Allometric equations for biomass estimation in central
African rain forests: state of the art and challenges
IUFRO 2014 World Congress
Salt Lake City, USA, October 511, 2014
Forests and climate change
Quantifying uncertainty in forest measurements and models
Nicolas Picard∗ (nicolas.picard@cirad.fr), Matieu Henry, Noël Fonton∗,
Josiane Kondaoule∗, Adeline Fayolle∗, Luca Birigazzi, Gaël Sola, Anatoli
Poultouchidou, Carlo Trotta, Hervé Maïdou∗
∗
Regional REDD+ project of the Forests Commission of Central Africa
IUFRO 2014 Allometric equations Friday Oct 10th 1 / 14
REDD+: Decision 4/CP.15 (CoP 15, Copenhagen, 2009)
2006 IPCC Guidelines
IUFRO 2014 Allometric equations Friday Oct 10th 2 / 14
Allometric equations in central Africa
5 studies in
central Africa
published since
2010
819 trees
measured
}unpublished data
}published data
mjhe n trees
IUFRO 2014 Allometric equations Friday Oct 10th 3 / 14
Chain of error propagation
IUFRO 2014 Allometric equations Friday Oct 10th 4 / 14
Inventory data
M'Baïki permanent sample plots in the Central African Republic
40 ha of permanent plots monitored since 1982
control plots, logged plots, logged + thinned plots, perturbation by
re
data of 1987 (after all treatments): dbh, species (→ wood density)
twelve 1-ha plots (pseudo-replicates) of undisturbed forest
twelve 1-ha plots (pseudo-replicates) of disturbed forest
Emission factor = (biomass of undisturbed plots)
− (biomass of disturbed plots)
IUFRO 2014 Allometric equations Friday Oct 10th 5 / 14
Biomass equations
4 biomass equations with the datasets used to t the models
Author Type Model n
Chave et al. (2014) pantropical B = f(D, H, ρ)
H = f(D, E)
4004
Ngomanda et al. (2014) local
(northeastern
Gabon)
B = f(D, ρ) 101
Djomo et al. (2010) local (southern
Cameroon)
B = f(D, ρ) 71
Henry et al. (2010) local (Ghana) B = f(D) 42
IUFRO 2014 Allometric equations Friday Oct 10th 6 / 14
Error propagation
Monte Carlo method for:
measurement error
error due to the uncertainty on the model coecients
residual error of the model
Error due to the model choice:
1 Models are considered equally likely
2 Or Bayesian model averaging (BMA) is used to assign dierent
weights to the 4 models
No tree biomass data available at M'Baïki
¯ Training data set for BMA: African data from Chave et al. dataset
(n = 1429)
IUFRO 2014 Allometric equations Friday Oct 10th 7 / 14
Biomass in the 24 plots at M'Baïki according to 4
biomass equations
5 10 15 20
100200300400500
Plot rank (basal area)
Biomass(tonneha−1
)
Chave et al. (2014)
Henry et al. (2010)
Djomo et al. (2010)
Ngomanda et al. (2014)
large error due to
the model choice
if the plot ×
model interaction
is null, this error
has no impact on
the estimation of
the emission factor
IUFRO 2014 Allometric equations Friday Oct 10th 8 / 14
The dierence of biomass between disturbed and
undisturbed plots depends on the biomass equation
Source Df Sum Sq Mean Sq F value p-value
model 3 391 484 130 495 175.426  0.001
plot type 1 512 373 512 373 688.790  0.001
model × plot type 3 42 415 14 138 19.006  0.001
residuals 88 65461 744
IUFRO 2014 Allometric equations Friday Oct 10th 9 / 14
Estimates of the emission factor for the dierent biomass
equations
50100150200250300350
Biomass model
Emissionfactor(tonneha−1
)
Error
sampling
measurement
coefficients
residual
Chave Henry Djomo Ngomanda
IUFRO 2014 Allometric equations Friday Oct 10th 10 / 14
Combining the dierent biomass equations to get a single
estimate of the emission factor50100150200250300
Biomass weights
Emissionfactor(tonneha−1
)
Error
model
sampling
measurement
coefficients
residual
Equal BMA
Weights
Model Equal BMA
Chave et al. 0.25 0.152
Henry et al. 0.25 0.001
Djomo et al. 0.25 0.639
Ngomanda et al. 0.25 0.207
IUFRO 2014 Allometric equations Friday Oct 10th 11 / 14
Conclusions
At M'Baïki, emission factor from intact to degraded forest (logging
+ thinning) is approximately 150 tonne ha−1, but with a very large
uncertainty
The choice of the allometric equation is the largest source of error
(40% of the square error) when estimating the emission factor
1-ha plot sampling (30%) and the uncertainty on the model
coecients (20%) are also important sources of errors
Improving the choice of the allometric equation will require
additional tree biomass measurements
Data base on allometric equations: Globallometree
(http://www.globallometree.org/)
IUFRO 2014 Allometric equations Friday Oct 10th 12 / 14
Thanks for your
attention
This study was supported
by the regional REDD+
project of the COMIFAC 
GEF trust fund grant n◦
TF010038  World Bank
project n◦
P113167
We thank
for access to the M'Baïki
data base
IUFRO 2014 Allometric equations Friday Oct 10th 13 / 14
Studies on allometric equations in central Africa
published since 2010 Map showing the locations of the studies
Djomo et al. (2010) Allometric equations for biomass estimations in
Cameroon and pan moist tropical equations including biomass data from
Africa. For Ecol Manage 260:1873-1885
Ebuy Alipade et al. (2011) Biomass equation for predicting tree
aboveground biomass at Yangambi, DRC. Journal of Tropical Forest
Science 23:125-132
Dorisca et al. (2011) Établissement d'équations entre le diamètre et le
volume total de bois des arbres, adaptées au Cameroun. Bois For Trop
65:87-95
Fayolle et al. (2013) Tree allometry in Central Africa: Testing the
validity of pantropical multi-species allometric equations for estimating
biomass and carbon stocks. For Ecol Manage 305:29-37
Ngomanda 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
IUFRO 2014 Allometric equations Friday Oct 10th 14 / 14

Nicolas picard

  • 1.
    Allometric equations forbiomass estimation in central African rain forests: state of the art and challenges IUFRO 2014 World Congress Salt Lake City, USA, October 511, 2014 Forests and climate change Quantifying uncertainty in forest measurements and models Nicolas Picard∗ (nicolas.picard@cirad.fr), Matieu Henry, Noël Fonton∗, Josiane Kondaoule∗, Adeline Fayolle∗, Luca Birigazzi, Gaël Sola, Anatoli Poultouchidou, Carlo Trotta, Hervé Maïdou∗ ∗ Regional REDD+ project of the Forests Commission of Central Africa IUFRO 2014 Allometric equations Friday Oct 10th 1 / 14
  • 2.
    REDD+: Decision 4/CP.15(CoP 15, Copenhagen, 2009) 2006 IPCC Guidelines IUFRO 2014 Allometric equations Friday Oct 10th 2 / 14
  • 3.
    Allometric equations incentral Africa 5 studies in central Africa published since 2010 819 trees measured }unpublished data }published data mjhe n trees IUFRO 2014 Allometric equations Friday Oct 10th 3 / 14
  • 4.
    Chain of errorpropagation IUFRO 2014 Allometric equations Friday Oct 10th 4 / 14
  • 5.
    Inventory data M'Baïki permanentsample plots in the Central African Republic 40 ha of permanent plots monitored since 1982 control plots, logged plots, logged + thinned plots, perturbation by re data of 1987 (after all treatments): dbh, species (→ wood density) twelve 1-ha plots (pseudo-replicates) of undisturbed forest twelve 1-ha plots (pseudo-replicates) of disturbed forest Emission factor = (biomass of undisturbed plots) − (biomass of disturbed plots) IUFRO 2014 Allometric equations Friday Oct 10th 5 / 14
  • 6.
    Biomass equations 4 biomassequations with the datasets used to t the models Author Type Model n Chave et al. (2014) pantropical B = f(D, H, ρ) H = f(D, E) 4004 Ngomanda et al. (2014) local (northeastern Gabon) B = f(D, ρ) 101 Djomo et al. (2010) local (southern Cameroon) B = f(D, ρ) 71 Henry et al. (2010) local (Ghana) B = f(D) 42 IUFRO 2014 Allometric equations Friday Oct 10th 6 / 14
  • 7.
    Error propagation Monte Carlomethod for: measurement error error due to the uncertainty on the model coecients residual error of the model Error due to the model choice: 1 Models are considered equally likely 2 Or Bayesian model averaging (BMA) is used to assign dierent weights to the 4 models No tree biomass data available at M'Baïki ¯ Training data set for BMA: African data from Chave et al. dataset (n = 1429) IUFRO 2014 Allometric equations Friday Oct 10th 7 / 14
  • 8.
    Biomass in the24 plots at M'Baïki according to 4 biomass equations 5 10 15 20 100200300400500 Plot rank (basal area) Biomass(tonneha−1 ) Chave et al. (2014) Henry et al. (2010) Djomo et al. (2010) Ngomanda et al. (2014) large error due to the model choice if the plot × model interaction is null, this error has no impact on the estimation of the emission factor IUFRO 2014 Allometric equations Friday Oct 10th 8 / 14
  • 9.
    The dierence ofbiomass between disturbed and undisturbed plots depends on the biomass equation Source Df Sum Sq Mean Sq F value p-value model 3 391 484 130 495 175.426 0.001 plot type 1 512 373 512 373 688.790 0.001 model × plot type 3 42 415 14 138 19.006 0.001 residuals 88 65461 744 IUFRO 2014 Allometric equations Friday Oct 10th 9 / 14
  • 10.
    Estimates of theemission factor for the dierent biomass equations 50100150200250300350 Biomass model Emissionfactor(tonneha−1 ) Error sampling measurement coefficients residual Chave Henry Djomo Ngomanda IUFRO 2014 Allometric equations Friday Oct 10th 10 / 14
  • 11.
    Combining the dierentbiomass equations to get a single estimate of the emission factor50100150200250300 Biomass weights Emissionfactor(tonneha−1 ) Error model sampling measurement coefficients residual Equal BMA Weights Model Equal BMA Chave et al. 0.25 0.152 Henry et al. 0.25 0.001 Djomo et al. 0.25 0.639 Ngomanda et al. 0.25 0.207 IUFRO 2014 Allometric equations Friday Oct 10th 11 / 14
  • 12.
    Conclusions At M'Baïki, emissionfactor from intact to degraded forest (logging + thinning) is approximately 150 tonne ha−1, but with a very large uncertainty The choice of the allometric equation is the largest source of error (40% of the square error) when estimating the emission factor 1-ha plot sampling (30%) and the uncertainty on the model coecients (20%) are also important sources of errors Improving the choice of the allometric equation will require additional tree biomass measurements Data base on allometric equations: Globallometree (http://www.globallometree.org/) IUFRO 2014 Allometric equations Friday Oct 10th 12 / 14
  • 13.
    Thanks for your attention Thisstudy was supported by the regional REDD+ project of the COMIFAC GEF trust fund grant n◦ TF010038 World Bank project n◦ P113167 We thank for access to the M'Baïki data base IUFRO 2014 Allometric equations Friday Oct 10th 13 / 14
  • 14.
    Studies on allometricequations in central Africa published since 2010 Map showing the locations of the studies Djomo et al. (2010) Allometric equations for biomass estimations in Cameroon and pan moist tropical equations including biomass data from Africa. For Ecol Manage 260:1873-1885 Ebuy Alipade et al. (2011) Biomass equation for predicting tree aboveground biomass at Yangambi, DRC. Journal of Tropical Forest Science 23:125-132 Dorisca et al. (2011) Établissement d'équations entre le diamètre et le volume total de bois des arbres, adaptées au Cameroun. Bois For Trop 65:87-95 Fayolle et al. (2013) Tree allometry in Central Africa: Testing the validity of pantropical multi-species allometric equations for estimating biomass and carbon stocks. For Ecol Manage 305:29-37 Ngomanda 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 IUFRO 2014 Allometric equations Friday Oct 10th 14 / 14