Forest Ecology and Management 257 (2009) 1684–1694

Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage: www.elsevier.com/locate/foreco

Allometric equations for estimating the above-ground biomass in tropical

lowland Dipterocarp forests
T.M. Basuki *, P.E. van Laake, A.K. Skidmore, Y.A. Hussin
Department of Natural Resources, International Institute for Geo-information Science and Earth Observation (ITC), P.O. Box 6, 7500 AA Enschede, The Netherlands

A R T I C L E I N F O A B S T R A C T

Article history: Allometric equations can be used to estimate the biomass and carbon stock of forests. However, so far the
Received 19 May 2008 equations for Dipterocarp forests have not been developed in sufficient detail. In this research, allometric
Received in revised form 12 January 2009 equations are presented based on the genera of commercial species and mixed species. Separate
Accepted 21 January 2009
equations are developed for the Dipterocarpus, Hopea, Palaquium and Shorea genera, and an equation of a
mix of these genera represents commercial species. The mixed species is constructed from commercial
Keywords: and non-commercial species. The data were collected in lowland mixed Dipterocarp forests in East
DBH
Kalimantan, Indonesia. The number of trees sampled in this research was 122, with diameters (1.30 m or
Destructive sampling
above buttresses) ranging from 6 to 200 cm. Destructive sampling was used to collect the samples where
Site specific equation
Carbon stock diameter at breast height (DBH), commercial bole height (CBH), and wood density were used as
predictors for dry weight of total above-ground biomass (TAGB). Model comparison and selection were
based on Akaike Information Criterion (AIC), slope coefficient of the regression, average deviation,
confidence interval (CI) of the mean, paired t-test. Based on these statistical indicators, the most suitable
model is ln(TAGB) = c + aln(DBH). This model uses only a single predictor of DBH and produces a range of
prediction values closer to the upper and lower limits of the observed mean. Model 1 is reliable for forest
managers to estimate above-ground biomass, so the research findings can be extrapolated for managing
forests related to carbon balance. Additional explanatory variables such as CBH do not really increase the
indicators’ goodness of fit for the equation. An alternative model to incorporate wood density must be
considered for estimating the above-ground biomass for mixed species. Comparing the presented
equations to previously published data shows that these local species-specific and generic equations
differ substantially from previously published equations and that site specific equations must be
considered to get a better estimation of biomass. Based on the average deviation and the range of CI, the
generalized equations are not sufficient to estimate the biomass for a certain type of forests, such as
lowland Dipterocarp forests. The research findings are new for Dipterocarp forests, so they complement
the previous research as well as the methodology of the Good Practice Guidance for Land Use and Land
Use Change and Forestry (GPG-LULUCF).
ß 2009 Elsevier B.V. All rights reserved.

1. Introduction emerging mechanisms, such as Reducing Emissions from Defor-

estation in Developing Countries (REDD), and they are likely to
The accurate estimation of biomass in tropical forests is crucial require temporally and spatially fine-gained assessments of carbon
for many applications, from the commercial exploitation of timber stock (UNFCCC, 2008).
to the global carbon cycle. Particularly in the latter context the Carbon stock is typically derived from above-ground biomass
estimation of the total above-ground biomass (TAGB) with an by assuming that 50% of the biomass is made up by carbon. The
accuracy sufficient to establish the increments or decrements in most accurate method for the estimation of biomass is through
carbon stored in the forest over relatively short periods of time (2– cutting of trees and weighing of their parts. This destructive
10 years) is increasingly important. Under the United Nations method is often used to validate others, less invasive and costly
Framework Convention on Climate Change (UNFCCC), countries methods, such as the estimation of carbon stock using non-
have to report regularly the state of their forest resources and destructive in-situ measurements and remote sensing (Clark et al.,
2001; Wang et al., 2003). Allometric equations developed on the
basis of sparse measurements from destructive sampling are
* Corresponding author. Tel.: +31 53 4874444; fax: +31 53 4874388. related to more easily collected biophysical properties of trees,
E-mail addresses: basuki@itc.nl, tmbasuki@yahoo.com (T.M. Basuki). such as diameter at breast height (DBH) and commercial bole

0378-1127/$ – see front matter ß 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.foreco.2009.01.027
 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694 1685

height (CBH). The estimation of carbon over large areas using branch dry weight was then used to model leaf dry weight. The
remote sensing is supported by correlating the reflection of the equations developed by Yamakura et al. (1986) are not practical for
canopy recorded at the sensor to the carbon measured directly or most uses since the equation does not directly predict the TAGB.
estimated indirectly on the ground (Chiesi et al., 2005; Gibbs et al., The lack of allometric equations for Dipterocarp forests is also
2007; Myeong et al., 2006; Tan et al., 2007). evident in the Good Practice Guidance for Land Use and Land Use
In this research, tree allometric equations are developed by Change and Forestry or GPG-LULUCF (IPCC, 2003; further abbre-
establishing the relationship between tree parameters such as viated as GPG). In this document, there are only two allometric
DBH, CBH, and wood density with above-ground biomass. Various equations for tropical forests, while in fact there are many
allometric equations have been developed for tropical rain forests differences in the characteristics of tropical forests. The accuracy
(Araújo et al., 1999; Brown, 1997; Chambers et al., 2001; Chave or uncertainty of models is an important aspect that is mentioned in
et al., 2001, 2005; Keller et al., 2001; Nelson et al., 1999). However, the GPG and the different instruments of the Kyoto Protocol. To
there are few allometric equations developed specifically for reduce uncertainty, accurate carbon accounting methods are
lowland Dipterocarp forests, despite the fact that this type of forest required. The development of new, species-specific allometric
covers extensive areas in tropical South-East Asia: around 59% of equations are necessary to achieve higher levels of accuracy, and
forests in Kalimantan and 53% of forests in Sumatra, Indonesia we present some new equations here to achieve a better estimation
(Tyrie, 1999), and that they are among the most commercialized of carbon stock for tropical lowland Dipterocarp forests.
hard wood species from South-East Asia. So far, allometric
equations for multi-species tropical forests of Indonesia have 2. Methods
been published in Brown (1997), Hashimotio et al. (2000),
Ketterings et al. (2001) and Yamakura et al. (1986). The 2.1. Study area
applicability of the equations needs to be affirmed before they
can be applied to monotypic Dipterocarp forests. The study was executed at two sites in Berau Regency, East
Brown (1997) developed allometric equations for tropical forests Kalimantan, Indonesia. The first site is located in Labanan (18450 to
using data collected from Kalimantan and other tropical regions. A 28100 North latitude and from 1168550 to 1178200 East longitude),
logistic curve based on tree age was constructed by Hashimotio et al. Teluk Bayur District. In this location, trees with the diameters (DBH
(2000) in fallow forests of East Kalimantan. The equation of or above buttresses) mostly ranging from 5 to 70 cm were felled. The
Hashimotio et al. (2000) cannot be applied to Dipterocarp forests second site is in the Puji Sempurna Raharja forest concession area,
as it requires stand age as explanatory variable which is not Merancang, around 60 km North-East of Labanan (28110 to 28200
available, and due to the differences in ecological environments North latitude and 1178380 to 1188110 East longitude), where trees
between fallow and natural forests. Ketterings et al. (2001) with diameters of 70 up to 200 cm were felled. These study areas are
established an allometric equation in mixed secondary forest in located within the same forest type, namely lowland mixed
Sumatra, but this forest was not classified as Dipterocarp forest. Dipterocarp. The dominant family in these forests is Dipterocarpa-
Yamakura et al. (1986) constructed allometric equations using data ceae. These forests consist of commercial, non-commercial and
collected from Dipterocarp forests from Sebulu, East Kalimantan. protected species. The map of the study area is presented in Fig. 1.
They used DBH and tree height to predict stem dry weight. The stem Based on the data from 1971 to 1997, the mean annual rainfall
dry weight was used to predict branch dry weight, and stem and in the study area is 2000 mm. The months from June to October are

Fig. 1. The locations of the sampling sites in Labanan (the first site) and Merancang (the second site).
1686 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694

dry, while the wet months are from November to May. The rainfall trunks with their barks were included (Nelson et al., 1999). Wood
during the dry season remains more than 100 mm per month. The density was measured by the water replacement method. To avoid
temperature ranges from 21 to 34 8C and the mean is 26 8C (BFMP, shrinkage during volume measurement, the samples were first
1999). saturated. In so doing, rehydration was conducted for 48 h. The
At the Labanan study area, the dominant soil types are red volume of each sample was determined from the volume of the
yellow podzolic or ultisols. However, the soil properties within a water displaced when submerged. The wood density was
soil type vary according to the terrain conditions. Minor soil types calculated as dry oven weight divided by saturated volume. The
include oxisols, vertisols and inceptisols (Mantel, 1999). As in results for the individual specimens are presented in Appendix A.
Labanan, the dominant soil of the Merancang study area is red The dry weight of the stumps, stems, and branches with the
yellow podzolic (66%), and the remainder is Alluvial. diameter of >15 cm was calculated by multiplying the fresh
volume of each section by wood density. For the other partitioned
2.2. Data collection trees, the dry weight was calculated through fresh weight
multiplied by dry weight/fresh weight ratio of the corresponding
Trees selection was based on species grouping according to samples. The total dry weight of a tree was obtained by summing
Gunawan and Rathert (1999), the inventory data of the conces- the dry weight of the stump, stem, branches, twigs, and leaves.
sioners, and a ground check. Forest inventories were conducted
within 83,240 ha of the areas managed by PT Hutan Sanggam 2.3. Data analysis
Labanan Lestari and 51,000 ha of the areas managed by PT Puji
Sempurna Raharja. To obtain the representative tree samples, Based on the data collected, several equations were developed.
diameter distributions and the species grouping were taken into Firstly, the equations were developed for four individual genera
account during tree selection. The data were collected twice: the Dipterocarpus, Hopea, Palaquium and Shorea. Secondly, these four
first data set was collected in March to April 2006, and the second genera were mixed to develop an equation for commercial species.
data set was collected in September to October 2006. The first data Finally, the four genera together with the second data set (Samalca,
set consisted of four commercial genera including Dipterocarpus 2007) were used to develop an allometric equation of mixed species.
(20 trees, five species), Hopea (20 trees, three species), Palaquium Before establishing the allometric equation, scatter plots were
(19 trees, three species) and Shorea (24 trees, nine species). The used to see whether the relationship between independent and
first data set was collected in Labanan and Puji Sempurna Raharja dependent variables was linear. Furthermore, several allometric
concession areas, Merancang. The second data set was collected in relationships between independent and dependent variables were
Labanan, covering commercial (excluding the four mentioned tested. The independent variables included DBH, CBH, and wood
genera) and non-commercial species that consisted of 39 trees density, whereas the dependent variable was the dry weight of the
from 27 species (Samalca, 2007). These two data sets were used for TAGB. Because the data exhibited heteroscedasticity, a power
destructive sampling. The diameters were 6.5–135 cm, 6.5–87 cm, function was an inappropriate model in this study, so we
6.3–74 cm, 6.5–200 cm, and 6–68.9 cm for Dipterocarpus, Hopea, transformed the data for linear regression using a natural
Palaquium, Shorea and the second data set, respectively. The overall logarithm. The transformation equalized the variance over the
DBH distributions were: >6–20 cm, 29 trees, >20–40 cm, 36 trees, entire range of biomass values which satisfies the prerequisite of
>40–60 cm, 28 trees, >60–80 cm, 12 trees, >80–100, 9 trees, linear regression (Sokal and Rohlf, 1995; Sprugel, 1983). However,
>100–120 cm, 5 trees and >120–200 cm, 3 trees. this transformation introduced a systematic bias in the calculation
Before conducting the destructive sampling, the DBH was which was corrected using a correction factor (CF) when back-
measured. Generally the DBH was measured at 1.30 m above- transforming the calculation into biomass (Chave et al., 2005; Sah
ground, but for trees with enlargement or buttresses, the diameter et al., 2004; Son et al., 2001; Sprugel, 1983).
was measured at 30 cm above the main enlargement (FAO, 2004). Model comparison and selection were based on average
After felling, the tree height was measured. Diameter was measured deviation (Brand and Smith, 1985; Cairns et al., 2003; Chave
at 2 m intervals for the stems and big branches with the diameters of et al., 2005; Nelson et al., 1999), slope coefficient of the regression
more than 15 cm. In addition, the stump height and its diameter (Nelson et al., 1999), Akaike Information Criterion (AIC) (Burnham
were also measured. These measurements were used to estimate the and Anderson, 2002; Chave et al., 2005), confidence interval (CI) of
volume and dry weight. The volume of each section was calculated the predictions, and paired t-test. Coefficients of determination (r2)
using Smalian’s formula as cited by De Gier (2003). The total volume more than 90% are reported in this paper.
was obtained by summing the volume of each section. Due to the The average deviation was computed from the absolute
difference in moisture content, the tree material was partitioned difference between predicted and observed dry weight and
into leaves, twigs (diameter <3.2 cm), small branches (diameter expressed as the percentage of observed dry weight, then all
3.2–6.4 cm), large branches (diameter >6.4 cm) and stem (Ketter- deviations were averaged (Brand and Smith, 1985; Cairns et al.,
ings et al., 2001). The fresh weight of the leaves, twigs, branches and 2003; Chave et al., 2005; Nelson et al., 1999). The equation to
stems with the diameters equal or less than 15 cm were weighed in calculate average deviation is shown in Eq. (1). The deviation was
the field using spring weighing scales of 50 and 25 kg capacity with calculated after the prediction was back-transformed to the unit
accuracy 1%. A table scale was used to weigh smaller specimens. The values and corrected using a CF. The average deviation has been
latter balance had a capacity of 2000 g with accuracy 0.5%. calculated as follows:
The samples from the partitioned trees were taken in three
replications and stored in sealed plastic bags, and then sent to the 100 X
n
jŶi  Y i j
S̄ð%Þ ¼ (1)
laboratory to determine their moisture content. In the laboratory, n i¼1 Yi
an analytical balance with capacity of 420 g and accuracy of 0.001 g
was used to weigh the samples. Dry weights were obtained by where S̄ is the average deviation, Yi = the observed dry weight,
drying the samples at a temperature of 105 8C until a constant Ŷi ¼ the predicted dry weight, n = number of observations.
weight was achieved (Stewart et al., 1992; Ketterings et al., 2001). The formula of AIC used as the criterion for model selection
To determine the wood density, samples were taken from the (Chave et al., 2005) is:
lower, middle and upper parts of the stems. The samples were
taken as a pie shape or cylinder, so the inner and outer parts of the AIC ¼ 2lnðLÞ þ 2 p (2)
 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694 1687

where L is the likelihood of the fitted model and p is the total on DBH measurements and wood density was applied to the
number of parameters in the model. The optimal model will current data. The equation is:
minimize the AIC value (Chave et al., 2005). AIC is used to compare 2
non-nested models. TAGB ¼ rexpð1:499 þ 2:148 lnðDBHÞ þ 0:207ðlnðDBHÞÞ
3
 0:0281ðlnðDBHÞÞ Þ (4)
2.4. Comparing the equations to previously published equations
3
where r = species-specific wood density (g/cm ). In addition to the
The proposed mixed species model was applied to the data of models mentioned above, the model of Ketterings et al. (2001) was
Ketterings et al. (2001). Besides that, we also employed the models chosen. Ketterings’ data consisted of 29 trees from 14 genera with
of Brown (1997), one of the pan-tropic equations of Chave et al. the diameters ranging from 7.6 to 48.1 cm. The allometric equation
(2005), and Ketterings et al. (2001) to the current data. developed by Ketterings et al. (2001) is:
The allometric equation developed by Brown (1997) for tropical
TAGB ¼ r ravg ðDBHÞ2þc (5)
moist forest is:
where r is a parameter that is constant over wide range of geog-
TAGB ¼ expð2:134 þ 2:53 lnðDBHÞÞ; (3)
raphical areas, ravg is the average wood density for the study areas,
where TAGB is total above-ground biomass in kg/tree and DBH is in and c is a parameter estimated from relationship between tree height
cm. The equation of Brown (1997) was constructed from the data and DBH. For the study areas, c is 0.397, r is 0.604 g/cm3, and r is 0.11.
collected by several authors from different tropical countries and
at different times. The diameters used to establish this equation 3. Results
ranged from 5 to 148 cm, and the number of sample trees was 170.
Besides the model of Brown (1997), one of the pan-tropic models of 3.1. Developing allometric equations
Chave et al. (2005) was chosen as a comparison of the proposed
models. Pan-tropic models were developed from various tropical The model initially developed is presented in Eq. (6):
forests based on the compilation of data since 1950s from 27 study
lnðTAGBÞ ¼ c þ alnðDBHÞ (6)
sites in America, Asia and Oceania (Chave et al., 2005). The samples
were collected from 2410 trees with DBH that ranged from 5 to where TAGB is in kg/tree, DBH is in cm, c is the intercept, and a is
156 cm. The best pan-tropic model for moist tropical forest based the slope coefficient of the regression. The values of the coefficients

Fig. 2. The linear regression of natural log transformation of DBH (cm) and total above-ground biomass (TAGB) (kg/tree) of Dipterocarpus (a), Hopea (b), Palaquium (c), Shorea
(d), commercial species (e) and mixed species (f). The number of trees for every regression is 20, 20, 19, 24, 83 and 122, respectively.
1688 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694

are presented in Table 1 and the regression lines are presented in was in the form of main stems and not in branches and leaves. It
Fig. 2. For this model, the adjusted r2 ranged from 0.963 to 0.989 accounted of 45–90% with the average of 67% of the TAGB. By
and the lowest was found for mixed species. The average deviation incorporating CBH (in meters) as the second predictor in model 2,
was therefore the highest for the mixed species. Among the genera, the model becomes a multiple linear regression as follows:
Shorea had the highest average deviation (19.6%). The allometric
lnðTAGBÞ ¼ c þ alnðDBHÞ þ blnðCBHÞ (7)
equation for every genus, commercial species, and mixed species
had a significant slope coefficient at p < 0.001 (Table 1). Since correlation between DBH and TAGB has been high, the
Model 1 uses only DBH as a predictor. In fact, tree biomass is addition of CBH in model 2 only increased slightly r2 and also
affected by its height and wood density as well. Therefore, CBH and reduced slightly the average deviation. Table 1 shows that at the
wood density were incorporated as additional independent genus level, the b coefficients from the inclusion of CBH are
variables. Even though measuring the total tree height was easy significant at p < 0.05 for Dipterocarpus and Hopea, and not
for felled trees, CBH was used instead of total height due to the statistically significant for Shorea (Table 1).
practical difficulty of measuring standing trees in the field and the As mentioned above, wood density is an important factor when
properties of the tree architecture in the study areas. Regarding calculating biomass (Baker et al., 2004; Chave et al., 2005;
tree architecture, the bulk of tree biomass measured in the field Nogueira et al., 2007), so it was added as a predictor variable.

Table 1
Model description for the estimation of the total above-ground biomass of Dipterocarp forests.

Species grouping N Allometric equation Coefficient Standard error of Adjusted r2 Standard error Average CF
the coefficient of residual deviation (%)
Symbol Value

Dipterocarpus 20 ln(TAGB) = c + aln(DBH) c 1.232*** 0.200 0.989*** 0.210 18.65 1.022

a 2.178*** 0.053
ln(TAGB) = c + aln(DBH) + bln(CBH) c 2.187*** 0.428 0.992*** 0.183 14.15 1.017
a 2.007*** 0.084
b 0.575* 0.236
ln(TAGB) = c + aln(DBH) + bln(WD) c 1.190** 0.336 0.989*** 0.213 18.84 1.023
a 2.175*** 0.057
b 0.082ns 0.488

Hopea 20 ln(TAGB) = c + aln(DBH) c 1.813*** 0.213 0.987*** 0.184 14.08 1.017

a 2.339*** 0.061
ln(TAGB) = c + aln(DBH) + bln(CBH) c 2.856*** 0.480 0.990*** 0.165 12.76 1.014
a 2.116*** 0.109
b 0.656* 0.277
ln(TAGB) = c + aln(DBH) + bln(WD) c 1.708*** 0.301 0.987*** 0.188 13.87 1.018
a 2.335*** 0.063
b 0.174ns 0.344

Palaquium 19 ln(TAGB) = c + aln(DBH) c 1.098*** 0.281 0.975*** 0.230 18.78 1.027

a 2.142*** 0.082
ln(TAGB) = c + aln(DBH) + bln(CBH) c 1.663*** 0.275 0.984*** 0.181 13.72 1.017
a 1.817*** 0.114
b 0.612** 0.177
ln(TAGB) = c + aln(DBH) + bln(WD) c 0.723* 0.286 0.980*** 0.201 16.86 1.020
a 2.145*** 0.071
b 0.704* 0.273

Shorea 24 ln(TAGB) = c + aln(DBH) c 2.193*** 0.253 0.984*** 0.2601 20.49 1.034

a 2.371*** 0.063
ln(TAGB) = c + aln(DBH) + bln(CBH) c 2.758*** 0.417 0.985*** 0.250 18.60 1.032
a 2.178*** 0.131
b 0.463ns 0.277
ln(TAGB) = c + aln(DBH) + bln(WD) c 1.533*** 0.405 0.986*** 0.244 19.14 1.030
a 2.294*** 0.070
b 0.560ns 0.278

Commercial species 83 ln(TAGB) = c + aln(DBH) c 1.498*** 0.127 0.981*** 0.252 21.61 1.032
a 2.234*** 0.034
ln(TAGB) = c + aln(DBH) + bln(CBH) c 2.266*** 0.213 0.985*** 0.227 18.63 1.026
a 2.030*** 0.056
b 0.542*** 0.125
ln(TAGB) = c + aln(DBH) + bln(WD) c 1.045*** 0.150 0.985 0.225 18.77 1.057
a 2.203*** 0.031
b 0.639*** 0.138

Mixed species 122 ln(TAGB) = c + aln(DBH) c 1.201*** 0.141 0.963 0.335 30.32 1.058
a 2.196*** 0.039
ln(TAGB) = c + aln(DBH) + bln(CBH) c 1.935*** 0.234 0.967*** 0.318 27.49 1.052
a 1.981*** 0.067
b 0.541*** 0.141
ln(TAGB) = c + aln(DBH) + bln(WD) c 0.744*** 0.154 0.970*** 0.303 26.51 1.047
a 2.188*** 0.035
b 0.832*** 0.157

Note: The statistical analyses are significant at 95% confidence interval. ***p < 0.001; **p < 0.01; *p < 0.05; and non-significant, nsp > 0.05. N; the number of tree samples,
TAGB; total above-ground biomass based on dry weight (kg/tree), DBH: diameter at breast height (cm), CBH: commercial bole height (m), WD: wood density (g/cm3), and CF:
correction factor.
 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694 1689

Adding wood density in the model is important in order to

estimate the biomass for mixed species and big trees, since
biomass estimates for larger DBH trees are more variable and have
a disproportionately large contribution to forest biomass.
By incorporating wood density in the third model, the equation
becomes:

lnðTAGBÞ ¼ c þ alnðDBHÞ þ blnðWDÞ (8)

where WD is wood density for every measured tree. The b

coefficient in model 3 was not statistically significant for
Dipterocarpus, Hopea and Shorea, whereas for the commercial
and mixed species equations, it was highly significant at p < 0.001
(Table 1).
To compare model 2 and 3, AIC was calculated and the results
are presented in Table 2. The AIC values for model 3 were lower
than that for model 2, except for Dipterocarpus and Hopea. Based on
the species grouping, the highest AIC values were found for the
mixed species followed by the commercial species and the lowest
were in the genera. Fig. 3. The wood density of 1: D. pacyphyllus (n = 8), 2: H. cernua (n = 18), 3: P. gutta
(n = 13) and 4: S. retusa (n = 6), the dominant species of Dipterocarpaceae, Hopea,
Our estimates of wood densities were compared with the data
Palaqium and Shorea, respectively. The cycle shapes with values of 0.76 and 0.43 are
from GPG and from Plant Resources of South East Asia (PROSEA) the outlier of H. cernua from the current data. The rectangle shapes indicate the
(Soerianegara and Lemmens, 1993). The results show that increasing highest and the lowest wood density of P gutta and H cernua from the data of PROSEA.
DBH is not followed by an increase in wood density. This finding is in
agreement with the previous research by Baker et al. (2004) and
Nogueira et al. (2005, 2007). The comparison of the wood density
values based on the current study and those from the published data
of GPG and PROSEA (Soerianegara and Lemmens, 1993) is presented
in Appendix A. In the GPG, for Shorea only Shorea spp. balau group,
Shorea spp. (dark red meranti), Shorea spp. (light red meranti), etc.
are mentioned but their species are not. In fact, the commercial
timber red meranti consists of many species, such as S. parvifolia, S.
smithiana, S. macroptera. The wood density for the dominant species
from every genus is superimposed with the wood density from the
corresponding species of the published data (PROSEA) and
presented in Fig. 3. Two outliers of wood density, indicated by
the circles, are found from Palaquium gutta of the current data. The
first is at a DBH of 50 cm with a wood density of 0.43 g/cm3, and the
second is at a DBH of 28.6 cm with a wood density of 0.76 g/cm3. The
data from the PROSEA are indicated by the rectangular shapes which
show the highest and the lowest values of wood densities of H.
cernua and P. gutta.
Fig. 4. The regression line in model 1 of the mixed species (122 sample trees) as
applied to the published data by Ketterings et al. (2001).
3.2. Comparing the equations to previously published equations

Model 1 of the mixed species regression was applied to the data

of Ketterings et al. (2001). The predicted TAGB using the first model
to the Ketterings’ data produced the same trend as the observed
data (Fig. 4).

Table 2
The Akaike Information Criterion (AIC) for model 2 and 3.

Species grouping Model AIC

Dipterocarpus 2 321
3 326

Hopea 2 282
3 303

Palaquium 2 271
3 264

Shorea 2 458
3 437

Commercial species 2 1490

3 1464

Mixed species 2 2176 Fig. 5. The DBH versus dry weight of the total above-ground biomass for the mixed
3 2157 species (122 sample trees) from the observed data and the prediction lines using
model 1, models of Brown (1997) and Ketterings et al. (2001).
1690 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694

Table 3 Table 5
The average deviation of various models. Paired t-test at 95% confidence interval of the mean of the total above-ground
biomass (kg) for mixed species (122 trees).
Species grouping The employed equation Average deviation (S̄)*(%)
Pair t-statistic Significance
Dipterocarpus Model 1 18.65 (one tailed)
Model 2 14.15
Model 3 18.84 Observed–model of Brown (1997) 4.184 0.000
Brown (1997) 62.78 Observed–model of Chave et al. (2005) 4.849 0.000
Chave et al. (2005) 89.67 Observed–model of Ketterings et al. (2001) 4.330 0.000
Ketterings et al. (2001) 46.81 Model 1–model 2 2.575 0.006
Model 1–model 3 0.921 0.179
Hopea Model 1 14.08 Model 1–model of Brown (1997) 3.978 0.000
Model 2 12.76 Model 1–model of Chave et al. (2005) 4.346 0.000
Model 3 13.87 Model 1–model of Ketterings et al. (2001) 6.669 0.000
Brown (1997) 42.94
Chave et al. (2005) 52.08
Ketterings et al. (2001) 49.46

Palaquium Model 1 18.78 4. Discussion

Model 2 13.72
Model 3 16.86 4.1. Allometric equations
Brown (1997) 43.94
Chave et al. (2005) 54.13
Ketterings et al. (2001) 50.71
When the first and the second data sets were used to generate the
equations for mixed species, the deviations of the predicted biomass
Shorea Model 1 20.49
was higher than the observed one. This is due to the higher variation
Model 2 18.60
Model 3 19.14
in tree characteristics among those species, and in the genera
Brown (1997) 107.16 between the first and the second data set. Based on the raw data, it
Chave et al. (2005) 94.44 can be observed that the second data set for the DBH of 6, 36, and
Ketterings et al. (2001) 32.15 68 cm has dry weights of 19, 1100, and 6300 kg biomass,
Commercial species Model 1 21.61 respectively, whereas for the first data set at those diameters the
Model 2 18.63 biomass is 10–14, 900, and 3800 kg, respectively. These results are in
Model 3 18.77 agreement with the research of Nelson et al. (1999) who found that
Brown (1997) 66.77
by using the same allometric equation with DBH as the independent
Chave et al. (2005) 73.85
Ketterings et al. (2001) 44.02 variable, the deviation in species-specific regressions varies from
10.9% to 14.7% for trees with the diameters from 5.1 to 38.2 cm, but if
Mixed species Model 1 30.32
Model 2 27.49
these species were mixed, the average deviation was 19.8%.
Model 3 26.51 In the equation based on the individual genus, Hopea has the
Brown (1997) 51.95 lowest average deviation, because the sample of this genus only
Chave et al. (2005) 55.09 consists of three species. On the other hand, within the Shorea
Ketterings et al. (2001) 51.89
group, the average deviation is the highest because it consists of 10
Pn
*
S̄ð%Þ ¼ 100
n i¼1 jŶi  Y i j=Y i , S̄ ¼ average deviation, Yi = observed values, species and the diameter range is wider (6.5–200 cm).
Ŷi ¼ predicted values. At the genus level for model 2, the lower significance level of b,
when compared to the a coefficient, may be caused by effect of
The average deviation for individual trees of the previously multicollinearity. In a multiple linear regression, multicollinearity
published models is always higher than that of models 1, 2, and 3 of causes partial regression coefficients for one or both independent
the mixed species (Table 3). At the genus level, prediction of TAGB variables to be less precise, and t-values to become less significant
using the equation of Brown (1997) resulted average deviation (Nelson et al., 1999). In the current data, the tolerance values for
ranging from 43% to 107% and from 52% to 94% for the equation of every genus and mixed species do not indicate multicollinearity.
Chave et al. (2005). However, based on Pearson correlations between DBH and CBH, it
When the equations of Chave et al. (2005) and Brown (1997) shows a fair correlation for every genus, except Palaquium. So, this
were applied to our data, the predicted values were over estimated. correlation induces multicollinearity in model 2, but the multi-
In contrast, the model of Ketterings et al. (2001) underestimated collinearity is weak because the b coefficient is still significant at
the TAGB. This evident can be seen from the CI values presented in p < 0.05, except for Shorea. Multicollinearity disappears for the
Table 4. At 95% CI, upper and lower limit of the mean TAGB from allometric equations of commercial and mixed species, because
the model of Brown (1997) and Chave et al. (2005) were higher these equations were constructed from a larger number of samples
than the observed values. In addition, paired t-test presented in where some of them have weak multicollinearity and the others do
Table 5 shows that for one tailed at a = 0.05, the mean of the not indicate multicollinearity at all.
observed data and the proposed models is significantly different The role of wood density in the allometric equation is more
from the predicted mean using the previously published models. prominent for the mixed species than in the genera. As presented in
Fig. 5 shows the observed values and the prediction line using Table 1, the b coefficients of commercial and mixed species equations
model 1, model of Brown (1997) and Ketterings et al. (2001). are significant at the 95% CI. The evidence that the variation of wood

Table 4
The confidence interval (CI) of the mean from various models for mixed species (122 trees).

Parameters Observed Model 1 Model 2 Model 3 Brown Chave Ketterings

Mean TAGB (kg) 2284.21 2416.69 2280.74 2458.60 3832.45 4180.43 1239.39
95% CI Lower limit of mean TAGB (kg) 1376.51 1456.35 1526.76 1570.11 2297.53 2581.58 759.25
95% CI Upper limit of mean TAGB (kg) 3191.90 3377.35 3034.73 3347.09 5367.38 5779.27 1719.25
The number of trees 122 122 122 122 122 122 122
 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694 1691

density among the genera is higher than within a single genus is also Kalimantan, as used by Brown (1997) and Chave et al. (2005),
supported by Baker et al. (2004) and Chave et al. (2006). are the same as the trees used in this study. As can be seen from
The importance of including wood density in biomass estima- Brown’s data, one of the trees from Kalimantan with the diameter
tion can be examined for big trees, such as Shorea superba and of 130 cm had dry weight of 42.8 tons (Brown, 1997), whereas
Shorea sp. The S. superba with a diameter of 170 cm has a TAGB dry from the current research, Shorea sp. with a diameter of 200 cm has
weight of 39.7 tons, whereas the other (Shorea sp.) with a diameter dry weight of 36 tons. However, the prediction line of Ketterings’
of 200 cm has a lower dry weight of TAGB, that is 36 tons. It is likely equation lies below the observed values and the prediction line of
that the differences in wood density and tree architecture explain model 1 (Fig. 5). This may be caused by the marked difference
the differences in the dry weight of these two species. Although S. between the sampled trees in this study and Ketterings’ data. The
superba has smaller DBH and shorter CBH, it has higher wood only species which are found both in Ketterings’ data and this
density compared to that of Shorea sp., resulting in the same dry study are Shorea and Alseodaphne, sp. The lower prediction of
weight of commercial bole (27.8 tons). The characteristics of these Ketterings’ equation is because the trees used to construct
species show that their wood density, DBH and CBH are 0.86 g/cm3, Ketterings’ equation were much smaller than those from the
170 cm and 26 m for S. superba, and for Shorea sp., they are 0. 57 g/ current study, as elaborated in the previous section. In addition to
cm3, 200 cm and 28 m, respectively. In this research, the genus of the explanations above, the inclusion of wood density in assessing
Shorea consists of several big trees, however, adding wood density biomass carbon will reduce uncertainty due to the variation among
to the model does not significantly influence b coefficient (Table 1). differences of sites (Chave et al., 2006; Baker et al., 2004).
The measurements of wood density of different species from Based on the application of the proposed model and the
the lowland Dipterocarp forests can be useful to complement the previously published data, for an accurate biomass estimation, one
proposed methodology for estimating carbon stock change. In must consider site specific equations. This finding is supported by
addition, for several of the species reported here no wood density Cairns et al. (2003) and Nelson et al. (1999) when they apply a
values have previously been published. In the GPG, only a single previously published equation to their data. Nelson et al. (1999)
value for wood density is given for certain species (see Appendix overestimated biomass prediction by 10–60% for trees with a DBH
3A.1. of GPG). With respect to the GPG methodology for estimating from 5 to 25 cm, and showed an even greater over-estimation for
carbon stock change, the major source of uncertainty for trees with larger DBH. In contrast with these results, Chave et al.
estimating carbon stock using the default method is related to (2005) and Gibbs et al. (2007) stated that for tropical forests, local
the applicability of these parameters for the diverse age and species-specific allometric equations are not needed; instead,
composition structure of specific stands (IPCC, 2003). Thus, the generalized allometric relationships must be employed. Moreover,
various values of wood density from different characteristics of grouping species by broad forests types or ecological zones is
forests must be considered. As an example, in this study the wood highly more effective than generating allometric equations for
density of Dipterocarpus grandiflorus is 0.56 g/cm3 for a DBH of local conditions or species-specific allometric equations, because
18.8 cm and 0.75 g/cm3 for a DBH of 44 cm (Appendix A), while in the local equations will not improve accuracy significantly.
the GPG the wood density of this species is 0.62 g/cm3, without any However, based on the analysis of the 95% CI of the mean, the
indication of the diameter. prediction using the pan-tropic allometric equation by Chave et al.
(2005) showed that the lower limit of the prediction is much
4.2. Fitting and applicability of the models higher than the observed values (Table 4). The lower boundaries of
the observed data, the prediction using model 1, and the prediction
Even though most AIC values of model 3 are lower than model 2 using pan-tropic model are 1376.5, 1456.3, and 2581.6, respec-
(Table 2), the values are relatively the same. Therefore for fitting tively. The average deviations of the lower limit CI of these two
the models, the emphasis will be given more on the coefficient of predictions from the observed data are 6% and 87%.
the regression, average deviation, the range of CI, and paired t-test. Paired t-test presented in Table 5 supports the CI discussed
Of the proposed models, model 1 gives a better prediction than above. At 95% CI, the mean of the observed and the proposed
model 2. This can be observed from the inclusion of CBH into the models are significantly lower than the mean of the models of
model. It increases r2 slightly and the average deviation decreases Brown (1997) and Chave et al. (2005), but higher compared to that
slightly, but the slope less significant and the standard error of a of Ketterings et al. (2001).
coefficient also increases. Based on the lower and upper limit CI of With regards to GPG (Appendix 4A.2), the developed model can
the mixed species, model 1 is closer to the range of the mean of the be used to complement the existing equations by Brown which are
observed values (Table 4). In contrast, the upper and lower limit of used in GPG.
CI of model 2 does not reach the range of CI of the observed values.
For the mixed species, model 1 is significantly different from model 4.3. Sources of error
2 (Table 5). Model 1 is therefore the preferred model for estimating
the TAGB of Dipterocarp forests. In developing the equations in this study several potential
Within the genus, model 3 does not give effect over model 1 sources of errors could be identified:
since b coefficient does not significantly influence the model
(Table 1). However, adding wood density as a predictor is an  Wood density differs among the tree sections: it is higher at
alternative to estimate the biomass of big trees and mixed species. breast height than at the top of bole (Nogueira et al., 2005) and
The b coefficient is significant when wood density is incorporated also higher at the base of the tree stem than that at the base of the
into mixed species and the average deviation decreases. The effect living crown (Cordero and Kanninen, 2002). In the current study,
of wood density to the dry weight is evident for big trees from the the samples for wood density analysis were taken from the
current data set as explained previously and from the different upper, middle and lower of the main trunk. However, these data
data set of Brown as explained in the next paragraph. were also used to estimate the weight of the big branches that
A possible explanation for higher prediction when applying the were impossible to be weighed. This causes over-estimation of
model of Brown (1997) and Chave et al. (2005) to the current data the weight for individual trees.
is the deference in wood density and tree architecture. Although,  The number of species for developing the allometric equations
some of Brown’s and Chave’s data were collected in Kalimantan, it may not be enough to represent the gamut of species present at
does not imply that the characteristics of the trees from the study areas. Based on the data from 90 plots of 500 m2,
1692 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694

Wahyuningum (2005) found 124 species in Labanan, whereas in an alternative model since wood density is an important factor for
this study, only 47 species were used. estimating the biomass for mixed species.
 The majority of samples had a diameter of less than 120 cm; Diameter is the only explanatory variable in model 1, it is easy
there were only 3 trees with a diameter of more than 120 cm, to measure and generally available in standard forest inventory.
which were 135, 170 and 200 cm. Ideally, there should be several The inclusion of CBH into DBH as predictor variable does not
trees which have diameters between 135 and 200 cm for improve the performance of model 1.
developing mixed species equation. Based on the application of the proposed model to the
previously published data and the application of the published
equation to the current data, it can be concluded that the
5. Conclusion application of site specific equation must be considered.

Model 1, ln(TAGB) = c + a ln(DBH), is the most suitable Acknowledgement

allometric equation for Dipterocarp forests. The range of 95% CI
of model 1 is closer to the observed values, so it can be applied to We would like to express our appreciation to the anonymous
studies of the carbon balance. This allometric equation can be used reviewers for their constructive comments on the manuscript. This
to improve and complement the GPG, especially for tropical forests work is supported by a grant from the Netherlands Organization
that are dominated by Dipterocarp species. In addition, model 3 is for International Cooperation in Higher Education (NUFFIC).

Appendix A. The wood density of the current study and the published data
No. Species This research GPG-LULUCF** PROSEA**
* 3 * 3
Diameter (cm) WD (g/cm ) Diameter (cm) WD (g/cm ) Diameter (cm) WD* (g/cm3)

1 Dipterocarpus convertus 11.5 0.52

2 D. crinitus 83.0 0.65 0.740–1.070
3 D. crinitus 104.0 0.77
4 D. grandiflorus 18.8 0.64 0.62 0.650–0.945
5 D. grandiflorus 44.0 0.75
6 D. grandiflorus 49.5 0.62
7 D. humeratus 14.0 0.54 0.730–0.800
8 D. humeratus 27.0 0.69
9 D. pacyphyllus 34.0 0.70
10 D. pacyphyllus 67.0 0.73
11 D. pacyphyllus 69.0 0.70
12 D. pacyphyllus 73.0 0.61
13 D. pacyphyllus 93.0 0.74
14 D. pacyphyllus 111.0 0.66
15 D. pacyphyllus 120.0 0.66
16 D. pacyphyllus 135.0 0.73
17 D. palmbanicus 6.5 0.67 0.595–0.785
18 D. palmbanicus 11.8 0.78
19 D. palmbanicus 19.8 0.65
20 D. palmbanicus 23.5 0.70
21 Hopea cernua 6.5 0.65 0.650–0.960
22 H. cernua 10.0 0.57
23 H. cernua 11.8 0.49
24 H. cernua 15.5 0.51
25 H. cernua 17.0 0.67
26 H. cernua 22.0 0.52
27 H. cernua 23.0 0.64
28 H. cernua 28.0 0.55
29 H. cernua 31.5 0.55
30 H. cernua 32.5 0.66
31 H. cernua 34.0 0.54
32 H. cernua 38.5 0.54
33 H. cernua 41.0 0.51
34 H. cernua 45.0 0.77
35 H. cernua 51.0 0.56
36 H. cernua 57.0 0.53
37 H. cernua 59.0 0.58
38 H. cernua 64.0 0.51
39 H.dryobalanoides 62.4 0.69 0.480–0.980
40 H. mengrawan 87.0 0.66 0.510–0.980
41 Palaquium gutta 10.0 0.50 0.610–0.910
42 P. gutta 19.0 0.65
43 P. gutta 24.0 0.65
44 P. gutta 28.6 0.76
45 P. gutta 29.0 0.65
46 P. gutta 30.0 0.61
47 P. gutta 36.0 0.61
48 P. gutta 40.5 0.65
49 P. gutta 48.0 0.59
50 P. gutta 50.0 0.43
51 P. gutta 54.0 0.63
52 P. gutta 55.0 0.53
 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694 1693

Appendix A (Continued )
No. Species This research GPG-LULUCF** PROSEA**
* 3 * 3
Diameter (cm) WD (g/cm ) Diameter (cm) WD (g/cm ) Diameter (cm) WD* (g/cm3)

53 P. gutta 65.0 0.63

54 P. rostratum 6.3 0.56 0.480–0.760
55 P. rostratum 15.8 0.58
56 P. rostratum 25.0 0.62
57 P. rostratum 45.0 0.67
58 Palaquium sp. 13.0 0.44
59 Palaquium sp. 74.0 0.38
60 Shorea agamii 100.0 0.50 0.665
61 S. atrinervosa 50.0 0.60 0.770–1.110
62 S. atrinervosa 100.0 0.71
63 S. macroptera 23.0.0 0.54 0.370–0.770
64 S. macroptera 45.0 0.48
65 S. parvifolia 31.5 0.51 0.290–0.835
66 S. parvifolia 57.0 0.52
67 S. parvifolia 102.0 0.61
68 S. parvifolia 110.0 0.60
69 S. parvistipulata 35.0 0.45
70 S. retusa 26.5 0.50
71 S. retusa 70.0 0.48
72 S. retusa 77.0 0.61
73 S. retusa 84.0 0.47
74 S. retusa 92.0 0.50
75 S. retusa 97.0 0.79
76 S. smithiana 10.2 0.32 0.300–0.720
77 Shorea sp. 6.5 0.55
78 Shorea sp. 12.5 0.39
79 Shorea sp. 38.1 0.50
80 Shorea sp. 43.0 0.54
81 Shorea sp. 82.0 0.38
82 Shorea sp. 200.0 0.57
83 S. superba 170.0 0.86 0.695–1.095
Shorea spp., balau group 0.70
Shorea spp., dark red meranti 0.55
Shorea spp., light red meranti 0.40
Shorea spp., white meranti 0.48
Shorea spp., yellow meranti 0.46
*
Wood density (g/cm3), and for this research it is expressed as oven dry weight (g) per saturated volume (cm3).
**
Published data.

References forests using ground and remote sensing measurements. Agricultural and
Forest Meteorology 136, 22–34.
Clark, D.A., Brown, S., Kicklighter, D.W., Chambers, J.Q., Thomlinson, J.R., Ni, J.,
Araújo, T.M., Higuchi, N., de Carvalho Júnior, J.A., 1999. Comparison of formulae for Holland, E.A., 2001. Net primary production in tropical forests: an evaluation
biomass determination in a tropical rain forest site the state of Pará, Brazil. and synthesis of existing field data. Ecological Application 11 (2), 371–384.
Forest Ecology and Management 177, 43–52. Cordero, L.D., Kanninen, M., 2002. Wood specific gravity and above ground biomass
Baker, T.R., Phillips, O.L., Malhi, Y., Almeida, S., Arroyo, L., Di Fiore, A., Killeen, T.J., of Bombacopsis quinata plantation in Costa Rica. Forest Ecology and Manage-
Laurance, S.G., Laurance, W.F., Lewis, S.L., Lloyd, J., Monteagudo, A., Neill, D.A., ment 165, 1–9.
Patiño, S., Pitman, N.C.A., Silva, N., Martı́nez, R.V., 2004. Variation in wood De Gier, A., 2003. In: Roy, P. (Ed.), A New Approach to Woody Biomass Assessment
density determines spatial patterns in Amazonian forest biomass. Global in Woodlands and Shrublands. Geoinformatics for Tropical Ecosystems, India,
Change Biology 10, 545–562. pp. 161–198.
Berau Forest Management Project (BFMP), 1999. The climatic and hydrology of FAO, 2004. National forest inventory: Field manual template, Rome. http://
Labanan concession. Ministry of Forestry, Jakarta, 27 pp. www.fao.org/docrep/008/ae578e/ae578e00.htm (accessed 02.01.07).
Brand, G.J., Smith, W.B., 1985. Evaluating allometric shrub biomass equations fit to Gibbs, H.K., Brown, S., Niles, J.O., Foley, J.A., 2007. Monitoring and estimating
generated data. Canadian Journal of Botany 63, 64–67. tropical forest carbon stocks: making REDD a reality. Environmental Research
Brown, S., 1997. Estimating biomass and biomass change of tropical forests: a Letters 2, 13, doi:10.1088/1748-9326/4/045023.
primer. FAO. Forestry Paper 134, Rome, 87 pp. Gunawan, A., Rathert, G., 1999. Monitoring, Data Management and Analysis of the
Burnham, K.P., Anderson, D.R., 2002. Model selection and multimodel inference. A BFMP Permanent Sample Plots (STREK plots) at Berau. Berau Forest Manage-
practical information-theoretic approach, 2nd ed. Springer Science + Business ment Project. European Union – Ministry of Forestry and Estate Crops, Jakarta,
Media, Inc., New York, 488 pp. 51 pp.
Cairns, M.A., Olmsted, I., Granados, J., Argaez, J., 2003. Composition and above- Hashimotio, T., Kojima, K., Tange, T., Sasaki, S., 2000. Changes in carbon storage in
ground tree biomass of a dry semi-evergreen forest on Mexico’s Yucatan fallow forests in the tropical lowlands of Borneo. Forest Ecology and Manage-
Peninsula. Forest Ecology and Management 186, 125–132. ment 126, 331–337.
Chambers, J.Q., dos Santos, J., Ribeiro, R.J., Higuchi, N., 2001. Tree damage, allometric Intergovernmental Panel on Climate Change (IPCC), 2003. In: Penman, J., Gystarsky,
relationship, and above-ground net primary production in central Amazon M., Hiraishi, T., Krug, T., Kruger, D., Pipatti, R., Buendia, L., Miwa, K., Ngara, T.,
forest. Forest Ecology and Management 152, 73–84. Tanabe, K., Wagner, F. (Eds.), Good Practice Guidance for Land Use, Land-Use
Chave, J., Riéra, B., Dubois, M., 2001. Estimation of biomass in a Neotropical Forest of Change and Forestry (GPG-LULUCF). IPCC National Greenhouse Gas Inventories
French Guiana: spatial and temporal variability. Journal of Tropical Ecology 17, Programme.
79–96. Keller, M., Palace, M., Hurtt, G., 2001. Biomass estimation in the Tapajos National
Chave, J., Andalo, A., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., Fölster, H., Forest, Brazil examination of sampling and allometric uncertainties. Forest
Fromard, F., Higuchi, N., Kira, T., Lescure, J.-P., Nelson, B.W., Ogawa, H., Puig, H., Ecology and Management 154, 371–382.
Riéra, B., Yamakura, T., 2005. Tree allometry and improved estimation of carbon Ketterings, Q.M., Coe, R., van Noordwijk, M., Ambagau, Y., Palm, C.A., 2001. Reducing
stocks and balance in tropical forests. Oceologia 145, 87–99. uncertainty in the use of allometric biomass equations for predicting above-
Chave, J., Muller-Landau, H.C., Baker, T.R., Easdale, T.A., Steege, H.T., Webb, C.O., ground tree biomass in mixed secondary forests. Forest Ecology and Manage-
2006. Regional and phylogenetic variation of wood density across 2,456 Neo- ment 146, 199–209.
tropical tree species. Ecological Applications 16, 56–2367. Mantel, S., 1999. Development of environmental framework: Soils and terrain
Chiesi, M., Maselli, F., Bindi, M., Fibbi, L., Cherubini, P., Arlotta, E., Tirone, G., conditions of Labanan. Summary Report. Berau Forest Management Project.
Matteucci, G., Seufert, G., 2005. Modelling carbon budget of Mediterranean European Union – Ministry of Forestry and Estate Crops, Jakarta, 14 pp.
1694 T.M. Basuki et al. / Forest Ecology and Management 257 (2009) 1684–1694

Myeong, S., Nowak, D.J., Duggin, M.J., 2006. A temporal analysis of urban forest carbon Sprugel, D.G., 1983. Correcting for bias in log-transformed allometric equations.
storage using remote sensing. Remote Sensing of Environment 101, 277–282. Ecology 64 (1), 209–210.
Nelson, B.W., Mesquita, R., Pereira, J.L.G., de Souza, S.G.A., Batista, G.T., Couta, L.B., Stewart, J.L., Dunsdon, A.J., Hellin, J.J., Hughes, C.E., 1992. Wood biomass estimation
1999. Allometric regressions for improved estimate of secondary forest biomass of Central American dry zone species. Tropical Forestry Papers 26. Oxford
in the Central Amazon. Forest Ecology and Management 117, 149–167. Forestry Institute, Department of Plant Sciences, University of Oxford.
Nogueira, E.M., Nelson, B.W., Fearnside, P.M., 2005. Wood density in dense forest in Tan, K., Piao, S., Peng, C., Fang, J., 2007. Sattellite-based estimation of biomass carbon
central Amazonia, Brazil. Forest Ecology and Management 208, 261–286. stocks for northeast China’s forests between 1982 and 1999. Forest Ecology and
Nogueira, E.M., Fearnside, P.M., Nelson, B.W., França, M.B., 2007. Wood density in Management 240, 114–121.
forests of Brazil’s ‘arc of deforestation’: Implications for biomass and flux of Tyrie, G., 1999. Ten years of tropical lowland rainforest research in Labanan, East
carbon from land-use change in Amazonia. Forest Ecology and Management Kalimantan the STREK plots. Berau Forest Management Project. European Union
248, 119–135. – Ministry of Forestry and Estate Crops, Jakarta, 7 pp.
Sah, J.P., Ross, M.S., Kaptur, S., Snyder, J.R., 2004. Estimating aboveground biomass of UNFCCC (United Nations Framework Convention on Climate Change), 2008. Report
broadleaved woody plants in the understory of Florida Keys Pine forests. Forest of the Conference of the Parties on its thirteenth session, held in Bali from 3 to
Ecology and Management 203, 319–329. 15 December 2007. Addendum, Part 2. Document FCCC/CP/2007/6/Add.1.
Samalca, I., 2007. Estimation of forest biomass and its error: A case in Kalimantan, UNFCCC, Bonn, Germany.
Indonesia. MSc thesis, ITC, Enschede, 74 pp. Wahyuningum, N., 2005. Foliage biomass estimation in tropical logged over forest
Soerianegara, I., Lemmens, R.H.M.J. (Eds.), 1993. Plant Resources of South-East Asia East Kalimantan, Indonesia. MSc Thesis, ITC, Enschede, 54 pp.
(PROSEA). Timber trees: Major Commercial Timbers, vol. 5 (1). Pudoc Scientific Wang, H., Hall, C.A.S., Scatena, F.N., Fetcher, N., Wu, W., 2003. Modeling the
Publisher, Wageningen, 610 pp. Spatial and temporal variability in climate and primary productivity across
Sokal, R.R., Rohlf, F.J., 1995. Biometry. The Principles and Practice of Statistic in the Luquillo mountains, Puerto Rico. Forest Ecology and Management 179,
Biological Research, 3rd ed. W.H. Freeman and Co., New York, 859 pp. 69–94.
Son, Y., Hwang, J.W., Kim, Z.S., Lee, W.K., Kim, J.S., 2001. Allometry and biomass of Yamakura, T., Hagihara, A., Sukardjo, S., Ogawa, H., 1986. Aboveground biomass
Korean pine (Pinus koraiensis) in Central Korea. Bioresource Technology 78, of tropical rainforest stands in Indonesian Borneo. Plant Ecology 68 (2),
251–255. 71–82.