Comparative ecology of 11 sympatric species of Macaranga in Borneo: tree distribution in relation to horizontal and vertical resource heterogeneity



1 Horizontal and vertical heterogeneity of resource availability, coupled with the specialized use of resources by tree species, results in complex patterns of tree species distributions in tropical rain forests. We studied the horizontal and vertical distributions of 4014 individuals in 11 species of early successional Macaranga (Euphorbiaceae) in tropical rain forest in Sarawak, Malaysia.

2 The horizontal distribution of individual trees was assessed with respect to crown light levels, establishment microsites, and broader scale variation in soil textural properties. Vertical distribution was assessed using an allometric approach to estimate maximum tree height (Hmax) and the slope of the sapling height–diameter relationship.

3 Average light levels intercepted and the proportion of individuals in each of five crown illumination classes varied significantly among the 11 species. Species ranged from extremely high-light demanding, to quite shade tolerant. Average light levels intercepted by trees generally increased through ontogeny, but the ranking of species did not change significantly.

4 Fewer individuals of the more shade-tolerant species established on disturbed microsites, irrespective of light levels. Among the more high-light demanding species, the proportion of trees on different types of disturbed sites varied.

5 Trees of seven species were significantly more common on clay-rich soils, two preferred sand-rich soils, and two were not strongly affected by soil texture.

6 Hmax ranged from 5.5 to 31.3 m and was negatively correlated with shade tolerance among species, although among the more high-light demanding species there was a wide range of tree sizes. Among species, Hmax was negatively correlated with both the slope and y-intercept of the sapling height–diameter relationship, indicating that small-statured species (also more shade tolerant) had more slender saplings than larger statured species.

7 Heterogeneity of resource availability leads to differences in horizontal and vertical tree distribution, which are important for the coexistence of 11 Macaranga species.


Tropical tree species vary in their capacity to utilize resources, and their distribution is strongly influenced by spatial and temporal heterogeneity of resource availability (Ricklefs 1977; Whitmore 1978; Bazzaz 1983; Denslow 1987; Fowler 1988; Kohyama 1993; Bazzaz & Wayne 1994). Spatial heterogeneity of soil resources influences broad-scale patterns of tree species distribution (Vitousek 1984; Vitousek & Denslow 1986; Denslow et al. 1990; Clark et al. 1998). Within forests, heterogeneity of light availability across the gap–understorey continuum can determine local distributions (Swaine & Whitmore 1988; Bazzaz 1991). Fine-scale variation in soil conditions, for example within tree-fall gaps, can also influence species distribution patterns (Orians 1982; Putz 1983; Vitousek & Denslow 1986; Núñez-Farfan & Dirzo 1988). Furthermore, gradients in light availability down through the forest canopy influence vertical distributions (Chazdon et al. 1988; Canham et al. 1994; Thomas 1995; Aiba & Kohyama 1996). Few studies, however, have considered an array of both horizontal and vertical components of resource availability and how they might interact to influence the distribution and coexistence of sympatric tropical trees (although see Clark & Clark 1992).

Many studies of horizontal heterogeneity of resource availability have focused on the importance of light levels, and the consequent variation in tree life histories in relation to the gap–understorey continuum (Denslow 1987). Pioneer and non-pioneer tree species have been shown to differ in distribution and performance at various light levels (Augspurger 1984; Bongers & Popma 1990; Clark et al. 1993), but few studies have examined ecologically ‘similar’ groups of species such as non-pioneers (but see Clark & Clark 1992; Denslow et al. 1998), and rarely have resources other than light been considered. Kohyama (1993) suggested that incorporating both horizontal and vertical structure into models of forest dynamics increases the conditions under which stable species coexistence occurs, even when the species are all non-pioneers. We investigated the distributions of 11 pioneer species of Macaranga (Euphorbiaceae) in Borneo, in relation to a range of horizontal and vertical components of resource heterogeneity. We argue that an understanding of life-history variation and its role in the maintenance of high diversity in tropical rain forests, must involve detailed analyses of species’ responses to the full spectrum of resource heterogeneity throughout their life cycles (see also Clark & Clark 1992).

We investigated whether horizontal tree distribution was, as expected, closely correlated with canopy openings and high light environments (Brokaw 1985; Alvarez-Buylla & Martinez-Ramos 1992). We also tested whether distributions were biased towards particular physical microsites for germination and establishment, for example tip-up mounds or nutrient-rich sites (as seen by Orians 1982; Brandani et al. 1988; Uhl et al. 1988; Denslow et al. 1990; Molofsky & Augspurger 1992; Itoh et al. 1995), or to slightly broader scale variation in soil texture.

Light availability declines two orders of magnitude from above the canopy to the understorey of a tropical forest (Yoda 1974; Chazdon 1986) and access to this vertical heterogeneity is largely determined by plant size and allocation patterns to above-ground organs (Kohyama 1993; Thomas 1995; Aiba & Kohyama 1996; Zipperlen & Press 1996). Maximum tree size differs among pioneer species (Sarukhán et al. 1985; Swaine & Whitmore 1988) and although this may be indicative of different longevities and hence temporal differences in the occupation of successional sites, the relationship to resource acquisition has not been considered. We quantified maximum tree size and sapling height–diameter allometry to investigate vertical distributions of Macaranga species.



Macaranga (Euphorbiaceae) is a genus of approximately 280 species, distributed from west Africa to the south Pacific islands, with the centre of diversity in New Guinea and Borneo (Whitmore 1969). Species range from shade-tolerant understorey shrubs and small trees to high-light demanding pioneers. In Borneo, Macaranga is a conspicuous component of the secondary forest and forest-gap flora, due both to the large number of species (c. 55 recorded) and high stem densities. It is not uncommon to find five to eight pioneer Macaranga species colonizing a single large forest gap.

Eleven species, within or closely related to section Pachystemon[Macaranga beccariana Merr., M. gigantea (Reichb. f. & Zoll.) Muell. Arg., M. havilandii Airy Shaw, M. hosei King ex Hook. f., M. hullettii King ex Hook. f., M. hypoleuca (Reichb. f. & Zoll.) Muell. Arg., M. kingii Hook. f. var. platyphylla Airy Shaw, M. lamellata Whitmore, M. trachyphylla Airy Shaw, M. triloba (Bl.) Muell. Arg., and M. winkleri Pax & Hoffm.], were selected because they were abundant in the study area, and initial observations suggested that they were pioneers of either large or small gaps. Macaranga brevipetiolata Airy Shaw and M. praestans Airy Shaw, which although common are not closely related to the study group and are understorey trees (S.J. Davies, personal observation), along with several related but rare species were excluded from the study. Nomenclature follows Whitmore (1975). Herbarium vouchers of all species are deposited in Kuching (SAR) and Harvard University (A) herbaria.


Lambir Hills National Park (Lambir NP) in Sarawak, east Malaysia (4°20′N, 113°50′E), includes 6800 ha of primarily lowland, tropical mixed dipterocarp forest (MDF), ranging from near sea-level to an altitude of 465 m, on sandy to clay-rich soils (Watson 1985). The heterogeneous forest canopy is 40–60 m tall and stem turnover rates are high (Phillips et al. 1994). Floristically, Lambir NP is exceptionally diverse (Hall 1991; Davies & Becker 1996), with 1175 tree species Ð1 cm diameter at breast height (d.b.h.) recently recorded in a 52-ha survey (LaFrankie et al. 1995).

Lambir NP receives approximately 3000 mm of rainfall a year, with all months averaging >100 mm (Watson 1985). Temperatures are typical for the region (mean daily maxima c. 32 °C and mean daily minima c. 24 °C) and show no substantial seasonal fluctuations.

In 1991, a long-term research project was initiated to monitor woody species in 52 ha of lowland MDF. The methodology was similar to that used on Barro Colorado Island, Panama (Hubbell & Foster 1983), and in Pasoh Forest Reserve, west Malaysia (Manokaran et al. 1990). Thus, all individuals Ð1 cm d.b.h. were tagged, mapped to ±10 cm, identified to species and their diameters measured to ±1 mm (Condit 1995).

During the initial census (November 1991–November 1992) all Macaranga individuals Ð1 cm d.b.h. (n = 2882) were identified in the field to a morphospecies. Trees in 8 ha were subsequently checked to ensure correspondence between the morphospecies and a taxonomic species. All extant individuals of the study species in the 52-ha plot were remeasured and given a correct species identification approximately 32 months after the initial census. However, 192 dead stems could not be confidently identified to species either because there was no remaining plant material or the species’ field names had been inconsistently used (Davies et al. 1995), and they were excluded from analyses.

Seven supplementary plots were established in other areas of primary and secondary forest at Lambir NP. Individuals of all size classes (including stems ≤1 cm d.b.h.) of the study species (n = 1132) in the additional plots were mapped and measured between November 1991 and January 1992. The second census of all plots was carried out during July and August 1994.


During the initial census the d.b.h. of all trees in all plots was measured. The heights of trees from 8 ha of the 52-ha plot, and all the supplementary plots, were also measured. The d.b.h. of all extant trees was remeasured in the second census. The heights of all extant trees initially measured for height were also remeasured, and the allometry data set was supplemented by height measurements in a further approximately 30 ha of the 52-ha plot.

For trees Ð1 cm d.b.h., diameter was measured to the nearest 1 mm with a metallic forester's diameter tape. For seedlings and saplings ≤1 cm d.b.h., diameter was estimated as the mean of two dial calliper measurements at approximately 10% of stem height. Tree diameters were remeasured in 4 ha of the 52-ha plot shortly after the initial census and accuracy was found to be very high. The height of trees ≤6 m tall was measured directly to the nearest 1 cm, while taller tree heights were estimated with a clinometer (SUUNTO, Finland).


All trees in the seven supplementary plots and subsamples of trees in the 52-ha plot were assessed. The light environment of each individual's canopy was characterized by its crown illumination index (CI index) on a scale of 1–5, as in Clark & Clark (1992).

Hemispherical fish-eye photographs were taken across a wide range of canopy-openness levels with a Spiratone fish-eye lens mounted on a tripod. Black and white negatives were scanned from a video image and analysed using SOLARCALC 6.03c (Chazdon & Field 1985), and were used to calibrate the CI index (Table 1).

Table 1.  Crown illumination (CI) index calibrated against percentage canopy openness and weighted percentage canopy openness estimated from (n) hemispherical fish-eye photographs. Different letters following mean values indicate significant differences at P = 0.001 for anova on log-transformed data
CI indexCrown light environment description% canopy opennessmean (SE)Weighted % canopyopenness mean (SE)n
5Crown ± completely exposed vertically and laterally (e.g. large forest opening)35.5 (3.9) a45.3 (5.2) a11
4Crown exposed to vertical and some lateral light (e.g. medium forest gap)15.8 (1.4) b19.2 (1.9) b13
3Crown exposed to some vertical and some lateral light (e.g. small forest gap)7.4 (0.3) c9.0 (0.5) c34
2No vertical light and medium lateral light (e.g. edge of small gap)5.1 (0.4) d6.7 (0.5) d24
1No vertical and minimal lateral light (e.g. forest understorey)2.9 (0.3) e4.1 (0.4) e19

The establishment microsite of each Macaranga tree was scored as: ‘undisturbed’ = on open ground without obvious disturbance; ‘logs’ = on or having at least some of the main tree roots directly in contact with logs and/or major branches derived from a tree or branch fall; ‘tip-up’ = on or associated with a tip-up mound or root-pit; ‘landslip’ = on a landslip; or ‘alluvial’ = in an alluvial site. This method underestimates the proportions of stems on disturbed microsites because evidence of disturbance disappears over time. As we had no estimates of the relative area of forest covered by each of the establishment types, statistical comparisons were only undertaken among species for microsite type.

A soil map was constructed for the entire 52-ha plot by sampling in the centre of each of the 1300 20 m × 20 m quadrats in the plot. Soil from 5–15 cm deep was assigned to one of four texture classes based on the apparent sand content, stickiness and slipperiness of the soil (Kimmins 1987). Soil samples (n = 145) spanning all four texture classes were analysed (Table 2) using a LaMotte Soil Texture Unit (LaMotte Co., Chestertown, MD).

Table 2.  Soil texture variation in the 52-ha plot at Lambir Hills National Park, Sarawak, Malaysia. Mean (±1 SE) percentage sand, silt and clay contents determined for four field soil texture classes. Sample sizes are given in parentheses. Different letters following rows indicate significant differences among soil texture classes (G-tests; P < 0.05)
Soil texture class% sand% silt% clay
139 ± 0.8 (56)25 ± 0.8 (51)36 ± 0.7 (51) a
245 ± 1.0 (23)37 ± 1.5 (22)19 ± 1.3 (22) b
351 ± 0.8 (35)22 ± 0.7 (33)27 ± 0.8 (33) ba
464 ± 1.4 (31)23 ± 1.1 (29)13 ± 1.0 (29) c


The allometric relationships between tree height and stem diameter were assessed using a series of statistical approaches. To test for linearity of height–diameter relationships, second order polynomials were fit using least squares (LS) regression of both untransformed and log-transformed variables (Niklas 1995). Relationships were considered non-linear when the second order term was significant, and asymptotic maximum height (Hmax) was then estimated as H = Hmax*[1 − exp (−aDb)] where H is tree height in m, D is d.b.h. in cm, and a and b are allometric constants that approach values of standard allometric constants for small values of H (Thomas 1995). This model could not be resolved for three species; there were too few large individuals for M. hypoleuca, and the height of M. beccariana and M. winkleri did not appear to reach an asymptote with respect to diameter. For these species, maximum tree height was estimated as the mean height of the 10 largest trees sampled (n = 6 for M. hypoleuca). For the other eight species this method gave a similar value of Hmax to that derived from the allometric model. The asymptotic allometric model with standard errors was estimated by non-linear least squares regression using the Nonlin module in systat (Wilkinson 1990). Because there were size-dependent changes in the slope of the height–diameter relationship, the slope of the initial linear portion (5–40 mm d.b.h.) was estimated using the allometric model: log10 H = log10 B + A*log10 D, where H and D are stem height and diameter in m, and B and A are the standard allometric constants. This model was estimated using both reduced major axis (RMA) and standard least squares regression; only the RMA results are shown as the patterns were essentially the same. Due to the difficulty of statistically comparing non-linear regression parameters (Ross 1981), comparisons of non-linear parameters were made by inspecting for overlap of 95% confidence intervals. Linear allometric slopes were compared with homogeneity of slopes tests (Sokal & Rohlf 1981).



Crown light environment

The CI index was strongly correlated with canopy openness as estimated with hemispherical fish-eye photographs (Spearman rank correlation, rs = 0.82). The five CI classes represented on average significantly different levels of canopy openness (Table 1).

Both mean CI index (range 4.2–2.0) and the proportion of trees in each CI class varied significantly among species (Fig. 1), although each species occurred in at least four classes. The more high-light demanding species had mean CI indices near 4.0 (M. gigantea, M winkleri and M. hosei) and the more shade-tolerant species had mean CI indices near 2.0 (M. lamellata and M. kingii), and were at either end of a continuum of distribution with respect to canopy light environment (Fig. 1).

Figure 1.

The percentage of individuals occurring in each of the five crown illumination classes (Table 1) for 11 Macaranga species. Species are arranged in order of decreasing mean CI index, with sample size for CI index estimates in parentheses. Species are indicated by a single letter: E, M. gigantea; W, M. winkleri; O, M. hosei; Y, M. hypoleuca; L, M. triloba; B, M. beccariana; A, M. trachyphylla; V, M. havilandii; U, M. hullettii; M, M. lamellata; and K, M. kingii.

Macaranga species also differed significantly in crown light environment at different tree sizes (Table 3). All species, with the exception of the small-statured M. havilandii, had significantly greater crown light exposure for larger trees. Among the size classes, however, differences among species in average CI index were maintained (at least below 10 cm d.b.h.; Table 3). The rank order of species’ CI indices did not change substantially with increasing tree size. The more shade-tolerant species had lower CI indices across all size classes (Table 3).

Table 3.  Mean crown illumination (CI) indices for 11 Macaranga species in seven diameter size classes. Species arranged as in Fig. 1 with sample sizes (n). Probability values (P*) refer to Kruskal–Wallis non-parametric anovas; values at the ends of rows are within-species comparisons among size classes, and values at the base of columns are among-species comparisons at each size class

Tree diameter size classes
0–2 cm2–4 cm4–6 cm6–8 cm8–10 cm10–15 cmd > 15 cm
gigantea3.084.084.5134.545.015.014.77 0.004**
winkleri3.6393.8264.294.484.6104.611 0.001**
hypoleuca2.6143.8114. 0.001**
havilandii2.8982.74 0.96
kingii1.7392.3232.762.73 0.004**





Establishment microsites

There were significant differences among Macaranga species in the proportion of trees in the five establishment classes (Table 4). Establishment on disturbed sites varied between 12% and 43% of trees among the 11 species. The four most shade-tolerant species, M. havilandii, M. hullettii, M. lamellata and M. kingii (Fig. 1), had greater proportions (80–88%) of individuals growing on apparently undisturbed sites than did the more light-demanding species (57–76%). Among species, the proportion of stems in low light (CI = 1) was strongly positively correlated with the proportion of stems on undisturbed soil (Pearson, r = 0.82, P = 0.002). This might be expected since canopy disturbances that result in increased light availability often result in disturbance at the soil surface. However, there were only two species (M. winkleri and M. beccariana) where trees on undisturbed sites had a significantly lower CI index than on disturbed microsites (Kruskal–Wallis tests, P < 0.01).

Table 4.  Percentage (%) and number (n) of individuals of 11 species of Macaranga in five establishment microsites. Probability (P*) refers to a chi-square test of among-species differences in the proportion of trees in ‘undisturbed’ microsites. Species arranged as in Fig. 1

Establishment microsite

Species were differentially distributed within disturbance types. Macaranga triloba had almost 30% of its stems in alluvial or periodically inundated sites, old landslips had relatively more stems of M. hypoleuca and M. beccariana than other species, and tip-up mounds were relatively rich in M. hypoleuca and M. winkleri (Table 4).

Trees of different sizes might occur in different establishment microsites because of either differential survivorship according to establishment site, or temporal fluctuations in the availability of establishment sites. Our analyses for species with large sample sizes (data not shown) suggested that M. beccariana had significantly more large individuals in landslips (P < 0.001), perhaps due to temporal variation since such sites are rare in Lambir NP. In contrast, M. hosei, M. triloba and M. trachyphylla on disturbed sites (on logs, tip-ups and landslips) had significantly greater proportions of individuals in small or intermediate size classes than in larger size classes (P < 0.01), suggesting differential survivorship. However, larger trees could well have established after a disturbance of which no evidence remained.

Soil texture

Field-determined soil textural classes were strongly related to differences in soil sand content, increasing from class 1 to 4 (Table 2). Approximately 65% of the 1300 quadrats in the 52-ha plot were classified in the sand-rich class 4, and 11% in the clay-rich class 1 (Table 5). The distribution of nine of the 11 species of Macaranga was significantly biased with respect to soil texture class (Table 5). Seven species (M. beccariana, M. hosei, M. hypoleuca, M. kingii, M. trachyphylla, M. triloba and M. winkleri) had significantly more individuals than expected in more clay-rich quadrats, whereas M. lamellata and M. havilandii had significantly more individuals in sand-rich quadrats.

Table 5.  The spatial distribution of 11 species of Macaranga with respect to soil texture (see Table 2) variation in Lambir Hills National Park, Sarawak, Malaysia. The proportion of the 1300 52-ha plot 20 m × 20 m quadrats in the four soil classes is listed in the first row. For each species the density of individuals in each soil class is listed (trees ha−1). N is the total number of quadrats in the plot, and the total number of individuals for each species. Probabilities (P*) refer to G-tests of whether species distributions differ from expected based on the frequencies of soil types in the plot. Species arranged as in Fig. 1

Soil texture classes  
1234N P*
52-ha plot11.114.29.465.41300
gigantea1. 0.16
havilandii0. 0.007
hullettii8. 0.053

For species with larger sample sizes we tested whether there was a relationship between soil texture class and light availability (estimated from CI index). In no case (n = 7 species) was mean CI index significantly different for individuals on different soil texture classes (P > 0.1, Kruskal–Wallis tests).


Maximum tree size and the allometry of tree height and diameter varied significantly among Macaranga species (Table 6). All 11 species showed significant size-dependent changes in the slope of the tree height–diameter relationship, as indicated by quadratic regression (P < 0.05). This is due to the tendency for tree height to reach an asymptote and diameter to be indeterminate (King 1990a). However, in M. winkleri and M. beccariana the asymptote was well beyond the observed data and was therefore considered an unreasonable estimate of maximum height.

Table 6.  Tree size and height–diameter allometry in 11 species of Macaranga. (a) Estimated maximum tree height, Hmax, and the allometric exponent, β, estimated using the asymptotic allometric model (Thomas 1995). Values in parentheses are standard errors. DBHmax is maximum observed d.b.h. For species marked with an asterisk, Hmax was calculated using the mean value of larger trees as described in the text. (b) Allometry of sapling height–diameter for 11 Macaranga species in the size range 5–40 mm d.b.h. The initial slope and y-intercept of height × diameter relationships were assessed with RMA regression of log10 transformed data. Values in parentheses are 95% confidence intervals for regression coefficients. Sample size (n) and r2 are listed for both analyses. Species arranged as in Fig. 1
(a) SpeciesHmax (m)DBHmax (cm)βr2n
gigantea29.30 (5.39)42.70.84 (0.053)0.94129
winkleri*21.54 (0.66)14.7135
hosei31.32 (2.62)55.10.98 (0.066)0.91159
hypoleuca*22.46 (1.74)35.542
triloba22.39 (3.58)14.80.93 (0.051)0.93336
beccariana*17.19 (0.36)14.4355
trachyphylla21.46 (1.32)23.20.99 (0.043)0.90421
havilandii5.63 (0.92)2.51.64 (0.255)0.7296
hullettii17.92 (3.34)13.71.03 (0.130)0.8682
lamellata14.99 (1.77)9.31.22 (0.124)0.84109
kingii14.92 (3.66)7.41.22 (0.137)0.8798
(b) SpeciesSlopey-interceptr2n
gigantea1.00 (0.12)2.17 (0.24)0.8836
winkleri1.10 (0.13)2.45 (0.24)0.8641
hosei0.76 (0.08)1.86 (0.15)0.7880
hypoleuca0.96 (0.29)2.29 (0.51)0.6519
triloba1.09 (0.10)2.41 (0.18)0.8090
beccariana1.01 (0.06)2.30 (0.11)0.80235
trachyphylla1.23 (0.07)2.68 (0.12)0.86184
havilandii1.33 (0.16)3.03 (0.32)0.6690
hullettii0.90 (0.12)2.16 (0.20)0.8440
lamellata1.03 (0.08)2.39 (0.13)0.8977
kingii1.30 (0.14)2.76 (0.24)0.8172

Estimated maximum tree height (Hmax) ranged from 5.5 to 31.3 m and maximum stem diameter ranged from 2.5 to 55.1 cm d.b.h. among the 11 Macaranga species (Table 6a). Macaranga gigantea and M. hosei were at one end of a continuum of tree size, with mature trees up to approximately 30 m in height and maximum stem diameters >40 cm. Macaranga hypoleuca is similar to these species (S. J. Davies, personal observation), although few large individuals occurred in our samples. Macaranga havilandii was at the other extreme, with an estimated Hmax of 5.6 m and maximum d.b.h. of only 2.5 cm. Macaranga lamellata and M. kingii were slightly larger trees with Hmax of 14–15 m and maximum stem diameters of 7–10 cm d.b.h. The other five species (M. winkleri, M. triloba, M. beccariana, M. trachyphylla and M. hullettii) were intermediate in size, with Hmax ranging from 17 to 23 m and maximum d.b.h. of 13–23 cm (Table 6a). Among species, Hmax was strongly positively correlated (r = 0.85, P = 0.001) with the proportion of trees in high-light environments (CI = 5).

Homogeneity of slopes tests indicated that there were significant differences among Macaranga species in both the slope and the y-intercept of the sapling height–diameter allometry model (Table 6b; tests not shown). In addition, the initial slope estimated from the asymptotic model differed significantly among species (Table 6a). Estimated maximum tree height was negatively correlated with both the y-intercept (r = −0.79, P = 0.003), and the initial slope (r = −0.68, P = 0.018) of the height–diameter relationship, indicating that saplings of the small-statured species (and in this case the more shade-tolerant species) were more slender than saplings of the large-statured species (Table 6).

As saplings of the more shade-tolerant species were on average living in lower light levels than the more high-light demanding species (Fig. 1), it is possible that different sapling allometries were due to light environment rather than species. Although sample sizes were limited, homogeneity of slopes tests revealed no significant differences in height–diameter relationships among light levels for five of the seven more abundant species (data not shown). Only M. hosei had significantly greater allometric slopes in lower light levels. These analyses suggest that species do differ, but larger sample sizes are required to assess the importance of light environment for Macaranga sapling allometries.



A strong correlation was found between fish-eye photographic estimates of canopy light levels and the arbitrary crown illumination index. Based on this index, there were significant differences among Macaranga species in the distribution of individuals with respect to light levels, suggesting a broad range of shade tolerance within this group of tropical rain forest trees. Our study species ranged from very high-light demanding pioneers typical of large forest gaps and secondary forest habitats (e.g. Macaranga gigantea), through what might be considered small-gap species (e.g. M. trachyphylla), to considerably shade-tolerant trees that persist in the understorey (e.g. M. kingii). However, all species had at least some individuals in four of the five CI classes, indicating substantial overlap, especially among the more high-light demanding species. In a previous analysis, using seedlings of nine of the 11 species grown in uniform environments, the index of shade tolerance was found to be strongly correlated with photosynthetic and other ecophysiological traits (Davies 1998).

Average light levels intercepted by individuals increased through ontogeny in 10 of the 11 Macaranga species. Similar results have been found in studies of non-pioneer species (Clark & Clark 1992). In most of the Macaranga species this ontogenetic increase was due both to higher mortality rates in low-light environments (Davies 1996), and to the increasing light levels intercepted as trees grow towards the canopy (Aoki et al. 1978; Chazdon 1986). The result was that CI index differed among species below 10 cm d.b.h. but not above this size (with smaller species excluded). There was no evidence for a decline in CI index at larger tree sizes, for any species, which might indicate over-topping due to canopy closure.

Qualitative estimates of crown light environment (such as the CI index) may not be comparable among studies, especially if the indices are not calibrated against some standard such as fish-eye photographs (Clark et al. 1993). However, we note that Clark & Clark (1992), using a similar but more refined analysis of light environment for species ranging from very shade-tolerant to quite light-demanding, demonstrated significant differences among species in microsite occupancy with respect to crown illumination. In addition, Clark et al. (1993) found significant differences between the pioneers Cecropia obtusifolia and C. insignis in sapling distribution with respect to light environment. Brokaw (1987) also described differences in gap size requirements among three pioneer species in Panama, and differences in the distributions of pioneer species within gaps (Popma et al. 1988) suggested different light requirements. Numerous other studies (e.g. Riddoch et al. 1991; Kitajima 1994; Reich et al. 1994) have found differences in ecophysiological responses to light among tropical pioneer species, but few have examined their consequences for forest spatial distributions (Davies 1998).

Macaranga species differed significantly in the proportions of stems on different establishment microsites. The more shade-tolerant species had significantly fewer stems in disturbed sites than the more high-light demanding species. This was not simply due to a correlation between undisturbed sites and lower light levels since trees of only two species had significantly lower mean CI indices on undisturbed than on disturbed sites. The high-light demanding species appear to prefer to establish on disturbed sites irrespective of light levels. Whether this is due to higher germinability of the smaller seeds (Davies 1996), differential seed rain (Levey 1988) or the requirement of a physical disturbance for germination is unknown. Within gaps, Ellison et al. (1993) found highest seedling densities for small-seeded Melastomataceae on the exposed mineral soil of root pits and mounds, even though light levels were lower and nutrient availability may have been lower in these microsites (Vitousek & Denslow 1986). Putz (1983) found higher pioneer species colonization of disturbed soil than undisturbed soil in gaps in Panama. Among the more high-light demanding Macaranga species, there were large differences in the proportion of trees on the different ‘disturbed’ microsites (logs, tip-ups and landslips). Núñez-Farfan & Dirzo (1988) found differences in the distribution and subsequent performance of the pioneer species Cecropia obtusifolia and Heliocarpus appendiculatus between crown and root zones of forest gaps, although the generality of this pattern needs further study as the zones were unreplicated. There were significant differences among species, but a finer scale approach is necessary to assess the relative importance of physical conditions for establishment and light levels in determining variation in horizontal distribution and performance for the more high light-demanding Macaranga species.

Typically, soil variation provides a coarser scale of resource heterogeneity in tropical forests than establishment conditions (Newbery & Proctor 1984; Baillie et al. 1987; Swaine 1996). The interbedding of sandstone and shale rocks at the study site, however, means that soils here vary on quite a small spatial scale (Watson 1985). Strikingly, the distributions of nine of the 11 Macaranga species were significantly biased with respect to soil texture, and only one of the species for which there was a large sample size, M. hullettii, was not strongly correlated (P = 0.053). Soil texture is correlated with both root mat depth and leaf litter thickness in the study site (Palmiotto 1995). While soil nutrient levels have been shown to be higher on the finer-textured (more clay-rich) soils at Lambir NP (Hall 1991; Ashton & Hall 1992), these soils also have a thinner root-mat and humus layer, and may also differ in soil moisture availability (Newbery et al. 1996). Experiments are therefore required to investigate the basis of soil preferences among these Macaranga species. In western Malesia, Macaranga is considerably more diverse and abundant in secondary succession on more nutrient-rich sites (shale- and basalt-derived soils) than on nutrient-poor sites (sandstone-derived soils), and is virtually absent from Adinandra-belukar, the very depauperate secondary successional communities of Malaya and Singapore (Wyatt-Smith 1963; Sim et al. 1992).


There were significant differences among species in maximum tree size (Hmax and Dmax), and the slope and y-intercept of the sapling height–diameter relationship (Table 6). Among species, Hmax was negatively correlated with shade tolerance; the three small-statured species were also the most shade-tolerant (Fig. 2). Thomas (1993) found a positive relationship between Hmax and light-saturated rates of photosynthesis for 25 species in four genera of Malaysian rain forest trees, although the relationships between photosynthetic rates and shade-tolerance and the distribution of seedlings and saplings with respect to light environment were not studied. Although the relationship between shade-tolerance and maximum tree size for tropical trees is by no means general, with numerous small statured high-light demanders (Swaine & Whitmore 1988; Clark & Clark 1992), the pattern seen in the four genera studied by Thomas (1993) is repeated in our results with Macaranga. Whether this pattern occurs in other genera (or clades) needs further study.

Figure 2.

Representation of the three-dimensional distribution of 11 Macaranga species with respect to estimated maximum tree height, the proportion of stems in high light levels (CI class 5) and the proportion of trees on sand-rich soil (soil texture class 4). Species are indicated by a single letter as in Fig. 1.

Even though Hmax was negatively correlated with shade tolerance, there was considerable variation in Hmax among the more shade-intolerant Macaranga species. Macaranga gigantea and M. hosei grow to 25–30 m tall and >40 cm d.b.h., whereas M. winkleri and M. beccariana rarely reach 20 m tall and 15 cm d.b.h., yet all four species are often sympatric in very high-light environments early in secondary succession. The four species establish early after gap formation, and grow extremely rapidly. Macaranga beccariana and M. winkleri initiate reproduction at tree sizes of approximately 5–6 cm d.b.h. and begin to senesce at around 10 cm d.b.h., while M. gigantea and M. hosei do not begin to reproduce until 10 cm d.b.h. and persist in forest gaps well after the mortality of neighbouring trees of M. winkleri and M. beccariana (Davies 1996). These closely related species obviously share similar resources, but the suggested role of temporal or vertical stratification of resources in species coexistence must be tested by a more detailed analysis of their differential resource use throughout ontogeny. The broader question of the evolutionary history of this divergence in life-history traits may also be addressed using a phylogenetic approach (Davies 1996).

Estimated maximum tree height was negatively correlated with the slope of the sapling height–diameter relationship among the 11 Macaranga species. Even though light availability did influence sapling allometries in some species, saplings of small statured species always tended to have more slender stems than saplings of larger statured species. This pattern was also found for 37 species of Malaysian non-pioneer rain forest trees (Thomas 1995). However, the opposite pattern was found for rain forest sapling allometries in Panama (King 1990b), where two understorey species were reported to have thicker stems at a given sapling height than four canopy species. It was argued that this was due to the requirement of the understorey species to support a larger crown biomass and foliage area in order to maximize light interception. Thomas (1995), on the other hand, suggests that the more slender stems of the small-statured understorey species in his study may be accounted for by their generally greater wood density and hence strength, and by the disproportionate benefit, in terms of light interception, that a sapling of an understorey species might get by maximizing height increment vs. diameter increment. In the case of the three small shade-tolerant Macaranga species in this study, all have rather simple architectures with a small number of medium to very large unlobed leaves on long petioles (S.J. Davies, personal observation). Further analyses of the interrelationships between leaf display, canopy architecture and sapling allometry, and how they respond to sapling light environment, are needed to assess the relative importance of maximizing stem diameter with respect to height (to enable large leaf biomass display) vs. maximizing height growth (to increase light interception).

Overall, the 11 species of Macaranga that are sympatric at the forest scale show a wide range of interspecific differences in both horizontal and vertical distribution patterns. Figure 2 provides a diagrammatic summary of these patterns focusing on three main axes of variation in distribution pattern, crown light levels (based on the proportion of trees in the highest CI index class), soil type distribution (based on the proportion of trees on the sand-rich soils) and estimated maximum tree height. There is a continuum of species occupancy of microsites, from the small statured and more shade-tolerant species (M. kingii, M. lamellata, M. hullettii and M. havilandii) which differ greatly in soil texture preferences, to the other seven species that have higher demands for both light and soil resources, but differ significantly in maximum tree size, and hence the time that individual species dominate succession. The heterogeneity of resource availability in the forest and the apparent high degree of specialization of these species demonstrates that all three axes of distribution are important in influencing the coexistence of this diverse group of early successional trees. How this variation in resource heterogeneity influences both individual performance and the regulation of the population dynamics of these species will have a strong influence on patterns of secondary succession in Bornean forests.


The 52-ha Long-Term Ecological Research Site was established as a collaborative research project of the Sarawak Forest Department, Malaysia, Harvard University, USA (under NSF award DEB 9107247 to P. S. Ashton), and Osaka City University, Japan. The authors would like to thank the Government of Sarawak for permission to conduct the project, and particularly for providing research permission to S. J. Davies, P. A. Palmiotto, P. S. Ashton and J. V. LaFrankie. We thank the National Parks and Forest Departments of Sarawak for assisting the project in many ways. S. J. Davies was supported by a graduate student fellowship from the Department of Organismic and Evolutionary Biology at Harvard University and a Deland Award for student research to the Arnold Arboretum at Harvard University. The staff of the Sarawak Forest Department and the people of Rumah Ajai are gratefully acknowledged for their invaluable field assistance. Peter Wayne, Sean Thomas, Glenn Berntson, David Ackerly, Fakhri Bazzaz and Peter Stevens assisted with analyses or made helpful comments on earlier drafts of the manuscript.

Received 10 November 1997revision accepted 5 February 1998