1. Lianas face a dilemma: how can they achieve a balance between the benefits they gain from light capture in their host canopies and the risk of falls resulting from the deleterious effects they have on the growth and survival of their host trees? To address this issue, we investigated leaf distribution patterns, canopy dynamics and the impact of four liana species on the growth of their hosts.
2. In the forest canopy, the majority of the leaves of Actinidia arguta (Actinidiaceae) received >80% irradiance relative to the canopy top. The leaf mass and the length of the canopy framework of this species increased linearly with time after it had reached the forest canopy (estimated from the number of growth rings in the main liana stem at 8 m height). In contrast, a much lower percentage irradiance was received by leaves of the three other species, Celastrus orbiculatus (40–80%, Celastraceae), Schisandra repanda (<40%, Schisandraceae) and Schizophragma hydrangeoides (<20%, Hydrangeaceae). In these species, canopy sizes did not change markedly with time. Species that intercepted more light acquired a larger number of host trees.
3. Growth-ring widths of the host trees of A. arguta and C. orbiculatus were smaller than those of liana-free trees; this difference was not significant in the two species that intercepted less light. The length of the basal stem between the rooting point and the point of attachment to the current host tree was greater in species that intercepted more light, suggesting the successful movement of these lianas to new hosts following the death of previous host trees.
4.Synthesis. Lianas have various ecological strategies for resolving their dilemma. They may be aggressive and rapidly spread in host canopies, intercepting much light, but reducing the risk of falls by acquiring many host trees to balance their top-heavy architecture. Alternatively, they may be commensal, whereby small liana canopies in lower positions in their host canopies acquire less light, but do not negatively affect the current hosts. Such variations reflect niche differentiation among species, and could be an important mechanism underlying the diversification and coexistence of liana species.
While many studies have investigated liana–tree interactions at the community level, we have limited information on how individual lianas or individual liana species interact with their host trees. Past studies reported that leaf distribution patterns varied among liana species (Castellanos 1992; Caballé 1993; Avalos et al. 2007). Some species place most of their leaves on the top of the host canopy, whereas others distribute leaves in patches within the host canopy (Castellanos 1992). Some form a dense, thick foliage canopy while others form a relatively open canopy with a thin leaf layer (Avalos et al. 2007). These differences in leaf distributions alter the light acquisition and productivity of lianas and have an impact on the growth and survival of the host trees. Host trees are also affected by the extent of crown infestation by liana canopies (Clark & Clark 1990; Ladwig & Meiners 2009; Ingwell et al. 2010). Lianas that rapidly expand larger crowns will do more harm to their host trees than those with smaller crowns; however, if the former lianas spread into many tree crowns, the negative impacts on individual host trees could be reduced and, if a host tree dies, a liana may have a reduced risk of falling to the ground as it has support from other trees. Thus, presumably lianas could have at least two contrasting strategies to compensate for their non-self-supported growth habit. One is to develop a small canopy in a lower or inner position of the host canopy, which will reduce the negative impact on the host trees but decrease light acquisition. The other is to extend and disperse a crown rapidly into many trees and thereby reduce the risk of falling.
In the present study, we investigated the leaf distribution patterns, canopy growth and number of host trees of four deciduous liana species in a cool-temperate forest in Japan. Leaf distribution patterns were evaluated by measuring the light availability on the liana canopy relative to that on the top of the host canopy. We also examined the impact of the lianas on the growth of their host trees. To acquire information about mortality of host trees in relation to liana growth, we examined the length of the basal part of the liana stems that hung from the host crown and trailed on the ground, which increases when previous host trees have died and the liana has successfully moved to new hosts. Based on these evaluations, we discuss the strategic significance of liana behaviours in the host canopies.
Materials and methods
Study site and species
The study was conducted in the Yakushidake National Forest [36°43′ N, 139°31′ E; c. 1000 m above sea level (a.s.l.)] in Nikko, Japan. The mean annual temperature and precipitation over the last 30 years were 6.7 °C and 2103 mm, respectively, at the Okunikko weather station (1293 m a.s.l.; 3.3 km north of the site), and 10.1 °C and 2187 mm, respectively, at Nikko Botanical Gardens (647 m a.s.l.; 7.0 km east of the site). The forest is in a cool-temperate zone and dominated by deciduous trees including Fagus japonica, Quercus crispula, Betula grossa, Carpinus japonica and Acer spp. A large part of the forest was approximately 15 m high with 50- to 80-year-old canopy trees (the ages were determined from the numbers of growth rings in core samples from 100 canopy trees that were used for evaluating the impact of lianas on the radial growth of host trees; described later). The following liana species were examined: Actinidia arguta (Siebold et Zucc.) Planch. ex Miq. (Actinidiaceae), Celastrus orbiculatus Thunb. (Celastraceae), Schisandra repanda (Siebold et Zucc.) Radlk. (Schisandraceae) and Schizophragma hydrangeoides Siebold et Zucc. (Hydrangeaceae). These are the most frequently occurring liana species at the site and all are widely distributed in East Asia. Actinidia arguta, C. orbiculatus and S. repanda climb hosts with twining stems, and S. hydrangeoides climbs with adhesive, adventitious roots. Root climbers usually do not extend to additional trees from the current host (Hegarty 1991); thus, S. hydrangeoides was expected to show an extreme case of liana strategies in which growth is restricted to a single host tree. All of the species shed their leaves in the winter (from November to April at our site). Large plants of C. orbiculatus often produce sprout shoots from long-extended below-ground organs (either roots or rhizomes, but not determined; R. Ichihashi, pers. obs.). This type of vegetative propagation from roots and stolons was not observed in the other species.
Current-year shoots of the liana species consist of climbing (searcher) and ordinary shoots (Ichihashi, Nagashima & Tateno 2009). Climbing shoots have the ability to twine (in stem twiners) or generate adventitious roots (in root climbers) and thereby acquire external support, whereas ordinary shoots are free-standing, short leafy shoots that do not twine or develop adventitious roots. Stem elongation ends in all species by late summer and new shoots are produced sympodially in spring. In subsequent years, climbing shoots produce both ordinary shoots and new climbing shoots on their stems, whereas ordinary shoots produce only new ordinary shoots (R. Ichihashi, pers. obs.). Successive productions of climbing shoots on former climbing shoots make structures that form the framework of a liana canopy (hereafter termed ‘axis’), whereas successive productions of ordinary shoots on former ordinary shoots make free-standing, fine-branched structures with dense foliage (termed ‘twig’).
Leaf distribution patterns
Leaf distribution patterns were evaluated by comparing light availability on the liana canopy with that on the host canopy. The measurements were conducted between 2006 and 2008, from late August to early October of every year, when stem elongation and leaf expansion had terminated. Leaf fall began after the sampling period. Five samples of each liana species were chosen from plants in the forest canopy, those that were not seriously damaged or senesced, those with main host trees that were not markedly shaded by neighbouring trees and those with main host trees that did not support other lianas in the canopy. The basal diameters (measured at 1.3 m from the rooting point; Gerwing et al. 2006) of the samples of A. arguta, C. orbiculatus, S. repanda and S. hydrangeoides were 2.8–5.8, 3.5–7.9, 2.8–4.6 and 3.2–7.7 cm, respectively; the age ranges (estimated from the numbers of growth rings at the 1.3 m) of the samples were 35–60, 33–71, 43–89 and 40–90 years, respectively. For three plants of A. arguta, which spread across several tree crowns, light environments were measured for the parts of their canopies that were located in and around the canopy of the main host tree (more than half of the whole liana canopy was included in the measurement). The remaining parts of these liana canopies were collected later for evaluation of canopy growth.
We climbed host trees using ropes to access the canopies. The canopy of each liana was divided into groups of spatially clustered foliage. Each of the spatial foliage groups were further divided into two or more groups if light environments varied within the initial group (e.g. when a part of the foliage group was shaded by a tree branch). When a cluster of foliage was continuously spread in a wide area without variations in the light environment (e.g. when the foliage was spread on the surface of the host canopy), the foliage was divided into groups with projected areas of 2 × 2 m, or 2 m in length if the foliage was spread linearly. We obtained 10–20 foliage groups per liana plant. Photosynthetically active photon flux densities (PPFDs) were measured at the centre of every foliage group and also on the top of the host canopy. The measurement was conducted under overcast conditions. The PPFDs were initially expressed as values relative to those at a nearby open site (instantaneous %PPFD; Messier & Puttonen 1995; Machado & Reich 1999), and then, these PPFDs were recalculated as values relative to those on the top of the host canopy. An extendable rod with a light sensor at the top was used to make the measurements on the liana canopies. At the measurement sites, the light sensor was set horizontally on the foliage. The light sensor consisted of a photodiode (S1133; Hamamatsu Photonics, Shizuoka, Japan) calibrated with a commercially manufactured light sensor (LI-190; LI-COR, Lincoln, NE, USA). We recorded the time of day at which each PPFD measurement was made to match it with the coincident PPFD measured at the open site. PPFDs at the open site were measured beside a nearby river with the LI-COR sensor and recorded automatically every minute with a data logger (LI-1000; LI-COR). Following PPFD measurements, foliage from each group was collected and the dry mass of the leaves was measured as described in the next subsection.
To test whether plant size affected the measured light environments, we examined relationships of the median and mode [the %PPFD class (when divided by the 20% interval range) that contained the largest proportion of the foliage] light environment to liana basal diameter. Since no significant correlations were found for either parameter in any species (data not shown), we did not consider plant size variation in further analyses.
Canopy growth of lianas
The canopy growth of lianas was evaluated in the five plants for which PPFDs were measured, and in one or two additional smaller plants (1.8–2.6 cm in basal diameter) of each species. Six plants were collected for A. arguta and C. orbiculatus, and seven plants for S. repanda and S. hydrangeoides. This sampling was conducted simultaneously with the PPFD measurements.
The basal stem of each liana was cut at a height of 8 m, which represented the lower boundary of the forest canopy; the part distal to the cut was regarded as the ‘canopy’ of the liana and used in the analysis. We counted the number of annual rings at the cut under a stereomicroscope (×30) after trimming the cross-sectional surface of the stem with razors. The number of rings represented the years since the liana reached the forest canopy and was identified as the ‘canopy age’. When the basal stem of a liana branched below 8 m, all branches were cut at a height of 8 m and all parts distal to the cuts were treated as the canopy. The number of rings was counted for every branch at a height of 8 m; the largest value was used as the canopy age.
In the laboratory, all current-year shoots were detached from sample plants. The stem lengths of all climbing shoots were measured to represent the total current-year extension length. The mean and maximum lengths of individual climbing shoots were also evaluated: the length of climbing shoots determines the potential distance that the liana can reach from its present position (Hegarty 1991; Putz 2010), and is thus used as a parameter related to the ability of lianas to reach new host trees. In stem-twining species, shoots with twining or coiled stems were regarded as climbing shoots. In the root-climbing species S. hydrangeoides, shoots with adventitious roots on the stems were regarded as climbing shoots. For all climbing and ordinary shoots, leaf and stem masses were determined after drying to a constant weight at 80 °C. Petioles were considered to be supporting organs and thus included in the stem category. In four smaller plants of S. hydrangeoides, a part of their leaves (8–27% of the whole leaf mass) was located below 8 m in height. Since we aimed to evaluate dynamics of lianas in the forest canopy, those leaves were excluded from the analyses. Results were largely unaffected by whether those leaves were included or excluded (data not shown).
For stems that were not current-year shoots, the lengths of the axes (stems forming the framework of a canopy, originating from former climbing shoots) were measured to represent the spatial scale of the canopy. For stem twiners, sympodial units (i.e. stem segments that had extended in a single year) that were <1 m long were regarded as twigs (fine-branched, free-standing structure; originating from former ordinary shoots), and thus excluded from the measurement. For the root climber S. hydrangeoides, stems that generated adventitious roots were regarded as the axes. The masses of all the stems were determined after drying to a constant weight at 80 °C.
Number of host trees per liana
In March 2006, we established a 2.6-ha study plot (six subplots of 0.3–0.6 ha each) at a relatively flat site along a stream where canopy lianas were abundant. For every liana >2 m in height, the basal diameter of the main stem was measured 1.3 m from the rooting point. In November 2006, the number of host trees for lianas with basal diameters >2 cm (canopy lianas) was determined visually from the forest floor. At the time of the measurements, most leaves had fallen from the forest canopy and we could easily follow the range of liana canopies. Any tree that was attached by more than a single stem of a target liana was regarded as a host.
Effects of lianas on the radial growth of host trees
To evaluate whether lianas affect the radial growth of host trees, we compared annual ring widths between host trees and control trees that bore no lianas in the canopy. Host trees that met the following criteria were used: trees that made up the forest canopy and were not markedly shaded by surrounding trees, trees that supported only one liana plant in the crown and where most of the liana canopy was supported by the tree, trees with basal stem diameters of 15–30 cm at 1.3 m from the ground that hosted lianas with basal stem diameters of 2.5–6 cm (these were the typical sizes of canopy trees and lianas in the site) and appropriate control trees that were found nearby. Control trees were the same species of canopy tree as the target host with a basal diameter ±10% that of the host. Controls were required to stand within 50 m of the host tree in an area with the same topology (e.g. in the middle of a slope, a relatively flat area or along a stream). We obtained 10–14 pairs of host and control trees for each liana species. Tree species were mainly B. grossa, C. japonica, Carpinus cordata, F. japonica and Acer amoenum. At first, comparisons were made in a pairwise manner, whereby all host–control pairs of the various tree species were analysed for each liana species. Then, comparisons were made among the hosts of each liana species and the controls of the same tree species. We chose the two dominant tree species, B. grossa and C. japonica, for these within-tree-species analyses. We collected extra host samples of these species regardless of the lack of controls nearby; controls from the pairwise analyses were used. In all analyses, basal diameters of neither lianas nor their host trees differed significantly among liana species (P > 0.05, one-way anova; mean ± SD of basal diameters of lianas and hosts used in the pairwise analyses are shown in Table 2); thus, plant size was not included in further analyses.
Table 2. Increments of annual ring widths of host trees of the liana species and the control trees during a 3-year period (2002–2004). The increments were compared between hosts and controls using a pairwise t-test (one-tailed test for Host < Control). Sample sizes (n) indicate the number of pairs of host and control. Neither basal diameters of lianas nor basal diameters of host trees differed significantly between liana species (P > 0.05, one-way anova)
Basal diameter (mean ± SD, cm)
Ring width increment (mean ± SD, mm)
4.13 ± 1.03
21.0 ± 5.2
2.29 ± 0.75
3.96 ± 1.53
4.66 ± 1.01
21.6 ± 5.0
2.83 ± 1.48
4.01 ± 1.41
3.96 ± 0.49
22.7 ± 4.2
3.45 ± 1.34
3.83 ± 1.36
4.04 ± 0.74
22.3 ± 2.6
3.50 ± 1.16
3.65 ± 1.06
In 2006 and 2007, we collected two 5.15-mm-wide increment cores with a borer from each target tree at a height of 1.3 m. At the laboratory, the surfaces of the core samples were trimmed with razors so that the annual rings were clearly visible under a stereomicroscope (×30). We measured the increments of annual ring width during a recent 3-year period (between 2002 and 2004) to the nearest 0.1 mm. For each tree, the mean of the values for the two core samples was used in the analyses.
Mortality of host trees in relation to liana growth
The basal part of the main stem of many lianas hangs from the host crown and trails on the ground, caused mainly by the death of former host trees. When a previous host tree (or a part of the host crown that supported the liana) dies and disappears after the liana has extended to a new host, the basal part of the stem that was once attached to the dead host hangs from the new host’s crown (Putz 2010). If a liana has passed through many host trees before reaching its current host tree, the basal stem length between the rooting point and the attachment point to the current host tree (SLroot-host) becomes longer. SLroot-host increases when a host tree dies but the liana avoids falling by acquiring new hosts in advance. SLroot-host is affected by two factors: the mortality of the host trees and the probability that the lianas will avoid falling when the host tree dies (the latter factor may parallel the ability of the lianas to acquire new hosts). We note that an increase of SLroot-host resulted not only from the death of a whole host tree but also from the death of host branches that supported lianas. Thus, this parameter may not directly reflect the mortality of whole host trees, but is useful for suggesting interspecific variations in impacts on host survival and the ability to avoid falling. Also, trailing on the forest floor (Peñalosa 1983, 1984) and self-supporting height growth (Putz 1984; Caballé 1998) before lianas start climbing may increase SLroot-host. However, none of the present liana species showed extensive trailing before climbing, and the self-supporting juveniles were 1 m in height at most (Ichihashi, Nagashima & Tateno 2010). Thus, growths before climbing had little effect on the SLroot-host of the canopy lianas.
Measurements were conducted in 2007 for lianas more than 2 m in height in the 2.6-ha study plot. For each liana plant, we determined (i) the original rooting point, (ii) the point at which the basal stem detached from the ground and (iii) the point at which the stem first attached to the host tree. The basal stems of lianas that trail on the ground were often buried in the litter layer and generated adventitious roots. We defined the position at which the stem completely intruded into the soil after the litter was removed as the original rooting point (i). Some lianas twined around the stem of a tree and extended not to the canopy of the first tree but to that of a different tree. In these cases, point (iii) was determined to be the position at which the liana first twined around a stem (i.e. the stem of the first tree), regardless of whether the liana also infested the canopy of the same tree. We measured the lengths of the stems that trailed on the ground between (i) and (ii). We approximated the stem lengths between (ii) and (iii) from the sum of the horizontal distance between the two points and the height of (iii). The height of (iii) was measured using a scaled rod when the height was less than 8 m, and the triangular method (i.e. to estimate the height of the target position from the horizontal distance and elevation angle from a measurement position) when the height was more than 8 m. SLroot-host was determined by the sum of the lengths between the points.
Leaf distribution patterns and canopy growth of lianas
The light environments in the liana canopies varied depending on the species (Fig. 1). In A. arguta, a large proportion of the leaves received more than 80% PPFD relative to the top of the host canopy, leaves of C. orbiculatus received 40–80% PPFD, leaves of S. repanda received < 40% PPFD and leaves of S. hydrangeoides received <20% PPFD. The light environments reflected the leaf distribution patterns of each species. Actinidia arguta tended to display its leaves on the upper surface of the host canopy, whereas S. repanda and S. hydrangeoides mainly distributed leaves within and under their host canopies, respectively, and rarely developed their canopies near the top of the host canopies (Fig. 2). Celastrus orbiculatus distributed leaves relatively evenly from the inside to the surface of the host canopy.
The canopy mass of the lianas increased with canopy age (i.e. years since the lianas reached the forest canopy; Fig. 3a). In A. arguta, which displayed leaves higher in the host canopy, a greater increase in canopy mass with canopy age occurred than in the other species (P < 0.01, equality of slopes of the regression lines analysed with Bonferroni post hoc multiple comparisons). Leaf mass, axis length and current-year extension length also increased linearly with canopy age in A. arguta. These parameters were not correlated with canopy age in the other species, except for the leaf mass of S. hydrangeoides, which gradually increased with canopy age (Fig. 3b–d). Thus, the canopy mass, leaf mass, axis length and current-year extension length of A. arguta were initially similar to those of the other species, but became much greater in older plants. These canopy traits were similar among the other three species, except that C. orbiculatus had a significantly larger canopy mass than S. repanda at the same canopy age (P < 0.001, ancova with Bonferroni post hoc multiple comparisons), the mean axis length was greater in S. repanda than in S. hydrangeoides (P < 0.05, anova with a Tukey–Kramer post hoc test) and the mean current-year extension length was larger in C. orbiculatus than in the other two species (P < 0.05, anova with a Tukey–Kramer post hoc test). Individual climbing shoots were longer in A. arguta and C. orbiculatus than in the other two species; long climbing shoots of the former species exceeded 3 m in length, whereas they were 1.0–1.5 m at most in the latter (Table 1).
Table 1. Lengths of individual climbing shoots of the liana species. Values are means ± SD of the means and maximum values per plant (n = 6 for A. arguta and C. orbiculatus, n = 7 for S. repanda and S. hydrangeoides). Differences between species were tested using one-way anova with a Tukey–Kramer post hoc test; values with different letters are significantly different (P < 0.05)
165 ± 46a
180 ± 40a
73 ± 9b
37 ± 16b
314 ± 71a
308 ± 56a
116 ± 31b
72 ± 17b
The number of host trees per liana (> 2 cm in basal diameter) increased with the basal diameter of A. arguta and C. orbiculatus (Fig. 4), but was not correlated with the basal diameter of the other species (P > 0.05). The slopes of the regressions for A. arguta and C. orbiculatus did not differ significantly (P = 0.08), although the relatively small P-value suggests that A. arguta tends to spread to more host trees as it grows in the forest canopy. The mean number of host trees per liana decreased in the order A. arguta [3.88 ± 1.93 (mean ± SD)], C. orbiculatus (2.56 ± 1.05), S. repanda (1.59 ± 0.75) and S. hydrangeoides (1.00 ± 0.00; each liana had a single host tree; Fig. 4). Thus, the species that displayed leaves higher in the host canopies had more host trees.
Effects on radial growth of host trees
In the pairwise comparisons, host trees of A. arguta and C. orbiculatus showed significantly smaller ring widths than control trees, but the differences were not significant for S. repanda and S. hydrangeoides (Table 2), although the relatively small P-value for S. repanda (0.08) suggests that host trees of this species also tend to have smaller ring widths than controls. In the comparisons of ring widths between hosts and controls within the same tree species, ring widths were smaller in the hosts of A. arguta than in the control trees of B. grossa (Fig. 5a), and in the hosts of A. arguta and C. orbiculatus than in the control trees of C. japonica (Fig. 5b). Overall, host trees of A. arguta and C. orbiculatus tended to show decreased diameter growth, but the tendency was not clear in hosts of S. repanda and S. hydrangeoides.
Mortality of host trees in relation to liana growth
Basal stem lengths between the rooting point and the attachment point to the host tree (SLroot-host) increased with the basal diameter of the lianas in A. arguta, C. orbiculatus and S. repanda, but were not correlated in S. hydrangeoides; most plants of S. hydrangeoides rooted just below the host trees regardless of the plant size (Fig. 6). Slopes of the regressions for log-transformed data [i.e. the relationship between log (SLroot-host) and log (basal diameter)] of the three species did not show significant differences (P > 0.05), but intercepts did (P < 0.001 for every pair of the species, ancova with Bonferroni post hoc multiple comparisons). Thus, SLroot-host decreased from A. arguta, C. orbiculatus, S. repanda, to S. hydrangeoides in plants of the same basal diameters, which paralleled the height order (higher to lower) of their canopy locations.
Leaf display patterns and canopy growth as strategies to resolve the dilemma associated with non-self-supporting growth
The main locations for leaf display varied among the liana species, from the well-lit upper surface of host canopies to the shaded subcanopy (Figs 1 and 2). Actinidia arguta spread and kept expanding on the well-lit surface of its host canopies (Fig. 3); therefore, after reaching the forest canopy, it acquired many host trees (Fig. 4). The other species spent decades in the forest canopy with few changes to their leaf masses and axes lengths (Fig. 3b,c). However, the sample plants of these species produced climbing shoots up to tens of metres in length (Fig. 3d), and a comparable amount of extension growth occurred every year; canopies of these species may be maintained by a continuous local turnover of shoots and branches. Climbing shoots of A. arguta and C. orbiculatus often exceeded 3 m in length, whereas those of the other species were much shorter (Table 1). The length of climbing shoots is a critical factor affecting ability to extend to neighbouring trees (Hegarty 1991; Putz 2010). The long climbing shoots of A. arguta helped it to spread into many host crowns. Celastrus orbiculatus also increased the number of host trees after reaching the forest canopy (Fig. 4), even though the apparent axes lengths did not show marked change (Fig. 3c). The leaf mass of the A. arguta canopy was 5.5 kg; it was 2 kg at most in the other species (Fig. 3b). The typical ranges of the basal diameters (15 ≤ D ≤ 30 cm) and heights (12 ≤ H ≤ 20 m) of the canopy trees at the site were assigned to an allometric equation: 1/Wl = 2320/D2H + 0.032, which was developed as a regression for deciduous trees in a cool-temperate forest in Japan (Ogino 1977). As a result, the leaf mass (Wl) of the canopy trees was estimated at 1.1–6.2 kg; A. arguta had a comparable amount of leaves as its host trees. The location of the leaves of A. arguta on the upper surface of the host canopies indicated that A. arguta was in strong competition with its hosts for light. The other species had a similar or smaller amount of leaves than the small canopy trees; therefore, the impact of these lianas on light acquisition by host trees was smaller than that of A. arguta.
Actinidia arguta and C. orbiculatus were located in better-lit, higher positions in the host canopies, and the ring widths of their host trees were smaller compared to those of the controls (Table 2). Schisandra repanda and S. hydrangeoides, which distributed leaves in shadier positions within the host canopies, did not have significant effects on the ring widths of their hosts (Table 2; Fig. 5). In addition to light interception by liana canopies, physical interactions between lianas and the host canopies can cause damage and degradation (Clark & Clark 1990; Schnitzer, Kuzee & Bongers 2005). This becomes severe if the lianas locate higher in the canopies because host shoots and twigs in the surface layer of the canopies are thin and susceptible to physical damage. The reduced growth of the host trees may have resulted from competition with lianas for below-ground resources (Dillenburg et al. 1993; Pérez-Salicrup & Barker 2000; Schnitzer, Kuzee & Bongers 2005; Toledo-Aceves & Swaine 2008). However, most of the host trees stood metres away from the rooting points of the lianas, and the distances were larger in the species that showed negative effects on host growth (mean horizontal distance to the host trees in A. arguta, C. orbiculatus, S. repanda and S. hydrangeoides were 8.7, 6.3, 3.1 and 0.2 m, respectively). Although distinguishing between the effects of above-ground and below-ground interactions is difficult in the field (Ingwell et al. 2010), our results suggest that light competition was the major cause of the negative effects of the lianas on host growth at the study site (see Clark & Clark 1990; van der Heijden & Phillips 2009; Ladwig & Meiners 2009). The study site was located at the bottom of slopes near a stream where water and nutrient availabilities are generally high (Roy & Singh 1994; Gessler et al. 2000). Thus, below-ground resources may not be a strong limiting factor at the site.
Among the four species of liana, S. repanda and the root climber, S. hydrangeoides, distributed most of their leaves under the shade of host canopies, and scarcely intercepted light on the surfaces of host canopies. In these species, spatial scales and leaf masses showed little change for decades in the forest canopy. Radial growths of the host trees were not significantly affected and their shorter SLroot-host (Fig. 6) suggests that they associated with fewer host trees (only one host in S. hydrangeoides) through their lifetimes than the other species. Their behaviours could be regarded as a strategy close to commensalism, in which they do not benefit much from light acquisition but instead reduce the risk of falls caused by the death of host trees. Schizophragma hydrangeoides is a root climber, in which adventitious roots adhere only after prolonged contact with a surface (Darwin 1867). It is probable that this climbing mechanism restricts the growth of S. hydrangeoides to a position inside the canopy of a single host tree: its climbing shoots (shoots that generate adventitious roots) may be unable to attach to thin host twigs and shoots in the top layer of the canopy or to the surfaces of other trees that are discontinuous from the present position, perhaps because the surfaces they come in contact with are unstable due to wind and/or the self-weight of the climbing shoots. On the other hand, S. repanda is a stem twiner and does not have such an apparent growth restriction associated with its climbing mechanism.
In contrast, A. arguta was a more aggressive competitor; it intercepted a large amount of light by displaying many leaves on the surface of its host canopies. Although this habit can result in suppression of the growth and survival of host trees, it enables the liana to greatly expand its canopy and spread into many tree crowns, which will distribute the leaves and the negative effects of the liana into many trees. In addition, if one of the host trees falls, the liana is less likely to fall due to the support offered by the other host trees. Compared to the other species, A. arguta had the greatest SLroot-host; although this implied high host tree mortality, it was also indicative that A. arguta can survive the loss of individual host trees. These data indicate that A. arguta benefits from canopy-level light acquisition, and the resultant risk of a fall is dispersed by spreading on the forest canopy.
Celastrus orbiculatus showed intermediate characteristics between A. arguta and S. repanda. This species distributed leaves from the inside to the surface of host canopies, and its canopy remained at a similar, relatively small scale for decades. Among the species studied, only C. orbiculatus showed extensive vegetative reproduction by below-ground organs (either roots or rhizomes). Large plants of C. orbiculatus in the forest canopy had these organs that extended horizontally just below the ground for more than 10 m; many sprout shoots were generated from the organ (R. Ichihashi, pers. obs.). Thus, C. orbiculatus spread along the forest floor after reaching the top of the forest canopy, rather than spreading aerially within the canopy.
Consequences of liana regeneration strategies
In previous work, we investigated habitat preferences and shoot production patterns in juveniles (< 8 m tall) of five liana species, including all four species of the present study species (Ichihashi, Nagashima & Tateno 2010). The aggressive species, A. arguta and C. orbiculatus, were more frequent at the forest edges than in the understorey, suggesting that these species mainly regenerate at the forest edges (and probably in sites with comparable environmental conditions, e.g. large gaps). The commensal species, S. repanda and S. hydrangeoides, were more abundant in the forest understorey. Among the species, biomass allocation to leaves and climbing stems showed a strong negative correlation: the aggressive species allocated a larger proportion of shoot biomass to climbing stems (because they produced long climbing shoots; cf. Table 1), whereas the commensal species allocated more to leaves. The result suggests that there may be a trade-off between the ability to search for hosts and/or well-lit environments (climbing stems) and current productivity (leaves).
In disturbed sites such as forest edges, plant size increases but density decreases rapidly due to intense competition among plants (Niklas, Midgley & Rand 2003; van Breugel, Martínez-Ramos & Bongers 2006). For lianas growing within such unstable vegetation, the ability to rapidly acquire new hosts is important; A. arguta and C. orbiculatus reach the forest canopy by repeatedly moving to the next tree by means of long climbing shoots. In the forest canopy, A. arguta remains an aggressive competitor and extends to the surfaces of many tree crowns. Celastrus orbiculatus becomes less aggressive in that it does not markedly expand its own canopy; however, it still intercepts much light in the forest canopy and negatively affects host growth. These species may be physiologically adapted to well-lit regeneration habitats and, if so, assimilation will be significantly suppressed in shady environments. For them to locate in the well-lit surface layer of the forest canopy and acquire many host trees (in case some of the host trees die) is preferable by means of long climbing shoots.
In contrast, the availability of light resources is limited, and the dynamics of growth and death of plants are less active in the forest understorey: shade-tolerant saplings and small trees can persist for long periods in the understorey with little growth (e.g. Lorimer 1980; Canham 1985; Kobe et al. 1995). For lianas growing in such environments, increasing current productivity by having many leaves is more important than increasing searching ability. The root climber, S. hydrangeoides, grows along the boles and branches of a single host tree. Schisandra repanda, a stem twiner, also grows slowly within the crown of an individual host tree. When they reach the top of the current host, S. hydrangeoides can only continue to grow with subsequent growth of the host, but S. repanda can extend to neighbouring trees. In cases where S. repanda does not find the next host within the range of their climbing shoot length, it persists at the present position for many years (in one case, a liana of this species did not show apparent height growth for more than 12 years, judging from bud scale scars on liana shoots at the top position, on a host sapling of 1.5 m in height; R. Ichihashi, pers. obs.). After reaching the forest canopy, S. repanda remained in shady positions in the canopy, similar to the root climber. The canopy lianas, however, received more light (Fig. 1c,d) than the understorey individuals, which typically receive less than 5% full sun (Ichihashi, Nagashima & Tateno 2010). Since photosynthesis of shade-tolerant species generally comes close to the maximum rate at relatively low light levels (e.g. Björkman 1981; Lambers, Chapin & Pons 2008), receiving further light by displaying their leaves on the well-lit surface of the forest canopy may not greatly enhance their productivity. In addition, the short climbing shoots of S. repanda are not suitable for acquiring many host trees. Therefore, for this species to remain under the shade of the forest canopy and reduce any negative effects on their current host trees is adaptive. In S. repanda, the climbing shoots of canopy lianas (Table 1) were shorter than those of juveniles [mean and maximum climbing shoot lengths of S. repanda juveniles were 108 ± 28 and 190 ± 26 cm, respectively (n = 5; Ichihashi, Nagashima & Tateno 2010); P < 0.05 for both parameters, t-test]; thus, canopy lianas reduced their searching ability compared to those in the active climbing stage of growth.
This study is the first to demonstrate great variations among liana species in leaf distribution patterns, canopy dynamics and their impacts on host trees. We suggest that these variations reflect different strategies for achieving a balance between the potential benefits of light acquisition and the possible risks associated with falls and injuries. Previously, most studies emphasized the harmful effects of lianas on trees in the context of forest dynamics or management. Our results suggest that lianas vary from aggressive to commensal in the way they relate to their host trees. Such variation, if commonly found in liana communities, is important because it may represent niche differentiation among species, and could be an important mechanism underlying the diversification and coexistence of liana species; in addition, this variation may increase the complexity of forest structures and dynamics, especially in tropical forests where lianas contribute 20% of the stem density and species composition, and 30% of the leaf area, in woody plants (Putz 1983; Gentry 1991).
We thank H. Nagashima, I. Terashima, K. Kitajima and two anonymous referees for useful comments on this manuscript, E. Kasuya for helpful advises on data analysis, and T. Tani, S. Nakayama, K. Sawakami, A. Miyashita, D. Sugiura and R. Minamino for valuable discussion and field assistance. We also deeply appreciate E. Ito for all the support on this work. This work (R.I.) was supported in part by Global COE Program (Integrative Life Science Based on the Study of Biosignaling Mechanisms), MEXT, Japan.