Forest Ecology and Management 334 (2014) 241–253

Contents lists available at ScienceDirect

Forest Ecology and Management

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

Tree biomass equations for tropical peat swamp forest ecosystems

in Indonesia
Solichin Manuri a,⇑, Cris Brack a, Nunung Puji Nugroho b, Kristell Hergoualc’h c, Nisa Novita d,
Helmut Dotzauer e, Louis Verchot d, Chandra Agung Septiadi Putra f, Eka Widyasari g
a
Fenner School of Environment and Society, The Australian National University, Linnaeus way Building 48, Acton, Canberra 0200, Australia
b
Research Institute for Forestry Technology on Watershed Management, Jl. Jenderal Ahmad Yani, P.O. Box 295, Pabelan, Kartasura, Surakarta 57102, Indonesia
c
Center for International Forestry Research (CIFOR) c/o Centro International de la Papa (CIP), Av. La Molina 1895, La Molina, Apdo postal 1558, 15024 Lima, Peru
d
Center for International Forestry Research (CIFOR), Jl. CIFOR, Situ Gede, Bogor 16115, Indonesia
e
Forests and Climate Change Programe, Deutsche Gesellschaft für Internationale Zusammenarbeit (FORCLIME-GIZ), Manggala Wanabakti Building, Block VII, 6th floor,
Jakarta, Indonesia
f
Tropical Forest Foundation (TFF), Komplek Cimanggu Permai, Jl. Tumapel Block O, IV No 17, Bogor 16164, Indonesia
g
Faculty of Forestry, Bogor Agricultural University, Dramaga Bogor, Indonesia

a r t i c l e i n f o a b s t r a c t

Article history: To assist countries to reduce emissions from deforestation and forest degradation, the United Nations has
Received 20 June 2014 introduced the REDD+ mechanism. This performance-based incentive mechanism requires accurate
Received in revised form 22 August 2014 quantification of carbon stock and emissions. However, currently there are limited existing local or regio-
Accepted 27 August 2014
nal equations for estimating aboveground biomass in peat swamp forests. The main objective of this
Available online 30 September 2014
study was to define the most accurate models for aboveground biomass estimation in Indonesian peat
swamp forests. We found that the pan-tropical equations performed better in estimating biomass of peat
Keywords:
swamp forests than did existing local equations. We developed new equations, based on 148 trees from
Aboveground biomass estimation
Destructive sampling
24 families with diameter at breast height in the range of 2–167 cm collected from peat swamp forests in
Tree allometry the western part of Indonesia. Statistical indicators showed that the best model form was the common
Wood density linear one using log-transformed data. Estimated biomass values needs to be back-transformed applying
Species grouping correction factors. The ratio estimator correction factor which derives from the ratio between the average
REDD+ of measured biomass and the average of predicted biomass, was found to provide the lowest mean devi-
ation. The existing pan-tropical equations performed similarly to our mixed species and dipterocarp
models but they systematically under- or over-estimated the biomass of certain species groups, espe-
cially non-dipterocarp trees. We also found that grouping by family (dipterocarp vs. non-dipterocarp)
and wood density class (hardwood vs. softwood) significantly improved the accuracy of biomass estima-
tion. In the absence of wood density values, wood density-class specific equations, instead of mixed-spe-
cies equations improved the accuracy of biomass estimates.
! 2014 Elsevier B.V. All rights reserved.

1. Introduction
Greenhouse gas emissions from land use, land-use change and
forestry (LULUCF) account for about 12.5% of global emissions
(Baccini et al., 2012; Houghton et al., 2012). In Indonesia, this sector
represents 47% of national emissions and emissions from peat fires
an additional 13% (Ministry of Environment, 2010). Peat swamp for-
ests (PSF) in particular suffered intensive deforestation in Indonesia
between 1990 and 2010 with around 4.6 million hectares being
cleared out of the 1990 8.7 million hectares (Miettinen et al.,
⇑ Corresponding author. Tel.: +61 2 6125 4588; fax: +61 2 6125 0651.
E-mail addresses: solichin.solichin@anu.edu.au, solichin.manuri@gmail.com
(S. Manuri).
2011). In 2011, only 31% of Indonesian PSF were still in pristine con-
dition (Ministry of Forestry, 2012). Peat swamp forest are carbon-
rich ecosystems (Jaenicke et al., 2008) which store on average
220 Mg C ha!1 in the phytomass (Hergoualc’h and Verchot, 2011)
and 670 Mg C ha!1 per meter depth of peat (Warren et al., 2012).
Deforestation of PSF and drainage for cultivation of agricultural
crops involve major shifts in the carbon and nitrogen cycles leading
to substantial greenhouse gas emissions; especially when land-
clearing fires are used (Hergoualc’h and Verchot, 2014). Peat
swamp forests could hence play a potentially important role in
climate change mitigation strategies involving forest conservation.
In 2005, a mechanism for reducing emissions from deforestation
and forest degradation (REDD) was discussed at the 11th annual
Conference of the Parties from the United Nations Framework

http://dx.doi.org/10.1016/j.foreco.2014.08.031
0378-1127/! 2014 Elsevier B.V. All rights reserved.
242 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253
Convention on Climate Change (UNFCCC-COP11). Subsequently at
COP13 in Bali, Indonesia, the parties agreed to emphasize the need
to include carbon stock enhancement, conservation and sustainable
management of forests into the mechanism and it became known as
REDD+. Under REDD + tropical developing countries will be incen-
tivized to decrease the emission rate successfully with respect to
that of the reference emission level. As the performance will be
based on quantified emission reduction, accurate biomass and car-
bon estimation is essential. Inaccurate biomass estimation at tree
and plot level tends to multiply uncertainties at the landscape level
and may result in substantial error (Houghton et al., 2012; Pelletier
et al., 2012). Appropriate allometric models that are precise and
unbiased for estimating forest biomass at tree level are required in
conjunction with remote sensing and field observation to quantify
landscape emission and to understand the role of tropical forests
in climate change.
Current aboveground biomass (AGB) assessments of tropical for-
ests commonly rely on general allometries, instead of specific equa-
tions (Jubanski et al., 2012; Kronseder et al., 2012). Bias, a
systematic over- or under-estimate of the true value, can be severe
when using allometric models from other regions or based on
unsuitable diameter at breast height (DBH) or wood density (WD)
ranges (van Breugel et al., 2011; Alvarez et al., 2012). Biomass esti-
mation using pan-tropical equations or models from another region
displayed a higher bias when compared to estimates derived from
locally-developed models, for instance in Columbia (Alvarez et al.,
2012), Kalimantan (Basuki et al., 2009), Sarawak (Kenzo et al.,
2009a), north-east Gabon (Ngomanda et al., 2014) and Peru
(Goodman et al., 2014). However, the pan-tropical models devel-
oped by Chave et al. (2005) were found to perform equally as local
models in the tropical forests of central Africa (Fayolle et al., 2013),
Madagascar (Vieilledent et al., 2012) and in the lowland dipterocarp
forests in Indonesia (Rutishauser et al., 2013).
In order to avoid the bias introduced by the use of general mod-
els research on forest specific allometric models has been con-
ducted in the past years for local and regional AGB estimation in
tropical forests (Cole and Ewel, 2006; Djomo et al., 2010). In South-
east Asia the focus has been not only on certain forest types,
including lowland dipterocarp forest (Yamakura et al., 1986;
Basuki et al., 2009) and mangrove forest (Komiyama et al., 2002),
but also on various land use and forest degradation levels, such
as agroforestry land (Ketterings et al., 2001), secondary forests
(Hashimoto et al., 2004; Kenzo et al., 2009b) and logged-over for-
ests (Kenzo et al., 2009a). Although some allometries specific to
PSF have been produced (see compilation by Krisnawati et al.,
2012), these were either developed from relatively small sample
sizes and restricted data ranges or based on other forest types
and regions.
The main objective of this study was hence to define which
models were the most accurate for ABG estimation in Indonesian
PSF. For this we used a large dataset of 148 trees, destructively
sampled in PSF of Sumatra and Kalimantan. Specific goals
included: (1) to evaluate the performance of existing pan-tropical
and local equations; (2) to develop new mix-species allometries
for Indonesian PSF testing different equation forms (linear, polyno-
mial and power) with single or multiple independent variables
(DBH, WD and height H) and evaluate their performance; and (3)
to identify effective grouping by family (dipterocarp vs. non dip-
terocarp), WD class (soft vs. hard wood) and island (Sumatra vs.
Kalimantan) for allometric equation development.
2. Materials and methods
The study was based on a relatively large dataset of 148 trees,
destructively sampled during a series of research projects. The
dataset included trees within a large range of diameters, heights
and wood densities (Table 1). The samples consisted of 24 families
collected from three sites of peat swamp forest in the Sumatra and
Kalimantan islands in western Indonesia.
2.1. Study sites
The data was compiled from three different PSF sites: Rokan
Hilir in Riau in Sumatra; Musi Banyuasin in South Sumatra; and
Kapuas Hulu in West Kalimantan, Indonesia (Fig. 1). The mean
annual rainfall in Rokan Hilir, Musi Banyuasin and Kapuas Hulu
is 2637 mm, 2454 mm and 4100 mm, respectively. The peat depths
were up to 5, 8.5 and 6.5 m in Rokan Hilir, Musi Banyuasin and
Kapuas Hulu, respectively. The forests at all sites had been previ-
ously logged by timber companies or by illegal loggers. The Rokan
Hilir study site was managed by a timber company (PT DRT) and
was selectively logged, while the two other sites were illegally
and moderately logged.
In Rokan Hilir the most frequent species were Eugenia sp.,
Palaquium obovatum, Ilex macrophylla, Horsfieldia glabra, and Sho-
rea uliginosa. In Musi Banyuasin, the most dominant tree species
were Melanorrhoea walicchii, Tetramerista glabra, Gonystylus banc-
anus, Hydnocarpus woodii, Palaquium sp. and Dyera lowii, which
represented 11%, 9%, 7%, 6%, 4% and 3% of the total trees measured
during the previous forest inventory. In Kapuas Hulu, the dominant
species were Melanorrheoa sp., Shorea sp., Callophyllum sp., Eugenia
sp. and Diallium patens.
2.2. Destructive sampling
We carried out the destructive sampling from August 2008 to
October 2012. A total of 148 trees with a DBH of between 2 cm
and 167 cm were felled and measured. Tree selection was based
on the dominance and abundance of tree species, derived from per-
manent sample plot data in Rokan Hilir or previous forest inven-
tory plots in Musi Banyuasin and Kapuas Hulu. Further tree
selection within the plots was applied based on diameter class rep-
resentation. We prioritised the selection of large tree before select-
ing small trees, as trees smaller than 20 cm in DBH were abundant.
We selected the trees which would most likely would get impact
from the felling of large trees. As such, we reduced further damage
to the remaining stand and avoided bias of selecting only certain
quality of trees.
Prior to felling, we measured the diameter at breast height
(DBH) or, whenever there was a buttress, 20 cm above it. We used
machetes or chainsaws (60 cm and 90 cm bar length) to cut down
the trees. Tree heights were measured after each tree was felled
using 50 m cloth tape. We collected specimens of unidentified
trees for species identification at the national herbarium (Lembaga
Ilmu Pengetahuan Indonesia-LIPI) or forestry research center (Pusat
Penelitian dan Pengembangan Kehutanan-Puslitbang) in Bogor.
Trees were divided into stems, branches, twigs and leaves as
described by Ketterings et al. (2001). Stems of small trees or
branches (mostly of diameter < 30 cm) were cut into pieces to
allow weighing using spring or digital scales with capacities of
50 kg, 100 kg or 200 kg. Commercial stems of large trees or very
large branches (diameter P 30 cm) were measured for their green
volume at 2 m intervals and the volume was calculated as a
frustum of a cone (except for Rokan Hilir and some of the Musi
Banyuasin dataset, in which we weighed them all).
2.3. Laboratory work
Sub-samples of wood were extracted from at least two sections
of each stem (at the base and the end of the sections) branches,
twigs and leaves for dry mass and/or green volume determination.
 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253 243

Table 1
Dimensions of trees sampled for this study by family. Dash means no data has been collected.

Family/species N Min DBH Max DBH Min Height Max Height Min WD Max WD Mean WD Max buttresses
(cm) (cm) (m) (m) (g cm!3) (g cm!3) (g cm!3) height (m)
Anacardiaceae 8 9.6 160.0 10.3 43.1 0.360 0.477 0.386 0
Annonaceae 6 13.0 46.8 18.3 29.7 0.471 0.690 0.610 0
Apocynaceae 3 7.5 15.5 7.9 14.3 0.320 0.342 0.331 0
Aquifoliaceae 6 6.8 54.4 8.7 27.9 0.560 0.560 0.560 -
Burseraceae 2 5.0 16.0 4.9 14.3 0.490 0.547 0.519 0
Chrysobalanaceae 1 2.0 2.0 3.8 3.8 0.803 0.803 0.803 0
Clusiaceae 6 6.0 31.2 6.9 23.8 0.594 0.760 0.711 0
Dipterocarpaceae 27 3.5 117.0 4.2 49.5 0.328 0.745 0.541 5.0
- Cotylelobium lanceolatum 2 27.0 34.0 27.7 28.5 0.498 0.594 0
- Dipterocarpus borneensis 1 20.0 20.0 20.7 20.7 0.620 0.620 0
- Dryobalanops rappa 4 40.0 72.5 29.0 40.7 0.570 0.639 1.9
- Shorea balangeran 1 117.0 117.0 47.8 47.8 0.704 0.704 5.0
- Shorea dasyphylla 5 4.0 102.9 6.3 49.5 0.328 0.553 4.5
- Shorea sp. 1 92.0 92.0 42.1 42.1 0.745 0.745 -
- Shorea teysmaniana 2 9.0 66.0 9.8 37.2 0.470 0.515 0.5
- Shorea uliginosa 11 3.5 84.0 4.2 43.4 0.350 0.580 3.2
Elaeocarpaceae 2 3.0 13.6 4.0 6.6 0.453 0.530 0.492 0
Euphorbiaceae 11 3.1 35.2 4.3 21.1 0.242 0.717 0.363 0
Fabaceae 3 2.8 46.5 3.2 31.1 0.320 0.651 0.502 0
Fagaceae 3 9.5 20.7 11.7 22.6 0.670 0.728 0.707 1.7
Icacinaceae 6 9.0 42.0 9.3 29.5 0.503 0.760 0.632 0
Lauraceae 6 4.5 29.5 4.8 22.7 0.338 0.618 0.472 0
Magnoliaceae 2 37.6 46.4 20.3 33.3 0.440 0.499 0.470 0
Moraceae 1 30.2 30.2 19.1 19.1 0.501 0.501 0.501 0
Myristicaceae 7 7.1 33.0 7.5 23.9 0.408 0.460 0.453 0
Myrtaceae 17 2.4 33.2 2.8 19.8 0.305 1.000 0.618 0
Phyllanthaceae 1 11.7 11.7 22.4 22.4 0.358 0.358 0.358 0
Rhizophoraceae 1 8.3 8.3 10.0 10.0 0.620 0.620 0.620 -
Rutaceae 1 5.7 5.7 4.4 4.4 0.463 0.463 0.463 0
Sapotaceae 10 6.5 51.5 7.8 29.3 0.435 0.940 0.626 0
Theaceae 6 14 167.0 12.6 36.2 0.436 0.910 0.590 0
Thymelaeaceae 6 5.2 62.7 8.1 39.4 0.530 0.530 0.530 -
All data 148 2.0 167.0 2.8 49.5 0.242 1.000 0.539 5.0

The wood sub-samples collected were either disk-shaped or
wedge-shaped with a 3–5 cm thickness. These sub-samples were
weighed in the field to measure fresh mass, and were then packed,
labelled and transported to a laboratory for dry mass and wood
density (WD) analysis. We dried the wood and leaf samples with
temperature of 105 "C for Rokan Hilir and Musi Banyuasin samples
and 80 "C for Kapuas Hulu samples, respectively. All samples were
dried until constant mass. The green volume of the samples were
measured using water displacement method.
The wood dry mass and volume measurements were carried out
at several laboratories, including the Sebelas Maret University for
the Rokan Hilir dataset; the Forestry Research Center in Bogor
and the CIFOR laboratory in Jambi for the Musi Banyuasin dataset;
and the Tanjung Pura University for the Kapuas Hulu dataset. We
calculated the ratio of dry mass and fresh mass of each subsample
(Rdm/fm). We estimated wood densities by dividing the dry mass by
the green volume of the wood samples. We calculated dry mass
(DM) of each tree components, using the following formulae:
DM ¼ FM # Rdm=fm ð1Þ
DM ¼ gVol # WD ð2Þ

Fig. 1. Maps of study area, yellow dots indicate the research sites.
244 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253

where FM is the fresh mass and gVol is the green volume measured
in the field. DM of each tree component was calculated using the
Rdm/fm or WD value from the same section of each component. We
then summed all component mass to calculate total AGB of the tree.
For determination of individual WD, we calculated the WD by
averaging the WD values taken from the base and the end of the
stems, and additionally from the middle part of the stem and the
large branches of large trees. We did not measure the WD from
the Rokan Hilir dataset, but we used instead the global WD dataset
from Zanne et al. (2009).
2.4. Assessment of existing equations
We compared the performance of available pan-tropical and
local equations, commonly used for estimating AGB in Indonesian
forests (Table 2) through correlation of estimated data with mea-
sured biomass data. For this analysis, we excluded the biomass
data from the two largest trees (160 cm and 167 cm of DBH) due
to their unexpected / abnormal low tree height and biomass value.
We further selected the best existing models in each model type to
compare with our new equations.
2.5. Selection of model forms for new allometric equations
We selected several candidate model forms to fit our data,
including common linear, polynomial and power forms (see
Table 4). We categorized the models into four types, depending
on the independent variables used, namely DBH, DBH–WD, DBH–
H and DBH–WD–H types. As such, we provided a wide range of
possibilities to estimate aboveground biomass in relation to the
availability of measured parameters.
2.6. Species, wood density and island groupings
We developed species-group models for estimating above-
ground biomass. We evaluated the influence of the island in esti-
mating AGB, as the datasets were collected from the two main
islands in western Indonesia, namely Sumatra and Kalimantan.
We generated dipterocarp and non-dipterocarp models. Our data-
set included eight species from four genera (Shorea, Dryobalanops,
Dipterocarpus and Cotylelobium) in the dipterocarp family.
Within the non-dipterocarp species, we further categorized the
trees into two classes based on their mean WD, that is light/soft
and hard wood. The WD values of non-dipterocarp trees ranged
from 0.242 to 1.000 g cm!3, with an average of 0.543 g cm!3. We
used the average WD as cut-off value between soft and hardwood.
We assessed the influence of WD class in estimating aboveground
biomass by generating soft and hard wood models for non-diptero-
carp species.
2.7. Statistical analysis
All statistical analyses were performed using JMP software. The
accuracy of existing models was evaluated by regressing the bio-
mass values of measured dataset and the values predicted by the
models. In a perfectly accurate relationship, the intercept would
be 0 and the slope equal to 1. An unbiased but imprecise relation-
ship would still go through 0 with a slope of 1 but observations

Table 2
Selected existing AGB models developed for pan-tropical and regional data (with correction factors), used for comparison. AGB is tree aboveground biomass (kg). DBH is diameter
at breast height (cm). WD is wood density (g cm!3). H is total tree height (m). ERH, EMB and EKH are environmental variables for Rokan Hilir, Musi Banyuasin and Kapuas Hulu,
respectively. Bs, Bb and Bl are stem biomass, branch biomass and leaf biomass, respectively.

Model Types Model Name Specification Sample data range AGB Equations
DBH Brown.DBH Mixed species, pan-tropical n = 371 exp (!2.289 + 2.649 & ln(DBH) –
(Brown et al., 2011) forests, including lowland DBH range: 5–148 cm 0.021 & (ln(DBH))2)
dipterocarp forests Asia and
Latin America
Basuki1.DBH Mixed species, lowland n = 122; exp (!1.145 + 2.196 & ln(DBH))
(Basuki et al., 2009) dipterocarp forests, Kalimantan DBH range: 6–200 cm
Basuki2.DBH Commercial species, dominated n = 83 exp (!0.698 + 2.234 & Ln (DBH))
(Basuki et al., 2009) by dipterocarp trees, lowland DBH range: 6–200 cm
dipterocarp forest, Kalimantan
Kenzo.DBH Mixed species, lowland n = 30 0.1525DBH2.34
(Kenzo et al., 2009a) secondary logged-over forests, DBH range: 2–44 cm
Sarawak
DBH–WD Chave.DBH–WD–E Mixed species, pan-tropical n = 4004 AGB = 0.0673 & (DBH2 & WD & H)0.976
(Chave et al., 2014) forests, Africa, America and Asia DBH range: 5–156 cm
H0 = exp(0.893 – Ei + 0.76 & ln(DBH) –
0.034 & (ln(DBH))2
ERH = !0.1033
EMB = !0.1199
EKH = !0.0599
Basuki1.DBH–WD Mixed species, lowland n = 122; exp
(Basuki et al., 2009) dipterocarp forests, Kalimantan DBH range: 6–200 cm (!0.698 + 2.188 & ln(DBH) + 0.832 & ln(WD)
Basuki2.DBH–WD Commercial species, dominated n = 83 exp
(Basuki et al., 2009) by dipterocarp trees, lowland DBH range: 6–200 cm (!0.989 + 2.203 & ln(DBH) + 0.639 & ln(WD)
dipterocarp forest, Kalimantan
Ketterings.DBH–WD Mixed species, agroforestry land, n = 29 0.11 WD & DBH2.62
(Ketterings et al., 2001) Sumatra
DBH range: 5–50 cm
H up to 34 m
DBH–H Yamakura.DBH–H. Mixed species, lowland n = 145 Bs = (0.02909 & (DBH2 & H)0.9813
(Yamakura et al., 1986) dipterocarp forests, Kalimantan
DBH range: 4.5–130 cm. H Bb = 0.1192 & (Bs)1.059
up to 70.7 m
Bl = 0.09146 & (Bs + Bb)0.7266
AGB = Bs + Bb + Bl
DBH–WD–H Chave.DBH–WD–H Mixed species, pan-tropical n = 4004 0.0673 & (DBH2 & WD & H)0.976
(Chave et al., 2014) forests, Africa, America and Asia DBH range: 5–156 cm
 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253 245

Fig. 2. Mean relative difference in% between WD global dataset of Zanne et al. (2009) and measured WD (n = 93). Error bars represent the standard deviation.

Fig. 3. Correlation between measured biomass (AGBm) in trees (n = 146) of peat swamp forests of Sumatra and Kalimantan and predicted biomass (AGBp) by allometric
equations from the literature. The circle marker is representing tree samples, while the cross marker represents tree outliers. The upper figures are the best existing equations
from each model type.
246 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253

Table 3 Scatter plots of independent (tree DBH and H) and dependent

Test effects for linear AGB models performed using selected model types. Each model (tree AGB) variables demonstrated heteroscedasticity of the data.
evaluates the relevancy of family (dipterocarp and non-dipterocarp), wood density
(WD) classes (hard and soft wood) and island (Sumatra and Kalimantan) grouping.
We transformed the data using a natural logarithm to distribute
Significance is indicated by ⁄⁄⁄ for P < 0.001, ⁄⁄ for P < 0.01 and ⁄ for P < 0.05. the variance evenly across biomass values. In order to transform
the data back to a biomass value, a correction factor to reduce
AGB model types Effects Prob > F
the systematic bias associated with log-transformations of biomass
DBH WD class <.001⁄⁄⁄ was applied (Baskerville, 1972). Two alternative correction factors
Family groups 0.005⁄⁄
Island 0.651
were used for the analysis: Correction Factor (CF) (Sprugel, 1983)
and Ratio Estimator (REst) (Snowdon, 1991), where RSE is residual
DBH–WD WD class 0.289
Family groups 0.002⁄⁄
standard error, yi and ŷi are observed and predicted biomass of
Island 0.662 treei, and n is number of tree samples:
DBH–H WD Class <.001⁄⁄⁄
Family groups 0.013⁄ RSE2
CF ¼ exp ð3Þ
Island 0.207 2
DBH–WD–H WD class 0.342
P
Family groups 0.009⁄⁄ y =n
Island 0.302 REst ¼ P i ð4Þ
y
^i =n
We performed multicollinearity test to evaluate correlation
between parameters. The multicollinearity was identified using
would be scattered about the line, while a biased relationship variance inflations factor (VIF). High VIF or more than 10 indicates
would have an intercept and/or slope significantly different from that the parameter closely related with other parameter. VIF of a
0 and 1 respectively. The best existing model with the highest cor- parameter was defined as in Eq. (5), where R2i is a coefficient of var-
relation (R2), intercept value closest to 0 and slope value closest to iation from the model regressing the ith parameter on the other
1 will be chosen from each model type. parameter.

Fig. 4. Plot residuals of mixed species models (upper figures) and species grouped models based on WD Class (lower figures). All species grouped models have normal
distribution of plot residuals (slope of the fit lines close to 0).

Table 4
Mixed species model description for the aboveground biomass estimation in peat swamp forests. df is degree of freedom. RSE is residual standard error. AICc is corrected Akaike
Information Criterion. REst is ratio estimator. CF is correction factor. Dev REst is mean deviation of AGB transformed back using Rest. Dev CF is mean deviation of AGB transformed
back using CF.

Model name Model parameters df RSE R2 AICc REst CF Dev REst Dev CF
a b c d e
MDBH ln (AGB) = a + b & ln (DBH)
!1.974 2.475 147 0.400 0.968 152 0.826 1.083 31.9 37.4
MDBH–polynomial ln (AGB) = a + b & ln (DBH) + c & ln (DBH))2
!2.520 2.893 !0.072 147 0.395 0.969 150 0.949 1.081 32.3 37.0
MDBH–WD ln (AGB) = a + b & ln (DBH) + c & ln (WD)
!1.383 2.441 0.746 147 0.345 0.977 110 0.842 1.061 28.1 31.0
MDBH–WD Polynomial ln (AGB) = a + b & ln (DBH) + c & (ln (DBH))2 + d & (Ln (DBH))3 + e & ln (WD)
!1.178 1.866 0.311 !0.045 0.739 147 0.331 0.979 100 1.038 1.056 29.1 29.7
MDBH–H ln (AGB) = a + b & ln (DBH) + c & ln (H)
!2.586 1.965 0.761 147 0.358 0.975 121 0.952 1.066 28.5 32.3
MDBH power H ln (AGB) = a + b & ln (DBH2H)
!2.705 0.926 147 0.359 0.974 121 0.982 1.067 29.2 32.4
MDBH–WD–H ln (AGB) = a + b & ln (DBH) + c & Ln (WD) + d & Ln (H)
!1.964 2.023 0.652 0.630 147 0.313 0.981 83 0.937 1.050 24.3 27.2
MDBH power–WD–H ln (AGB) = a + b & ln (DBH2 & WD & H)
0.152 0.904 147 0.328 0.979 94 0.531 1.144 26.1 28.3
 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253 247

1 were not available in the database. However, we found that the

VIF i ¼
1 ! R2i
The effect of dipterocarp/non-dipterocarp family, wood density
class and island grouping on the models was evaluated by linear
regression analysis. We assessed the significance of each factor
using the least squares fitting.
We chose the best form of the mixed-species and species
grouped models based on the highest coefficient of determination
(R2) and lowest residual standard error (RSE), the corrected Akaike
Information Criterion (AICc), mean relative error (mean RE) and
mean deviation (mean Dev). Compared to the original Akaike Infor-
mation Criterion (AIC), the AICc has less bias and is more suited to
model selection (Hurvich and Tsai, 1991). AICc is defined as
follows:
2kðk þ 1Þ
AIC c ¼ !2 log likelihood þ 2k þ
ðn ! k ! 1Þ
where k is the number of estimated parameters, including intercept
and error terms in the model, and n is the number of observations in
the dataset.
To calculate mean RE (%) and mean Dev (%) of the AGB estimates,
we used the Eqs. (7) and (8), where AGBp and AGBm are predicted
and measured AGB, respectively.
100 X AGBp ! AGBm
mean RE ¼
n AGBm
! !
100 X !!AGBp ! AGBm !!
mean Dev ¼ ! !
n AGBm
Finally, we plotted the relative error (%) of each AGB estimates
against DBH distribution to assess and compare the performance of
the best existing equations with our equations. We generated the
kernel smoother lines using the spline method (with lambda
100,000).
3. Results
3.1. Wood density
Similarly to Fayolle et al. (2013), we found almost 70% of total
tree species in the WD global database. A total of 52 out of 58 tree
species had been identified at species level, and only 14 species
ð5Þ
correlation between the measured dataset and the global database
was very low (R2 = 0.169; n = 93). The WD values of small trees
(DBH < 50 cm) in the global database tended to be higher than
the measured ones (Fig. 2).
3.2. Selection of the best existing equations
Most of the estimation using local equations with only DBH as
predictor showed good correlation but tended to underestimate
the measured AGB of peat swamp trees, with slope far less than
1 (Fig. 3b–d). Only Brown.DBH equation had a slope of more than
0.9 in this model type (Fig. 3a). Although WD was added as an
additional parameter, the local equations Basuki1.DBH-WD and Basu-
ki2.DBH-WD did not significantly improve the Basuki1DBH and Basu-
ki2.DBH models, respectively. The other local equation developed
ð6Þ from a lowland dipterocarp forest (Yamakura.DBH-H), on the other
hand, had good performance, in spite of its complexity (Fig. 3i).
Both new pan-tropical equations (Chave.DBH–WD–E and Chave.DBH–
WD–H) performed well with slopes around 1 (Fig 3e and j). However,
when we included the two largest trees (represented by cross mar-
ker in Fig 3) in the correlation analysis, all Chave’s and Brown’s
equations tended to overestimate the predicted AGBs, while Basu-
ki’s equations provided more accurate estimates (Fig. 3f and g).
ð7Þ 3.3. Important factors in assessing tree aboveground biomass
Island grouping was not a significant factor (p > 0.1) for all
ð8Þ model types, which we assume is due to the fact that each site
shares similar characteristics, such as average peat depth and dip-
terocarp species domination. However, this study found that fam-
ily group was a significant factor for biomass estimation in PSF. It
influenced the aboveground biomass estimation in all model types.
WD Class factor was significant only in the models without WD
parameter (Table 3).
Plot residuals of mixed species models with WD as parameter
has normal distribution across WD values (Fig. 4.2), while mixed
species models without WD (Fig. 4.1and 4.3) have trending resid-
uals distribution with slope very significantly different from 0. In
contrast, plot residuals of species-grouped models based on WD
class have slightly normal residual distribution (Fig. 4.4 and 4.6),
even though without WD as parameter. While the species grouped
model with WD, has slope very close to 0 (Fig. 4.5) (see Table 4).

Table 5
Species grouped models for the aboveground biomass estimation in peat swamp forests. df is degree of freedom. RSE is residual standard error. AICc is corrected Akaike
Information Criterion. REst is ratio estimator. CF is correction factor. Dev REst is mean deviation of SGB estimates calculated using Rest. Dev CF is mean deviation of AGB estimates
calculated using CF.

Model name Species groups Model parameters df RSE R2 AICc REst CF Dev REst Dev CF
a b c d
MDBH Dipterocarpaceae !2.155 2.562 26 0.330 0.984 22 0.929 1.056 23.4 28.6
Non-dipterocarpaceae !1.888 2.437 120 0.409 0.963 132 27.3 29.8
– Hard wood !1.954 2.537 56 0.337 0.975 42 0.972 1.058 26.6 29.6
– Light wood !1.831 2.348 63 0.351 0.973 52 0.796 1.063 28.5 30.4
MDBH–WD Dipterocarpaceae !1.882 2.525 0.231 26 0.335 0.984 24 0.928 1.058 23.3 28.2
Non-dipterocarpaceae !1.343 2.423 0.776 120 0.343 0.974 90 27.0 28.5
– Hard wood !1.630 2.543 0.841 56 0.317 0.978 37 0.942 1.052 24.4 27.4
– Light wood !1.423 2.339 0.441 63 0.342 0.975 50 0.780 1.060 29.6 29.4
MDBH–H Dipterocarpaceae !3.113 1.733 1.206 26 0.275 0.989 14 1.032 1.038 20.6 20.8
Non-dipterocarpaceae !2.400 2.016 0.626 114 0.371 0.970 104 24.2 25.8
– Hard wood !2.886 1.815 1.085 56 0.280 0.983 22 1.020 1.040 21.9 22.5
– Light wood !2.132 2.103 0.376 63 0.336 0.976 47 0.881 1.058 25.8 29.4
MDBH–WD–H Dipterocarpaceae !2.716 1.662 0.352 1.230 26 0.275 0.990 16 1.034 1.038 20.2 20.4
Non-dipterocarpaceae !1.827 2.092 0.662 0.496 114 0.318 0.978 69 23.2 24.5
– Hard wood !2.562 1.871 0.669 1.008 56 0.265 0.985 17 0.998 1.036 20.1 20.8
– Light wood !1.731 2.099 0.427 0.369 63 0.327 0.977 45 0.862 1.055 25.9 27.5
248 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253

Fig. 5. Relative error (%) in aboveground biomass estimates of the peat swamp models and existing models. From left to right row: mixed species, dipterocarp,
non-dipterocarp hardwood and non-dipterocarp softwood model.
 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253 249

3.4. New allometric equations for peat swamp forest
The models for mixed species, which use DBH, H and WD as
parameters, explained 98% of tree biomass variation (see Table 4),
while those which use only DBH as predictor only explained 97% of
the variation. We found that the Ratio Estimator (REst) is better
than the correction factor (CF) for transforming back to biomass
value, as it provides lower deviations for all equations.
The common log linear model forms, performed better than
polynomial log linear and DBH-power log linear models. The mul-
ticolinearity test showed that Ln DBH parameter has multicolin-
earity problem with (Ln DBH)2 and (Ln DBH)3, with VIF of 297,
1279 and 395, respectively. Therefore, for each model type we sug-
gest the use of common log-linear form models (MDBH-; MDBH–WD;
MDBH–H and MDBH–WD–H). Accordingly, we further used the model
forms for developing specific species group models.
All species grouped models for the dipterocarp family performed
better (higher R2, lower AICc and RSE) than did mixed species models
(Table 5). Similarly, the WD-based grouping for non-dipterocarp
species reduced the RSE and increased the R2 for most models.
group MDBH models is less than 1%, which is much better than the
mixed species model which has a total error of 12.7%. The mean
Dev of species group models is less than that of both the existing
and our mixed species models. Without using an additional predic-
tor, the MDBH species group models could achieve a mean Dev sim-
ilar to that of the new pan-tropical Chave.DBH–WD–H.
None of the existing equations, including ours, can estimate
non-dipterocarp softwood trees accurately, due to two large outlier
trees. The outliers were the only samples representing large trees
with DBH > 150 cm. By including them, the equations tend to
underestimate the other normal trees in the mixed and softwood
models. We therefore updated the equations by excluding the out-
liers for mixed and softwood models (Fig. 6, Table 7). The new
models performed very well for mixed species and non-diptero-
carp softwood trees. Despite of the outliers exclusion, most of
the existing models were not valid for estimating non-dipterocarp
softwood trees (Fig. 6d, h and l), except for the new pan-tropical
equation from Chave.DBH–WD–H, which has relatively flat slope close
to 0 (Fig. 6p). The Chave.DBH–WD–H equation performed quite well
for large trees, but slightly underestimated small trees. All our
mixed species models performed very well in estimating AGB.

3.5. Accuracy comparison between new equations and the best 4. Discussion
existing equations
4.1. Factors affecting accuracy in biomass estimation
The Brown.DBH and Yamakura.DBH–H equations were the best fit
in estimating biomass of dipterocarp trees. This may be because Our results demonstrated that dipterocarp and non-dipterocarp
most of the sample trees used to develop the equations were col- grouping is an additional significant factor in AGB estimation in
lected from lowland dipterocarp forests of Indonesia and Malaysia. peat swamp forest. This finding is in agreement with the results
However, the equations tended to underestimate the non- of (Basuki et al., 2009), that taxon-based species grouping can
dipterocarp hard wood trees and overestimate the softwood trees improve the performance of AGB model. It explains AGB variation
(Fig. 5e–h and u-5x). beyond the traditional parameters (DBH, WD and H) can explain.
Both the Chave.DBH–WD–E and Chave.DBH–WD–H equations showed We developed 3 species groupings according to their major
similar mean Dev to our MDBH–WD and MDBH–WD–H mixed species family (dipterocarp and non-dipterocarp) and WD classes. By
models (Table 6). They both performed very well for large diptero- doing so, we kept the sample number high, rather than grouping
carp trees but underestimated dipterocarp trees with DBH < 50 cm them by genera or family. In fact, it is more practical to use fewer
(Fig. 5n and ad). They are also valid for estimating AGB of non- equations, rather than accommodating all genera or family, in esti-
dipterocarp hardwood trees. mating biomass in highly diverse tropical forest (Kartawinata,
Tree height was the most important estimator after DBH when 1990; Slik et al., 2003; Phillips et al., 2002). The significancy level
estimating non-dipterocarp soft wood trees of peat swamp forest. of WD parameters became less when the model was developed
Our models without H as predictor (MDBH and MDBH–WD) tended to based on WD class grouping. In the absence of WD as an indepen-
underestimate the non-dipterocarp softwood trees (see Fig. 5d and dent variable, the use of WD class grouping models can reduce the
l) when compared to the more accurate MDBH–H and MDBH–WD–H variation in estimating the biomass of mixed species models.
models (Fig. 5t and ac). In an opposite trend, the existing models Tree height was the most important parameter after DBH when
without H as variable (Brown.DBH and Chave.DBH–WD–E) tended to estimating non-dipterocarp soft wood trees of peat swamp forests.
overestimate the biomass of non-dipterocarp softwood trees Several studies outlined the lack of tree height data in most histor-
(Fig. 5h and p). ical inventories in tropical forest. Up-to-date tree height invento-
The use of species-group models increases the overall accuracy ries remain scarce due to visual limitations in dense tropical
of the estimates. They perform better than the best existing models forests (Basuki et al., 2009). However, LiDAR has recently been uti-
and our mixed species models (Table 6). The total error of species lized to determine not only forest canopy height (Asner et al.,
2012), but also individual tree height and height growth
Table 6
Error and deviation of the model estimates using existing model, mixed model and (Andersen et al., 2011). In addition, H–DBH models have also pro-
species-group model equations. Total Error is mean total predicted AGB subtracted by vided an alternative solution (Feldpausch et al., 2010; Feldpausch
mean total measured AGB then divided by mean total measured AGB. et al., 2012; Rutishauser et al., 2013; Chave et al., 2014). Develop-
Model name Total error (%) Mean Dev (%) Mean RE (%) ing a tree height model using handheld laser technology has also
been tested, with promising accuracy for biomass estimation in
Brown.DBH 25.23 33.9 7.9
MDBH mixed species 12.70 32.2 0.9 tropical forests (Rutishauser et al., 2013). Another possibility is
MDBH species group 0.34 26.5 !5.7 the use of bole height instead of total tree height, which is more
Chave.DBH–WD–E 14.89 28.1 !9.4 easily measured in the dense tropical forests, as suggested by
MDBH–WD mixed species 0.26 28.0 !10.6 Basuki et al. (2009), or in conjunction with the Terrestrial Laser
MDBH–WD species group !0.18 26.1 !8.8
Scanning application (Yang et al., 2013).
Yamakura.DBH–H 13.48 29.5 !3.6
MDBH–H mixed species !0.04 28.7 1.4
MDBH–H species group 0.08 24.4 0.9 4.2. Biomass equations for peat swamp forests
Chave.DBH–WD–H 7.31 25.7 !12.6
MDBH–WD–H mixed species 0.02 23.6 4.7
The biomass in large trees with DBH > 100 cm was highly vari-
MDBH–WD–H species group !0.08 22.4 4.0
able in the peat swamp forests studied. Such a variation in a few
250 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253

Fig. 6. Relative error (%) in aboveground biomass estimates of the models and existing models for mixed species and non-dipterocarp soft wood trees. The models excluded
outliers.
 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253 251

Table 7
Corrected equations of mixed species and species group developed without outliers and include correction factors. df is degree of freedom. RSE is residual standard error. AICc is
corrected Akaike Information Criterion.

Model Name Equations df RSE R2 AICc

2.513
MDBH mixed species 0.136 ⁄ DBH 145 0.385 0.969 139
MDBH species group Dipterocarp: 0.108 ⁄ DBH2.562 26 0.330 0.984 22
Non-dipterocarp hardwood: 0.138 ⁄ DBH2.537 56 0.337 0.975 42
Non-dipterocarp softwood: 0.149 ⁄ DBH2.399 61 0.342 0.97 47
MDBH–WD mixed species 0.242 ⁄ DBH2.473 ⁄ WD0.736 145 0.331 0.979 96
MDBH–WD species group Dipt: 0.141 ⁄ DBH2.525 ⁄ WD0.231 26 0.335 0.984 24
Non-dipterocarp hardwood: 0.185 ⁄ DBH2.4543 ⁄ WD0.841 56 0.317 0978 37
Non-dipterocarp softwood: 0.224 ⁄ DBH2.391 ⁄ WD0.485 61 0.331 0.973 44
MDBH–H mixed species 0.081 ⁄ DBH2.049 ⁄ H0.672 145 0.355 0.974 116
MDBH-H species group Dipt: 0.046 ⁄ DBH1.733 ⁄ H1.206 26 0.275 0.989 14
Non-dipterocarp hardwood: 0.057 ⁄ DBH1.815 ⁄ H1.085 56 0.28 0.983 22
Non-dipterocarp softwood: 0.124 ⁄ DBH2.192 ⁄ H0.296 61 0.335 0.972 46
MDBH–WD–H mixed species 0.15 ⁄ DBH2.095 ⁄ WD0.664 ⁄ H0.552 145 0.307 0.981 76
MDBH–WD–H species group Dipt: 0.068 ⁄ DBH1.662 ⁄ WD0.352 ⁄ H1.230 26 0.275 0.99 16
Non-dipterocarp hardwood: 0.077 ⁄ DBH1.871 ⁄ WD0.669 ⁄ H1.008 56 0.265 0.985 17
Non-dipterocarp softwood: 0.187 ⁄ DBH2.190 ⁄ WD0.474 ⁄ H0.287 61 0.323 0.974 43

trees is difficult to model accurately using a single predictor in
mixed-species models. The inclusion of the two largest old non-
dipterocarp trees but with relatively low AGB, Tetramerista glabra
with a DBH of 160 cm and Mangifera quadrifida with a DBH of
167 cm, resulted in compromised equations that avoided the large
positive deviation caused by large trees but allowed for the negative
deviation on smaller trees. The low biomass of these large DBH trees
may be the result of their relatively short height, which is similar to
that of emergent Shorea trees that have a DBH of only about 90 cm. A
study in lowland dipterocarp forest in East Kalimantan found simi-
lar characteristics, where some large trees with unexpected very
low height occurred in logged-over forests, but not in primary for-
ests (Rutishauser et al., 2013). However, this is in contradiction with
those in the lowland forest of Peru, where large but short trees
found to have relatively large biomass stored in the canopy, i.e.
44% of total AGB (Goodman et al., 2014).
We found that the equation of Kenzo et al. (2009a) underesti-
mated the measured AGB. This may be due to the fact that the study
was conducted in a secondary forest dominated by small diameter
trees. The equation failed to accurately estimate AGB of large trees
with a DBH beyond its DBH range. The new pan-tropical equations
(Chave et al., 2014) and the new Brown’s equations (Brown et al.,
2011) performed better than the existing local equations developed
by Basuki et al. (2009) and Ketterings et al. (2001), which underes-
timated measured AGB. This underestimation is in agreement with
previous studies in Indonesian tropical forests by Brown et al.
(2011) and Rutishauser et al. (2013). It is certainly that Ketterings’s
equation unable to predict AGB beyond its DBH range and species
composition. However, Basuki’s equations, which developed using
large DBH range, sample size and similar species composition
(dipterocarps dominated) did not perform well to our dataset. The
reason might be there were some differences in soil properties
(i.e. mineral soil vs. organic soil) and forest type (lowland diptero-
carp forest vs. peat swamp forest), or maybe due to unexpectedly
short and low in biomass of the large sample trees used for
allometric development, since the local equations from Basuki per-
formed better than all pan-tropical equations when estimating AGB
of the large but short trees in our peat swamp forest datasets.
Our results showed that the new pan-tropical equations,
Chave.DBH–WD–E and Chave.DBH–WD–H performed almost similar to
our model for mixed-species trees in tropical peat swamp forests.
The representativeness of Indonesian tropical forests as well as
large in DBH range and sample size, used in Chave et al. (2014)
study, could be the reason of this similar performance. The new
pan-tropical equations performed quite well in estimating AGB of
large dipterocarp trees, but was less accurate in the case of small
trees. This inaccuracy may fade away at plot and landscape level,
if the tree WD variations are similar to those in mixed tropical for-
ests. In the absence of local allometric equations, the existing pan-
tropical equations are still reliable for estimating AGB in tropical
forests, even for peat swamp forests. However the pan-tropical
equations need to be used carefully for estimating AGB of large
but short trees in logged peat swamp forests, from which most
commercial large emergent dipterocarp trees have been extracted.
We found that AGB models developed based on species grouping
performed better than mixed model. This is because the species
grouped models taken into account variation within dipterocarp
and non-dipterocarp families and wood density class.
4.3. The use of wood density in models
The inclusion of WD as a predictor in allometric models could
generate inaccurate results, especially when the identification of
tree species is ambiguous or based on commercial inventories
instead of scientific research, which is common in most forest
inventories in developing countries. In the absence of field data it
is a common practice to use WD global databases in estimating
tree biomass, which has large uncertainties in our study. Some
studies even use an averaged WD value in estimating AGB in a for-
est ecosystem. This may induce errors due to variations in WD, not
only between species but also among individual trees within spe-
cies (Henry et al., 2010) and among tree sections (Iida et al., 2012).
Although it is noted that WD variation within species is less than
between species.
Henry et al. (2010) demonstrated that the variation in WD is not
only influenced by species and light-demanding factors, but also by
the tree size and distance of wood sample to the bark. The low corre-
lation we found between measured WD and WD values in the global
database of Zanne et al. (2009) illustrates well the existing high
variation in WD. It is likely that the data used to compile this global
database originated from commercial-sized trees, thus that the data-
base under-represents smaller trees. Finally, methodological differ-
ences (i.e. percentage of water content or average WDs vs. WD at a
DBH point) may also explain discrepancies between WD values from
field measurements and those from the global database (Reyes et al.,
1992). Defining WD classes and integrating these into biomass esti-
mation models would be more practical and cost-efficient and
would help to narrow WD variation-derived errors.
252 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253
5. Conclusions
This study provides more accurate AGB estimates in PSF
through species-grouping models which were developed from a
datasets with wide range of species, DBH, WD and H. Species-
group equations based on dipterocarp and non-dipterocarp family
and WD class improve the performance of mixed-species
equations in AGB estimation. In the absence of WD values, WD
classes-based equations will improve the accuracy of the estima-
tion. Tree height is an additional key parameter in estimating
aboveground biomass in peat swamp forests, especially for large
non-dipterocarp trees. Although all of the existing equations per-
formed similarly to our mixed-species and dipterocarp models,
they systematically underestimated or overestimated the AGB of
certain species-groups, especially for non-dipterocarp trees. How-
ever, if an accurate and validated local equation is not available,
the new pan-tropical equations developed by Chave et al. (2014)
are more reliable than local equations which developed using
limited samples and DBH range.
Acknowledgements
The authors thank the German International Cooperation –
Merang REDD Pilot Project (GIZ-MRPP), GIZ-FORCLIME, KfW-FORC-
LIME projects and ACIAR for funding the field research. We highly
appreciate the support and permit granting from PT RHM, PT DRT
and PT SMD for the field data collection. For arranging permits and
access for tree cutting, we also wish to thank Mr. Zulfikhar from
South Sumatra Forestry Service and Mr. Indra Kumara from Kapuas
Hulu Forestry Service. For the hard work during the field data col-
lection, we extended personal thanks to all field crews and field
staff from KMPH Merang, PT DRT, Saka Wanabhakti and Forestry
Service of Kapuas Hulu. We also would like to thank Julie Watson
from Fenner School of the Australian National University and two
anonymous reviewers for suggestions to improve the manuscript.
Solichin Manuri was supported by the Australian Award
Scholarship and received financial support from Indonesian
Peatland Network initiated by ICCC/DNPI, CIFOR and USFS.
References
Alvarez, E., Rodríguez, L., Duque, A., Saldarriaga, J., Cabrera, K., de las Salas, G., del
Valle, I., Lema, A., Moreno, F., Orrego, S., 2012. Tree above-ground biomass
allometries for carbon stocks estimation in the natural forests of Colombia. For.
Ecol. Manage. 267, 297–308.
Andersen, H.-E., Strunk, J., Temesgen, H., 2011. Using airborne light detection and
ranging as a sampling tool for estimating forest biomass resources in the Upper
Tanana Valley of Interior Alaska. West. J. Appl. For. 26, 157.
Asner, G.P., Mascaro, J., Muller-Landau, H.C., Vieilledent, G., Vaudry, R., Rasamoelina,
M., Hall, J.S., van Breugel, M., 2012. A universal airborne LiDAR approach for
tropical forest carbon mapping. Oecologia 168, 1147–1160.
Baccini, A., Friedl, M.A., Samanta, S., Houghton, R.A., Goetz, S.J., Walker, W.S.,
Laporte, N.T., Sun, M., Sulla-menashe, D., Hackler, J., Beck, P.S.A., Dubayah, R.,
2012. Estimated carbon dioxide emissions from tropical deforestation improved
by carbon-density maps. Nat. Clim. Change 2, 182.
Baskerville, G., 1972. Use of logarithmic regression in the estimation of plant
biomass. Can. J. For. Res. 2, 49–53.
Basuki, T.M., van Laake, P.E., Skidmore, A.K., Hussin, Y.A., 2009. Allometric equations
for estimating the above-ground biomass in tropical lowland Dipterocarp
forests. For. Ecol. Manage. 257, 1684–1694.
Brown, S., Casarim, F., Grimland, S., Pearson, T., 2011. Carbon Impacts from Selective
Logging of Forests in Berau District, East Kalimantan, Indonesia.
Chave, J., Kira, T., Lescure, J.P., Nelson, B.W., Ogawa, H., Puig, H., Riéra, B., Yamakura,
T., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., Fölster, H.,
Fromard, F., Higuchi, N., 2005. Tree allometry and improved estimation of
carbon stocks and balance in tropical forests. Oecologia 145, 87–99.
Chave, J., Réjou-Méchain, M., Búrquez, A., Chidumayo, E., Colgan, M.S., Delitti, W.B.,
Duque, A., Eid, T., Fearnside, P.M., Goodman, R.C., Henry, M., Martínez-Yrízar, A.,
Mugasha, W.A., Muller-Landau, H.C., Mencuccini, M., Nelson, B.W., Ngomanda,
A., Nogueira, E.M., Ortiz-Malavassi, E., Pélissier, R., Ploton, P., Ryan, C.M.,
Saldarriaga, J.G., Vieilledent, G., 2014. Improved allometric models to estimate
the aboveground biomass of tropical trees. Glob. Change Biol..
Cole, T.G., Ewel, J.J., 2006. Allometric equations for four valuable tropical tree
species. For. Ecol. Manage. 229, 351–360.
Djomo, A.N., Ibrahima, A., Saborowski, J., Gravenhorst, G., 2010. Allometric
equations for biomass estimations in Cameroon and pan moist tropical
equations including biomass data from Africa. For. Ecol. Manage. 260, 1873–
1885.
Fayolle, A., Doucet, J.-L., Gillet, J.-F., Bourland, N., Lejeune, P., 2013. Tree allometry in
Central Africa: testing the validity of pantropical multi-species allometric
equations for estimating biomass and carbon stocks. For. Ecol. Manage. 305, 29–
37.
Feldpausch, T.R., Bird, M., Brondizio, E.S., Camargo, P.D., Chave, J., Djagbletey, G.,
Domingues, T.F., Drescher, M., Fearnside, P.M., França, M.B., Fyllas, N.M., Banin,
L., Lopez-Gonzalez, G., Hladik, A., Higuchi, N., Hunter, M.O., Iida, Y., Silam, K.A.,
Kassim, A.R., Keller, M., Kemp, J., King, D.A., Phillips, O.L., Lovett, J.C., Marimon,
B.S., Marimon-Junior, B.H., Lenza, E., Marshall, A.R., Metcalfe, D.J., Mitchard,
E.T.A., Moran, E.F., Nelson, B.W., Nilus, R., Baker, T.R., Nogueira, E.M., Palace, M.,
Patiño, S., Peh, K.S.H., Raventos, M.T., Reitsma, J.M., Saiz, G., Schrodt, F., Sonké, B.,
Taedoumg, H.E., Lewis, S.L., Tan, S., White, L., Wöll, H., Lloyd, J., Quesada, C.A.,
Affum-Baffoe, K., Arets, E.J.M.M., Berry, N.J., 2010. Height-diameter allometry of
tropical forest trees. Biogeosci. Discuss. 7, 7727–7793.
Feldpausch, T.R., Affum-Baffoe, K., Alexiades, M., Almeida, S., Amaral, I., Andrade, A.,
Aragão, L.E.O.C., Murakami, A.A., Arets, E.J.M.M., Arroyo, L., C, G.A.A., Lloyd, J.,
Baker, T.R., Bánki, O.S., Berry, N.J., Cardozo, N., Chave, J., Comiskey, J.A., Dávila,
E.A., Oliveira, A.D., DiFiore, A., Djagbletey, G., Lewis, S.L., Domingues, T.F., Erwin,
T.L., Fearnside, P.M., França, M.B., Freitas, M.A., Higuchi, N., C, E.H., Iida, Y.,
Jiménez, E., Kassim, A.R., Brienen, R.J.W., Killeen, T.J., Laurance, W.F., Lovett, J.C.,
Malhi, Y., Marimon, B.S., Marimon-Junior, B.H., Lenza, E., Marshall, A.R.,
Mendoza, C., Metcalfe, D.J., Gloor, E., Mitchard, E.T.A., Nelson, B.W., Nilus, R.,
Nogueira, E.M., Parada, A., Peh, K.S.H., Cruz, A.P., Peñuela, M.C., Pitman, N.C.A.,
Prieto, A., Mendoza, A.M., Quesada, C.A., Ramírez, F., Ramírez-Angulo, H.,
Reitsma, J.M., Rudas, A., Saiz, G., Salomão, R.P., Schwarz, M., Silva, N., S.ilva-
Espejo, J.E., Lopez-Gonzalez, G., Silveira, M., Sonké, B., Stropp, J., Taedoumg, H.E.,
Tan, S., Steege, H.t., Terborgh, J., Torello-Raventos, M., Heijden, G.M.F.v.d.,
Vásquez, R., Banin, L., Vilanova, E., Vos, V., White, L., Wilcock, S., Woell, H.,
Phillips, O.L., Salim, K.A., 2012. Tree height integrated into pan-tropical forest
biomass estimates. Biogeosci. Discuss. 9, 2567–2622.
Goodman, R.C., Phillips, O.L., Baker, T.R., 2014. The importance of crown dimensions
to improve tropical tree biomass estimates. Ecol. Appl. 24, 680–698.
Hashimoto, T., Tange, T., Masumori, M., Yagi, H., Sasaki, S., Kojima, K., 2004.
Allometric equations for pioneer tree species and estimation of the
aboveground biomass of a tropical secondary forest in East Kalimantan.
Tropics 14, 123–130.
Henry, M., Besnard, A., Asante, W., Eshun, J., Adu-Bredu, S., Valentini, R., Bernoux,
M., Saint-Andre, L., 2010. Wood density, phytomass variations within and
among trees, and allometric equations in a tropical rainforest of Africa. For. Ecol.
Manage. 260, 1375–1388.
Hergoualc’h, K., Verchot, L.V. 2011 Stocks and fluxes of carbon associated with land-
use change in Southeast Asian tropical peatlands: a review. Glob. Biochem.
Cycles, 25, GB2001, doi: 2010.1029/2009GB003718.
Hergoualc’h, K., Verchot, L.V., 2014. Greenhouse gas emission factors for land use
and land-use change in Southeast Asian peatlands. Mitig. Adapt. Strateg. Glob.
Change 19, 789–807.
Houghton, R., House, J., Pongratz, J., van der Werf, G., DeFries, R., Hansen, M., Quéré,
C.L., Ramankutty, N., 2012. Carbon emissions from land use and land-cover
change. Biogeosciences 9, 5125–5142.
Hurvich, C.M., Tsai, C.-L., 1991. Bias of the corrected AIC criterion for underfitted
regression and time series models. Biometrika 78, 499–509.
Iida, Y., Poorter, L., Sterck, F.J., Kassim, A.R., Kubo, T., Potts, M.D., Kohyama, T.S.,
2012. Wood density explains architectural differentiation across 145 co-
occurring tropical tree species. Funct. Ecol. 26, 274–282.
Jaenicke, J., Rieley, J.O., Mott, C., Kimman, P., Siegert, F., 2008. Determination of the
amount of carbon stored in Indonesian peatlands. Geoderma 147, 151–158.
Jubanski, J., Ballhorn, U., Kronseder, K., Franke, J., Siegert, F., 2012. Detection of large
above ground biomass variability in lowland forest ecosystems by airborne
LiDAR. Biogeosci. Discuss. 9, 11815–11842.
Kartawinata, K., 1990. A review of natural vegetation studies in Malesia, with
special reference to Indonesia. The plant diversity of Malesia. Proc. symposium,
Leiden 1989, 121–132.
Kenzo, T., Furutani, R., Hattori, D., Kendawang, J.J., Tanaka, S., Sakurai, K., Ninomiya,
I., 2009a. Allometric equations for accurate estimation of above-ground
biomass in logged-over tropical rainforests in Sarawak, Malaysia. J. For. Res.
14, 365.
Kenzo, T., Takahashi, N., Okamoto, M., Tanaka-Oda, A., Sakurai, K., Ninomiya, I.,
Ichie, T., Hattori, D., Itioka, T., Handa, C., Ohkubo, T., Kendawang, J.J., Nakamura,
M., Sakaguchi, M., 2009b. Development of allometric relationships for accurate
estimation of above- and below-ground biomass in tropical secondary forests in
Sarawak, Malaysia. J. Trop. Ecol. 25, 371–386.
Ketterings, Q.M., Coe, R., van Noordwijk, M., Ambagau’, Y., Palm, C.A., 2001.
Reducing uncertainty in the use of allometric biomass equations for predicting
above-ground tree biomass in mixed secondary forests. For. Ecol. Manage. 146,
199–209.
Komiyama, A., Jintana, V., Sangtiean, T., Kato, S., 2002. A common allometric
equation for predicting stem weight of mangroves growing in secondary
forests. Ecol. Res. 17, 415–418.
Krisnawati, H., Adinugroho, C., dan Imanuddin, R., 2012. Monograf Model-Model
Alometrik untuk Pendugaan Biomassa Pohon pada berbagai tipe ekosistem
 S. Manuri et al. / Forest Ecology and Management 334 (2014) 241–253
hutan di Indonesia. Badan Penelitian dan Pengembangan Kehutanan, Pusat
Penelitian dan Pengembangan Konservasi dan Rehabilitas, Bogor.
Kronseder, K., Ballhorn, U., Böhm, V., Siegert, F., 2012. Above ground biomass
estimation across forest types at different degradation levels in Central
Kalimantan using LiDAR data. Int. J. Appl. Earth Observ. Geoinform. 18, 37–48.
Miettinen, J., Shi, C., Liew, S.C., 2011. Influence of peatland and land cover
distribution on fire regimes in insular Southeast Asia. Region. Environ. Change
11, 191–201.
Ministry of Environment, 2010. Indonesia Second National Communication under
the United Nations Framework Convention on Climate Change (UNFCCC).
Ministry of Environment of the Republic of Indonesia, Jakarta.
Ministry of Forestry, 2012. Digital land cover map of Indonesia for year 2012.
Spatial Planning Agency. Ministry of Forestry of the Republic of Indonesia,
Jakarta.
Ngomanda, A., Engone Obiang, N.L., Lebamba, J., Moundounga Mavouroulou, Q.,
Gomat, H., Mankou, G.S., Loumeto, J., Midoko Iponga, D., Kossi Ditsouga, F.,
Zinga Koumba, R., 2014. Site-specific< i> versus</i> pantropical allometric
equations: Which option to estimate the biomass of a moist central African
forest? For. Ecol. Manage. 312, 1–9.
Pelletier, J., Kirby, K.R., Potvin, C., 2012. Significance of carbon stock uncertainties on
emission reductions from deforestation and forest degradation in developing
countries. For. Policy Econ. 24, 3–11.
Phillips, P.D., Yasman, I., Brash, T.E., van Gardingen, P.R., 2002. Grouping tree species
for analysis of forest data in Kalimantan (Indonesian Borneo). For. Ecol. Manage.
157, 205–216.
Reyes, G., Brown, S., Chapman, J., Lugo, A.E., 1992. Wood densities of tropical tree
species.
Rutishauser, E., Noor’an, F., Laumonier, Y., Halperin, J., Rufi’ie, Hergoualc’h, K.,
Verchot, L., 2013. Generic allometric models including height best estimate
forest biomass and carbon stocks in Indonesia. For. Ecol. Manage. 307, 219–225.
253
Slik, J.W.F., Poulsen, A.D., Ashton, P.S., Cannon, C.H., Eichhorn, K.A.O., Kartawinata,
K., Lanniari, I., Nagamasu, H., Nakagawa, M., Van Nieuwstadt, M.G.L., Payne, J.,
Purwaningsih, , Saridan, A., Sidiyasa, K., Verburg, R.W., Webb, C.O., Wilkie, P.,
2003. A floristic analysis of the lowland dipterocarp forests of Borneo. J.
Biogeogr. 30, 1517–1531.
Snowdon, P., 1991. A ratio estimator for bias correction in logarithmic regressions.
Can. J. For. Res. 21, 720–724.
Sprugel, D., 1983. Correcting for bias in log-transformed allometric equations.
Ecology 64, 209–210.
van Breugel, M., Ransijn, J., Craven, D., Bongers, F., Hall, J.S., 2011. Estimating carbon
stock in secondary forests: decisions and uncertainties associated with
allometric biomass models. For. Ecol. Manage. 262, 1648–1657.
Vieilledent, G., Vaudry, R., Andriamanohisoa, S.F.D., Rakotonarivo, O.S.,
Randrianasolo, H.Z., Razafindrabe, H.N., Rakotoarivony, C.B., Ebeling, J.,
Rasamoelina, M., 2012. A universal approach to estimate biomass and carbon
stock in tropical forests using generic allometric models. Ecol. Appl. 22, 572–
583.
Warren, M.W., Kauffman, J.B., Murdiyarso, D., Anshari, G., Hergoualc’h, K.,
Kurnianto, S., et al., 2012. A cost-efficient method to assess carbon stocks in
tropical peat soil. Biogeosci. Discuss. 9, 7049–7071.
Yamakura, T., Hagihara, A., Sukardjo, S., Ogawa, H., 1986. Aboveground biomass of
tropical rain forest standsin Indonesian Borneo. Plan Ecol. 68, 71–82.
Yang, X., Strahler, A.H., Schaaf, C.B., Jupp, D.L.B., Yao, T., Zhao, F., Wang, Z., Culvenor,
D.S., Newnham, G.J., Lovell, J.L., Dubayah, R.O., Woodcock, C.E., Ni-Meister, W.,
2013. Three-dimensional forest reconstruction and structural parameter
retrievals using a terrestrial full-waveform lidar instrument (Echidna#). Rem.
Sens. Environ. 135, 36–51.
Zanne, A., Lopez-Gonzalez, G., Coomes, D., Ilic, J., Jansen, S., Lewis, S., Miller, R.,
Swenson, N., Wiemann, M., Chave, J., 2009. Global wood density database.
Dryad. Identifier: http://hdl. handle. net/10255/dryad 235.