from a spatio-temporal comparison

Because macroscopic plant remains are not usually transported very far from their point of origin, macrofossil assemblages are closely related to the local vegetation around a site (Birks 2001). The interpretation of sedimentary plant macrofossil records may therefore be particularly affected by taphonomic processes such as decomposition, post-mortem transport, reworking, and final sedimentation in lake or mire deposits (Birks 2001). In addition, macrofossils are rather rare in sediments and this often impedes quantitative analyses (Lang 1994). The spatio-temporal reproducibility of our results is illustrated in Fig. 2 and the good correspondence between the two independently dated macrofossil records (GR 2 and GR 6/7) reveals that processes such as transport, distribution and preservation acted in a rather uniform way across the lake basin. Even minor fluctuations in macrofossil numbers, such as the peak of Betula fruits and the contemporary decline of Pinus cembra needles at c. 9300–9200 cal. bp, are depicted in both Gouillé Rion stratigraphies, indicating that low counts and (local) taphonomic processes did not significantly affect their reproducibility. The course of the curves is much more likely to have been controlled by vegetational change, an interpretation corroborated by similar Holocene macrofossil stratigraphies at two neighbouring sites in the central Swiss Alps (Gobet 2004).

Figure 2.

Comparison between macrofossil concentrations of different cores at Gouillé Rion. On the left the core GR 2 (Tinner et al. 1996; Wick & Tinner 1997) and on the right the cores GR 6/7. For this comparison, six neighbouring samples of GR 6/7 were amalgamated throughout the profile, the resulting sum (finds per 132 cm3) was divided by six (GR 2 recorded finds per 25 cm3). The two records have independent chronologies: the depth-age model for GR 2 is described in Tinner et al. (1996) and Wick & Tinner (1997).

macrofossil concentrations and percentages

The analyses of macrofossil concentration and of macrofossil percentages gave comparable results (Figs 3 and 4), showing that macrofossil percentages may provide useful information about past vegetation, provided the range of values does not differ greatly among the taxa (see Discussions in Watts & Winter 1966 and Birks 2001). In our case, the macrofossil sum is dominated by whole needles of coniferous species (e.g. Larix decidua, Pinus cembra and Juniperus nana) and/or leaves from Alpine shrubs and dwarf shrubs (e.g. Salix herbacea and Dryas octopetala, Figs 3 and 4). The high macrofossil concentrations, the good preservation and the similar provenance of macrofossils included in the macrofossil sum may explain the good performance of percentage analysis. The use of percentages has the advantage that it does not presuppose uniform sedimentation (Watts & Winter 1966; Faegri & Iversen 1989) and, in spite of distortion effects (in particular, internally created changes such as one taxon influencing unvarying taxa), is usually preferred in palaeoecology (e.g. pollen, diatom, chironomid and cladocera analysis). Sums as low as 45–50 are sufficient to produce reliable percentage results (e.g. Hofmann 1986; Heiri & Lotter 2001).

Figure 3.

Plant-macrofossil concentrations at Gouillé Rion (finds per 22 cm3). LMAZ = local macrofossil-assemblage zones, AN = anthers, B = buds, BS = bud scales, CS = cone scales, F = fruits, FS = fruit scales, L = leaves, MB = male blossoms, N = needles, P = periderm, NU = nuts, S = seeds, SH = short shoots, SW = seed wings, W = wood. Macrofossils belonging to the groups ‘Trees’, ‘Shrubs’ and ‘Creeping Shrubs and Herbs’ were included in the macrofossil sum with the exception of periderm (P) and wood (W).

Figure 4.

Plant-macrofossil concentrations and percentages at Gouillé Rion compared with DCA axes 1 and 2. Selected types. N = needles, FS = Fruit scales, F = fruils, L = leaves, B = buds, co = macrofossil concentrations, % = macrofossil percentages.

MACROFOSSIL-INFERRED VEGETATION HISTORY

At the Late-Glacial/Holocene transition the sediment deposited in the lake basin at Gouillé Rion changed from silt to silty gyttja. During the Younger Dryas (12 600–11 550 cal. year bp according to the GRIP chronology; Schwander et al. 2000) the previously established alpine meadows (c. 14 400–13 000 cal. year bp) were broken up and almost destroyed (see Tinner et al. 1996). Our new macrofossil data suggest that the subsequent recovery of alpine vegetation started at about 11 600 cal. year bp (Salix herbacea leaves and Salix buds recorded at 208.5–207.5 cm, LMAZ R67-A, Fig. 3). Larix macrofossil concentrations started to increase about 250 years later, indicating the expansion of forest vegetation at the modern day tree line (beginning of LMAZ R67-B, 206 cm, 11 350 cal. year bp). The percentage diagram suggests that until 10 700 cal. year bp (end of LMAZ R67-B, 202.5 cm, Fig. 4), Gouillé Rion was situated in the tree line ecotone, probably characterized by a mosaic of open Larix decidua stands, Juniperus nana shrublands and alpine meadow, snowbed and debris communities (Caryophyllaceae, Rumex t., Salix and Dryas octopetala).

The first appearance of Pinus cembra macrofossils (beginning of LMAZ R67-C, Figs 3 and 4) coincided with a partial closing of the forest at 10 700 cal. year bp (Σ tree macrofossil > 80%, Fig. 4). This process, which is also documented in the Gouillé Rion pollen record (Σ tree pollen and P. cembra pollen, see Wick & Tinner 1997 and Heiri et al. 2004), lasted only 100 years before it was interrupted by the expansion of Juniperus nana and a temporary disappearance of Pinus cembra until 10 150 cal. year bp (Figs 3 and 4). Juniperus nana macrofossil concentrations and percentages reached maximum values in the subsequent period (10 200–10 000 cal. year bp), indicating a strong expansion of shrublands around the lake. Neither Juniperus nana nor Dryas octopetala tolerate shading by taller plants (Ellenberg et al. 1992), so it is unlikely that the shrub communities were growing in a closed forest. Towards the end of LMAZ R67-C (194–192 cm, 9600–9450 cal. bp), mixed Larix decidua and Pinus cembra stands became increasingly dense, as documented by the sudden increased concentrations of (coniferous) forest indicators (e.g. conifer scales, conifer periderm, indeterminate wood and periderm, Fig. 3) and the decrease of silt deposition in the lake basin.

During LMAZ R67-D (192–179 cm, 9600–8450 cal. year bp) Pinus cembra was more important than before, although its macrofossil concentrations and percentages declined temporarily between 9200 and 8900 cal. year bp (Figs 3 and 4). The decline was preceded by an expansion of tree birches (most probably Betula pendula). Fruits of Betula trees were also found regularly before 9400 cal. year bp (Figs 3 and 4), although, owing to the wide dispersal of the winged seeds, fruit scales are more indicative of the local presence of this taxon (Tinner & Theurillat 2003). Betula trees were rather common around Gouillé Rion between 9400 and 8800 cal. year bp (regular findings of fruit scales, Figs 3 and 4), probably indicating transient openings of the Pinus cembra-Larix forests. In fact, in contrast to Pinus cembra, which is long-lived and the most shade-tolerant species among the tree and shrub taxa recorded at Gouillé Rion, Betula trees are rather short-lived and light-demanding and thus often restricted to early successional stages (Ellenberg et al. 1992; Ellenberg 1996). Forest openings indicated by the expansion of light-demanding Betula mainly occurred at the cost of Pinus cembra. The macrofossil record suggests that, after the recovery of Pinus cembra and the decline of Betula trees at c. 8800 cal. bp, Larix declined gradually. This development continued until the end of LMAZ R67-D, when Alnus viridis occurred for the first and only time in the macrofossil record and Betula trees re-expanded.

During the subsequent zone, LMAZ R67-E (179–177.5 cm, 8450–8300 cal. year bp), Larix decidua macrofossils disappeared completely (Figs 3 and 4), suggesting an abrupt collapse of the tree species. After this fall, which is also clearly documented in the Gouillé Rion pollen record (Wick & Tinner 1997), Larix never regained its early Holocene importance, as suggested by concentration and percentage values (zone R67-F, 177.5–172 cm, 8300–7800 cal. year bp) and the subsequent Holocene vegetation history (Tinner et al. 1996; Kaltenrieder 1999).

Discussion

vegetational dynamics in response to temperature changes at the late glacial/holocene transition

Detailed comparisons of Late Glacial and Holocene oxygen-isotope records from lakes in central Europe with the Greenland GRIP ice-core record show many close similarities, suggesting that past climatic changes caused almost simultaneous responses in the signature in both regions (von Grafenstein et al. 1998, 1999, 2000; Schwander et al. 2000). The correspondence is so pronounced that prominent Holocene events such as the Preboreal oscillation and the 8200 cal. year bp event could easily be matched. It is assumed that changes in δ18O mainly reflect variations in past mean annual air temperature, leading to a relationship of a 3 °C change being reflected in a 1 change in δ18O for Greenland and a 1.7 °C/ δ18O relationship for central Europe (von Grafenstein et al. 1998, 2000). We compare the chronology of the more accurate GRIP record (Dansgaard et al. 1993) with our macrofossil data (Fig. 5) in order to correlate vegetational responses with early Holocene climatic changes. We assume a close relationship between annual and summer climate, as is currently observed in the study region.

Figure 5.

Comparison of selected macrofossil types (percentages) and macrofossil DCA axis 1 and 2 with independent proxies of environmental change. The curve of DCA axis 1 was inverted for better comparison with tree macrofossil percentages. The GRIP oxygen-isotope record (Dansgaard et al. 1993) is highly correlated with alpine and central European oxygen-isotope series (von Grafenstein et al. 1998; 1999; Schwander et al. 2000) and therefore is used as a proxy for air-temperature change. Original values (dots) were smoothed with LOWESS for better comparison with the macrofossil record. Reconstructed June insolation for 60° N (Berger & Loutre 1991) shows an early Holocene maximum resulting form orbital variations (Milankovitch cycles). The residual Δ14C record (Stuiver & Braziunas 1991) is used as a proxy for decadal to centennial-scaled variations of solar radiation (Muscheler et al. 2000). CE = central European cold-humid phases according to Haas et al. (1998).

According to the GRIP oxygen-isotope record, the Younger Dryas ended at 11 535 cal. year bp (Schwander et al. 2000), and oxygen-isotope values then rose to early Holocene (Preboreal) levels within 48 years. The rise in temperature around Gouillé Rion during this transitional period can be estimated to be c. 3–4 °C (c. 2 δ18O in and around the Alps, see Schwander et al. 2000; von Grafenstein et al. 2000). Temperatures had increased by a further 1–2 °C (0.5–1 δ18O) by 11 350 cal. year bp, when a first early Holocene maximum was reached, followed by climatic cooling of the Preboreal Oscillation (c. −1 to −2 °C, −0.5 to −1 δ18O). According to the GRIP chronology, the Preboreal Oscillation ceased at c. 11 100 cal. year bp (Schwander et al. 2000; von Grafenstein et al. 2000) and the temperature in Greenland then increased steadily by about 4.5 °C (1.5 δ18O) between then and 9800 cal. year bp. Early Holocene climatic change was probably a lot more moderate in Europe, as evidenced by the Ammersee record (von Grafenstein et al. 1999), which indicates a gradual temperature increase of about 1–1.5 °C (0.5–0.8 δ18O) between 11 100 and 9800–9600 cal. year bp. After 9600 cal. year bp, oxygen-isotope values do not show further pronounced increases in temperature in either European or Greenland records (Dansgaard et al. 1993; Grootes et al. 1993; von Grafenstein et al. 1999; Willemse & Törnqvist 1999).

The early Holocene patterns found in the oxygen isotope records from Europe and Greenland are mirrored in the macrofossil-inferred vegetation history of Gouillé Rion, provided allowance is made for some dating uncertainties (Figs 3–5). Alpine vegetation (Salix herbacea) was established at the beginning of the Holocene at 11 600 cal. bp, when temperature increased abruptly. Forest vegetation (Larix decidua) expanded at 11 350 cal. year bp during the first inferred maximum of temperature, before gradually replacing alpine and shrubland communities. The concentration and percentage diagrams (Figs 3–5) show that this process was not completed until 9800–9600 cal. year bp (final displacement of Dryas octopetala and Juniperus nana, see also tree percentage curve), in agreement with the continuing temperature increase indicated by δ18O until 9800–9600 cal. year bp (Fig. 5). The gradual adjustment of mountain vegetation to climatic change is summarized by the tree percentage curve and axis 1 of the de-trended correspondence analysis (DCA). As axis 1 represents the greatest amount of variation in the macrofossil data set (Kovach 1995), we assume that gradual (summer) temperature changes during the early Holocene were the main determinant for vegetation dynamics at Gouillé Rion. The long-term nature (c. 1600–1800 years) of the vegetational changes observed (Salix herbacea expansion followed by Dryas octopetala, Larix decidua, Juniperus nana, and finally Pinus cembra) excludes ordinary successional processes as an explanation. In fact, L. decidua and P. cembra are able to build dense forests on glacial moraines within 150 years after ice retreat (Ellenberg 1996).

The site is located at the present day tree line (single individuals of Pinus cembra < 5 m) and it is therefore likely that forest established only when summer temperatures reached similar levels to those currently observed. Oxygen-isotope (Schwander et al. 2000; von Grafenstein et al. 2000) and chironomid records (Heiri et al. 2003) suggest that temperature thresholds allowing tree growth at Gouillé Rion were reached at about 11 350–11 200 cal. year bp. Larix decidua expanded at about this time, suggesting that vegetation responded almost immediately to climatic change at the beginning of the Holocene, probably within the sample resolution of the study (51 years) and certainly within the time-scale of normal successional processes (100–150 years). The timberline ascended 800 m in 200–300 years, as it was located at about 800 m below Gouillé Rion during the Younger Dryas (Welten 1982; Tobolski & Ammann 2000). Moreover, the macrofossil record suggests that alpine communities were established in response to climatic warming at 11 550 cal. year bp, i.e. 200 years before trees expanded, and we thus conclude that vegetation was in equilibrium with climate, even during the pronounced climatic warming at the Late Glacial/Holocene transition.

a comparison with previously detected centennial-scaled vegetation oscillations after the preboreal oscillation

Previous palaeobotanical studies at Gouillé Rion have highlighted the sensitivity of the site to Holocene century-scaled climatic oscillations (Tinner 1994; Tinner et al. 1996; Wick & Tinner 1997; Haas et al. 1998; Tinner & Ammann 2001; Heiri et al. 2004). Most fluctuations of tree line vegetation at Gouillé Rion were identified from oscillations in the pollen record and their climatically driven nature, indicated by comparison with palaeocological studies in other areas of the Alps. Our new high-resolution records confirm both the previously established macrofossil stratigraphy of the site (Tinner et al. 1996; Kaltenrieder 1999) and the existence of the cold phases indicated by previous pollen and macrofossil results. The cold phase Rion-1 (Fig. 2) was primarily defined from the pollen profile of core GR 2, where the decline of Pinus cembra and increase of Larix decidua lasted from 10 950 to 10 300 cal. year bp, with the zone delimited by the expansion of Juniperus nana and the subsequent expansion of Pinus cembra (Wick & Tinner 1997). In the new record (GR 6/7) this biostratigraphical change occurred between 11 000 cal. year bp (first J. nana macrofossils) and 10 150 (expansion of P. cembra) (Figs 3 and 4). Rion-1 has the same age as CE-1 of Haas et al. (1998) (see Fig. 5).

According to Wick & Tinner (1997) the following cold phase, Rion-2, lasted from 9500 to 9150 cal. year bp and was characterized by a marked increase of Betula and a decrease of Pinus cembra macrofossil concentrations (Fig. 2), with P. cembra pollen also declining. In the new cores, this phase is confined to 9300–8900 cal. year bp (Figs 3 and 4). Rion-3, the last cold zone of the early Holocene, was preceded by a strong Larix decidua decline, with its start at about 8350 cal. year bp marked by the recovery of L. decidua and the onset of a decrease in P. cembra needles and its end in the mid Holocene at 7550 cal. year bp (see Wick & Tinner 1997). In our new diagrams (Figs 3 and 4) the biostratigraphical changes used to define the beginning of Rion-3 are dated at 8350–8300 cal. year bp. Rion-2 and Rion-3 are coeval with CE-2 and CE-3 of Haas et al. (1998), and the later period may correspond to the 8200 cal. year bp event recorded in the Greenland ice cores (Wick & Tinner 1997; Haas et al. 1998). Centennial-scaled cold periods synchronous with Rion-1, Rion-2 and Rion-3 were also detected in Alpine chironomid records (Heiri 2001; Heiri et al. 2003; Heiri et al. 2004). Smoothed chironomid-based records suggest that inferred mean July air temperatures decreased by c. 1 °C during these cooling episodes.

In general, previous data on vegetational changes during the cold periods suggest centennial-scaled shifts towards colder-adapted alpine vegetation (e.g. expansion of Juniperus nana at c. 11 000–10 150 cal. year bp) and opening of timberline forests (e.g. expansion of Betula trees at c. 9500–9150 cal. year bp). This overall interpretation is still valid, but the markedly higher resolution of the new Gouillé Rion record offers additional possibilities of comparison with independent, decadal-scaled proxies of past environmental conditions. DCA axis 2 (Fig. 4) shows a pattern of vegetational change that can be explained by the cumulative appearance of heliophilous woody taxa such as Dryas octopetala, Juniperus nana, Larix decidua, Salix herbacea and Betula pendula (ranked according to decreasing shade intolerance, Ellenberg et al. 1992; Fig. 4). The intensity of solar radiation may therefore have influenced the composition of vegetation directly, as well as indirectly via its effects on the Holocene climate. To detect potential linkages between solar irradiance, climate and vegetation we therefore compare our records with reconstructed changes in solar radiation. However, close connections have been also proposed between Holocene meltwater pulses to the North Atlantic, Alpine climatic oscillations and tree line oscillations (Heiri et al. 2004), and this alternative climatic forcing factor is also considered.

the ecological role of changes in solar radiation and moisture availability

Variations in the earth's orbit (Milankovitch cycles) affect the distribution of solar energy over its surface, resulting in millennial-scaled changes in seasonality and climate. During the early Holocene increased summer solar radiation in the northern hemisphere (Fig. 5) and reduced winter solar radiation (Berger & Loutre 1991) resulted in a greater seasonal contrast in temperature than today, with a prevalence of hot summers and cold winters (Kutzbach & Webb 1993). Higher-frequency fluctuations of solar irradiance are superimposed on this long-term trend, as suggested by pronounced correspondences between the fluctuations of cosmogenic nuclides 14C and 10Be during the early Holocene (Muscheler et al. 2000). The three biggest Δ14C excursions during the entire Holocene occurred around 11 100, 10 100, and 9400 cal. year bp, and the corresponding production-rate maxima resemble those associated with solar-activity minima similar to the Maunder minimum (Muscheler et al. 2000; Fig. 5). According to these authors, the 10Be and Δ14C concentration increases are possibly related to cold periods at 11 250 and 10 300 cal. year bp, which are synchronous with the Preboreal Oscillation (11 360–11 100 cal. year bp; Schwander et al. 2000) and the end of CE-1 (10 950–10 300 cal. year bp, Haas et al. 1998; Fig. 5). Moreover, the oscillation at 9400 cal. year bp falls into the cold phase CE-2 (9550–9000 cal. year bp), possibly indicating that the radiation decrease caused substantial cooling over the Alps and central Europe. Marked correlations between 14C, 10Be and proxies of drift-ice movements in the North Atlantic (Bond et al. 2001), as well as records of past lake productivity from Alaska (Hu et al. 2003), suggest that solar-radiation changes influenced climate at hemispherical or global scales.

Despite these relationships, the climatic cold phases CE-1 (10 950–10 300 cal. year bp) and CE-3 (8400–7700 cal. year bp) are better correlated with rerouting of meltwater from the remnant North American ice sheet to the Northern Atlantic at 10 700–10 300 cal. year bp and 8400 cal. year bp (Barber et al. 1999; Teller et al. 2002). Indeed, the latter event was followed by several centuries of increased freshwater flux through Hudson Street (Clark et al. 2001; Teller et al. 2002). Together, these correspondences may suggest that the most prominent Alpine Holocene cooling episodes (CE-1 and CE-3) resulted from slow-downs of the thermohaline circulation rather than from changes in solar radiation (Heiri et al. 2004). In this case, the temporary exclusion of Pinus cembra from tree line at 10 600–10 200 cal. year bp and the striking Larix collapse at 8400 cal. year bp (see also Tinner et al. 1996; Wick & Tinner 1997, for GR 2, Figs 4 and 5) can probably be attributed to climatic cooling in response to slowing of thermohaline circulation at 10 700–10 500 and 8400 cal. year bp.

Effects of variability in solar radiation on vegetation may not be limited to responses related to air temperature. For instance, variations in radiation can modify the predominant influence of temperature, thus leading to substantial deviations of the timberline from isotherms (Tranquillini 1979). In the Central Alps, low air temperatures are, in part, compensated by higher solar radiation, so that plant species can reach higher altitudes than in the Northern and Southern Alps, despite mean air temperatures being comparable (Landolt 1992; Ellenberg 1996). Solar radiation favours warming of shoots and hence maturation of plant needles and leaves, and, although such effects are most pronounced for creeping plants (e.g. Dryas octopetala and Salix herbacea), they are also important for timberline-forming tree species, as the growing season is significantly shorter at altitude than in lowland areas (Ellenberg 1996; Körner 1999). In combination with the altitudinal decrease of air humidity, high solar radiation can lead to soil-water deficits in summer, influencing alpine plants and vegetation (Landolt 1992). Solar radiation also promotes late-winter desiccation, which can cause substantial damage to, for example, young Pinus cembra trees near timberline (Tranquillini 1979; Körner 1999). Depending on the individual heat requirements for maturation, and hence development of cuticular protection during the growing season (Ellenberg 1996; Körner 1999), late-winter desiccation can codetermine vegetation, as drought-resistant species may ascend to greater altitudes (Tranquillini 1979).

These ecological observations show the relevance of solar radiation for alpine environments and may aid understanding of the correlation between the abundance of heliophilous woody taxa, as summarized by DCA axis 2, and the early Holocene maximum of summer radiation (Figs 4 and 5). The presence or dominance of extremely heliophilous taxa such as Dryas octopetala, Juniperus nana and Larix decidua is best explained by the absence of dense Pinus cembra forests until 9500 cal. year bp (Figs 4 and 5). In fact, P. cembra forests are rather dark, so that shade-intolerant species are excluded within decades (shrubs) or a few centuries at most (long-lived Larix trees, Ellenberg 1996) after canopy closure. As P. cembra is not more limited by summer temperature than L. decidua, its late expansion at Gouillé Rion (c. 2000 years after establishment of first L. decidua stands) requires an alternative explanation. AMS-radiocarbon dated pollen studies from Gouillé Rion suggest that P. cembra was present in the surrounding area from at least 14 000 cal. year bp (Tinner et al. 1996) and macrofossil findings of the species in the Valais region (Simplon, 2005 m a.s.l., c. 50 km distant) confirm its Late-Glacial and early Holocene presence at lower altitudes (Lang & Tobolski 1985), thus excluding migrational delays as a cause of late expansion. In other regions of the Central Alps, Pinus cembra was important earlier at lower altitudes (e.g. since 10 500 cal. bp in the Engadine at 1800 m a.s.l., Gobet 2004) or where continentality is less pronounced than in the Valais region (e.g. since 11 200 cal. bp in the San Giacomo Valley at 2250 m a.s.l.; Wick & Tinner 1997). However, tree-ring studies document the dominance of Larix decidua at tree line forests in Tyrol during the early Holocene until c. 9000 cal. bp (K. Nicolussi, personal communication), suggesting that the pattern we observed was not confined to the Gouillé Rion area. Although less extreme in terms of hygric continentality, the climate of Tyrol is comparable with that at Gouillé Rion, i.e. it is of the Central Alpine dry type (Ozenda 1988).

The late appearance of Pinus cembra might be explained by continentality effects related to relatively high summer solar radiation, including low winter temperatures and drought. Pinus cembra is less (summer-) drought-adapted than Larix decidua, Dryas octopetala and Juniperus nana (Ellenberg et al. 1992) and while the shrubs are protected by snow packs during winter and L. decidua loses its needles, minimizing winter desiccation damage, P. cembra individuals are fully exposed to late-winter drought stress.

Siberian sites with an ultracontinental climate may be used as a summer drought and winter low-temperature analogue for the early Holocene situation in the most continental valleys of the Alps. Under such extreme conditions, Pinus sibirica, a very close relative of P. cembra (Gugerli et al. 2001), is excluded from the polar timberline (Walter 1974; Anenkhonov & Chytry 1998; Ermakov et al. 2002) and the only tree species present is Larix gmelinii, a species closely related to L. decidua (Semerikov et al. 2003; Wei & Wang 2003). In addition to marked continental conditions (dry summers, very cold winters), permafrost may also help to exclude Pinus sibirica from polar Larix timberline in western Siberia, even though continentality is less pronounced (Mudrik & Vilchek 2001). In general, it seems that dry climatic conditions strongly favour Larix sibirica near tree line (Walter 1974). For instance, codominance of Pinus sibirica and Larix sibirica has been recorded on more humid west aspects east around Lake Baikal, whereas Pinus sibirica occurs only sporadically in tree line forests on the dry, eastern aspect.

Drought and low winter temperatures as a consequence of higher summer and lower winter insolation than today may therefore explain the dominance of Larix decidua and the insignificance of Pinus cembra near Gouillé Rion during the early Holocene. The site has a south-east aspect with high solar radiation exposure, and is located in one of the driest and most continental regions of the Alps (Ozenda 1988; Landolt 1992). Decreases of solar radiation, as indicated by oscillations of 14C and 10Be, provide the opportunity to check whether decadal-scaled changes coincided with tree line vegetation on shorter time-scales. If the general early Holocene dominance of Larix decidua was caused by solar-triggered dry conditions, decadal-scaled decreases of summer solar radiation at 11 100, 10 100, and 9400 cal. year bp (Muscheler et al. 2000) should have resulted in temporary declines. Short-term decreases of Larix decidua percentages did occur at around 11 000, 10 100 and 9500 cal. year bp (Fig. 5), suggesting a link with increased moisture availability. Pinus cembra, which would be favoured by moister conditions, is present or increased at around 10 700 (isolated occurrence), 10 100, and 9500 cal. year bp. The concentrations of Larix decidua at 11 000, 10 100 and 9500 cal. year bp are stable, indicating that the percentage oscillations were primarily caused by temporary increases of cold-adapted species such as Dryas octopetala (11 000 cal. year bp) and Juniperus nana (10 100 cal. year bp), and of Pinus cembra (9500 cal. year bp). Temporary climatic coolings, in response to diminished solar activity, may have reinforced the ecological effects of increases in moisture availability. Even taking into account the dating uncertainties, the expansion of Juniperus nana at c. 10 100 cal. bp seems too late to be related to the re-routing of meltwater from the remnant North American ice sheet at 10 700–10 300 cal. bp (Barber et al. 1999; Teller et al. 2002) and the coeval cooling in the Alps (CE-1, Haas et al. 1998; Heiri et al. 2004). Nevertheless it is conceivable that tree line vegetation in the Alps was influenced by synergetic effects of changes of solar radiation and the thermohaline circulation at 10 700–10 000 cal. bp.

Conclusions

Comparison with the general course of oxygen-isotope records from the Alps suggests that, within the limits of our time resolution (50 year) and precision, tree line vegetation was in dynamic equilibrium with climate, even during periods of extraordinarily rapid climatic change. In general, timberline position was very sensitive to temperature changes (e.g. at 11 500, 10 400 and 8400–8200 cal. bp), whereas vegetation composition in the Gouillé Rion area was mainly a result of air and soil-moisture conditions in response to high summer radiation. Vegetation patterns at Gouillé Rion appear to be due to a combination of extreme environmental conditions and a range of factors causing high climatic variability over both long-term and decadal scales.

Our data confirm previous reports of the high sensitivity of vegetation to changing climatic conditions (e.g. Ammann et al. 2000; Tinner & Lotter 2001) and imply that the predicted global warming of 1.4–5.8 °C may trigger large-scale displacements of plant species as well as rapid upslope movements of tree line. Such rapid changes would be especially dangerous for alpine biodiversity, given the already restricted extent of high mountain habitats. As emphasized by Overpeck et al. (2003), palaeoecological studies may be able to contribute to our understanding of responses to rapid climatic change and thus help to avert an unprecedented ecological disaster in the Alps and elsewhere.

Acknowledgements

We are very grateful to B. Ammann for funding radiocarbon dating and part of the macrofossil analysis by P.K. We thank K. van der Borg, van de Graaff laboratory, Utrecht University for radiocarbon analysis. Improvements of the manuscript and helpful advices by H. H. Birks, L. Haddon, P. Moore, H. E. Wright Jr. and an anonymous reviewer are gratefully acknowledged.

Ancillary