1Elevational tree line change in the southern Swedish Scandes was quantified for the period 1915–2007 and for two sub-periods 1915–1975 and 1975–2007. The study focused on Betula pubescens ssp. czerepanovii, Picea abies and Pinus sylvestris at a large number of sites distributed over an 8000-km2 area. The basic approach included revisitations of fixed sites (elevational belt transects) and measurements of tree line positions (m a.s.l.) during these three periods.
2Over the past century, tree lines of all species rose at 95% of the studied localities, with means of 70–90 m. All three species displayed maximum upshifts by about 200 m, which manifests a near-perfect equilibrium with instrumentally recorded air temperature change. This magnitude of response was realized only in particular topographic situations, foremost wind-sheltered and steep concave slopes. Other sites, with more wind-exposed topoclimatic conditions, experienced lesser magnitudes of upshifts. Thus, spatial elevational tree line responses to climate change are markedly heterogeneous and site-dependent. Modelling of the future evolution of the forest-alpine tundra transition has to consider this fact. Even in a hypothetical case of substantial climate warming, tree lines are unlikely to advance on a broad front and a large proportion of the alpine tundra will remain treeless.
3During the period 1975–2007, the tree lines of Picea and Pinus (in particular) advanced more rapidly than that of Betula towards the alpine region. These species-specific responses could signal a potential trajectory for the evolution of the ecotone in a warmer future. Thereby a situation with some resemblance with the relatively warm and dry early Holocene would emerge.
4Substantial tree line upshifts over the past two to three decades coincide with air and soil warming during all seasons. This implies that both summer and winter temperatures have to be included in models of climate-driven tree line performance.
5Synthesis. Maximum tree line rise by 200 m represents a unique trend break in the long-term Holocene tree line regression, which has been driven by average climate cooling for nearly 10 000 years. Tree line positions are well-restored to their pre-Little Ice Age positions. Recent tree line ascent is a truly anomalous event in Holocene vegetation history and possibly unprecedented for seven millennia.
Beyond broad conceptions such as unfavourable heat balance, no generally accepted detailed mechanism has yet been identified that predicts the precise elevation and structure of the tree line. As evident from recent research, relevant mechanisms seem to be critically dependent on spatial and temporal scales of consideration (Tranquillini 1979; Walsh et al. 1994; Grace et al. 2002; Holtmeier & Broll 2005), but maybe the very idea of a fundamental and globally valid tree line theory is untenable (Miehe & Miehe 2000; Crawford 2008). From a practical and predictive point of view, the importance of defining regional or local drivers and modulators should be more stressed. Besides, the balance between summer and winter climatic conditions (temperature and wind) needs further investigation, since some studies suggest growth and survival of tree line trees are not exclusively related to summer conditions (cf. Grace & Norton 1990; Moiseev & Shiyatov 2003; Sturm et al. 2005; Kapralov et al. 2006; Kullman 2007a; Rickebusch et al. 2007). This aspect is of special relevance as warming during the present century is projected to particularly increase during the winter period (IPCC 2007b).
An understanding of tree line function, based on real-scale observations, is critical to the generation of realistic ecosystem models for devising possible future climate change scenarios of either climate warming or cooling (Wieser & Tausz 2007). Effective instruments are formed by detailed long-term (decadal to secular) observation series (‘natural experiments’) of tree line dynamics against a background of instrumental meteorological records (cf. Kallio et al. 1986; Woodward 1987; Holtmeier 2003).
The present study reports and analyses the results from a multi-site, regional tree line monitoring effort, spanning almost a century (1915–2007). In addition, multi-site Holocene paleo-tree line information from the same area contributes to putting recent tree line dynamics into a broader historical context (Kullman 2001a; Kullman & Kjällgren 2006; Öberg 2008).
Assuming a perfect tree line-climate equilibrium and based on the literature cited above, we make the following general predictions for positional tree line evolution over the past 100 years, in accordance with an observed summer air temperature increase by 1.4 °C (see below) and assuming a lapse rate of 0.6 °C per 100 m of altitude (Wieser & Tausz 2007):
1In ideal cases, tree lines of all species would ubiquitously shift upslope by about 200 m.
2Due to local topoclimatic effects, most localities will display significantly more modest or no advances at all, relative to the ideal case 1.
3Over shorter periods, there will be species-specific performances with respect to the pace of tree line shift.
4Winter temperatures are co-drivers of tree line performance.
The study area embraces the southern Scandes (Caledonides) in Sweden within the counties of Jämtland and Dalarna (Fig. 1). More than 100 sites are distributed over an area of about 8000 km2, ranging from 63°25′ to 61°05′ N and from 12°03 to13°11′ E.
The mountain summits reach altitudes of 1000 to 1800 m a.s.l., while the mainly U-shaped valleys have their floors at 600–800 m a.s.l. The bedrock consists of amphibolites, quartzites, sparagmites and calcareous tectonites. More or less well-drained podzolic soils and peat have developed on prevailing glacial deposits.
The study area harbours the southernmost mountain glaciers in the Swedish Scandes, with their lower fronts situated about 400 m above the highest tree line. Currently, permafrost exists neither in nor close to the tree line ecotone.
The regional climate ranges from weakly maritime in the northwest to a more continental character in the eastern and southern parts of the study region. Mean annual temperatures as recorded by official meteorological stations range between 1.0 and –2.0 °C and precipitation varies between 400 and 1000 mm per year (Raab & Vedin 1995). Data collected at two meteorological stations, Storlien/Visjövalen (642 m a.s.l.) and Särna (435 m a.s.l.), represent climatic conditions within the northern and southern part of the study area, respectively. For Storlien/Visjövalen, the mean temperatures for January, July and the year are –7.6, 10.7 and 1.1 °C, respectively. Corresponding data for Särna are –12.1, 13.3 and 0.8 °C. The annual precipitation is 857 and 601 mm, respectively. All data have been obtained from the Swedish Meteorological and Hydrological Institute and represent the period 1961–1990.
the tree line ecotone
The upper boreal forest with the alternating dominants Picea abies (Norway spruce) and Pinus sylvestris (Scots pine) gradually changes into the sub-alpine belt, which is characterized by Betula pubescens ssp. czerepanovii (mountain birch). This belt has an elevational range of 50 to 300 m, with its largest extension in the westernmost and most snow-rich parts of the study area. As a rule, the highest tree line is formed by Betula, followed by Picea and Pinus. Locally and particularly in the southern and most continental and snow-poor regions of the study area, Pinus or Picea reaches higher than Betula.
The ground-cover vegetation within the ecotone is dominated by ericaceous dwarf-shrub heaths (Vaccinium myrtillus, Vaccinium uliginosum, Betula nana, Empetrum hermaphroditum) with some low herbs, sedges and grasses, alternating with open mires and snow-bed communities. A more detailed account of the tree line ecotone and its geoecological context is provided by Kullman (2005b).
During the past 30–40 years at least, the tree line ecotone as a whole has not been subjected to stand-destroying natural disturbances, such as fire, insect outbreaks, avalanches, slope failure or wind-throw, except for on a strictly local level.
At the tree line, Betula and Picea reproduce regularly and predominantly by vegetative means (basal sprouts and layering, respectively). Picea may form clonal groups with ages of more than 9000 years (Kullman 2001a; Öberg 2008), while maximum ages of Betula clones are unknown, but ages of about 500 years have been inferred (Kullman 2005c). In congruence with shifting climatic conditions both Picea and Betula display striking phenotypic plasticity and swiftly alternate growth form between krummholz (stunted or prostrate individuals) and arborescent.
Typically, the transition between the closed upper birch-forest belt and the alpine tundra, that is the tree line ecotone, is diffuse, gradual and structurally unique for each mountain side, which makes a stringent ‘forest line’ definition impossible. Thus, the elevation of the uppermost tree (tree line) at a specific site bears a more straightforward relation to the regional climate and climate change than does any forest line. This distinction is particularly relevant when seeking efficient bioindicators of global climate change (Kullman 1998).
The use of the tree line as a monitoring target for the detection of ecological responses to climate change necessitates a sharp, practical and homogeneous definition. Henceforth, the tree line is defined for each tree species as the maximum elevation (m a.s.l.), at a given site, of single- or multi-stemmed trees with a minimum height of 2 m (cf. Wardle 1974; Hustich 1979; Camarero & Gutiérrez 2004). This critical height facilitates correlation between tree line and standard meteorological variables particularly as the concerned tree line marker does not become entirely snow-covered. In the present case, the decision for the use of this specific tree line definition was inevitable, since the basic historical records (Smith 1920) were made on the same premises.
With the current definition, the tree lines of all species rise by c. 300 m in elevation along a 250 km north–south gradient in the study area, which primarily reflects increasing climatic continentality. Along this gradient, tree lines reach their highest elevations on south-facing and wind-protected slopes (Kjällgren & Kullman 1998, 2002).
Above the tree line of birch, a pool of low-growing and sparsely scattered saplings of birch, pine and spruce extends 500–700 m higher (Kullman 2007b). The uppermost individual of each species constitutes the tree species line.
Grazing and trampling associated with summer reindeer pastoralism (free-ranging) has a long tradition on the high mountains (cf. Eriksson et al. 2007). It is well-known that reindeer affect the subalpine birch vegetation and alpine tundra as a more or less chronic disturbance (Cairns et al. 2007), with quantitatively similar impacts from prior wild and present semi-domesticated animals, as discussed by Helle (2001). The intensity appears to have fluctuated modestly around a stable mean over the first 75 years of the past century (Moen & Danell 2003). Thereafter, reindeer density in the study area has remained on a significantly higher level (Öberg 2002; Eriksson et al. 2007). Existing evidence suggests that reindeer grazing has not been a decisive obstacle to birch tree line rise during the past century, since tree line positions and dynamics do not significantly differ between areas of high and low reindeer land use (Kullman 1979, 2005c; Holmgren & Tjus 1996; Kjällgren & Kullman 1998).
Given the present definition, the tree line has not been particularly sensitive to past human impact, which for several reasons, for example, the trade-off between accessibility and reward, has tended to decline in intensity above the closed forest (cf. Blüthgen 1960).
monitoring and sampling
The fundament of a regional tree line monitoring network was established around 1915 in the southern Swedish Scandes by Dr Harry Smith, a well-respected Swedish botanist (Smith 1920). This network is the source of unique baseline information, consisting of almost 400 data collection points and covering c. 8000 km2. During the years 1915 and 1916, Smith hiked along the tree line from north to south in the study area. For approximately each 2 km interval along this route, the highest position (m a.s.l.) of the mountain birch tree line (defined above) was determined by use of an aneroid barometer, repeatedly calibrated against distinct points on the best available ordnance maps.
Around 1975, repeated tree line measurements by a Paulin aneroid barometer were carried out by the senior author for 213 of Smith's 391 original sites, which could be unambiguously re-localized. The accuracy of Smith's elevational measurements was tested by tree age determinations using root collar tree ring counts on living stems. These assessments were performed in belt transects with a width of c. 500 m, centred on the best point-estimate of Smith's locations. Each transect extended downslope to the highest position (m a.s.l.) where living and dead tree-sized birches could ascertain tree growth for the years around 1915 (Kullman 1979). This can be taken as a minimum altitude for the contemporary tree line. For 85% of the sites, the difference between Smith's tree line measurements and the modern tree ring based estimates of tree line positions around 1915 was 25 m or less. In all cases of incongruence, the highest value was chosen for comparison with the later positional measurements. Thus, the original baseline data of this monitoring system are approved to be of high quality. These surveys yielded accurate and quantitative assessments of elevational tree line change or stability over a period of c. 60 years (Kullman 1979).
For spruce and pine and a few birch sites, no records existed before 1975. Tree line positions by 1915 were estimated solely by aging of trees at and below the current tree line, to find out the highest level for trees with more than 60 tree rings above the ground level (Kullman 1981, 1986).
During the re-surveys in the mid-1970s, the uppermost trees within each transect (henceforth termed ‘tree line marker’), were tallied and photographed for future recognition. This ensures a particularly high positional accuracy for later re-surveys.
The most recent update of the data set was carried out by the authors between 2005 and 2007. Due to funding restrictions only about 50% of the sites investigated by the mid-1970s were randomly selected and re-surveyed. Tree line positions (altitude, latitude, longitude) representing the years around 1915 and the mid-1970s were re-localized and now even more accurately positioned, using modern topographical maps and a GPS navigator (Garmin 60CS). The elevational accuracy of these measurements was repeatedly checked against distinct points on the topographical maps (summits, lakes, etc.). In almost all cases the difference was less than 5 m. The precision of the measurements was certainly higher than previously when data were obtained by barometer measurements. As a consequence, some tree line positions reported in the mid-1970s were corrected by 5–10 m.
Each transect (see above) was extended upslope and carefully scrutinized for the presence or absence of trees above the position (m a.s.l.) held in the mid-1970s. Tree line markers were tallied with respect to size and vitality and their surrounding landscapes were re-photographed. Only a minor fraction of these images are shown here. Some of the raw data contained in this study have been published as a fundament for continuous and detailed environmental monitoring (Öberg 2008).
The sampling design, constrained by the original data provided by Smith (1920), implies a certain selection bias, which precludes rigorous randomized-based testing of the data contained in this study. For example, there is a preponderance of sites facing south and the density of re-localized sites decreases towards the south in the study area. Thus, only descriptive statistics, that is mean, S.D. and range, can be computed and displayed. However, the large sample size provides evidence of secular response patterns in an array of different and representative types of physiographic and climatic settings.
By the methodological approach outlined above, elevational tree line shifts (or stasis) of Betula, Picea and Pinus can be quantified separately for the periods 1915–1975, 1975–2007 and for the entire period 1915–2007. The obtained values of tree line advance have been rounded to the nearest 5 m. These data constitute the base of the present study, which has the character of an assemblage of individual case histories, each one reporting observations of reality, integrating climate change with unique site characteristics and history.
The botanical nomenclature follows Mossberg & Stenberg (2003). Henceforth, the following acronyms are used: BTL, birch tree line; STL, spruce tree line; and PTL, pine tree line.
Electronic files accounting for the position of each study site and its past and present tree line elevations are stored at the Department of Ecology and Environmental Science, Umeå University and can be obtained on request from both authors.
Regional climate evolution over the past century
Since an absolute Holocene thermal nadir in the late 19th century, that is the final phase of the so-called Little Ice Age (Grove 1988; Bradley & Jones 1993), the investigated region has experienced secular climate warming (Alexandersson 2006), which is larger than the European average (Moberg et al. 2005; IPCC 2007b). Figure 2 depicts the mean annual temperatures as recorded by the meteorological station Sveg (360 m a.s.l.) at the southern fringe of the study area. This station, which holds the longest instrumental temperature record in the region, manifests a positive linear trend of 1.4 °C over the past 131 years, with the most consistent warming between 1876 and 1940. In line with this, the concerned region has experienced a long-term decreasing trend in days with snow cover (Moberg et al. 2005).
The records of air temperature and precipitation since 1915, that is from the time of the earliest tree line records, are given separately for the meteorological stations Storlien/Visjövalen (adjusted for station move) and Särna. With respect to summer temperatures (June to August), both stations display significantly rising linear trends by 1.3 and 1.0 °C, respectively (Figs 3 and 4). Likewise, winter temperatures (December to February) increased by 1.4 and 0.9 °C, respectively, over the same period of time (Figs 3 and 4). Closer analysis based on 4th order polynomial curve fitting (Figs 2–4) reveals that the thermal trends (both summer and winter) are composed of two distinct warming plateaux, viz. the 1930s and 1988–2007, respectively, and separated by a slightly cooler interval. In addition, it appears that the magnitude of the most recent phase of warming was clearly larger for the winter period than for the summer period.
Because of inconsistent instrumentation and local effects, long-term records of precipitation are inherently uncertain (Alexandersson 2006). With this in mind, data from neither Storlien/Visjövalen nor Särna display a clear trend for summer and winter precipitation from 1915 to 2007 (not shown). However, data for a larger region, encompassing the study area, indicate that annual precipitation has increased by 5–10% over the past century, with a somewhat larger rise during the winters of the past 25 years (Alexandersson 2006).
Perhaps the most conspicuous physical landscape-scale manifestation of climate change over the past century is frontal recession and thinning of mountain glaciers and reduced late-summer snow pack on the alpine landscape. After a slowdown and minor advance during some cooler decades before the 1990s (Lundqvist 1969), glacier retreat and earlier snowmelt have resumed, particularly since 2000–2001 (Kullman 2003, 2004a,c, 2007b). The lower terminus of the glacier Sylglaciären has retreated by 155 m in altitude since the early 20th century.
A large number of semi-perennial snow patches in the low alpine zone has disappeared earlier than before in the summers of the past 10–15 years. In consequence, alpine and subalpine soils have tended to become drier in late summer (Kullman 2004b,c, 2005c, 2007b,c). This process was initiated in the study area more than 50 years ago, in close relation to the first warming peak of the 20th century (Smith 1957; Nordhagen 1964).
Prior to the mid-1990s, the study area harboured the southernmost outposts in Sweden of sporadic permafrost (Smith 1920; Kullman 1989), which have subsequently disappeared (Kullman 2007c).
A phenological event of great relevance for the present study is leaf unfolding of mountain birch. During the years 1977–2007, this process has been observed and photographically documented by the first author for the mountain birch forest belt (up to the tree line) on the east-facing slope of Mt. Lillulvåfjället (Fig. 5). The latest observed date during this phase was 19 June 1987. In all other years, leaf unfolding occurred between 5 and 10 June. For the same area, Smith (1920) has provided dates for the period 1908–1919, when leafing occurred around 20–25 June. Overall, it appears that both snow melt and leafing took place at least 2 weeks earlier during the past two decades relative to the early 20th century.
Over the period 1985–2007, maximum summer (June to August) and minimum winter (December to February) soil temperatures (30 cm depth) have been measured by thermistors close to the tree lines of birch, spruce and pine on the SE-facing slope of Mt. Storsnasen. For both these parameters, positive trends with a range of 2.2–3.4 °C were obtained. A tendency for relatively more pronounced warming in the winter period can be discerned (Kullman 2007a,c).
positional tree line change
Site-specific data on tree line upshift or stability over the period 1915–2007 and its sub-periods 1915–1975 and 1975–2007 are reported for mountain birch, spruce and pine, separately (Figs 6–8). Irrespective of species, about 95% of all studied locations experienced an upward shift in tree line elevation between 1915 and 2007 (Table 1). Only one case of tree line descent was recorded and that was at a site where birch retracted by 15 m during the period 1975–2007.
Table 1. Species-specific frequencies of investigated localities with tree line rise during different periods of time
Common for all studied species, the largest advances (1915 to 2007) were about 200 m, with means of 70–90 m (Figs 6–8). Mountain birch achieved most of its total elevational expansion during the period 1915–1975, while spruce and pine ascended relatively more during the period 1975–2007.
The displacement data from all the individual sites are computed as means of annual advance rates for each tree species and for each of the above-mentioned three periods (Table 2). For the period 1915 to 1975, PTL rose clearly less rapidly than BTL and STL. Advance of BTL was less pronounced than that of STL and PTL over the period 1975–2007, when PTL displayed the highest upshift rate. The entire study period 1915–2007 was characterized by the largest rise for STL, followed by PTL and BTL. Thus, on the secular scale, the coniferous tree lines seem to have slightly approached the birch tree line. That course of change was particularly pronounced during the period 1975–2007.
Table 2. Species-specific rates of altitudinal tree line shifts during different periods of time
Altitudinal shift (m year−1) Mean ± SD
0.75 ± 0.58
0.65 ± 0.78
0.74 ± 0.49
0.78 ± 0.63
1.34 ± 1.41
0.98 ± 0.61
0.35 ± 0.45
1.66 ± 1.25
0.81 ± 0.47
From an analytical and predictive point of view, the range of elevational displacements is of particular interest. Over the entire study region and for all three periods, sites with very large upshifts (maximum 220 m) intermingle, seemingly irregularly, with those displaying modest, minor or no advance. These data, arranged as a north–south sequence, give no indication of a relationship between magnitude of tree line rise and latitudinal position within the study area (Figs 6–8).
In order to understand the striking site-dependence of response patterns in greater detail (Figs 6–8), physiographic characteristics of sites with the largest advances (≥100 m) were contrasted with those displaying no or only insignificant change (0–25 m). For that purpose, snow cover and wind exposure were assessed for each tree line site of the three tree species. These site classifications draw on Kullman (1979, 1981) and have subsequently been refined and updated by field-assessment over several years and during different seasons. The species-specific proportions of sites representing different classes of this parameter were computed for the two contrasting size categories of tree line change (Table 3). Thereby it appears that the largest upshifts of BTL (1915–2007) occurred predominantly in lee slopes prone to large snow accumulation. STL achieved maximum rise under intermediate snow and wind conditions, while somewhat more exposed slopes with a sparse snow cover were relatively most conducive to large advances of PTL. Typically, BTL, which generally reaches higher elevations than STL and PTL and meets the strongest winds, has advanced most readily in steep concave terrain with a particularly rugged micro relief. These slopes tend to be relatively warm, contain relatively deep snow and soil packs and are frequently discharge sites, well-watered throughout the summer (cf. Sundqvist et al. 2008).
Table 3. Species-specific proportions of sites representing three different categories with respect to physiographic site characteristics and subdivided into two contrasting groups of tree line change. 1 – Wind-exposed, level or convex terrain with a thin snow cover. 2 – Intermediate between 1 and 3. 3 – Concave lee slopes with a relatively deep snow cover
n = 0
n = 7
n = 0
n = 5
n = 10
n = 0
n = 4
n = 16
n = 27
n = 4
n = 8
n = 3
n = 25
n = 0
n = 8
n = 0
n = 1
n = 5
Consistently, at most sites where the birch and spruce tree line rises have not attained the predicted maximum value of about 200 m, the slope is entirely level or there is a sharp discontinuity, a ‘knee’, in the slope morphology right at the ‘stop elevation’ for further tree line ascent. Here concavity transforms into level and wind-swept terrain and this shift effectively halts further tree line advance. Pine, with its generally lower tree line, is less affected by this type of topographic effects and constraints.
character of the new tree line
The vast majority of tree line markers (all species) are 2–3 m tall and it is obvious from the architecture of their crowns that they have only recently attained the critical size to fit the tree line definition. Pinus in particular has displayed exceptionally long leaders during the past 5 years or so. The same kind of response has been observed also for young saplings growing high above the tree line.
In most cases, BTL and STL have advanced by means of accelerated in situ growth form transformation from old-established krummholz into erect, arborescent modes (Fig. 9). This is generally evident from striking disproportions between stout basal and substantially more slender higher parts of the main trunks and supported by tree ring counts (Kullman 1986, 1993a, 1995a). Case studies have disclosed that such individuals, which mainly reproduce by basal sprouts and layering, may have become established long before the 20th century (Kullman 1993a, 2001a, 2005c; Öberg 2008). As a further testimony of this circumstance, Smith (1920) and Kilander (1955) report that low-growing birches and spruces (<1 m high) existed at some of the study sites, roughly up to the position of the currently advanced tree line.
Krummholz-spruces and spruces that recently converted to upright trees and have been reproductively ‘silent’ for centuries or more, have started to reproduce sexually over the past 10 years or so (Kullman 2006). In the case of Pinus, the new tree line markers have produced some offspring in their vicinity during the past decade or so (Kullman 2007b).
Obviously also a present-day anomaly, sparsely scattered young saplings of all concerned tree species now exist high above the current tree line as a result of a recent upward ‘surge’ of the tree species line, which has reached 500–700 m higher than the lines (Kullman 2001b, 2007b,c). Case studies indicate that the majority of these saplings has become established since the late-1980s in concert with a substantially increased seed viability of tree line trees (Kullman 2007a,b,c). This implies a remarkable rate of upslope migration, in the order of 30 m year−1. The density decreases with altitude and range between 2 and 10 individuals ha−1 (Kullman 2004c, 2007b). Typically, these specimens take advantage of minor topographic irregularities, e.g. boulders or concavities, which provide maximum solar radiation, wind-shelter and ample soil moisture conditions. Only rarely do these individuals show signs of reindeer browsing.
In great contrast to the situation prevailing between the mid-1970s and the late-1980s (Kullman 1981, 1993b), it was particularly evident during the field-work 2005–2007 that most (95%) specimens of pine and spruce in the tree line advance zone displayed virtually no new winter-desiccation injuries to needles and buds (cf. Kullman 2007a,c).
structural change of the tree line ecotone
In general, the new tree line markers are not isolated outliers, but closely accompanied by similar individuals growing slightly below. Concomitant with general tree line ascent, stands of more or less closed tree birch vegetation have advanced upslope by densification of sparse founder populations (cf. Kulllman 2007c). This is particularly evident in the most maritime and snow-rich parts of the study area. Quite frequently, the solitary tree line markers of the early-20th century are currently found growing in a matrix of predominantly younger trees. This is a local ecosystem shift which also manifests as the emergence of a more sylvatic ground-layer flora, spreading upslope from lower elevations, both to newly established ‘tree islands’ and to treeless alpine tundra (Kullman 2007c). Minor ‘embryonic’ birch woods are rapidly forming in restricted slope concavities where excess of snow previously precluded any growth of birch (Fig. 10), and where earlier annual melt-out during the past 10–15 years has enabled birch establishment and growth (Kullman 2007b). This change has taken place despite fairly strong reindeer grazing and trampling. On the other hand, at strongly wind-exposed sites in the tree line ecotone, birch and occasionally conifers are locally withering and retreating in response to direct wind action and wind-driven soil erosion (Kullman 2005a).
The spatial structure of the birch tree line advance zone is categorized into tree main types (Kullman 1979, updated): (i) wedges, lobes or bands of trees protruding uphill along sheltered ravines or incised watercourses, (ii) small and dense groves in minor topographic concavities and (iii) scattered solitary trees. Only rarely does the newly established tree vegetation border the alpine tundra as a sharp line of closed forest. Possibly, these patterns indicate trajectories for further arboreal advance in a hypothetically warmer future.
In those parts of the study area where the climate is most continental and where tree line birch is less expansive (see above), a sparse belt of pine trees has leap-frogged the narrow subalpine birch belt. This may represent the onset of an entirely new zonation pattern, which mimics the situation in the early Holocene (Kullman 2004b). In these regions, many lower mountains with summits of alpine tundra have become entirely covered with mixed stands of spruce, pine and birch during the course of the past 80–90 years, with peak regeneration during the warmest decades. One example is Mt. Tandövala, which was Sweden's most southerly spot of alpine tundra until recently (Kullman 2005e).
In principle, all our basic predictions have been validated.
Tree lines of the principal boreal tree species in the southern Scandes have shifted upslope during the past 100 years in concert with rising trends of summer and winter temperatures. Maximum advance by about 200 m for all species, concomitant with 1.4 °C summer and winter warming, matches almost perfectly the predicted value, based on a lapse rate of 0.6 °C per 100 m altitude and the assumption of a near-perfect tree line-climate equilibrium. This supports a theoretical projection of about 700 m elevational tree line advance in response to a hypothetical 5 °C warming over the present century (Grace et al. 2002). The potential for such a magnitude and rate of upslope shift is further substantiated from the remarkable advance of the tree species line over the past 20 years, that is 30 m year−1.
The operational efficiency of a common broad-scale driver, that is climate change, is indicated by analogously large maximum upshifts of tree species vegetation in Swedish Lapland, far north of the study region (Kullman 1991a,b; Kullman & Kjällgren 2006; Sundqvist et al. 2008). A further indication in the same direction is provided by dendroecological case studies of tree line populations in the study area, which display regeneration peaks and lows in congruence with periodic shifts in temperature (Kullman 1993a, 2000, 2004b, 2005e, 2007c).
A tight relationship between climate variability and tree line performance is supported also by records from the present monitoring network, showing that tree line rise was halted or locally even slightly reversed in response to relatively cold periods during the 1960–1980s. In conjunction with that phase, low winter temperatures (air and soil) and a sparse snow cover caused severe landscape-scale defoliation and even mortality of both older trees and young saplings (review by Kullman 1997).
Current regional-scale upshifts of tree lines are paralleled in the same region with roughly equally large uphill migrations during the past 50–60 years of several resident ground-layer species, many with woodland (boreal) affinities (Kullman 2001b, 2007b,c). Analogous species displacements are recorded in the European Alps (e.g. Walther et al. 2005).
Thus, the balance of evidence presented above indicates an ongoing and profound all-level (trees to ground-layer) transformation of the subalpine-alpine tundra system, most likely attributable to climate forcing (cf. Sundqvist et al. 2008). This also implies that tree line performance can be used as a marker for more widespread ecological change of stability (Kullman 2007c).
The ability of rapid positional responsiveness of tree lines to altered climatic conditions, as evidenced in this study, is reflected also by paleorecords (megafossils) from the same area. These encompass the entire Holocene and show an average tree line descent (Pinus sylvestris) of 500–600 m (Kullman 1995b, 2001a, 2004a; Kullman & Kjällgren 2006). Similarly, a dynamic climate – tree line equilibrium is inferred from paleobotanical data representing the European Alps and covering the entire Holocene (Tinner & Kaltenrieder 2005).
During the entire observation period (1915–2007) all species have advanced to broadly the same extent, both with respect to mean and maximum upshifts. This suggests that in one and the same climatic region, tree lines of species with different ecologies respond to climate change according to a common mechanism. For shorter sub-periods, however, there are inter-specific differences. Over the period of 1915–1975, Betula and Picea displayed larger and more rapid upshifts than Pinus. Eventually, Pinus has lined up with Betula and Picea. Conceivably, this dichotomy is not only a consequence of a delayed response of Pinus owing to its stronger reliance on genotypic spread, but also suggests that the climate has gradually become particularly conducive to pine. Only pine has current tree line markers, which originated from seeds which germinated as recently as the 1970s (Kullman 2000).
Magnani et al. (2007) launched a general theory that boreal ecosystems, by implication including tree line ecotones, are nitrogen-limited (cf. Sveinbjörnsson et al. 1992; Weih & Karlsson 1999) and thereby indirectly controlling the carbon balance. Thus, it could be argued that the current tree line rise was merely a response to air-deposition of anthropogenic nitrogen, although quite modest in the study area (Akselson & Westling 2000), rather than an effect of rising temperatures. However, several circumstances contradict such a supposition. First, tree line rise was initiated early in the 20th century, prior to any extra anthropogenic deposition of nitrogen in remote mountain regions of northern Sweden. Second, the complex and topoclimatically related pattern of tree line shift is reasonably not paralleled with a comparable spatial deposition pattern of airborne nitrogen. Third, tree lines tended to decline during the cold periods between1960s and 1980s, when nitrogen loads were on the rise (Kullman 1997). Fourth, there is a strong case that the tree line phenomenon is not a consequence of a negative carbon balance (Körner 1998, 2003; Grace et al. 2002). Fifth, it is doubtful that high levels of nitrogen deposition could further tree line rise by increasing winter survival of tree seedlings as suggested (Weih & Karlsson 1999). This is because the ‘bottle-neck’ in the generation of tree line trees is well-beyond the seedling stage (Kullman 1981; Grace et al. 2002).
One of the most important aspects of the present study is the large variation between sites with respect to the magnitude of tree line rise, that is, 0–220 m. Only quite infrequently have tree lines reached the potential positions as prescribed by regional ambient temperature conditions. Ultimately, this discrepancy relates to the highly heterogeneous and complex mountain landscape, where topoclimatic conditions vary considerably over short distances and thereby mediate responses to climate change. Thus, over most parts of the landscape local site conditions, foremost wind exposure and associated snow distribution patterns, have prevented trees to expand up to their potential elevation provided by regional climatic factors. Slopes with a complex and large-scale concave relief have offered the best preconditions for full realization of potential climate-driven tree line rise (cf. Kullman 1979). Optimal conditions in this respect arise as the combined effect of wind-speed reduction, stable snow cover protection and accumulation of ample soil, nutrients and moisture. On the other hand, sites with a more simple and wind-exposed topography are substantially less apt for tree line rise in response to climate warming and likely also to climate cooling. Thus, there exists in the Scandes a broad spectrum of natural tree line sites with various degrees of responsiveness to climate change. An analogous pattern has been hypothesized to prevail also in other parts of the world (Luckman & Kavanagh 2000; Lloyd 2005; Holtmeier & Broll 2005; Butler et al. 2007). These circumstances have implications for proper modelling of future landscape evolution, but also for reconstruction and interpretation of the paleolandscape and the paleoclimate. The present results caution against paleo-studies based on just one or a few sites, which may provide entirely misleading conclusions in high mountain landscapes.
On a broader time-scale, these strictly site-specific responses of tree lines in the study area are consistent with paleo tree line performances as reconstructed from megafossil remains. During the warmest and driest phase of the Holocene, that is 9000–10 000 radiocarbon years ago, birch, pine and spruce trees grew 400–600 m higher than today (Kullman & Kjällgren 2006). It appears, however, that this was not a broad-front phenomenon since megafossil remnants of these trees are mainly confined to particular habitats that are of similar character to those where trees expand to the greatest extent at the present day. In the case of mountain birch, these sites were often concave segments of the terrain, were some wind shelter and suitable moisture conditions prevailed. For example, glacier and snow niches, without permanent ice during this early period of the Holocene, offered particularly congenial conditions for growth of scattered and extra-zonal birch clusters at very high elevations (Kullman 2004a,c). Alpine mires and lake sediments in more flat and wind-exposed terrain are generally found to be devoid of subfossil tree remnants and the same seems to hold true for steep avalanche and scree slopes (Kullman & Kjällgren 2006). This indicates that trees never grew in these locations and that large parts of the alpine tundra have a very long treeless history, even encompassing intervals warmer than at present. Taken together, these circumstances suggest that in a hypothetical case of continued climate warming, elevational tree expansion, particularly in the case of birch, is likely to become patchy and fragmentary. Conceivably, a major part of today's alpine tundra will remain treeless, although its ground cover flora and vegetation may change (cf. Kullman 2007b,c). This tentative and empirically based projection challenges simplistic theoretical models, which envisage a future extensive birch forest cover over most of the Scandes (Moen et al. 2004). However, pine, which raised its tree line by establishment of new individuals, may display a more broad-front expansion in the future, replacing much of the current subalpine birch belt. The relatively fast advance of the pine tree line since 1975, and its prevalence as upper tree line marker in the warmer early Holocene (Kullman 1995b), provides some support for such a course of future landscape transformation. In case of a continued trend of late-summer snow-scarcity and associated soil drought, the birch belt may become reduced in extent in favour of the more drought-resistant pine (Kullman 2007b,c; Öberg 2008). The observed reduced vigour of many birch tree line markers, particularly in the most continental and driest parts of the study area (Kullman 2007b; Öberg 2008), supports this contention. Growth ring analyses reveal that mountain birch has suffered from drought stress during particularly warm summers of the past century (Kirchhefer 1996). Obviously, this fits with a more common trend of summer drought in boreal regions throughout Eurasia (MacDonald et al. 2007).
The spruce tree line, which hitherto has made the largest total advance since 1915, is likely to lose its lead since the rise has been almost entirely accomplished by transformation of old krummholz-individuals to erect trees. The pool of such individuals above the new tree line will soon be depleted and therefore the potential for further substantial spruce tree line rise is small. Moreover, tree line rise by spruce may become rapidly reversed at any time in response to just one or a few harsh winters nested in a longer warming trend (cf. Kullman 1997).
One aspect of relevance for general tree line ecology, which originates from this study, concerns the role of winter temperature conditions. As evidenced, tree lines of conifers have risen substantially more rapidly than the birch tree line during the past 2–3 decades. This course of change coincides with air and soil temperature increases, which were most significant for the winter period. Increased survival rates during this period seem to be the consequence of a substantially reduced incidence of foliage dieback due to winter desiccation (Kullman 2007a). In addition, relatively high winter temperatures may stimulate growth rates by means of larger microbial activity in warmer soils and thereby enhanced nutrient availability (Sturm et al. 2005). Besides considering summer temperatures, temperatures and wind conditions during the winter period have to be taken into account as well when modelling tree line responses to future climate change. This view is gaining support from several studies carried out in various parts of the arctic and boreal world (Tranquillini 1979; Grace & Norton 1990; Grace & James 1993; Hofgaard & Wilmann 2002; Kharuk et al. 2005, 2006; Linderholm & Gunnarson 2005; Kapralov et al. 2006; Payette 2007; Rickebusch et al. 2007).
The larger tree line movements (1975–2007) recorded for Pinus and Picea relative to Betula comply with the generalization that rising winter temperatures may favour evergreen coniferous species at the expense of broadleaved deciduous species (cf. MacDonald et al. 2007).
Tree line rise during the past century followed on the Little Ice Age (Grove 1988), when climate cooling had caused widespread tree mortality and regeneration failure, thereby depressing tree lines to progressively lower elevations, with consequent expansion of the alpine tundra (Kellgren 1891; Gavelin 1910; Olafsen 1911; Nordhagen 1923; Kullman 1987, 2005d). The general societal perception and fear of continuation of this phenomenon is clear from the fact that a national park (Sonfjället) was set aside in the study area in 1909, with the outspoken purpose to provide a basis for long-term studies of this course of landscape evolution (Lönnberg 1912).
Centennial tree line rise by a common maximum for all species of about 200 m represents the effect of a reversal of a long-term cooling, forced predominantly by Earth's orbital patterns. This trend has prevailed since the early Holocene (Kullman & Kjällgren 2006) and is largely consistent with Greenland ice core paleotemperature data (Johnsen et al. 2001; Oldfield 2005). To the best of our present-day knowledge, the maximum upshifts of the tree lines seem to have reached slightly above the positions held during the Medieval Warm Period (AD 900–1300) and several preceding millennia (Kullman 2003, 2004b; Kullman & Kjällgren 2006). This is compatible with Northern Hemisphere temperature reconstructions (Esper et al. 2002; Moberg et al. 2005, Oldfield 2005) and represents a full positional recovery from Little Ice Age tree line regression. This contrasts with reports of less swift and more modest responses in some other northern regions (MacDonald et al. 2007), as discussed above. It appears, however, that in the study region the woodland structure (stand stocking) has not been entirely re-built (Kullman 2005d).
Recent tree line evolution constitutes a truly anomalous event in Holocene vegetation history, possibly unsurpassed during the past 7000 years (Kullman 2004a; Kullman & Kjällgren 2006).
This study was defrayed by the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS). We thank two anonymous referees for valuable comments on the manuscript.