See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/378741406

Tree belowground biomass in Congo Basin forests: allometric equations and

scaling with aboveground biomass

Article in Forestry · March 2024

DOI: 10.1093/forestry/cpae009

CITATIONS READS

0 35

7 authors, including:

Alain Franck Kossi Ditsouga Quentin Moundounga Mavouroulou

National Center for Scientific and Technological Research Institut de Recherche en Ecologie Tropicale, Libreville, Gabon
4 PUBLICATIONS 211 CITATIONS 7 PUBLICATIONS 296 CITATIONS

SEE PROFILE SEE PROFILE

Adeline Fayolle Nicolas Picard

University of Liège 188 PUBLICATIONS 6,511 CITATIONS
179 PUBLICATIONS 3,695 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Alain Franck Kossi Ditsouga on 08 March 2024.

The user has requested enhancement of the downloaded file.

 Forestry: An International Journal of Forest Research, 2024, 1–10
https://doi.org/10.1093/forestry/cpae009
Original Article

Tree belowground biomass in Congo Basin forests:

allometric equations and scaling with aboveground

Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024

biomass
Alain Franck Kossi Ditsouga1,2 , Quentin Moundounga Mavouroulou1 , Cynel Gwenael Moundounga1 , Adeline Fayolle3,4 ,
Nicolas Picard5, *, Akinobu Sato6 , Alfred Ngomanda1
1 Institutde Recherche en Ecologie Tropicale (IRET), Centre National de la Recherche Scientifique et Technologique (CENAREST), Libreville, Gabon
2 Université des Sciences et Techniques de Masuku, Franceville, Gabon
3 Gembloux Agro-Bio Tech, University of Liège, Gembloux 5030, Belgium
4 Cirad Forêts et Sociétés, Campus International de Baillarguet, Montpellier 34398, France
5 GIP ECOFOR, Paris 75116, France
6 Japan Forest Technology Association (JAFTA), Forest Information Group, Tokyo 102-0085, Japan

*Corresponding author. GIP ECOFOR, 42 rue Scheffer, Paris 75116, France. E-mail: nicolas.picard@gip-ecofor.org

Abstract
Many allometric models to predict tree aboveground biomass have been developed in tropical moist forests, but few models are available
for tree belowground biomass. Theory predicts that belowground biomass scales in an isometric way with aboveground biomass.
Estimates of belowground biomass could then be derived from aboveground biomass using the root:shoot ratio. Using a dataset of
118 tropical trees for which both aboveground and belowground biomass and other tree and species characteristics were measured
in Gabon and Cameroon, we found a near isometric, yet significantly allometric, relationship between belowground biomass (B, in
kilograms) and aboveground biomass (A, in kilograms): B = 0.324 A0.939 . The root:shoot ratio was 0.20–0.22, regardless of tree size. An
efficient model to predict belowground biomass from tree diameter (D, in centimeters), height (H, in meters) and wood density (ρ, in
grams per cubic centimeter) was B = 0.0188 (ρD2 H)0.977 . A significant residual effect of species and leaf habit was found in this model,
indicating that further tree and species characteristics are likely to explain additional variation in belowground biomass. Yet, the future
development of belowground allometric models can benefit from the many models already developed for aboveground biomass. On the
basis of this unprecedented sampling effort on tree belowground biomass in the dense tropical forests of the Congo Basin, we conclude
that the scaling of belowground biomass with aboveground biomass should be the relationship to focus on.

Keywords: biomass; belowground; Congo Basin; forests; allometric; scaling

information is then converted into biomass using allometric

Introduction
Tropical forests are critical for the global climate system as they
contribute 30% of the primary productivity of terrestrial ecosys-
tems and represent between 40 and 50% of terrestrial carbon
stocks (Pan et al., 2011; Lewis et al., 2015). Although aboveground
biomass of living trees in tropical forests stands out intuitively as a
major share of terrestrial carbon stocks, non-negligible f luxes of
carbohydrate photosynthates are also allocated to belowground
roots. Roots control plant and ecosystem functioning through
their involvement in carbon and nutrient cycles and water uptake
(Malhi et al., 2011; Malhi, 2012). However, our understanding of
how much carbon is partitioned between above- and belowground
biomass pools at individual tree and stand levels in tropical
forests is still limited. Indeed, measuring root biomass is time-
consuming and thus costly. It also remains technically challeng-
ing, especially for fine roots.
Various methods can be used to estimate carbon stocks
in tropical forests (Gibbs et al., 2007). Nevertheless, all these
methods are ultimately based on forest inventory data, where
trees are measured and identified in sample plots, and this
models and summed over all trees (Clark and Kellner, 2012;
Chave et al., 2014). Aboveground biomass is finally converted into
carbon using estimates of carbon fraction (ca. 0.48; Martin et al.,
2018). Allometric models to estimate the aboveground biomass of
tropical trees have received great interest across the global tropics
(Chave et al., 2005; Chave et al., 2014) and in central Africa (Fayolle
et al., 2013; Ngomanda et al., 2014; Fayolle et al., 2018). While
pantropical models offer a solution for consistently estimating
tree aboveground biomass on a large scale, locally fitted models
have demonstrated variations in aboveground biomass allometry
at the local level (Wang et al., 2023). However, much less knowledge
is available for tree belowground biomass in tropical forests.
Few allometric equations predicting the belowground biomass
from tree and species characteristics have been developed in the
tropics, using trees sampled in agricultural landscapes of Western
Kenya (Kuyah et al., 2012), miombo woodlands of Mozambique
(Ryan et al., 2011) and Tanzania (Mugasha et al., 2013), and in
primary dipterocarp forests of Malaysia (Niiyama et al., 2010) and
Viet Nam (Nam et al., 2016; Kralicek et al., 2017). As far as we know,
only one allometric equation has been developed for the tropical

Handling editor: Prof. Arshad Ali

Received: June 24, 2023. Revised: February 2, 2024. Accepted: February 6, 2024
© The Author(s) 2024. Published by Oxford University Press on behalf of Institute of Chartered Foresters. All rights reserved. For permissions,
please e-mail: journals.permissions@oup.com.
2 | Kossi Ditsouga et al.
moist forests of central Africa in regenerating semi-deciduous
forest of Cameroon, but with few trees and very few large trees
with a stem diameter >70 cm (Deans et al., 1996). Therefore,
pending questions are whether the same variability in allometry
exists for belowground biomass as for aboveground biomass and
whether it can be well captured by the same predictors: tree
diameter, total height, and species-specific wood density.
Species variability has been a central question in the devel-
opment of biomass allometric equations (Paul et al., 2019). In
tropical forests where tree species diversity is huge and many
species are rare, the statistical challenge of obtaining enough
samples is added to the biological questions of between-species
differences. Since the work of Brown et al. (1989), it has been
acknowledged that developing species-specific allometric equa-
tions for aboveground biomass is not the appropriate approach in
tropical forests. Substituting species-dependent functional traits
for the species identity is both statistically more efficient and
biologically more relevant. Wood density is the species-specific
functional trait most often used as a substitute for species identity
in aboveground biomass equations (Chave et al., 2014). It was
found to well capture between-species differences in Cameroon
(Fayolle et al., 2013) as earlier suggested (Ketterings et al., 2001).
We assume that the same logic applies to belowground biomass.
Nonetheless, developing mixed-species allometric equations on
the basis of functional traits does not preclude for testing for a
residual species effect.
Whereas allometric models that predict tree belowground
biomass from tree and species characteristics are useful to
convert inventory data into belowground biomass estimates,
aboveground biomass plays a distinct role as a predictor of
belowground biomass. The metabolic scaling theory predicts an
isometric scaling of tree belowground biomass with aboveground
biomass (Niklas and Enquist, 2002; Robinson, 2004). This isometric
scaling implies that a root:shoot biomass ratio can be defined
for trees irrespective of their size and species. It also justifies
scaling up root:shoot ratios from the tree to the stand level
(Cairns et al., 1997; Mokany et al., 2006). Intergovernmental
Panel on Climate Change (IPCC) guidelines stipulate that stand-
level belowground biomass can be estimated by multiplying the
aboveground biomass by the default root:shoot ratio. This ratio
is 0.37 for tropical wet forests. For tropical moist forests, it is
0.20 for stand-level aboveground biomass less than 125 Mg ha−1
and 0.24 for biomass greater than 125 Mg ha−1 . It is not always
clear to distinguish tropical wet and moist forests, especially
in Africa where tropical forests are found under drier and
more seasonal climates than in other tropical regions (Guan
et al., 2015). It is thus difficult to know which root:shoot ratio
to use. In miombo woodlands (savanna biome), it has been
shown that the tree root:shoot ratio varied from 0.27 to 0.58,
with a mean of 0.42, slightly higher than the mean (0.35)
reported for 18 other savanna sites with comparable aboveground
biomass (Ryan et al., 2011). In contrast, the IPCC guidelines
suggest a value of 0.56 for aboveground biomass less than
20 Mg ha−1 and 0.28 for aboveground biomass greater than
20 Mg ha−1 .
In addition to the discrepancies between values reported for
the root:shoot ratio, testing the assumption of isometric scal-
ing between above and belowground biomass using observations
worldwide has brought conf licting results. When fitting a power
model to above and belowground biomass data, Cheng et al. (2015)
found an exponent that was not significantly different from one.
In contrast, Kuyah et al. (2012) at the tree level, and both Cairns
et al. (1997) and Mokany et al. (2006) at the stand level, found
an exponent smaller than one. Given the implications it has for
defining root:shoot ratios, the theoretical prediction of isometric
scaling remains a central hypothesis to be tested.
When focusing on the belowground biomass of tropical trees,
the size of the stump needs to be considered. Stump size can
not only be a predictor of belowground biomass but also of
aboveground biomass. One interest is to predict the biomass of
trees that have been cut and removed, so that their stump is
the only available indicator of their size (Corral-Rivas et al., 2007;
Manyanda et al., 2019). Another interest is to use a predictor that
Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024
is closely related to the tree diameter at ground level, which is a
common predictor in drylands or in forests dominated by multi-
stemmed trees. Using stump size rather than diameter at breast
height then facilitates the comparison of models across biomes
(Paul et al., 2019).
The overall objective of this study was to improve the esti-
mation of carbon stocks in tropical forests, with a focus on the
belowground biomass of tropical trees and its relationship with
aboveground biomass. The following research questions were
addressed in central Africa, a tropical region that remains vastly
undersampled and for which there is only little information on
tree roots (Freycon et al., 2015): (i) How much tree biomass is
allocated to the roots? (ii) What are the best allometric mod-
els to predict the belowground biomass from tree and species
characteristics? (iii) How does the belowground biomass scale
with the aboveground biomass? We specifically tested the null
hypothesis that belowground biomass scales in an isometric way
with aboveground biomass or equivalently that the root:shoot
ratio or the allocation of biomass to roots does not vary with tree
size. (4) Can we use stump information to reconstruct the above-
and belowground biomass of logged trees?
Material and method
Destructive biomass data
In the frame of two regional projects for capacity building related
to the REDD+ (Reducing Emissions from Deforestation and forest
Degradation, and the role of conservation, sustainable manage-
ment of forests and enhancement of forest carbon stocks in devel-
oping countries) mechanism in central Africa, field campaigns to
acquire destructive biomass data were conducted in three sites
and forest types. Above and belowground biomass, along with
other measurements prior to tree felling (stem diameter, total
height, crown radius, shape of the stem basis) and after tree felling
(stump basal area, wood density, root depth, and lateral spreads)
were available for 97 trees of 15 species sampled in the evergreen
highland forest of Gabon, one of the six sites of the projet régional
REDD+ (PREREDD+) project (Fayolle et al., 2018; Momo et al.,
2020; Mankou et al., 2021). Additional data from Japan’s Grant Aid
Project, collected in the evergreen inland forest in the south-east
of Gabon (15 trees) and in the semideciduous-evergreen transition
forest of Cameroon (6 trees) were included, leading to a final
dataset with aboveground and coarse root biomass information
for 118 trees of 18 species (Table 1). The target species were
selected because they are locally abundant and characteristic of
each forest type, according to forest management inventory data
of logging concessions in the area. The leaf habit of each species
(deciduous, leaf exchanger, and evergreen) was also available from
an earlier trait compilation (Fayolle et al., 2014; Ouédraogo et al.,
2016).
Aboveground data were collected following a standardized
protocol adapted from Fayolle et al. (2013) and Ngomanda
et al. (2014). Before felling, the stem diameter at breast height
(130 cm above ground level) was measured. For trees with
trunk irregularities, the diameter was measured 30 cm above

Table 1. Above and belowground information available in three sites in central Africa in the frame of the PREREDD+ project and
Japan’s Grant Aid Project. n is the number of trees, Bi is the observed belowground biomass of the ith tree (i = 1, . . . , n), Ai its observed
aboveground biomass, Di its diameter, Hi its height, Ci the radius of its crown, ρ i its wood density, and Si its stump equivalent diameter.
Numbers in brackets are the minimum and maximum.
PREREDD+ project
Site names Site 1 in Gabon
Longitude 11.131◦ E
Latitude 0.505◦ N
Forest type Atlantic highland evergreen
n 97
Bi (kg) [20, 1694]
Ai (kg) [73, 10 348]
Di (cm) [16, 82]
Hi (m) [13, 43]
ρ i (g cm−3 ) [0.312, 0.849]
Si (cm) [16, 94]
Ci (m) [1.56, 8.55]
deformations (Metcalf et al., 2009), which occurred for 13 trees out
of 118. For all trees, the shape of the stem base was assigned to
one of the six following categories: cylindrical, low-buttress, high-
buttress, wheelbase, f luted, or aerial roots. The total tree height
was measured before and after tree felling, with a vertex and a
tape. The cross-sectional area of the stem base, corresponding
to the beginning of the root collar, was measured either with a
tape or indirectly through georeferenced photographs (applied
for all trees collected in the PREREDD+ project; see Fayolle
et al., 2018). The aboveground compartments, divided into stem,
branches, and foliage, were then weighed or cubed in the
field.
Belowground data were collected following a protocol adapted
from Niiyama et al. (2010). The coarse root system of each tree
was extracted stepwise. The primary lateral roots were initially
exposed through successive excavation prior to hoisting the root
collar. To achieve this, soil trenches were meticulously dug 1–3
m away from the root collar of each tree, reaching a maximum
depth of 5 m, using a mechanical tractor digger. The coarse root
system was then pulled slowly out with a chain connecting the
stump to the tractor. The stem and root portions were separated
from the root collar, and roots were divided into large (diame-
ter ≥ 10 cm), medium (5 ≤ diameter < 10 cm), and small (diame-
ter < 5 cm) roots. Large and medium broken roots were completely
excavated, and soil was carefully removed using brushes and
picks, while acknowledging that fine roots (<2 mm) were probably
undersampled.
Subsamples were collected in each above- and belowground
compartment. Two opposite disk sectors were sampled on the
stump and at both ends of the bole of each tree. Three disks
were sampled in large, medium, and small branches. Three ran-
dom samples were collected in twigs, leaves, and f lowers and
fruits. Subsamples were weighed and dried in an oven until a
constant mass (at 80◦ C and 105◦ C for foliage and wood samples,
respectively) to determine their moisture content (ratio between
the mass of water and the green mass) and wood density (ratio
of the oven-dried dry mass per unit green volume; Gao et al.,
2017). The fresh biomass of weighed compartments (medium
and small branches and roots) and the volume of cubed com-
partments (stem, large branches, and roots) were then converted
into dry mass using the moisture content and the wood density,
respectively.
Tree belowground biomass in Congo Basin forests | 3
Japan’s Grant Aid Project
Site 2 in Gabon Site 3 in Cameroon
13.7131◦ E 14.5452◦ E
1.4209◦ N 3.433◦ N
Atlantic inland evergreen Semideciduous-evergreen transition
Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024
15 6
[4, 453] [9, 1856]
[23, 2276] [51, 11 528]
[10, 56] [12, 98]
[13, 35] [11, 43]
[0.378, 0.605] [0.475, 0.883]
[11, 58] [15, 112]
Not measured Not measured
Analysis of the biomass allocation to tree
compartments
Biomass allocation to tree compartments was computed as the
proportion of biomass in roots, root collar, stem, branches, and
foliage. The latter compartment included f lowers and fruits
when present. Biomass allocation to the root and shoot was
calculated by aggregating belowground (viz. roots and root
collar) and aboveground (viz. stem, branches, and foliage)
compartments. Biomass allocations were computed for all trees
and for each of seven diameter classes: [10, 20) cm, [20, 30) cm, . . . ,
[60, 70) cm, and ≥70 cm. Differences in biomass allocation across
diameter classes were tested using the analysis of variance for
compositional data (Van den Boogaart and Tolosana-Delgado,
2013). Proportions were transformed using the isometric log-
ratio transformation. Log-ratios were compared using analysis
of variance (for two compartments) or multivariate analysis of
variance (for more than two compartments). The analysis was
done using the ‘compositions’ package of R (Van den Boogaart
and Tolosana-Delgado, 2013).
Models to predict the belowground biomass
Models that predict tree belowground biomass were fitted using
linear regression and ordinary least squares. Both belowground
biomass and predictors were log-transformed prior to model fit-
ting. In what follows, notations with a subscript i referring to the
ith tree are observations, whereas notations without any subscript
are model predictions. The different models fitted are listed in
Table 2. They can all be written as
log(Bi ) = log(xi )T β + εi (1)
where Bi is the belowground biomass of the ith tree, xi is the vector
of some characteristics of the ith tree, β is the vector of the model
coefficients, and the residual ε i is distributed as a normal dis-
tribution with zero mean and standard deviation σ . When back-
transforming predictions, a bias correction factor CF = exp(σ 2 /2)
was applied (Parresol, 1999). Models were compared on the basis of
their Akaike information criterion (AIC), their Bayesian informa-
tion criterion (BIC), their root mean square error (RMSE), and their
relative bias (RB). RMSE and RB were estimated using leave-one-
out cross-correlation. Each observation i in turn was dropped from
 4|

Table 2. Allometric equations relating tree belowground biomass (B in kilograms) to tree diameter (D in centimeters), tree height (H in meters), tree crown radius (C in meters), species
wood density (ρ in grams per cubic centimeter), tree foliage biomass (F in kilograms), tree aboveground biomass (A in kilograms), or the stump equivalent diameter (S in centimeters).

Model Coefficients Parameters of model performance

a b c d e n R2 σ AIC BIC RMSE (kg) RB (%)

Kossi Ditsouga et al.

(i) log(B) = a + b log(ρD2 H) −4.018∗∗∗ [−4.49, 0.929∗∗∗ [0.88, 118 0.93 0.31 63.5 71.8 120 10.6∗∗
−3.55] 0.98]
(ii) log(B) = a + b log(D) + c −3.344∗∗∗ [−4.18, 2.069∗∗∗ [1.86, 0.514∗∗ [0.15, 0.997∗∗∗ [0.79, 118 0.93 0.31 62.2 76 113 10.4∗∗
log(H) + d log(ρ) −2.51] 2.28] 0.88] 1.2]
(iii) log(B) = a + b log(D) + c log(ρ) −2.4∗∗∗ [−2.91, 2.302∗∗∗ [2.17, 1.093∗∗∗ [0.9, 118 0.93 0.32 68 79.1 109 10.8∗∗
−1.89] 2.43] 1.29]
(iv) log(B) = a + b log(A) −1.178∗∗∗ [−1.52, 0.939∗∗∗ [0.89, 118 0.92 0.32 72.6 80.9 107 11.7∗∗
−0.83] 0.99]
(v) log(B) = a + log(A) −1.59∗∗∗ [−1.65, 118 - 0.33 76.4 81.9 133 12∗∗
−1.53]
(vi) log(B) = a + b log(D) + c −4.132∗∗ [−7.09, 3.297∗∗∗ [1.63, −0.141 [−0.38, 1.073∗∗∗ [0.87, 118 0.93 0.31 68.6 82.5 118 10.8∗∗
log(D2 ) + d log(ρ) −1.19] 4.97] 0.1] 1.28]
(vii) log(B) = a + b log(D) + c −13.298. [−28.27, 11.474. [−1.72, −2.536 [−6.38, 0.23 [−0.14, 0.6] 1.072∗∗∗ [0.87, 118 0.93 0.31 69 85.6 110 10.8∗∗
log(D2 ) + d log(D3 ) + e log(ρ) 1.67] 24.67] 1.31] 1.27]
(viii) log(B) = a + b log(ρ) + c log(S) −2.671∗∗∗ [−3.23, 1.072∗∗∗ [0.86, 2.319∗∗∗ [2.18, 118 0.92 0.34 82.6 93.7 99 12.4∗∗
−2.11] 1.28] 2.46]
(ix) log(B) = a + b log(D2 H) −5.106∗∗∗ [−5.82, 0.977∗∗∗ [0.91, 118 0.88 0.41 128.3 136.6 164 18.9∗∗∗
−4.39] 1.04]
(x) log(B) = a + b log(D) −3.467∗∗∗ [−4.13, 2.408∗∗∗ [2.22, 118 0.85 0.45 149 157.3 183 22.5∗∗∗
−2.8] 2.59]
(xi) log(B) = a + b log(F) 2.834∗∗∗ [2.38, 3.29] 0.755∗∗∗ [0.62, 118 0.5 0.82 293.1 301.4 301 97.3∗∗∗
0.9]
(xii) log(B) = a + b log(ρD2 H) + c −3.760∗∗∗ [−4.30, 0.902∗∗∗ [0.84, 0.021 [−0.15, 97 0.93 0.28 33.9 44.2 123 8.2∗∗
log(C) −3.22] 0.96] 0.19]
(xiii) log(B) = a + b log(D) + c −2.980∗∗∗ [−3.89, 2.03∗∗∗ [1.79, 0.464∗ [0.09, −0.024 [−0.2, 1.001∗∗∗ [0.80, 97 0.93 0.28 32.2 47.6 120 7.9∗∗
log(H) + d log(C) + e log(ρ) −2.07] 2.27] 0.84] 0.15] 1.20]
(xiv) log(B) = a + b log(D) + c −2.078∗∗∗ [−2.65, 2.239∗∗∗ [2.06, −0.051 [−0.23, 1.089∗∗∗ [0.90, 97 0.93 0.28 36.5 49.4 119 8.3∗∗
log(C) + d log(ρ) −1.51] 2.42] 0.13] 1.28]
(xv) log(B) = a + b log(C) 3.337∗∗∗ [2.71, 3.97] 1.384∗∗∗ [0.95, 97 0.30 0.87 251.6 259.3 285 118.4∗∗∗
1.82]
(xvi) log(A) = a + b log(ρ) + c log(S) −1.240∗∗∗ [−1.83, 1.096∗∗∗ [0.87, 2.367∗∗∗ [2.32, 118 0.91 0.35 93.1 104.1 642 14.1∗∗
−0.65] 1.32] 2.52]

Equations are fitted to 118 trees measured in Gabon and Cameroon when C is not a predictor, or 97 trees when C is used. R2 is the coefficient of determination of the model, σ is the residual standard error, AIC and BIC
are, respectively, the Akaike and Bayesian Information Criteria, RMSE is the root mean square error, and RB is the relative bias. Brackets give the 95% confidence interval of model’s coefficients. Asterisks indicate the
significance of the test of the null hypothesis with ∗ ∗ ∗ for P-value <.001; ∗ ∗ for P-value <.01, and ∗ for P-value <.05. Equations are sorted depending on the response variable (the last model is the only one using
aboveground biomass as the response variable), by the sample size n used for model fitting, and finally by increasing BIC value.

Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024


the dataset, and the model was refitted to the dataset without
observation i. Let m−i (x) be the belowground biomass predicted
by this model for a tree with characteristics x, and let Bi be the
observed
 belowground biomass of the ith tree. Then, RMSE =
1 n
 2    contrast, biomass allocation to root and shoot did not significantly
n i=1 m−i (xi ) − Bi and RB = 1n ni=1 m−i (xi ) − Bi /Bi . For the
best model, i.e. the model with the lowest values of AIC and
BIC, the additional effect of a categorical variable—viz. site, stem
basis shape, leaf habit, or species—was tested by adding it as a
predictor in the model. The full model (with the added predictor)
was tested against the initial model (without the predictor) using
an F-test. If a significant effect was found, groups of categories
were determined using a finite mixture of linear regressions. The
finite mixture of model (1) considers that there are K groups,
each associated with parameters β k and σ k . Each observation
is assumed to be generated by first choosing a group k with
probability π k and then generating the observation from model
(1) with parameters β k and σ k . The different model equations (1)
thus resulted in different groupings of the categories.
Without loss of generality, we describe the method when the
categories are the species. Thus, species within a group were
supposed to have the same allometry for belowground biomass,
while species in different groups were supposed to have different
allometries. Assuming that there are K groups of species, the finite
mixture is written:

K 
ns
 timeters and H in meters, where the multiplicative coefficient
log (Bs ) ∼ πk N log (xsi )T β k , σk (2)
k=1 i=1
where s indexes the species, ns is the number of observations
for species s, Bs is the vector of the ns observed belowground
biomasses for trees of species s, ‘∼’ means ‘is distributed as’,
xsi are the tree characteristics of the ith tree of species s, N(m,
σ ) is the density of the normal distribution with mean m and
standard deviation σ , π k is the prior probability of the kth group,
and β k and σ k are the parameters of the kth group. The posterior
probability that species s belongs to group k is given  by wsk =
 
πk i N log (xsi )T β k , σk / T
j πj i N log (xsi ) β j , σj . Species s is
assigned to the group with the greatest posterior probability. The
number K of groups was determined as the one bringing the lowest
P-value in the F-test. The finite mixture regression was fitted using
the ‘f lexmix’ package of R (Leisch, 2004).
Models to predict biomass from stump size
Models that predict aboveground or belowground tree biomass
from the stump characteristics were fitted in the same way as
above (see section “Models to predict the belowground biomass”),
and are written:
log(Yi ) = a + b log(ρi ) + c log(Si ) + εi (3)
where Yi is the aboveground or belowground biomass of the ith
tree; ρ i is its wood density; Si is the equivalent diameter of its
stump (i.e. the diameter of the disk having the same area as the
stump); a, b, and c are the model coefficients; and the residual ε i
is distributed as a normal distribution with zero mean and stand
deviation σ .
Results
Biomass allocation to tree compartments
Stem was the most important biomass compartment (57% on
average), followed by branches (25%), roots (15%), root collar
(2%), and foliage (2%; Fig. 1a). Biomass allocation to roots, root
Tree belowground biomass in Congo Basin forests | 5
collar, stem, branches, and foliage (including f lowers and fruits)
significantly differed among diameter classes (multivariate anal-
ysis of variance for compositional data: η2 = 0.57, P-value <.001;
Fig. 1b) and among leaf habits (η2 = 0.21, P-value = .001; Fig. 1c). In
differ among diameter classes (analysis of variance for composi-
tional data: F = 0.82, P-value = .55) or among leaf habits (F = 0.81,
P-value = .45). The proportion of biomass in the crown, and specif-
ically in the branches, increased with increasing diameter to the
Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024
detriment of the proportion of biomass in the stem. Belowground
biomass (including coarse roots and root collar) represented 17%
of the total biomass of a tree while shoot represented 83% of it.
These proportions resulted in a root:shoot ratio of 0.20 that might
be slightly underestimated due to the difficulties associated with
the sampling of fine roots in the field.
Belowground biomass allometric equation
To predict tree belowground biomass from tree characteristics,
two models performed best based on AIC and BIC [Table 2, models
(i) and (ii), Fig. 2a]:
B = 0.01888 (ρD2 H)0.929 (4)
B = 0.03702 ρ 0.997 D2.069 H0.514 (5)
with B in kilograms, ρ in grams per cubic centimeter, D in cen-
here includes the bias correction factor. It is remarkable that the
exponents of ρ, D, and H in model (5) are very close to 1, 2, and 0.5,
respectively, corresponding to geometrical expectations. When
compared to their values in model (4), the main difference comes
from the allometric scaling of belowground biomass with tree
height. Adding crown radius C as a predictor or substituting C for
H did not significantly improve the allometric models. The smaller
AIC or BIC values of the models that included C were a mere
consequence of the smaller number of observations available
for model fitting. The coefficient associated with C was never
significantly different from zero, except for the model having C as
its sole predictor. However, this latter model had poor predictive
performance (Fig. 2d and h).
The allometric model (4) was not significantly improved when
adding a site effect (F = 0.35, P-value = .56) or a stem basis shape
effect (F = 0.94, P-value = .45). However, species (F = 2.12, P-
value = .01) and leaf habit (F = 5.22, P-value <.01) had a significant
additional effect. The same results were obtained with model (5).
The ranking of the finite mixture models that assigned the 18
different species to three groups differed from the ranking of these
models (Supplement 1, Table S1). The best model when account-
ing for species groups was the model: log(B) = ak + bk log(D) + ck
log(ρ) (Table 3, Fig. 2a). The species groups thus obtained were
consistent across all models that used tree diameter as a pre-
dictor. However, there was no significant relationship between
these groups and those obtained based on the model that used
foliage (χ 2 -test: χ 2 = 3.0, P-value = .57) or aboveground biomass
(χ 2 = 6.7, P-value = .15) as a predictor. The species groups were not
significantly related either to the leaf habit of the species (χ 2 = 3.6,
P-value = .45) nor to phylogenetic groups (whether at the family,
order, or class level).
Models (iv), (v), and (xi) in Table 2 are of theoretical interest to
see how belowground biomass (B) scales with total aboveground
biomass (A) and the biomass of tree foliage (F). In model (iv),
the 95% confidence interval of the exponent of the power model
B = 0.324 A0.939 excludes 1, which shows that belowground and
aboveground biomasses slightly but significantly deviate from
isometric scaling (Fig. 2b). Consistently, the isometric model (v)
6 | Kossi Ditsouga et al.

Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024

Figure 1. Biomass allocation to tree compartments (a) for all trees, (b) depending on diameter class and (c) depending on the leaf habit for 118 tropical
trees measured in tropical forests of Gabon and Cameroon. Compartments are roots, root collar, stem, branches, and foliage. The number of trees
sampled in each class is given on top of each bar.

Table 3. Finite mixture of linear regressions log(B) = a + b log(D) + c log(ρ) that assigns 118 trees from 18 different species to three
groups of species, where B is tree belowground biomass (in kilograms), ρ is wood density (in grams per cubic centimeter), and D is tree
diameter (in centimeters).

Group a b c n σ Species

1 −2.196∗∗∗ [−3.11, −1.28]
2 −3.938∗∗∗ [−4.89, −2.98]
3 −2.216∗∗∗ [−2.71, −1.72]
2.349∗∗∗ [2.1, 2.59] 1.668∗∗∗ [1.22, 2.12] 37 0.33 Dialium pachyphyllum (6), Distemonanthus
benthamianus (5), Drypetes gossweileri (1), Lophira
alata (7), Pterocarpus soyauxii (3), Scyphocephalium
mannii (14), Streblus usambarensi (1)
2.581∗∗∗ [2.33, 2.83] 0.307 [−0.08, 0.69] 37 0.29 Aucoumea klaineana (24), Leptonychia raynalionum (1),
Octoknema bakossiensis (1), Plagiostyles Africana (11)
2.26∗∗∗ [2.13, 2.39] 1.346∗∗∗ [1.15, 1.54] 44 0.17 Anonidium mannii (5), Desbordesia glaucescens (1),
Petersianthus macrocarpus (5), Pycnanthus angolensis
(5), Santiria trimera (6), Staudtia kamerunensis (15),
Xylopia aethiopica (7)

n is the number of observations, and σ is the residual standard error. The number in parentheses after each species indicates the number of trees measured
for that species.

obtained by setting the exponent to 1, B = 0.215 A, has lower Models to predict above and belowground biomass from
predictive performance than the allometric power model in terms stump equivalent diameter
of AIC, but similar BIC values. The coefficient of 0.215 can be The models that predict aboveground biomass A or belowground
interpreted as the geometric mean of individual root:shoot ratios. biomass B from the characteristics of the stump are [Table 2,
There was much more variability in the scaling of belowground models (xvi) and (viii)]
biomass to foliage biomass [R2 = 0.50 and σ = 0.82 for the
fitting of model (xi)] than to aboveground biomass [R2 = 0.92 A = 0.3078 ρ 1.096 S2.367 (6)
and σ = 0.32 for the fitting of model (iv)]. The exponent of B = 0.0732 ρ 1.072 S2.319 (7)
the allometric scaling of belowground biomass with respect to
foliage biomass was very close to and not significantly different with A and B in kilograms, ρ in grams per cubic centimeter, and S
from 3/4. in centimeters, where the multiplicative coefficient here includes
 Tree belowground biomass in Congo Basin forests | 7

Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024

Figure 2. Observed tree belowground biomass versus observed predictors (top panel) or versus predicted belowground biomass when using this
predictor (bottom panel) for 118 trees measured in tropical forests of Gabon and Cameroon. Predictors are (a, e) the tree dendrometric quantity ρ i Di 2 Hi
(where ρ i is the wood density of the ith tree, Di its diameter, and Hi its height), (b, f) tree aboveground biomass, (c, g) tree diameter, and (d, h) tree crown
radius. In the top panel, the solid line shows the fitted power model between belowground biomass and the predictor, with its confidence interval as
the gray area. In the bottom panel, the dashed line shows the 1:1 line. In (a), the symbol of the dots corresponds to three species groups obtained by a
finite mixture regression (see Table 3). In (b), the dashed line shows the fitted isometric model, with its confidence interval as the shaded area.

the bias correction factor. The exponents of ρ and S are close
to 1 and 2.3, respectively. These exponents are very similar in
the two models, which is consistent with the almost-isometric
relationship between aboveground biomass A and belowground
biomass B.
Discussion
Belowground biomass scales in a near-isometric
way with aboveground biomass
At the tree level, the scaling of belowground biomass with above-
ground biomass deviated slightly, yet significantly, from isometry,
as shown by the exponent 0.939 of the power model that predicted
the belowground biomass of a tree from its aboveground biomass.
The same result was reported in other studies, e.g. by Kuyah et al.
(2012) who also found an exponent <1 for the same power model.
It may seem paradoxical that the relationship between below-
ground biomass and aboveground biomass was concave, while
the biomass allocation to root and shoot did not significantly
differ among diameter classes. Indeed, an exponent α < 1 in the
power model B ∝ Aα implies that both the tree-level root:shoot
ratio (∝ Aα—1 ) and the proportion of biomass allocated to root
decrease with increasing tree size. Nevertheless, this paradox can
be understood given the near-isometric relationship (α very close
to 1) and the fact that tree biomass has a much wider range (from
23 to 11 528 kg) than the tree diameter (from 10 to 98 cm; Table 1).
The regression takes account of the full continuous spread of
biomass to estimate the slope, while the division into diameter
classes breaks this continuity.
The allometric scaling of belowground biomass with above-
ground biomass has both theoretical and practical implications.
From a theoretical point of view, it challenges the prediction of
the metabolic scaling theory that belowground biomass scales in
an isometric way with aboveground biomass (B ∝ A; Niklas and
Enquist, 2002). To be more precise, the metabolic scaling theory
predicts a near-isometric relationship with a deviation from isom-
etry toward convexity (A ∝ B + cB3/4 ), whereas we observed a devia-
tion toward concavity (Fig. 2b, black line). Moreover, the metabolic
scaling theory predicts an exponent of 4/3 for the power rela-
tionship between belowground biomass and leaf biomass while
we obtained an exponent of 3/4, though with a large uncer-
tainty. Robinson (2004) argued that the deviation of the scaling of
belowground biomass from isometry could be due to the under-
estimation of the root mass for large trees for which it is difficult
to excavate or properly sample fine roots. Accordingly, Kuyah et al.
(2012), by extrapolating the biomass of unsampled roots from the
sampled ones, estimated that 48.4% of the root biomass was miss-
ing. After correcting for unexcavated root biomass, they found a
scaling of belowground biomass with aboveground biomass closer
to isometry than before the correction.
From a practical point of view, the allometric scaling of below-
ground biomass with aboveground biomass calls into question
the use of the root:shoot ratio in carbon estimation studies. An
isometric scaling means that the belowground biomass relates
to the aboveground biomass in the same way for a tree or for a
stand consisting of trees with different sizes. It also means that
root:shoot ratios can be equally estimated from or used for data
collected at the tree or at the stand level, irrespective of the size
8 | Kossi Ditsouga et al.
Table 4. Comparison of the predictive performance of published models to predict the belowground biomass of trees with the
predictive performance of the same models fitted to 118 trees measured in Gabon and Cameroon.
Reference Model Original
performance
RB (%)
Kuyah et al. (2012) B = 0.095 ρ 0.575 D2.357 H–0.165 9.5
B = 0.048 D2.303 10.2
B = 0.087 ρ 0.611 D2.257 10.1
Kralicek et al. (2017) B = 0.012564 (D2 H)0.913188 −15.24
B = 0.026438 D2.35221 −29.1
RB is relative bias, and RMSE is the root mean square error. ‘Original performance’ is the performance of the published model for the data of the publication.
‘Performance with the data of this study’ is the performance of the published model for the 118 trees in Gabon and Cameroon. ‘Performance of the
corresponding model of this study’ is the performance of the same model fitted to the 118 trees for these same 118 trees.
structure of the stand. In contrast, an allometric scaling implies
that (i) stand-level root:shoot ratios depend on the size structure
of the stand and (ii) models to predict belowground biomass
from aboveground biomass will differ whether they are fitted to
tree data or stand data. For a large range of size structures, a
concave tree allometry (i.e. B ∝ Aα with α < 1) brings a decreasing
relationship between the stand-level root:shoot ratio and the
stand aboveground biomass. It also brings a concave allometry
between the stand belowground biomass and the stand above-
ground biomass (Supplement 2, Fig. S1a). Such relationships at the
stand level have been earlier reported (Cairns et al., 1997; Mokany
et al., 2006), thus indirectly confirming the nonisometric scaling of
tree belowground biomass with tree aboveground biomass.
Biomass allocation in a tree
On average, the tropical trees in our study allocated 17% (arith-
metic mean) or 18% (geometric mean) of their biomass to roots.
The allocation of biomass to roots is known to depend on water
availability. Across large ecological gradients that cover several
biomes, it thus depends on rainfall, with smaller root:shoot ratios
in wetter biomes (Cairns et al., 1997; Mokany et al., 2006; Qi et al.,
2019). The average values of the root:shoot ratio obtained here
(0.20 using the arithmetic mean or 0.22 using the geometric mean)
were consistent with this pattern. It was lower than IPCC values
for wet forests (0.37) but close to IPCC values for moist forests
(0.20–0.24). It was also close to the values reported in various
studies on tropical rain forests: 0.205–0.235 (Mokany et al., 2006),
0.22 (Fonseca et al., 2012), 0.24 (Cairns et al., 1997), and 0.26
(Kuyah et al., 2012). While allocation to roots was not found to
vary significantly across diameter classes, allocation to the crown
(branch + foliage) was found to increase with stem diameter, as
already reported (Ploton et al., 2016; Mankou et al., 2021).
Allometric models to predict tree belowground
biomass
The allometric models fitted in this study to predict tree below-
ground biomass had a predictive performance similar to those
fitted in other studies. For instance, Kuyah et al. (2012) obtained
a relative bias of 9.5% for the model B = 0.095 ρ 0.575 D2.357 H–0.165
fitted to their data, compared to 10.4% for model (4) in this study.
When applying Kuyah et al.’s model to the data of this study, an
RMSE of 140 kg was obtained, which is greater than the RMSE
of 113 kg for model (4). Similar results were obtained with other
published models that predict the belowground biomass of trees
in tropical rainforests (Table 4). A direct comparison between tree-
based allometric models, such as the ones developed in this study,
and stand-based allometric models, exemplified by Cairns et al.
(1997), is not feasible due to differences in both explanatory and
Performance for the data of Performance of the corresponding
this study model of this study
RMSE (kg) RB (%) RMSE (kg) RB (%)
139 8.49 113 10.4
177 16.49 183 22.5
128 17.31 109 10.8
Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024
171 15.74 164 19.0
231 −23.52 183 22.5
response variables. Nevertheless, across a broad spectrum of tree
size structures and for stand-level aboveground biomass ranging
from 0 to 300 Mg ha−1 , model (iv) and Cairns et al.’s model exhibit
consistency (Supplement 2). For higher aboveground biomass
values, Cairns et al.’s model yields lower predictions than model
(iv), with differences reaching up to 20% (Supplement 2, Fig. S1b).
It has already been pointed out that the prediction of aboveground
biomass depends critically on the choice of the allometric model
used, with local equations possibly bringing high bias (Picard
et al., 2015). Here, it is also true for the prediction of belowground
biomass.
However, contrary to allometric models for aboveground
biomass where tree characteristics consistently have the same
effect on biomass in different studies, the effect of tree height
on belowground biomass varies between studies. In this study, a
positive exponent close to 0.5 was found for tree height in model
(4). In other words, for a given diameter, the more slender a tree
was, the greater its belowground biomass. In contrast, Kuyah et al.
(2012) found a negative exponent for tree height, meaning that
the more slender a tree was, the smaller its belowground biomass.
The relatively good predictive performance of models that do not
use tree height as a predictor (Table 2) also calls into question the
effect of tree height on belowground biomass. Further studies will
be needed to clarify this effect.
A significant residual effect of species or leaf habit was found
in models (4) and (5), even though wood density already accounted
for some species differences. Nonetheless, no relationship was
found between species groups and leaf habit or phylogeny. Given
the low number of observations for each species, we cannot
exclude that the residual species effect was due to some unidenti-
fied confounding factor. Therefore, additional tree characteristics
or species traits should be explored as additional predictors in
the models. Kralicek et al. (2017) used crown area as an addi-
tional predictor of belowground biomass but without meaning-
ful improvement compared to models based on diameter, wood
density, and height. We came to the same conclusion with crown
radius.
Finally, predicting biomass from the equivalent diameter of the
stump can be useful in situations where trees have been logged
and only the stump remains and can be measured in the field.
As far as we know, this study is the first to provide a model to
predict tree aboveground biomass from the equivalent diameter
of the stump for tropical forests. As regards belowground biomass,
the equivalent diameter of the stump is actually quite an efficient
predictor. Kuyah et al. (2012) also fitted a model to predict below-
ground biomass from the diameter of the stump: B = 0.024 S2.283 .
But they got a relative bias of 35% with this model, indicating a
lower predictive performance than the model fitted in this study.

Conclusion
Based on an unprecedented data set in the dense tropical forests
of the Congo Basin, the belowground biomass of trees scaled in a
near-isometric, yet significantly allometric, way with aboveground
biomass. Further work is needed to determine if this deviation
from the isometric scaling was due to fine roots that were most
likely undersampled in this study. The biomass allocation to
roots, with a root:shoot ratio of 0.20–0.22, conformed to what was
expected for tropical moist forests. Allometric models that predict
tree belowground biomass from tree and species characteristics
seemed to be as variable as allometric models for aboveground
biomass, with an additional uncertainty on the effect of tree
height on belowground biomass. To circumvent the development
of many allometric models for belowground biomass and take
advantage of the many models already developed for above-
ground biomass, we recommend focusing on the relationship
between belowground and aboveground biomass.
Acknowledgements
The authors thank the students, technicians, and colleagues that
were involved in the data collection, data quality controls, and
data management.
Author contributions
Alain Franck Kossi Ditsouga (Conceptualization, Formal analysis,
Investigation, Methodology), Quentin Moundounga Mavouroulou
(Conceptualization, Investigation, Methodology), Cynel Gwenael
Moundounga (Conceptualization, Investigation, Methodology),
Adeline Fayolle (Data curation, Formal analysis, Project adminis-
tration, Writing—original draft), Nicolas Picard (Formal analysis,
Writing—original draft), Akinobu Sato (Conceptualization,
Investigation, Methodology), Alfred Ngomanda (Conceptualiza-
tion, Formal analysis, Investigation, Methodology, Supervision,
Writing—original draft)
Supplementary data
Supplementary data are available at Forestry online.
Conf lict of interest: None declared.
Funding
This work was supported by the PREREDD+ regional project,
funded by a gift from the Global Environment Facility [TF010038],
administered by the World Bank to the COMIFAC. Japan’s
Programme Grant Aid for The Forest Preservation Programme
[FY 2009 to A.S.] for the Governments of the Gabonese and
Cameroonian Republics also contributed to the support. The
component 2b of the PREREDD+, which aimed at ‘Building
allometric equations for the forests of the Congo basin,’ was led by
the ONFi/TEREA/Nature+ consortium. Data collection in Gabon
was carried out by IRET with logistic support necessary for the
field and laboratory measurements provided by Rougier Gabon.
Japan’s Grant Aid Project was led by JAFTA, with logistic and
technical support in field and laboratory measurements provided
by IRET, the Minister of Forests, and Rougier in Gabon, and the
Minister of Forests in Cameroon.
Data availability
The data underlying this article will be shared on reasonable
request to the last author.
Tree belowground biomass in Congo Basin forests | 9
References
Brown, S., Gillespie, A.J.R. and Lugo, A.E. 1989 Biomass estima-
tion methods for tropical forests with applications to for-
est inventory data. For Sci 35, 881–902. https://doi.org/10.1093/
forestscience/35.4.881.
Cairns, M.A., Brown, S., Helmer, E.H., et al. 1997 Root biomass alloca-
tion in the world’s upland forests. Oecologia 111, 1–11. https://doi.
org/10.1007/s004420050201.
Chave, J., Andalo, C., Brown, S., et al. 2005 Tree allometry and improved
Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024
estimation of carbon stocks and balance in tropical forests.
Oecologia 145, 87–99. https://doi.org/10.1007/s00442-005-0100-x.
Chave, J., Réjou-Méchain, M., Búrquez, A., et al. 2014 Improved allo-
metric models to estimate the aboveground biomass of tropi-
cal trees. Glob Chang Biol 20, 3177–3190. https://doi.org/10.1111/
gcb.12629.
Cheng, D., Zhong, Q., Niklas, K.J., et al. 2015 Isometric scaling of above-
and below-ground biomass at the individual and community
levels in the understorey of a sub-tropical forest. Ann Bot 115,
303–313. https://doi.org/10.1093/aob/mcu238.
Clark, D.B. and Kellner, J.R. 2012 Tropical forest biomass estimation
and the fallacy of misplaced concreteness. J Veg Sci 23, 1191–1196.
https://doi.org/10.1111/j.1654-1103.2012.01471.x.
Corral-Rivas, J.J., Barrio-Anta, M., Aguirre-Calderón, O.A., et al. 2007
Use of stump diameter to estimate diameter at breast height and
tree volume for major pine species in El Salto, Durango (Mexico).
Forestry 80, 29–40. https://doi.org/10.1093/forestry/cpl048.
Deans, J.D., Moran, J. and Grace, J. 1996 Biomass relationships for tree
species in regenerating semi-deciduous tropical moist forest in
Cameroon. For Ecol Manage 88, 215–225. https://doi.org/10.1016/
S0378-1127(96)03843-1.
Fayolle, A., Doucet, J.L., Gillet, J.F., et al. 2013 Tree allometry in Central
Africa: Testing the validity of pantropical multi-species allomet-
ric equations for estimating biomass and carbon stocks. For Ecol
Manage 305, 29–37. https://doi.org/10.1016/j.foreco.2013.05.036.
Fayolle, A., Ngomanda, A., Mbasi, M., et al. 2018 A regional allometry
for the Congo basin forests based on the largest ever destructive
sampling. For Ecol Manage 430, 228–240. https://doi.org/10.1016/j.
foreco.2018.07.030.
Fayolle, A., Picard, N., Doucet, J.L., et al. 2014 A new insight in the
structure, composition and functioning of central African moist
forests. For Ecol Manage 329, 195–205. https://doi.org/10.1016/j.
foreco.2014.06.014.
Fonseca, W., Alice, F.E. and Rey-Benayas, J.M. 2012 Carbon accumula-
tion in aboveground and belowground biomass and soil of differ-
ent age native forest plantations in the humid tropical lowlands
of Costa Rica. New Forests 43, 197–211. https://doi.org/10.1007/
s11056-011-9273-9.
Freycon, V., Wonkam, C., Fayolle, A., et al. 2015 Tree roots can pen-
etrate deeply in African semi-deciduous rain forests: evidence
from two common soil types. J Trop Ecol 31, 13–23. https://doi.
org/10.1017/S0266467414000595.
Gao, S., Wang, X., Wiemann, M.C., et al. 2017 A critical analy-
sis of methods for rapid and nondestructive determination of
wood density in standing trees. Ann For Sci 74, 1–13. https://doi.
org/10.1007/s13595-017-0623-4.
Gibbs, H.K., Brown, S., Niles, J.O., et al. 2007 Monitoring and estimating
tropical forest carbon stocks: making REDD a reality. Environ Res
Lett 2, 1–13. https://doi.org/10.1088/1748-9326/2/4/045023.
Guan, K., Pan, M., Li, H., et al. 2015 Photosynthetic seasonality of
global tropical forests constrained by hydroclimate. Nat Geosci 8,
284–289. https://doi.org/10.1038/ngeo2382.
Ketterings, Q.M., Coe, R., van Noordwijk, M., et al. 2001 Reduc-
ing uncertainty in the use of allometric biomass equations for
10 | Kossi Ditsouga et al.
predicting above-ground tree biomass in mixed secondary
forests. For Ecol Manage 146, 199–209. https://doi.org/10.1016/
S0378-1127(00)00460-6.
Kralicek, K., Huy, B., Poudel, K.P., et al. 2017 Simultaneous estima-
tion of above-and below-ground biomass in tropical forests of
Viet Nam. For Ecol Manage 390, 147–156. https://doi.org/10.1016/
j.foreco.2017.01.030.
Kuyah, S., Dietz, J., Muthuri, C., et al. 2012 Allometric equations
for estimating biomass in agricultural landscapes: II. Below-
ground biomass. Agric Ecosyst Environ 158, 225–234. https://doi.
org/10.1016/j.agee.2012.05.010.
Leisch, F. 2004 FlexMix: A general framework for finite mixture
models and latent class regression in R. J Stat Softw 11, 18. https://
doi.org/10.18637/jss.v011.i08.
Lewis, S.L., Edwards, D.P. and Galbraith, D. 2015 Increasing human
dominance of tropical forests. Science 349, 827–832. https://doi.
org/10.1126/science.aaa9932.
Malhi, Y. 2012 The productivity, metabolism and carbon cycle of trop-
ical forest vegetation. J Ecol 100, 65–75. https://doi.org/10.1111/
j.1365-2745.2011.01916.x.
Malhi, Y., Doughty, C. and Galbraith, D. 2011 The allocation of ecosys-
tem net primary productivity in tropical forests. Philos Trans
R Soc Lond B 366, 3225–3245. https://doi.org/10.1098/rstb.2011.
0062.
Mankou, G.S., Ligot, G., Panzou, G.J.L., et al. 2021 Tropical tree allom-
etry and crown allocation, and their relationship with species
traits in central Africa. For Ecol Manage 493, 119262. https://doi.
org/10.1016/j.foreco.2021.119262.
Manyanda, B.J., Mugasha, W.A., Nzunda, E.F., et al. 2019 Biomass and
volume models based on stump diameter for assessing degrada-
tion of miombo woodlands in Tanzania. Int J For Res 2019, 1876329.
https://doi.org/10.1155/2019/1876329.
Martin, A.R., Doraisami, M. and Thomas, S.C. 2018 Global patterns in
wood carbon concentration across the world’s trees and forests.
Nat Geosci 11, 915–920. https://doi.org/10.1038/s41561-018-
0246-x.
Metcalf, C.J.E., Clark, J.S. and Clark, D.A. 2009 Tree growth inference
and prediction when the point of measurement changes: mod-
elling around buttresses in tropical forests. J Trop Ecol 25, 1–12.
https://doi.org/10.1017/S0266467408005646.
Mokany, K., Raison, R.J. and Prokushkin, A.S. 2006 Critical analysis of
root: shoot ratios in terrestrial biomes. Glob Chang Biol 12, 84–96.
https://doi.org/10.1111/j.1365-2486.2005.001043.x.
Momo, S.T., Ploton, P., Martin-Ducup, O., et al. 2020 Leveraging sig-
natures of plant functional strategies in wood density profiles
of African trees to correct mass estimations from terrestrial
laser data. Sci Rep 10, 2001. https://doi.org/10.1038/s41598-020-
58733-w.
Mugasha, W.A., Eid, T., Bollandsås, O.M., et al. 2013 Allometric models
for prediction of above- and belowground biomass of trees in
the miombo woodlands of Tanzania. For Ecol Manage 310, 87–101.
https://doi.org/10.1016/j.foreco.2013.08.003.
View publication stats
Nam, V.T., Van Kuijk, M. and Anten, N.P.R. 2016 Allometric equations
for aboveground and belowground biomass estimations in an
evergreen forest in Vietnam. PloS One 11, e0156827. https://doi.
org/10.1371/journal.pone.0156827.
Ngomanda, A., Engone Obiang, N.L., Lebamba, J., et al. 2014 Site-specic
versus pantropical allometric equation: Which option to estimate
the biomass of a moist central african forest? For Ecol Manage 312,
1–9. https://doi.org/10.1016/j.foreco.2013.10.029.
Niiyama, K., Kajimoto, T., Matsuura, Y., et al. 2010 Estimation of
Downloaded from https://academic.oup.com/forestry/article/doi/10.1093/forestry/cpae009/7618952 by CIRAD-DIC user on 05 March 2024
root biomass based on excavation of individual root systems
in a primary dipterocarp forest in Pasoh Forest Reserve, Penin-
sular Malaysia. J Trop Ecol 26, 271–284. https://doi.org/10.1017/
S0266467410000040.
Niklas, K.J. and Enquist, B.J. 2002 Canonical rules for plant organ
biomass partitioning and annual allocation. Am J Bot 89, 812–819.
https://doi.org/10.3732/ajb.89.5.812.
Ouédraogo, D.Y., Fayolle, A., Gourlet-Fleury, S., et al. 2016 The determi-
nants of tropical forest deciduousness: disentangling the effects
of rainfall and geology in central Africa. J Ecol 104, 924–935.
https://doi.org/10.1111/1365-2745.12589.
Parresol, B.R. 1999 Assessing tree and stand biomass: a review with
examples and critical comparisons. For Sci 45, 573–593. https://
doi.org/10.1093/forestscience/45.4.573.
Pan, Y., Birdsey, R.A., Fang, J., et al. 2011 A large and persistent
carbon sink in the world’s forests. Science 333, 988–993. https://
doi.org/10.1126/science.1201609.
Paul, K.I., Larmour, J., Specht, A., et al. 2019 Testing the general-
ityof below-ground biomass allometry across plant functional
types. For Ecol Manage 432, 102–114. https://doi.org/10.1016/j.
foreco.2018.08.043.
Picard, N., Rutishauser, E., Ploton, P., et al. 2015 Should tree biomass
allometry be restricted to power models? For Ecol Manage 353,
156–163. https://doi.org/10.1016/j.foreco.2015.05.035.
Ploton, P., Barbier, N., Momo, S.T., et al. 2016 Closing a gap in tropical
forest biomass estimation: accounting for crown mass variation
in pantropical allometries. Biogeosciences 12, 19711–19750. https://
doi.org/10.5194/bg-13-1571-2016.
Qi, Y., Wei, W., Chen, C., et al. 2019 Plant root-shoot biomass allocation
over diverse biomes: A global synthesis. Global Ecol Conserv 18,
e00606. https://doi.org/10.1016/j.gecco.2019.e00606.
Robinson, D. 2004 Scaling the depths: below-ground allocation in
plants, forests and biomes. Funct Ecol 18, 290–295. https://doi.
org/10.1111/j.0269-8463.2004.00849.x.
Ryan, C.M., Williams, M. and Grace, J. 2011 Above-and below-
ground carbon stocks in a miombo woodland landscape of
Mozambique. Biotropica 43, 423–432. https://doi.org/10.1111/
j.1744-7429.2010.00713.x.
Van den Boogaart, K.G. and Tolosana-Delgado, R. 2013 Analyzing
compositional data with R. Use R! Springer, Heidelberg.
Wang, Z., Huang, X., Li, F., et al. 2023 Global patterns of allomet-
ric model parameters prediction. Sci Rep 13, 1550. https://doi.
org/10.1038/s41598-023-28843-2.