Tree Growth and Competition in a Betula platyphylla–Larix cajanderi Post-fire Forest in Central Kamchatka

Abstract

• Background and Aims Fire is the dominant disturbance in central Kamchatka boreal forests, yet patterns and mechanisms of stand recovery have not been investigated.

• Methods Measurements were made of 1433 stems ≥1·3 m height and annual radial increments of 225 randomly selected trees in a 0·4-ha plot of a 53-year-old fire-origin mixed-species stand to examine the spatio-temporal variation in establishment, growth, size inequality and the mode of competition among individual trees. Growth variations were related to tree size, age and local interference with neighbours.

• Key Results Betula platyphylla formed the main canopy following a fire in 1947, with Larix cajanderi and Pinus pumila progressively reinvading the lower tree and shrub stratum. Most B. platyphylla originated from sprouts in small patches (polycormons) during the first 15 post-fire years. Betula platyphylla had normal distributions of diameter and age classes, but negatively skewed height distribution, as expected from shade-intolerant, pioneer species. Larix cajanderi had fewer tall and many short individuals. The smaller and younger B. platyphylla grew disproportionately more in diameter than larger trees from 1950 to 1975, and hence stem size inequalities decreased. The reverse trend was observed from 1995 to 2000: larger trees grew more, indicating an increasing asymmetry of competition for light. Betula platyphylla had steady diameter growth in the first 25 post-fire years, after which the growth declined in smaller trees. Neighbourhood analysis showed that the decline resulted from increased competition from taller neighbours.

• Conclusions The observed growth patterns suggest that mode of interactions altered during stand development from early stages of weak competition for soil resources released by fire to later stages of asymmetric competition for light. Asymmetric crown competition started later than reported in other studies, which can be attributed to the lower stem density leaving much space for individual growth, greater relative importance of below-ground competition in this site of nutrient-poor volcanic soil, and the vegetative origin of B. platyphylla. Larix cajanderi growing under B. platyphylla had steady diameter growth during the first 20 years, after which growth declined. It is suggested that early succession fits the tolerance model of succession, while inhibition dominates in later stages.

INTRODUCTION

Fire is the dominant disturbance in many boreal forests, where it profoundly alters the composition and structure of plant communities (Shugart et al., 1992). The last three decades have seen major advances in description and modelling of the development of post-fire tree populations. There is a large body of empirical studies that has related patterns of forest recovery to the nature of the fire, i.e. its intensity, duration and return period (Bergeron and Dansereau, 1993), characteristics of the pre-fire community (Halpern, 1988), local environmental conditions (Sirois, 1993; Little et al., 1994), and the reproductive characteristics of the available species (Wein and MacLean, 1983). Fewer studies have examined how the pattern of forest recovery depends on growth and local interactions of pioneer trees early in stand development (Lieffers et al., 1996). The present study examines the spatio-temporal patterns in establishment, growth and survival of Betula platyphylla and Larix cajanderi in a young post-fire stand in central Kamchatka in relation to local interference among neighbours.

In the coniferous zone of central Kamchatka, frequent forest fires maintain a large portion of the landscape in early and mid-seral stages. The landscape is a patchy mosaic of stands at various stages of recovery, often sharply delimited by abrupt changes in tree composition and/or canopy height. These stands include species such as Japanese white birch (Betula platyphylla), poplar (Populus tremula), Manchurian alder (Alnus hirsuta), larch (Larix cajanderi), Siberian dwarf pine (Pinus pumila) and Ajan spruce (Picea ajanensis, syn. Picea jezoensis), which are a common feature in the taiga of the Russian Far East (Kolbek et al., 2003). In central Kamchatka, they occur from low to middle elevations that belong to the driest biogeoclimatic zone in the province (Kojima, 1997). Following disturbance by fire these forests usually initially become dominated by B. platyphylla and/or P. tremula, and in the prolonged absence of fire there is an increasing domination of conifers (Khomentovsky, 1998). Betula platyphylla and P. tremula are described as rapidly colonizing burned sites by a combination of good seed dispersal with frequent resprouting from pre-burn trees (Goldammer and Furyaev, 1996), followed by reinvasion of more shade-tolerant Larix cajanderi, Pinus pumila and Picea ajanensis.

General descriptions of forest successions in central Kamchatka (e.g. Krestov, 2003) have not been followed by more detailed quantitative studies on the patterns and mechanisms of species abundance and replacement after the disturbance. Efforts have been made recently to fill the gaps in the understanding of natural forest dynamics in the region (Khomentovsky, 1998; Takahashi et al., 2001). This includes gathering data on their demography, as well as monitoring the response of tree species to changes in the environment, using the system of permanent research plots (Homma et al., 2003). This study reconstructs the development of pioneer B. platyphylla stand and its progressive recolonization by coniferous species over a 50-year period following fire by combining spatial pattern data with supplementary information (age, diameter growth) about individual trees. In the search for mechanisms responsible for the replacement of tree species, attempts have been made to relate the turnover of species to their life-history strategies, especially to their mode of regeneration, growth and biotic interactions. To achieve this we examined: (a) spatial and temporal patterns of tree establishment; (b) tree growth pattern and size-hierarchy development; and (c) several potential causes of growth decline. Variations in growth of individual trees during the 50 years of stand development were related to stem size, age and local crowding. As trees respond to these factors with similar changes in growth increment, records of past influences are retained in their tree rings and can be unravelled through dendroecological techniques (Schweingruber, 1996). Studying several potential factors in the decline of B. platyphylla growth can help us understand the mechanisms that drive their pioneer stands toward a system dominated by L. cajanderi.

MATERIALS AND METHODS

Study site

Research was conducted approx. 30 km east of the village Esso in the Bystraya river lowland in the Central Kamchatka Depression. The climate is sub-continental boreal, with large intra-annual temperature variations, short warm summers and long cold winters (Martyn, 1992). According to climatic data from Esso (1940–1995), mean annual temperature is −3·1 °C, with mean for January −18·7 °C and, and for July 13·2 °C. Mean annual precipitation is 398 mm, with a 55-year range of 244–571 mm. Seasonal distribution of precipitation is approx. 19 % in winter, 12 % in spring, 42 % in summer and 27 % in autumn. April receives the lowest amount of rainfall and July the highest. The site is a well-drained fluvial terrace. The shallow soils (andisol) are derived from fluvial gravel overlain by multiple Holocene volcanic ash deposits and aeolian dust. Litter thickness ranges from 3 to 6·5 cm. The forest studied established naturally after fire burned an old-growth stand dominated by Larix cajanderi Mayr. in 1947. The numerous burnt remnants of L. cajanderi, including fallen logs and stumps with diameter up to 70 cm, were found on the forest floor during fieldwork. The post-fire stand is dominated by Betula platyphylla Sukaczev (Fig. 1), with significant components of invading L. cajanderi and Pinus pumila (Pall.) Regel., and minor components of Alnus hirsuta (Spach.) Rupr., A. fruticosa, Salic bakko and S. caprea (Table 1). The stand has a well-developed undergrowth stratum in which the most common species are Lonicera caerulea, L. chamissoi, Vaccinium vitis-idea, Linnaea borealis, Epilobium angustifolium and Ledum palustre.

Fig. 1.

A 53-year-old fire-origin Betula platyphylla-dominated stand in central Kamchatka.

Table 1.

Tree species composition, number of individuals in a 0·4-ha plot, their basal area, maximum breast height diameter, and maximum stem height

Species	Count	%	Basal area (m2)	%	Max. dbh (cm)	Max. height (m)
Betula platyphylla	1033	74·5	6·86	76·6	23·5	19·7
Larix cajanderi	216	15·6	1·25	14·0	31·8	21·3
Alnus hirsuta	35	2·5	0·41	4·6	25·6	19·2
Alnus fruticosa	31	2·2	0·31	3·5	20·7	14
Salix bakko	40	2·9	0·05	0·6	7·9	8·1
Salix caprea	32	2·3	0·07	0·8	8·9	12·8
Total	1387	100	8·95	100

Data collection

Field measurements were conducted during the summer and autumn 2000 and 2001 in a 50 m × 80 m plot (0·4 ha) selected to be uniform in stand and environmental characteristics. Each live and dead tree >1·3 m tall within the plot was marked with a numbered plastic tag, its spatial coordinates determined within a grid of 40 10 m × 10 m sub-plots, and its breast height diameter (dbh), total tree height and crown base height measured. Cores taken from 225 randomly selected trees were used to age the trees and to reconstruct their growth histories. Two cores were extracted on opposite sides of each tree 20–30 cm above the ground. The cores were dried, mounted, sanded and inspected for fire injuries, reaction wood and other aberrant features. Rings were counted from pith to bark and their widths measured to the nearest 0·01 mm with the aid of a microscope interfaced to a computer. For approx. 24 % of the cores from which pith was missing (as easily happens in trees like B. platyphylla with an asymmetric stem) the number of missing rings was estimated from the diameter of the innermost tree-ring and the average width of the five following rings. The ring-sequences were cross-dated visually using the pattern of wide and narrow rings, and verified using COFECHA, a computer-assisted-tree-ring analysis program (Holmes, 1983).

Data analysis

We used univariate Ripley's K-function and its derived variable the L-function (Ripley, 1977), with the edge correction given in Diggle (1983), to determine the type and intensity of spatial pattern of stem distribution in the plot. L-function has zero expectation when the pattern is random, and positive and negative values when plants are clumped or over-dispersed, respectively. Calculating the values of L(r) was started at a radius of 1 m and then at 1-m intervals up to 25 m, i.e. up to half the length of the shorter side of the plot. Two null hypotheses concerning the spatial stem distribution were tested. The first hypothesis was that there is no deviation from Poisson random distribution, in which case random coordinates of the same number of trees that were recorded for a given species in the plot were generated 99 times using Monte Carlo simulation, and the lowest and highest values of L(r) used to construct approx. 99 % confidence limits. The sample statistics above and below the limits indicated significantly clumped or regular stem distribution, respectively (Diggle, 1983). The second hypothesis was that of distribution of a species is independent of other species in the community. Of particular interest was to find out if the spatial distribution of L. cajanderi and P. pumila was independent of B. platyphylla local population density. Each simulation of density-dependence of two species consisted of randomly assigning the target species identity values to a set of pooled real coordinates, and calculating values of L-function. This approach was also used to test for a relationship between stem mortality and local population density (see Kenkel, 1988). Random mortality was simulated by removing trees at random from the combined (live plus dead trees) data set. The simulation was done by removing the same number of individuals as there were dead trees, and determining values of L(r) for the remaining individuals.

Growth variations and size-structure dynamics

Temporal size-dependent growth responses were examined by relating radial increments (absolute growth rate, AGR) for a specified time-period (e.g. 1995–2000) to stem size at the beginning of this period using log–log linear regression:

where a and q are intercept and slope parameters specific to a given growth interval, and c the error term. This approach has been used frequently to delineate between linear, size-proportional and non-linear, size-disproportional increase in growth rate, and to infer an alternative mode of plant interference (Westoby, 1982; Hara, 1988; Newton and Jolliffe, 1998). It is assumed that if plants grow in proportion to their sizes (q₁ = 1), thereby depleting limited resources without any individual obtaining a monopoly, competitive interactions among them are weak and symmetric, whereas if larger individuals grow disproportionately more than others (q₁ > 1), thereby pre-emptying resources at the expanse of smaller plants, competition is assumed to be intense and size-asymmetric, further increasing size inequality. The size inequality at the end of each growing interval was expressed by the coefficient of variation (CV), based on the variance of cumulative increments (stem diameters) obtained from tree-ring analysis. The growth was further examined as a function of both size and age of a tree using multiple regression:

where a partial regression coefficient q₂ significantly different from zero indicates that trees of the same size have different growth rates depending on their age.

To investigate the influence of neighbours on target stem growth, a simple measure of crowding (W, or index of competition) was calculated for every B. platyphylla and L. cajanderi for which growth data were available, as the sum of individual basal areas of neighbours divided by their distances from the target tree within a circle of specified radius around the target stem. The radii of 3 m and 6 m were used in this study. Edge effects were avoided by including only target trees from the inner square r metres of the plot border. Significant negative correlation between growth and neighbourhood indices can be considered as an indirect evidence of competition. To describe the intraspecific interference in the species-mixed stand, the indices were calculated using only conspecifics as neighbours, and to describe the effect of interspecific competition, W was calculated using neighbours of all species or heterospecific neighbours. To examine the asymmetry of competitive effect on growth, i.e. if neighbours larger than target tree have more than size-proportional negative effect on growth, the modified indices were calculated using the formula:

where k is the effect of neighbour, S_i is the size (basal area) of ith neighbour, S_t is the size of the target tree, d_i is the distance between ith neighbour and target stem, and A is the measure of asymmetry representing the proportion by which the effects of smaller neighbours are discounted (Thomas and Weiner, 1989). By discounting the effects of smaller neighbours on target tree growth, what degree of asymmetry in the neighbourhood model accounts for the maximum amount of growth variation was tested. The neighbour values were square-root-transformed to achieve normality of the residuals and homogeneity of variances. Although tree mortality was low, the neighbourhood analysis was restricted for the last 5 years, during which the arrangement of stems was likely to be the same for the entire period.

In 2000, it was assumed that some trees were still growing actively, and thus had no apparent asymptote, while others showed a decline in radial growth. To quantify this, individual cumulative growth curves, representing general size–age relationship, were fitted with the simple linear model and the non-linear Richards model (Richards, 1959):

where D_t is stem diameter at a given age, A is the asymptotic or maximum diameter, K defines ‘speed’ of growth, δ determines the more or less sigmoid shape of the curve (shape parameter is the rate at which diameter approaches the asymptote at inflection point), t is age at which size would be zero and t^* is age of growth inflexion. Many attempts have been made to develop functions which are able to describe patterns of individual growth. The Richards growth model is the generalization of the classical growth curves: Gompertz (δ = 1), logistic (δ = 2) and von Bertalanffy (δ = 2/3) and can be compared against these simpler models (Damgaard et al., 2002). The parameters in the individually based growth models were determined and then tested to see if trees with different growth patterns differ in size, age and intensity of crowding they experienced.

RESULTS

Composition and stand structure

A total of 1387 live trees (3468 ha⁻¹) taller than 1·3 m and 46 standing dead trees (115 ha⁻¹) were mapped in the 0·4-ha plot. Betula platyphylla comprised 74·5 % of the trees and 76·6 % of the basal area, and formed the main canopy with L. cajanderi, Salix spp., Alnus spp. and Pinus pumila occurring from low to moderate density in the lower canopy and understorey (Table 1). Betula platyphylla had normal distributions of age and diameter classes (Fig. 2), typical of pioneer species that establish as approximately even-aged cohorts in a large forest opening. Latix cajanderi had a negative exponential or inverse-J pattern, typical of an uneven-aged population, with most trees <5 cm dbh. The estimated density of P. pimula clumps taller than 1 m is 324 ha⁻¹. Pinus pumila had an average height of 2·1 m.

Fig. 2.

Size-class distributions of breast height diameters and tree heights.

All tree species had clumped distribution at one or more spatial scales. Betula platyphylla stems were significantly clumped at distances of up to 6 m, were randomly dispersed between 6 and 14 m, and were clumped at scales greater than 14 m (Fig. 3). The highest clumping intensity found between 1 and 2 m is an approximate indicator of average patch radius. Mean number of conspecific neighbours (± s.d.) within a circle of 3-m and 6-m radius around B. platyphylla was, respectively, 9·4 ± 4·9 and 27·8 ± 9·6 individuals. Larix cajanderi stems were clumped at all scales analysed. The mean conspecific density (±s.d.) in the neighbourhood 0–3 and 0–6 m around Larix cajanderi was 3·4 ± 2·7 and 10·3 ± 7·1 individuals. The mean number of B. platyphylla stems within a circle 3 m around the target L. cajanderi was 6·9 individuals when all neighbours were counted, and 5·0 individuals when larger neighbours were considered. The analysis of spatial distribution of L. cajanderi in relation to B. platyphylla showed no deviation from the null hypothesis of density independence. The only significant result was a negative dependence between P. pumila and B. platyphylla at short distances of 2 m. The analysis of spatial distribution of standing dead trees, most of which (80·4 %) were B. platyphylla, in relation to the surviving trees, showed no deviation from the null model of random, density independent mortality.

Fig. 3.

L(r) showing spatial distribution pattern of Betula platyphylla stems (solid line) over 25 m distances. The thin lines give approx. 99 % confidence interval for null hypothesis of Poisson random distribution.

Size-dependent growth responses in pioneer B. platyphylla population

Figure 4 and Table 2 present the results of univariate regressions relating radial increment of B. platyphylla during the 5-year period to the stem diameter at the beginning of this period. Slope parameter estimates for relationship between AGR and size on a log–log scale were significantly greater than zero but less than one in eight out of ten 5-year periods analysed. The slope of <1 for the log AGR–log size relationship means that the relationship between size and relative growth rate (RGR) is negative, i.e. the relative increment of smaller trees is higher than that of larger trees, and hence size inequality will decrease over time. There were two temporal trends in the slope and intercept values obtained from regressions on a log-transformed growth-size data (Table 2): (1) a decline in slope values with increasing intercept values during 1950–1970, and (2) an increase in slope values with decreasing intercept values during 1970–2000. During the early phase of tree colonization, the stand was composed of many small B. platyphylla and a few older individuals, some of which may have survived the fire. The abundant smaller trees were growing faster than larger trees during 1950–1970, which caused size differences to decrease rapidly during that period. During the period 1970–1995, the stand becomes increasingly saturated: as the average radial increment declines, tree size variability decreases only slightly to remain almost constant during the period 1980–1995 (‘period of weak, symmetric competition’). From 1990 to 1995, trees grew in proportion to their sizes, i.e. relative growth rate was unrelated to size, whereas from 1995 to 2000, larger trees grew disproportionately more than smaller trees (the slope of >1 in eqn 1), and hence stem size inequality increased in this period (Table 2). Although stem size at the beginning of a time interval was most important determinant of radial growth during that interval, the age of a tree also showed a significant effect. When size was held constant in the multiple regression model, the age of the tree had a significant negative effect on radial growth rate, i.e. younger trees grew faster than older trees of the same size.

Fig. 4.

Least-squares regressions of Betula platyphylla stem diameter increment during a 5-year growth period on stem size at the beginning of this period (see Table 2).

Table 2.

Changes in intercept and slope parameter estimates obtained from linear regression of log AGR on log stem diameter (with explained variance and significance levels), and average values of absolute growth rate (AGR) of stem diameter of B. platyphylla, size inequality of stem diameters (measured as coefficient of variation, CV_d), and spatial variability in radial increments (CV_AGR)

Period	Intercept	Slope		AGR (mm)	CVAGR	CVd
1950–55	0·47	0·32	0·23**	4·98	0·53	0·89
1955–60	0·69	0·20	0·14*	7·80	0·58	0·99
1960–65	0·71	0·26	0·35***	9·53	0·53	0·66
1965–70	0·74	0·19	0·20***	10·08	0·47	0·56
1970–75	0·58	0·21	0·11***	8·37	0·47	0·48
1975–80	0·36	0·28	0·10**	7·10	0·51	0·46
1980–85	0·17	0·33	0·13***	5·86	0·54	0·43
1985–90	−0·35	0·53	0·22***	4·63	0·64	0·41
1990–95	−0·95	0·88	0·32***	4·04	0·71	0·40
1995–00	−1·61	1·21	0·42***	3·32	0·85	0·43

^*P < 0·05,

^**P < 0·01,

^***P < 0·001.

Growth and local competition

The relative growth rate of B. platyphylla from the 1995 to 2000 was negatively correlated with all neighbourhood interference indices calculated (Table 3). Strengths of the linear regressions increased as the neighbourhood radius included in the calculations increased from 3 to 6 m, and when the effects of smaller neighbours on target tree growth were gradually discounted. Improved r² of regressions discounting the effects of a smaller neighbour suggests that asymmetric competition was occurring among B. platyphylla stems from 1995 to 2000. The competition indices that accounted for the most variation incorporated both intra- and interspecific effects and an asymmetry value of 0·8–1 for both the 3- and 6-m radius (Fig. 5B). In the multiple regression model, where AGR of B. platyphylla from 1995 to 2000 was related to the stem diameter and age in 1995, and local competition, the most variation was explained by stem size (partial correlation r_part = 0·55), followed by competition (r_part = −0·38), while the effect of age on growth was no longer significant. Stem size and competition together accounted for 56·9 % of the variation in radial growth rate from 1995 to 2000.

Table 3.

Explained variance (adjusted r²) for univariate regression of relative growth rates of Betula platyphylla and Larix cajanderi during 1995–2000 on several measures of local interference, for two neighbourhood radii, and several levels of competitive asymmetry (see text)

			Degree of competitive asymmetry
Target	Neighbour	Radius	0	0·2	0·4	0·6	0·8	1
L.c.	L.c.	3 m	–	–	–	–	0·04	0·05
		6 m	–	–	–	–	0·02	0·05
L.c.	B.p.	3 m	0·21**	0·25***	0·28***	0·31***	0·33***	0·30***
		6 m	0·27***	0·32***	0·36***	0·37***	0·36***	0·30***
L.c.	All spp.	3 m	0·09*	0·15*	0·19**	0·23**	0·26**	0·29**
		6 m	0·16*	0·24**	0·29**	0·32***	0·33***	0·33***
B.p.	B.p.	3 m	0·09***	0·13***	0·16***	0·20***	0·24***	0·23***
		6 m	0·13***	0·18***	0·24***	0·28***	0·31***	0·33***
B.p.	All spp.	3 m	0·13***	0·17***	0·21***	0·25***	0·27***	0·26***
		6 m	0·20***	0·27***	0·32***	0·37***	0·38***	0·38***

L.c., Larix cajanderi; B.p., B. platyphylla.

^*P < 0·05.

^**P < 0·01.

^***P < 0·001.

Fig. 5.

Relationship between relative growth rate (RGR) of stem diameter and the intensity of neighbourhood competition (crowding); (A and B) neighbourhoods were defined by considering all trees as neighbours from distances of 3 m; (C) only L. cajanderi from distances of 6 m; and (D) B. platyphylla as neighbours from distances of 3 m. Note that asymmetry values differed.

Regressions of L. cajanderi relative growth rate on measures of local interference that incorporated only conspecific neighbours were nonsignificant for both neighbourhood radii and all measures of competitive asymmetry (Table 3 and Fig. 5C). The growth of L. cajanderi was negatively correlated with measures of neighbourhood competition incorporating only B. platyphylla as neighbours (Fig. 5D). This indicates that for L. cajanderi interspecific competition was much stronger than intraspecific competition. The neighbourhood radius of 6 m and asymmetry values of 0·6 resulted in the highest r², although models with zero asymmetry were also significant (Table 3).

The growth curves of most B. platyphylla trees (91 %) were better fitted by the multiparameter non-linear Richards growth model (r² > 0·98) than the linear model. The Richards model, which includes the additional shape parameter (δ), gave equal or better fit than the single logistic or Gompertz model. Based on the estimated parameters in the Richards model, two group of trees could be distinguished: inferior and dominant (Fig. 6). The former had the smaller size at which radial growth began to deviate from exponential and, hence, the smaller maximum size (diameter) that could be achieved (Table 4). The inferior trees (39 % of 152 trees fitted by the non-linear model) were not different in age from dominant trees, but they had significantly smaller height, shorter crown and slenderer stem, and experienced stronger crowding (Table 4).

Fig. 6.

Growth curves (obtained from tree-ring analysis) of stem diameter for the individual trees of B. platyphylla (A and B) and L. cajanderi (C and D). In both species there were suppressed (A and C) and dominant (B and D) trees.

Table 4.

Mean values of parameter estimates (A, K, δ, age, t^*) in the Richards growth model (eqn 4), and mean stem height (m), age (years), SSI (index of stem slenderness), RCL (relative crown length), crown length (m) and intensity of crowding (W) calculated for two groups of B. platyphylla trees

Tree status	A	K	δ	t*	Height (m)	Age (years)	SSI	RCL	CL	W
Inferior	75·5	0·147	1·99	12·0	10·3	41·2	1·31	0·46	5·6	88·9
Dominant	141·6	0·108	1·94	16·4	14·9	43·2	1·53	0·52	7·3	47·7

Influence of neighbours on stem allometry

Size distributions of tree measurements in 2000 reflected growth patterns of the respective species. In B. platyphylla, dbhs were more variable than heights compared with L. cajanderi (Fig. 2). Obviously, B. platyphylla stems grew taller rather than thicker. When stem slenderness (height: dbh ratio) was regressed on indices of local crowding, there were significant positive correlations, indicating that B. platyphylla with close and larger neighbours developed slenderer stems. Interspecific correlations with indices where neighbourhoods included all species were stronger than intraspecifc correlations, and further increased when the neighbourhood radius increased from 3 to 6 m, and the levels of asymmetry from 0 (r = 0·17, P = 0·07) to 1 (r = 0·48, P < 0·001). The relative crown length (crown length : stem height ratio) of B. platyphylla was negatively correlated with measures of local interference, and the relationship was stronger for intraspecific interference, i.e. when only B. platyphylla individuals were used to define a local neighbourhood, and when the levels of competitive asymmetry increased from 0 (r = −0·208, P = 0·024) to 1 (r = −0·263, P = 0·004). Mean height (±s.d.) of B. platyphylla stems was 10·9 ± 4·5 m and mean crown length 5·5 ± 2·9 m. Crown length increased as tree height increased (Fig. 7B); however, given individuals of the same height, those with thicker stem had a longer crown than individuals with a thinner stem (log crown length₂₀₀₀ = 0·19 log dbh₂₀₀₀ + 0·78 log tree height₂₀₀₀ − 0·231, adjusted r² = 0·83, P < 0·001, n = 1033). Larix cajanderi stem allometry was also significantly influenced by neighbour abundance and size. The significant negative relationship between stem slenderness and the crowding indices was due primarily to a negative effect of taller B. platyphylla on L. cajanderi (r = 0·45). On the contrary, the relative crown length was best correlated with indices that incorporated all tree species larger than the target tree (r = −0·29). Larix cajanderi in 2000 were shorter than B. platyphylla of the same dbh, but had a longer crown than B. platyphylla of the same height (ANCOVA, both P < 0·001; Fig. 7). Obviously, L. cajanderi allocated more biomass to diameter growth and crown maintenance than height growth compared with B. platyphylla.

Fig. 7.

Relationship between biometrical measures of Betula platyphylla (shaded circles, thin line) and Larix cajanderi (filled circles, thick line). (A) B. platyphylla: y = 0·95x + 3·22, r² = 0·83; L. cajanderi: y = 0·71x + 2·52, r² = 0·89. (B) B. platyphylla: y = 0·56x − 0·62, r² = 0·76; L. cajanderi: y = 0·64x − 0·15, r² = 0·91.

DISCUSSION

Forest stages similar to our stand were observed at other sites in central Kamchatka, where B. platyphylla is consistently the early dominant of burned areas, followed by the reinvasion of more shade-tolerant L. cajanderi, P. pumila and Picea ajanensis. Succession pathways appear to depend on the hydrology of the site and intensity of fire disturbances (Krestov, 2003). Dryer, excessively drained soils with frequent fires facilitate a shift to dominance by L. cajanderi (Goldmmer and Furyaev, 1996). Dryer sites discourage establishment from seeds so that hardwoods regenerate largely from sprouts after fire, but sprouting is usually not abundant enough to result in dominance, and opened stands are soon invaded by conifers (J. Doležal, unpubl. res.). Dryer sites, where fire danger is high, succeed usually to a light larch–lichen tajga, whereas mesic sites succeed to dark larch or spruce tajga. In the long-term absence of fire, the future mature forest at the study site is likely to be dominated by L. cajanderi, with P. pumila understorey, and a smaller component of hardwood trees.

The majority of B. platyphylla trees established in the first 15 post-fire years, and little regeneration occurred after this period. Larix cajanderi establishment peaked 20–30 years after fire. Many studies documented that the first 10–30 years, before full development of tree canopy and herb understorey, is most favourable for establishment of tree populations (Payette, 1992). The lack of trees <4 m tall indicates that B. platyphylla is not continuing to establish in this burn, and hence its stand density was determined by the relative success of seedling establishment during the early stage of stand development. Post-fire stand density and structure may depend on the size, intensity and seasonality of the burn, pre-fire floristic composition and propagule availability (Bergeron and Dansereau, 1993). Rapid birch establishment is favoured when there is an exposed mineral seedbed and an abundant supply of seeds (Johnson, 1992). The low intensity fire resulting in poor seedbed conditions, or the lack of seed sources, lead to sporadic or delayed recruitment restricted to areas of high intensity burn (Shafi and Yarranton, 1973). The main obstacles restricting tree regeneration at the studied site and other comparable sites are the numerous burnt remnants of L. cajanderi on the forest floor and the adverse effects of competition from understorey vegetation (Fig. 1). Since most fires in the region are low-intensity ground fires that occur in spring when understorey plants have enough resources stored in below-ground organs, regrowth of above-ground parts, flowering and restoration of ground-cover communities may occur shortly after fire (cf. Komarova, 1992). A low light level in the understorey then results in a short recruitment period for trees. Betula platyphylla had a relatively low stand density and a strongly clumped stem distribution, indicating that vegetative reproduction (sprouting from pre-burn trees) prevailed over seed regeneration. Seedling establishment in this burn is likely to have occurred only in some favourable seedbeds lacking wood debris and deciduous litter that prohibit germination.

After the initial phase of colonization, the B. platyphylla saplings of different diameter and age tended to converge giving a population of similar-sized individuals. The size differences that arose from differences in age decreased most rapidly during the first two decades as younger trees were attaining greater RGR than older trees. This suggests that resources such as water and soil nutrients released by fire have been depleted by many trees without any individual obtaining dominance. If there was any competition early in stand development, its effect seems to be greater on older than younger trees (see regression coefficients in Table 2). It is possible that trees that originated from sprouts grew faster than those emerging from seeds. Sprouts might gain competitive advantage by utilizing resources of stem remnants they originated from. Another possible explanation for decreasing size inequality is relatively low tree density in this stand leaving much space for individual growth. A steady decrease in size variability during the 45-year period of Pinus sylvestris stand development was attributed to a weak competition among individuals growing under low density on infertile peat bog soil (Stoll et al., 1994). In the case described here, the steady decrease in size variability was followed by a period of relatively stable size structure, but the last decade of stand development was characterized by increasing inequality in stem size. An intuitive interpretation is that with increasing stand cover and light limitation, the tallest individuals begun to disproportionately pre-empt the incoming solar radiation because of its unidirectional nature, thereby suppressing the growth of smaller trees. Such asymmetric effects should give rise to positive correlation between RGR and stem size (Weiner, 1990), and this is what was observed during the period 1995–2000.

Further evidence for increasing size-asymmetry of competition in the later stages of stand development was demonstrated by neighbourhood analysis. Our measure of local competition, together with stem size, accounted for 56·9 % of the variation in B. platyphylla radial growth from 1995 to 2000. The neighbourhood model fitted best when the effect of smaller neighbours on target stem growth was discounted 80–100 %. Moreover, B. platyphylla growth was inhibited by both the neighbouring conspecific and heterospecific trees of larger sizes (see improved r² of regression incorporating the effects of taller L. cajanderi and Alnus hirsuta; Fig. 5). These results support the existing generalization that the mode of plant interactions alter during stand development, from early stages of weak competition for soil resources released by fire to later stages of asymmetric competition for light (Kenkel, 1988). However, compared with studies where stand density or productivity were higher (Hara et al., 1991; Kenkel et al., 1997; Wichmann, 2001; Dolezal et al., 2004), the first evidence of asymmetric competition was seen after four to five decades. The later start of asymmetric competition in this stand can be explained by (a) the lower density of natural establishment leaving much space for individual growth, (b) greater relative importance of below-ground competition at this site of nutrient-poor volcanic soil, and (c) B. platyphylla's preference for height-growth which is common for shade-intolerant, pioneer species (Sumida and Komiyama, 1997).

The study reported here suggests that the effect of B. platyphylla canopy on the establishment of invading L. cajanderi was not inhibitory. Where inhibition is predominant, there should be spatial separation between the dominant and subsequent invader species. In northern Michigan, pioneer aspen and invading white pine were negatively correlated as a result of dense aspen clones restricting white pine regeneration into canopy openings (Peterson and Squiers, 1995). In our case, spatial associations between B. platyphylla and L. cajanderi were neutral, as predicted, as B. platyphylla tolerates conifer colonization. Detailed inspection of the stand map revealed that most L. cajanderi did not establish in canopy gaps but rather in the vicinity of B. platyphylla clumps and few older conspecific trees, which can be explained by reduced herb and shrub cover beneath tree canopies. The B. platyphylla stand was relatively sparse and non-uniform (clumped) in structure. It remained transparent for light, providing good conditions for undergrowth development, especially in gaps between canopy patches. Larix laricina regeneration in northern Québec occurred exclusively in the vicinity of older trees because of reduced lichen cover beneath their canopies and a more favourable microclimate (Morin and Payette, 1984). It is likely that the few adult L. cajanderi trees found in the plot represented the initial point for species advance that peaked 20–30 years after fire, before the full development of B. platyphylla canopy. Peak establishment coincided with the onset of a decline in the growth in B. platyphylla that can be attributed to intensifying intraspecific competition for light with canopy closure. Growing in the summer shade of the B. platyphylla, L. cajanderi had steady diameter growth during the first 10–20 years, after which radial growth declined somewhat (Fig. 6). Subalpine fir growing under birch in the British Columbia sub-boreal forest likewise showed slow but steady height growth during the first 10–15 years, after which growth declined (Wang et al., 2000). The results of this study support the hypothesis that interactions among species involved in succession vary with the life-stage or successional phase. Little evidence of a negative effect of B. platyphylla canopy was found on the establishment of L. cajanderi but there was an apparent local inhibition of L. cajanderi growth by B. platyphylla canopy in the the later stages of stand development. As an appropriate successional model for species replacement it is suggested that there is tolerance in early and inhibition in later stages of stand development (Connell and Slatyer, 1977).

It is concluded that, although the influence of competition on growth and stem allocation patterns of the species was clearly detectable, it had little influence on their population size and spatial structure. Competition led to suppression but not to exclusion. After 53 years of stand development, mortality was low and less than half of B. platyphylla trees showed growth decline as a result of competitive stress. It can be assumed that, if trees continue to compete asymmetrically, the resultant dominance and suppression will cause increases in size inequality and mortality rates in shaded trees. It is predicted that the stand will become more uneven-sized over time, and more open as a result of self-thinning, and will become increasingly dominated by L. cajanderii and P. pumila, since they currently occur at higher density in the sapling and shrub stratum. In contrast, B. platyphylla, Alnus spp. and Salix spp., which are infrequent in the lower tree strata, will decline over time.