One of the central goals of ecology is to describe and explain patterns of distributions and abundances of species. This makes it natural to focus on distributional limits and to ask questions such as ‘Why does this species occur here and not there?’ (cf. Qian & Ricklefs 2004). One of the most striking distributional limits in high mountains is the upper limit of forests. This so-called tree line may take on various forms: it may form an abrupt border where large trees in closed stands change over to tree-less alpine vegetation (Fig. 1a), or it may gradually change from large trees to shrubs and to stunted individuals, i.e. krummholz (Fig. 1b).
In a recent article, Dullinger et al. (2004) report the results of a model that they developed to investigate the response of an Austrian Pinus mugo tree line to climate change, dispersal limitations and the resistance of non-woody vegetation to invasion. As a point of departure for their work they cite the equivocal response of tree lines to changes in climate over the past century. Because the tree line is traditionally considered to be a thermally limited ecotone, the increases in temperature evident over the past century should have resulted in an upslope expansion of the ranges of the sub-alpine tree species present at tree lines. The model described by Dullinger et al. (2004) illustrated that dispersal and invasibility of the tundra matrix rivalled temperature change for importance in controlling range expansion, and they interpret this as an explanation for the unevenness of responses of tree lines under similar regional climate changes. Here we would like to propose an additional controlling factor that may further explain variation in regional tree-line responses to improving climatic conditions at tree lines: herbivory.
There may be several factors that influence the positions of tree lines. It is clear that tree lines in general ultimately depend on an increasingly unfavourable heat balance with increasing altitude (Tranquillini 1979; Körner 1998 and references therein). As the usage of ‘tree line’ and ‘timberline’ is rather confused in the literature, we use ‘tree line’ in a general sense for the entire border zone from closed forest-stands to the upper limit for trees (cf. Körner 1998). A tree line can be seen as a balance between the opposing forces of a slow tendency to advance upslope and more dramatic retreats due to infrequent disturbances (Slatyer & Noble 1992). These disturbances are usually discussed in terms of the effects of frost-drought or wind-shear on established individuals (e.g. Tranquillini 1979; Gamache & Payette 2004), and competition or climatic effects on earlier life-stages (Grace 1989; Cuevas 2000; Dullinger et al. 2004).
Ellenberg (1988) discussed transitional and abrupt tree lines, and called them ‘natural’ and ‘anthropo-zoogenic’ tree lines, respectively. He argued that in a purely climatic-driven tree line, trees would become progressively smaller and smaller as the climate became more unfavourable, until eventually the species could only survive as stunted individuals in the krummholz zone. An opposing theory suggests that gradual tree lines of this type are the result of human activities thinning the forest (Tranquillini 1979). However, contemporary thinking (cf. Holtmeier 2003) regards the latter theory as untenable in the majority of situations. Ellenberg (1988) further argued that in situations where livestock (or natural herbivores) were present, browsing would deplete resources from individuals already living on the margin and thus accelerate their death. This would result in an abrupt tree line where any established trees taller than the browsing line would not be affected, while everything under the browsing line would be killed if the browsing pressure is high enough. In such tree lines, one can usually find single established trees above the actual contiguous forest boundary, and Ellenberg (1988, p. 392) states that: ‘where one tree grows then others can grow near it if man and animals permit, provided the soil is deep enough. A change of the general climate cannot be so abrupt that directly beside a living tree, several meters tall, which has been growing for centuries, conditions could be too rough for the development of another tree’. The key element here is ‘if man and animals permit’. Trees above the tree line can become established in periods of relaxed herbivory pressure and, if they become large enough, can withstand browsing of the lower branches. A browsing-induced tree line could also be depressed below its maximum level; Ellenberg (1988) showed an example based on pollen analyses from the Alps where cutting and grazing since the Iron Age has depressed the current tree line more than 200 m below the potential altitude. Similarly, contemporary tree lines in Scotland have shown remarkable increases in the establishment of seedlings up to and above the regional climatically controlled tree line subsequent to the reduction of fire frequency and exclusion of herbivores (French et al. 1997; Bayfield et al. 1998).
It is well documented that herbivorous mammals may change successional patterns of trees. For instance, deer (Odocoileus virginianus) browsing significantly slows down invasion of old fields by trees in the nutrient poor soils of Cedar Creek, Minnesota (Inouye et al. 1994). Moose (Alces alces) browsing in the boreal parts of the world also has strong effects on forest succession patterns after fire or clear-cutting, as they feed selectively on the early successional deciduous species, thus accelerating the return to coniferous forest (Andrén & Angelstam 1993; Kielland & Bryant 1998; Pastor et al. 1999). Aspen (Populus tremuloides) regeneration has been linked to elk (Cervus elaphus) populations in western North America (Hessl & Graumlich 2002; Hessl 2002). The interaction between herbivores and vegetation under a changing climate is complex and may proceed along several potential pathways (Fig. 2).
It has also been shown that herbivores may affect tree growth and reproduction at the tree line. In Fennoscandia, the tree line is formed by mountain birch (Betula pubescens ssp. czerepanovii), which is browsed by reindeer (Rangifer tarandus) in the summer. Studies have shown that browsing by both reindeer and sheep may affect both the position and the structure of the tree line (Oksanen et al. 1995). In areas with summer browsing, the tree line was typically abrupt and consisted of large trees, while the tree line was of the transitional type where browsing was less intensive. Some birch saplings are usually found at several hundred metres above the tree line, but typically they are all browsed and resemble coppiced trees, except in sheltered positions where browsing is not possible (Oksanen et al. 1995; Hofgaard 1997) or where reindeer densities are low (Kullman 2002). Reindeer browsing has also been shown to be an important mortality factor for birch saplings in mountain birch forests defoliated by Epirrita autumnata outbreaks (Lehtonen & Heikkinen 1995).
In addition to consumption of foliage, animals may also inhibit the migration of tree lines through trampling and seed predation. Trampling of seedlings has been shown to reduce seedling survival in a wide variety of environments (den Herder et al. 2003; Li et al. 2003; Castro et al. 2004). Animal activity can also reduce seedling survival due to soil compaction (Kozlowski 1999). The restriction of migration due to seed predation results from a reduction in the number of potential migrants into the new environment. Seed predation is hypothesized to be a major limitation to regeneration of Scots pine (Pinus sylvestris) at Spanish tree lines (Castro et al. 1999), and analogs from other ecotones indicate that seed predation is a factor in limiting migration (Garcia 2001; Muñoz & Arroyo 2002).
Effects of herbivory on tree lines are seen worldwide. In addition to the above-mentioned grazing by livestock in the Alps (Ellenberg 1988), at the other end of the world, herbivory by guanaco (Lama guanicoe) affects seedling growth at Nothofagus tree lines on Tierra del Fuego (Cuevas 2002). Between 29 and 57% of seedlings in the alpine zone showed signs of browsing (Cuevas 2002), and the browsing pressure was even higher near the tree line where herbivory almost completely inhibited regeneration (Rebertus et al. 1997). It is even suggested that guanaco herbivory prevents an advance of the tree line as a response to higher temperatures (Cuevas 2002).
The negative effects of herbivory on tree-line migration, however, are mediated by other potentially positive effects. For example, although trampling by large ungulates may have a negative impact on seedlings in some situations, trampling may also reduce existing vegetation cover and thereby increase soil temperatures and facilitate vascular plant growth (e.g. van der Wal & Brooker 2004). At tree lines low soil temperatures are hypothesized to be in part responsible for setting the tree-line location worldwide (Körner 1998). Therefore, increased soil temperatures beyond the tree line could result in upslope tree-line migration.
Competition between existing vegetation and invading tree seedlings has been shown to limit the ability of trees to invade beyond the current tree line (Hobbie et al. 1998; Moir et al. 1999; Dullinger et al. 2003) and experimental tests have shown that simulated grazing can reduce the intensity of this competition, thereby allowing tree seedlings to establish (Castro et al. 2002). Therefore, if the upslope field layer plants are relatively more palatable as forage, the presence of herbivores may have a net positive effect on tree migration.
Herbivores may also promote upslope migration at tree line through their role as seed dispersers. Tree lines dominated by species that are wind dispersed would not be affected by herbivores in this way, but ecto- and endo-zoochorous species would benefit from herbivore activity. Birds are the primary dispersers of seeds from tree-line trees in this regard (Tranquillini 1979).
To incorporate the positive and negative impacts of herbivory on the migration potential of the tree line, we propose the following conceptual model in which three general patterns of tree-line response to herbivory pressure become evident (Fig. 3). Type I responses are characteristic of systems where the relative palatability of field layer and arboreal vegetation is shifted towards herbivores consuming the field layer vegetation. Low herbivory pressure in these systems fails to release the invading tree seedlings from competition and therefore migration rates are low. One example of this type of system would be the Pinus sylvestris tree lines that experience reindeer herbivory and are locally dominant at some Scandinavian locations. Type II tree lines occur when the foliage of the arboreal vegetation is preferred forage for the herbivores and migration rates are therefore highest under low herbivory pressure. Scandinavian mountain birch tree lines may be an example of this type. The shapes of response for Types I and II tree lines are modified slightly due to the impacts of trampling, seed dispersal and seed predation but where these effects are high relative to the foliage consumption of the two adjacent vegetation types, Type III tree lines are observed, characterized by the highest migration rates occurring under moderate herbivory pressure. This type of tree line is most likely to be found where there is not a single dominant herbivore, but where multiple animal species have a significant impact on the system or where the negative impacts of arboreal foliage consumption are balanced by the positive effects of the consumption of upslope vegetation leading to preparation of establishment sites.
To illustrate the potential effects of herbivores we provide the following example from northern Sweden where small birch trees are a favoured forage for reindeer during the early summer months. According to our conceptual model (Fig. 3), mountain birch tree lines should exhibit Type II characteristics, and sites that experience different herbivore densities should have different size structures of trees. During the summer of 2003, we collected diameter data at tree-line sites across northern Sweden, and here we present data from two of those sites that have different levels of herbivory. The sites are similar with regard to slope and elevation (when adjusted for latitude) but differ in aspect and reindeer use. The Tjidtjak site is on a south-west facing slope and the Kärkevagge site faces north. Reindeer have been shown to avoid areas in the vicinity of hiking trails and major roads (Nellemann et al. 2000; Vistnes et al. 2004) and as the Kärkevagge site is close to both, the number of reindeer that encounter this site is likely to be low. The Tjidtjak site, on the other hand, is on a hillside within the calving grounds for the local reindeer herding district and the proximity to herding cottages indicates that the presence of reindeer in the area is high. The two sites are therefore likely to experience, respectively, low and high levels of reindeer browsing.
Due to recent warming at tree lines in northern Sweden (Kullman 1990, 2002) we expect that under low levels of herbivory there should be large numbers of stems in the smallest diameter classes. This is indeed the case at the Kärkevagge site, where approximately 60% of the stems were < 1 cm in diameter at the base (Fig. 4). Sites with higher levels of herbivory should exhibit a much lower proportion of stems for the smallest diameter classes. The Tjidtjak site does have a number of small stems, thereby indicating that reproduction is occurring, but the proportion of stems in the smallest diameter classes at the site is considerably lower than at the Kärkevagge site. Although our sites do differ in aspect, moisture limitations are not important at these tree-line sites and the increased number of small stems should not therefore be related to aspect. We hypothesize that the difference in size class structure between the two sites is the result of reindeer herbivory.
The data that we present here are not intended to be conclusive. Considerably more sites need to be evaluated to allow for a thorough testing of the hypothesis and to control for other environmental conditions, such as elevation, slope and aspect. The data do, however, suggest that herbivory needs to be considered as a potentially significant local control on the position and abruptness of alpine tree lines in northern Sweden, and perhaps also for other tree-line sites. Furthermore, it thus seems likely that even fairly moderate browsing may be able to slow, or even halt, the upward spread of the tree line by feeding on seedlings and saplings that have already been weakened by adverse climatic conditions and by competition from field layer plants. In the generally unfavourable climate for trees at alpine tree lines, plant species are slow growing and long-lived and even infrequent disturbances may have long-lasting effects (Slatyer & Noble 1992; Hofgaard 1997).
The conceptual model presented here can act as a guide for future studies of tree-line responses to herbivory under a changing climate and should be applicable to other ecotones as well. To improve our understanding of tree-line response to climate change under varying local conditions we suggest that a combination of field- and model-based studies should be undertaken. We suggest two specific hypotheses that should be tested. (i) The relative palatability of forage available at tree line influences the response of arboreal vegetation to climate change leading to either a Type I or a Type II response. (ii) Tree-line environments subjected to multiple herbivores with different modes of interaction with the vegetation (both arboreal and non-arboreal) will respond more quickly to an improved climate at intermediate levels of herbivore activity (Type III). To conclude, we suggest that herbivory should be considered as an additional hypothesis when explanations for the equivocal response of tree lines to climate change are sought.