Content uploaded by Charles A Nock
Author content
All content in this area was uploaded by Charles A Nock
Content may be subject to copyright.
Available via license: CC BY-NC 2.0
Content may be subject to copyright.
Wood density and its radial variation in six canopy tree species
differing in shade-tolerance in western Thailand
Charles A. Nock1,*, Daniela Geihofer2, Michael Grabner2, Patrick J. Baker3,
Sarayudh Bunyavejchewin4and Peter Hietz1
1
Institute of Botany, University of Natural Resources and Applied Life Sciences, 33 Gregor Mendel Strasse, Vienna 1180,
Austria,
2
Institute of Wood Science and Technology, University of Natural Resources and Applied Life Sciences, Vienna 1180,
Austria,
3
Monash University, School of Biological Sciences, Clayton, Victoria 3800, Australia and
4
National Parks, Wildlife, and
Plant Conservation Department, Chatuchak Bangkok 10900, Thailand
Received: 19 February 2009 Returned for revision: 17 March 2009 Accepted: 9 April 2009 Published electronically: 19 May 2009
†Background and Aims Wood density is a key variable for understanding life history strategies in tropical trees.
Differences in wood density and its radial variation were related to the shade-tolerance of six canopy tree species
in seasonally dry tropical forest in Thailand. In addition, using tree ring measurements, the influence of tree size,
age and annual increment on radial density gradients was analysed.
†Methods Wood density was determined from tree cores using X-ray densitometry. X-ray films were digitized
and images were measured, resulting in a continuous density profile for each sample. Mixed models were
then developed to analyse differences in average wood density and in radial gradients in density among the
six tree species, as well as the effects of tree age, size and annual increment on radial increases in Melia
azedarach.
†Key Results Average wood density generally reflected differences in shade-tolerance, varying by nearly a factor
of two. Radial gradients occurred in all species, ranging from an increase of (approx. 70%) in the shade-intolerant
Melia azedarach to a decrease of approx. 13% in the shade-tolerant Neolitsea obtusifolia, but the slopes of
radial gradients were generally unrelated to shade-tolerance. For Melia azedarach, radial increases were most-
parsimoniously explained by log-transformed tree age and annual increment rather than by tree size.
†Conclusions The results indicate that average wood density generally reflects differences in shade-tolerance in
seasonally dry tropical forests; however, inferences based on wood density alone are potentially misleading for
species with complex life histories. In addition, the findings suggest that a ‘whole-tree’ view of life history and
biomechanics is important for understanding patterns of radial variation in wood density. Finally, accounting for
wood density gradients is likely to improve the accuracy of estimates of stem biomass and carbon in tropical trees.
Key words: Radial gradients, shade-tolerance, tree biomass estimates, tree rings, tropical trees, wood density.
INTRODUCTION
Wood density is related to a number of plant functional traits and
is an important indicator of the mechanical properties of woods
(Panshin and de Zeeuw, 1980; Chave et al., 2009). A direct
relationship between wood density and tree growth is expected
because the volume of wood produced for a given unit biomass
is inversely proportional to its density (King et al., 2005). In tro-
pical forests the growth, survival and reproduction of light-
demanding tree species is dependent on an ability to avoid or
escape prolonged periods of low light (e.g. Ackerley, 1996).
Thus, light-demanding species attain rapid rates of height
growth, in part by investing in low density wood that is cheap
to construct (King et al., 2006), but as a consequence of low
stem strength, have high mortality rates due to stem breakage
(Putz et al., 1983; van Gelder et al., 2006; Poorter, 2008). In
contrast, shade-tolerant tree species grow more slowly and
invest in dense, strong and damage-resistant wood that in turn
lowers their mortality rates (Putz et al., 1983; Muller-Landau,
2004; van Gelder et al., 2006). A growth-mortality trade-off is
likely to be common in trees of diverse tropical forests, and
wood density may be one of the best predictors of species differ-
ences along this axis of variation (Poorter et al., 2008; Chave
et al., 2009).
In addition to interspecific variation in wood density in
forests, within individual trees wood density often varies ver-
tically along the main axis of the stem and/or radially from the
pith to the bark (Panshin and de Zeeuw, 1980; Grabner and
Wimmer, 2006). For example, in gap-colonizing heliophiles
of tropical wet forests with very low-density juvenile wood,
increases in wood density from the pith to the bark as large
as 200– 300% have been documented (Whitmore, 1973;
Wiemann and Williamson, 1988), and in drier and montane
tropical forests increases ranged from 20% to 100%
(Wiemann and Williamson, 1989b). It is thought that radial
increases in wood density result from a shift in allocation
from low density wood and rapid height growth early on in
tree development to denser wood and structural reinforcement
as trees increase in size, age and height and are exposed to
* For correspondence. E-mail charles.nock@gmail.com
#2009 The Author(s).
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://
creativecommons.org/licenses/by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium,
provided the original work is properly cited. 297
Annals of Botany 104: 297– 306, 2009
doi:10.1093/aob/mcp118, available online at www.aob.oxfordjournals.org
increasing wind speeds within the forest (Wiemann and
Williamson, 1989b).
Radial gradients in wood density are found in a range of tree
species of different successional stages and from different
forests, suggesting that they may occur frequently in trees
(Panshin and de Zeeuw, 1980; Wiemann and Williamson,
1989a,b; Omolodun et al., 1991; Hernandez and Restrepo,
1995; Parolin, 2002; Woodcock and Shier, 2002). In tropical
wet and dry forest, increases were observed in 80% and 60%
of the tree species, respectively (Wiemann and Williamson,
1989a,b), although species were not a random sample of the
local tree community. Documented increases are greatest in
wet tropical forest and become progressively less in drier tro-
pical forests, montane rain forest and temperate forest
(Wiemann and Williamson, 1989b; Woodcock and Shier,
2002), a pattern which is consistent with differences in compe-
tition for light, which may be greater in tropical wet forest
compared with drier forests (Wiemann and Williamson,
1989a; Coomes and Grubb, 2000; Markesteijn et al., 2007).
Understanding the relative influence of stem age versus stem
size on radial gradients in wood density is important for infer-
ences of their adaptive value (e.g. Rosell and Olsen, 2007; de
Castro et al., 1993). In a previous study of the effects of tree
size and age on radial gradients in wood density, de Castro
et al. (1993) presented evidence that tree age was likely to
be the driver of radial increases in wood density for a single
cohort of Joannesia princeps grown in a Brazilian plantation:
for trees of the same age but different diameters the slopes
from a regression of wood density versus distance from the
pith were generally greater in smaller, slower-growing trees
because similar pith to bark changes in wood density occurred
across a smaller stem radius. However, these results remain to
be verified for trees varying in both age as well as size, and
growing in a natural forest setting. At least for tropical trees
such data are not available, because quantifying ages in most
tropical trees remains difficult. Understanding the drivers of
radial gradients in wood density is also important for deter-
mining their implications for tree biomechanics; mechanical
models of tree stability have thus far not incorporated the poss-
ible effects of radial variation in stem wood density (e.g.
Sterck and Bongers, 1998). Furthermore, estimates of carbon
stocks in forests are strongly affected by accurate estimates
of wood density (Chave et al., 2005), but few studies have
addressed the potential bias arising from radial variation in
wood density within stems (Nogueira et al., 2005, 2008).
In seasonal tropical forests a growing number of species are
known to produce reliable annual growth rings, thus allowing
one to determine tree ages as well as growth patterns in select
species (Worbes, 2002; Baker et al., 2005; Baker and
Bunyavejchewin, 2006). By employing X-ray densitometry,
which produces a continuous and high resolution (,1mm)
wood density profile from tree cores, annual growth boundaries
can be accurately delineated using digitized images of the X-ray
films (e.g. Grabner et al., 2005). Tree age and annual increment
can therefore be examined as potential explanatory variables in
models of wood density variation in addition to tree size.
The overarching goal of this study was to characterize radial
variation in wood density in sympatric tropical tree species that
differed in relative shade-tolerance. Using data collected from
six tropical canopy tree species from a seasonal dry evergreen
forest in Thailand, the aim was to address the following ques-
tions. (1) How common are radial gradients? (2) How are mean
wood density and shade-tolerance related? (3) Do radial gradi-
ents decrease with increasing shade-tolerance? (4) Does tree
size or age most-parsimoniously explain radial increases in
wood density and what is the effect of radial increment?
MATERIALS AND METHODS
Study site and species
This study was conducted in a 50-ha forest dynamics plot estab-
lished in seasonally dry evergreen forests in the Huai Kha
Khaeng Wildlife Sanctuary in west-central Thailand, located
at 158400N, 998100E(Bakeret al., 2005; Bunyavejchewin
et al., 2001, 2002). Although logging occurred in the region
prior to 1970, the forest dynamics plot was established in an
area of forest with no evidence of prior logging. The climate
in the region is monsoonal with ,100 mm of rainfall per
month in the dry season between November and April and a
mean annual rainfall of approx. 1500 mm. Mean July and
January temperatures are 278Cand198C, respectively. Soils
in the area are highly weathered ultisols (Lauprasert, 1988).
Six canopy tree species – Afzelia xylocarpa Craib,
Neolitsea obtusifolia Merrill, Vitex peduncularis Wall. ex
Schauer, Toona ciliata M.Roem., Melia azedarach L. and
Chukrasia tabularis A. Juss. (¼C. velutina M.Roem.) –
were selected because they form annual growth rings and com-
prise an important component of the forest community
measured by their basal area or frequency (Baker, 2001;
Baker et al., 2005). The shade-tolerance of Afzelia was
assigned on the basis of regeneration patterns in the vicinity
of the forest dynamics plot. Afzelia has been observed to
regenerate exclusively in gaps and juvenile diameter growth
rates are high (.10 mm year
21
; Baker et al., 2005). For the
other species, shade-tolerance was determined primarily
from the literature; recent studies and plot observations are
consistent with the literature categorization (Troup, 1921;
Baker et al., 2005; Baker and Bunyavejchewin, 2006). The
selected species thus represent the full range of canopy tree
life histories that occur in the forest (Table 1).
Field and laboratory procedure
Canopy trees were cored in April 2007 with a 5-mm-diameter
increment borer at approx. 1 m height. Due to the limited
number of individuals of each species on the plot, trees were
selected non-randomly and to represent the full range of diam-
eters present, excluding very irregular stems or severely leaning
individuals. The number of trees sampled, diameter ranges and
mean diameters for each species are given in Table 1.
X-ray densitometry was used to measure the wood density
of one core per tree. Samples were allowed to air dry in the
laboratory, sawn along their length to a uniform thickness of
1.4 mm with a double-bladed saw and then placed on film
and exposed to X-rays (10 kV, 24 mA) for 25 min (Grabner
et al., 2005). X-ray films were then digitized using a custom-
built scanner equipped with a stepwise driven motor and a
line-camera on an incident-light microscope, resulting in a res-
olution of 3.52 mm pixel
21
. Wood density was then measured
Nock et al. — Wood density in Thailand298
from the digital images (256 grey levels) using SigmaScan
Pro
w
version 5.0 (Systat Software Inc., San Jose, CA, USA)
and the data for each tree were summarized for analysis by cal-
culating means for 1-cm intervals. Absolute density values
were obtained by including a density standard made of cellu-
lose acetate with each X-ray film during exposure and later
calibrating the greyscale.
Although annual growth rings were identifiable in all six
species, those in Melia were the most readily identifiable in
density images because of conspicuous rows of large vessels
at the start of rings and a pronounced intra-annual density gra-
dient. Melia was therefore chosen to compare the effects of
age, radial distance from the pith and annual growth on
wood density, and in addition, the average wood density in
each growth ring, the radial distance from pith at the end of
each growth ring and annual increment were measured. Then
the average growth rate was calculated for each tree.
Not all coring attempts reached the pith of the tree, so
missing distances to the pith were estimated from the curvature
of the last complete inner ring using Duncan’s geometric
method (Duncan, 1989). Hereafter, these corrected data are
referred to as the radial distance from the pith. By extension,
the number of missing inner tree rings for each Melia individual
was calculated by dividing the missing distance (cm) by the
average growth rate (cm year
21
) in the adjacent five growth
rings, and tree ages adjusted for the missing number of rings.
Statistical analysis
Models of variation in wood density. Two sets of linear
mixed-effects models were developed to investigate the
relationships among wood density and the explanatory
variables. The first set of models examined differences in
average wood density and radial variation in wood density
among the six species. The second set examined the impor-
tance of radial distance from the pith, tree age and annual
increment in models of radial variation in wood density for
Melia. Below, the models are described and the methods
used to select the most-parsimonious models outlined.
Interspecific differences in average wood density and in radial
gradients in wood density. Average wood density and gradients
in wood density with radial distance were expected to vary
with shade-tolerance, and thus fit the following linear
mixed-effects model to explain variation in wood density
among the six species:
WDijk ¼ð
b
0þ
b
1Dij þ
b
2Skþ
b
3SkDijÞþð
m
0iþ
m
1iDij
þ1ijÞð1Þ
where WD
ijk
is the wood density for ith 1-cm interval in the jth
tree of the kth species; D
ij
is the radial distance in cm from the
pith and S
k
is the species of tree. b
0
,b
1
,b
2
,b
3
are the fixed
effects. Initially, both a random intercept, m
0i
, and a random
slope, m
1i
were included to account for random variation at
the tree level. In order to account for dependence among
1
ij
due to temporal autocorrelation within-trees, a variety of
error autocorrelation structures were evaluated. To determine
the most-parsimonious model, a series of reduced models
which varied in their inclusion of the fixed and random
effects terms were examined (Table 2).
Effects of tree age, radius and annual increment on radial
increases in Melia.A separate series of linear mixed-effects
models was developed solely for Melia. Exploratory data
TABLE 1. Species examined, shade-tolerance, number of trees sampled (n), diameter at breast height (DBH) of sample trees and
average percentage change in wood density (from pith to bark) for six tropical trees in western Thailand
Species Family Shade-tolerance nDBH mean (range) (cm) Percentage change in wood density*
Afzelia xylocarpa Fabaceae (Caesalp.) Intolerant 9 58.8 (12.1–84.4) 24
Chukrasia tabularis Meliaceae Intolerant/intermediate 22 41.5 (11.0–67.2) 38
Melia azedarach Meliaceae Very intolerant 11 51.2 (16.9–62.7) 70
Neolitsea obtusifolia Lauraceae Tolerant 26 44.5(8
.5–69.7) 213
Toona ciliata Meliaceae Intolerant 9 64.4 (36.0–76.1) 27
Vitex peduncularis Lamiaceae Intermediate 12 28.8 (14.0–43.4) 36
* Values from a linear mixed-effects model (Table 2, Model 1; Table 4).
TABLE 2. Comparison of the fitted models for six tree species from western Thailand including the predictor variables examined,
corresponding maximum log-likelihoods, AIC values and differences in AIC relative to the model with the lowest AIC
(Model 2;
D
AIC)
Model DSDSRandom intercept Random slope Log- likelihood AIC DAIC*
1xxx x 27615.76 15261.53 0.9
2xxx x x 27613.31 15260.63 0
3xxx x 27621.41 15272.81 12.18
4xx x x 27645.76 15315.52 54.89
5x x x 27696.08 15406.16 145.53
Predictor variables included distance from the pith in cm (D) and species (S). The inclusion of a term is indicated by an x.
* Note: DAIC ,2 indicates little difference in support for competing models (Burnham and Anderson, 2002).
Nock et al. — Wood density in Thailand 299
plots indicated a log-transformation was appropriate in order to
obtain a linear relationship between wood density and tree age,
and to meet assumptions of linearity and normality in var-
iance. Candidate models ranged in complexity from simple
models that included age or size only (either the ith ring
from the pith or the ith centimetre from the pith) to more
complex models that included all of the terms as well as inter-
actions among variables (Table 3). The final linear
mixed-effects model form was:
WDij ¼ð
b
0þ
b
1Aij þ
b
2Iij þ
b
3AijIij Þþð
m
0iþ
m
1iAij
þ1ijÞð2Þ
where WD
ij
is the value of the wood density for the ith of n
i
observations in the jth tree; A
ij
is the log-transformed tree
age, I
ij
is the annual increment. m
0i
,m
1i
are random effects
for each tree and
1
ij
is the error. A variety of error autocorrela-
tion structures were evaluated to determine the most suitable
for the model.
Model selection and evaluation. To select the most-
parsimonious model among competing models the significance
of the fixed effects terms was first evaluated using approximate
Wald tests and likelihood-ratio tests – the latter via compari-
son with a reduced model which differed by the term being
tested (Pinheiro and Bates, 2000). Secondly, the significance
of the random-effect terms and error autocorrelation were eval-
uated using likelihood-ratio tests. To select among error auto-
correlation structures, nested models were compared using
likelihood-ratio tests and non-nested models were compared
using information criteria statistics (Pinheiro and Bates,
2000). Finally, models were compared by calculating the
difference in Akaike’s information criteria (AIC) between
the model with the lowest AIC, indicating greatest parsimony,
and the AIC of another candidate model (Tables 2 and 4;
Burnham and Anderson, 2002).
In addition to using AIC to indicate which model most-
parsimoniously explained radial gradients in Melia, compari-
son of our analysis with that of de Castro et al. (1993) was
facilitated by examining the relationship between the increase
in wood density with distance (slope term from model 2,
Table 3, including the random effect) and average growth
rate in Melia trees.
For each linear mixed-effects model assumptions – constant
variance, homogeneity of group variances and normality of the
within-group errors – were assessed graphically (Pinheiro and
Bates, 2000). Plots of predicted values versus the observed
data were used to assess model fit (see Supplementary Data,
available online). All analyses were conducted in R version
2.6.2 (R Foundation for Statistical Computing, Vienna,
TABLE 3. Comparison of the fitted linear mixed models for
Melia azedarach examining the effects of radial distance from
the pith (D), log-transformed age (A) and annual increment (I)
on wood density including the predictor variables examined and
corresponding maximum log-likelihoods, AIC values and
differences in AIC among models (
D
AIC)*
Model number Fixed effects terms AIC Log-likelihood DAIC
1AþIþA:I 3122.37 21552.18 0
2DþIþD:I 3126.49 21554.24 4.12
3A3130.26 21558.13 7.89
4AþI
n.s.
3131.29 21557.65 8.92
5AþD
n.s.
3132.09 21558.04 9.72
6D3138.63 21562.31 16.26
7DþI
n.s.
3138.91 21561.46 16.54
n.s.
, Non-significant terms at P,0.05.
* Note: DAIC values between 3 and 7 indicate considerably less support
for the model (Burnham and Anderson, 2002).
TABLE 4. Maximum likelihood parameter estimates and confidence limits (95%) for a linear mixed-effects model describing changes
in wood density with distance from the pith (D) for six canopy tree species (Spp) in western Thailand (Table 2, model 1)
Parameter Estimate (s.e.) t-value P-value Confidence interval
Intercept 809.6 (32.0) 25.31 ,0.001 747.1, 872.1
D5.0(1
.5) 3.26 0.001 2.0, 8.0
Spp
Ct 2193.7 (38.1) 25.08 ,0.001 2269.2, 2118.3
Ma 2320.7 (42.5) 27.54 ,0.001 2404.3, 2236.5
No 37.2 (37.6) 1.00 0.325 237.2, 111.6
Tc 2336.6 (47.2) 27.13 ,0.001 2430.1, 2243.2
Vp 2109.7 (43.7) 22.51 0.014 2196.2, 223.2
DSpp
DCt 3.8(2
.0) 1.90 0.056 20.1, 7.7
DMa 6.0(2
.1) 2.79 0.005 1.8, 10.2
DNo 28.2(1
.9) 24.30 ,0.001 211.9, 24.5
DTc 21.2(2
.2) 20.53 0.589 25.4, 3.1
DVp 10.1(3
.2) 3.17 0.002 3.9, 16.3
Variance components s.d. Likelihood-ratio P-value Confidence interval
Intercept 55.616
.19 ,0.001 41.7, 74.1
Residual 76.7 62.1, 80.3
AR1 error term 0.76 436.51 ,0.001 (0.70, 0.80)
P-values for Spp and DSpp test are for significant differences relative to Afzelia xylocarpa.
Species abbreviations: Ct,Chukrasia tabularis;Ma,Melia azedarach;No,Neolitsea obtusifolia;Tc,Toona ciliata;Vp,Vitex peduncularis.
Nock et al. — Wood density in Thailand300
Austria) using the package nlme (Pinheiro and Bates, 2000).
Error rates for multiple comparisons of average wood
density among species (Fig. 1) were controlled by using the
Tukey multiple comparison procedure within the R package
multcomp (Hothorn et al., 2008).
RESULTS
Species differences in average wood density and radial gradients
in wood density
Two models including the main effects of species and distance
from the pith as well as their interaction on wood density for
the six species had equal support (models 1 and 2; Table 2).
There was little support for alternative models which either
excluded a random intercept or the interaction between
species and distance from the pith (models 3–5; Table 2).
Likelihood-ratio tests confirmed the statistical significance of
the random intercept term (Tables 2 and 4). Inspection of
regressions of wood density versus distance from the pith con-
ducted for the individuals of each species suggested testing for
random variation in the slope and intercept (data not shown),
but the inclusion of a random slope term did not significantly
improve model fit (models 1 and 2; Table 2). Thus, the simpler
of the two models (model 1) was selected. An error autoregres-
sive correlation structure of order one provided the most-
parsimonious fit. The best model explained 79% of the vari-
ation in the observed data and provided a good fit (model 1;
see Fig. S1 in Supplementary Data].
Mean wood density for the six tree species varied by nearly
a factor of two, from approx. 470 kg m
23
to approx.
850 kg m
23
(Fig. 1). The rank order of species from highest
to lowest mean wood density was Neolitsea,Afzelia,Vitex,
Chukrasia,Melia,Toona (Fig. 1). With the exception of
Afzelia, this order corresponded to species differences in
shade-tolerance (Fig. 1 and Table 1). Significant differences
in mean wood density were present, although species that
were relatively similar in shade-tolerance did not significantly
differ (Fig. 1).
Radial gradients in wood density were present in all of the
six species (Fig. 2 and Table 4). In five of the six species
wood density increased with distance from the pith (Fig. 2),
although large confidence intervals indicated a marginally
insignificant slope for Toona (Table 4). Individual tree plots
of wood density versus distance from the pith suggested that
despite the curvilinear appearance of the aggregated data in
some species (e.g. Melia; Fig. 2), individual trees trends
were generally linear (data not shown). The most shade-
tolerant species, Neolitsea, showed a radial decrease in wood
density (Fig. 2); however, the rank order of slopes
(Neolitsea,Toona,Afzelia,Chukrasia,Melia,Vitex; see
Table 4) did not reflect differences in shade-tolerance (Fig. 2
and Table 1). Unexpectedly, radial increases in the more
shade-tolerant species Chukrasia and Vitex were greater than
in Toona and, despite similar average wood densities and
shade-tolerances, radial gradients were quite different
between Toona and Melia.
Radial gradients are often reported in terms of a percentage
change from the pith to the bark, whereas slopes describe the
rate of change in wood density independent of size differences
among trees. Expressed in percentage change, the greatest
increase occurred in Melia at 70% (Table 1).
Effects of size, age and increment on radial shifts in Melia
Increases in wood density in Melia with tree age and tree
diameter were similar (Figs 2 and 3F). However, the AIC com-
parisons indicated the model including log-age and annual
increment provided a better fit to the data than the equivalent
model with distance from the pith. The most-parsimonious
model (model 1; Table 3) included terms for log-age, annual
increment and their interaction, and explained 74% of the vari-
ation in the observed data. This model fit well, with the excep-
tion of some model bias at the lowest and highest wood densities
(see Fig. S2 in Supplementary Data, available online). Delta
AIC values indicated considerably less support for the alterna-
tive models (Table 3). As in the previous model for all six tree
species, an error autoregressive correlation structure of order
one provided the most-parsimonious fit.
Size and age varied substantially among the 11 Melia trees
(Fig. 3A). As expected, there was a negative relationship
between annual increment and log-age (Fig. 3D), and
between annual increment and wood density (Fig. 3C).
Wood density increased linearly with log-transformed tree
age (Fig. 3E). A significant but small interaction effect was
found between annual increment and log-age on wood
density: slower growth in young trees was associated with
greater wood density but this difference diminished with tree
age (Fig. 3B). Finally, a negative relationship was also
TABLE 5. Maximum likelihood parameter estimates and confidence limits (95%) for the linear mixed model describing variation in
wood density with log-transformed tree age (A) and annual increment (I) for Melia azedarach (Table 3, model 1)
Parameter Estimate t-value P-value Confidence interval
Intercept 450.5 (60.9) 7.40 ,0.001 331.5, 569.6
A177.9 (50.6) 3.51 ,0.001 78.9, 276.9
I2124.3 (34.5) 23.60 ,0.001 2191.7, 256.8
A:I116.3 (33.5) 3.47 ,0.001 50.7, 181.8
Variance components Standard deviation Likelihood-ratio P-value Confidence interval
Intercept 138.613
.63 0.001 73.3, 261.9
A115.69
.48 0.009 50.4, 265.1
Residual 64.1 57.3, 71.7
AR1 error term 0.34 24.38 ,0.001 0.22, 0.48
Nock et al. — Wood density in Thailand 301
observed between the slope of the increase in wood density
with distance from the pith (predicted from model 2;
Table 3) and the average growth rate of individuals (Fig. 4).
DISCUSSION
Shade-tolerance and mean wood density
It was found that the variation in mean wood density corre-
sponded with differences in shade-tolerance for all species
except for the intolerant Afzelia, which was closer in density
to the shade-tolerant Neolitsea than to Toona or Melia, the
other intolerant species (Fig. 1 and Table 1). Baker and
Bunyavejchewin (2006) examined canopy ascension patterns
using tree ring analyses for five of the species in the present
study (excluding Afzelia) and found that rapid growth into
the canopy in high light was most common in Melia and
declined substantially with increasing shade-tolerance,
whereas the most common pattern amongst shade-tolerant
trees was slow growth in low-light beneath the canopy.
These patterns for a seasonally dry rainforest in Thailand are
consistent with recent results from a wide range of tropical
forests, indicating a central role for wood density in the
growth strategies of trees, with low wood density facilitating
rapid canopy ascension in shade-intolerant trees and higher
wood density contributing to higher survival in shade-tolerant
trees which grow slowly beneath a canopy (Ackerley, 1996;
van Gelder et al., 2006). Chave et al. (2009) noted a possible
exception to the general negative trend of lower growth rates
and higher wood density, suggesting that in seasonally dry cli-
mates denser wooded plant species may be better equipped to
cope with strongly negative xylem potentials and water stress
(resistance to xylem implosion; Hacke et al., 2001) and
hence exhibit greater annual growth rates than lighter-wooded
species. This could in part explain the association of relatively
high growth rates with high wood density in Afzelia in the
present study.
Potential causes of radial gradients in wood density in Melia
The AIC comparisons (Table 3) revealed the most-
parsimonious model included the effects of tree age with a
minor effect of annual growth (Fig. 2B), demonstrating that
wood density does indeed vary with tree age independently
of tree size (radial distance from the pith) in Melia. In addition,
the negative relationship between the slope of the change in
wood density with distance and mean growth rate (Fig. 4)
shows that the density gradient depends on tree age: in slower-
growing trees the increase in wood density with age occurs
across a smaller radius, resulting in greater slopes. De Castro
et al. (1993) documented the importance of tree age in a pre-
vious study, though whether their results could be generalized
to natural forests and adult canopy trees was uncertain for two
reasons: (1) the study trees were young (17 years) and con-
sisted of a single cohort grown in a plantation; and (2)
growth and density could not be annually resolved as the
species did not produce annual growth rings.
Radial gradients in wood density have been primarily inter-
preted as an adaptation for structural support in tropical
pioneer trees which exhibit rapid height and diameter growth
by producing wood of low density as juveniles but require
greater stability later in development, possibly due to greater
exposure to wind (Wiemann and Williamson, 1988, 1989a,
b). As previously noted by de Castro et al. (1993) given this
interpretation, one would expect a measure of tree size such
as diameter or height rather than tree age to be most strongly
related to radial gradients in wood density. In a recent study,
Woodcock and Shier (2003) found some evidence for an influ-
ence of tree height on radial gradients in temperate tree
species, although this was from a correlation of the ‘current’
height of individuals with their radial gradient in wood
density and not from an analysis with tree height as an inde-
pendent variable, thus it is difficult to derive inferences from
these results. The present data are novel for tropical trees in
that the effect of age, stem size (distance from the pith) and
growth rates on wood density can be tested; however,
because it was not possible to obtain retrospective data on
tree heights, it cannot be ruled out that age may be a proxy
for some other measure of size such as tree height. Finally,
higher density wood is also more resistant to pathogens
(Augspurger, 1984), thus increasing wood density with tree
age may function as a defence mechanism against increasing
pathogen loads over time.
Association between mean wood density or shade-tolerance and
radial gradients in wood density
Woodcock and Shier (2002) proposed a model for the
degree and direction of radial trends in wood density (specific
gravity) in tree species. In their model, large positive slopes
are associated with pioneer species with low wood density
and slopes progressively decline and ultimately become nega-
tive in late-successional species with specific gravities ranging
from approx. 0.6to1
.0. The present observations were not
200
Neolitsea
Afzelia
Vitex
Chukrasia
Melia
Toona
400
a
ab
bc
c
d
d
600
Wood density (kg m–3)
800
1000
FIG. 1 . Wood density for six canopy trees species in western Thailand: pairs
with the same letter are not significantly different (P.0.05). Values were cal-
culated from a linear mixed effects model of the variation in wood density
(Table 4) and error inflation controlled using Tukey’s multiple comparison
procedure.
Nock et al. — Wood density in Thailand302
found to be consistent with this model: Vitex and Afzelia had
relatively high-density wood and positive radial gradients
and both had much higher wood densities and greater slopes
than the light-wooded, shade-intolerant Toona. In addition,
mean wood density, as well as shade-tolerance was quite
similar between Toona and Melia, but their radial gradients
were markedly different. In another recent study of radial gra-
dients in wood density which examined 35 tree species of the
Amazonian floodplains, the mean wood density of a species
and the average increase in wood density from pith to bark
were not significantly related, although generally increases
were greatest and most common in fast-growing pioneers
(Parolin, 2002). Similarly, though differences between pio-
neers and very shade-tolerant species were evident in the
present study (compare Melia and Neolitsea; Figs 1 and 2),
radial gradients in wood density did not correspond to mean
wood density or shade-tolerance across the range of shade-
tolerance categories (Table 1).
It is suggested that a ‘whole-tree’ perspective – such as that
proposed by Givnish (1988) for the interpretation of leaf level
traits within the context of whole-plant adaptation to light
environments – is likely key to understanding the relationships
among radial gradients in wood density, tree biomechanics and
successional status/shade-tolerance. For example, recent
research has documented that species strategies for dealing
with wind disturbance vary: some species may lose canopy
elements during severe winds to mitigate stress on the main
stem, but retain enough elements to recover post-disturbance,
while others retain their canopy at higher wind speeds, but
risk catastrophic failure if winds become strong enough
(Metcalfe et al., 2008). Maximum adult stature is also impor-
tant for understanding species differences in wood density, tree
400
Neolitsea obtusifolia
600
800
1000
1200
Toona ciliata
400
Vitex peduncularis
600
800
1000
1200
Chukrasia tabularis
400
0 10203040
Distance from pith (cm) Distance from pith (cm)
0 10203040
Afzelia xylocarpa
600
800
1000
1200
Melia azedarach
Wood density (kg m–3) Wood density (kg m–3)Wood density (kg m–3)
FIG. 2 . Variation in wood density with distance from the pith for six tree species in western Thailand with fitted values predicted by a linear mixed effects model
fit by maximum likelihood (Table 2, Model 1).
Nock et al. — Wood density in Thailand 303
structural stability and likely radial gradients in wood density.
Recent studies indicate that, among pioneer species, wood
density, stiffness factors, safety factors and wood strength all
increase with maximum adult stature (Falster and Westoby,
2005; van Gelder et al., 2006). Finally, longevity is an impor-
tant life history trait that could lead to differences in wood
density and its radial trends among two pioneer species with
similar requirements for regeneration; in the present study
Afzelia and Melia are both shade-intolerant and have high
juvenile growth, but individuals of Afzelia reach ages in
excess of 250 years, whereas Melia rarely exceed approx. 70
years of age (Baker et al., 2005).
Further evidence for the co-ordination of wood density with
traits across the whole-plant, including leaf size, minimum leaf
water potential and possibly rooting depth was recently pre-
sented in an extensive review by Chave et al. (2009) in
which the existence of a ‘wood economics spectrum’ was pro-
posed. The present results suggest that radial gradients in wood
density – which likely represent an allocational shift from
rapid height growth to mechanical support throughout
400
01234
Increment (cm year–1)
Log10(age + 1) Log10(age + 1)
200
600
800
1000
1200
CF
BE
Wood density (kg m–3) Wood density (kg m–3) Stem radius (cm)
Wood density (kg m–3) Increment (cm year–1)
400
0 1020 5060
Age (years)
200 4030
600
800
1000
1200
Wood density (kg m–3)
400
200
600
800
1000
1200
Mean increment s.d.
Mean increment + s.d.
Mean increment – s.d.
400
600
800
1000
1200
0·0 0·5 1·0 1·5 2·0 0·0 0·5 1·0 1·5 2·0
Age (years) Log10(age + 1)
AD
10
0
20
30
40
0
1
2
3
4
02010 30 40 50 0·0 0·5 1·0 1·5 2·0
FIG. 3 . Relationships for Melia azedarach: (A) stem radius and tree age; (B) predicted wood density and the interaction of annual increment and tree age; (C)
annual increment and wood density; (D) increment and log-tree age; (E) wood density and log-tree age; ( F) wood density and tree age. For (B) values for mean
increment, mean increment plus one standard deviation, and mean increment minus one standard deviation were calculated with the observed data for the follow-
ing year classes: 0– 9, 10 – 19, 20 – 29, 30 – 39, 40 –50. These values were then used to calculate predicted values of wood density at each age from a linear mixed
model (Model 1, Table 3).
Nock et al. — Wood density in Thailand304
development in response to environmental conditions are rel-
evant to understand aspects of the ‘wood economics
spectrum’.
Significance for biomass estimates
Variation in wood density occurs at multiple scales in forest
ecosystems, ranging from within trees, to regional changes in
mean wood density among forest communities (Hernandez
and Restrepo, 1995; Muller-Landau, 2004; Chave et al.,
2006; Grabner and Wimmer, 2006; Swenson and Enquist,
2007). Understanding and documenting the great degree of
variation in wood density is thus an important challenge in cal-
culating biomass and carbon pools in tropical forests, a critical
area of research on global carbon cycles (Brown, 2002;
Woodcock and Shier, 2002; Nogueira et al., 2005). Studies
of destructively sampled trees have shown that, in predicting
the mass of a tree, wood density is the second most important
parameter after tree diameter (Chave et al., 2005). Where
radial gradients are present, mean stem density increases or
decreases with stem size and thus estimates of stem biomass
and in turn carbon based on a mean species density become
inaccurate (Nogueira et al., 2008). In the case of the six
species in the present study, simple calculations of the
biomass of a stem disc 40 cm in diameter using mean
density values of the cores resulted in an overestimate of
approx. 8% in Neolitsea due to its radial decrease in density,
to underestimates ranging from approx. 12% to approx. 31%
in the other five species due to radial increases. Errors even
will be larger when stem-wood density is determined from
the density of outer wood, such as when short tree cores are
taken. Overall, errors in carbon estimation resulting from
radial gradients will depend mainly on the slope of the
change in wood density, as well as the proportion of individ-
uals with increasing and decreasing densities. Thus, studies
designed specifically to assess how frequent radial gradients
in wood density occur in different forests are needed.
Previous studies have examined radial gradients in wood
density by individually quantifying the wood-specific gravity
of small segments of cores (e.g. 1 cm, Woodcock and Shier,
2003; de Castro et al., 1993), while providing less-refined
data sets and opportunities for anatomical analysis than x-ray
densitometry, gravimetric analysis would likely require less
time and specialized equipment.
SUPPLEMENTARY DATA
Supplementary data are available online at http://aob.oxford-
journals.org/ and consist of Fig. S1: the goodness-of-fit for
the linear mixed model for the wood density of six tree
species fitted by maximum likelihood (see Table 2, model
1); and Fig. S2: the goodness-of-fit for the radial variation in
the wood density of Melia azederach fitted by maximum like-
lihood (see Table 3, model 1).
ACKNOWLEDGEMENTS
We thank Manop Keawfoo and the staff of HKK for help in the
field. We also acknowledge the Thai Royal Forest Department
and the Center for Tropical Forest Science of the Smithsonian
Tropical Research Institute for establishing the 50-ha plot. This
work was supported by a grant from the Austrian Science
Foundation (Grant P19507-B17). The authors would also
like to thank S. C. Thomas and D. G. Woolford for suggestions
on statistical approaches and to D. Metcalfe and J. Norghauer
for helpful comments. Funding to pay the Open Access publi-
cation charges for this article was provided by the Austrian
Science Fund (FWF).
LITERATURE CITED
Ackerly DD. 1996. Canopy structure and dynamics: integration of growth pro-
cesses in tropical pioneer trees. In: Mulkey S.S., Chazdon RL, Smith AP.
eds. Tropical forest plant ecophysiology. New York, NY: Chapman and
Hall, 619–659.
Augspurger CK. 1984. Light requirements of Neotropical seedlings: a com-
parative study of growth and survival. Journal of Ecology 72: 777– 795.
Baker PJ. 2001. Age structure and stand dynamics of a seasonal tropical
forest in western Thailand. PhD Thesis, University of Washington.
Baker PJ, Bunyavejchewin S. 2006. Suppression, release and canopy recruit-
ment in five tree species from a seasonal tropical forest in western
Thailand. Journal of Tropical Ecology 22: 521– 529.
Baker PJ, Bunyavejchewin S, Oliver CD, Ashton PS. 2005. Disturbance
history and historical stand dynamics of a seasonal tropical forest in
western Thailand. Ecological Monographs 75: 317– 343.
Brown S. 2002. Measuring carbon in forests: current status and future chal-
lenges. Environmental Pollution 116: 363– 372.
Bunyavejchewin S, Baker PJ, Lafrankie JV, Ashton PS. 2001. Stand struc-
ture of a seasonal evergreen forest at the Huai Kha Khaeng Wildlife
Sanctuary, western Thailand. Natural History Bulletin of the Siam
Society 49: 89– 106.
Bunyavejchewin S, Baker PJ, Lafrankie JV, Ashton PS. 2002. Floristic
composition of a seasonal evergreen forest at the Huai Kha Khaeng
Wildlife Sanctuary, western Thailand. Natural History Bulletin of the
Siam Society 50: 125–134.
Burnham KP, Anderson DR. 2002. Model selection and multi-model infer-
ence: a practical information-theoretic approach. New York, NY:
Springer-Verlag.
de Castro F, Williamson GB, Dejesus RM. 1993. Radial variation in the
wood specific-gravity of Joannesia princeps: the roles of age and diam-
eter. Biotropica 25: 176– 182.
0
0·2 0·4 0·6 0·8 1·0
Tree growth rate (cm year–1)
1·2 1·4 1·6
5
10
Predicted change in wood density with
distance from the pith (kg m–3 cm–1)
15
20
25
FIG. 4 . Slope of the increase in wood density with distance from the pith pre-
dicted from a linear mixed model fit by maximum likelihood versus average
growth rate for Melia azedarach trees in western Thailand (Model 2, Table 3).
Nock et al. — Wood density in Thailand 305
Chave J, Andalo C, Brown S, et al. 2005. Tree allometry and improved esti-
mation of carbon stocks and balance in tropical forests. Oecologia 145:
87– 99.
Chave J, Muller-Landau HC, Baker TR, Easdale TA, Ter Steege H, Webb
CO. 2006. Regional and phylogenetic variation of wood density across
2456 Neotropical tree species. Ecological Applications 16: 2356– 2367.
Chave J, Coomes D, Jansen S, Lewis SL, Swenson NG, Zanne AE. 2009.
Towards a worldwide wood economics spectrum. Ecology Letters 12:
351–366.
Coomes DA, Grubb PJ. 2000. Impacts of root competition in forests and
woodlands: a theoretical framework and review of experiments.
Ecological Monographs 70: 171– 207.
Duncan RP. 1989. An evaluation of errors in tree age estimates based on
increment cores in kahikatea (Dacrycarpus dacrydioides). New Zealand
Natural Sciences 16: 31– 37.
Falster D, Westoby M. 2005. Alternative height strategies among 45 dicot
rain forest species from tropical Queensland, Australia. Journal of
Ecology 93: 521– 535.
van Gelder HA, Poorter L, Sterck FJ. 2006. Wood mechanics, allometry,
and life-history variation in a tropical rain forest tree community. New
Phytologist 171: 367– 378.
Givnish TJ. 1988. Adaptation to sun and shade: a whole-plant perspective.
Australian Journal of Plant Physiology 15: 63– 92.
Grabner M, Wimmer R. 2006. Variation of different tree-ring parameters as
present in a complete Norway spruce stem. Dendrochronologia 23:
111–120.
Grabner M, Wimmer R, Gierlinger N, Evans R, Downes G. 2005.
Heartwood extractives in larch and effects on X-ray densitometry.
Canadian Journal of Forest Research 35: 2781– 2786.
Hacke UG, Sperry JS, Pockman WT, Davis SD, McCulloch KA. 2001.
Trends in wood density and structure are linked to prevention of xylem
implosion by negative pressure. Oecologia 126: 457–461.
Hernandez RE, Restrepo G. 1995. Natural variation in wood properties of
Alnus acuminata H.B.K. grown in Colombia. Wood and Fiber Science
27: 41– 48.
Hothorn T, Bretz F, Westfall P. 2008. Simultaneous inference in general
parametric models. Biometrical Journal 50: 346– 363.
King DA, Davies SJ, Supardi MNN, Tan S. 2005. Tree growth is related to
light interception and wood density in two mixed dipterocarp forests of
Malaysia. Functional Ecology 19: 445– 453.
King DA, Davies SJ, Tan S, Noor NSM. 2006. The role of wood density and
stem support costs in the growth and mortality of tropical trees. Journal of
Ecology 94: 670– 680.
Lauprasert M. 1988. The creation of a permanent sample plot in dry ever-
green forest of Thailand and investigations of a suitable plot size for per-
manent sample plot programs. MSc Thesis, International Institute for
Aerospace Survey and Earth Sciences, the Netherlands.
Markesteijn L, Poorter L, Bongers F. 2007. Light-dependent leaf trait vari-
ation in 43 tropical dry forest tree species. American Journal of Botany
94: 515– 525.
Metcalfe DJ, Bradford MG, Ford AJ. 2008. Cyclone damage to tropical rain
forests: species- and community-level impacts. Austral Ecology 33:
432–441.
Muller-Landau HC. 2004. Interspecific and inter-site variation in wood
specific gravity of tropical trees. Biotropica 36: 20– 32.
Nogueira EM, Fearnside PM, Nelson BW. 2008. Normalization of wood
density in biomass estimates of Amazon forests. Forest Ecology and
Management 256: 990– 996.
Nogueira EM, Nelson BW, Fearnside PM. 2005. Wood density in dense
forest in central Amazonia, Brazil. Forest Ecology and Management
208: 261– 286.
Omolodun OO, Cutter BE, Krause GF, McGinnes EA. 1991. Wood quality
in Hildegardia barteri (Mast.) kossern – an African tropical pioneer
species. Wood and Fiber Science 23: 419– 435.
Panshin AJ, de Zeeuw C. 1980. Textbook of wood technology. New York,
NY: McGraw-Hill Publishing Co.
Parolin P. 2002. Radial gradients in wood specific gravity in trees of Central
Amazonian floodplains. IAWA Journal 23: 449– 457.
Pinheiro JC, Bates DM. 2000. Mixed effects models in S and S-plus.
New York, NY: Springer-Verlag.
Poorter L. 2008. The relationships of wood-, gas- and water fractions of tree
stems to performance and life history variation in tropical trees. Annals of
Botany 102: 367– 375.
Poorter L, Wright SJ, Paz H, et al. 2008. Are functional traits good predic-
tors of demographic rates? Evidence from five Neotropical forests.
Ecology 89: 1908– 1920.
Putz FE, Coley PD, Lu K, Montalvo A, Aiello A. 1983. Uprooting and snap-
ping of trees – structural determinants and ecological consequences
Canadian Journal of Forest Research 13: 1011– 1020.
Rosell JA, Olson ME. 2007. Testing implicit assumptions regarding the age
vs. size dependence of stem biomechanics using Pittocaulon (similar to
Senecio)praecox (Asteraceae). American Journal of Botany 94:
161–172.
Sterck FJ, Bongers F. 1998. Ontogenetic changes in size, allometry, and
mechanical design of tropical rain forest trees. American Journal of
Botany 85: 266– 272.
Swenson NG, Enquist BJ. 2007. Ecological and evolutionary determinants of
a key plant functional trait: wood density and its community-wide vari-
ation across latitude and elevation. American Journal of Botany 94:
451–459.
Troup RS. 1921. The silviculture of Indian trees. Oxford: Clarendon Press.
Whitmore JL. 1973. Wood density variation in Costa Rican balsa. Wood
Science 5: 223–229.
Wiemann MC, Williamson GB. 1988. Extreme radial changes in wood
specific-gravity in some tropical pioneers. Wood and Fiber Science 20:
344–349.
Wiemann MC, Williamson GB. 1989a.Radial gradients in the specific-
gravity of wood in some tropical and temperate trees Forest Science
35: 197–210.
Wiemann MC, Williamson GB. 1989b.Wood specific-gravity gradients in
tropical dry and montane rain-forest trees. American Journal of Botany
76: 924–928.
Woodcock DW, Shier AD. 2002. Wood specific gravity and its radial vari-
ations: the many ways to make a tree. Trees – Structure and Function
16: 437–443.
Woodcock DW, Shier AD. 2003. Does canopy position affect wood specific
gravity in temperate forest trees? Annals of Botany 91: 529– 537.
Worbes M. 2002. One hundred years of tree-ring research in the tropics – a
brief history and an outlook to future challenges. Dendrochronologia
20: 217–231.
Nock et al. — Wood density in Thailand306
