Content uploaded by Philip Fearnside
Author content
All content in this area was uploaded by Philip Fearnside on Oct 15, 2017
Content may be subject to copyright.
The text that follows is a PREPRINT.
Please cite as:
Feldpausch, T. R., Lloyd, J., Lewis, S. L., Brienen, R. J. W., Gloor, E.,
Monteagudo Mendoza, A., Lopez-Gonzalez, G., Banin, L., Abu
Salim, K., Affum-Baffoe, K., Alexiades, M., Almeida, S., Amaral,
I., Andrade, A., Aragão, L. E. O. C., Araujo Murakami, A., Arets, E.
J. M. M., Arroyo, L., G. A. Aymard C., Baker, T. R., Bánki, O. S.,
Berry, N. J., Cardozo, N., Chave, J., Comiskey, J. A., Dávila, E. A.,
de Oliveira, A., DiFiore, A., Djagbletey, G., Domingues, T. F.,
Erwin, T. L., Fearnside, P. M., França, M. B., Freitas, M.A.,
Higuchi, N., E. Honorio C., Iida, Y., Jiménez, E., Kassim, A. R.,
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., Mitchard, E. T. A., Nelson, B. W., Nilus, R.,
Nogueira, E. M., Parada, A., Peh, K. S.-H., Pena Cruz, A., Peñuela,
M. C., Pitman, N. C. A., Prieto, A., Quesada, C.A., Ramírez, F.,
Ramírez-Angulo, H., Reitsma, J. M., Rudas, A., Saiz, G., Salomão,
R. P., Schwarz, M., Silva, N., Silva-Espejo, J. E., Silveira, M.,
Sonké, B., Stropp, J., Taedoumg, H. E., Tan, S., ter Steege, H.,
Terborgh, J., Torello-Raventos, M., van der Heijden, G. M. F.,
Vásquez, R., Vilanova, E., Vos, V., White, L., Wilcock, S., Woell,
H. & Phillips, O. L. 2012: Tree height integrated into pan-tropical
forest biomass estimates, Biogeosciences (in press: accepted 23-07-
12)
Copyright: Copernicus Publications on behalf of the European Geosciences Union.
The original publication will be available at: www.biogeosciences.net
Note: This replaces: Biogeosciences Discussions, 9, 2567-2622,
doi:10.5194/bgd-9-2567-2012. Available at:
http://www.biogeosciences-discuss.net/9/2567/2012/bgd-9-
2567-2012.html
Manuscript prepared for Biogeosciences Discuss.
with version 3.5 of the L
A
T
E
X class copernicus discussions.cls.
Date: 12 July 2012
Tree height integrated into pan-tropical
forest biomass estimates
T. R. Feldpausch1, J. Lloyd1,2, S. L. Lewis1,3, R. J. W. Brienen1, E. Gloor1,
A. Monteagudo Mendoza4, G. Lopez-Gonzalez1, L. Banin1,5, K. Abu Salim6,
K. Affum-Baffoe7, M. Alexiades8, S. Almeida9,†, I. Amaral10, A. Andrade10,
L. E. O. C. Arag˜
ao11, A. Araujo Murakami12 , E. J. M. M. Arets13, L. Arroyo12,
G. A. Aymard C.14, T. R. Baker1, O. S. B ´
anki15, N. J. Berry16 , N. Cardozo17,
J. Chave18, J. A. Comiskey19, E. A. D ´
avila20, A. de Oliveira10 , A. DiFiore21,
G. Djagbletey22, T. F. Domingues23 , T. L. Erwin24, P. M. Fearnside10,
M. B. Franc¸a10, M.A. Freitas9, N. Higuchi10, E. Honorio C.1, Y. Iida25, E. Jim´
enez26,
A. R. Kassim27, T. J. Killeen28, W. F. Laurance29, J. C. Lovett30, Y. Malhi31 ,
B. S. Marimon32, B. H. Marimon-Junior32, E. Lenza32, A. R. Marshall33,34,
C. Mendoza35, D. J. Metcalfe36, E. T. A. Mitchard37, D. A. Neill38, B. W. Nelson39,
R. Nilus40, E. M. Nogueira10 , A. Parada12, K. S.-H. Peh41, A. Pena Cruz42,
M. C. Pe˜
nuela26, N. C. A. Pitman43 , A. Prieto44, C.A. Quesada10 , F. Ram´ırez17,
H. Ram´ırez-Angulo45, J. M. Reitsma46, A. Rudas47, G. Saiz48 , R. P. Salom ˜
ao9,
M. Schwarz1, N. Silva49, J. E. Silva-Espejo50, M. Silveira51, B. Sonk ´
e52,
J. Stropp53, H. E. Taedoumg52, S. Tan54, H. ter Steege55, J. Terborgh43,
M. Torello-Raventos2, G. M. F. van der Heijden56,57, R. V ´
asquez42, E. Vilanova58,
V. Vos59,60, L. White61,62,63, S. Wilcock1, H. Woell64, and O. L. Phillips1
1School of Geography, University of Leeds, Leeds, LS2 9JT, UK
2School of Earth and Environmental Science, James Cook University, Cairns, Qld 4870,
Australia
3Department of Geography, Univ. College London, UK
1
4RAINFOR/Jard´
ın Bot´
anico de Missouri, Peru
5School of Environmental Sciences, University of Ulster, Cromore Road, Coleraine, BT52 1SA,
UK
6Biology Programme, Faculty of Science, Universiti Brunei Darussalam, Tungku Link Road
BE1410, Brunei Darussalam
7Resource Management Support Centre, Forestry Commission of Ghana, P.O. Box 1457,
Kumasi, Ghana
8New York Botanical Garden, New York City, New York 10458, USA
9Museu Paraense Emilio Goeldi, Av. Magalh˜
aes Barata, 376 – S˜
ao Braz, CEP: 66040-170,
Bel´
em, PA, Brazil
10National Institute for Research in Amazonia (INPA), C.P. 478, Manaus, Amazonas, CEP
69011-970, Brazil
11Geography, College of Life and Environmental Sciences, University of Exeter, Rennes Drive,
Exeter, EX4 4RJ, UK
12Museo de Historia Natural Noel Kempff Mercado, Universidad Autonoma Gabriel Rene
Moreno, Casilla 2489, Av. Irala 565, Santa Cruz, Bolivia
13Centre for Ecosystem Studies, Alterra, Wageningen University and Research Centre, P.O.
Box 47, 6700 AA Wageningen, The Netherlands
14UNELLEZ-Guanare, Programa de Ciencias del Agro y el Mar, Herbario Universitario
(PORT), Mesa de Cavacas, Estado Portuguesa 3350, Venezuela
15IBED, University of Amsterdam, POSTBUS 94248, 1090 GE Amsterdam, The Netherlands
16School of GeoSciences, University of Edinburgh, Edinburgh, EH9 3JN, UK
17Universidad Nacional de la Amazon´
ıa Peruana, Iquitos, Loreto, Per´
u
18Universit ´
e Paul Sabatier, Laboratoire EDB, bˆ
atiment 4R3, 31062 Toulouse, France
19Mid-Atlantic Network, Inventory and Monitoring Program, National Park Service, 120
Chatham Lane, Fredericksburg, VA 22405, USA
20Jardin Botanico de Medellin, Colombia
21Department of Anthropology, University of Texas at Austin, 1 University Station, SAC 5.150
Mailcode C3200, Austin, TX 78712, USA
22Ecosystem and Climate Change Division (ESCCD) Forestry Research Institute of Ghana
(FORIG), U.P. Box 63, KNUST-Kumasi, Ghana
23Instituto de Astronomia, Geof´
ısica e Ciˆ
encias Atmosf´
ericas – Universidade de S˜
ao Paulo,
05508-090, Brasil
2
24Department of Entomology, Smithsonian Institute, P.O. Box 37012, MRC 187, Washington,
DC 20013-7012, USA
25Graduate School of Environmental Science, Hokkaido University, Sapporo, 060-0810, Japan
26Universidad Nacional de Colombia, Kil ´
ometro 2 Via Tarapac ´
a, Leticia, Amazonas, Colombia
27Forest Research Institute Malaysia (FRIM), 52109 Kepong, Selangor Darul Ehsan, Malaysia
28Conservation International, 2011 Crystal Drive, Suite 500, Arlington, VA 22202, USA
29Centre for Tropical Environmental and Sustainability Science (TESS) and School of Marine
and Tropical Biology, James Cook University, Cairns, Queensland 4878, Australia
30CSTM, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
31Environmental Change Institute, School of Geography and the Environment, University of
Oxford, UK
32Universidade do Estado de Mato Grosso, Campus de Nova Xavantina, Caixa Postal 08,
CEP 78.690-000, Nova Xavantina, MT, Brazil
33CIRCLE, Environment Department, University of York, York, UK
34Flamingo Land Ltd., Kirby Misperton, YO17 6UX, UK
35FOMABO (Manejo Forestal en las Tierras Tropicales de Bolivia), Sacta, Bolivia
36CSIRO Ecosystem Sciences, Tropical forest Research Centre, P.O. Box 780, Atherton, QLD
4883, Australia
37School of GeoSciences, University of Edinburgh, Drummond St, Edinburgh, EH8 9XP, UK
38Universidad Estatal Amaz ´
onica, Facultad de Ingenier´
ıa Ambiental, Paso lateral km 2 1/2 via
Napo, Puyo, Pastaza, Ecuador
39National Institute for Research in Amazonia (INPA), Environmental Dynamics Department,
C.P. 478, Manaus, Amazonas, CEP 69011-970, Brazil
40Forest Research Centre, Sabah Forestry Department, Sandakan, 90715, Malaysia
41Department of Zoology, University of Cambridge, Downing Street, CB2 3EJ, UK
42Jard´
ın Bot´
anico de Missouri, Oxapampa, Pasco, Peru
43Center for Tropical Conservation, Duke University, Box 90381, Durham, NC 27708, USA
44Doctorado Instituto de Ciencias Naturales, Universidad Nacional de Colombia
45Universidad de Los Andes, Facultad de Ciencias Forestales y Ambientales, M´
erida,
Venezuela
46Bureau Waardenburg bv, P.O. Box 365, 4100 AJ Culemborg, The Netherlands
47Instituto de Ciencias Naturales, Universidad Nacional de Colombia, Colombia
48Karlsruhe Institute of Technology, Garmisch-Partenkirchen, Germany
3
49UFRA – Universidade Federal Rural da Amazˆ
onia, Brasil
50Universidad Nacional San Antonio Abad del Cusco, Av. de la Cultura No. 733. Cusco, Peru
51Universidade Federal do Acre, Rio Branco AC 69910-900, Brazil
52Department of Biology, University of Yaound ´
e I, P.O. Box 047, Yaound ´
e, Cameroon
53European Commission – DG Joint Research Centre, Institute for Environment and
Sustainability, Via Enrico Fermi 274, 21010 Ispra, Italy
54Sarawak Forestry Corporation, Kuching, Sarawak, Malaysia
55NCB Naturalis, PO Box, 2300 RA, Leiden, The Netherlands
56 University of Wisconsin-Milwaukee,Milwaukee, Wisconsin, Department of Biological
Sciences, PO Box 413, 53201, USA
57 Smithsonian Tropical Research Institute, Apartado 2072, Balboa, Republic of Panama
58Instituto de Investigaciones para el Desarrollo Forestal (INDEFOR), Universidad de Los
Andes, M´
erida, Venezuela
59PROMAB, Casilla 107, Riberalta, Beni, Bolivia
60Universidad Autonoma del Beni, Campus Universitario, Av. Ej ´
ercito Nacional, final, Riberalta,
Beni, Bolivia
61Agence National des Parcs Nationaux, Libreville, Gabon
62Institut de Recherche en Ecologie Tropicale (CENAREST), Gabon
63School of Natural Sciences, University of Stirling, UK
64Sommersbergseestr. 291, 8990 Bad Aussee, Austria
†deceased
Correspondence to: T. R. Feldpausch (t.r.feldpausch@leeds.ac.uk)
4
Abstract
Above-ground tropical tree biomass and carbon storage estimates commonly ignore
tree height (H). We estimate the effect of incorporating Hon forest biomass estimates
using 37 625 concomitant Hand diameter measurements (n= 327 plots) and 1816
harvested trees (n= 21 plots) tropics-wide to answer the following questions:
1. For trees of known biomass (from destructive harvests), which H-model form and
geographic scale (plot, region, and continent) most substantially reduces biomass
estimate uncertainty?
2. To what extent does the inclusion of Hestimates derived in (1) reduce uncertainty
in biomass estimates across 327 plots spanning four continents?
3. What effect does the inclusion of Hin biomass estimates have on plot- and
continental-scale forest biomass estimates?
The mean relative error in biomass estimates of destructively harvested trees was half
(mean 0.06) when including H, compared to excluding H(mean 0.13). The power- and
Weibull-Hasymptotic model provided the greatest reduction in uncertainty, with the re-
gional Weibull-Hmodel preferred because it reduces uncertainty in smaller-diameter
classes (≤40 cm D) that store about one-third of biomass per hectare in most forests.
Propagating the relationships from destructively harvested tree biomass to each of the
327 plots from across the tropics shows errors are reduced from 41.8 Mg ha−1(range
6.6 to 112.4) to 8.0 Mg ha−1(−2.5 to 23.0) when including H. For all plots, above-
ground live biomass was 52.2 (±30.8 Mg ha−1bootstrap 95% CI) or 13 % (±10 %
bootstrap 95 % CI) lower when including Hestimates, with the greatest relative re-
ductions in estimated biomass in forests of the Brazilian Shield, east Africa, and Aus-
tralia, and relatively little change in the Guyana Shield, central Africa and southeast
Asia. Appreciable different stand structure was observed among regions across the
tropical continents, whereby, some regions store significantly more biomass in small
5
diameter stems, which crucially affects height model selection to reduce uncertainty,
and biomass reductions due to H. After accounting for variation in H, total biomass
per hectare is greatest in Australia, the Guyana Shield, Asia, central and east Africa,
and lowest in east-central Amazonia, W. Africa, W. Amazonia, and the Brazilian Shield
(descending order). Thus, if tropical forests span 1668 million km2and store 285 Pg C
(estimate including H), then applying our regional relationships implies that carbon
storage is overestimated by 35 Pg,C ( ±10 bootstrap 95 % CI) if His ignored, assum-
ing that the sampled plots are an unbiased statistical representation of all tropical forest
in terms of biomass and height factors. Our results show that tree His an important
allometric factor that needs to be included in future forest biomass estimates to reduce
error in estimates of pantropical carbon stocks and emissions due to deforestation.
1 Introduction
Accurate estimates of tropical tree biomass are essential to determine geographic pat-
terns in carbon stocks, the magnitudes of fluxes due to land-use change, and to quan-
tify avoided carbon emissions via mechanisms such as REDD+ (Reducing Emissions
from Deforestation and forest Degradation). Global estimates of tree carbon in tropical
forests vary between 40 to 50 % of the total carbon in terrestrial vegetation (Watson
et al., 2000; Kindermann et al., 2008), indicating considerable uncertainty. Such un-
certainty is the consequence of linking individual tree measurements to large-scale
patterns of carbon distribution, as well as definition as to what constitutes “forest.”
The estimation of tree-, plot-, regional-, or global-level mass of tropical trees re-
quires first harvesting and weighing trees (Fittkau and Klinge, 1973), and subsequently
estimating biomass on a larger population by measuring tree stem diameter (D) and
converting Dto biomass based on allometric equations developed using the destruc-
tive harvest data (Brown et al., 1989; Overman et al., 1994; Ogawa et al., 1965).
Biomass can also be estimated using active (e.g. radar and from light detection and
ranging (LiDAR) remote sensing-based methods (e.g. Drake et al., 2002; Mitchard et
6
al., 2011, Morel et al. 2011; Saatchi et al. 2011). Calibration of remotely-sensed
biomass requires ground-based biomass estimates derived from stem diameter mea-
surements and allometric equations (either calibrated “on-site” or from the literature to
“ground-truth” data (e.g. Lucas et al., 2002; Mitchard et al., 2009)). Both ground- and
space-borne biomass estimates have uncertainty, and scaling from plots to regions
introduces additional uncertainty. For example, carbon stock estimates for Amazonia
based on spatial interpolations of direct measurements, relationships to climatic vari-
ables, and remote sensing data have an uncertainty of ±20 % (Saatchi et al. 2007;
Houghton 2010). It is therefore necessary to generate accurate allometric models that
reduce uncertainty in tree and plot-level estimates, in order that remote sensing meth-
ods can provide biomass estimates more accurately over large spatial scales.
The most widely used allometric equation for tropical forest biomass ground-based
estimates and validation of satellite-based estimates are based on ∼1300 harvested
and weighed moist forest trees (Chave et al., 2005; Chambers et al., 2001), and with
no destructive biomass samples from Africa included. The small sample size and ge-
ographical limits of this dataset are due to the tremendous efforts required to work in
remote forests dissecting and determining mass of trees, some of which may weigh
over 20 Mg. Such a lack of calibration data may bias estimates of carbon stocks in
tropical forests (Houghton et al., 2000; Malhi et al., 2004). One major uncertainty in
carbon stock estimates is related to architectural differences in tropical trees. For ex-
ample, across plots, regions and continents there is significant and systematic variation
in tropical forest tree height (H) for a given diameter (Feldpausch et al., 2011; Banin et
al. 2012). This applies both to multispecies equations and to those restricted to individ-
ual species (Nogueira et al., 2008b). Hence, accounting for H:Dallometry may reduce
uncertainty associated with tropical forest biomass estimates from plot to pan-tropical
scales.
Improving the accuracy of such estimates is important as almost all tropical forest
regions of the world are currently undergoing major changes, which alter biomass and
carbon stocks. For example, it is now apparent that many remaining intact tropical
7
forests are not at carbon equilibrium, but rather are accumulating biomass (Lewis et al.,
2009; Phillips et al., 1998), but an accurate quantification of this apparent pantropical
sink hinges on, amongst other factors, unbiased biomass estimates for individual trees.
Similarly, quantifying changes in global carbon stocks and emissions where much of
the active deforestation occurs (e.g. arc of deforestation in Brazil, INPE, 2009) can
be overestimated when ignoring the effect of tree Hin biomass estimates, because
trees tend to be shorter for a given Din transitional forests where the most active de-
forestation fronts often occur (Nogueira et al., 2008b). As a result, carbon emissions
from tropical deforestation (INPE, 2009) may be biased. More generally, incorpora-
tion of Hin biomass estimates may help to account for variation in carbon stocks and
could represent potential changes in calculated carbon emissions under deforestation
(INPE, 2009), selective logging (Pinard and Putz, 1996; Feldpausch et al., 2005), sinks
caused by forest regrowth (Uhl and Jordan, 1984; Feldpausch et al., 2004) and car-
bon valuation under Reducing Emissions from Deforestation and Degradation (REDD)
(Aragao and Shimabukuro, 2010; Asner et al., 2010; Gibbs et al., 2007).
Along with wood specific gravity (ρW) (Baker et al., 2004b), tree Hhas already
been incorporated into some regional and pantropical forest biomass allometric mod-
els (Brown et al., 1989; Chave et al., 2005). Biomass estimation is then based on a
four-step process:
1. measure individual tree D;
2. estimate ρWat the finest taxonomic level available from ρWdatabases (Chave et
al., 2009; Fearnside, 1997);
3. measure or estimate Hfrom allometric models based on the relationship between
Hand Dalone (Brown et al., 1989) or with additional forest structure and climate
variables to parameterise Hestimates (Feldpausch et al., 2011);
4. use these data to calculate biomass for individual trees from allometric equations
based on D,ρW, and H.
8
Despite the early recognition of the importance of Hin biomass estimates (Crow,
1978; Ogawa et al., 1965), in practice Hhas less frequently been accounted for in
pantropical biomass estimates due to lack of data. Nevertheless, where data have
been available inclusion of Hhas been shown to appreciably reduce errors in the esti-
mation of destructively sampled biomass. For example, the standard error in estimating
stand biomass for a destructively sampled dataset of trees ≥10 mm Dwas 12.5 % if
an equation including Hwas used, but 20 % for an equation derived without H(but
calibrated on the same dataset) was applied (Chave et al., 2005). This same study
showed that Hwas more important than a precipitation-based forest categorisation
(dry, moist, wet) in more accurately estimating biomass.
Thus, allometric model choice, rather than sampling error or plot size, may then be
the most important source of error in estimating biomass (Chave et al., 2004). With
the pantropical destructive biomass dataset sample size restricted by sampling cost
and effort, Hestimates from regional or continental-scale H:Dmodels may provide a
simple way to improve aboveground biomass estimates. Selection of the “best ” model
form to represent Hin biomass models is not straightforward, however, with numer-
ous statistical forms, geographical and environmental parameterisations, separations
by growth form having been tested (e.g., Fang and Bailey, 1998; Feldpausch et al.,
2011; Rich et al., 1986; Thomas and Bazzaz, 1999; Banin et al., 2012). In a global
tropical analysis using multi-level models to examine the relationship between Hand
diameter, Feldpausch et al. (2011) grouped plots into regions and found that after tak-
ing into account the effects of environment (annual precipitation coefficient of variation,
dry season length, and mean annual air temperature) and forest basal area, there were
two main regional groups differing in their H:Drelationships. Forests in Asia, E., W.
and Central Africa and the Guyana Shield are all similar in their H:Dallometry, but
with trees in the forests of much of the Amazon Basin and tropical Australia typically
being shorter at any given diameter. Using a similar dataset, but excluding drier forests,
Banin et al. (2012), conducted a continental-scale analysis and showed significantly
different asymptotic maximum Hamong continents, after accounting for differences in
9
environment, forest structure and wood specific gravity. These results suggest that ei-
ther continental, or sub-continental geographic H:Dpatterns may, in addition to model
form, be important in reducing error in biomass estimates.
Here, using a large dataset of tree H, destructive biomass data (i.e. actual tree
biomass is known) and pantropical permanent plot data (where information on Hand D
is known, but not the true biomass of a plot), we provide a first pantropical evaluation of
the effects of Hon biomass estimates, including by geographical location (plot, region,
and continent). Specifically, we address the following questions:
1. Which is the best H-model form and geographic scale for inclusion in biomass
models to minimise site-level uncertainty in estimates of destructive biomass?
2. What is the reduction in uncertainty in plot-level biomass estimates based on
census data from permanent plots across the tropics?
3. How does inclusion of Hin biomass estimation protocols modify plot- and
continental-level biomass estimates across the tropics?
2 Methods
We developed above-ground forest biomass estimates and evaluated biases using
tree diameter (D), wood specific gravity (ρW) and Hbased on destructive sampling
and permanent-plot census data. This assessment was completed through this pro-
cess, (1) compiled pantropical destructive biomass, tree H, and permanent sample
plot census data; (2) computed new pantropical biomass models that included or ex-
cluded tree H; (3) developed models to estimate Hfrom D; (4) using the destructive
data, evaluated the effect of inclusion or exclusion of actual or simulated Hin biomass
estimates; (5) applied the new biomass models and error estimate from destructive
biomass estimates to pantropical plot-based tree census data to (6) determine how
biomass estimates change when including H; (7) determined the error associated with
10
biomass estimates for pantropical permanent plots, (8) assessed regional and conti-
nental changes in biomass estimates due to Hintegration in biomass estimates.
Destructive biomass data were compiled from published and non-published data
from 21 plots in 10 countries (described below). Hand Dmeasurements are identical
to those in Feldpausch et al. (2011). The tree census data reported here (Fig. 1;
Supplement Table S1) are from permanent sample plots primarily from the RAINFOR
(Peacock et al., 2007; Baker et al., 2004a; Phillips et al., 2009) and AfriTRON (Lewis et
al., 2009) networks across South America and Africa respectively, the TROBIT network
of forest-savanna transition sites (Torello-Raventos et al., 2012), the CSIRO network
in Australia (Graham, 2006), and data from Asia (Banin, 2010) curated in the www.
forestplots.net data repository (Lopez-Gonzalez et al., 2011). In addition, for each plot,
mean annual precipitation, annual precipitation coefficient of variation, and dry season
length were obtained from WorldClim global coverage at 2.5 min resolution based on
meteorological station data from 1950–2000 (Hijmans et al. 2005).
2.1 The destructive dataset
To determine the efficacy of biomass models to predict biomass, we assembled a de-
structively sampled tree biomass dataset (n= 1816 trees) based on actual cut and
weighed tropical forest trees (Chave et al., 2005; Nogueira et al., 2008a; Hozumi et
al., 1969; Ara ´
ujo et al., 1999; Mackensen et al., 2000; Brown et al., 1995; Lescure
et al., 1983; Yamakura et al., 1986; Djomo et al., 2010; Henry et al., 2010; Deans
et al., 1996; Ebuy et al., 2011; Ketterings et al. 2001; Samalca, 2007). We here-
after refer to this as the “destructive data”. The destructive data are pantropical but
with relatively few samples from Africa (n= 116). The main differences between the
dataset used by Chave et al. (2005) are that we excluded mangrove and dry forest
biomass data from Chave et al. (2005) from our analysis; and, we included recently
published destructive biomass datasets from Africa (Ghana, the Democratic Republic
of Congo, and Cameroon) (Djomo et al., 2010; Henry et al., 2010; Deans et al., 1996;
Ebuy et al., 2011), Kalimantan, Indonesia (Samalca, 2007) and Brazil (Nogueira et al.,
11
2008a). To classify sites, climate data for the destructive dataset were extracted from
the WorldClim data based on plot coordinates. For the destructive site data, mean an-
nual precipitation ranged from 1520 to 2873 mm, dry season length 0 to 6 months, D
from 1.2 to 1800 mm, and Hfrom 1.9 to 70.7 m.
2.2 Tree height measurements
Tree height (H) had been previously measured at many of the permanent census plots
from each of the four continents. Methodology and sites are specified in Feldpausch et
al. (2011). To summarise the methods, in general a minimum of 50 trees per plot were
sampled for H(total tree Habove the ground) from 100 mm binned diameter classes
(i.e., 100 to 200, >200 to 300, >300 to 400 mm, and >400 mm). For some plots
every tree was measured for HTree Hwas measured using Vertex hypsometers (Ver-
tex Laser VL400 Ultrasonic-Laser Hypsometer III, Hagl¨
of Sweden), laser range-finders
(e.g. LaserAce 300, LaserAce Hypsometer, Leica Disto-5), mechanical clinometers,
physically climbing the tree with a tape measure, or by destructive methods. To exam-
ine how tree architectural properties related to stem D, independent of external factors
such as trees damaged by treefalls, trees known to be broken or with substantial crown
damage were excluded from analyses. A recent comparison of ground-based methods
found that trigonometric methods resulted in no systematic bias (non-laser method), or
resulted in a small underestimate of actual tree height (ground laser-based methods)
compared to heights measured from an observational tower (M. Larjavaara and H.
Muller-Landau, in prep). A second study reported a Pearson correlation of r2=0.977 for
trigonometry versus laser rangefinder estimates of height (Marshall et al. 2012).
2.3 Biomass calculations
Above-ground biomass of trees for each destructively sampled site or permanent sam-
ple plot was calculated from a combination of variables. Wood specific gravity, ρW,
was extracted from a global database (http://datadryad.org/handle/10255/dryad.235;
12
Zanne et al., 2010; Chave et al. 2009). Where species-specific values were unavail-
able, we applied genus-level values. Likewise when genus-level values were missing,
we applied family level values. Where tree identification was lacking, we applied the
mean ρWfrom all stems in the plot. Based on the moist forest biomass model form
proposed by Chave et al. (2005), we developed bootstrapped biomass model (1) as
described below to estimate biomass based on either just the measured diameter and
estimated ρW(i.e., excluding tree H)using the model form:
B= exp(a+bln(D)+ c(ln(D))2−d(ln(D))3+eln(ρW)),(1)
Alternatively, using the H:Ddatabase developed by Feldpausch et al. (2011) we in-
ferred Husing a range of H:Dallometric models, and then used that inferred value in
a bootstrapped biomass model (2) based on the form proposed by Chave et al. (2005)
as described below. The model parameterisation, which includes Hin addition to di-
ameter and ρWis:
B= exp(a+bln(ρWD2H)) (2)
2.4 Biomass error estimation with and without height
From the destructive dataset, we evaluated the ability of a range of models to esti-
mate biomass (kg) from a combination of Dand ρW, or D,ρWand H, also examining
error distributions across diameter classes and sites. To develop the H:Dallometric
relationships for inclusion in biomass models we used Hmeasurements for individual
trees made in 283 plots in 22 countries representing 37 625 individual concurrent H
and Dmeasurements. Because the global destructive tree biomass dataset was small
compared to this and with the distribution of trees in the destructive dataset not nec-
essarily similar in biomass/size distribution of a natural forest, we applied a three-step
approach to scale biomass estimates and their associated errors from the destructive
dataset to permanent plots and landscape.
1. When biomass models included H, we recomputed the regional and continental
Hmodels of Feldpausch et al. (2011) to test for their efficacy to reduce error
13
in biomass estimates. These Hmodels were either a non-linear 3-parameter
exponential (Pinheiro et al. 1994; Banin et al. 2012) viz:
H=a−b(exp(−cD)),(3)
or, a model where Hscales with Daccording to a simple power function as in:
H=aDb,(4)
or, alternatively a Weibull function, which takes the form of (Bailey, 1979):
H=a(1−exp(−bDc),(5)
As there is good evidence of a large difference between different geographical areas
in H:Dallometry (Feldpausch et al., 2011; Banin et al. 2012), we derived region- and
continent-specific parameterisations for each H:Dequation and report the residual
standard error and Akaike Information Criterion for the selected models (Akaike, 1974).
We then tested how these parameterisations of Hincreased or decreased biomass
estimates.
1. To test the effect of the inclusion of Hestimates on biomass estimates, we com-
puted a biomass model of all sites with destructively harvested trees, except the
site which we wished to estimate. We then estimated the biomass of the trees in
the site that was excluded from the model. We then repeated dropping a different
single site until all sites were excluded once from model development. For each
dropped site, the mean relative error in estimated biomass was calculated for a
site, where relative error was represented as: (BP–BM)/BM, where BPis the
predicted biomass of a tree (with or without Hmodel) and BMis the biomass
measured by destructive sampling of individual trees.
2. To evaluate how the error from the destructive dataset related to the distribution
of trees found in pantropical forests, we estimated biomass for 327 plots from the
14
forest permanent-plot database as described above by locale for tree-diameter
classes, providing a biomass distribution by diameter class for each geographic
locale (note that the destructive data came from “sites” – sample areas that may
not have defined boundaries—while the permanent plot data come from defined-
area sample “plots”). We then propagated error from (ii) from the destructive
dataset to each diameter bin by geographical location and report the mean rel-
ative error for each region. The log-transformation of tree Dand biomass data
produces a bias in final biomass estimation so that uncorrected biomass esti-
mates are theoretically expected to underestimate the real value (Sprugel, 1983;
Baskerville, 1972). This effect can be corrected by multiplying the estimate by a
correction factor:
CF= exp RSE2
2(6)
which is always a number greater than 1, and where RSE is the residual standard error
of the regression model.
2.5 Permanent plot tree census data
To determine how Hintegration alters biomass estimates and affects error in biomass
estimates, we compiled a pantropical dataset of permanent sample plots (Supplement
Table S1). All plots occurred in intact (minimal recent direct anthropogenic influence)
forest, with a minimum plot size of 0.2 ha (mean = 0.95; max = 9 ha), area using stan-
dardised sampling methodologies across all sites. Diameters of all live trees and palms
(≥100 mm diameter at breast height (D)) were measured to the nearest 1 mm at 1.3 m
above the ground or 0.5m above any buttresses or stilt-roots following international
standards of permanent sampling plot protocol (Phillips et al., 2010). Trees were iden-
tified by a local botanist. For unknown species, vouchers were collected, later identified
and archived.
15
2.5.1 Africa
African permanent sample plots (n= 62) were grouped into three geographical regions:
Western, Eastern and Central Africa. Measurements were made in West Africa in
Ghana and Liberia (Lewis et al. 2009; Feldpausch et al. 2011). Central African sites
were sampled in central and southern Cameroon, and Gabon (Lewis et al. 2009;
Feldpausch et al. 2011). Eastern African sites were established in the Eastern Arc
Mountains of Tanzania (Marshall et al., 2012). The number of months with precipitation
<100 mm per month, the estimated average monthly evapotranspiration of a tropical
forest (Shuttleworth, 1988) and a widely used index of dry season length (Malhi and
Wright, 2004), varies from 1 to 7 months across all sites.
2.5.2 Asia
We classified forests in Asia (n=14) as one region for this study, with the division be-
tween Asian and Australasian plots according to Lydekker’s line (Lohman et al., 2011).
Wet and moist forests were sampled in Brunei and Malaysian Borneo (Banin, 2010;
Banin et al., 2012). These sites have zero months with mean precipitation <100 mm
per month.
2.5.3 Australasia
Trees were sampled in tropical forest permanent plots (n= 26) in northern Australia
(Graham, 2006; Torello-Raventos et al., 2012). Precipitation varies over very short
distance from coastal to inland sites, with the dry season ranging from 4 to 10 months.
2.5.4 South America
Tree censuses conducted in plots (n= 225) (Baker et al. 2009; Feldpausch et al. 2011;
Nogueira et al. 2008) in South America are here grouped into four regions based
on geography and substrate origin (e.g. Fittkau 1971; Schobbenhaus and Bellizzia,
16
2001): Western Amazonia (Colombia, Ecuador and Peru), with soils mostly originating
from recently weathered Andean deposits (Quesada et al., 2009); Southern Amazo-
nia encompassing the Brazilian shield (Bolivia and Brazil); on the opposite side of
the Basin to the north the Guyana shield (Guyana, French Guiana, Venezuela), and
Eastern-Central Amazonia (Brazil) which is mostly comprised of old sedimentary sub-
strates derived from the other three regions. The number of months with precipitation
<100 mm per month ranges from 0 to 9 months.
2.6 Patterns and revision of biomass and carbon stocks
We used a Monte Carlo approach to quantify uncertainty in biomass estimates with
and without H, and to extrapolate biomass estimates from plots to the landscape.
We accounted for the uncertainty in wood specific gravity (ρW) and Hmeasurements
in biomass estimates. For our analysis, we calculated a mean biomass (or carbon)
estimate and 95 % confidence interval for each plot, region, and continent from 1000
realisations of biomass estimates for individual plots. These estimates were based
on 1000 realisations of biomass estimates for individual trees in each plot based on
the normal distribution of values of the standard error drawn from a random sample
for each tree. To estimate biomass for each tree, we used our new biomass models
and generated 1000 realisations for each tree by adding error to the ρWand Hwhere
applicable. The ρWof each tree including the error terms was estimated as ˆρW=ρW
+χρW∗SE ρW, and Hfor each tree including the error terms was estimated as ˆ
H=
H+χH* SEH, where the hat symbol ”ˆ” represents values that include estimates of
error , χrepresents a random value sampled from a distribution with mean = 0 and
standard deviation = 1, and SE represents the standard error of ρWor Hfor a plot.
For the realisations of biomass stocks based on forest area, we drew 1000 times from
the sample plots for each region. The 95 % confidence interval was calculated as CI =
(C97.5- C2.5)/2, where C97.5and C2.5are the 97th and 2.5th percentiles, respectively, of
the 1000 realisations of each estimate.
Spatial patterns in plot-level biomass estimates with and without Hwere exam-
17
ined by region and continent. Based on the regional tropical forest area estimates
of broadleaf deciduous open and closed and evergreen tree cover classification from
GLC2000 (Global Land Cover Map 2000) (Bartholom´
e and Belward, 2005) reclassified
in ArcGIS®(ESRI, 2010), we scaled bootstrapped regional biomass estimates and
uncertainty tropics-wide. Our estimates of tropical forest area are lower than those re-
ported by Mayaux et al. (2005) since we excluded the more open vegetation classes.
Biomass was converted to carbon values using a conversion factor of 0.5 (Chave et
al., 2005). Statistical analyses were conducted using the R statistical platform (R De-
velopment Core Team 2011). Biomass and Hmodels were developed using the lme
and nlme functions of R (Pinheiro et al., 2011).
3 Results
Using our expanded pantropical destructive biomass dataset (Fig. 2a), we first examine
how estimates of real (destructive) biomass data using boot-strapped biomass models
(Table 1) are affected by different Hmodel forms and regional or continental parame-
terisations by examining the relative error by diameter bin (Fig. 2b) and overall bias in
biomass estimates by destructively sampled site (Table 2). We next examine how the
selected Hmodels (Table 3) affect biomass estimates (Fig. 3) and uncertainty (Fig. 4)
as a result of regional variation in forest structure (Supplement Table S2) and distri-
bution of biomass among diameter classes for trees measured in pantropical perma-
nent sample plots (Supplement Table S1), and finally extrapolate our results to assess
the influence of incorporating variations in H:Dallometry on regional/continental and
global biomass estimates (Table 4 and 5).
18
3.1 How much does the inclusion of height reduce uncertainty in destructive
biomass estimates?
The distribution of destructively sampled above-ground tree dry mass from the avail-
able pantropical dataset was roughly equally sampled across the 50 mm increment
diameter classes from 250 mm< D ≤500 mm but, although involving many more indi-
vidual trees, somewhat less for D <250mm (Fig. 2a). Although relatively few trees had
been sampled for large diameter classes (e.g. 17 trees ≥1000 mm diameter), these
larger trees clearly accounted for a significant proportion of the total biomass to be
simulated within the dataset. The biomass in Fig. 2a represents the nearly 1500 Mg of
biomass destructively sampled to date in moist tropical forest which we use to assess
the effect of Hin biomass estimates. Most of these data have been used in the param-
eterisation of currently used pantropical biomass models (e.g. Chave et al., 2005), but
with newly published data from Africa, Asia, and Brazil included in our analysis.
3.1.1 Measured heights
The effect of the inclusion of Husing the biomass model forms of Chave et al. (2005)
as applied to our dataset are presented in Table 1, where our allometric equations both
with and without Hincluded (i.e. Eqs. 1 and 2) are compared. This shows that applying
Eq. (1) (which excludes H) resulted in a considerably higher residual standard error
(RSE) and Akaike information criteria (AIC) estimates than for when Hwas included
(Eq. 2).
3.1.2 Simulated heights
The effects of substituting estimates of Hfrom Eqs. (3–5) into Eq. (2) are shown in
Table 2. The inclusion of Himproved site-level estimates of aboveground biomass,
bringing them closer to the known destructive harvest values, with a relative error of,
e.g. 0.06 for both the Weibull-Hregion and continent-specific Hmodels (Table 2).
19
Excluding Htended to produce overestimated aboveground biomass estimates, with
a relative error of 0.13. Regionally derived Hestimates were non-significantly better
than continental scale-derived Hestimates at predicting site-level biomass (Table 2).
Specifically, the Weibull-H(Eq. 5) (Table 3) consistently reduced the relative error in
biomass estimates over all diameter classes as compared to the non-Hestimates. This
contrasted with the power-Hmodel (Eq. 4) which, although reducing error even further
in some diameter classes, had greater error for other diameter classes, even greater
than those derived from Eq. (1) which excludes H(Fig. 2b), The power model also had
greater error for small diameter classes. Hence, overall, we consider that the Weibull
model modestly outperformed the other two function forms of H:Drelationships, and
utilise this relationship (Table 2).
3.2 Improving biomass estimates from permanent sample plots
3.2.1 Effect of including height in biomass estimates
Integration of the region-specific Weibull-H, on average, reduced estimated biomass
per plot ( ˆ
B) relative to excluding Hin biomass estimates, on average by
−52.2±30.8 Mg dry mass ha−1bootstrap 95 % CI (Figs. 1b and 3, Table 4). As shown
by the cumulative biomass curves in Fig. 3, including Hin biomass estimates did not
affect all regions equally. For South America, including Hsignificantly reduced biomass
estimates for all regions (e.g. by −55.9, −66.6, and −48.0 Mg ha−1for the Brazilian
Shield, east-central Amazonia and western Amazonia, respectively) (paired t-test, p
<0.001). East and West Africa, and Northern Australia also had significantly lower
biomass estimates when including H(e.g. −107.9, −44.2,−116.5 Mg ha−1, respec-
tively). Southeast Asia, central Africa, and the Guyana Shield of South America had
small, but significant reductions in biomass estimates when including H(paired t-test,
p <0.001). No region had significantly higher biomass estimates after including H(see
Supplement, Table S1, for ∆biomass estimates for all plots).
20
3.2.2 Differences in biomass distribution among regions
Forests store a large fraction of total biomass in smaller diameter stems, with appre-
ciable differences in the biomass distribution among diameter classes reflecting strong
regional patterns (Fig. 3). For example, forests of the four regions of South America
had a significantly (p < 0.05) larger fraction of total biomass in smaller size classes (≤
40 cm D) compared to the three regions of Africa and Asia. This is shown graphically in
Figure 3 by the cumulative biomass curves, where forests of some regions approach an
asymptote in cumulative biomass at larger diameter classes. The vertical dashed line
in Figure 3 represents the mid-point in biomass storage above and below the indicated
diameter bin.
It is because of the skewed biomass distributions of Figure 3 with a concentration
of biomass in smaller diameter classes in most forests (e.g. ≤40 to 60 cm D), that
in Sect 3.1 we chose the Weibull-Hmodel, which has lower relative error in small
diameter classes (in contrast to the power-Hmodel and three-parameter exponential
model), and therefore has the greatest plot-level effect in reducing uncertainty. After
accounting for regional tree Hdifferences, total biomass per hectare is thus estimated
to be greatest in Australia, the Guyana Shield, and Asia and lowest in W. Africa, W.
Amazonia, and the Brazilian Shield (descending order) (Table 4).
3.2.3 Estimating effects of Hon errors in permanent sample plot biomass
estimates
To estimate error in permanent plots due to error in destructive measurements, we mul-
tiplied the relative error from the diameter bin from the small sample of destructive mea-
surements for the Weibull-Hmodel (Eqs. 2 and 5) as shown in Fig. 2b, by the biomass
of the equivalent size-class in each of the pantropical permanent plots. This relative
error in field-based plots was greater when the same procedure was undertaken for the
“no-H” Eq. (1) (Fig. 4). Specifically, by including H, the error in estimates is reduced
in small diameter-classes, but not large diameter-classes. This is because of the in-
21
creasing absolute errors of the Weibull−Hmodel for the larger trees.The mean error in
biomass estimates for all regions when including Weibull-Hin biomass estimates was
an overestimate of 8.0 Mg ha−1; a value considerably less than the calculated overes-
timate of 41.8 Mg ha−1when Hwas excluded (Fig. 4). The alternative two Hmodels
of Eqs. (3) and (4) were also tested and found to underestimate biomass by −8.2 and
−5.5 Mg ha−1, respectively. Overall, inclusion of Weibull-H(Eq. 5) in biomass esti-
mates for tropical forest plots resulted in a smaller mean bias in biomass estimates
compared to when Hwas omitted. Specifically the bias with Hincluded ranged from 6
to 9.5 Mg ha−1(South America), 10.1 to 10.6 Mg ha−1(Asia and Australia), and 5.3 to
7.3 Mg ha−1(Africa), as compared to estimation without H, which had biases of 28.6
to 47.2 Mg ha−1(South America), 48.9 to 63.2 Mg ha−1(Asia and Australia), and 40.5
to 49.4 Mg ha−1(Africa) (Fig. 4).
3.3 Effect on pantropical carbon estimates
Based on published estimates of tropical forest area (GLC2000), and biomass and
carbon estimated in our permanent plot networks, we calculated the change in regional
and continental above-ground live tree carbon stocks due to integration of Hin biomass
models. Using GLC2000 (Bartholom ´
e and Belward, 2005) tropical forest categories
and mean carbon storage in each region from the plot data, the tropical Americas
had the largest relative reduction (−14 % ±5 % bootstrap 95 % CI) in estimated carbon
storage due to H, and with Asia (−2 %±2 % bootstrap 95 % CI) the smallest. Inclusion
of region-specific Hmodels to estimate carbon reduced tropics-wide estimates of total
carbon in tropical forests from 320 (±71 bootstrap 95 % CI) to 285 Pg C (±64 bootstrap
95 % CI), a reduction of 35.2 PgC, or 13 % (±10 % bootstrap 95 % CI), relative to when
Hwas included (Table 5).
22
4 Discussion
We show that (1) including Hsignificantly improves the accuracy of estimation of tropi-
cal forest aboveground biomass, (2) failing to include Husually causes an overestimate
of biomass, (3) such overestimates may have globally significant implications; here we
estimate that carbon storage in tropical forests may be overestimated by 13±10 %,
and; finally (4) we recommend continental or regional-specific asymptotic Weibull H:D
functions to be included in future estimates of biomass to reduce uncertainty in above-
ground biomass estimates in tropical forests. Below, we discuss some of the sources
of variability in biomass and Hestimates, limitations of these models and implications
for pantropical scaling and carbon valuation under REDD.
4.1 Compensating for imperfect biomass models
4.1.1 Representing height in biomass estimates
In this study we selected the Hmodel based on the region-specific parameterisation
of the Weibull-H(Eq. 5) model because it reduced error in estimating biomass for the
smaller diameter classes (Fig. 2b), and with these classes constituting a large part
of the plot-level biomass (Fig. 3). Although the Weibull-Hform is less than ideal for
trees 800–1000 mm diameter, the three-parameter exponential (Eq. 3) and power-H
models (Eq. 4) were not significantly better biomass estimators for the largest trees
(Fig. 2b). This may be because the parameterisation of the Weibull-Hmodel should
theoretically account for some of the asymptotic nature of tree growth more than the
power or 3-parameter-exponential-Hmodel. In general, however, asymptotic Hhas
not been detected as often as may be expected among species growing in tropical
forest (Poorter et al., 2006; Chave et al., 2003; Davies et al., 1998; Thomas, 1996;
Iida et al., 2011), where only one-fourth of species in sites sampled in Bolivia reached
an asymptote (Poorter et al., 2006). However, asymptote detection is likely to be, in
part, sample size dependent. Unlike the power model, the 3-parameter-exponential
23
and Weibull functions for tree Hhave an additional biologically meaningful parameter,
with a term for maximum tree height (hmax) here being applied at the plot, regional, or
continental (as opposed to species) level, and it is for this reason that the hmax should
be interpreted carefully. For example, when pooling the transitional forests from our
study for the Brazilian Shield of Amazonia, the Weibull-Hmodel converged on a hmax
beyond the observable tree size range and thus spurious, large hmax parameters may
be obtained (e.g. >200 m). This model, however, gives an estimate of 11 and 47
m for trees of 100 and 1600 mm diameter, respectively, demonstrating that although
the model provides realistic values, use of hmax alone to describe stand properties
could give erroneous interpretations. For some forests, the power-Hmodel provides
a better fit for large-diameter trees (Feldpausch et al., 2011) and in the current study
the power model resulted in a lower mean error in estimating destructive tree biomass
(Supplement Table S1). With a goal of reducing error in stand biomass estimates, the
asymptotic model form – which reduces error in small-diameter trees – outperforms
the power model because of the skewed distribution of stand-level biomass found in
smaller-diameter trees, and was, therefore, chosen (Fig. 3).
Independent of Hmodel form, no current regional-scale Hmodels are parame-
terised to account for successional variation of tropical forest trees. Secondary forest
trees are frequently taller for a given D(Montgomery and Chazdon, 2001). Mechanical
effects can also modify small patches of forest over large areas, where, for example,
bamboo can modify H:Drelationships (Griscom and Ashton, 2006) and wind may alter
forest structure (Laurance and Curran, 2008). Our Hmodels were developed from the
most comprehensive dataset to date, which includes a range of forest types including
bamboo and liana forests. Developing site- or forest-specific Hmodels is one alter-
native to account for localised variations in forest structure, but requires substantial
cost and field time to develop. Development of plot-level basal area-weighted height
estimates (i.e. Lorey’s height) would also aid in validating remote sensing biomass
estimates (e.g. Saatchi et al. 2011).
24
4.1.2 Modelling destructive biomass data
Examination of Fig. 2b raises two questions: “Why does exclusion of Hin biomass
estimates largely overestimate true biomass?” and “Why are biomass models unable
to reduce error in large trees?” It was previously noted that pantropical biomass mod-
els overestimate biomass in large trees (Chave et al. 2005). Some of this error was
attributed to the lack of sampling in large trees (Chave et al., 2004); however, close in-
spection of Fig. 4 in Chave et al. (2005) shows that biomass of the smallest trees (e.g.,
<100 mm diameter) is also underestimated (with these trees having the largest sample
size). This suggests a different biomass model functional form may be necessary to
remove the positive bias of trees ≥100mm diameter either with or without including
H. Other studies have confirmed that the model functional form we use (Eqs. 1 and 2)
provides a better fit than other parameterisations (e.g. Vieilledent et al., 2011).
The challenge to reduce uncertainty in biomass estimates of large-diameter trees
(e.g. ≥800 mm diameter) can be understood by examining the destructively sampled
trees. Trees from this diameter class have an enormous variation in mass, from 4.6 to
70.2 Mg (mean 15.3 Mg) and similarly, a wide range of wood specific gravity, 0.26 to
0.9 g cm3(mean 0.56), and vary in Hfrom 32 to 71 m (mean 46). These differences
may represent the substantial variation in life-strategies among “emergent” canopy
species, where large diameter low-density light demanding trees coexist with shade
tolerant species. Thus, not only larger sample sizes of large size trees are needed, but
in the future perhaps two differing equations, for differing life history strategies will be
required (e.g. see Henry et al. (2011), for some data analysed in this way).
Clearly, greater collaboration is required to sample trees and unify the many de-
structively sampled tree datasets into one database to improve regional or pan-tropical
biomass equations with inclusion of H. Our study provides a first step in dissecting one
component of this variation (regional H:Drelationships) to revise tropical biomass es-
timates, e.g. we show that regions differ in their distribution of biomass among Dclass
(Fig. 3), and that as a result, effects of inclusion of Hestimates on predicted biomass
25
values vary strongly from region to region (Table 5).
4.1.3 Regional and continental differences
While noting the limited sample sizes for some regions, we show that forest biomass,
after taking Hinto account was highest in Australian forests. Biomass was as high
in the Guyana Shield as in SE Asian forests, which is inconsistent with the view from
previous studies that have suggested that aboveground biomass storage is higher in
Southeast Asia (e.g. Slik et al., 2010). In addition, regional adjustments in biomass
estimates due to elevation and tree Hmay be necessary for some areas. For example,
tree Hvaries with elevation in Tanzania, with the tallest trees at mid-elevation (Marshall
et al., 2012).
We found substantial different biomass distribution among diameter classes between
the forests of South America and Australia, and Africa and SE Asia, which affected
error propagation and determined Hmodel selection. Forests of South America have
a greater proportion of the total biomass in the smaller size classes ≤40 cm D; flatter
distributions are found in Africa and Asian forests, with East African forests showing the
lowest proportion of biomass in small size-classes (22 %)(Fig. 3). With the exception of
the Guyana Shield, these regional patterns broadly correspond to reported differences
in the relationship in H:Dallometry (Feldpausch et al. 2011). Larger sample sizes are
needed to assess whether these biomass distributions differences are consistent when
expanded beyond the regional clusters. The Weibull Hmodel was selected because
it reduced uncertainty in the smallest diameter size-classes, which for most forests
hold a large part of the biomass. As a result of the large region-to-region variation
in biomass distribution among diameter classes (Fig. 3), future work my indicate that
other Hforms are more effective in reducing uncertainty in forests that contain different
biomass distribution among diameter classes.
Feldpausch et al. (2011) used a similar regional analysis, and showed a group of
tall-stature forests (African, Asia and Guyana Shield) and other lower-statured forests
(Amazon and Australia), while Banin et al. (2012) reported significant differences in
26
maximum heights among continents. Intriguingly, the biomass distribution by diameter
class results appear to follow a continental split, not a forest stature split, with the
Guyana shield forests grouping with the rest of South America and not African forests.
The reasons for this are unclear, but may be related to the interaction between stem
density and H. Their studies showed that H:Drelationships were related to stem
density, with forests with higher stem density having shorter trees for a given diameter.
Trees of the Guyana Shield, for example, have the lowest stem density for plots in South
America, and also are on average taller and have the highest biomass stocks for the
continent (Table 4; Supplement Table S2). Our current results indicate that the inclusion
of Hin biomass estimates for the Guyana Shield, Asia and Central Africa, forests with
trees on average taller for a given D, do not substantially modify estimates compared
to estimates based on the no-HEq. (1), but that including Hin biomass estimates
for those regions reduces the bias in destructive estimates relative to excluding H
(Table 2). These results showing substantial variation in biomass distribution and forest
structure among regions and continents indicate that future biomass models based on
continents and regions may prove more robust than pantropical models.
4.1.4 Climate and biogeography
Furthermore, the patterns that emerge in tree Hvariation as a function of region,
climate and, forest structure suggest alternative structuring is needed for pantropi-
cal Biomass:Diameter tree allometric models rather than basing them solely on forest
moisture class (e.g., dry, moist, wet). For example, H:Drelationships vary not only
according to climate (e.g., taller trees in moist climates), but also by forest structure
(e.g. taller trees in higher basal area forests), soil quality, and geography (e.g. taller
trees for a given diameter in the Guyana Shield, Africa and Asia than in the rest of
South America and Australia; Feldpausch et al., 2011). Biomass:Diameter allometry
for most published large-scale biomass models, however, is fixed by region (e.g. Ama-
zonia, Chambers et al. 2001) or is pan-tropical (e.g., Chave et al., 2005), or is based
on broad classifications of forest moisture (e.g., dry, moist, wet forest: Chave et al.
27
2005) or vegetation (e.g., diptercarp, secondary forest (Basuki et al., 2009; Nelson et
al., 1999)). These models therefore lack parameters to account for possible climate-
driven or biogeographic variation in Biomass:Diameter relationships. However, the
clear biogeographical differences amongst SE Asian and forests on other continents
(dominance by the Dipterocarpaceae) are not shown here, and were not the proximate
reason for differences in H:Dallometry in Asia versus elsewhere (Banin et al., 2012).
Formation of region-specific Hmodels provides a first step in parameterising regional
biomass estimates based on reported variation in tree Hallometry (Nogueira et al.,
2008b; Feldpausch et al., 2011).
Current pantropical biomass models are also unable to account for regional or
forest-specific variation in crown diameter, where wider crowns may impart greater
biomass for a given diameter. Based on high-resolution remote-sensing data, Bar-
bier et al. (2010) indicated that crown size increases by ∼20% from the wetter to the
more-seasonal regions of Amazonia. The regional Hpatterns showing shorter trees in
southern Amazonia (Nogueira et al., 2008b; Feldpausch et al., 2011) that would result
in reduced biomass stocks, may be partially offset by wider crowns that contain more
mass for a given diameter. Such possible effects remain to be tested with field data.
4.1.5 Intra-species, diameter-specific and regional wood density variation
Tree wood specific gravity (ρW)variation is another parameter that biomass models
may inadequately represent. Wood specific gravity is highly variable across regions,
is a key determinant of large-scale tree biomass spatial patterns (Baker et al., 2004b;
Chave et al., 2006), is a more important predictor than tree height in biomass models
(Chave et al., 2005), and therefore accounting for it holds a central role in reducing
uncertainty in biomass estimates. Current biomass calculations use ρWdatabases to
assign the finest taxonomic value to an individual (e.g., species-specific ρW)indepen-
dent of stem diameter, and our bootstrapped estimates account for uncertainty in ρW(in
addition to H). Data from Barro Colorado Island, Panama showed significantly lower
ρWin large-diameter trees than in smaller trees (Chave et al., 2004), while Pati˜
no
28
et al. (2009) showed, using branch wood density (which may be more plastic than
stem wood density), that there is considerable plot-to-plot variation in wood specific
gravity. Additionally, mean tree ρWis significantly lower in some regions of Amazo-
nia (Nogueira et al., 2007). In addition, engineering theory suggests that trees with
low density wood have an advantage in both Hgrowth and in mechanical stability as
compared to high-wood-density trees (Anten and Schieving, 2010; Iida et al., 2012);
in contrast to vertical growth, high-density wood imparts greater efficiency for horizon-
tal expansion. Together, these results suggest that biomass models may benefit from
greater parameterisation.
Variation in the wood carbon fraction is another source of uncertainty in estimating
regional and pantropical carbon stocks. Many studies, as in the current study, take the
wood carbon fraction as 0.5 to convert estimated biomass to carbon (e.g. Lewis et al.,
2009; Malhi et al., 2004; Clark et al., 2001). However, carbon content varies regionally
(Elias and Potvin, 2003), where, for example, a forest in Panama has mean carbon
values of 0.474±0.025, which would result in an overestimate of 4.1–6.8Mg C ha−1if
the assumed 0.5 carbon content were used (Martin and Thomas, 2011). Accounting
for such variation may assist in refining future pantropical carbon estimates.
4.1.6 Limited spatial extent
A further concern is the use of spatially limited destructively sampled biomass data
forming the base of biomass models used to estimate biomass for trees in other re-
gions. Until only recently, destructive data were unavailable for Africa, so that large-
scale biomass estimates for this continent were based on data from elsewhere. Re-
gional biomass equations may yield site-specific bias. For example, the Chambers
et al. (2001) equation, which is based on data from a small area north of Manaus,
Brazil, yet by necessity has been used to estimate biomass across the Amazon Basin
(Baker et al., 2004a; Malhi et al., 2004, 2006), an area with important variation in tree
architecture (Nogueira et al., 2008b; Feldpausch et al., 2011; Barbier et al., 2010),
taxonomy (Pitman et al., 1999) and wood density (Baker et al., 2004b). Application
29
of this model to southern Amazonia requires a height factor to down-scale biomass
estimates to account for shortertrees (Nogueira et al., 2008b; Nogueira et al., 2007).
Country-level assessments of biomass model-effects on estimates indicate that appli-
cation of generic pantropical biomass models (e.g. Brown et al., 1989; Chave et al.,
2005) should be evaluated prior to application, especially those that lack Hparameter-
isation (Alvarez et al., 2012; Vieilledent et al., 2011; Marshall et al., 2012). Our current
results showed tropics-wide geographical variation in biomass distribution among D
classes in permanent plots, which, together with tropics-wide variation in H:Drelation-
ships (Feldpausch et al., 2011), may not be represented when forming small regional
subsets or pooling pantropical destructive data without accounting for H.
4.2 Implications for remote sensing
Observed tropical forest H:Dallometry differences in ground-based studies (Feld-
pausch et al., 2011; Nogueira et al., 2008b; Banin et al., 2012) and their associated
regional effects on biomass estimates shown here, will be important for improving re-
trieval of biomass estimates from LiDAR (e.g. Drake et al., 2002; Lefsky et al., 2005;
Asner et al., 2010) and radar (Geoscience Laser Altimeter System, GLAS; e.g. Saatchi
et al. 2011), techniques that estimate a canopy H, or are used to estimate forest
structure (full waveforem LiDAR), either of which is then translated into a biomass es-
timate. Current pan-tropical remote-sensing biomass estimates (e.g. Saatchi et al.
2011) transform remotely-sensed estimates of canopy height into biomass estimates
based on the relationship between basal-area weighted H(Lorey’s H) for a ground-
based plot and biomass estimates for trees within those plots or based on the rela-
tionship between carbon density estimated from allometric models (e.g. Chave et al.
2005) and remotely-sensed estimates of canopy height (e.g. Baccini et al., 2012). Min-
imising error in estimating biomass for trees within plots will likewise reduce error when
calibrating remotely sensed estimates of biomass via LiDAR or radar. Height inclusion
in the allometry used to estimate biomass for those plots reduces uncertainty, as we
have shown here. Future remote sensing biomass estimates that address regional
30
variations in Hand its effect on biomass estimates when calibrating remotely sensed
Hto estimate biomass should therefore assist in evaluating potential bias and be able
to provide tropical biomass estimates of improved accuracy.
4.3 Implications for carbon sink and estimates of nutrient turnover
Permanent plot data indicate that intact apparently mature tropical forests are not in
biomass equilibrium, but have tended to gain biomass per unit area. Tree recruitment
has outpaced mortality (Phillips et al., 2004) and total tree above-ground biomass has
increased over recent decades (Phillips et al., 1998, 2009; Lewis et al., 2009). It
has been estimated that, on average, trees in tropical forests add 0.49 Mg C ha−1in
above ground mass each year, implying a carbon sink in aboveground live biomass of
0.9 Pg C yr−1(Lewis et al., 2009). This process, however, is susceptible to drought,
and for Amazonia the 2005 drought at least temporarily reduced the long-term above-
ground carbon sequestration (Phillips et al., 2009).
Our lower mean biomass estimates from forest plots implies that the calculated net
carbon sink or the magnitude of any reversal or reduction in the sink due to drought
may also be reduced for some regions as a direct result of Hparameterisations us-
ing current pantropical biomass models. This assumes that the proportional sink re-
mains unchanged. Furthermore, biomass estimates for individual trees are frequently
used to estimate nutrient stocks such as nitrogen and phosphorus in trees and stands
(Feldpausch et al., 2004, 2010; Buschbacher et al., 1988) based on component tis-
sue concentrations (Martinelli et al., 2000). Downscaling biomass estimates due to H
will therefore reduce the total estimated above ground nutrient stocks and flux due to
land-use change (e.g., selective logging, deforestation, forest regrowth and fire).
4.4 Comparison with global emissions
The biomass and carbon downscaling due to Halso affects estimates of carbon emis-
sions. The recent IPCC estimate of global emissions contribution of tropical deforesta-
31
tion estimates a net annual emission from this source of 1.6 Pg C (range 1.0–2.2 Pg C)
(Denman et al., 2007) based on the mean of estimates by DeFries et al. (2002) and
Houghton et al. (2003) from the 1980s and 1990s. The recent “unofficial” estimate with
the same methodology is 1.47 Pg C yr−1for the 2000–2005 period (Houghton, 2008).
Our new results incorporating Hinto these estimates imply that this is an overestimate
of ∼0.1 Pg C yr−1, this being based on the more recent number for the values used in
the estimate for emissions from below-ground biomass and uptake of secondary forest
regeneration, the contribution of live aboveground biomass cut in tropical deforestation
is 0.85 Pg C yr−1, and a 0.13 downward adjustment for tree H(Table 5). For compar-
ison, the last national inventory of the UK under Climate Convention indicates a total
emission in 2007 of 0.17 Pg yr−1of CO2-equivalent carbon (UK Department of Energy
and Climate Change, 2009).
4.5 Repercussions for carbon estimation and REDD
Integration of Hinto biomass estimates reduces estimates of tropical carbon storage
by 13 % (±10 % bootstrap 95 % CI). This estimated decrease has potential economic
implications based on the calculated high carbon storage of pantropical forests un-
der Reducing Emissions from Deforestation and Degradation (REDD) carbon-payment
schemes (Miles and Kapos, 2008). In monetary terms, our calculated decrease in car-
bon storage represents a reduction in value, in some regions, per unit area of tropical
forests based on current carbon market prices (e.g. Chicago Climate Exchange, Euro-
pean Climate Exchange) if previous estimates utilised published pantropical allometry
and excluded Hmeasurements. However, we stress, (i) the actual carbon storage of
these forests has not changed, only the estimated amount; (ii) the large-scale RAIN-
FOR South American estimates of biomass and change (e.g. Malhi et al., 2006; Phillips
et al., 2009) used the Baker et al. (2004b) regional biomass model; for Africa, Weibull
asymptotic continental-scale Hequations were used to estimate height in the Chave
et al. (2005) pantropical allometric equations (Lewis et al., 2009); hence, the effect of
accounting for Hin their estimates remains unexplored; (iii) that our adjustments in
32
plot-based estimates are sensitive to the current pantropical biomass equations as dis-
cussed above. Future improvement and inclusion of additional data (e.g. from Africa),
and harvested trees of larger diameter will further reduce uncertainty in estimates over
a heterogeneous landscape and at a variety of scales. New models may eventually
show that such downscaling is unnecessary; iv) tree Hintegration can reduce un-
certainty in biomass estimates (Figs. 2b and 4), which should benefit REDD; (v) our
extrapolations to regions and the tropics are based on necessarily limited sample sizes.
Furthermore, the tier-I estimation method of forest carbon density issued in support of
REDD by the Intergovernmental Panel on Climate Change (IPCC) is based on average
carbon values for biomes (IPCC, 2006), not plot-based estimates. The approach out-
lined in the present study, coupled with better measurement of H(e.g., using LiDAR)
can help generate accurate, verifiable biomass estimates which will ultimately increase
confidence in large-scale carbon estimates, potentially increasing the area receiving
carbon credits, , and greater investment per unit of carbon (Asner et al., 2010).
5 Conclusions and future considerations
Based on these results, it is possible to make a number of recommendations:
1. A major initiative is needed to improve the pantropical destructive tree data to
support global carbon modelling and policy: additional sampling is needed from
under-represented regions, forest types, growth forms (e.g., palms), and tree di-
ameter classes to represent the full diversity of tropical forests. We showed dis-
tinct differences in the biomass distribution among diameter classes of tropical
forests in different regions across the tropics, and such apparently important dif-
ferences will only be fully accounted for in biomass estimates when we have im-
proved understanding through destructive sampling.
2. Pantropical permanent forest plots, some monitored since the 1970s, are now a
baseline standard by which scientists and policymakers understand forest dynam-
33
ics and potential changes in net carbon gain, with implications for carbon valuation
under REDD+. There is known large variation in Hamong these plots. To account
for this variation and make full use of permanent-plot data, we recommend a strat-
ified random sample of Hmeasurements. If possible, Hmeasurements of every
tree are desirable. Where local H-diameter relationships are not known, using
those described in this paper is recommended.
Biomass estimates of tropical forests are prone to error because of the very small
destructive dataset, biomass models, Hmodels and also because of uncertainty in un-
ambiguously defining an area of tropical forest. Our study has explored the uncertainty
associated with current biomass estimates and shown the importance of accounting
for tree-level variation in H:Drelationships for scaling to more precise regional and
global biomass estimates. By reducing uncertainty in pantropical estimates, we make
a step forward in providing realistic, verifiable carbon estimates for models and policy
instruments such as REDD+.
pdf
Acknowledgements. Research was supported by the RAINFOR and AfriTRON networks with
additional support from the Gordon and Betty Moore Foundation; NERC through the AMA-
ZONICA and TROBIT and AfriTRON projects. LB was supported by a NERC PhD Studentship
and Henrietta Hutton Grant (RGS-IBG); SLL is supported by a Royal Society University Re-
search Fellowship; some African data were collected under a NERC New Investigator Award
(AfriTRON); Additional support was provided by EScFund grant of the Malaysian Ministry
of Science, Technology and Innovation (MOSTI); Tropenbos International and the European
Commission; Large-scale Biosphere Atmosphere Experiment in Amazonia (LBA) under the
leadership of the Brazilian Ministry of Science and Technology (MCT); PELD-CNPq (Proc. Nr.
558069/2009-6) and PROCAD-CAPES; the Brazilian National Council for Scientific and Tech-
nological Development (CNPq) and, the Tropical Ecology Assessment and Monitoring (TEAM)
Network, a collaboration between Conservation International, the Missouri Botanical Garden,
the Smithsonian Institution, and the Wildlife Conservation Society, and partially funded by
these institutions, the Gordon and Betty Moore Foundation, and other donors. For provision
of, or help in collecting data, we thank A. W. Graham, M. G. Bradford, A. Ford, D. Wilson,
34
K. Davies, M. Johnson, J. Grace, P. Meir, CSIRO and the Australian Canopy Crane Research
Station, James Cook University (Australia), E. Chavez, A. Sota, M. Steininger, J. S. Taborga,
(Bolivia); Rohden Ind ´
ustria Lignea Ltda, J. Barroso, W. Castro, E. Couto, C. A. Passos (de-
ceased), P. Nunes, D. Sasaki, D. M. de Freitas, M. Keller, G. B. da S. Oliveira, I. O. V. C. Feld-
pausch, L. Maracahipes, E. A. Oliveira, H. A. Mews, D. V. Silv ´
erio, M. Palace, M. Hunter,
M. Keller, Instituto de Pesquisa da Amazˆ
onia (IPAM), Woods Hole Research Center (WHRC)
and Grupo Amaggi at Fazenda Tanguro (Brazil); Guyana Forestry Commission, Iwokrama
International Centre for Rainforest Conservation and Development, M. Drescher (Guyana);
J. H. Ovalle, M. M. Sol´
orzano (Peru); R. Sukri, M. Salleh A. B. (Brunei); D. Burslem, C. May-
cock (Sabah); L. Chong, S. Davies, R. Shutine, L. K. Kho (Sarawak); for logistical aid and
maintenance of the large scale forest dynamics plots at Pasoh Forest Reserve, Malaysia
and Lambir Hills National Park, Sarawak, Malaysia, we thank, respectively, the Forest Re-
search Institute Malaysia (FRIM) and the Sarawak Forestry Corporation, Malaysia, the Cen-
ter for Tropical Forest Science – Arnold Arboretum Asia Program of the Smithsonian Tropical
Research Institute and Harvard University, USA and Osaka City University, Japan and their
funding agencies; V. O. Sene, J. Sonk ´
e, K. C. Nguembou; M.-N. Djuikouo K., R. Fotso and
Wildlife Conservation Society, Cameroon, ECOFAC-Cameroon, Cameroon Ministry Scientific
Research and Innovation, Cameroon Ministry of Forests and Fauna (MINFOF; Cameroon);
A. Moungazi, S. Mbadinga, H. Bourobou, L. N. Banak, T. Nzebi, K. Jeffery, SEGC/CIRMF/WCS
Research Station Lop´
e (Gabon); K. Ntim, K. Opoku, Forestry Commission of Ghana (Ghana);
A. K. Daniels, S. Chinekei, J. T. Woods, J. Poker, L. Poorter, Forest Development Authority
(Liberia), Valuing the Arc Program, Leverhulme Trust, N. Burgess, A. Balmford, A.R. Marshall,
P. K. T. Munishi (Tanzania). This research benefitted from the enthusiastic help of many field
assistants from across Africa, Asia, Australia, and South America.
References
Akaike, H.: A new look at the statistical model identification, IEEE Trans. Autom. Contr., 19,
716–723, 1974.
Alvarez, E., Duque, A., Saldarriaga, J., Cabrera, K., de las Salas, G., del Valle, I., Lema,
A., Moreno, F., Orrego, S., and Rodr´
ıguez, L.: Tree above-ground biomass allometries for
35
carbon stocks estimation in the natural forests of Colombia, For. Ecol. Manage., 267, 297–
308, http://dx.doi.org/10.1016/j.foreco.2011.12.013doi:10.1016/j.foreco.2011.12.013, 2012.
Anten, N. P. and Schieving, F.: The role of wood mass density and mechan-
ical constraints in the economy of tree architecture, Am. Nat., 175, 250–260,
http://dx.doi.org/10.1086/649581doi:10.1086/649581, 2010.
Aragao, L. E. and Shimabukuro, Y. E.: The incidence of fire in Amazonian forests with implica-
tions for REDD, Science, 328, 1275–1278, 10.1126/science.1186925, 2010.
Ara´
ujo, T. M., Higuchi, N., and Andrade de Carvalho J´
unior, J.: Comparison of formulae for
biomass content determination in a tropical rain forest site in the state of Par´
a, Brazil, For.
Ecol. Manage., 117, 43–52, 1999.
Asner, G. P., Powell, G. V. N., Mascaro, J., Knapp, D. E., Clark, J. K., Ja-
cobson, J., Kennedy-Bowdoin, T., Balaji, A., Paez-Acosta, G., Victoria, E.,
Secada, L., Valqui, M., and Hughes, R. F.: High-resolution forest carbon stocks
and emissions in the Amazon, Proc. Natl. Acad. Sci. USA, 107, 16738–16742,
http://dx.doi.org/10.1073/pnas.1004875107doi:10.1073/pnas.1004875107, 2010.
Baccini, A., Goetz, S. J., Walker, W. S., Laporte, N. T., Sun, M., Sulla-Menashe,
D., Hackler, J., Beck, P. S. A., Dubayah, R., Friedl, M. A., Samanta, S.,
and Houghton, R. A.: Estimated carbon dioxide emissions from tropical defor-
estation improved by carbon-density maps, Nature Clim. Change, 2, 182-185,
http://www.nature.com/nclimate/journal/v2/n3/abs/nclimate1354.htmlsupplementary-
information, 2012.
Bailey, R. L.: The potential of Weibull-type functions as flexible growth curves: discussion, Can.
J. For. Res., 10, 117–118, 1979.
Baker, T. R., Phillips, O. L., Malhi, Y., Almeida, S., Arroyo, L., Di Fiore, A., Erwin, T.,
Higuchi, N., Killeen, T. J., Laurance, S. G., Laurance, W. F., Lewis, S. L., Monteagudo,
A., Neill, D. A., Vargas, P. N., Pitman, N. C., Silva, J. N., and Martinez, R. V.: Increasing
biomass in Amazonian forest plots, Philos. Trans. R. Soc. Lond. B Biol. Sci., 359, 353–365,
http://dx.doi.org/10.1098/rstb.2003.1422doi:10.1098/rstb.2003.1422, 2004a.
Baker, T. R., Phillips, O. L., Malhi, Y., Almeida, S., Arroyo, L., Di Fiore, A., Erwin, T., Killeen,
T. J., Laurance, S. G., Laurance, W. F., Lewis, S. L., Lloyd, J., Monteagudo, A., Neill,
D. A., Patino, S., Pitman, N. C. A., Silva, J. N. M., and Martinez, R. V.: Variation in
wood density determines spatial patterns in Amazonian forest biomass, Global Change
Biol., 10, 545–562, http://dx.doi.org/10.1111/j.1529-8817.2003.00751.xdoi:10.1111/j.1529-
36
8817.2003.00751.x, 2004b.
Baker, T. R., Phillips, O. L., Laurance, W. F., Pitman, N. C. A., Almeida, S., Arroyo, L., DiFiore,
A., Erwin, T., Higuchi, N., Killeen, T. J., Laurance, S. G., Nascimento, H., Monteagudo, A.,
Neill, D. A., Silva, J. N. M., Malhi, Y., Gonzalez, G. L., Peacock, J., Quesada, C. A., Lewis,
S. L., and Lloyd, J.: Do species traits determine patterns of wood production in Amazonian
forests?, Biogeosciences, 6, 297-307, 2009.
Banin, L.: Cross-continental comparisons of tropical forest structure and function, PhD, School
of Geography, PhD Thesis. University of Leeds, Leeds, UK, 2010.
Banin, L., Feldpausch, T. R., Phillips, O. L., Baker, T. R., Lloyd, J., Affum-Baffoe, K., Arets, E.
J. M. M., Berry, N. J., Bradford, M., Breinen, R. J. W., Davies, S., Drescher, M., Higuchi, N.,
Hilbert, D., Hladik, A., Iida, Y., Silam, K. A., Kassim, A. R., King, D. A., Lopez-Gonzalez,
G., Metcalfe, D., Nilus, R., Peh, K. S.-H., Reitsma, J. M., Sonk ´
e, B., Taedoumg, H., Tan, S.,
White, L., W¨
oll, H., and Lewis, S. L.: What controls forest architecture? Testing environmen-
tal, structural and floristic drivers, Global Ecol. Biogeogr., in press, 2012.
Barbier, N., Couteron, P., Proisy, C., Malhi, Y., and Gastellu-Etchegorry, J. P.: The
variation of apparent crown size and canopy heterogeneity across lowland Ama-
zonian forests, Global Ecol. Biogeogr., 19, 72–84, http://dx.doi.org/10.1111/j.1466-
8238.2009.00493.xdoi:10.1111/j.1466-8238.2009.00493.x, 2010.
Bartholom ´
e, E. and Belward, A. S.: GLC2000: a new approach to global land cover mapping
from Earth observation data, Int. J. Remote Sens., 26, 1959–1977, 2005.
Baskerville, G.: Use of logarithmic regression in the estimation of plant biomass, Can. J. For.
Res., 2, 49–53, 1972.
Basuki, T. M., van Laake, P. E., Skidmore, A. K., and Hussin, Y. A.: Al-
lometric equations for estimating the above-ground biomass in tropi-
cal lowland Dipterocarp forests, For. Ecol. Manage., 257, 1684–1694,
http://dx.doi.org/10.1016/j.foreco.2009.01.027doi:10.1016/j.foreco.2009.01.027, 2009.
Brown, I. F., Martinelli, L. A., Thomas, W. W., Moreira, M. Z., Ferreira, C. A. C., and Victoria,
R. A.: Uncertainty in the biomass of Amazonian forests: an example from Rondonia, Brazil,
For. Ecol. Manage., 75, 175–189, 1995.
Brown, S., Gillespie, A. J. R., and Lugo, A. E.: Biomass estimation methods for tropical forests
with applications to forestry inventory data, For. Sci., 35, 881–902, 1989.
Brown, S.: Estimating biomass and biomass change of tropical forests: a primer, FAO Forestry
Paper, 134, 1997.
37
Buschbacher, R., Uhl, C., and Serr˜
ao, E. A. S.: Abandoned pastures in eastern Amazonia. II.
Nutrient stocks in the soil and vegetation, J. Ecol., 76, 682–699, 1988.
Carvalho, J. A. J., Higuchi, N., Ara´
ujo, T. M., and Santos, J. C.: Combustion completeness in
a rainforest clearing experiment in Manaus, Brazil, J. Geophys. Res., 103, 13195–113199,
1998.
Chambers, J. Q., dos Santos, J., Ribeiro, R. J., and Higuchi, N.: Tree damage, allometric
relationships, and above-ground net primary production in central Amazon forest, For. Ecol.
Manage., 152, 73–84, 2001.
Chave, J., Condit, R., Lao, S., Caspersen, J. P., Foster, R. B., and Hubbell, S. P.: Spatial and
temporal variation of biomass in a tropical forest: results from a large census plot in Panama,
J. Ecol., 91, 240–252, 2003.
Chave, J., Condit, R., Aguilar, S., Hernandez, A., Lao, S., and Perez, R.: Error propagation and
scaling for tropical forest biomass estimates, Philos Trans. R. Soc. Lond. B Biol. Sci., 359,
409–420, http://dx.doi.org/10.1098/rstb.2003.1425doi:10.1098/rstb.2003.1425, 2004.
Chave, J., Andalo, C., Brown, S., Cairns, M. A., Chambers, J. Q., Eamus, D., Folster, H., Fro-
mard, F., Higuchi, N., Kira, T., Lescure, J. P., Nelson, B. W., Ogawa, H., Puig, H., Riera,
B., and Yamakura, T.: Tree allometry and improved estimation of carbon stocks and bal-
ance in tropical forests, Oecologia, 145, 87–99, http://dx.doi.org/10.1007/s00442-005-0100-
xdoi:10.1007/s00442-005-0100-x, 2005.
Chave, J., Muller-Landau, H. C., Baker, T. R., Easdale, T. A., ter Steege, H., and Webb, C. O.:
Regional and phylogenetic variation of wood density across 2456 Neotropical tree species,
Ecol. Appl., 16, 2356–2367, 2006.
Chave, J., Coomes, D., Jansen, S., Lewis, S. L., Swenson, N. G., and Zanne, A.
E.: Towards a worldwide wood economics spectrum, Ecol. Lett., 12, 351–366,
http://dx.doi.org/10.1111/j.1461-0248.2009.01285.xdoi:10.1111/j.1461-0248.2009.01285.x,
2009.
Clark, D. A., Brown, S., Kicklighter, D. W., Chambers, J. Q., Thomlinson, J. R., and Ni, J.:
Measuring net primary production in forests: concepts and field methods, Ecol. Appl., 11,
356–370, 2001.
Crow, T. R.: Notes: common regressions to estimate tree biomass in tropical stands, For. Sci.,
24, 110–114, 1978.
Davies, S. J., Palmiotto, P. A., Ashton, P. S., Lee, H. S., and Lafrankie, J. V.: Comparative ecol-
ogy of 11 sympatric species of Macaranga in Borneo: tree distribution in relation to horizontal
38
and vertical resource heterogeneity, J. Ecol., 86, 662–673, 1998.
Deans, J. D., Moran, J., and Grace, J.: Biomass relationships for tree species in regenerating
semi-deciduous tropical moist forest in Cameroon, For. Ecol. Manage., 88, 215–225, 1996.
DeFries, R. S., Houghton, R. A., Hansen, M. C., Field, C. B., Skole, D., and Town-
shend, J.: Carbon emissions from tropical deforestation and regrowth based on satel-
lite observations for the 1980s and 1990s, Proc. Natl. Acad. Sci., 99, 14256–14261,
http://dx.doi.org/10.1073/pnas.182560099doi:10.1073/pnas.182560099, 2002.
Denman, K., Brasseur, G., Chidthaisong, A., Ciais, P., Cox, P., Dickinson, R., Hauglustaine,
D., Heinze, C., Holland, E., Jacob, D., Lohmann, U., Ramachandran, S., Pedro Leite da
Silva, D., Wofsy, S., and Zhang, X.: Couplings between changes in the climate system
and biogeochemistry, in: Climate Change 2007: The Physical Science Basis, Contribution of
Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate
Change, edited by: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B.,
Tignor, and Miller, H. L., Cambridge University Press, 499–587, 2007.
Djomo, A. N., Ibrahima, A., Saborowski, J., and Gravenhorst, G.: Allomet-
ric equations for biomass estimations in Cameroon and pan moist tropical equa-
tions including biomass data from Africa, For. Ecol. Manage., 260, 1873–1885,
http://dx.doi.org/10.1016/j.foreco.2010.08.034doi:10.1016/j.foreco.2010.08.034, 2010.
Drake, J. B., Dubayah, R. O., Clark, D. B., Knox, R. G., Blair, J. B., Hofton, M. A., Chazdon, R.
L., Weishampel, J. F., and Prince, S. D.: Estimation of tropical forest structural characteristics
using large-footprint lidar, Remote Sens. Environ., 79, 305–319, 2002.
Ebuy, J., Lokombe, J. P., Ponette, Q., Sonwa, D., and Picard, N.: Allometric equation for pre-
dicting aboveground biomass of three tree species, J. Trop. For. Sci., 23, 125–132, 2011.
Elias, M. and Potvin, C.: Assessing inter- and intra-specific variation in trunk carbon concentra-
tion for 32 neotropical tree species, Can. J. For. Res., 33, 1039–1045, 2003.
ESRI: ArcGIS Desktop: Release 10, http://www.esri.com/, Environmental Systems Research
Institute, Redlands, CA, USA, 2010.
Fang, Z. and Bailey, R. L.: Height-diameter models for tropical forests on Hainan Island in
southern China, For. Ecol. Manage., 110, 315–327, 1998.
Fearnside, P. M.: Wood density for estimating forest biomass in Brazilian Amazonia, For. Ecol.
Manage., 90, 59–87, 1997.
Feldpausch, T. R., Rondon, M. A., Fernandes, E. C. M., Riha, S. J., and Wandelli, E.: Carbon
and nutrient accumulation in secondary forests regenerating on pastures in central Ama-
39
zonia, Ecol. Appl., 14, S164–S176, http://dx.doi.org/10.1890/01-6015doi:10.1890/01-6015,
2004.
Feldpausch, T. R., Jirka, S., Passos, C. A. M., Jasper, F., and Riha, S.
J.: When big trees fall: damage and carbon export by reduced im-
pact logging in southern Amazonia, For. Ecol. Manage., 219, 199–215,
http://dx.doi.org/10.1016/j.foreco.2005.09.003doi:10.1016/j.foreco.2005.09.003, 2005.
Feldpausch, T. R., Coutoz, E. G., Rodrigues, L. C., Pauletto, D., Johnson, M.
S., Faheyk, T. J., Lehmann, J., and Riha, S. J.: Nitrogen aboveground turnover
and soil stocks to 8 m depth in primary and selectively logged forest in south-
ern Amazonia, Global Change Biol., 16, 1793–1805, http://dx.doi.org/10.1111/j.1365-
2486.2009.02068.xdoi:10.1111/j.1365-2486.2009.02068.x, 2010.
Feldpausch, T. R., Banin, L., Phillips, O. L., Baker, T. R., Lewis, S. L., Quesada, C. A., Affum-
Baffoe, K., Arets, E. J. M. M., Berry, N. J., Bird, M., Brondizio, E. S., de Camargo, P., Chave,
J., Djagbletey, G., Domingues, T. F., Drescher, M., Fearnside, P. M., Franca, M. B., Fyllas,
N. M., Lopez-Gonzalez, G., Hladik, A., Higuchi, N., Hunter, M. O., Iida, Y., Salim, K. A.,
Kassim, A. R., Keller, M., Kemp, J., King, D. A., Lovett, J. C., Marimon, B. S., Marimon, B.
H., Lenza, E., Marshall, A. R., Metcalfe, D. J., Mitchard, E. T. A., Moran, E. F., Nelson, B. W.,
Nilus, R., Nogueira, E. M., Palace, M., Pati˜
no, S., Peh, K. S. H., Raventos, M. T., Reitsma,
J. M., Saiz, G., Schrodt, F., Sonke, B., Taedoumg, H. E., Tan, S., White, L., Woll, H., and
Lloyd, J.: Height-diameter allometry of tropical forest trees, Biogeosciences, 8, 1081–1106,
http://dx.doi.org/10.5194/bg-8-1081-2011doi:10.5194/bg-8-1081-2011, 2011.
Fittkau, E. J.: Esboc¸o de uma divis ˜
ao ecolgica da regi˜
ao amazˆ
onica, Proc. Symp. Biol. Trop.
Amaz., , Florencia y Leticia, 1969, 1363-1372,1971.
Fittkau, E. J. and Klinge, H.: On biomass and trophic structure of the central Amazonian rain
forest ecosystem, Biotropica, 5, 2–14, 1973.
Gibbs, H. K., Brown, S., Niles, J. O., and Foley, J. A.: Monitoring and estimating tropical forest
carbon stocks: making REDD a reality, Environ. Res. Lett., 2, 1–13, 2007.
Graham, A. W.: The CSIRO rainforest permanent plots of North Queensland : site, structural,
floristic and edaphic descriptions, Accessed from http://nla.gov.au/nla.cat-vn3708155, Rain-
forest CRC, James Cook University, Cairns, Australia, 227 pp., 2006.
Griscom, B. W. and Ashton, P. M. S.: A self-perpetuating bamboo dis-
turbance cycle in a neotropical forest, J. Trop. Ecol., 22, 587–597,
http://dx.doi.org/10.1017/S0266467406003361doi:10.1017/S0266467406003361, 2006.
40
Henry, M., Besnard, A., Asante, W. A., Eshun, J., Adu-Bredu, S., Valentini, R., Bernoux, M.,
and Saint-Andre, L.: Wood density, phytomass variations within and among trees, and al-
lometric equations in a tropical rainforest of Africa, For. Ecol. Manage., 260, 1375–1388,
http://dx.doi.org/10.1016/j.foreco.2010.07.040doi:10.1016/j.foreco.2010.07.040, 2010.
Higuchi, N., dos Santos, J., Ribeiro, R. J., Minette, L., and Biot, Y.: Biomassa da parte a ´
erea da
vegetac¸ ˜
ao da floresta tropical ´
umida de terra-firme da Amaz ˆ
onia brasileira, Acta Amazon.,
28, 153–166, 1998.
Houghton, R.: Carbon flux to the atmosphere from land-use changes: 1850–2005, in:
TRENDS: a compendium of data on global change, Carbon Dioxide Information Analysis
Center, US Department of Energy, Oak Ridge, Tenn., USA, 2008.
Houghton, R. A., Skole, D. L., Nobre, C. A., Hackler, J. L., Lawrence, K. T., and Chomentowski,
W. H.: Annual fluxes of carbon from deforestation and regrowth in the Brazilian Amazon,
Nature, 403, 301–304, 2000.
Houghton, R. A.: How well do we know the flux of CO2from land-use change?, Tellus B, 62,
337-351, 10.1111/j.1600-0889.2010.00473.x, 2010.
Houghton, R. A.: Revised estimates of the annual net flux of carbon to the at-
mosphere from changes in land use and land management 1850–2000, Tellus
B, 55, 378–390, http://dx.doi.org/10.1034/j.1600-0889.2003.01450.xdoi:10.1034/j.1600-
0889.2003.01450.x, 2003.
Hozumi, K., Yoda, K., Kokawa, S., and Kira, T.: Production ecology of tropical rain forests in
southwestern Cambodia. I. Plant biomass, Nature and Life in Southeast Asia, 6, 1–51, 1969.
Iida, Y., Kohyama, T. S., Kubo, T., Kassim, A. R., Poorter, L., Sterck, F., and
Potts, M. D.: Tree architecture and life-history strategies across 200 co-occurring
tropical tree species, Funct. Ecol., 25, 1260–1268, http://dx.doi.org/10.1111/j.1365-
2435.2011.01884.xdoi:10.1111/j.1365-2435.2011.01884.x, 2011.
Iida, Y., Poorter, L., Sterck, F. J., Kassim, A. R., Kubo, T., Potts, M. D., and Ko-
hyama, T. S.: Wood density explains architectural differentiation across 145 co-
occurring tropical tree species, Funct. Ecol., 26, 274–282, http://dx.doi.org/10.1111/j.1365-
2435.2011.01921.xdoi:10.1111/j.1365-2435.2011.01921.x, 2012.
INPE: Monitoramento da floresta Amazˆ
onica Brasileira por sat´
elite 2009, available online at:
http://www.inpe.br, Instituto Nacional de Pesquisas Espaciais (INPE), S ˜
ao Jos´
e dos Cam-
pos, Brazil, 2009.
IPCC: 2006 IPCC guidelines for national greenhouse gas inventories, Institute for Global Envi-
41
ronmental Strategies, Hayama, Japan, 2006.
Kindermann, G., Obersteiner, M., Sohngen, B., Sathaye, J., Andrasko, K., Rametsteiner, E.,
Schlamadinger, B., Wunder, S., and Beach, R.: Global cost estimates of reducing carbon
emissions through avoided deforestation, Proc. Natl. Acad. Sci. USA, 105, 10302–10307,
http://dx.doi.org/10.1073/pnas.0710616105doi:10.1073/pnas.0710616105, 2008.
Ketterings, Q. M., Coe, R., van Noordwijk, M., Ambagau, Y., and Palm, C. A.: Reducing uncer-
tainty in the use of allometric biomass equations for predicting above-ground tree biomass
in mixed secondary forests, For. Ecol. Manage., 146, 199-209, 2001.
Laurance, W. F. and Curran, T. J.: Impacts of wind disturbance on fragmented tropical
forests: A review and synthesis, Austral Ecol., 33, 399–408, http://dx.doi.org/10.1111/j.1442-
9993.2008.01895.xdoi:10.1111/j.1442-9993.2008.01895.x, 2008.
Lefsky, M. A., Harding, D. J., Keller, M., Cohen, W. B., Carabajal, C. C., Espirito-Santo, F. D. B.,
Hunter, M. O., and de Oliveira Jr., R.: Estimates of forest canopy height and aboveground
biomass using ICESat, Geophys. Res. Lett., 32, 1–4, 2005.
Lescure, J.-P., Puig, H., Ri ´
era, B., Leclerc, D., Beekman, A., and B ´
en´
eteau, A.: La phytomasse
´
epig´
ee d’une forˆ
et dense en Guyane franc¸ aise, Acta Oecol., 4, 237–251, 1983.
Lewis, S. L., Lopez-Gonzalez, G., Sonke, B., Affum-Baffoe, K., Baker, T. R., Ojo, L. O., Phillips,
O. L., Reitsma, J. M., White, L., Comiskey, J. A., Djuikouo, K. M., Ewango, C. E., Feld-
pausch, T. R., Hamilton, A. C., Gloor, M., Hart, T., Hladik, A., Lloyd, J., Lovett, J. C., Makana,
J. R., Malhi, Y., Mbago, F. M., Ndangalasi, H. J., Peacock, J., Peh, K. S., Sheil, D., Sun-
derland, T., Swaine, M. D., Taplin, J., Taylor, D., Thomas, S. C., Votere, R., and Woll,
H.: Increasing carbon storage in intact African tropical forests, Nature, 457, 1003–1006,
http://dx.doi.org/10.1038/nature07771doi:10.1038/nature07771, 2009.
Lohman, D. J., de Bruyn, M., Page, T., von Kintelen, K., Hall, R., Ng, P. K. L., and von Rintelen,
T.: Biogeography of the Indo-Australian archipeligo, Ann. Rev. Ecol. Syst., 42, 205–206,
2011.
Lopez-Gonzalez, G., Lewis, S. L., Burkitt, M., and Phillips, O. L.: ForestPlots.net:
a web application and research tool to manage and analyse tropical forest plot
data, Journal of Vegetation Science, 22, 610–613, http://dx.doi.org/10.1111/j.1654-
1103.2011.01312.xdoi:10.1111/j.1654-1103.2011.01312.x, 2011.
Lucas, R. M., Honzak, M., Amaral, I. D., Curran, P. J., and Foody, G. M.: Forest regeneration
on abandoned clearances in central Amazonia, Int. J. Remote Sens., 23, 965–988, 2002.
Mackensen, J., Tillery-Stevens, M., Klinge, R., and F ¨
olster, H.: Site parameters, species com-
42
position, phytomass structure and element stores of a terra-firme forest in East-Amazonia,
Brazil, Plant Ecol., 151, 101–119, 2000.
Malhi, Y., Baker, T. R., Phillips, O. L., Almeida, S., Alvarez, E., Arroyo, L., Chave, J., Czimczik,
C. I., Di Fiore, A., Higuchi, N., Killeen, T. J., Laurance, S. G., Laurance, W. F., Lewis, S. L.,
Montoya, L. M. M., Monteagudo, A., Neill, D. A., Vargas, P. N., Patino, S., Pitman, N. C. A.,
Quesada, C. A., Salomao, R., Silva, J. N. M., Lezama, A. T., Martinez, R. V., Terborgh, J.,
Vinceti, B., and Lloyd, J.: The above-ground coarse wood productivity of 104 Neotropical
forest plots, Global Change Biol., 10, 563–591, 2004.
Malhi, Y. and Wright, J.: Spatial patterns and recent trends in the climate of tropical rainforest
regions, Phil. Trans. R. Soc. Lond. B., 359, 311–329, 2004.
Malhi, Y., Wood, D., Baker, T. R., Wright, J., Phillips, O. L., Cochrane, T., Meir, P., Chave, J.,
Almeida, S., Arroyo, L., Higuchi, N., Killeen, T. J., Laurance, S. G., Laurance, W. F., Lewis, S.
L., Monteagudo, A., Neill, D. A., Vargas, P. N., Pitman, N. C. A., Quesada, C. A., Salomao,
R., Silva, J. N. M., Lezama, A. T., Terborgh, J., Martinez, R. V., and Vinceti, B.: The regional
variation of aboveground live biomass in old-growth Amazonian forests, Global Change Biol.,
12, 1107–1138, 2006.
Marshall, A. R., Willcock, S., Platts, P. J., Lovett, J. C., Balmford, A., Burgess, N. D., Latham,
J. E., Munishi, P. K.T., Salter, R., Shirima, D. D., and Lewis, S. L.: Measuring and modelling
above-ground carbon and tree allometry along a tropical elevation gradient. Biol. Conserv.,
http://dx.doi.org/10.1016/j.biocon.2012.03.017, 2012.
Martin, A. R. and Thomas, S. C.: A reassessment of car-
bon content in tropical trees, PLoS ONE, 6, e23533,
hrefhttp://dx.doi.org/10.1371/journal.pone.0023533doi:10.1371/journal.pone.0023533,
2011.
Martinelli, L. A., Almeida, S., Brown, I. F., Moreira, M. Z., Victoria, R. L., Filoso, S., Ferreira,
C. A. C., and Thomas, W. W.: Variation in nutrient distribution and potential nutrient losses
by selective logging in a humid tropical forest of Rondonia, Brazil, Biotropica, 32, 597–613,
2000.
Mayaux, P., Holmgren, P., Achard, F., Eva, H., Stibig, H., and Branthomme, A.:
Tropical forest cover change in the 1990s and options for future monitoring, Philo-
sophical Transactions of the Royal Society B-Biological Sciences, 360, 373–384,
http://dx.doi.org/10.1098/rstb.2004.1590doi:10.1098/rstb.2004.1590, 2005.
Miles, L. and Kapos, V.: Reducing greenhouse gas emissions from deforestation and for-
43
est degradation: global land-use implications, Science, 320, 1454–1455, 10.1126/sci-
ence.1155358, 2008.
Mitchard, E. T. A., Saatchi, S. S., Woodhouse, I. H., Nangendo, G., Ribeiro, N. S.,
Williams, M., Ryan, C. M., Lewis, S. L., Feldpausch, T. R., and Meir, P.: Using
satellite radar backscatter to predict above-ground woody biomass: a consistent re-
lationship across four different African landscapes, Geophys. Res. Lett., 36, L23401,
http://dx.doi.org/10.1029/2009GL040692doi:10.1029/2009GL040692, 2009.
Mitchard, E. T. A., Saatchi, S. S., Lewis, S. L., Feldpausch, T. R., Woodhouse, I. H., Sonk´
e,
B., Rowland, C., and Meir, P.: Measuring biomass changes due to woody encroachment
and deforestation/degradation in a forest–savanna boundary region of central Africa us-
ing multi-temporal L-band radar backscatter, Remote Sens. Environ., 115, 2861–2873,
http://dx.doi.org/10.1016/j.rse.2010.02.022doi:10.1016/j.rse.2010.02.022, 2011.
Montgomery, R. A. and Chazdon, R. L.: Forest structure, canopy architecture, and light trans-
mittance in tropical wet forests, Ecology, 82, 2707–2718, 2001.
Morel, A. C., Saatchi, S. S., Malhi, Y., Berry, N. J., Banin, L., Burslem, D., Nilus, R.,
and Ong, R. C.: Estimating aboveground biomass in forest and oil palm plantation in
Sabah, Malaysian Borneo using ALOS PALSAR data, For. Ecol. Manage., 262, 1786-1798,
10.1016/j.foreco.2011.07.008, 2011.
Nelson, B. W., Mesquita, R., Pereira, J. L. G., de Souza, S. G. A., Batista, G. T., and Couto, L.
B.: Allometric regressions for improved estimate of secondary forest biomass in the central
Amazon, For. Ecol. Manage., 117, 149–167, 1999.
Nogueira, E. M., Fearnside, P. M., Nelson, B. W., and Franca, M. B.: Wood den-
sity in forests of Brazil’s ’arc of deforestation’: implications for biomass and flux
of carbon from land-use change in Amazonia, For. Ecol. Manage., 248, 119–135,
http://dx.doi.org/10.1016/j.foreco.2007.04.047doi:10.1016/j.foreco.2007.04.047, 2007.
Nogueira, E. M., Fearnside, P. M., Nelson, B. W., Barbosa, R. I., and Keizer, E. W. H.: Estimates
of forest biomass in the Brazilian Amazon: new allometric equations and adjustments to
biomass from wood-volume inventories, For. Ecol. Manage., 256, 1853–1867, 2008a.
Nogueira, E. M., Nelson, B. W., Fearnside, P. M., Franc¸ a, M. B., and Oliveira, ´
A. C. A. d.: Tree
height in Brazil’s ’arc of deforestation’: Shorter trees in south and southwest Amazonia imply
lower biomass, For. Ecol. Manage., 255, 2963–2972, 2008b.
Ogawa, H., Yoda, K., and Kira, T.: Comparative ecological studies on three main types of forest
vegetation in Thailand: II. Plant biomass, Nature and Life in South East Asia, 4, 49–80, 1965.
44
Overman, J. P. M., Witte, H. J. L., and Saldarriaga, J. G.: Evaluation of regression models
for above-ground biomass determination in Amazon rainforest, J. Trop. Ecol., 10, 207–218,
1994.
Pati˜
no, S., Lloyd, J., Paiva, R., Baker, T. R., Quesada, C. A., Mercado, L. M., Schmerler, J.,
Schwarz, M., Santos, A. J. B., Aguilar, A., Czimczik, C. I., Gallo, J., Horna, V., Hoyos, E. J.,
Jimenez, E. M., Palomino, W., Peacock, J., Pe˜
na-Cruz, A., Sarmiento, C., Sota, A., Turriago,
J. D., Villanueva, B., Vitzthum, P., Alvarez, E., Arroyo, L., Baraloto, C., Bonal, D., Chave, J.,
Costa, A. C. L., Herrera, R., Higuchi, N., Killeen, T., Leal, E., Luiz ˜
ao, F., Meir, P., Monteagudo,
A., Neil, D., N´
u˜
nez-Vargas, P., Pe ˜
nuela, M. C., Pitman, N., Priante Filho, N., Prieto, A., Panfil,
S. N., Rudas, A., Salom˜
ao, R., Silva, N., Silveira, M., Soares deAlmeida, S., Torres-Lezama,
A., V´
asquez-Mart´
ınez, R., Vieira, I., Malhi, Y., and Phillips, O. L.: Branch xylem density
variations across the Amazon Basin, Biogeosciences, 6, 545–568, 10.5194/bg-6-545-2009,
2009.
Peacock, J., Baker, T. R., Lewis, S. L., Lopez-Gonzalez, G., and Phillips, O. L.: The RAINFOR
database: monitoring forest biomass and dynamics, Journal of Vegetation Science, 18, 535–
542, 2007.
Phillips, O. L., Malhi, Y., Higuchi, N., Laurance, W. F., N ´
u˜
nez, P. V., V´
asquez, R. M., Laurance,
S. G., Ferreira, L. V., Stern, M., Brown, S., and Grace, J.: Changes in the carbon balance of
tropical forests: evidence from long-term plots, Science, 282, 439–442, 1998.
Phillips, O. L., Baker, T. R., Arroyo, L., Higuchi, N., Killeen, T. J., Laurance, W. F., Lewis, S. L.,
Lloyd, J., Malhi, Y., Monteagudo, A., Neill, D. A., Vargas, P. N., Silva, J. N. M., Terborgh, J.,
Martinez, R. V., Alexiades, M., Almeida, S., Brown, S., Chave, J., Comiskey, J. A., Czimczik,
C. I., Di Fiore, A., Erwin, T., Kuebler, C., Laurance, S. G., Nascimento, H. E. M., Olivier, J.,
Palacios, W., Pati˜
no, S., Pitman, N. C. A., Quesada, C. A., Salidas, M., Lezama, A. T., and
Vinceti, B.: Pattern and process in Amazon tree turnover, 1976–2001, Phil. Trans. Roy. Soc.
Lnd. S. B-Bio Sci., 359, 381–407, 2004.
Phillips, O. L., Aragao, L. E., Lewis, S. L., Fisher, J. B., Lloyd, J., Lopez-Gonzalez, G., Malhi,
Y., Monteagudo, A., Peacock, J., Quesada, C. A., van der Heijden, G., Almeida, S., Ama-
ral, I., Arroyo, L., Aymard, G., Baker, T. R., Banki, O., Blanc, L., Bonal, D., Brando, P.,
Chave, J., de Oliveira, A. C., Cardozo, N. D., Czimczik, C. I., Feldpausch, T. R., Freitas,
M. A., Gloor, E., Higuchi, N., Jimenez, E., Lloyd, G., Meir, P., Mendoza, C., Morel, A.,
Neill, D. A., Nepstad, D., Pati˜
no, S., Penuela, M. C., Prieto, A., Ramirez, F., Schwarz, M.,
Silva, J., Silveira, M., Thomas, A. S., Steege, H. T., Stropp, J., V´
asquez, R., Zelazowski,
45
P., Alvarez Davila, E., Andelman, S., Andrade, A., Chao, K. J., Erwin, T., Di Fiore, A.,
Honorio, C. E., Keeling, H., Killeen, T. J., Laurance, W. F., Pena Cruz, A., Pitman, N. C.,
N´
u˜
nez Vargas, P., Ramirez-Angulo, H., Rudas, A., Salamao, R., Silva, N., Terborgh, J., and
Torres-Lezama, A.: Drought sensitivity of the Amazon rainforest, Science, 323, 1344–1347,
http://dx.doi.org/10.1126/science.1164033doi:10.1126/science.1164033, 2009.
Phillips, O. L., Baker, T. R., Brienen, R., and Feldpausch, T. R.: Field manual for plot establish-
ment and remeasurement, available online at: http://www.geog.leeds.ac.uk/projects/rainfor/,
last access: February 2012, 2010.
Pinard, M. A. and Putz, F. E.: Retaining forest biomass by reducing logging damage, Biotropica,
28, 278–295, 1996.
Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., and the R Development Core Team: nlme:
Linear and nonlinear mixed effects models, R package version 3.1-102, 2011.
Pinheiro, J.C., Bates, D.M. Lindstrom, M.J.: Model building in nonlinear mixed effects models.
Department of Statistics, Technical Report 931, University of Wisconsin, Madison. 1994.
Pitman, N. C. A., Terborgh, J., Silman, M. R., and N ´
u˜
nez, P.: Tree species distributions in an
upper Amazonian forest, Ecology, 80, 2651–2661, 1999.
Poorter, L., Bongers, L., and Bongers, F.: Architecture of 54 moist-forest tree species: Traits,
trade-offs, and functional groups, Ecology, 87, 1289–1301, 2006.
Quesada, C. A., Lloyd, J., Schwarz, M., Baker, T. R., Phillips, O. L., Pati ˜
no, S., Czimczik, C.,
Hodnett, M. G., Herrera, R., Arneth, A., Lloyd, G., Malhi, Y., Dezzeo, N., Luiz˜
ao, F. J., Santos,
A. J. B., Schmerler, J., Arroyo, L., Silveira, M., Priante Filho, N., Jimenez, E. M., Paiva,
R., Vieira, I., Neill, D. A., Silva, N., Pe ˜
nuela, M. C., Monteagudo, A., V ´
asquez, R., Prieto,
A., Rudas, A., Almeida, S., Higuchi, N., Lezama, A. T., L ´
opez-Gonz´
alez, G., Peacock, J.,
Fyllas, N. M., Alvarez D´
avila, E., Erwin, T., di Fiore, A., Chao, K. J., Honorio, E., Killeen, T.,
Pe˜
na Cruz, A., Pitman, N., N´
u˜
nez Vargas, P., Salom ˜
ao, R., Terborgh, J., and Ram´
arez, H.:
Regional and large-scale patterns in Amazon forest structure and function are mediated by
variations in soil physical and chemical properties, Biogeosciences Discuss., 6, 3993-4057,
http://dx.doi.org/10.5194/bgd-6-3993-2009doi:10.5194/bgd-6-3993-2009, 2009.
R Development Core Team: R: A language and environment for statistical computing, R Foun-
dation for Statistical Computing, URL http://www.R-project.org, Vienna, Austria, 2011.
Rich, P. M., Helenurm, K., Kearns, D., Morse, S. R., Palmer, M. W., and Short, L.: Height and
stem diameter relationships for dicotyledonous trees and arborescent palms of Costa Rican
tropical wet forest, Bull. Torrey Bot. Club, 113, 241–246, 1986.
46
Saatchi, S. S., Harris, N. L., Brown, S., Lefsky, M., Mitchard, E. T., Salas, W., Zutta, B. R.,
Buermann, W., Lewis, S. L., Hagen, S., Petrova, S., White, L., Silman, M., and Morel, A.:
Benchmark map of forest carbon stocks in tropical regions across three continents, Proc.
Natl. Acad. Sci. USA, 108, 9899–9904, 10.1073/pnas.1019576108, 2011.
Samalca, I. K.: Estimation of forest biomass and its error: a case in Kalimantan, Indone-
sia, MSc., International Institute for Geo-information Science and Earth Observation, The
Netherlands, 74 pp., 2007.
Schobbenhaus, C., and A. Bellizzia (coord.): Geological Map of South America, 1:5,000,000,
CGMW-CPRM-DNPM-UNESCO, Paris, 2001.
Shuttleworth, W. J.: Evaporation from Amazonian rainforest, Proc. Roy. Soc. Ldn. Ser. B. Biol.
Sci., 233, 321–346, http://dx.doi.org/10.1098/rspb.1988.0024doi:10.1098/rspb.1988.0024,
1988.
Slik, J. W. F., Aiba, S.-I., Brearley, F. Q., Cannon, C. H., Forshed, O., Kitayama, K., Nagamasu,
H., Nilus, R., Payne, J., Paoli, G., Poulsen, A. D., Raes, N., Sheil, D., Sidiyasa, K., Suzuki, E.,
and van Valkenburg, J. L. C. H.: Environmental correlates of tree biomass, basal area, wood
specific gravity and stem density gradients in Borneo’s tropical forests, Global Ecol. Bio-
geogr., 19, 50–60, http://dx.doi.org/10.1111/j.1466-8238.2009.00489.xdoi:10.1111/j.1466-
8238.2009.00489.x, 2010.
Sprugel, D.: Correcting for bias in log-transformed allometric equations, Ecology, 64, 209–210,
1983.
Steininger, M. K.: Satellite estimation of tropical secondary forest above-ground biomass: data
from Brazil and Bolivia, Int. J. Remote Sens., 21, 1139–1157, 2000.
Thomas, S. C.: Asymptotic height as a predictor of growth and allometric characteristics in
Malaysian rain forest trees, Am. J. Bot., 83, 556–566, 1996.
Thomas, S. C. and Bazzaz, F. A.: Asymptotic height as a predictor of photosynthetic character-
istics in Malaysian forest trees, Ecology, 80, 1607–1622, 1999.
Torello-Raventos, M., Feldpausch, T. R., Veenendaal, E. M., Schrodt, F., Saiz, G., Domingues,
T. F., Djagbletey, G., Ford, A., Kemp, J., Marimon, B. S., Marimon-Junior, B. H., Lenza, E.,
Ratter, J. A., Maracahipes, L., Sasaki, D., Sonk ´
e, B., Zapfackv, L., Villarroel, D., Schwarz,
M., Quesada, C. A., Ishida, F. Y., Nardoto, G. B., Affum-Baffoe, K., Arroyo, L., Bowman, D. J.
M. S., Compaore, H., Davies, K., Diallo, A., Fyllas, N. M., Gilpin, M., Hien, F., Johnson, M.,
Killeen, T. J., Metcalfe, D., Miranda, H. S., Steininger, M., Thomson, J., Sykorac, K., Bird, M.
I., Grace, J., Lewis, S. l., Phillips, O. L., and Lloyd, J.: Characterising and classifying tropical
47
vegetation types with an emphasis of forest/savanna transitions, Plant Ecology and Diversity,
in review, 2012.
Uhl, C. and Jordan, C. F.: Succession and nutrient dynamics following forest cutting and burning
in Amazonia, Ecology, 65, 1476–1490, 1984.
UK Department of Energy and Climate Change: 5NC: The UK’s Fifth National Communication
under the United Nations Framework Convention on Climate Change, Department of Energy
and Climate Change, London, UK, 153, 2009.
Vieilledent, G., Vaudry, R., Andriamanohisoa, S. F. D., Rakotonarivo, S. O., Randrianasolo, Z.
H., Razafindrabe, H. N., Bidaud Rakotoarivony, C., Ebeling, J., and Rasamoelina, M.: A
universal approach to estimate biomass and carbon stock in tropical forests using generic
allometric models, Ecol. Appl., http://dx.doi.org/10.1890/11-0039.1doi:10.1890/11-0039.1,
preprint, 2011.
Watson, R. T., Noble, I. R., Bolin, B., Ravindranath, N. H., Verardo, D. J., and Dokken, D. J.:
Land use, land-use change, and forestry: a special report of the Intergovernmental Panel
on Climate Change, edited by: Watson, R. T., Noble, I. R., Bolin, B., Ravindranath, N. H.,
Verardo, David, J., and Dokken, D. J., Intergovernmental Panel on Climate Change, http:
//www.grida.no/publications/other/ipcc sr/, Cambridge Univ Press, Cambridge, UK, 377 pp.,
2000.
Yamakura, T., Hagihara, A., Sukardjo, S., and Ogawa, H.: Aboveground biomass of tropical
rain forest stands in Indonesian Borneo, Plant Ecol., 68, 71–82, 1986.
Zanne, A. E., Westoby, M., Falster, D. S., Ackerly, D. D., Loarie, S. R., Arnold, S. E.
J., and Coomes, D. A.: Angiosperm wood structure: Global patterns in vessel anatomy
and their relation to wood density and potential conductivity, Am. J. Bot., 97, 207–215,
http://dx.doi.org/10.3732/ajb.0900178doi:10.3732/ajb.0900178, 2010.
48
Table 1. Pantropical models to estimate dry biomass (kg) from, Eq. (1) diameter (D, cm) and
wood specific gravity (ρW, g cm−3), and Eq. (2) also including tree height (H, m) for trees in
pantropical forests, including the residual standard error (RSE), Akaike information criterion
(AIC) and number of trees (n) based on destructively sampled moist forest tree data from
Africa, Asia, and South America.
Model ab c d e RSE R2AIC n
Eq. (1): ln(B)=a+bln (D)+c(ln (D))2+d(ln (D)3+eln(ρW)
−1.8222 2.3370 0.1632 −0.0248 0.9792 0.3595 0.973 1444 1816
Eq. (2): ln(B)=a+bln(D2ρWH)
−2.9205 0.9894 – – – 0.3222 0.978 1044 1816
49
Table 2. Efficacy of biomass models including or excluding tree Hto predict true (destructively)
sampled biomass for trees ≥10 cm Dfor individual sites excluded from model formulation.
Values represent bootstrapped mean relative error, or bias ((Bpredicted–Bmeasured )/Bmeasured) for
a site, in dry biomass estimated from a biomass model excluding H(Eq. 1) and biomass
including H(Eq. 2) using various Hmodels (Eqs. 3–5) based on region- and continent- specific
Hmodels. Values in bold indicate the model with the lowest mean relative error (bias) for a
site (this excludes the power model, which although has the lowest overall bias and standard
deviation, fails to reduce error in the small diameter classes).∗
3PE Weibull Power No Ht Data source
Dropped Site∗∗ Location Region nContinent Region Continent Region Continent Region
BraCot Cotriguac¸u,
Par´
a, Brazil
Brazilian
Shield
151 0.01 −0.02 0.01 −0.09 −0.04 −0.07 0.09 Nogueira et al. (2008)
BraJuruena Juruena, Mato
Grosso, Brazil
Brazilian
Shield
49 −0.04 −0.06 −0.05 −0.13 −0.08 −0.11 0.05 Nogueira et al. (2008)
BraMan1 Manaus, Ama-
zonas, Brazil
E.-central
Amazonia
315 0.01 −0.07 −0.05 −0.14 −0.05 −0.13 −0.01 Chave et al. (2005)
BraMan2 Manaus, Ama-
zonas, Brazil
E.-Central
Amazonia
123 0.05 −0.03 0.04 −0.06 0.00 −0.09 0.13 Chave et al. (2005)
BraNPro Novo Progesso,
Mato Grosso,
Brazil
Brazilian
Shield
64 −0.22 −0.23 −0.25 −0.30 −0.25 −0.28 −0.20 Nogueira et al. (2008)
BraPara1 Tom ´
e Ac¸u,
Par´
a, Brazil
Brazilian
Shield
127 −0.04 −0.12 −0.02 −0.10 −0.08 −0.16 0.07 Araujo et al. (1999)
BraPara3 Belem, Par ´
a,
Brazil
Brazilian
Shield
21 −0.14 −0.21 −0.09 −0.16 −0.18 −0.25 0.01 Chave et al. 2005
BraRond Rˆ
ondonia,
Brazil
Brazilian
Shield
8−0.50 −0.53 −0.46 −0.53 −0.52 −0.54 −0.39 Brown et al. 1995
FrenchGu Piste St. Elie,
French Guiana
Guyana
Shield
360 0.48 0.77 0.37 0.53 0.40 0.73 0.47 Chave et al. (2005)
Llanosec Llanos
secondary
Western
Amazonia
24 0.47 0.79 0.45 0.66 0.40 0.73 0.61 Chave et al. 2005
Llanosol Llanos
old-growth
Western
Amazonia
27 0.10 0.35 0.17 0.35 0.07 0.35 0.32 Chave et al. (2005)
CamCampo-Ma’an Campo-Ma’an,
Cameroon
Central
Africa
71 0.15 0.34 −0.01 0.22 0.03 0.24 0.13 Djomo et al. (2010)
CamMbalmayo Mbalmayo,
Cameroon
Central
Africa
40.09 0.11 0.15 0.29 0.04 0.05 0.33 Deans et al. (1996)
50
Table 2. Continued.
3PE Weibull Power No Ht Data source
Dropped Site∗∗ Location Region nContinent Region Continent Region Continent Region
DRCY
angambi Yangambi,
Democratic
Republic of
Congo
Central
Africa
12 −0.07 −0.04 −0.01 0.12 −0.13 -0.11 0.13 Ebuy et al. (2011)
GhaBoiTano Boi Tano,
Ghana
Western
Africa
41 −0.18 −0.14 −0.13 −0.13 −0.14 −0.10 −0.01 Henry et al. (2010)
IndoMala South-east
Asia
119 0.55 0.55 0.37 0.37 0.45 0.45 0.53
Kaliman1 Kalimantan,
Balikpapan,
Indonesia
South-east
Asia
23 −0.04 −0.04 −0.02 −0.02 −0.07 −0.07 0.01 Chave et al. (2005)
Kaliman2 Kalimantan,
Sebulu,
Indonesia
South-east
Asia
69 −0.11 −0.11 −0.18 −0.18 −0.15 −0.15 −0.13 Yamakura et al. (1986)
Kaliman3 PT Hutan
Labanan
Sanggam
Lestari, Kali-
mantan,
Indonesia
South-east
Asia
40 −0.08 −0.08 −0.07 −0.07 −0.12 −0.12 −0.03 Samalca 2007
Pasoh-01 Pasoh,
Malaysia
South-east
Asia
139 −0.07 −0.07 −0.13 −0.13 −0.11 −0.11 −0.09 Chave et al. (2005)
Sumatra Sepunggur,
Sumatra,
Indonesia
South-east
Asia
29 0.27 0.27 0.26 0.26 0.22 0.22 0.33 Ketterings et al. (2001)
Relative mean error 0.03 0.05 0.06 0.06 −0.01 0.02 0.13
Std. Dev. 0.25 0.33 0.22 0.29 0.23 0.32 0.25
∗Biomass estimated from models based on tree diameter, wood density (Eqn. 1) and where
applicable, H(Eqn. 2). Height is estimated from models developed from the pantropical tree
H−Ddatabase of Feldpausch et al. (2011).
∗∗ Efficacy of the biomass model to predict biomass was independently assessed for each
“dropped site” which was exlcuded from the development of the biomass model.
51
Table 3. Coefficients for Weibull-Hregion-, continent-specific and pantropical models (H=
a∗(1−exp(−b∗Dc))) to estimate tree height (H, m) from diameter (D, cm) ≥10 cm in pantropical
forests, including the residual standard error (RSE), Akaike information criterion (AIC), and
number of trees (n)∗.
Continent Region a b c RSE AIC n
Africa 50.096 0.03711 0.8291 5.739 75 422 11 910
C. Africa 50.453 0.0471 0.8120 6.177 16 671 2572
E. Africa 43.974 0.0334 0.8546 5.466 10 343 1658
W. Africa 53.133 0.0331 0.8329 5.165 47 020 7680
S. America 42.574 0.0482 0.8307 5.619 121 167 19 262
Brazilian Shield 227.35∗∗ 0.0139 0.5550 4.683 20639 3482
E. C. Amazonia 48.131 0.0375 0.8228 4.918 39688 6588
Guyana Shield 42.845 0.0433 0.9372 5.285 32 491 5267
W. Amazonia 46.263 0.0876 0.6072 5.277 24 201 3925
Asia S. E. Asia 57.122 0.0332 0.8468 5.691 18 623 2948
Australia N. Australia 41.721 0.0529 0.7755 4.042 48 073 8536
Pantropical 50.874 0.0420 0.784 5.479 266 169 42656
∗Models adapted from the pantropical tree H:Ddatabase of Feldpausch et al. (2011).
∗∗ Although an unrealistic asymptotic maximum Hcoefficient (a), trees of 10 and 160 cm
diameter would have an estimated Hof 11.0 and 47.2m, respectively, with this model.
52
Table 4. Pantropical live tree above ground dry biomass (B) estimates (bootstrap mean
(Mg ha−1)±95% CI) when calculating as column (a) biomass estimated as per most published
studies excluding Husing our recalculation of the Chave et al. (2005) model with new pub-
lished data; (b) biomass estimated based on height (H) integration from a regional Hmodel;
(c) shows the difference (b–a) in biomass due to Hintegration for 327 plots.
Continent Region nplots (a) no H∗(b) with H∗(c) ∆Bdue to H
Africa C. Africa 16 392.9±223.9 379.4±211.5 −13.5±13.0
E. Africa 20 470.3±273.6 362.5±214.2 −107.9±59.4
W. Africa 26 374.4±114.5 330.2±102.8 −44.2±12.5
S. America Brazilian Shield 35 250.3±119.5 194.5±100.3 −55.9±22.4
E. C. Amazonia 44 410.7±161.9 344.1±136.2 −66.6±25.3
Guyana Shield 45 441.1±234.7 434.4±218.6 −6.7±24.0
W. Amazonia 101 299.6±161.3 251.7±125.2 −48.0±38.0
Asia S. E. Asia 14 434.6±201.5 424.2±198.0 −10.5±4.9
Australia N. Australia 26 571.8±359.1 455.3±281.0 −116.5±78.0
Grand mean 405.1±205.6 352.9±179.4 −52.2±30.8
∗Biomass estimated from the moist forest pantropical model based on tree diameter and ρW
or based on tree diameter, ρWand H, where His estimated from Weibull region-specific tree
Hmodels based on the pantropical tree H:Ddatabase from Feldpausch et al. (2011).
53
Table 5. Stocks and change in estimated pantropical C in above ground live trees (bootstrap
mean and 95% CI) due to Hintegrated into biomass estimates based on region-specific esti-
mates of tree H, compared to the pantropical forest biomass model that excludes H∗.
without height with height ∆C due to height –
Continent Region Area Total C Total C Total C Relative
reduction
(106ha) (Pg) (Pg) (Pg)
Africa C. Africa 422.6 83.0±48.3 80.2±45.6 −2.9 −0.03±0.02
E. Africa 123.1 29.0±19.6 22.3±15.5 −6.6 −0.23±0.19
W. Africa 69.8 13.1±4.6 11.5±4.2 −1.5 −0.12±0.05
Total 615.6 125.0±52.3 114.0±48.3 −11.0 −0.13±0.07
South-
Central
America
Brazilian Shield 220.9 27.7±18.1 21.5±16.0 −6.2 −0.22±0.18
E. C. Amazonia 106.2 21.8±8.7 18.3±7.3 −3.5 −0.16±0.08
Guyana Shield 148.3 32.7±17.8 32.2±16.6 −0.5 −0.02±0.01
W. Amazonia 286.4 42.9±24.0 36.0±18.9 −6.9 −0.16±0.11
Total 761.9 125.1±36.0 108.0±30.7 −17.1 −0.14±0.05
Asia S.E. Asia 185.0 40.2±21.2 39.2±20.8 −1.0 −0.02±0.02
Australia N. Australia 105.1 30.1±23.0 23.9±18.1 −6.1 −0.20±0.19
Total 1667.5 320.4±70.8 285.2±63.6 −35.2 −0.13±0.10
∗Tree height estimated from region-specific Weibull-Hmodels adapted from the pantropical
tree H:Ddatabase of Feldpausch et al. (2011). Mean ∆C values (0.5 of biomass values) from
each region in Table 4 were applied. Region geographic extent is shown in Fig. 1. Tropical forest
area was estimated for each region based on the broadleaf deciduous open and closed and
evergreen tree cover classification from GLC2000 (Global Land Cover Map 2000) (Bartholom´
e
and Belward, 2005).
54
Biomass (Mg/ha)
45 to 150
150 to 300
300 to 400
400 to 500
500 to 750
Biomass increase (Mg/ha)
+1.6 to 5
+5 to 10
Biomass decrease (Mg/ha)
-195 to -150
-150 to -100
-100 to -50
-50 to -25
-25 to 0
Amazonia, Brazilian Shield
Amazonia, W.-C. America
Amazonia, E.-Central
Australasia
Asia
Africa, W.
Africa, central
Africa, E.
Amazonia, Guyana Shield
Fig. 1. Location of the pantropical permanent plots, including (a) biomass stocks ( Mg ha−1),
and (b) change in biomass (Mg ha−1) due to inclusion of Hin biomass (B)estimates (relative to
exclusion of H) for forests (BH–BNo Ht.) in Africa, Asia, Australia and South America. Symbols
indicate an increase (blue ) or decrease (red ) in biomass estimates after including Hin biomass
estimates compared to our biomass model Eq. (1) that excludes H. See Supplement Table S1
for plot details. Biomass estimated from the moist forest pantropical models (Table 1) based
on tree diameter and wood density, and when H(where applicable), with Hestimated from
Weibull region-specific tree Hmodels (Eq. 5) based on the pantropical tree H−Ddatabase
from Feldpausch et al. (2011). Coloured shading indicates forest cover and different regions
used in Figs. 3 and 4.
55
Figure 2
Total destructively sampled biomass (Mg)
0 100 200 30 0 400
0 500 1000 1500
Cumulative biomass (Mg)
Total biomass (Mg)
Cumulative biomass (Mg)
411 192 122 87 52 53 39 32 21 25 29 22 27 17
10-15 15-20 20-25 25-30 30-35 35-40 40-45 45-50 50-55 55-60 60-70 70-80 80-100 100+
Diameter (cm)
Relative error in biomass estimate
-0.2 -0.1 0.0 0.1 0.2
without H
Continental p ower-H
Regiona l power-H
Continental 3 PE-H
Regiona l 3PE-H
Continental Wei bull-H
Regiona l Weibull-H
b)
a)
Fig. 2. (a) Distribution of destructively sampled above ground tree dry mass (bars) by diameter
class (cm) and cumulative biomass (line) on the second axis. Numbers above the bars indicate
the number of trees sampled. The dataset represents the pantropical destructive data to date
used to form biomass allometric models, including additional data from Africa, Asia, and South
America; and (a) Relative error associated with estimating the true (destructively) sampled
above ground tree dry mass ((Bestimated –Bmeasured)/Bmeasured )for the same dataset estimated
with and without estimated Hin the biomass model by diameter class (cm). Height estimated
by three model forms and either a continental or regional parameterisation. Positive values
indicate the biomass model overestimates true destructively sampled mass.
56
Figure 3
E-CAm azonia
AGB(M g ha
-1
)
0
20
40
60
80
0 10 0 200 300 400 500
W Amazonia
0
20
40
60
80
0 10 0 200 300 400 500
Brazilian S hield
0
20
40
60
80
0 10 0 200 300 400 500
Cumulative AGB (Mg ha -1)
without height
with Weibull regional height
Guyana Shield
AGB (Mg ha
-1
)
0
20
40
60
80
0 10 0 200 300 400 500
SE Asia
0
20
40
60
80
0 10 0 200 300 400 500
N Australia
0
20
40
60
80
0 10 0 200 300 400 500
Cumulative AGB (Mg ha -1)
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80-90
90-100
100-110
110-120
120-130
130-140
140-150
150-160
160+
C Africa
AGB (Mg ha
-1
)
0
20
40
60
80
0 100 200 300 400 500
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80-90
90-100
100-110
110-120
120-130
130-140
140-150
150-160
160+
E Africa
0
20
40
60
80
0 100 200 300 400 500
10-20
20-30
30-40
40-50
50-60
60-70
70-80
80-90
90-100
100-110
110-120
120-130
130-140
140-150
150-160
160+
W Africa
0
20
40
60
80
0 100 200 300 400 500
Cumulative AGB (Mg ha -1)
Fig. 3. (a) Biomass (Mg ha−1)distribution (bars) among diameter class (cm) by region with
cumulative AGB (Mgha−1) on the second axis (lines) for trees in pantropical permanent plots.
Tree-by-tree biomass was estimated by model (1) without Hor model (2) with Weibull (Eq. 5)
region-specific H. See Table 4 for differences in biomass estimates due to Hintegration.
57
E-CAma zonia
Error in biomass estimate (Mg ha -1)
-15 -10 -5 0 5 10 15
W Amazonia Brazilian S hield
without height
with Weibull regional height
Guyana Shield
Error in biomass estimate (Mg ha -1)
-15 -10 - 5 0 5 10 15
SE Asia N Australia
10-15
15-20
20-25
25-30
30-35
35-40
40-45
45-50
50-55
55-60
60-70
70-80
80-100
100+
C Africa
Error in biomass estimate (Mg ha -1)
-15 -10 -5 0 5 10 15
10-15
15-20
20-25
25-30
30-35
35-40
40-45
45-50
50-55
55-60
60-70
70-80
80-100
100+
E Africa
10-15
15-20
20-25
25-30
30-35
35-40
40-45
45-50
50-55
55-60
60-70
70-80
80-100
100+
W Africa
Fig. 4. Error in biomass estimates (Mg ha−1)for trees in pantropical permanent plots due
to biomass model inputs excluding or including H(relative error propagated from destructive
data). Tree-by-tree biomass was estimated by model (1) without Hor model (2) with Weibull
(Eq. 5) region-specific H.
58
