1We used forest stand history reconstruction to infer the relative roles of disturbance and climate warming on the population dynamics of Nothofagus menziesii (silver beech) dominated tree lines in north Westland, South Island, New Zealand.
2Stem recruitment in tree line forests over the last 300 years has been episodic, has tended to occur in small, scattered patches, and has been dominated by the production of new stems from existing trees. Pulses of stem recruitment also coincide with episodes of abrupt decline in the radial growth of established trees. These patterns suggest that infrequent natural disturbances form localized canopy openings, damage trees that survive the event, and initiate the establishment of new trees and the production of new stems by surviving trees that fill these openings.
3Climate warming in New Zealand since 1950 has had little effect on the recruitment of Nothofagus close to the tree line. There is a large seedling pool within the tree line forests, but recruitment will probably require a disturbance-related canopy opening.
4Natural disturbances drive the population dynamics of Nothofagus tree lines and may modify their response to climate warming. Unlike many Northern Hemisphere tree lines, there has been no recent upward movement of the tree line or increase in seedling establishment. This difference could reflect the greater importance of natural disturbance for recruitment in the abrupt closed-canopy tree lines formed by light-demanding Nothofagus species in the Southern Hemisphere. However, given the ubiquity of disturbance effects in low-altitude forests in both hemispheres, future studies need to consider their role when investigating any tree line response to climate warming.
Global temperatures have increased by up to 0.6 °C over the past century and temperatures are predicted to rise a further 1.5–4.5 °C by 2100 ad (Houghton et al. 1996). Faced with increasing temperatures, ecologists and conservation managers need to understand the effects of global warming on species, communities and ecosystems. Tree lines occur at the point where carbon gained during the summer is equal to, or only just greater than, that lost due to respiration or tissue mortality through lack of hardening to withstand winter temperatures (Wardle 1965). Because summer temperatures and the length of the growing season strongly influence carbon production and shoot hardening, higher temperatures should increase tree productivity and survival at the tree line (Brubaker 1986).
Tree line responses to warming over the last century have been detected in many parts of the Northern Hemisphere (see Kullman 1990; Rochefort et al. 1994 for reviews) with seedling invasion of subalpine meadows and recent establishment above the current tree line. Other studies have found upsurges in establishment within tree line forests (e.g. Szeicz & MacDonald 1995; Taylor 1995). The relatively rapid response of Northern Hemisphere tree lines to climate warming has reinforced the belief that temperature is the dominant control on tree line dynamics.
This may however, overlook the potential importance of natural disturbance events. Natural disturbances drive the regeneration dynamics of most closed-canopy forests by creating opportunities for the establishment of new individuals through canopy opening (Pickett & White 1985) and have been shown to play a role in the regeneration dynamics of tree line forests in some areas (e.g. Veblen et al. 1981). Increases in tree establishment and changes in tree line position could, therefore, primarily be a response to disturbance, rather than climate warming (Slayter & Noble 1992). Fire, for example, can control the position of North American tree lines (Vale 1981; Baker 1992), and regeneration above the current tree line is often a recovery to the tree line position before the fire (Shankman 1984). At or below the present forest tree line, disturbances, ranging from small-scale tree-fall gaps (Taylor et al. 1996) to catastrophic forest destruction by earthquakes (Veblen 1985), lead to periods of increased tree establishment. Events following an appropriately timed disturbance could mimic the expected responses to climate warming. Disturbances could also modify the response of tree line populations to climate change. Light-demanding tree species that rely on disturbances for recruitment may be unable to respond to climate warming unless it coincides with canopy opening (Brubaker 1986), whereas the response of tree line forests to warming may be enhanced during periods of disturbance (Overpeck et al. 1990).
There are few studies examining the response of Southern Hemisphere tree lines to climate warming. Warming in New Zealand has occurred mostly in the last 50 years, with an increase of 1.0 °C in average summer temperatures since 1950 (Salinger et al. 1993). Since the limit of New Zealand beech tree lines reflects both summer temperatures and the length of the growing season (Wardle 1965) we would expect that changes in position and density would be similar to those in the Northern Hemisphere. In fact, Wardle & Coleman (1992) suggest that Nothofagus tree lines in New Zealand have responded to the recent warming with a minor tree line advance (c. 10 m), however, any interpretation of tree line dynamics in response to climate warming should at least consider the role of natural disturbance (e.g. Veblen 1985; Baker 1992). Our aim was to examine the relative roles of natural disturbance and climate warming in influencing the population dynamics of Nothofagus tree lines in South Island, New Zealand.
Warming is likely to have affected New Zealand forests only within the last 50 years, but we aimed to reconstruct the population dynamics of Nothofagus tree line forests further back in time by ageing trees and dating periods of past tree establishment. An understanding of tree establishment patterns over the past 200–300 years provides a context within which to interpret recent changes as inferred from the present patterns of tree regeneration. To achieve this we examined: (i) the temporal and spatial patterns of tree establishment in current tree line forests; (ii) evidence for recent establishment above the tree line; and (iii) tree growth patterns.
Nothofagus species commonly form the tree line in the Southern Hemisphere, including New Zealand, where Nothofagus menziesii (Hook. f.) Oerst (silver beech) is dominant on the wetter western mountains and N. solandri var. cliffortioides (Hook. f.) Poole (mountain beech) is dominant in drier eastern areas (Ogden et al. 1996). Nothofagus tree lines generally vary little in their local altitude, and are abrupt and floristically simple (Fig. 1), providing a relatively simple system in which to investigate climate and disturbance influences. Our study was carried out in the Rahu Saddle area (42°19′ S, 172°07′ E), north Westland, South Island, New Zealand. Here tree lines dominated by silver beech are typical of Nothofagus tree lines in the Southern Hemisphere in general, where closed-canopy forest extends right up to an abrupt tree limit (Wardle 1965).
The climate of New Zealand, and the South Island in particular, is dominated by a prevailing west to south-west airflow (Salinger 1988). These winds are moisture-laden due to New Zealand's oceanic location. The main axial mountain range in South Island runs from the south-west to the north-east, creating a significant barrier to airflow and generating a steep rainfall gradient from the west to the east coast (Salinger 1988). The nearest long-term temperature station, Hokitika (c. 120 km south-west of the study site), shows slight seasonal variation in temperature, with February the warmest month (15.6 °C) and July the coldest (7.1 °C). The total annual precipitation at Reefton (the nearest rainfall station, 40 km west of the study area and at 700 m altitude) is 1930 mm. Precipitation at the tree line is likely to be higher because of the greater elevation (c. 1200 m), although much of this will fall as snow during winter.
We selected three sites near Rahu Saddle and established seven 0.05 ha plots at these sites to investigate tree line responses to climate warming and disturbance. Three north-east facing plots were established at Mt Haast (Ht1–3, 42°19′ S, 172°05′ E), three south-facing plots at Rahu Valley (Rv1–3, 42°17′30″ S, 172°08′ E) and one south-facing plot at Rahu Spur (Rs1, 42°18′ S, 172°07′ E, Table 1). We subjectively located each plot in areas that avoided obvious avalanche tracks or areas with large canopy openings that indicated recent major disturbance. Each plot was 10 × 50 m, with the longer side extending downslope at 90° from the tree line. Tree line altitude, aspect and slope were recorded for each plot (Table 1).
Table 1. Number of silver and mountain beech seedlings (< 1 cm diameter), saplings (1–10 cm diameter) and stems (≥ 10 cm diameter), and total basal area (b.a.) of each species in each 0.05 ha tree line plot
Plot started with a 4° slope, and after c. 25 m downslope changed to a 40° slope.
Interpretation of tree population dynamics requires information on the temporal patterns of tree establishment, often inferred from age and size distributions (Henry & Swan 1974; Stewart & Rose 1990). A distinct pulse of tree establishment evident in an age-class distribution is often an indication of regeneration in response to canopy-opening disturbance (Oliver & Stephens 1977; Stewart 1986). In contrast, we might expect a sustained upsurge in tree establishment, synchronous with warmer conditions, in response to recent climate warming.
In the tree line forests that we studied, multistemmed individuals that branched just above the forest floor were common. We defined a tree as any individual with one or more stems ≥ 10 cm diameter at 50 cm above the forest floor. We recorded the species, number and diameter of all stems ≥ 10 cm diameter at 50 cm for each tree rooted within the plot. We defined a sapling as any individual with all stems < 10 cm diameter at 50 cm, but with one or more stems ≥ 1 cm diameter at the forest-floor level. Sapling diameters were measured at forest-floor level, rather than at 50 cm, because sapling stems often branched extensively just above the forest floor. We defined a seedling as any individual < 1 cm diameter at forest-floor level. All seedlings in each plot were counted. We constructed size-class frequency distributions from the above data. Because silver beech was much more abundant in the tree line plots than mountain beech (Table 1), we only present the results for silver beech.
To construct age-class frequency distributions and to develop tree-ring growth chronologies, we cored all stems ≥ 10 cm diameter, at a height of 50 cm above the forest floor. We usually extracted one core per stem, but some stems were re-cored if the first core missed the pith. Cores were mounted and sanded using progressively finer grades of sandpaper until the growth rings were clearly visible. Only cores which either passed through the stem pith or close by, so that the arcs of the innermost rings were visible (‘arc’ cores), were retained. Growth rings were counted under a binocular microscope and ring widths were measured using a Henson bench, binocular microscope and the trims measuring program (Madera Software 1988). For the ‘arc’ cores, the number of missing rings was estimated using the geometric model of Duncan (1989) which estimates the distance from the largest visible arc to the pith, and a mean cumulative radial growth curve to estimate the growth rate of the missing rings, using only those cores that passed through the pith (see Villalba & Veblen 1997).
The time to reach coring height (50 cm) was estimated from linear regressions of forest-floor level age (which ranged from 10 to 80 years) against total height (which ranged from 0.25 m to 2.30 m) for silver beech saplings collected adjacent to, but outside, each plot. Due to small sample sizes the collections from around each plot were combined by site so that the sample sizes were: Mt Haast, n = 31, Rahu Valley, n = 26 and Rahu Spur, n = 26. Saplings in large canopy gaps or under a dense canopy were avoided to obtain an estimate of the time to reach coring height for saplings growing in small openings. Twenty-two, 15 and 19 years were added to the ages of silver beech stems at Mt Haast (sapling age–height regression, r2 = 0.55), Rahu Valley (r2 = 0.68) and Rahu Spur (r2 = 0.63), respectively. Because of the variation in tree growth rates to coring height, variation in the height at which trees were cored (due to low-growing branches), and error in estimating the number of rings missed in cores that missed the pith, we grouped tree and stem ages into 20-year age classes. The age of multistemmed trees was taken to be the age of the oldest stem of that tree.
Stems < 10 cm diameter were not cored and their ages could not be reliably estimated from age-diameter regressions for stems ≥ 10 cm diameter because the spread of ages around a particular diameter was large (e.g. 10 cm diameter stems ranged in age from 75 to 150 years). Instead, we inferred the temporal patterns in establishment of stems < 10 cm diameter from size-class distributions.
We used the New Zealand national temperature series, compiled from seven climate stations around New Zealand (Salinger 1980), to relate tree-line establishment patterns to temperature trends. Because tree lines are likely to respond only to changes in summer temperature, we used the average of the mean monthly temperatures from December to February to produce an average summer series. In using the New Zealand temperature series we assume that the recent increase in the national average summer temperature (1.0 °C since 1950) has also occurred at the tree line. This is likely, as warm and cool periods have been synchronous throughout New Zealand (Salinger 1980).
SPATIAL PATTERNS OF TREE LINE FORESTS
Positive spatial autocorrelation at small distances in tree or stem ages would suggest that trees or stems were established in even-aged patches, a pattern that could reflect establishment in localized canopy openings in response to disturbance (Duncan & Stewart 1991; Frelich & Lorimer 1991). Alternatively, tree ages may not be patchy but show a directional trend, such as increasing age below the tree line, which might indicate a progressive increase in tree line elevation in response to climate warming.
To examine the spatial patterns of tree establishment, we recorded the position (as x, y co-ordinates to the nearest 0.1 m) of all stems ≥ 10 cm diameter at 50 cm above the forest floor, and of all seedlings and saplings at forest-floor level in each plot. The ages of silver beech trees and stems ≥ 10 cm diameter in each plot were analysed for spatial autocorrelation following the methods in Duncan & Stewart (1991), using Moran's I coefficient (Moran 1950). Correlograms were constructed to examine how the degree of autocorrelation in tree and stem ages varied with increasing distance, and the type of spatial structure in tree or stem ages was inferred from the correlogram shape (Legendre & Fortin 1989). Each correlogram was tested for overall significance using a global test at the α = 0.05 level. When correlograms indicated significant spatial structure, we used agglomerative hierarchical clustering (upgma in the package patn, Belbin 1989) to identify even-aged patches of trees or stems in each plot by grouping trees or stems on the basis of their similarity in age and spatial location (see Duncan & Stewart 1991). Groups identified from the cluster analysis were only accepted as even-aged patches if they were spatially discrete, had a relatively even-aged structure, and had a different age structure from neighbouring patches (Duncan & Stewart 1991).
In each plot, the spatial distribution (either clumped, random or regular) of silver beech seedlings and saplings was assessed using the K(t) function (Ripley 1977; with the edge correction given in Diggle 1983), at 0.5 m intervals up to half the length of the shortest side of each plot. Using Monte Carlo simulation, approximate 99% confidence envelopes were constructed from the high and low values obtained from 1000 simulations of a random point process. Positive or negative values of K(t) that fall outside these confidence envelopes indicate a significantly clumped or regular stem distribution, respectively.
We tested for a directional trend in recent recruitment by regressing the size of seedlings and saplings against distance below the tree line. A trend towards smaller (and presumably younger) stems at the tree line could indicate a progressive advance of the tree line and an active zone of regeneration at the current boundary, a possible response to recent climate warming.
GROWTH RELEASES AND SUPPRESSIONS
Temporal patchiness in tree establishment is often a response to past disturbance (Stewart 1986; Taylor et al. 1996) but, at the tree line, periods of increased establishment could also be a response to climate warming. To help interpret the patterns of tree establishment, we searched for abrupt, short-term (< 5 years) changes in the radial growth of trees. Rapid increases in radial growth are often a response to canopy opening, where the death of a canopy tree leads to the ‘release’ of adjacent canopy or understorey trees (Lorimer 1980). Alternatively, an abrupt decline in tree growth (suppression) can be a response to a severe event in which trees were damaged and growth restricted (Jacoby et al. 1997).
To detect abrupt growth changes, we applied growth suppression and release enhancement filters (see Kitzberger et al. 1995) to chronologies developed for each plot using the ring-width measurements of individual silver beech stems. Ring-width series from individual stems were cross-dated using a combination of visual cross-dating and cofecha (Holmes 1983). Chronologies were standardized (using arstan; Cook 1985) to remove the biological growth trend using a negative exponential curve, or a negative linear regression, or if both of these failed, a horizontal line was fitted which does not remove any growth trend. Autocorrelation, where growth in previous years affects subsequent growth, was not removed because this would limit our ability to detect disturbance-related suppressions or releases. Chronologies were truncated at the year where the number of ring-width series fell below five. The suppression/release filters are designed to remove most annual growth variability, leaving only periods of strong suppression or release (Kitzberger et al. 1995). For each year, the indices were calculated based on the growth of the preceding and following 2 years, with strong suppressions or releases producing larger spikes.
RECENT ESTABLISHMENT ABOVE THE TREE LINE
To determine if the tree line altitude is increasing, we extended each plot directly above the current tree line in a 10-m wide transect and searched upslope for trees until no further trees were found. For any trees present, we recorded the species, diameter (cm) at the base if > 1 cm (otherwise recorded as seedling; stems were measured at their bases as most were prostrate and did not grow taller than 50 cm), and distance from the tree line (m). A core was extracted from any stem > 5 cm diameter near ground level.
Current tree line in the plots ranged from 1140 m a.s.l. (metres above sea level) to 1255 m a.s.l. (Table 1). Silver beech formed the tree line and was the dominant canopy species in all seven plots. Mountain beech was present in high numbers in plot Ht3, low numbers in plot Rs1, and as one seedling in plot Ht1.
OVERALL AGE AND SIZE STRUCTURES
The combined age data for all seven tree line plots suggest that silver beech stems ≥ 10 cm diameter have been recruited episodically in the last 300 years. There was a general increase in stem numbers starting around 1740 ad, but there was also a distinct pulse of stem recruitment over the period 1880–1920 (Fig. 2a). In Fig. 2(a), the lack of stems that were recruited after 1920 is partly an artifact of not coring young stems < 10 cm diameter. Nevertheless, the size distribution of all stems also implies a distinct recruitment pulse, with fewer saplings 1–10 cm diameter than stems 10–20 cm diameter (Fig. 2b). Hence it appears that there was an upsurge in stem recruitment in the period 1880–1920 that has not been sustained in recent decades. Significantly, the main peak in this recruitment pulse occurs during a period when the average summer temperature in New Zealand was low (1890–1915; Fig. 2a).
INDIVIDUAL PLOT AGE AND SIZE STRUCTURES, AND ABRUPT GROWTH CHANGES
A period of high stem recruitment in the period 1880–1920 is evident in the age structure of all individual plots (Fig. 3). In two plots (Ht2 and Rs1) this is evident as a discrete pulse of stem recruitment, but in other plots it appears to be part of a longer period of stem recruitment, sometimes with a distinct pulse initiated earlier than 1880 (e.g. 1860 in Ht3 and Rv3).
At least three widespread episodes of radial growth suppression occurred between 1880 and 1920 (indicated by arrows on Fig. 3 and Fig. 4). The most consistent, and usually the largest of these suppressions, started in 1898 (Rv1–3) or 1899 (Ht1, Ht3 and Rs1). The two remaining episodes are evident as one or two suppressions in the late 1870s to early 1880s, the most consistent of which started in 1881 (Rv1–3, Rs1 and less noticeably in Ht3), and as a group of several suppressions in the late 1910s, the most consistent of which started in 1915 in all chronologies except Rs1 (Fig. 4). Although periods of growth release are evident in most plots (Fig. 5), few are synchronous between plots, suggesting that there were few widespread episodes of growth release. The most consistent growth release started in 1950 (Rv2 and Rv3) or 1951 (Ht1–3, Rv1 and Rs1). The only other consistent growth release occurred in 1990 (Ht2–3, Rv1–3).
Although there was a major pulse of stem recruitment after 1880, earlier pulses in recruitment are also evident in some plots (Fig. 3). In Ht1, Ht3 and Rv2, for example, there was a smaller pulse of stem recruitment between 1820 and 1860, and in plots Rv3 and Rs1 there was a small recruitment pulse between 1800 and 1840. In other plots older pulses could have occurred, but they may have been destroyed by more recent disturbances, such as those which initiated the recruitment pulse starting around 1880. Although most of the chronologies did not extend far back enough to determine whether the older recruitment pulses are also associated with suppression events, Rv2 shows an 1820 recruitment pulse following major growth suppression around 1815 (Fig. 4e).
Sapling diameter distributions suggest that recent recruitment of silver beech has continued to be episodic (Fig. 6). Plots Rv1 and Rv2 show a peak at 6–10 cm diameter, in addition to a larger peak 3–5 cm seen in all plots, which contrasts with the few saplings 1–2 cm diameter and the large number of seedlings in all plots. This implies little recent recruitment from what may represent a suppressed ‘seedling pool’ (Wardle 1984).
SPATIAL PATTERNS OF ESTABLISHMENT
Tree ages (taken as the age of the oldest stem) showed little evidence of spatial structure. Only the correlograms for plots Ht2 and Rv1 were globally significant at the 0.05 level (not shown), but their shape did not suggest a spatial structure consistent with establishment in even-aged patches.
In contrast, the ages of individual stems were significantly spatially structured. The correlograms for all plots, except Ht2 (where the stem ages were all very similar), were globally significant and all showed significant positive autocorrelation in the smallest distance classes (up to 6 m), with a switch to significant negative autocorrelation in larger distance classes (9 m up to 24 m). These patterns are consistent with stem establishment in even-aged patches, with significant positive autocorrelation in small distance classes reflecting the clumping of stems of similar age within patches, and significant negative autocorrelation in larger distance classes reflecting the separation between patches of different age.
The cluster analysis of stem ages and locations failed to distinguish obvious, large even-aged patches. Rather it appeared that stems were grouped in small (2–4 m diameter) patches of similar age. Occasionally these consisted entirely of stems from a single tree, suggesting that a tree had established and rapidly produced many stems to fill an available opening. More often, however, stems came from several trees, suggesting that many trees produced stems simultaneously to fill an adjacent canopy opening. Because of the small size of even-aged patches, coupled with the ability of trees to produce many stems, individual trees often had stems in several different age groups, probably reflecting responses to multiple canopy openings. Seedlings and saplings were significantly clumped at distances up to 5 m in all plots (not shown), consistent with recruitment into small canopy openings.
There is little indication of an active zone of regeneration at the tree line boundary. The only significant relationships between sapling diameter and distance below tree line were found for plots Rv2 and Rv3 (r2 = 0.50, P = 0.024 and r2 = 0.60, P = 0.001, respectively), but these relationships were negative, meaning that larger stems were found closer to the tree line.
RECENT ESTABLISHMENT ABOVE THE TREE LINE
Silver beech had established above the tree line in only three plots: Ht1, Ht2 and Rv3 (Table 2), with most establishment in Ht2. Most individuals had established within a few metres of the current tree line, although one individual was found 20 m above the tree line at Rv3. Only one stem in each plot was large enough to be cored and their ages suggest that establishment above the tree line has occurred within the last 30–40 years (Table 2; the ages are not adjusted for years to reach coring height).
Table 2. Establishment of silver beech stems in 10 m wide transects above the tree line plots
Average diameter in cm (range)
Maximum distance above tree line (m) (average)
Highest alt. above tree line (m a.s.l.)
Age (diameter in cm)
n.r., not recorded.
3.9 (< 1–9.8)
DISTURBANCE INFLUENCES ON TREE-LINE POPULATION DYNAMICS
Two lines of evidence suggest that natural disturbance has a major influence on Nothofagus population dynamics at the tree line in our study area. Recruitment has been episodic, which is a feature of disturbance-affected forests (Oliver & Stephens 1977; Stewart 1986), and episodes of growth suppression often coincided with upsurges in stem recruitment. In particular, the largest growth suppression, evident in all plots and starting in 1898–1900, coincided with the peak in stem recruitment across all plots in 1880–1920. We suggest this is most likely caused by natural disturbances that formed small canopy openings through stem breakage, thus damaging the surviving trees while allowing the recruitment of new stems into the openings. Because the openings were generally small and many trees survived the disturbance, the major response was the production of new stems by surviving trees to fill the canopy openings, leading to patches of even-aged stems. The exception was plot Ht2 where it appears that a catastrophic disturbance around 1860 initiated widespread synchronous establishment of trees throughout the plot.
Although the major growth suppression starting in 1898 occurred during a period of low summer temperatures in New Zealand, it is unlikely that this is a temperature response because the suppression was no longer evident when temperatures reached their lowest (1903–04). Rather, the suppression is most likely due to a severe south-easterly windstorm recorded from the adjacent Maruia Valley in 1898 (Foster 1931; Conway 1952). This gale coincided precisely with the start of the suppression at our tree line sites and with an episode of small-scale gap formation and growth release in Nothofagus fusca (red beech) forests (Stewart et al. 1991). Because a delay of 1–2 years before the onset of growth suppression following a disturbance event is common (Jacoby et al. 1997), the growth suppressions starting 1898 and 1899 in all plots are probably synchronous and almost certainly reflect this windstorm. Other pulses of stem recruitment and growth suppression in the tree line plots (Figs 3 and 4) most likely reflect the impact of similar disturbances.
Overall, our results suggest that natural disturbances, most probably from windstorms or possibly snow damage, have been a major factor in initiating stem and tree recruitment in Nothofagus tree lines in our study area over the last 100–150 years, and probably back as far as 300 years. Previous research has also suggested that disturbance is important for forest regeneration at high altitudes in New Zealand (Allen & Wardle 1984), North America (Shankman 1984) and China (Taylor et al. 1996). In particular, Nothofagus-dominated tree line forests in parts of South America show a very similar regeneration pattern to our sites, with forests dominated by a mosaic of small even-aged patches (Veblen et al. 1981) and episodic regeneration (Cuevas in press). Recruitment of Nothofagus tree lines may also be limited by poor seed germination and seedling survival (see Cuevas 2000).
CLIMATE WARMING AT THE TREE LINE
Tree line position is not undergoing a substantial advance, with only limited localized recruitment above the present forest limit (mostly within 10 m), despite the 1.0 °C increase in New Zealand average summer temperature since 1950 (Salinger et al. 1993). In contrast, recent establishment above current tree limits in response to climate warming has been found in many parts of the Northern Hemisphere (e.g. Daly & Shankman 1985; Kullman 1991; Luckman & Kavanagh 1998). There is a large number of seedlings above the current tree line in plot Ht2. However, this plot is not typical since most stems and trees below were recruited as a distinct pulse after 1860, suggesting re-colonization following a stand-destroying disturbance, most likely an avalanche, and the tree line may be returning to a pre-avalanche state, as seen with fire-affected tree lines (e.g. Shankman 1984).
Wardle & Coleman (1992) also found only a limited advance (c. 10 m) of silver beech tree lines at other sites in New Zealand, which they attributed to warming. Given the presence of trees > 100 years old at the tree line boundary, the low numbers of seedlings found above the tree line and their scattered distribution in our study, we favour their alternative interpretation that limited recruitment above the tree line represents a permanent low-density population with continuous turnover (see also Szeicz & MacDonald 1995). In contrast, within 10 m of Chilean tree lines there can be relatively high numbers of Nothofagus pumilio seedlings, but the restriction of seed dispersal mostly to within 20 m of the tree line and poor seed viability above the tree line (Cuevas 2000) suggests that establishment beyond 10 m sufficient to create a new tree line is unlikely.
There has been no recent, continuous upsurge in Nothofagus seedling establishment in the tree line forests we studied; rather, recruitment continues to be episodic. This is in contrast to the many Northern Hemisphere studies which document increased seedling establishment below the tree line in response to climate warming (e.g. Scott et al. 1987; Magee & Antos 1992; Szeicz & MacDonald 1995; Taylor 1995; Rochefort & Peterson 1996). A recent study in South America (Daniels 2000) also found no increase in regeneration within Nothofagus pumilio tree line forests despite higher temperatures since 1970, an outcome that may be a consequence of the overriding importance of moisture availability for successful N. pumilio establishment. We suggest that a lack of response to climate warming below the tree line in the Nothofagus forests that we studied is a consequence of the overriding importance of natural disturbance for stem recruitment. Regeneration of Nothofagus in these tree line forests appears to depend on the creation of canopy openings and is therefore unlikely in the absence of disturbance.
Our results may reflect a general difference between the Hemispheres in tree line structure and physiognomy, and potentially the role of disturbance in driving regeneration processes at the tree line. Southern Hemisphere tree lines are often dominated by light-demanding species such as Nothofagus, which form a tall, closed-canopy forest right up to an abrupt limit (Wardle 1965). At least in our study area, disturbance is required for substantial recruitment within these closed-canopy forests and a response to climate warming is unlikely unless warming coincides with canopy disturbance. In contrast, tree lines in the Northern Hemisphere are typically dominated by coniferous species and often grade from closed-canopy forest to small clumps of trees to scattered individuals until the tree limit is finally reached (Kullman 1990). This more open structure may mean that coniferous tree lines are less dependent on canopy opening by disturbance for successful establishment, so that recruitment responds more rapidly to changes in temperature.
Nevertheless, recent work from Switzerland has suggested that the subalpine limit of Pinus sylvestris has shown little response to climate warming because seedling establishment relies on disturbance, mostly from fire (Hättenschwiler & Körner 1995). This suggests that there are tree lines in the Northern Hemisphere where disturbance could affect responses to climate warming in a similar way to that found in our study. Furthermore, the open coniferous tree lines of the Northern Hemisphere are not typical of forests below the tree line in both hemispheres, in which disturbance is usually important for forest regeneration. We suggest that understanding the likely impact of climate warming on the dynamics of forests in general will require careful consideration of the interactions between climate warming and disturbance impacts.
This research was funded by Lincoln University, the Lotteries Commission, the Robert C. Bruce Trust, and the Canterbury Botanical Society. Financial support for the first author was provided by a Lincoln University Doctoral Scholarship. We would like to thank Amanda Ridley, Bert Borger, Maartje de Deugd, Nancy Willems, Gillian Vaughan, Pavla Honzícková and Gillian Dennis for their tireless help in the field and two anonymous referees for their helpful comments. We also thank the Department of Conservation for permission to core trees in the study plots.