Study Sites

We selected two linear transects 9 km apart through the altitudinal treeline ecotone in the Central Swiss Alps (Fig. 1). The treeline ecotone was defined as the level situated between the end of the closed forest and the last upright tree (height > 2 m and diameter at breast height (DBH) > 7 cm). Forests are dominated by Pinus cembra and Larix decidua Miller, with an undergrowth of Rhododendron ferrugineum L. and Vaccinium myrtillus L. The first transect lies above the village of Grächen (46°11′N, 7°51′E), between 2200 and 2580 m a.s.l., on a large rock field. Because the surface is very irregular, it is unlikely that any significant timber harvest or cattle grazing has taken place in recent centuries and therefore the position of the treeline ecotone is most likely natural. P. cembra is the only species at the top end of the transect. The second transect lies above Ginals (46°15′N, 7°47′E), between 2100 and 2350 m a.s.l. This site is mostly on deep soils and the site has been exploited for a long time, with timber harvesting and summer cattle grazing, as in most of the Alps (Bätzing, 1991). P. cembra and L. decidua both occur along the whole transect. Both sites are on siliceous bedrock, with a western to northwestern aspect and a similar slope (ca. 30°).

Figure 1

Location of the study sites and related meteorological stations in Switzerland.

Climate

Temperature and precipitation data from 1901 to 2000 were collected at the MeteoSwiss station at Grächen (1620 m a.s.l.). Missing monthly values (62 between 1900 and 1932) were estimated (Fritts, 1976) using linear regressions from the years where there were also data from Saas Fee (1785 m, 12 km from Grächen) for temperature and Zermatt (1610 m, 20 km from Grächen; Fig. 1) for precipitation.

In Grächen, the average annual precipitation (±1 SD) for 1901–2000 was 577 ± 116 mm, with the winters drier (mean of the sum for December to February 112 ± 57 mm) than the summers (June to August 160 ± 46 mm). The data were very variable, with the total annual precipitation ranging between 297 and 881 mm and total monthly values between 1 and 256 mm. The coefficient of variation ranged from 83 to 85% in winter months and 45 to 53% in summer months.

The mean annual temperature was 4.7 ± 0.8°C (−3.5 ± 2.1°C in January and 13.8 ± 1.6°C in July). During the 20th century, the mean annual temperature increased from 4.29 to 5.24°C (linear regression significant, p  =  0.0001). This increase took place mainly during two periods (1900–1921 and 1980–2000), separated by a more or less stable mean temperature between 1921 and 1980. Warming was especially marked at the end of the century, with an increase of 1.9°C in winter in the period from 1966 to 2000 (linear regression, p  =  0.0075) and of 2.4°C in summer (Fig. 2; p < 0.0001).

Figure 2

Tree-ring width (age-detrended) in Grächen and Ginals and average growth season temperature in Grächen (Central Swiss Alps, 1620 m a.s.l.). The tree-ring curves are smoothed means of tree-ring widths after removing age effects (Paulsen et al., 2000), and the growth season temperature is the average of monthly values from May to September.

Core Sampling and Measurements

At each site, three stands were selected along the transect at different altitudes: in the forest (just under the treeline ecotone), in the lower part of the treeline ecotone, and among the last trees in upper part of the treeline ecotone (Table 1). Only trees with 2 m in height and DBH > 7 cm were considered. In the forest and lower treeline stands, 20 Pinus cembra trees were randomly selected along horizontal lines (closest tree every 20 m). In the upper treeline stands, we chose all the trees growing at the end of the transect (only 14 were available in Ginals). Trees were measured for height and DBH. In September 2000, two cores were taken, one upslope and the second perpendicularly at breast height in the forest and lower treeline stands at 50 cm above the root collar for upper treeline stands. The characteristics of the stands and trees are summarized in Table 1.

Table 1

Elevation and tree characteristics (mean ± 1 SD) for the sampled stands (A) and statistics for the tree-ring width residual chronologies (B). The proportion of corrected rings is the percentage of missing or false rings in the cores; the mean sensitivity is the relative change in ring-width from year to year; the mean correlation between cores is a coefficient calculated from the pairwise comparison of all series in the chronology; the variance explained in first eigenvector is the percentage of common variance among ring-width series explained by the first principal component (Fritts, 1976); the signal-to-noise ratio and expressed population signal are both measures of chronology confidence (Wigley et al., 1984).

The cores were mounted and sanded, and tree-ring widths were measured with an accuracy of within 0.01 mm. We followed standard visual cross-dating procedures with the help of the software XMATCH98 (Fowler, 1998) and evaluated the results with COFECHA (Holmes, 1983) in ITRDBLIB (Grissino-Mayer et al., 1997). Cross-dating showed that missing or false rings were rare (Table 1). Trees were aged by counting rings, and missing rings to the pith were estimated when necessary using Duncan's (1989) method. Tree age is underestimated when coring above the stem base. According to Motta and Nola (2001), Pinus cembra needs about 20 years to grow to 50 cm in height at 2200 m a.s.l.

Analyses

Two detrending procedures were used. The first one (Paulsen et al., 2000) aimed to compare absolute tree-ring growth through time and between stands. The effect of age on growth was first removed, as the age structure of the stands differed. For this, the cores which included the pith or were close to it were selected to calculate the age effect curve following Paulsen et al. (2000). The 83 selected series were aligned according to cambial age, and a cubic polynomial function was fitted to the first 150 years, as tree rings older than 150 years did not show any age effect: f(age)  =  3 * 10⁻⁷ * age³ − 2 * 10⁻⁴ 0.0002 * age² + 0.0167 * age + 0.6194. The series were individually detrended with this function between 20 and 150 years and multiplied by the average width of tree rings older than 150 years (0.46 mm). The whole correction was: ringwidth′  =  0.46 * ringwidth * f(age)⁻¹. Tree rings between 0 and 20 years old were not used as they showed more irregular values. The other trees, whose cores were not reliable for age estimation, were not detrended, and only the flat part of the series was used. The annual mean ring growth of the stands was calculated by aligning the series (one core per tree) with the correct calendar year, and annual growth was averaged for every stand and smoothed with a 13-year reciprocal low-pass filter (Fritts, 1976) in order to remove high-frequency signals. Only years with four or more series were considered.

The second detrending procedure used a classical dendrochronological method (Fritts, 1976) to produce a standardized chronology for each stand and to compare tree-ring growth with climate. As many young trees did not cover the complete 20th century, we selected the longest and best cross-dated series for this period and standardized them by fitting a negative exponential curve (33.8% of series), a straight line with a negative slope (13.8%), or, if both of these failed, a horizontal line (52.5%) which equals the mean ring-width value and so does not remove any growth trends. We used the software ARSTAN (Cook, 1985) for the standardization, which simultaneously calculates descriptive statistics of the chronology (Table 1) and a residual chronology where autocorrelation is removed to enhance the climate signal. The residual chronologies were compared with the mean monthly temperature and total monthly precipitation from August of the previous year to September of the growth year. Bootstrapped response function analyses (Fritts, 1976; Briffa and Cook, 1990) were performed with PRECON software (Fritts, 1999) using the principal components, linearly independent of climatic data, with 1000 replications and without previous years, as autocorrelation had already been removed from the chronologies.

Age Structure

At each transect level, the age structure was similar in both transects, but with temporal differences (Fig. 3). In the forests, we found old trees with a regular regeneration over centuries; at the lower treeline ecotones, the majority of trees were young, with a few old trees, and at the upper treeline we found only young trees (established after 1800). However, the two transects differed with respect to the age of the most recent colonization. It began at the lower and upper treeline stands mainly around 1840 in Grächen, and about 60 years later in Ginals. Numerous saplings and seedlings at the treeline indicated that the colonization is ongoing, and the last visible age class in Figure 3 certainly did not correspond to the end of the colonization. As all the trees in upper treeline stands were cored, tree limit expanded upslope after 1800 in both transects.

Figure 3

Tree age structure of six stands in the Central Swiss Alps: number of new trees per 20-year classes. Grächen stands were undisturbed and Ginals stands were influenced by timber harvest and cattle grazing.

Growth Trend

The growth curves had similar patterns at both sites (Fig. 2). Both forest stands showed a growth depression around 1815–1822, followed by better growth, particularly visible in Ginals, where growth reached its highest values. Another growth depression is visible in the forest and lower treeline stands around 1915, mainly in Ginals. After a period with varying patterns, growth increased since the 1980s in all stands except in the forest of Ginals. Annual rings were found to be the widest at the end of the 1990s in these three stands. In both upper treeline stands there was a very visible decrease in growth in the late 1970s, with a return to a normal growth at the end of the measured period.

Descriptive Statistics of Chronologies

The best part of the residual chronology, with a sufficient number of cores, was the longest in the forest stands (only the 20th century was retained here). In the treeline ecotone stands, the high proportion of young trees allowed only rather short chronologies (Table 1). The variance of residual chronologies accounted for by the first component (common variance among trees included in the chronology; Fritts, 1976) ranged between 31 and 45%. The mean sensitivity was lower than 0.2 for all the stands, but increased in the upper treeline. The expressed population signal exceeded 0.85 in four stands and was closed in the forest stand of Grächen (Table 1). This value is considered as a threshold for reliable chronology (Wigley et al., 1984), but it was much too low in the upper treeline stand of Ginals.

The yearly correlations of the standardized chronologies between the stands were mostly highly significant (Table 2). Correlations between the Grächen stands were clearly better than those between the Ginals stands. In both transects, correlations between closer stands were better than between distant ones. Between transects, the highest correlations generally occurred between stands at the same levels.

Table 2

Annual correlations of residual chronologies in the Grächen and Ginals sites for their common periods. Pearson correlation coefficients: *** highly significant (p < 0.001); * significant (p < 0.05); n.s. not significant.

Climate Influence on Radial Growth

In the 20th century, temperature and growth were related in two different seasons. The first one was the positive correlation between growth and summer temperatures (Table 3). This was particularly visible in July, but several signals appear in May, June, and August in some stands. Autumn temperatures of the previous year were positively correlated with radial growth as well. Correlations were apparent mainly in November in forest stands, and to a lesser extent, in October or September in treeline stands (Table 3).

Table 3

Results of the response function analyses of tree radial growth in relation to climate data in six stands (Central Swiss Alps). Analyses were performed on the the full chronology span period (Table 1). Significant coefficients are indicated with symbols.

The main correlation between precipitation and growth was in the autumn of the previous year. High rainfall, especially in October, was positively related to growth in four stands (Table 3). The stronger correlation observed in lower stands extended toward the end of the year in some stands. Another positive correlation was recorded in August of the growing season in the two lower stands in Grächen.

Treeline Ecotone Densification and Expansion

Since 1840, the treeline ecotone increased in density and expanded upslope in Grächen. The strong recruitment in forest and lower treeline stands happened in the middle of already existing trees (Fig. 3). However, there was a real upslope shift of the tree limit, as all the trees in the upper treeline stand were younger than 1840 (except one at 2540 m, 40 m lower than the current limit). Simultaneous forest expansion has already been observed in California (Taylor, 1995), the North American mountains (Brink, 1959; LaMarche, 1973; Bekker, 2005), the American boreal forest (Payette and Filion, 1985), Sweden (Kullman, 1986), and in the Ural Mountains (Shiyatov, 2003). At the same time, herbaceous alpine plants also expanded upslope (Braun-Blanquet, 1957). This expansion is generally associated with higher temperatures after the end of the Little Ice Age around 1850 (Esper et al., 2002).

However, tree recruitment in treeline ecotones does not depend only on higher temperature, and other factors may inhibit forest expansion. For example, Daniels and Veblen (2004) observed in the Andes that low moisture availability could limit tree recruitment. In the Spanish Pyrenees, Camarero and Gutiérrez (2004) observed an upslope expansion of the tree limit between 1850 and 1900, but no change later. They attributed this stability to the high interannual variability of March temperatures, which prevents seedling survival. Treeline expansion is sometimes limited by the weak seed dispersal and the better microclimatic conditions among existing trees, which promotes densification of the stands (Mazepa, 2005).

Little is known about necessary conditions for seedling establishment and survival of Pinus cembra. The nutcracker (Nucifraga caryocatactes L.) disperses the seeds and hides them in humus as winter caches, frequently above the treeline (Contini and Lavarelo, 1982). This species thus has a high potential for responding quickly to climate warming (Brubaker, 1986). Ridges, slopes, and rock ledges are preferred by the nutcracker to depressions as hiding places because of the smaller accumulation of snow (Tranquillini, 1979; Contini and Lavarelo, 1982). This is favorable to seedling survival as well, as long snow cover increases the risk of fungal parasitism in spring by Gremmeniella abietina (Lagerb.) Morelet or Phacidium infestans L. (e.g., Contini and Lavarelo, 1982; Schönenberger et al., 1990; Senn, 1999). Seedlings and saplings outside of snow cover (e.g., on wind-exposed ridges) are particularly sensitive to winter desiccation, when long-lasting soil freezing conditions are coupled with intense radiative warming (Tranquillini, 1979; Oberhuber, 2004). From these observations, we can conclude that survival of P. cembra seedlings is likely favored by a warm spring for a quick melting of snow and early snow falls at the beginning of winter to limit deep freezing of the soil. Long snow cover in spring is probably the most limiting factor, and the temperature increase during the growth season seems to have been the necessary and sufficient change to trigger the upslope expansion of the treeline after the Little Ice Age. Relatively humid summers could be important as well, as is the case for Pinus uncinata in the dry Spanish Pyrenees, but experiments showed that P. cembra seedlings are very tolerant of droughts (Hättenschwiler and Körner, 1995).

Treeline Expansion in the Alps

In the Alps, treeline expansion has already been observed but, up to now, it was difficult to interpret because the end of the Little Ice Age was almost simultaneous with the decrease in mountain agriculture (Motta and Nola, 2001). However, the delay of establishment in Ginals stands, an anthropogenic treeline, about 60 years after the establishment in Grächen stands, a natural treeline, showed that both events were not perfectly synchronous in this region (Fig. 3). The continuous decrease in agriculture in Swiss mountain regions started in the second half of the 19th century (Bätzing, 1991), but was sensitive in tree recruitment only at the beginning of the 20th century (Fig. 3). These periods of expansion are consistent with previous results. For example, Payette and Filion (1985) and Taylor (1995) found an expansion since about 1850 in natural conditions, while Stützer (1999) and Motta and Nola (2001) observed it mainly since 1920 in pastured areas. Under such conditions, it is likely that the already warmer climate facilitated the recolonization of subalpine grasslands, as the young seedlings have to withstand harsher microclimate conditions in the open areas than in the middle of trees at similar altitudes in Grächen stands (Holtmeier and Broll, 2005). Nowadays, the altitude difference of the treeline ecotones in both transects shows that the Ginals treeline is not yet in balance with climate and will certainly carry on its expansion.

Tree Growth and Climate

Cold temperatures are certainly the main limiting factor for tree growth in these treeline ecotones. All three growth depressions observed in radial growth were linked to cold periods (Fig. 2). The first one, around 1815–1822, was already recorded by Büntgen et al. (2005). It followed the Tambora eruption in Indonesia, which cooled the whole Northern Hemisphere (Briffa et al., 1998) during the already cold Dalton Minimum period (Esper et al., 2004). The year 1816 was characterized in the Swiss Alps by very bad weather and harvests (Seppey, 1986). LaMarche and Fritts (1971) found that the second growth depression visible around 1915 was correlated with the advance of many glaciers. Finally, the last one, observed in both upper treeline stands in the late 1970s, was simultaneous to lower summer temperatures (Fig. 2; Beniston et al., 1997). Recently established trees growing in conditions slightly above their physiological threshold have probably been stressed by a succession of colder summers, while trees in lower stands have not been influenced (Kullman, 2001). Conversely, the radial growth increase observed in three stands (both lower treelines and forest in Grächen) coincided with a strong temperature increase (Fig. 2; Paulsen et al., 2000; Luterbacher et al., 2004; Rebetez and Dobbertin, 2004) and corresponds to a similar change observed in Larix decidua chronologies in the Swiss and French Alps (Petitcolas, 1998; Büntgen et al., 2005). Longer observations will be necessary to confirm this growth increase and to see if the upper treeline stands follow this trend.

The importance of high temperatures for growth was confirmed by the response function analyses. At both sites, and at almost all altitudes, the correlation between July temperatures and tree-ring growth was positive (Table 3), as has often been reported for trees at subalpine altitudes (e.g., P. cembra: Eckstein and Aniol, 1981; Petitcolas, 1998; and Picea abies: Lingg, 1986; Petitcolas, 1998). Warm temperatures during the previous autumn were favorable as well (Table 3). This period is after the end of tree-ring growth and corresponds to the accumulation of reserves for the next season (Tranquillini, 1979; Contini and Lavarelo, 1982). This period is also at the end of needle cuticle development (Tranquillini, 1979; Oberhuber, 2004), which is an important stage for building up good protection against winter frost, drought, or wind.

However, in this dry part of the Alps, precipitation was positively correlated with growth as well (Table 3). The positive correlation with precipitation during the previous November–December period is probably related to snow cover. In these months, precipitation is mainly snow. An abundant, early snow cover provides good protection against soil frost and may help trees to avoid frost desiccation at the end of winter, when transpiration increases with temperature (Oberhuber, 2004). A light drought effect was visible during the growing season as well, with a positive correlation with rainfall in two stands in August and in four stands in the previous October (Table 3). Petitcolas (1998) observed similarly that low rainfall during the previous August affected growth of conifers in the inner French Alps. Tree rings are still growing in August in subalpine areas, and photosynthesis is possible for P. cembra until the end of October (Tranquillini, 1979). Körner (1995) suggested that trees are not directly sensitive to drought, but a lack of water stops microbiological activity in the soil, which decreases the availability of nutrients. However, in the dry inner Alps, it is possible that atmospheric humidity is low enough to reduce photosynthesis, which is greatly inhibited when air humidity falls below 50% (Tranquillini, 1979). The mean monthly air humidity values in Grächen are regularly around 60% (minimum 48% in August 1989; MeteoSwiss), and daytime values are even lower.

Future Trends

Since 1980, radial growth increased in five of the six considered P. cembra stands following the increase of summer temperatures. Annual ring growth at upper treeline is now as fast as in the forest stands (Fig. 2; Paulsen et al., 2000). As seedling recruitment and radial growth seem to respond similarly to warmer conditions, like those observed at the end of the Little Ice Age (Figs. 2 and 3), it is probable that colonization is ongoing at both sites, although we did not count and cut saplings to prove it. Phenological observations have shown that, in Switzerland, P. cembra is the best adapted conifer to colonize the lower alpine belt thanks to the decrease of the necessary degree-days for development with increasing elevation (Theurillat and Schlüssel, 2000).

Future colonization depends on tree establishment. Saplings have to cope with natural hazards for up to 50 years to ensure a successful recruitment (Szeicz and MacDonald, 1995) and hence are sensitive to a high climatic variability and extreme events like long snow cover in spring (Schönenberger et al., 1990; Taylor, 1995; Millar et al., 2004). Exceptional conditions have become more frequent since 1980 and are expected to increase significantly with global warming (Beniston et al., 1997; Rebetez et al., 1997; Schär et al., 2004). Thus, predictions on future recruitment in the treeline ecotone are difficult and need supplementary experiments and observations on sapling survival under exceptional conditions.

Even a drought-tolerant subalpine species, such as P. cembra, could increasingly suffer a moisture deficit. Drought stress is already visible in response function analyses (Table 3), and climate change models for the end of the 21st century (Mitchell et al., 2004) are predicting drier summers (19–36% decrease for June–August months) and warmer temperatures (annual mean increase of 3.4–6.5°C, increase for August of 5.9–11.3°C). It is highly probable that radial growth will be limited, or even reduced, by these dry summer conditions, because of lower rainfall and higher evapotranspiration (Barber et al., 2000; Daniels and Veblen, 2004; Rebetez and Dobbertin, 2004; Holtmeier and Broll, 2005). Data on the consequences of the 2003 drought (Schär et al., 2004) would be very instructive on the behavior of P. cembra in very dry conditions.

Conclusion

Climate change and land-use have influenced treeline dynamics in the Alps. At the end of the Little Ice Age, Pinus cembra establishment and radial growth in the treeline ecotone were simultaneously favored by warmer climate. This ecotone quickly shifted upslope in a natural treeline, while a 60-year delay was observed in an anthropogenic treeline, and only after a decrease of cattle activity. Radial growth is strongly linked to summer temperature along the whole altitudinal gradient we considered, but our results suggest that growth could be limited by long periods of drought in the lowest stands.

The treeline ecotone expansion will certainly continue, at least in the previously pastured areas where present ecotone elevation is lower than in undisturbed situations. Further studies will be necessary to precisely identify the consequences of continued climate warming. Tree establishment depends on a series of consecutive favorable years while predicted decreasing summer precipitation may reduce tree growth. Future studies should include investigations of the influence of extreme events, like the 2003 summer drought.