ArticlePDF Available

An estimate of above-ground carbon stock in tropical rainforest on Manus Island, Papua New Guinea

  • New Britain Palm Oil Limited

Abstract and Figures

Forest carbon emission mitigation schemes seek to protect tropical forest, combat effects of climate change, and offer potential cash and development opportunities. Reducing emissions from deforestation and degradation (REDD+) projects based on a foundation of accurate carbon stock assessment provide such an opportunity for Papua New Guinea. The objective of this study was to quantify the carbon stock of the central forests of Manus Island, Papua New Guinea, and identify factors that underpin any observed variation within it. We employed the Winrock Standard Operating Procedures for Terrestrial Carbon Measurement for plots and associated measurements. In 75 variable-radius nested plots (total area ≤14.4ha), we assessed above-ground and total carbon stock of stems ≥5cm diameter at breast height via general linear models in a model-selection framework. The top models described variation in average carbon stock at 95% lower and upper confidence interval in above-ground biomass solely in terms of forest type: primary hill forest 165.0MgCha-1 (148.3-183.7, n≤48), primary plain forest 100.9MgCha-1 (78.0-130.6, n≤10) and secondary hill forests 99.7MgCha-1 (80.9-122.9, n≤17). To a lesser extent, above-ground carbon stock increased with slope and varied idiosyncratically by the nearest village. Our estimates are comparable with published studies for Papua New Guinea and the wider tropical region. These data should strengthen pre-existing knowledge and inform policies on carbon accounting for REDD+ projects in the region.
Content may be subject to copyright.
An estimate of above-ground carbon stock in tropical
rainforest on Manus Island, Papua New Guinea
Arison Arihafa
,Sebastian Dalgarno
and Ezra Neale
Wildlife Conservation Society, Papua New Guinea Program, PO Box 277, Goroka,
Eastern Highlands Province 441, Papua New Guinea.
Corresponding author. Email:
Abstract. Forest carbon emission mitigation schemes seek to protect tropical forest, combat effects of climate change,
and offer potential cash and development opportunities. Reducing emissions from deforestation and degradation
(REDDþ) projects based on a foundation of accurate carbon stock assessment provide such an opportunity for Papua
New Guinea. The objective of this study was to quantify the carbon stock of the central forests of Manus Island, Papua New
Guinea, and identify factors that underpin any observed variation within it. We employed the Winrock Standard Operating
Procedures for Terrestrial Carbon Measurement for plots and associated measurements. In 75 variable-radius nested plots
(total area ¼14.4 ha), we assessed above-ground and total carbon stock of stems $5 cm diameter at breast height via
general linear models in a model-selection framework. The top models described variation in average carbon stock at 95%
lower and upper confidence interval in above-ground biomass solely in terms of forest type: primary hill forest
165.0 Mg C ha
(148.3–183.7, n¼48), primary plain forest 100.9 Mg C ha
(78.0–130.6, n¼10) and secondary
hill forests 99.7 Mg C ha
(80.9–122.9, n¼17). To a lesser extent, above-ground carbon stock increased with slope and
varied idiosyncratically by the nearest village. Our estimates are comparable with published studies for Papua New Guinea
and the wider tropical region. These data should strengthen pre-existing knowledge and inform policies on carbon
accounting for REDDþprojects in the region.
Additional keywords: aboveground carbon, biomass, environmental variables, Papua New Guinea, plain forest, primary
hill forest, REDDþ, secondary hill forest.
Received 15 June 2015, accepted 10 November 2015, published online 4 December 2015
Tropical deforestation is a major contributor to greenhouse gas
emissions (Malhi and Grace 2000;Bryanet al. 2010a;Harris et al.
2012), largely resulting from agriculture and forestry activities
(Van Laake and Sanchez-Azofeifa 2004;Ferraz et al. 2005;
Cayuela et al. 2006). Recently, tropical countries have been able
to take advantage of a developing mitigation system that aims to
reduce emissions from deforestation and degradation (REDDþ)
as a means of protecting forests and combating the effects of
climate change. Schemes like REDDþprovideanimportant
opportunity for lowering carbon emissions within tropical coun-
tries while generating income from forest retention at the same
time (Laurance 2007). The prospect of REDDþin Papua New
Guinea (PNG) offers the ability for many rural communities to
access cash and development opportunities without going down
the path of resource extraction (GPNG 2007). PNG has a large
rural population, and is unusual globally in that most of land
remains under community ownership (GPNG 2007). Despite
this, PNG has long had difficulty in safeguarding its forest estate
(e.g. Saulei 1990;Shearman et al. 2008;Laurance et al. 2011).
Yet, before any emission structures can be developed, quantifi-
cation of carbon stock content is needed as a prerequisite.
Carbon stocks are quantified through forest inventories,
destructive sampling, allometric relationships and remotely
sensed data (Malhi and Grace 2000;Chave et al. 2003;Bryan
et al. 2010b;Fox et al. 2010). Measurements of carbon stock is
useful for developing climate change mitigation policies (Chave
et al. 2003;Fox et al. 2010), and at the same time it can be used
for monitoring the global carbon cycle (Phillips and Gentry
1994). Some important predictors affecting overall estimates in
tropical forest carbon stocks include appropriate classifications
of forest type (e.g. dry, moist or wet), specific tree variables
(e.g. tree diameter, total tree height and wood specific gravity),
slope, drainage class, land-use history, elevation and soil type
(Edwards and Grubb 1977;Yamakura et al. 1986;Chave et al.
2005;Gibbs et al. 2007;Lewis et al. 2009;Fox et al. 2010).
Although precise estimates of forest carbon pools are
required for participation in REDDþschemes, quantification
for both above- and below-ground carbon pools in PNG has been
scant (Edwards and Grubb 1977;Abe 2007;Bryan et al. 2010a,
2010b;Fox et al. 2010;FPCD and IGES 2013;Vincent et al.
2015). The central forests of Manus Island in PNG have been
identified as a prospective REDDþsite; consequently, the
Wildlife Conservation Society (WCSPNG) Program was funded
Pacific Conservation Biology, 2015, 21, 307–314
Journal compilation ÓCSIRO 2015
by the Australian Department of Foreign Affairs and Trade to
carry out a REDDþreadiness project in this area. As part of this
project, we quantified the carbon stock in above-ground biomass
of these forests and identified factors that underpinned major
patterns of variation.
Study site
We conducted this study from May 2013 to April 2014 within
lowland forests of central Manus Island, PNG (Fig. 1). Manus
Island is the major island within the Admiralty Group of islands
and covers one of the largest remaining forest areas in the
Bismarck Archipelago. The study was undertaken within lands
belonging to nine consenting landowner clans at five villages:
Tulu 1, Saha, Pelipowai, Piri and Djekal (Fig. 1). The general
climate of Manus Province is equatorial with high daily tem-
peratures at 25–328C. The annual rainfall is between 3000 and
4000 mm, with two noticeable wetter seasons corresponding to
the times of the trade winds (Croft 1983): the south-easterly trade
winds from the middle half to the end of the year and the north-
west trade winds for the remainder of the year (Kisokau 1974).
The forest on central Manus Island is a typical lowland tropical
forest (sensu Paijmans 1976)dominatedbyCalophyllum spp.
(Calophyllaceae) and Dillenia papuana (Dilleniaceae), and
common tree families include Sapindaceae, Lauraceae,
Myristicaceae, Myrtaceae, Euphorbiaceae and Apocynaceae. The
main vegetation types include: hill forest, plain forests, swamp
forests (commonly covered with sago palms), and mangrove
forest alongestuaries and coastlines. Secondary forests atvarying
ages in regeneration, from both natural and anthropogenic
disturbances (e.g. shifting cultivation, settlement), are common
Forest stratification
We collected data from three main forest types: primary hill forest
(intact forest on well-raised slopes with good drainage), primary
plain forest (intact forest on flat lands and at low elevation, and
may be subject to periodic inundation during the wet season), and
secondary hill forest (original forest structure and species com-
position are altered, and occur on well-raised slopes with good
drainage). As stratification increases survey efficiency and cap-
tures major variations among forest types (Gibbs et al. 2007), our
forest classifications were consistent with the forest classes pro-
duced by the Japanese International Cooperation Agency and the
government of the PNG Forest Authority: hill forest, plain forest
and non-forest. We differentiated the hill forest into primary and
secondary forests; however, we did not observe any secondary
forests within the plain forest. Secondary forests elsewhere
resulted mostly from anthropogenic activities. Secondary forests
may also be classified into different ages (e.g.young, medium and
old) after disturbance for effective sampling. Here we sampled
only old secondary hill forest estimated at $40 years old on the
basis of local knowledge and our field observations of forest
structure and species composition present at that time. This limit
was set to avoid sampling in young forest areas, which are rou-
tinely disturbed by shifting cultivations and subsistence use.
Determination of plots
On a digitised 1980 Australian Survey Corps topographic map,
random plot locations were mapped within the sampling area
Manus Island
Manus Island
20 km
Koroji Butchou
Papua New Guinea
Tulu 1
Fig. 1. Map of Papua New Guinea showing Manus Island and location of 75 forest carbon plots (black dots)
assessed in five villages of central Manus.
308 Pacific Conservation Biology A. Arihafa et al.
using ArcGIS software with a minimum of 500 m distance
between each plot. The points were projected within ArcGIS
Editor, using a random sample point generator tool via a strat-
ified random sampling technique. Buffer zones were created –
30 m from rivers, 30 m from a clan boundary, and 500 m away
from any settlement – to eliminate chances that a plot was
located outside of the three forest types. Sampling points were
then created on clan boundary maps and printed for use before
undertaking field work.
Plot design and tree data measurement
The protocol for setting up plots and taking measurements fol-
lowed Walker et al. (2012). Seventy-five plots (primary hill ¼48,
secondary hill ¼17 and primary plain ¼10) were sampled
within E71 737 ha of forested landscape identified as a possible
REDDþproject. We were limited to sampling only on the land
of consenting clans. The number of plots targeted for each forest
type was deemed sufficient by the Winrock Carbon Stock Cal-
culation Tool (WCSCT) algorithm to achieve a 95% confidence
level within 5% of the true mean of the above-ground carbon
(AGC) (Walker et al. 2012), and it was sufficient to account for
the variability in each forest type. Plot locations were identified
in the field (from preselected points) with handheld GPS units.
A plot was relocated using a random compass bearing (0–3608)
and distance (determined by dividing the bearing by 4) if the plot
centre lay on a steep slope, cliff or within a non-forested area.
Non-forested areas, defined as .90% canopy opening or .10%
of the plot covered by grassland, bamboo or scrambling fern
(Dicranopteris linearis) were not sampled. A variable-radius
nested plot (each consisting of three concentric circular plots)
design was employed to take measurements (diameter at breast
height (DBH) and height) for both live and dead standing trees,
as in Table 1. Within each plot, we recorded DBH at 1.3 m above
ground for all stems $5 cm, the location (distance and bearing)
of each tree from the plot centre, slope and elevation. Each dead
standing tree in the plot was classified as either Condition ‘1’ or
‘2’ (1: with branches and twigs and resembles a live tree except
for leaves; 2: those containing small and large branches or only
the bole remaining) (sensu Walker et al. 2012). Height for dead
standing trees was measured using a clinometer. DBH and
length of coarse woody debris (CWD) lying on the ground was
measured along two transects crossing perpendicularly to each
other at the centre of the circular plot, giving a total transect
length of 80 m. Density class was estimated (i.e. sound, inter-
mediate, rotten) for each piece of CWD on the transect.
We used two groups of locally trained field assistants who
assisted with undertaking measurements in the field (sensu Butt
et al. 2015): WCS community facilitators (CFs), and interested
members of the community from each village. CFs were
community members trained in community conservation and
supervised by a WCS staff member throughout. Six CFs were
trained in forest inventory along with more than 40 local
community members for this project. In this inventory,
above-ground carbon stocks (both dead and live) for stems
$5 cm DBH were quantified, and below-ground carbon was
calculated post hoc using the WCSCT (Walker et al. 2012).
Time and cost prohibited measuring of carbon in other carbon
pools: soil, litter and above-ground non-trees (e.g. trees ,5cm
DBH, herbs, epiphytes, palms, pandanus and lianas), and
consequently these carbon stock were not quantified. However,
established above-ground biomass (AGB) ratios of trees are
readily available and often used to estimate other above-ground
carbon pools and below-ground carbon (e.g. Mokany et al.
2006;Fox et al. 2010).
Quantifying carbon stock
Allometric models are commonly used in biomass studies to
estimate forest carbon stock if a biomass expansion factor
method is not used (Pedroni 2007). We used the allometric
equation of Chave et al. (2005) for estimating above-ground
biomass for tropical moist forests that receive rainfall of 1500–
3500 mm annually. Their analysis (based on series of allome-
tric equations) relied upon compilation of tree harvested studies
from 27 datasets across a broad range of tropical forests. We
have been consistent with other studies (e.g. Bryan et al. 2010b;
Fox et al. 2010) in using the allometric equation of Chave et al.
(2005) to estimate carbon stock in tropical forests of PNG. With
an average error in the allometric models on the estimation of a
tree’s biomass to be 5%, Chave et al. (2005) have assured that
these regression models can reliably be used to predict carbon
across forest types in tropical forests.
We used the average wood density estimate of 0.477 g cm
for PNG lowland forests (Fox et al. 2010) to calculate the AGC
of all trees in the stand. Complete identifications of all trees
recorded within our sampling plots were not available at the time
of publication, but preliminary analysis carried out revealed that
the average wood density of 0.476 g cm
across the known
species was comparable to the estimate of Fox et al. (2010).
Average wood density estimates have been used in other studies
either at species, genus or stand level when such data were
unavailable (Chave et al. 2003;Fox et al. 2010;Butt et al. 2015).
AGB (in kilograms), based on Chave et al. (2005), was calcu-
lated as:
AGB ¼wood density expð1:499 þ2:148 ln DBHðÞ
þ0:207ðln DBH
20:0281ðln DBH
where wood density is in grams per cubic centimetre (g cm
and DBH is in centimetres (cm).
Estimations of AGB (kg ha
) were converted to mega-
grams of AGC per hectare (Mg C ha
fraction of 47% (Walker et al. 2012). Following Walker
et al. (2012), below-ground carbon was calculated using a
Table 1. Size classes of trees for both live and dead-standing trees
measured within the circular nested plot design adapted from Walker
et al. (2012)
Total area sampled was 14.4 ha
Subplot size Radius
DBH size class
Total area
Small 4 50.3 5–19.9 0.4
Medium 14 615.7 20–49.9 4.6
Large 20 1256.6 $50 9.4
Total 14.4
Carbon stock in tropical rainforest on Manus Island Pacific Conservation Biology 309
root : shoot AGB ratio for tropical moist forest as in Mokany
et al. (2006):
BGC ¼0:235 AGC if above-ground biomass carbon
BGC ¼0:205 AGC if above-ground biomass carbon
where BGC ¼below-ground carbon, and AGC ¼above-ground
Statistical analysis
Five explanatory variables were modelled to determine local
spatial effects on the most likely factors affecting variation in
AGC stock: forest class, forest type, slope, elevation and loca-
tion. AGC was calculated from AGB estimation; ‘forest class’
was hill or plain forest; ‘forest type’ was primary, secondary
or plain forests; ‘slope’ was the average inclination of plot;
‘elevation’ was height above sea level (m); and ‘location’ was
the name of the clan (or village) that owned a specific forest area.
‘Location’ allows interpretations of differences in carbon stock
due to village management and related local attributes. A simple
intercept model was also run within the models to reveal a
baseline for poor performance. We used an information theoretic
approach to assess 12 linear models in terms of parsimony (sensu
Burnham and Anderson 2002). The candidate models reflected
our a priori hypotheses about how differing factors could
influence AGC and combined carbon (CC) stocks (total carbon
of above-ground, below-ground, dead standing and CWD).
All analysis was carried out using R ver. 3.0 (R Development
Core Team 2013). As the residuals of our carbon stock measure-
ments did not meet the requirement of normality, AGC and CC
were log-transformed before analysis. We report means with
95% (lower and upper) confidence intervals (CI) derived from
our top-ranked models.
The only independent variable for the most parsimonious
model for AGC stocks was forest type (model weight 64%:
Table 2). There was also some lower support for models with
forest type and slope (model weight 23%) and forest type and
location (model weight 13%) as explanatory variables
(Table 2). Mean carbon stocks in AGB were: primary hill forest
165.0 Mg C ha
(95% CI: 148.3–183.7, n¼48), primary plain
100.9 Mg C ha
(78.0–130.6, n¼10) and secondary hill
forests 99.7 Mg C ha
(80.9–122.9, n¼17) (Fig. 2). AGC
increased with increasing slope while location indicated cor-
relations with population pressure on forests and carbon stock
levels. Summary coefficients for the top-ranked AGC model
are presented in Table 3.
The same top three models (forest type, forest type and slope,
and forest type and location) were the top-ranked candidate
models and had similar support in terms of model weight for CC
(Table 4) compared with AGC (Table 2). Mean CC stocks were:
primary hill 215.4 Mg C ha
(95% CI: 194.8–238.2, n¼48),
primary plain 128.8 Mg C ha
(100.5–165.1, n¼10) and
secondary hill forests 130.4 Mg C ha
(106.6–159.5, n¼17)
(Fig. 2). Summary coefficients for the top-ranked CC model are
presented in Table 5.
Factors affecting carbon stock variation
The top model describing AGC as being best explained in terms
of different forest types suggests the presence of three forest
strata: primary hill, primary plain and secondary hill forests
(Table 2). The mean carbon stock in AGB for primary hill forest
(165.0 Mg C ha
) estimated in this study was within a similar
range (105–180 Mg C ha
) reported from other primary forests
in PNG (Edwards and Grubb 1977;Bryan et al. 2010a;Fox et al.
2010;FPCD and IGES 2013;Vincent et al. 2015) and the
tropical forests (Aiba and Kitayama 1999;Chave et al. 2003;
IPCC 2006;Butt et al. 2015)(Table 6). However, one study in
primary lowland forest of PNG has reported a higher above-
ground carbon stock range of 207.5–296 Mg C ha
Table 2. Twelve candidate models selected for determining appropriate AGC stock level for trees $5 cm DBH in primary hill, secondary hill and
primary plain forests ranked by AICc
Key: K, number of parameters; AICc, Akaike’s information criterion with a small sample correction; DAICc, difference between candidate model and top
model in terms of AICc; LL, maximised value of the log-likelihood function (LL)
Rank Candidate model K AICc DAICc Model weight LL
1 AGC ,forest type 4 52.55 0 0.64 30.56
2 AGC ,forest type þslope 5 50.51 2.05 0.23 30.69
3 AGC ,forest type þlocation 12 49.34 3.21 0.13 39.19
4 AGC ,forest class 3 34.57 17.98 0 20.45
5 AGC ,forest class þlocation 11 34.46 18.10 0 30.32
6 AGC ,slope 3 33.60 18.96 0 19.97
7 AGC ,forest class þslope 4 33.56 18.99 0 21.07
8 AGC ,simple intercept 2 30.75 21.80 0 17.46
9 AGC ,elevation 3 30.33 22.22 0 18.33
10 AGC ,slope þelevation 5 29.94 22.61 0 20.41
11 AGC ,location 10 28.04 24.51 0 25.74
12 AGC ,slope þlocation 19 23.64 28.91 0 37.73
310 Pacific Conservation Biology A. Arihafa et al.
destructive sampling; however, the data are based on sampling
only 37 trees of variable size from five dominant species (Abe
2007). Several other tropical studies have also reported higher
carbon estimates (210–324 Mg C ha
) from destructive sam-
pling (e.g. Rai and Proctor 1986). Differences in setting mini-
mum limits in measuring DBH of trees could affect variation in
carbon estimates among studies. For example, the minimum
limit of $5 cm DBH of trees in our study could have produced
higher carbon estimates compared with most other studies in
PNG with minimum limits of $10 cm DBH (Table 6). The mean
AGC estimated in secondary hill forests in our study was higher
(99.7 Mg C ha
) than that of Fox et al. (2010) and Bryan et al.
(2010a)determined from logged over secondary forests
(66.3 Mg C ha
and 82.0 Mg C ha
respectively) in PNG
(Table 6). Successional stages of forest or severity of forest
fragmentation will affect carbon stock levels; thus sampling
only older anthropogenic secondary forests ($40 years old), as
done in this study, could have resulted in higher carbon stock
estimates compared with other logged over secondary forests.
Our mean estimates for AGC for primary plain forests
(100.9 Mg C ha
) and secondary hill forest (99.7 Mg C ha
were practically indistinguishable; from this, we infer that there
are only two categories of forest type on Manus: primary hill
forest and other forests.
A potential secondary factor identified from our modelling
process was forest type and slope (23% model support), which
500 AGC
Primary hill forest
Primary plain forest
Primary plain forest
Secondary hill forest
Secondary hill forest
Primary hill forest
Forest type
Mg C ha1
Fig. 2. Estimates of above-ground carbon (AGC) and combined carbon (CC) stocks across three
different forest types on Manus Island drawn from the top-ranked candidate models. Black dots represent
average values, error bars represent 95% confidence intervals, and grey dots represent range.
Table 3. Summary coefficients for the top-ranked AGC model
Forest type Log
(estimate) Log(s.e.) Mean carbon (Mg C ha
) (95% CI)
Primary hill forest (intercept) 2.22 0.02 165.0 (148.3–183.7)
Primary plain forest 0.21 0.06 100.9 (100.5–165.1)
Secondary hill forest 0.22 0.05 99.7 (106.6–159.5)
Carbon stock in tropical rainforest on Manus Island Pacific Conservation Biology 311
Table 4. Twelve candidate models selected for determining CC stock level for trees $5 cm DBH in primary hill, secondary hill and primary plain
forests ranked by AICc
Key: K, number of parameters; AICc, Akaike’s information criterion with a small sample correction; DAICc, difference between candidate model and top
model in terms of AICc; LL, maximised value of the log-likelihood function (LL)
Rank Model selection K AICc DAICc Model weight LL
1CC,forest type 4 58.18 0 0.69 33.37
2CC,forest type þslope 5 56.17 2.01 0.25 33.52
3CC,forest type þlocation 12 53.18 5.00 0.06 41.10
4CC,forest class 3 38.97 19.21 0 22.65
5CC,forest class þslope 4 38.06 20.12 0 23.31
6CC,forest class þlocation 11 37.66 20.52 0 31.92
7CC,slope 3 37.54 20.64 0 21.94
8CC,elevation 3 34.18 24.00 0 20.26
9CC,slope þelevation 5 34.15 24.03 0 22.51
10 CC ,simple intercept 2 34.00 24.18 0 19.08
11 CC ,location 10 30.45 27.73 0 26.94
12 CC ,slope þlocation 19 27.13 31.05 0 39.47
Table 5. Summary coefficients for the top-ranked CC model
Forest type Log
(estimate) Log(s.e.) Mean carbon (Mg C ha
) (95% CI)
Primary hill forest (intercept) 2.33 0.02 215.4 (194.8–238.2)
Primary plain forest 0.22 0.06 128.8 (194.8–238.2)
Secondary hill forest 0.22 0.04 130.4 (106.6–159.5)
Table 6. Average above-ground carbon stock (Mg C ha
) estimated for PNG forest trees compared with some regional (Asia) and biome averages
for tropical rainforests
Dead wood consists of standing dead and fallen trees. Estimations of AGB (in kg or t ha
) have been converted to AGC (Mg C ha
) using a conversion factor
of 0.5. Logged forests are all selectively logged
Forest type Location AGC
(Mg C ha
Dead wood
(Mg C ha
(Mg C ha
Tree definition Citation
Lowland primary hill forest PNG 165.0 14.5 47.5 $5 cm DBH This study
Lowland secondary hill forest PNG 99.7 6.1 30.4 $5 cm DBH This study
Lowland primary plain forest PNG 100.9 10.8 1.3 $5 cm DBH This study
Lowland primary forest PNG 120.1 7.6 n.a. $5 cm DBH FPCD and IGES (2013)
Lowland primary forest PNG 106.3 10.3 n.a. $10 cm DBH Fox et al. (2010)
Primary lower montane forest PNG 141.1 14.1 n.a. $10 cm DBH Fox et al. (2010)
Lowland selectively logged forest PNG 66.3 16.6 n.a. $10 cm DBH Fox et al. (2010)
Lower montane selectively logged PNG 58.8 14.7 n.a. $10 cm DBH Fox et al. (2010)
Lowland primary forest PNG 111.4 n.a. n.a. Extrapolated Bryan et al. (2010a)
Lowland selectively logged forest PNG 82.0 n.a. n.a. Extrapolated Bryan et al. (2010a)
Lowland primary forest PNG 207.5–296 n.a. n.a. $5 cm DBH Abe (2007)
Primary mid-montane forest PNG 147.5 n.a. 20 $10 cm DBH Edwards and Grubb (1977)
Lowland primary forest PNG 105.4 n.a. n.a. .1 cm DBH Vincent et al. (2015)
Dipterocarp forest SE Asia 243 n.a. n.a. $10 cm DBH Yamakura et al. (1986)
Dipterocarp forest SE Asia 256 n.a. n.a. $10 cm DBH Aiba and Kitayama (1999)
Wet tropical forest Africa 202 n.a. n.a. $10 cm DBH Lewis et al. (2009)
Moist tropical forest Guyana 153 n.a. n.a. $10 cm DBH Butt et al. (2015)
Neotropical forest Panama 140 n.a. n.a. $1 cm DBH Chave et al. (2003)
Lowland primary forest India 210–324 n.a. n.a. $5 cm DBH Rai and Proctor (1986)
Lowland forest Tropical 180 n.a. n.a. Default value IPCC (2006)
312 Pacific Conservation Biology A. Arihafa et al.
showed that AGC and combined carbon values increased with
increasing slope. This relationship most likely results from
forests on steeper slopes being less accessible to humans than
those on lower slopes, and they are thereby less disturbed by
anthropogenic activities (Gibbs et al. 2007). Trees in the uphill
primary forests were generally taller and denser than the primary
plain forests on the flatlands (sensu Butt et al. 2015). Addition-
ally, diverse environmental factors (e.g. water stress and the
effect of sunlight on aspect) could also affect carbon stock
levels, making trees on well raised slopes with good drainage
accumulate more biomass than trees on lower slopes or from the
plains (Aiba and Kitayama 1999). Another potential model for
consideration was forest type and location (13% model support).
We suspect that location in this study is associated with human
accessibility to forests and the land-use history (Gibbs et al.
2007), as secondary forests were located closer to existing
villages and deserted settlements. Indeed, we know that one of
our sampling sites near Pelipowai village (Karowan forest)
(Fig. 1) had lower carbon stock levels compared with other sites
as a result of the area being a colonial–post settlement in the
1940–50s, which resulted in much of the forests being cleared
during that period.
Use of established AGB ratios in estimating
other carbon pools
Established AGB ratios for trees are often used to estimate other
above-ground carbon pools (e.g. non-trees including herbs,
climber and epiphytes, and litter and dead wood including fallen
flowers, fruits, leaves and small branches) and below-ground
carbon. For example, Fox et al. (2010) estimated from the litera-
ture in an undisturbed lowland forest in PNG that unmeasured
components of non-trees ,10 cm DBH,litter and deadwood (both
standing and fallen) accounted for 5%, 2.5% and 10% of AGB of
trees $10 cm DBH respectively. Similarly, below-ground bio-
mass is determined through the application of a regression model
following Walker et al. (2012). Thus, BGC of root biomass in this
study was estimated from established root to shoot ratios as either
23% or 20% of AGC if AGB carbon was .62.5 Mg C ha
62.5 Mg C ha
respectively (Mokany et al. 2006).
Implications for community participation
in estimating carbon stocks
The last remaining forests of central Manus Island are under
threat from logging activities in a manner similar to other low-
land areas of PNG (Shearman et al. 2009;Bryan et al. 2010b). It
is necessary to develop and test standardised methodologies for
quantifying carbon stocks to inform management decisions and
policies on REDDþand other forest carbon mitigation schemes
in PNG (Shearman et al. 2009). Yet the systematic sampling of
the proposed Manus REDDþproject area (E71 737 ha) in this
study would have been impractical due to issues related to:
remoteness of the inland parts of the area, identification of
landowners for consent and clan land disputes. This study was
carried out on traditional lands, and sampling was carried out
only in consented areas. With this situation, we added value to
the project by engaging local landowners as locally trained field
assistants and community facilitators in our forest inventory and
set the stage for future REDDþwork, which could be of a more
extensive nature. Participation of trained local assistants in
estimations of AGC in biomass can be efficient compared with
other methods of quantifying forest carbon stocks (e.g. remotely
sensed data: Butt et al. 2015). On this basis, we advocate such
community participatory approaches to others working in PNG
on traditional lands not only as a means for social inclusion and
capacity building but to foster trust within the community. Our
findings regarding the forests of Manus Island are consistent
with pre-existing studies, which demonstrate that PNG forests
hold ,100–200 Mg C ha
of above-ground tree carbon stock
(Table 6). With a better understanding of carbon assessments
based on accurate estimates, carbon accounting for REDDþ
projects would be feasible.
This study was funded by the Australian Department of Foreign Affairs and
Trade as part of an activity under the climate change mitigation project grant
agreement no. 63060. We thank the WCSPNG Program for making this
study possible. We are grateful to Nathan Whitmore for help with data
analysis, and Nathan together with Richard Cuthbert for reviewing manu-
script drafts and for critical comments. We appreciate the efforts of
WCSPNG Manus staff Daniel Charles, June Polomon and Julian Benjamin
for logistical support and the community facilitators Matawai Pondrilei,
Benson Lian, Misu Nick, Bryan Ausung, Pondrilei Sori and the late David
Posa who assisted in carrying out the field work. We are grateful to the
Manus Provincial Government and POBUMA and PNKA Local Level
Governments in supporting this project. Finally we thank the community
leaders, ward councillors and the people of Tulu 1, Saha, Djekal, Piri and
Pelipowai villages of Manus Provincefor allowing us to work in their forests.
Abe, H. (2007). Forest management impacts on growth, diversity and
nutrient cycling of lowland tropical rainforest and plantations, Papua
New Guinea. Ph.D. thesis. University of Western Australia, Perth.
Aiba, S., and Kitayama, K. (1999). Structure, composition and species
diversity in an altitude–substrate matrix of rain forest tree communities
on Mount Kinabalu, Borneo. Plant Ecology 140, 139–157. doi:10.1023/
Bryan, J., Shearman, P., Ash, J., and Kirkpatrick, J. B. (2010a). Impact of
logging on aboveground biomass stocks in lowland rainforest, Papua
New Guinea. Ecological Applications 20, 2096–2103. doi:10.1890/09-
Bryan, J., Shearman, P., Ash, J., and Kirkpatrick, J. B. (2010b). Estimating
rainforest biomass stocks and carbon loss from deforestation and
degradation in Papua New Guinea 1972–2002: best estimates, uncer-
tainties and research needs. Journal of Environmental Management 91,
995–1001. doi:10.1016/J.JENVMAN.2009.12.006
Burnham, K. P., and Anderson, D. R. (2002). ‘Model Election and Multi-
model Inference: A Practical Information-theoretic Approach.’ 2nd edn.
(Springer-Verlag: New York.)
Butt, N., Epps, K., Overman, H., Iwamura, T., and Fragoso, J. M. V. (2015).
Assessing carbon stocks using indigenous peoples’ field measurements
in Amazonian Guyana. Forest Ecology and Management 338, 191–199.
Cayuela, L., Benayas, J. M. R., and Echeverria, C. (2006). Clearance and
fragmentation of tropical montane forests in the Highlands of Chiapas,
Mexico (1975–2000). Forest Ecology and Management 226, 208–218.
Chave, J.,Condit, R., Lao, S.,Caspersen, J. P., Foster, R. B., andHubbell, S. P.
(2003). Spatial and temporal variation of biomass in a tropical forest:
results from a large census plot in Panama. Journal of Ecology 91,
240–252. doi:10.1046/J.1365-2745.2003.00757.X
Carbon stock in tropical rainforest on Manus Island Pacific Conservation Biology 313
Chave, J., Andalo, C., Brown, S., Cairns, M. A., Chambers, J. Q., Eamus, D.,
Falster, H., Fromard, F.,Higuchi, N., Kira, T., Lescure, J. P.,Nelson, B. W.,
Ogawa, H., Puig, H., Riacra, B., and Yamakura, T. (2005). Tree
allometry and improved estimation of carbon stocks and balance in
tropical forests. Oecologia 145, 87–99. doi:10.1007/S00442-005-0100-X
Croft, J. R. (1983). An historical survey of botanical exploration in the
Admiralty Islands, Manus Province, Papua New Guinea. Science in
New Guinea 10, 1–15.
Edwards, P. J., and Grubb, P. J. (1977). Studies of mineral cycling in a
montane rain forest in New Guinea; the distribution of organic matter in
the vegetation and soil. Journal of Ecology 65, 943–969. doi:10.2307/
Ferraz, S. F. D., Vettorazzi, C. A., Theobald, D. M., and Ballester, M. V. R.
(2005). Landscape dynamics of Amazonian deforestation between 1984
and 2002 in central Rondonia, Brazil: assessment and future scenarios.
Forest Ecology and Management 204, 67–83.
Fox, J. C., Yosi, C. K., Nimiago, P., Oavika, F., Pokana, J. N., Lavong, K.,
and Kennan, R. J. (2010). Assessment of aboveground carbon in primary
and selectively harvested tropical forest in Papua New Guinea. Biotro-
pica 42, 410–419. doi:10.1111/J.1744-7429.2009.00617.X
FPCD and IGES (2013). FPCD-IGES community-based forest monitoring
project, April 2013 Report. (Eds H. Scheyvens, L. Poruschi, Y. A. Bun,
T. Fujisaki and R. Avtar.) (FPCD & IGES: Papua New Guinea.)
Gibbs, H. K., Brown, S., Niles, J. O., and Foley, J. A. (2007). Monitoring and
estimating tropical forest carbon stocks: making REDDþa reality.
Environmental Research Letters 2, 045023. doi:10.1088/1748-9326/2/
Government of PNG (GPNG) (2007). The national land development
taskforce report: land administration, land dispute settlement, and
customary land development. National Research Institute Monograph
39, Port Moresby.
Harris, N. L., Brown, S., Hagen, S. C., Saatchi, S. S., Petrova, S., Salas, W.,
Hansen, M. C., Potapov, P. V., and Lotsch, A. (2012). Baseline map of
carbon emissions from deforestation in tropical regions. Science 336,
1573–1576. doi:10.1126/SCIENCE.1217962
IPCC (2006). IPCC Guidelines for National Greenhouse Gas Inventories.
(Eds H. S. Eggleston, L. Buendia, K. Miwa, T. Ngara and K. Tanabe.)
(National Greenhouse Gas Inventories Programme, IGES: Japan.)
Kisokau, K. (1974). Fauna survey of Manus District field report. Science in
New Guinea 2, 172–175.
Laurance, W. F. (2007). A new initiative to use carbon trading for tropical
forest conservation. Biotropica 39, 20–24. doi:10.1111/J.1744-7429.
Laurance, W. F., Kakul, T., Keenan, R. J., Sayer, J., Passingan, S., Clements,
G. R., Villegas, F., and Sodhi, N. S. (2011). Predatory corporations,
failing governance and the fate of forests in Papua New Guinea.
Conservation Letters 4, 95–100. doi:10.1111/J.1755-263X.2010.
Lewis, S. L., Lopez-Gonzalez, G., Sonke, B., and Affum-Baffoe, K. (2009).
Increasing carbon storage in intact African tropical forests. Nature 457,
1003–1006. doi:10.1038/NATURE07771
Malhi, Y., and Grace, J. (2000). Tropical forests and atmospheric carbon
dioxide. Trends in Ecology & Evolution 15, 332–337. doi:10.1016/
Mokany, K., Raison, R. J., and Prokushkin, A. S. (2006). Critical analysis
of root : shoot ratios in terrestrial biomes. Global Change Biology 12,
84–96. doi:10.1111/J.1365-2486.2005.001043.X
Paijmans, K. (1976). ‘New Guinea Vegetation.’ (CSIRO & Australian
National University Press: Canberra.)
Pedroni, L. (2007). Methodology for sub-national project activities reducing
greenhouse gas emissions from deforestation and forest degradation.
BioCarbon Fund, World Bank, Washington DC.
Phillips, O. L., and Gentry, A. H. (1994). Increasing turnover through time in
tropical forests. Science 263, 954–958. doi:10.1126/SCIENCE.263.
R Development Core Team (2013). ‘R: A Language and Environment for
Statistical Computing.’ (R Foundation for Statistical Computing:
Vienna, Austria.)
Rai, S. N., and Proctor, J. (1986). Ecological studies on four rainforests in
Karnataka, India: environment, structure, floristics and biomass. Journal
of Ecology 74, 439–454. doi:10.2307/2260266
Saulei, S. M. (1990). Forest research and development in Papua New
Guinea. Ambio 19, 379–382.
Shearman, P., Bryan, J., Ash, J., Hunnam, P., Mackey, B., and Lokes, B.
(2008). The state of the forests of Papua New Guinea. University of
Papua New Guinea, Port Moresby.
Shearman, P. L., Ash, J., Mackey, B., Bryan, J. E., and Lokes, B. (2009).
Forest conversion and degradation in Papua New Guinea 1972–2002.
Biotropica 41, 379–390. doi:10.1111/J.1744-7429.2009.00495.X
Van Laake, P. E., and Sanchez-Azofeifa, G. A. (2004). Focus on deforesta-
tion: zooming in on hot spots in highly fragmented ecosystems in Costa
Rica. Agriculture, Ecosystems & Environment 102, 3–15. doi:10.1016/
Vincent, J. B., Henning, B., Saulei, S., Sosanika, G., and Weiblen, G. (2015).
Forest carbon in lowland Papua New Guinea: local variation and the
importance of small trees. Austral Ecology 40, 151–159. doi:10.1111/
Walker, S., Pearson, T., Harris, N. L., Petrova, S., Swails, E., Netzer, M.,
Goslee, K. M., and Brown, S. (2012). ‘Standard Operating Procedures
for Terrestrial Carbon Measurement, Version 2012.’ (Winrock
Yamakura, T., Hagihara, A., Sukardjo, S., and Ogawa, H. (1986). Above-
ground biomass of tropical rain forest stands in Indonesian Borneo.
Vegetatio 68, 71–82.
314 Pacific Conservation Biology A. Arihafa et al.
... The large decrease in the area of primary forest, as reported in these official figures, is of key importance for the context of REDD+ in PNG because these areas will have held considerably higher carbon stocks than the regenerating forest areas that have replaced them. This is the case even for relatively old (>40 years) secondary forest which contain on average 60% of the above ground biomass (ABG) of comparable primary forest areas (see Table 6 of Arihafa et al. 2015 for this and other studies). These reported decreases in the extent of primary forest (and their replacement with regenerating forest) will have resulted in a substantial reduction in forest biomass (degradation) in PNG's forest over the last 25 years. ...
... According to Bryan and Shearman (2015) this forest area is declining at a rate of around 0.5% per year, although this is contradicted by FAO (2015) figures for PNG. While the FAO (2015b) report no major change in forest area, their data does indicate a decrease of 2.3% per year in the area of primary forest (see Annex 4) which will have caused an overall degradation of forest biomass due to the lower carbon biomass in secondary forests (Arihafa et al. 2015) that have replaced them. Bryan and Shearman (2015) report that deforestation and forest degradation are occurring as a result of large-scale industrial logging, large-scale clearance to produce agricultural commodities, and small-scale clearance for gardens and subsistence agriculture ( Figure 8). ...
Full-text available
Efforts to incentivize the reduction of carbon emissions from deforestation and forest degradation require accurate carbon accounting. The extensive tropical forest of Papua New Guinea (PNG) is a target for such efforts and yet local carbon estimates are few. Previous estimates, based on models of neotropical vegetation applied to PNG forest plots, did not consider such factors as the unique species composition of New Guinea vegetation, local variation in forest biomass, or the contribution of small trees. We analysed all trees >1 cm in diameter at breast height (DBH) in Melanesia's largest forest plot (Wanang) to assess local spatial variation and the role of small trees in carbon storage. Above-ground living biomass (AGLB) of trees averaged 210.72 Mg ha−1 at Wanang. Carbon storage at Wanang was somewhat lower than in other lowland tropical forests, whereas local variation among 1-ha subplots and the contribution of small trees to total AGLB were substantially higher. We speculate that these differences may be attributed to the dynamics of Wanang forest where erosion of a recently uplifted and unstable terrain appears to be a major source of natural disturbance. These findings emphasize the need for locally calibrated forest carbon estimates if accurate landscape level valuation and monetization of carbon is to be achieved. Such estimates aim to situate PNG forests in the global carbon context and provide baseline information needed to improve the accuracy of PNG carbon monitoring schemes.
Full-text available
Tree turnover rates were assessed at 40 tropical forest sites. Averaged across inventoried forests, turnover, as measured by tree mortality and recruitment, has increased since the 1950's, with an apparent pantropical acceleration since 1980. Among 22 mature forest sites with two or more inventory periods, forest turnover also increased. The trend in forest dynamics may have profound effects on biological diversity.
Full-text available
Reducing carbon emissions from deforestation and degradation in developing countries is of central importance in efforts to combat climate change. Key scientific challenges must be addressed to prevent any policy roadblocks. Foremost among the challenges is quantifying nations' carbon emissions from deforestation and forest degradation, which requires information on forest clearing and carbon storage. Here we review a range of methods available to estimate national-level forest carbon stocks in developing countries. While there are no practical methods to directly measure all forest carbon stocks across a country, both ground-based and remote-sensing measurements of forest attributes can be converted into estimates of national carbon stocks using allometric relationships. Here we synthesize, map and update prominent forest biomass carbon databases to create the first complete set of national-level forest carbon stock estimates. These forest carbon estimates expand on the default values recommended by the Intergovernmental Panel on Climate Change's National Greenhouse Gas Inventory Guidelines and provide a range of globally consistent estimates.
(1) As a preliminary to studies on mineral cycling in a tropical montane system, where show release of minerals from organic matter may be an important feature of the environment for plants, estimates have been made of the amounts of organic matter in a Lower Montane Rain forest and its soil at 6°S and altitude c. 2500 m in New Guinea. (2) A detailed destructive study was made of a plot 20 × 20 m, and derived estimates of biomass were made for various other plots using a regression of dry weight on the square of trunk diameter at breast height. On a 0.24 ha plot considered representative of the forest at large the total biomass was estimated at c. 350 t ha-1, made up of c. 40 t ha-1 in roots, 295 t ha-1 in above-ground parts of trees of gbh ⩽ 30 cm, c. 4 t ha-1 in climbers and scramblers, c. 2 t ha-1 in epiphytes, c. 2 t ha-1 in the 0-1 m layer and c. 7 t ha-1 in the other plants. (3) Determinations of specific leaf area were made for seventeen tree species (range 37-127 cm2 g-1, mean 58 cm2 g-1), and for four common climbers (45-81 cm2 g-1) and a scrambling bamboo (216 cm2 g-1). The leaf area index was estimated as. c 5.5 m2 m-2 on the plot of 20 × 20 m. (4) The ash contents of 140 representative plant samples were determined; they ranged from c. 10% for the 0.1 m layer to 2% in the trunks. The organic matter content of the soil was taken to be twice the organic carbon content, which was c. 19% dry weight at 0-2 cm and c. 13% at 2-10 cm, falling to c. 9% at 100 cm depth. The organic matter content of the plants on the 0.24 ha plot was estimated at c. 340 t ha-1, and that of the soil at c. 1200 t ha-1 1200 t ha-1. About 1 t ha-1 of soil (4% ash) had accumulated in the crowns of the larger trees, formed largely from the remains of epiphytes. (5) Comparisons are made with other tropical forests, and the inadequacy of sampling in most studies is emphasized. (6) The ratio of organic matter in the soil to that in the plants is 3-4:1 in the New Guinean forest. This may be compared with probable values of 0.7-4.0:1 in other Lower Montane forests and measured values within the range 0.2-0.6:1 in Lowland forests. The very large amount of organic matter in the soil of the New Guinean forest is tentatively related to fixation by amorphous clays.
(1) The environment, structure and floristics of four evergreen rainforest sites at 575-800 m altitude in Karnataka, southern India, are described. (2) The mean annual rainfall on the four sites ranges from 5310 to 7670 mm. Most of the rainfall occurs from June to September and there is an intense dry period from December to April. All the sites occur on oxisols overlying hornblendic rocks. (3) The forests at three of the sites are species-rich with an important contribution from the Dipterocarpaceae but one site is unusual and has an almost monospecific dominance by Poeciloneuron indicum (Guttiferae). (4) The total aboveground tree biomass was 420-649 t ha-1. The root fraction was 13.9-20.2 t ha-1. (5) Girth increment data over 35 years were available for one site and these were used with biomass estimates to calculate the approximate mean annual increase of above-ground and root (⩾ 5-cm girth) biomass in the four sites. These were 6.4-11.1 t ha-1 for aboveground material and 0.2-0.4 t ha-1 for roots. (6) For small trees (⩽ 5 cm dbh) and herbs, biomass was estimated by destructive sampling in one plot. The combined above-ground biomass of these fractions was 7.2 t ha-1
Accurate estimations of carbon stocks across large tracts of tropical forests are key for participation in programs promoting avoided deforestation and carbon sequestration, such as the UN REDD+ framework. Trained local technicians can provide such data, and this, combined with satellite imagery, allows robust carbon stock estimation across vegetation classes and large areas. In the first comprehensive survey in Guyana conducted by indigenous people, ground data from 21 study sites in the Rupununi region were used to estimate above ground tree carbon density across a diversity of ecosystems and land use types. Carbon stocks varied between village sites from 1 Tg to 22.7 Tg, and these amounts were related to stem density and diameter. This variation was correlated with vegetation type across the region, with savannas holding on average 14 MgC ha−1 and forests 153 MgC ha−1. The results indicated that previous estimates based on remotely sensed data for this area may be inaccurate (under estimations). There were also differences in carbon densities between village sites and uninhabited control areas, which are presumably driven by community use. Recruiting local technicians for field work allowed (a) large amounts of ground data to be collected for a wide region otherwise hard to access, and (b) ensured that local people were directly involved in Guyana’s Low Carbon Development Strategy as part of REDD+. This is the first such comprehensive survey of carbon stocks, carbon density and vegetation types over a large area in Guyana, one of the first countries to develop such a program. The potential inclusion of forests held by indigenous peoples in REDD+ programs is a global issue: we clearly show that indigenous people are capable of assessing and monitoring carbon on their lands.
I describe a new initiative, led by a coalition of developing nations, to devise a viable mechanism for using carbon trading to protect old-growth tropical forests. I highlight some of the practical and political hurdles involved in forest-carbon trading, and explain why this initiative is rapidly gaining broad-based political support.