Plant foraging and dynamic competition between branches of Pinus sylvestris in contrasting light environments
Abstract
1 The morphological plasticity of sun and shade branches of Pinus sylvestris growing at the edge of a ‘tree-patch’ created in 1989, when surrounding trees were removed, was compared with that of branches of trees growing in the centre of the patch.
2 In 1992, we selected 10-year-old branches. Yearly growth increments (long-shoots) and emerging buds were individually marked, architectural parameters measured, the branch topology mapped, and the number of male and female cones counted. In 1993, the survival of growth increments and buds was recorded and the branches were harvested to determine dry mass of growth increments and needles.
3 The size of growth increments produced after 1989, i.e. their length, total dry mass and needle dry mass, decreased in the order sun branches of edge trees > branches of centre trees > shade branches of edge trees. Thus, the growth increments produced on the shade branches of trees that also had branches in the sun were consistently smaller than the growth increments on the branches of centre trees growing completely in the shade.
4 The number of new growth increments produced after 1989 was highest in sun branches and lowest in shade branches of edge trees. Survival of growth increments and buds was higher in edge than in centre trees; no difference was found between sun and shade branches of edge trees.
5 In shade branches of edge trees, branching angles between first- and second-order growth increments were highest and increased from older to younger growth increments. This was interpreted as ‘bending’ towards the edge of the patch.
6 Production of female cones was almost totally restricted to edge trees and higher in sun than in shade branches.
7 One prediction of an optimal foraging strategy, i.e. the production of more growth increments and buds in higher light, was supported by the data, whereas the other prediction, i.e. decreased length of growth increments, could only be supported when it was expressed per unit dry mass. Thus, the hypothesis of an adaptive foraging strategy in plants was rejected in favour of a ‘passive’ growth null hypothesis.
8 The results suggest that both growth increments within branches and branches within trees are physiologically integrated and their ‘foraging behaviour’ can only partly be understood in terms of their local environment. We interpreted the observed differences between shade branches of edge and centre trees as correlative growth inhibition.
Introduction
The acquisition of heterogeneously distributed resources by foraging individuals has been studied extensively in animals (reviewed by Pyke 1984) and also in clonal plants (reviewed by Hutchings & de Kroon 1994; de Kroon & Hutchings 1995). Theories of optimal foraging consider how an individual organism can optimize the amount of food intake through its behaviour. The basic assumption is that an individual that consumes food more efficiently than others grows bigger and thereby increases its individual fitness (Maynard Smith 1978; Pyke 1984). For example, in plants there is a clear positive relationship between size and reproductive output (Samson & Werk 1986; Hutchings 1988; Weiner 1988; Schmid et al. 1994). Plants are sessile organisms (except as seeds) but an analogy to animal foraging can be made if we substitute growth for movement and resources for food. Thus, growth plasticity in plants, e.g. increased vegetative growth in response to high availability of local resources, can be seen as analogous to (non-random) movement in animals. This analogy has been applied to branching patterns in clonal plants with a modular architecture, where vegetatively produced ramets with leaves and roots are considered to be ‘feeding sites’ that are placed by ‘spacers’, e.g. rhizomes or stolons (Bell 1984; Hutchings 1988; Schmid 1990).
Whether morphological plasticity in branching patterns of plants follows some general rules (i.e. adaptive growth plasticity) or may be explained simply in terms of a growth response of vegetative ramets to the availability of local resources (growth null hypothesis) is still an open question (Cain 1991, 1994; Oborny 1991, 1994; Oborny & Cain 1997; Stoll et al. 1998).
Clonal growth is only one conspicuous expression of modularity. Another expression can also be found in the three-dimensional architecture of trees (Harper 1985; Franco 1986) whose crowns are comprised of three basic shoot types: the main stem, the branches or long shoots that frequently have indeterminate growth, and short shoots that may carry most of the foliage (Ford 1985). Like clonal plants, trees grow by ‘vegetative reproduction’ of modules, where a module is defined as ‘a repeated unit of multicellular structure, normally arranged in a branch system’ (Harper 1981). The set of all modules derived from a single zygote is defined as genet (Harper & White 1974). Assimilate transport, as known from clonal plants in the context of physiological integration (e.g. Lötscher & Hay 1997; Derner & Briske 1998), can potentially take place at two hierarchical levels of structural elements: the growth increments within branches and the branches within the trees. In contrast to many clonal plants in which the plagiotropic connections between modules or ramets are in close contact with the soil and usually only persist for a few years, trees offer the advantage that the woody aerial connections persist as long as the modules they bear, thus keeping a record of their ‘foraging history’ over time.
Although carbon allocation patterns in trees have been investigated intensively in the context of gap dynamics (Shugart 1984; Young & Hubbell 1991) and branch autonomy (Sprugel et al. 1991), the parallel with foraging strategies through morphological plasticity of spacers (e.g. long-shoots) has rarely been made. This is surprising for two reasons. First, pioneering studies (Franco 1985; Jones 1985; Jones & Harper 1987a,b) showed that modules of trees like those of herbaceous plants respond to their local environment. These studies used the module demography to describe tree growth in a mechanistic but integrated way (see also Flower-Ellis 1980; Maillette 1982a,b). Secondly, considerable knowledge is available at the level of needles (on short shoots). For example, it is known that needle packing, i.e. the number of needles per unit stem length, is increased in sun compared with shade shoots in several conifer species (Carter & Smith 1985) and that shade shoots of subalpine fir (Abies lasiocarpa) have a more planar arrangement of non-overlapping needles than sun shoots (Smith et al. 1991). For obvious reasons, such patterns have been studied in the context of maximizing (stem) wood production. However, as Ford (1985) pointed out in discussing timber production, branching structures may have evolved, at least in part, to ensure the efficient display of reproductive organs (Despland & Houle 1997) or to avoid shading by neighbouring plants (Franco 1986).
A more comprehensive approach is to combine the study of allocation patterns with either or both of architecture (Canham 1988; King 1997) and module demography. For example, Küppers (1989a,b) links plant growth, as evidenced by increment of biomass, with architecture, i.e. deployment of biomass in space, to evaluate the efficiency of different crown forms and the effects of their development on tree succession. Very few such combined analyses have been made, although Sorrensen-Cothern et al. (1993) showed that including plasticity at the branch level in a model of competing trees fundamentally changes the effect of competition at the population level. If plasticity through modular growth of branches and foliage was included within a spatially explicit stand model, the whole stand had a greater leaf area index and individuals grew taller compared with a model without plasticity. Thus, for a more mechanistic analysis of competition, interactions between plant modules and their local resource environments should be modelled (Sorrensen-Cothern et al. 1993; Stoll et al. in press) in addition to interactions between whole plants (e.g. Mack & Harper 1977; Silander & Pacala 1985; Miller & Weiner 1989; Bonan 1991; Stoll et al. 1994).
Model system and hypothesis
In the present study, branches of Pinus sylvestris L. growing in different light environments were investigated. A management plan to re-establish the former vegetation on a degenerated raised bog had prescribed cutting down the naturally established, more or less even-aged population of P. sylvestris. However, one patch of trees was deliberately left and this allowed the growth, architectural relationships, and demography of buds and yearly growth increments (long shoots) to be recorded and compared between branches that were exposed to full sunlight at the edge of the patch and shaded branches of the same trees. The results from shade branches could then be compared with those obtained from trees growing in the centre of the patch.
We asked two questions. (1) Do branches in the sun behave differently from branches in the shade? (2) Do shade branches attached to trees that also have some branches in the sun behave differently from shade branches attached to trees that are entirely in the shade?
The first question was used to contrast a hypothesis of optimal foraging strategy in plants (Sutherland & Stillman 1988) with the alternative hypothesis that plant parts in the sun simply grow more because they have more photosynthate available (growth null hypothesis). For branch architecture, the foraging hypothesis predicts that: (i) growth increments should be shorter and (ii) branching frequency higher in branches exposed to full sunlight, but (iii) branching angles should not be altered due to differences in local light availability. The first prediction can be used to discriminate between the foraging and the growth null hypothesis because the latter would predict longer growth increments on sun compared to shade branches of edge trees.
In addition to these predictions regarding branch architecture, we expected that (iv) total dry mass of growth increments would be higher in the sun and (v), as a consequence of predictions (i) and (iv), the length per unit dry mass would be greater in the shade. We further hypothesized that (vi) not only the production but also the survival of growth increments and buds would be lower in the shade. If there is integration within branches, (vii) those parts of the branches that had been produced before cutting created the patch should become heavier in branches afterwards exposed to sunlight than in those still growing in shade, because of assimilates received from their progenies in the sun. Finally, if resource acquisition is indeed optimized at the level of vegetative growth increments this should also benefit sexual structures and (viii) reproductive output should therefore be higher in the sun.
If we can assume that shade branches of both edge and centre trees are growing in similar light environments, then the theory of branch autonomy (Sprugel et al. 1991) would predict no differences between the shade branches of edge and centre trees. However, responses could also potentially involve either import or export of assimilates or resource competition between branches of edge trees (integration within tree), as demonstrated in experiments with herbaceous aclonal (Novoplansky et al. 1989; Sachs & Novoplansky 1997) and clonal plants (Stuefer et al. 1994) and even with connected shoots of genetically different individuals of the facultative hemiparasite Rhinantus serotinus (Prati et al. 1997).
Materials and methods
Study site and tree species
The study site ‘Hagenmoos’ is a degenerated raised bog 15 km south of Zürich (47°14′N, 8°31′E; altitude 600 m). During the Second World War the bog was drained and used for peat cutting but, after being abandoned in 1945, the bog was invaded by P. sylvestris, a common pioneer tree in such situations. During the next 40 years the trees grew without direct human influence. In February 1989, in an attempt to restore the raised bog, the old drainage ditches were blocked and, with the exception of one patch (c. 90 m2), all trees were cut (for a detailed description of the site see Lüönd & Göttlich 1982; Stoll et al. 1994).
Pinus sylvestris is a light-demanding tree with a regular, monopodial, determinate branching pattern (Kozlowski 1971) and a single annual flush of growth (Kikuzawa et al. 1996). Buds are initiated in autumn of the previous year and elongate during a relatively short period of about 3 weeks in spring (Kozlowski & Winget 1964; StaÅhl 1984). Thereafter, no new development of leaf tissue occurs, although cambial (diameter) growth continues for several months (Sprugel et al. 1991). In contrast to Picea, in which some new growth increments are produced by branch segments older than 1 year (Gruber 1987; Schill 1989; Gruber 1990; Pankow et al. 1991), such older branch segments never produce buds or new growth increments in P. sylvestris. It is therefore possible to date every segment within a branch. Male and female cones are produced on the same individual tree. Flower development from differentiation of floral initials up to seed maturity takes about 2.5 years and depends on the temperature sum (Chung 1981). Short shoots bear two needles each and are attached to long shoots. The annual increase in length of these long shoots are here referred to as growth increments.
Sampling strategy
In 1992, four edge trees and four centre trees were selected by visual assessment of their placement in relation to the border of the patch (Fig. 1). Two 10-year-old branches were selected in each tree and, since clearing of trees around the patch took place in 1989, all had experienced the pre-clearing light regime. On edge trees, one branch was taken from the outside, i.e. had been exposed to full sunlight after clearing, and one branch was taken from the inside, i.e. had remained in the shade. The two branches of each centre tree were both growing in the shade and served as controls. Notice that we subsequently refer to the branches of edge trees as sun and shade branches, respectively, and to the branches of centre trees as control branches. Where the selected branches were subsequently discovered to have lost their main axes before clearing, they were replaced by 11-year-old branches [two on each of tree numbers 3 and 5 (Fig. 1), and one on each of tree numbers 6 and 7].
Map of the study site with the patch of trees. Trees growing around the patch were cut in February 1989. This treatment exposed ‘sun branches’ of edge trees to full sunlight, whereas ‘shade branches’ of edge trees and ‘control branches’ of centre trees remained in the same local light environment. Branches on trees are drawn according to actual measurements (see scale bar). Insertion height of branches on the trunk is about 6 m above ground. Unmeasured branches are not included in the figure and unmeasured trees are only indicated. Tree numbers (within circles) are used to refer to particular trees in the text.
The growth increments building up the main axes of the branches were termed first-order, those on primary side branches second-order, and so on (Fig. 2). Over 3800 growth increments (summing up to a total length of 140 m branch wood) were labelled with thin, coloured plastic wires and their topological relationships (parent–offspring) mapped. The length of growth increments and their branching angles relative to parent growth direction were measured from 18 to 25 May in 1992, and the presence and absence of cones were recorded. At the tip of the branches, all growth increments of the four youngest cohorts (i.e. those formed between 1989 and 1992 after the exposure to light) were sampled, whereas the older cohorts were only fully sampled for first- and second-order growth increments. The remaining orders were sampled at least once per second-order parent. From 10 to 15 May in 1993 all branches were harvested and the survival of growth increments was determined by comparing the maps of the previous year with the harvested branches. The measurement dates in both years were chosen after needle expansion was well underway and growth increments were assumed to be fully elongated (see below). The length of buds newly elongating in 1993 was measured and all others were remeasured. However, out of the more than 3800 growth increments, only three measurements of 1-year-old and 23 of older increments showed an increase of more than 5 mm between the two dates. Thus, elongation was restricted to a short period in spring and was finished completely after 1 year. The entire harvested branches were stored and dried for half a year in a dark and dry cellar (c. 15 °C). The growth increments were then cut off from the branches, divided into stem tissue and needles, and all parts weighed.
Sample branch to illustrate numbering and dating scheme. The main axis consists of first-order growth increments and bears the next higher, second-order growth increments on side branches, and these again bear the third-order growth increments, etc.
Light measurements
To quantify the different light environments, we made measurements on each branch for 1 day (07.00 to 19.00) between 30 June and 3 July in 1992. The instantaneous photosynthetically active photon flux density (PPFD) was simultaneously measured using 21 quantum sensors on each day. Five sensors were placed from the tip to the trunk on each of four branches: sun and shade branches of one edge and two branches of a centre tree. An additional sensor, placed in the middle of the tree patch at the same height as the branches at the same location on all days, was used as a reference. Values were averaged over 1-min intervals and recorded on a data logger (CAMPBELL CR10). In order to avoid artefacts due to particular sun flecks but nevertheless establish the anticipated differences in PPFD for our purposes, we summarized the measurements at the highest possible level. That is, minute averages of PPFD were integrated over individual 12-h measurement days, which had different weather conditions. Integrated values of the five sensors per branch were then averaged and summed over the 4 days. PPFD (percentage of reference sensor) were as follows: sun branches of edge trees received more light (133.4%) than shade branches of edge trees (85.9%) and these received slightly more light than branches of centre trees (78.0%).
Statistical analysis
The analysis of growth data should take serial correlations into account when differences due to different local light availability are tested for significance in an analysis of variance. However, because single-year analyses gave very similar results as repeated-measures analyses (Stoll 1995) we only report the single-year analyses but would like to emphasize that type I errors could be seriously inflated in other situations (Potvin et al. 1990). We used a mixed-model analysis of variance (calculated separately for each year) with the residual maximum-likelihood (REML) method to calculate predicted means, which where then used to present the results graphically. Tree and branch were used as random terms to account for individual differences not attributable to light. The REML method was used because the design was unbalanced at the level of growth increments and we wanted to use all the available information, including the variability between growth increments within branches. Comparisons between the analyses with branch means and the analyses without averaging revealed very minor differences. Wald test statistics with an asymptotic chi-squared distribution were used to test the significance of the fixed model terms (edge vs. centre trees and sun vs. shade within edge trees) as they were added into the model. Where standard errors of the means are shown they were taken from the same analyses.
The variables length and number of daughters were square-root transformed prior to analysis, whereas dry mass of growth increments, needles and specific length (i.e. length per unit dry mass) were log-transformed to normalize error variance. All analyses were carried out using the statistical package GENSTAT (Payne et al. 1993).
Results
The differences in total needle dry mass (including all orders from the completely sampled branch tips) of sun and shade branches were very large (Fig. 3). More surprisingly, although shade branches received more light than control branches, shade branches had less total needle dry mass (26.3 ± 3.6 g) than control branches (43.0 ± 3.5 g) on growth increments produced in 1992. Regardless of the light environment, there were very few 3-year-old needles (i.e. on growth increments produced in 1990) on any branches.
Total needle dry mass of Pinus sylvestris branches in contrasting light environments. Each point for sun and shade branches represents the total of four branches (± 1 SE) from the completely sampled branch tips. The standard errors for shade branches are smaller than the plot symbols.
New growth increments produced after the edge trees were exposed were shortest in shade branches and longest in sun branches (Fig. 4). The length of growth increments produced in the years before light exposure did not differ between branches that subsequently became sun and shade branches. The first- and second-order growth increments of shade branches of edge trees produced after 1989 were consistently (although not significantly) shorter than those of control branches (Fig. 4). The growth response to light exposure was remarkably similar in first- and second-order growth increments. We have not reported responses of higher-order growth increments because they also showed qualitatively similar patterns. Mean length of growth increments at the tree level, i.e. averaged over sun and shade branches of edge trees or averaged over the two control branches of centre trees, was very similar after 1988 but in the beginning (1985–87) longer growth increments on average were produced on edge compared to control trees. This difference was significant in 1986 for first- and second-order growth increments (χ2 = 4.5 and 5.5, respectively; d.f. = 1; P < 0.05).
Length of growth increments produced by Pinus sylvestris branches from 1985 to 1993 for (a) first- and (b) second-order growth increments. Plot symbols are (back-transformed) predicted means (±1 SEM) from analysis of variance. The thin line without symbols represents the control branches. Stars indicate significant differences between sun and shade branches of edge trees in separate analyses for each date (*P < 0.05; **P < 0.01; ***P < 0.001). The arrow indicates the year of light exposure of sun branches of edge trees in 1989. (These explanations also apply to 5-7).
There was no difference in the survival probability of growth increments between sun and shade branches (Table 1). The survival of growth increments for all branch orders was higher than 95% in edge trees but lower than 90% for second- and third-order growth increments in centre trees. For third-order growth increments this difference was significant (F1/6 = 8.2; P < 0.05). After the sun branches had been exposed to full sunlight, they produced more daughter growth increments than shade branches (Table 2). The differences were significant in 1991 for first-order (χ2 = 6.9; d.f. = 1; P < 0.01) and in 1990 for second-order growth increments (χ2 = 21.2; d.f. = 1; P < 0.001), and marginally significant for both orders in 1992 (χ2 = 3.6; d.f. = 1; P = 0.058). In the years 1990–92 shade branches always produced less first- and second-order growth increments than control branches such that there was no difference in the average number of daughter growth increments at the tree level.
| Tree position | |||||||
|---|---|---|---|---|---|---|---|
| Edge | Centre (Control branches) | ||||||
| Light environment of branch | |||||||
| Sun | Shade | ||||||
| Order of growth increments | % | n | % | n | Average % | % | n |
| 1 | 95.7 | 46 | 100.0 | 51 | 97.9 | 100.0 | 97 |
| 2 | 96.2 | 522 | 95.7 | 469 | 96.0 | 85.4 | 773 |
| 3 | 98.3 | 543 | 100.0 | 319 | 99.0 | 89.3 | 625 |
| 4 | 100.0 | 102 | 98.7 | 75 | 99.4 | 95.4 | 87 |
| All | 97.4 | 1213 | 97.7 | 917 | 97.6 | 88.4 | 1583 |
| Tree position | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Edge | Centre (Control branches) | ||||||||||
| Light environment of branch | |||||||||||
| Sun | Shade | Average | |||||||||
| Order of growth increments | Year | Mean | SE | Mean | SE | Mean | SE | n | Mean | SE | n |
| 1 | 85 | 4.25 | 0.75 | 4.25 | 0.25 | 4.25 | 0.37 | 8 | 4.12 | 0.58 | 8 |
| 86 | 4.50 | 0.29 | 4.25 | 0.48 | 4.37 | 0.26 | 8 | 3.75 | 0.25 | 8 | |
| 87 | 4.25 | 0.25 | 4.00 | 0.41 | 4.12 | 0.23 | 8 | 3.50 | 0.50 | 8 | |
| 88 | 4.00 | 0.41 | 3.25 | 0.48 | 3.62 | 0.32 | 8 | 3.62 | 0.26 | 8 | |
| 89 | 3.75 | 0.25 | 3.50 | 0.29 | 3.62 | 0.18 | 8 | 2.62 | 0.32 | 8 | |
| 90 | 2.00 | 0.41 | 2.00 | 0.41 | 2.00 | 0.27 | 8 | 2.62 | 0.42 | 8 | |
| 91 | 3.00 | 0.00 | 1.50 | 0.50 | 2.14 | 0.40 | 7 | 2.37 | 0.46 | 8 | |
| 92 | 2.50 | 0.50 | 1.25 | 0.25 | 1.67 | 0.33 | 6 | 1.75 | 0.31 | 8 | |
| 2 | 85 | 2.50 | 0.27 | 2.58 | 0.14 | 2.54 | 0.15 | 39 | 1.71 | 0.13 | 24 |
| 86 | 2.34 | 0.17 | 2.32 | 0.14 | 2.33 | 0.11 | 63 | 1.68 | 0.13 | 40 | |
| 87 | 2.27 | 0.13 | 2.12 | 0.10 | 2.19 | 0.08 | 88 | 1.74 | 0.10 | 61 | |
| 88 | 2.30 | 0.11 | 2.11 | 0.10 | 2.21 | 0.08 | 112 | 1.99 | 0.09 | 79 | |
| 89 | 2.10 | 0.10 | 1.97 | 0.09 | 2.03 | 0.07 | 118 | 1.79 | 0.08 | 98 | |
| 90 | 1.79 | 0.08 | 1.44 | 0.08 | 1.62 | 0.06 | 130 | 1.55 | 0.07 | 104 | |
| 91 | 1.36 | 0.07 | 1.23 | 0.06 | 1.30 | 0.05 | 132 | 1.32 | 0.06 | 105 | |
| 92 | 1.12 | 0.05 | 0.87 | 0.04 | 1.01 | 0.04 | 134 | 1.03 | 0.05 | 111 | |
The average branching angles between first- and second-order growth increments produced after 1989 were higher in shade branches than in sun branches or branches of centre trees (Fig. 5). Irrespective of the year in which a growth increment was formed, branching angles in shade branches remained fairly constant. In contrast, branching angles progressively decreased from older to younger growth increments in both sun and control branches after 1989.
Branching angles between first- and second-order growth increments produced by Pinus sylvestris branches from 1985 to 1993.
Given the different light environments, the consistent differences in the dry mass of growth increments between sun and shade branches was not surprising (Fig. 6). However, we also found that after the sun branches had been exposed to full sunlight, dry mass was lower for both first- and second-order growth increments of shade branches compared with the control branches. Although this difference was small and not significant, it was consistent. It also contrasted with 1988 and 1989 second-order growth increments, which had exactly the same dry mass in shade and control branches, and with older growth increments, which for shade branches were considerably heavier than those of control branches. Whereas the length of first- and second-order growth increments already present in 1989 was almost the same in sun and shade branches (cf. Figure 4), the dry mass of the same growth increments was heavier in sun branches (notice the log-scale in Fig. 6).
Dry mass of growth increments produced by Pinus sylvestris branches from 1985 to 1993 for (a) first- and (b) second-order growth increments.
We regressed the number of vegetatively produced daughters on the dry mass of growth increments to test whether the slope (or intercept) differed in the contrasting environment (results not shown). There were no significant differences in the parameters, although shade branches of edge trees tended to have the shallowest slope, i.e. produced less daughters at a given dry mass than either sun or control branches. In fact, the slope for sun and control branches (0.47 ± 0.03, 0.48 ± 0.02, respectively) and the intercept for sun and control branches (2.48 ± 0.07, 2.51 ± 0.06, respectively) differed remarkably little.
The differences in the size of growth increments were associated with different reproductive output (Table 3). Sun branches produced almost exclusively female cones, whereas the average sex ratio (female: male) on shade branches for all orders of growth increments was 1:1.3. In contrast to edge trees, branches on control trees in the centre of the patch produced almost exclusively male cones. The number of male cones was slightly lower in branches of centre trees than in shade branches of edge trees. In total, edge trees produced much more cone dry mass (32.6 g) than control branches (0.4 g).
| Tree position | ||||||
|---|---|---|---|---|---|---|
| Edge | Centre: Total | |||||
| Light environment of branch | ||||||
| Sun | Shade | |||||
| Order of growth increments | Female | Male | Female | Male | Female | Male |
| 1 | 2 | 0 | 2 | 0 | 1 | 3 |
| 2 | 32 | 2 | 20 | 11 | 0 | 10 |
| 3 | 42 | 1 | 1 | 15 | 2 | 9 |
| 4 | 3 | 0 | 0 | 3 | 0 | 0 |
| 5 | 0 | 0 | 0 | 1 | 0 | 0 |
| All | 79 | 3 | 23 | 30 | 3 | 22 |
Although the absolute length was higher in sun branches (Fig. 4), the length per unit dry mass (Fig. 7) of growth increments was longer in shade branches, because the (positive) response of dry mass to increased light availability was greater than the (positive) response of length. This difference was significant as far back as 1986, i.e. in growth increments produced 3 years before the actual light exposure in 1989 (Fig. 7a,b). Again, growth increments of control branches had an intermediate specific length after light exposure, i.e. there were no overall differences in mean values between edge and centre trees.
Specific length (length per unit dry mass) of growth increments produced by Pinus sylvestris branches from 1985 to 1993 for (a) first- and (b) second-order growth increments.
Discussion
The present study reveals considerable plasticity of yearly growth increments (long shoots) that make up P. sylvestris branches. Individual growth increments on sun branches of edge trees responded to increased light by increased length, branching (vegetative reproduction), dry mass and sexual reproduction. As a result, sun branches of edge trees overall were bigger than shade branches. Thus, resource acquisition of branches was increased at the level of vegetative growth increments and this in turn led to increased sexual reproduction. Among all the variables investigated, there was only one, namely survival probability (over 1 year; Table 1), that was not increased in sun branches.
At the tree level, i.e. averaging over sun and shade branches of edge trees and averaging over the two control branches of centre trees, there were remarkably few differences because shade branches of edge trees had consistently smaller (reduced length, dry mass, branching intensity) growth increments than control branches although the former received, if anything, more light because of their location close to the edge. This consistent difference between the shade branches of edge trees and centre trees was interpreted as ‘negative’ physiological integration between branches of edge trees. Such a correlative inhibition, in which sun branches seem to obtain resources that, had the light environment been homogenous, would have been available to the shade branches of edge trees, has previously been observed in herbaceous plants (Novoplansky et al. 1989; Prati et al. 1997).
If we first consider the observed growth responses in the context of strategies based on experiments with clonal plants (Schmid 1986; Hutchings 1988; de Kroon & Hutchings 1995) and simulations (Sutherland & Stillman 1988), the following conclusions can be drawn. (i) The predicted decrease in length of growth increments (e.g. resulting in a beneficial accumulation of feeding sites) in response to increased light availability was not observed. On the contrary, sun branches of edge trees produced longer growth increments after light exposure (cf. Figure 4). (ii) The prediction of increased branching frequency under higher light conditions was supported. (iii) Branching angles, which are not expected to be altered according to foraging theory, were highest in shade branches of edge trees. Over all, the growth increments within branches of P. sylvestris responded to the contrasting (at the branch level, but heterogeneous at the tree level) light environment in a way that was consistent with a growth null hypothesis (de Kroon & Hutchings 1995; Cain et al. 1996; Stoll et al. 1998). That is, local resource availability affected the growth of the plant at the level of growth increments. Nevertheless, the branching angles of shade branches of edge trees responded in a way consistent with ‘active’ foraging because it enhanced the probability that foliage was placed in localities with higher light availability (see below).
Viewed as transport systems for the feeding sites, the aerial branches of trees explore and place the needles in space (analogous to stolons or rhizomes in herbaceous plants). While there has been considerable work regarding how light interception and height growth can be optimized in homogeneous environments at the tree or even population level (Mäkelä 1985; Mäkelä & Sievänen 1992) and with respect to the successional stage of communities (cf. Henry & Aarssen 1997), the link to foraging strategies through morphological plasticity in heterogeneous light environments has rarely been made, although trees leave a much longer record of their foraging movements in the form of woody branch systems than herbaceous plants do. In clonal herbaceous plants the decrease in the length of growth increments predicted by plant foraging theory is only rarely observed, and normally absent in plagiotropic spacers like rhizomes (Sutherland & Stillman 1988; Dong 1994; Hutchings & de Kroon 1994; Stoll et al. 1998). Therefore, de Kroon & Hutchings 1995) suggested that plasticity in feeding sites might be more important than plasticity of (clonal) spacers, especially if spacers serve as storage organs. In contrast to herbaceous plants, trees offer yet another possibility. It is well known that the needles of conifers not only acquire but also store most of the food that contributes to subsequent shoot growth (Kozlowski 1964). For example, in P. resinosa, girdling of the terminal leader precluded phloem translocation of reserves in needles more than 1 year old to the terminal buds or to expanding shoots, but it allowed these reserves as well as those in the phloem to be used in proximal parts of the branch. When the contribution of the 1-year-old branches (exclusive of needles) was accounted for, the resources supplied by 1-year-old needles alone contributed to almost two-thirds of the total elongation of the next year's leader (Kozlowski 1964). This means that a relatively tight feed-back between spacers and feeding sites connects the current year's shoot growth to resource acquisition in the previous year. Thus, the growth increments in sun branches of P. sylvestris were very probably longer simply because of higher light availability.
Branching frequency is usually considered to be the most consistent plant foraging response (de Kroon & Hutchings 1995) and was also found in 13 of the 14 studies used by Sutherland & Stillman (1988) to test predictions of their simulation. However, an increased branching frequency does not reject the growth hypothesis (de Kroon & Hutchings 1995) if it can be shown that branching frequency is a function of dry mass and the relationship is the same in the sun and in the shade. We found increased branching that was a result of two processes, i.e. higher production of growth increments in sun compared with shade branches and, although there was no difference in mortality between sun and shade branches (prediction vi in the Introduction), reduced mortality compared to control branches. When we analysed the relationship between the dry mass of a growth increment and the number of daughters it produced, we found that the relationship was the same in sun and shade branches. That is, the increase in branching frequency was not greater as would be expected from the increased dry mass of growth increments.
The third architectural parameter, the branching angle, showed an apparent ‘bending’ of growth increments of shade branches of edge trees towards the incoming light from the edge (higher branching angles than sun or control branches). The most probable interpretation for this response would be phototropism (Aphalo & Ballaré 1995; Ballaréet al. 1995). Generally, phototropic responses have been demonstrated to result in the projection of shoots towards light gaps (Novoplansky et al. 1990; Ballaréet al. 1992). Because it is often possible to form compression wood, the cessation of branch elongation in trees does not necessarily mean that branches lose their ‘mobility’ (Wilson & Archer 1981).
To summarize the results of our study in the foraging context, we could not reject the growth null hypothesis but have to reject the foraging hypothesis (see also Silvertown & Gordon 1989; Oborny 1991; Huber & Stuefer 1997), at least in its original formulation. Nevertheless, as the biomass allocation patterns (predictions iv, v and vii in introduction) and sexual reproduction (viii) suggest, architectural or morphological plasticity did occur in response to increased light availability, clearly shown by the specific length of growth increments.
Branch autonomy and dynamic competition
The differences in terms of biomass and reproduction between sun and shade branches of edge trees were not surprising. What was surprising, however, was the observation of longer and heavier growth increments of control compared to shade branches after 1989, despite the somewhat higher light level in shade branches of edge trees. One possible explanation of this difference would be dynamic competition (Sachs & Novoplansky 1997) between branches within edge trees. Such correlative inhibition has been demonstrated in peas (Pisum sativum) by Novoplansky et al. (1989). When these authors grew two connected shoots in different light conditions, the shaded shoot was inhibited and eventually even withered and died. It elongated and became etiolated only if the shoot in the high light was removed.
This compensatory physiological integration contradicts, or at least questions, the hypothesis of branch autonomy (Sprugel et al. 1991), which seems to be restricted to periods outside spring when young and vigorous growth increments are not actively elongating. For example, Sprugel & Hinckley (1990) exposed the foliage of individual branches of Abies amabilis to 14CO2 before shoot growth and harvested the trees 3 or 30 days later. In trees harvested before bud break, labelled carbon was found in the trunk both above and below the labelled branch. No 14C was found in any branches other than the labelled ones. However, in trees harvested after shoot expansion, newly produced foliage was labelled everywhere on the tree. Thus, especially during the very active time of shoot elongation in spring (with a daily rhythm in which more elongation occurs at night than during the day; Kozlowski 1964) competition for carbohydrates and nutrients among shoots may be important, especially in older trees (Wareing 1956, cited in Kozlowski 1964).
Branching patterns and angles in woody plants play important roles in establishing both crown form and leaf position (Fisher & Hibbs 1982; Fisher 1986; Ceulemans et al. 1990). In addition, most species show varying degrees of phenotypic plasticity in their branching parameters, and a tree's capacity to respond opportunistically to heterogeneous environments may well be more adaptively significant than its inherited or deterministic form (Fisher 1986). Although the size and slow growth of trees make experimental studies on the structural parameters of branching and their phenotypic plasticity more difficult than for herbaceous plants, a more general empirical basis that includes trees could help to elaborate theories of foraging strategies in plants. It might be a useful conceptual framework for future research to compare vertical and horizontal growth strategies in herbaceous and woody plants.
Acknowledgements
We thank H.P. Rusterholz for field assistance, M. Schwarze, D. Schneider, P. Lindemann-Matthies, and E. Schreier for help at harvest, and J. Weiner, C. Lavigne, H. de Kroon, T. Hara, L. Haddon and an anonymous referee for helpful comments on earlier versions of the manuscript. Financial support by the Swiss National Science Foundation, grant no. 31–30041.90 to B. Schmid, and a postdoctoral fellowship from the Roche Research Foundation and the Swiss National Science Foundation to P. Stoll, are gratefully acknowledged.
References
Received 7 May 1997revision accepted 28 April 1998
