Where does uncertainty stem from?
The role of wood density in carbon pool assessments of tropical forests
Johannes Dietz 1 , Christian Wirth 1, Annette Freibauer 1, Joe N. Pokana 2
1 Max-Planck-Institute for Biogeochemistry, Jena, Germany2 Papua New Guinea Forest Research Institute, Lae, Papua New Guinea
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
ScopeScope
Accurate accounting of carbon stocks from tropical forests is also limited by uncertainties in estimates of carbon stock in forest
Crucial factor in the conversion of volume to mass
Highly variable important source for uncertainty
Look-up in databases if floristic data is available
http://www.worldagroforestrycentre.org/sea/Products/AFDbases/WD/ http://www.prosea.nl/
m3 tρ
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
MotivationMotivation
Errors may be caused by:
− Inadequate protocols for wood density determination− Inconsistency with standard terminology (incompatibility)− Low reliability of species identifications (e.g. common names)− Gaps in the data structure (missing values)− Issues in upscaling to forest stand or regional level
Recent studies reveal a substantial bias in upscalingprocedures to regional levels. It is suggested that the emission from the Brazilian Amazon alone could have been overestimated by at least 23 million tons of CO2eq. annually.
(Nogueira et al. 2007)
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Related studiesRelated studies
Region Country
Borneo Malaysia King et al. 2005
Borneo Indonesia Slik 2006
Amazonia Brazil Fearnside 1997
Amazonia Baker et al. 2004
Central Amazonia Brazil Nogueira et al. 2005
Amazonia Brazil Nogueira et al. 2007
Neotropics Chave et al. 2006
Reference
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Case studyCase study
Collaboration with the PNG Forest Research Institute
93 permanent 1 ha sample plots maintained by FRI offer a wealth of information on their forests
Full and repeated inventories of all trees Ø ≥ 10 cm
Floristic data available
Modeling approach to compensate for data gaps- Hierarchical model drawing on Bayesian inference
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Papua New GuineaPapua New Guinea
Sour
ce:
JRC
/ M
PI-B
GC
2006
LowlandRainforest
PermanentSample Plot
Forest cover:76 %(34.6 million ha)
Lowland RF:58 %
93 PSPs activepredominantlyin lowlands
Source: FAO FRA 2005
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Modeling approachModeling approach
Model drawing on Bayesian inferenceInherits predictive power from hierarchical structureSuitable for hierarchical structure of taxonomy
http://www.mobot.org/MOBOT/Research/APweb/welcome.html
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Available dataAvailable data
Species
Genus
FamilyNo data
30,000 Trees
Species
GenusFamily Order
466 Species
Wood density known for the majority of speciesData gaps on species level through:- Missing species wood density data- Identification of trees only at genus level
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Modeling effectModeling effect
Input of variable records from different databasesLinkage to superior taxonomic structure solidifies the estimate already on species level
Species
Measured Wood Density (kg m-3)
Mod
eled
Woo
d D
ensi
ty(k
g m
-3)
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Modeling effectModeling effect
The effect of higher precision reduces uncertainty also on genus and family levelOnly extreme values appear critical
Genera
Averaged Wood Density (kg m-3)
Mod
eled
Woo
dD
ensi
ty(k
g m
-3)
Families
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Genus levelGenus level
Model performance for all genera with > 3 speciesWith modeling, the precision of species level data predicted from genus improves while the effect on residuals is minimal.
R2 = 0.73
0
200
400
600
800
1000
0 200 400 600 800 1000
Species Wood Density (kg m-3)
Aver
age
Gen
us W
ood
Den
sity
(kg
m-3
) R2 = 0.71
0
200
400
600
800
1000
0 200 400 600 800 1000
Species Wood Density (kg m-3)
Mod
eled
Gen
us W
ood
Den
sity
(kg
m-3
)
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
R2 = 0.43
0
200
400
600
800
1000
0 200 400 600 800 1000
Species Wood Density (kg m-3)
Mod
eled
Fam
ily W
ood
Den
sity
(kg
m-3
)R2 = 0.52
0
200
400
600
800
1000
0 200 400 600 800 1000
Species Wood Density (kg m-3)
Ave
rage
Fam
ily W
ood
Den
sity
(kg
m-3
)Family levelFamily level
Prediction from family level is less accurate but still significantly more precise.On family level average approach preferable?
R2 = 0.43
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
Forest type levelForest type level
Average vs. modeling approachMean (93 lowland rainforest plots): + 0.6 %Conservative lower limit: + 4.6 %
0.0
0.5
1.0
1.5
2.0
2.5
0 2 4 6 8 10
Average Approach Error (%)
Mod
elin
g Ap
proa
ch E
rror
(%
)
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
ConclusionsConclusions
Challenges− Dynamic taxonomy (e.g. synonyms)− Databases with inconsistent protocols and standards− Discrepancies between density measures (e.g. air dry density,
wood specific gravity, oven dry weight of green volume)
Perspectives− Knowledge on wood density can reduce the error of carbon pool
estimates from > 20 % even to < 2 %.− Smaller error bars generally result in higher conservative values.− Modeling approach to filling the gaps in wood density data
clearly preferable on the genus level.− Only basic capacities required and efforts very manageable.
Forest Day, COP 13, BaliDecember 8, 2007
Max-Planck-Institutefor Biogeochemistry Wood density implications for carbon pool assessments
• Johannes Dietz Joe N. Pokana− MPI-BGC, Jena, Germany PNG FRI, Lae, Papua New Guinea− [email protected] [email protected]
Thank You for Your AttentionThank You for Your Attention