Architectural differences associated with adult stature and wood density in 30 temperate tree species


  • Masahiro Aiba,

    Corresponding author
    1. Tomakomai Research Station, Hokkaido University Forests, Takaoka, Tomakomai, Hokkaido 053-0035, Japan; and
      *Correspondence author. E-mail:
    Search for more papers by this author
  • Tohru Nakashizuka

    1. Graduate School of Life Sciences, Tohoku University, Aoba 6-3, Aramaki, Aoba-ku, Sendai 980-8578, Japan
    Search for more papers by this author

*Correspondence author. E-mail:


  • 1Tree architecture is a major determinant of performance, such as height growth, light capture, and mechanical stability. Studies both in temperate and tropical forests have shown significant architectural differences associated with adult stature and light demand.
  • 2However, studies in temperate forests have not been as thorough in examining these relationships with respect to phylogeny and ontogeny, thus preventing a complete understanding of the patterns in temperate forests and limiting comparisons of the relationship between tropical and temperate forests. Therefore, we performed a community-level analysis of the relationship between tree form and ecology in a temperate forest with statistical consideration of phylogeny and ontogeny.
  • 3The height–diameter relationship throughout tree development was asymptotic in most species. Crown diameter and depth increased allometrically with tree height, with no asymptote. The tree height, crown diameter, and crown depth of small trees were estimated using these relationships and were similar to those reported for tropical species.
  • 4Taller species had more slender stems at any reference size and narrower crowns at small reference sizes, whereas crown depth was relatively independent of adult stature. Light-wooded species had narrower and shallower crowns at medium to large reference heights. Stem thickness was virtually independent of wood density throughout the size range.
  • 5Our results support the hypothesis that the architecture of short or shade-tolerant species is optimized for light capture and mechanical stability, whereas that of tall or light-demanding species is optimized for height growth. These relationships generally agree with results from studies in tropical rain forests, although considerable differences exist, and may potentially promote the stable coexistence of the species.


Partitioning of horizontal and vertical light gradients by tree species is an important mechanism that enables the stable coexistence of tree species in a forest (e.g. Kohyama 1992; Kobe 1999). Coexistence of tree species that differ in adult stature and light demand is observed in many forests, suggesting the global significance of this mechanism. Niche partitioning is necessarily accompanied by interspecific differences in functionally important traits, and many studies have attempted to identify the traits underlying the stable coexistence of tree species along light gradients. Plant architecture, as well as physiological traits, has been a major candidate for such traits (e.g. Horn 1971; Givnish 1988).

Studies on seedlings and saplings have mainly discussed the functions of biomass distribution and leaf traits (e.g. Kitajima 1994), whereas studies on larger trees have focused on stem and crown architecture, which may determine tree height growth, mechanical stability, and light capture. Various hypotheses concerning the associations between these architectural traits and tree ecology have been proposed. For example, whereas shorter tree species can reproduce in lower, shaded layers of forests, canopy species initiate reproduction only after reaching upper, brighter layers (Thomas 1996b). As a result, the traits essential for successful regeneration are assumed to be efficient height growth in taller species but effective assimilation and high survival under shaded conditions in shorter species (Kohyama et al. 2003). Therefore, taller species are expected to have slender stems and narrow crowns that minimize extension costs. In contrast, shorter species should have thick stems that are mechanically strong and wide, shallow crowns that enhance their ability to intercept light. Similarly, light-demanding species must grow rapidly to maintain a leading position in forest gaps and should thus have slender stems and narrow crowns compared with shade-tolerant species, which place preference on light capture (Givnish 1988). Light-demanding species are also expected to have long crowns with several leaf layers to maximize light interception under high radiation (Horn 1971).

Recently, these hypotheses have been examined in intensive studies focusing on numerous sympatric species at multiple growth stages in tropical rain forests (Kohyama et al. 2003; Poorter et al. 2003; Bohlman & O’Brien 2006; Poorter et al. 2006; Osunkoya et al. 2007). Although earlier reports focusing on these hypotheses are available (e.g. Kohyama 1987; King 1990), the recent studies are valuable, because the overall relationships between architecture and ecology may be sensitive to the species studied (Poorter & Werger 1999) and ontogeny (Sterck & Bongers 2001). The number of species examined must be large enough to allow a general understanding of the ecological roles of architecture at the community level. Analyses that consider the ontogeny of tree species are also essential, because tree architecture varies substantially with growth, possibly to cope with the changing environment and life stage. The results of these recent studies are relatively consistent and generally support the hypotheses described above. As predicted, taller species had slender stems (Kohyama et al. 2003; Poorter et al. 2003; Bohlman et al. 2006; Poorter et al. 2006) and narrow crowns (Poorter et al. 2003; Bohlman & O’Brien 2006; Poorter et al. 2006, but see Osunkoya et al. 2007) at least at some growth stages; however, contrary to predictions, taller tree species had more shallow crowns (Kohyama et al. 2003; Poorter et al. 2003; Bohlman et al. 2006; Poorter et al. 2006). Similarly, light-demanding species had slender stems (Poorter et al. 2003; Bohlman et al. 2006) and narrow (Poorter et al. 2006) but shallow crowns (Bohlman et al. 2006; Poorter et al. 2006).

Research relating tree architecture to ecology also has a long history in temperate forests (e.g. Kohyama 1987; King 1991; Aiba & Kohyama 1997; Millet et al. 1999). These studies have shown that the relationships between tree architecture and ecology more or less support the above hypotheses and that such relationships play a potentially important role in temperate forest dynamics. Other studies have found that environmental factors characteristic of temperate forests, such as strong winds (King 1986) and snow cover (Homma 1997), are also important in understanding the relationship between tree form and ecology. However, most of these studies, especially those in deciduous forests, were not as intensive as the more recent studies in tropical rain forests: community-level studies are rare, and analyses were performed only on a single growth stage. This situation has made it difficult to thoroughly understand the roll of tree architecture in temperate forests and to compare the results of tropical forests with that of temperate forests. In this study, we comparatively analysed the tree allometry and architecture of 30 sympatric species in a Japanese temperate deciduous forest and examined the relationships among architecture and two important axes representing tree growth strategies: maximum height (Hmax) and wood density. Wood density is a good predictor of species’ light demand, growth and survival in some forests (Williamson 1975; King et al. 2005; van Gelder et al. 2006; Poorter et al. 2008). Furthermore, wood density directly affects tree architecture as an important material property (McMahon 1973). Wood density is strongly correlated with modulus of elasticity and rupture, which enable the production of thinner structures of the same strength (van Gelder et al. 2006). However, this advantage may be negated, because dense wood adds extra weight for a given length of trunk and branch (Valladares 1999). Actual relationships between tree architecture and wood density have only been tested in a few studies. For example, Sterck et al. (2006) demonstrated that dense-wooded species have wider crowns than light-wooded species in a Bolivian rain forest. In Malaysian rain forests, King et al. (2006) showed that dense-wooded species have thinner trunks.

To increase the generality of our results, we performed ontogenetic analyses using phylogenetic independent contrasts (PIC; Felsenstein 1985), which reduce the risk of spurious correlations arising from evolutionary conservation and phylogenetically biased species selection. Based on previous studies, we specifically expected, when species are compared at a given size, that (i) to tolerate strong winds and snow cover, temperate tree species have thicker trunks and/or narrower crowns compared to tropical species, (ii) taller species have thinner trunks and narrower crowns to increase height growth rates, and (iii) dense-wooded species have wider crowns because of the biomechanical potential and ecological requirements required for efficient light capture, similar to tropical species. We did not make predictions regarding the relationship between stem slenderness and wood density, because both positive and negative relationships are possible.


study site

This study was conducted in a 6-ha permanent plot in the 100-ha Ogawa Forest Reserve (OFR), located in the southern part of the Abukuma Mountains, central Japan (36°56′ N, 140°35′ E, elevation 610–660 m). Mean monthly temperature is 9·0 °C, with the highest monthly mean of 20·5 °C in August and the lowest of –1·6 °C in February. Annual precipitation is 1750 mm, and in winter, snow depth occasionally reaches 50 cm. OFR is an old-growth, cool temperate deciduous forest. Quercus spp. and Fagus spp. dominate the canopy layer, and Acer spp. and Carpinus spp. are abundant in the subcanopy layer. In the 6-ha study plot, more than 50 woody species (diameter at breast height [DBH]≥5 cm) have been recorded (Masaki 2002). The canopy height is ca. 20 m, and some tall trees reach 30 m (Tanaka & Nakashizuka 1997). The reserve has been free of human disturbance for at least 80 years. The structure and dynamics of the community have been described in detail by Nakashisuka & Matsumoto (2002).

data collection

We focused on 30 common tree species for which individuals from a broad size range were available (Table 1). These species belong to 12 families, made up 96% of the basal area recorded in the plot, and differed in adult stature and light demand. Several light-demanding species, e.g. Castanea crenata, Betula grossa, and Quercus spp. successfully regenerate only after large-scale disturbances, and their populations are currently diminishing (Nakashisuka & Matsumoto 2002). For each species, 22–32 individuals (804 in total) with DBH ≥1 cm were randomly selected from the entire size range. For some light-demanding species that had failed to regenerate in the plot, we collected data on small individuals from the surrounding secondary forests. Trees with obvious damage or abnormal leaning were rejected.

Table 1.  A list of the 30 tree species included in the study with their maximum height, wood density and size distribution index (SDI)
FamilyGenusSpeciesAuthorityMaximum Height (m)Wood Density (g/cm3)SDI
AceraceaeAcerrufinerveSieb. et Zucc.18·10·45−0·006
AceraceaeAcertenuifolium(Koidz.) Koidz.15·70·58−0·061
AraliaceaeAcanthopanaxsciadophylloidesFranch. et Savat.17·80·32−0·040
BetulaceaeBetulagrossaSieb. et Zucc.21·90·55−0·007
BetulaceaeCarpinuslaxiflora(Sieb. et Zucc.) Bl.17·50·58−0·063
ClethraceaeClethrabarvinervisSieb. et Zucc.11·60·47−0·035
CornaceaeBenthamidiajaponica(Sieb. et Zucc.) Hara9·10·66−0·052
CornaceaeSwidacontroversa(Hemsl.) Soj.20·50·47−0·016
FagaceaeCastaneacrenataSieb. et Zucc.22·90·440·003
FagaceaeQuercusserrataThunb. ex. Muuray27·00·61−0·011
OleaceaeFraxinuslanuginosaKoidz. f. serrata (Nakai) Murata14·60·59−0·034
RosaceaePrunusverecunda(Koidz.) Koehne22·00·61−0·026
RosaceaeSorbusalnifolia(Sieb. et Zucc.) C. Koch20·20·60−0·063
RosaceaeSorbusjaponica(Decne.) Hedlund16·80·53−0·048
SabiaceaeMeliosmamyrianthaSieb. et Zucc.12·50·37−0·058
StyracaceaeStyraxjaponicaSieb. et Zucc.9·90·56−0·025
StyracaceaeStyraxobassiaSieb. et Zucc.15·90·51−0·038

Tree architecture was measured in November 2006. The angle to the top leaf, lowest leaf, stem base, and the distance to the tree were measured using a laser range finder (TruPulse, Laser Technology) to evaluate tree height and crown depth (vertical distance from the lowest leaf to the top leaf). Lower, small branches derived from adventitious buds were only considered part of a crown when bearing a non-negligible number of leaves (c. 1% of all leaves). Crown width in two directions at right angles (including the widest point) was also measured, and crown area was approximated as an ellipse using these two measurements. For small trees, height, crown depth, and width were directly measured using a measuring pole or tape measure. DBH was measured using a tape measure or callipers at a height of 1·3 m.

To determine wood density, cores were taken from three trees per species using a borer in June 2007. DBH of the sampled trees was ca. 20 cm (smaller or larger individuals were included if individuals of this size were not available; range, 6–28 cm), and the sampling height was 1–1·5 m. The lengths of cores were roughly equal to half the diameter of the trees. The resulting holes were filled with synthetic resin. Cores were stored in plastic bags until returned to the field station and were then trimmed into cylinders. The length and volume of the cylindrical cores were calculated from the length and inner diameter (0·515 cm) of the borer (Muller-Landau 2004). After oven-drying at 60 °C for 4 days, wood density was calculated as oven dry mass/fresh volume and was averaged for each species. To verify the validity of wood density as an index of light demand, we calculated Spearman's rank correlation between wood density and the size distribution index (SDI). SDI is the third moment of the DBH distribution around the midpoint of the DBH range (Masaki 2002). A large SDI indicates that the population structure is biased toward large individuals and thus the species is possibly light-demanding (Wright et al. 2003). Wood density ranged from 0·32 g cm−3 for Acanthopanax sciadophylloides to 0·66 g cm−3 for B. japonica (Table 1). Wood density was weakly negatively correlated with the SDI in OFR (Spearman's rank correlation, r = −0·26, P = 0·21), indicating that the population structure of light-wooded species tended to be biased toward large individuals and that these species are currently regenerating unsuccessfully in OFR. This negative correlation was especially strong when two oak species, which are failing to regenerate in OFR even though their wood densities are relatively high, were excluded (r = −0·42, P < 0·05). The exclusion of the two oak species had limited effects on subsequent analyses and thus we only presented results that include all species. These results indicate that wood density is a reasonable indicator of light demand of the species.

data analysis

For convenience of comparison, we followed the analytical procedure of Poorter et al. (2006) with some modifications. For diameter–height relationships, an equation with an asymptote, H = Hmax[1 –exp(−a[DBH]b)], rather than a regular allometric equation such as H=a[DBH]b, often provides a better fit (where H is tree height, Hmax is the asymptotic maximum height, and a and b are constants). To determine the equation that best explained the diameter–height relationships for each species, we first regressed height against DBH and DBH2. A significantly negative coefficient for DBH2 suggests that the asymptotic equation is more appropriate for a species. Based on the result, tree height was regressed against diameter using either the asymptotic equation or the regular allometric equation. Nonlinear regression (in the asymptotic equation) or the standardized major axis (SMA) method after log-transformation (in the regular allometric equation) was used to estimate the parameters. SMA is more suitable than type I regression for describing the relationships between two variables that are not free from error (Warton et al. 2006). Because a few species did not show a clear asymptote, we used height at the maximum observed DBH predicted by species-specific regression equations, rather than the estimated asymptote, as the maximum height. However, the heights were virtually identical to the estimated asymptotes for most species with asymptotes, and thus this Hmax would be very close to the actual Hmax in most species, although the actual maximum DBH may be larger than the observed maximum DBH.

There were no asymptotes for crown area or depth, and thus these were analysed using the allometric equation. DBH and tree height were the best predictors of crown area and crown depth, respectively. However, for concordance with Poorter et al. (2006), we analysed the height–crown area and height–crown depth relationships. The relationships between height and crown architecture were assessed using SMA after log-transformation of both variables. The interspecific differences of the relationships were tested using the common slope test of the SMATR package available in the R statistical environment (Warton 2007).

Based on the species-specific regressions, tree height at 1-cm DBH intervals and crown area and depth at 1-m height intervals were calculated. For this purpose, type I regression, rather than SMA, was used. Partial correlations between these values and Hmax (with wood density as a covariable) and wood density (with Hmax as a covariable) were calculated using both raw values and PIC when >10 species were available at a given size. Partial correlation was used to separate the effects of Hmax and wood density on tree architecture. PIC analysis helps confirm the related evolution of the traits and thus reinforces the results. The phylogenetic tree of the 30 species used in the PIC calculations (see Fig. S1 in Supporting Information) was constructed using the web application Phylomatic (Webb & Donoghue 2005) and previously published data (Ackerly & Donoghue 1998; Suh et al. 2000; Yoo & Wen 2002). The genus Ostrya was treated as an inner group of the genus Carpinus, based on Yoo & Wen (2002). All branch lengths were set at 1, and a polytomy was resolved by inserting very short branches to calculate PICs. The PICs were calculated using the APE package for the R environment (Bolker et al. 2007). Although we also calculated the bivariate correlation between architecture and Hmax or wood density, the results were not considerably different from those using partial correlation; thus, we only present results of the partial correlation. All statistical analyses were performed using the software package R version 2·50 (R Development Core Team 2007).



The diameter–height relationships were significantly nonlinear for 29 of the 30 species. The exception was the second-smallest understorey species, Clethra barvinervis. The relationship was asymptotic in most species, except for Magnolia obovata and C. barvinervis (Fig. 1a). The goodness of fit of the asymptotic equation to the observed diameter–height relationships was high (mean of r= 0·88, range = 0·76–0·97). The Hmax predicted for the largest observed diameter ranged from 9·1 m for Benthamidia japonica to 27·0 m for Quercus serrata (Table 1). The height at 15 cm DBH varied from 8·6 m for B. japonica to 15·0 m for Betula grossa (cf. Fig. 2a).

Figure 1.

Allometric relationships in the architecture of 30 temperate tree species. (a) Height vs. DBH, (b) crown area vs. height, and (c) crown depth vs. height. Each line represents a species and the ranges of lines reflect the observed size range of each species.

Figure 2.

Relationships between tree architecture (height, crown area, and crown length) and Hmax over all size ranges. Bivariate relationships for trees of 15 cm DBH (for height) or 10 m height (for crown area and depth) are shown in the left panels (a, d, g). Partial correlation coefficients (with wood density as a covariable) for different reference DBH (for height) or height (for crown area and depth) are shown in the middle panels (b, e, h). The right panels (c, f, i) show the same analysis using phylogenetic independent contrasts. Asterisks in the left panels indicate significant correlations (*P < 0·05, **P < 0·001). In the middle and the right panels, solid symbols indicate significant correlations (P < 0·05) and open symbols indicate nonsignificant correlations. Solid lines indicate the number of species used in the calculations for each reference height. All the species were included in the analysis only at intermediate reference sizes because small saplings of some light-demanding species and large individuals of understorey species were not available.

Crown area increased allometrically with tree height (Fig. 1b). The regular allometric equation fit well for many but not all species (mean r2 = 0·66, range = 0·12–0·88). The slopes of the height–crown area relationships for the 30 species were significantly different (test for common SMA slope, likelihood ratio = 140, d.f. = 29, P < 0·001). Crown area at maximal height ranged from 7·8 m2 for C. barvinervis to 216 m2 for Fagus crenata. The crown area at 10 m varied from 2·6 m2 for Castanea crenata to 33 m2 for B. japonica (cf. Fig. 2d).

Crown depth also increased allometrically with tree height (Fig. 1c). The allometric equation fit well for most species (mean r2 = 0·69, range = 0·32–0·89). The slopes of the height–crown depth relationships for the 30 species were significantly different (test for common SMA slope, likelihood ratio = 84, d.f. = 29, P < 0·001). Crown depth at maximal height varied among the 30 species from 3·5 m for Styrax japonica to 16 m for Fagus japonica. Crown depth at 10 m ranged from 2·0 m for C. crenata to 7·0 m for F. japonica (cf. Fig. 2g).

relationships between architecture and hmax

Figure 2 summarizes the partial correlation coefficients between the three architectural traits (tree height, crown area, and crown depth) and Hmax (using wood density as a covariable) for trees of different reference diameter (for height) or height (for crown area and depth).

Tree height at 15 cm DBH was positively correlated with Hmax, indicating that species with larger Hmax have slender trunks at this size (Fig. 2a). The tendency was consistent throughout the size range, and the partial correlation coefficients gradually increased from 0·50 at 2 cm DBH to 0·87 at 56 cm DBH (Fig. 2b). The pattern of the partial correlation coefficients between tree height and Hmax for different DBH sizes was quite similar even in the PIC analysis (Fig. 2c). Therefore, species with larger Hmax have slender trunks, independent of both ontogeny and phylogeny.

Crown area was negatively correlated with Hmax, indicating that species with larger Hmax have narrower crowns at small reference heights (3–10 m, Fig. 2d,e). However, the strength of the correlation consistently decreased and was not significant at larger reference heights (≥11 m). The pattern was similar in the PIC analysis (Fig. 2f).

The correlation between crown depth and Hmax was never significant, although the value was consistently negative at intermediate reference sizes (Fig. 2g,h). The pattern was similar in the PIC analysis, although the negative correlation in larger individuals (16–17 m) was occasionally significant (Fig. 2i).

relationships between architecture and wood density

A significant negative correlation between stem slenderness and density was found at small reference DBH (2–4 cm), whereas the correlation was consistently near zero for larger trees (Fig. 3a,b). When phylogeny was taken into account, the correlation coefficient was more sensitive to which species were included in the analyses and was occasionally significantly negative (Fig. 3c).

Figure 3.

Relationships between tree architecture (height, crown area, and crown length) and wood density over all size ranges. Bivariate relationships for trees of 15 cm DBH (for height) or 10 m height (for crown area and depth) are shown in the left panels (a, d, g). Partial correlation coefficients (with Hmax as a covariable) for different reference DBHs (for height) or height (for crown area and depth) are shown in the middle panels (b, e, h). The right panels (c, f, i) show the same analysis using phylogenetic independent contrasts. Asterisks in the left panels indicate significant correlations (*P < 0·05). In the middle and right panels, solid symbols indicate significant correlations (P < 0·05) and open symbols indicate nonsignificant correlations. Solid lines indicate the number of species used in the calculations for each reference height. All the species were included in the analysis only at intermediate reference sizes because small saplings of some light-demanding species and large individuals of understorey species were not available.

The correlation between crown area and wood density was significantly positive at all but the smallest reference heights (Fig. 3d,e), indicating that species with heavy wood tend to have larger crowns. Although the correlation was similarly strong at intermediate to large reference heights in cross-species analyses, it was stronger at larger reference heights in the PIC analysis.

The correlation between crown depth and wood density gradually increased from –0·11 at 4 m reference height to 0·56 at 19 m reference height, and the relationship was significant at larger reference heights (≥15 m, Fig. 3g,h). In the PIC analysis, the increment was more prominent (Fig. 3i). The correlation consistently increased from significantly negative (–0·37) for a 3-m tall sapling to significantly positive (0·83) for a 19-m tall tree. The correlation was consistently significantly positive for larger reference heights (≥15 m).


allometry and architecture of temperate species

The diameter–height relationships were asymptotic for 28 of the 30 species (Fig. 1a), in contrast to results from tropical forests, where generally one-fourth of the studied species do not exhibit clear asymptotic relationships (Thomas 1996a; Chave et al. 2003; Poorter et al. 2006). These contrasting results may be explained by compositional differences between tropical rain forests and temperate forests. Generally, fewer single-stemmed understorey species occur in temperate forests, although bushy shrubs, which were excluded from this study, are abundant (Turner 2001). This pattern also occurs in OFR, and all of our study species reach at least 9 m (Table 1), whereas many species without asymptotes are small trees with Hmax < 7 m in Poorter et al. (2006). Heights at 5 cm DBH and 30 cm DBH were positively but weakly correlated (r = 0·37), indicating that the rank order of stem slenderness changes to a certain extent with growth. This shift in slenderness may occur because understorey species initiate reproduction earlier and afterward grow slowly in height, whereas canopy species must continue to grow in height to reach the canopy. Height at 15 cm DBH was similar to that of Liberian (Poorter et al. 2003) and Bolivian (Poorter et al. 2006) species, whereas the range was slightly wider in tropical species (cf. Fig. 2a). The crown architectures at 5 and 20 m height were strongly positively correlated (crown area: r = 0·69, crown depth: r = 0·81). Therefore, the rank order of crown architecture is highly maintained throughout ontogeny, although there is some crossover. Both crown area and depth of 10 m tall trees were similar to that of Liberian (Poorter et al. 2003), Bornean (Kohyama et al. 2003), and Bolivian (Poorter et al. 2006) rain forest species, whereas the crown area of a Liberian species (64 m2 at 10 m height) was substantially larger than that in other forests.

Although we expected temperate tree species in OFR, which are subject to strong winds (including typhoon-related winds) and occasional snow cover during winter, to have thicker stems and smaller crowns to reduce the risk of breakage, the trunk and crown architecture of temperate forest trees did not differ substantially from that of tropical forest trees. This result may imply that interspecific competition and evolutionary pressure for diversification, rather than responses to the physical environment, determine the architectural range of forest trees. On the other hand, canopy heights of Japanese temperate forests including OFR rarely exceed 30 m (cf. Table 1, Fig. 1) and are generally much lower than that of tropical rain forests at least partly because of the environmental differences. Therefore, wind and snow load may be not so severe in lower layers even in temperate forests and only affect architecture of canopy trees, which are directly exposed to the stresses.

taller species have more slender stems and narrow crowns

The partial correlations between slenderness, i.e. tree height for each DBH, and Hmax were consistently positive and significant as well as independent of both ontogeny and phylogeny (Fig. 2a–c). Positive correlations between slenderness and Hmax at larger reference heights are not surprising, because shorter species retard height growth once they begin reproduction, whereas taller species continue to grow. On the other hand, positive correlations in the smallest size class, in which even understorey species have not yet reached reproductive heights, is much more ecologically interesting. We found a positive correlation between slenderness and Hmax even for small trees (DBH ≥ 2 cm), whereas such relationships were only observed in larger-sized individuals in tropical forests (King 1996; Thomas 1996a; Bohlman et al. 2006; Poorter et al. 2006) and an evergreen temperate forest (Aiba & Kohyama 1997). This result suggests that small species allocate more to radial growth, possibly to reduce the risk of stem breakage, at the expense of slow height growth and delayed achievement to reproductive height. In contrast, taller species with slender trunks, whereas they realize rapid height growth, are expected to suffer higher mortality from stem breakage. This architectural and possibly strategic differentiation at earlier life stages may be due to the severe environment specific to temperate forests, e.g. snow cover in winter. Future studies should test whether this relationship is common in temperate forests at other sites.

Crown area was negatively correlated with Hmax only at small to medium reference heights (3–10 m, Fig. 2d–f), as found in tropical studies (Bohlman et al. 2006; Poorter et al. 2006). Although the correlation shifted to being significantly positive at large reference heights in Poorter et al. (2006), it was nearly zero in our study. Considering that the rank order of crown area is rather stable through ontogeny, as noted above, such size dependence of the relationship is mostly attributable to differences in the species analysed rather than to ontogenetic changes in crown architecture. Namely, a negative correlation between crown area and Hmax is found only when comparing the shortest understorey species with canopy species. Intensive lateral expansion would enable understorey species to capture more light, even though they might exhibit inferior height growth.

The correlation between crown depth and Hmax was negative for medium-sized trees but rarely significant (Fig. 2g–i), whereas in tropical forests, this relationship is consistently negative for medium-sized trees (Kohyama et al. 2003; Poorter et al. 2003; Bohlman et al. 2006; Poorter et al. 2006). Our result also differs from the observed significant positive correlations for medium-sized trees in a warm temperate evergreen forest (Aiba & Kohyama 1997). Considering that significant negative correlations have only been found in equatorial forests, and long crowns are advantageous for capturing lateral light (Kuuluvainen 1992), the relationship between crown depth and Hmax may depend on latitude. Negative correlations in tropical rain forests may be explained as follows: in the tropics, where trees receive most sunlight from the zenith, light captured by older, lower branches may be limited and thus sustainment of the branches may be costly for juveniles of taller species growing to the canopy. On the other hand, in the temperate zone, long crowns may be more efficient and not be disadvantageous even for taller species. Although crown depth of the OFR species was similar to that of tropical species, the correlation would be weak if some temperate canopy species adopt longer crowns.

dense-wooded species have wider and longer crowns

The partial correlation between slenderness and wood density was generally weak and occasionally significantly negative (Fig. 3a–c). Although wood density can either positively or negatively affect stem strength, biomechanical studies in tropical forests have shown that dense-wooded species can have more slender stems of the same strength as a result of strong positive correlations between wood density and modulus of elasticity (King et al. 2006; van Gelder et al. 2006). On the other hand, from an ecological perspective, species with lighter wood should have more slender stems to maintain positions at the top of forest gaps, because wood density is negatively correlated to light demand (King et al. 2005; Falster 2006; Poorter et al. 2008). Although further analyses of the biomechanical properties of these species are essential, our result suggests that stems of light-wooded species are likely to be more vulnerable to physical damage. Because most light-wooded species are light-demanding species that regenerate only after large-scale disturbances (Nakashisuka & Matsumoto 2002; cf. Table 1), it seems reasonable for them to preferentially invest in height growth to outcompete neighbouring individuals, despite the risk of stem breakage. This result is consistent with reports from tropical rain forests that species requiring well-lit conditions have more slender trunks (Poorter et al. 2003; Bohlman et al. 2006). In contrast, King et al. (2006) found a weak but significant negative correlation between wood density and stem volume (in other words, a positive correlation between wood density and stem slenderness) for Malaysian rain forest species. To gain a general understanding of this relationship, additional studies at other sites are required, because direct examinations have only been attempted by King et al. (2006) and this study.

Crown area was significantly positively correlated with wood density at larger reference sizes (≥5 m, Fig. 3d–f). Although evidence for correlations between crown size and light demand is limited in tropical forests (Sterck & Bongers 2001; Poorter et al. 2003; Bohlman & O’Brien 2006; but see Poorter et al. 2006), a positive correlation is reasonable from both a physical and ecological viewpoint. Sterck et al. (2006) found that dense-wooded tropical species can have longer and more slender branches of the same strength as a result of a strong correlation between wood density and the modulus of rupture, which supports our results if this pattern holds true for temperate species. Furthermore, from an ecological perspective, many dense-wooded species are shade-tolerant and thus require larger crowns to efficiently use the limited light of the understorey. In contrast, narrow crowns enhance the height growth rate of light-wooded, possibly light-demanding species. Kohyama (1987) hypothesized that species with narrow crowns are ‘optimists’, which grow faster with a higher risk of mortality and outcompete other individuals following gap creation, whereas species with wide crowns are ‘pessimists’, which grow slower with a higher survival rate, operating on the premise that shaded conditions will continue. These two types of species can stably coexist under temporally heterogeneous light environment. A positive correlation between crown area and wood density found in this study leads to a positive correlation between crown area and stem support mass. The correlation would enhance the difference in height growth rate between the two types and thereby contribute to coexistence. Additional studies of the relationship between wood density and modulus of rupture in these species would reinforce our understanding of the relationship between crown width and wood density.

Crown depth was positively related to wood density at larger reference sizes (Fig. 3g–i), contrary to the prediction that shade-tolerant species should have shallow crowns to reduce self-shading (Horn 1971). Bohlman & O’Brien (2006) and Poorter et al. (2006) also found relationships contrary to the classic prediction, and the discrepancy was explained as follows. Leaves of shade-tolerant species, the light-compensation points of which are relatively low, can maintain a positive carbon balance even under partially shaded conditions. Thus, the development of a deep, multilayered crown by retaining old branches with partially shaded leaves maximizes carbon gain of shade-tolerant species. In contrast, light-demanding species should discard lower, shaded branches, the carbon balance of which becomes negative with growth. This hypothesis coincides with our finding that the positive correlation between crown depth and wood density is stronger at larger reference heights. This relationship might increase the carbon gain of shade-tolerant species and thus enable them to live longer than light-demanding species.


Our community-level analysis of tree architecture in a temperate forest shows that the stem slenderness and crown architecture of small trees were similar to those in tropical forests, even though the physical environment and constituent taxa greatly differ. Furthermore, linkages between tree architecture and ecology were generally similar to those in tropical forests. The architecture of taller species and of light-wooded, and possibly more light-demanding, species is optimized for height growth, whereas that of shorter and dense-wooded species maximizes light capture and tolerance to damage at least at some reference sizes. A significant positive correlation between maximum height and stem slenderness in small saplings, which may indicate an adaptation of understorey species to snow cover, is a unique trend that has been never found in the tropics. Another distinction is the lack of a negative correlation between maximum height and crown depth in temperate OFR species. Lower sun angle in temperate forests may be responsible for this difference. All these relationships potentially structure the temperate forest and promote the stable coexistence of species along horizontal and vertical light gradients.


We thank Dr. H. Tanaka and other members of the Forestry and Forest Products Research Institute for the opportunity to conduct research in the Ogawa Forest Reserve and for their advice in the field. Helpful comments were provided by Dr. L. Poorter and two anonymous reviewers. This study was supported by a Research Fellowship for Young Scientists from the Japan Society for the Promotion of Science. This study complies with the current laws of Japan.