Risk and Uncertainty in
Forest Carbon
Sequestration Projects
Cris Brack
Forestry, ANU
Risk and Uncertainty
Optimal decisions change when
risk or uncertainty explicitly
recognised
Risk: multiple outcomes with
known probability (contrast with
the loss function defined by
Statistical Decision Theory)
Uncertainty: outcome probabilities
unknown
Categories of risk / uncertainty
Forest dynamics or growth
Inventory or stock
Preference function
Internal sources
– Simplifications required for models
– Inaccuracies in data or projections
External sources
– Changing nature of desired state
– Improper specifications of returns
Carbon Pools
Tree biomass
– Bole
– Bark, twigs, leaves
– Roots
Soil
Litter and debris
Products (off-site)
Change or Standing Inventory
Amount sequested between 2008
and 2012
Amount present in 2008 and 2012
Independence of estimates
CAMFor
Carbon Account Modelling for
Forests
Developed by NCAS (AGO)
Based on FORTRAN code of
CO2Fix
– Modifications to number of pools and
management activities
Inputs to CAMFor
Bole volume increment (CAI m3ha-1yr-1)
Relative allocation to branches,
bark, leaves, twigs, roots
Rates of transfers between pools
and atmosphere
Density and Carbon Content
Soil
Inputs to CAMFor
Management regime
– Intensity and timing of harvests
– Products
– Area established by year
Fire
Schematic of CAMFor
Growt h
St emVolIncTbl( Rot Age)
xBasicDens
x Sit eAdjust St em
Decompo
Trees
Product s
St ems
St emM, CFracMain
xBranMIncTbl( Rot Age)
xSit eAdjust Bran
Branches
BranM, CFracMain
xBarkMIncTbl( Rot Age)
xSit eAdjust Bark
Bark
BarkM, CFracLit
xLeaf MIncTbl( Rot Age)
xSit eAdjust Leaf
Leaves & Twigs
Leaf M, CFracLit
xRoot MIncTbl( Rot Age)
xSit eAdjust Root
Decomposit ion
DcyM x DcmpFracDcy
Root s
Root M, CFracDcy
Turnover
BranM x BranTurnFrac
Turnover
BarkM x BarkTurnFrac
Turnover
Leaf M x Leaf TurnFrac
Decay Pool
DcyM, CFracDcy
Lit ter
Lit M, CFracLit
Humif icat ion
Lit M
x Humf FracLit
x Xf erFracLit Hum
Encapsulat ion
HumM
x EncpFracHum
x Xf erFracHumInrt
Inert Pool
Inrt M, CFracInrt
Pulp and Paper
PaprM, CFracMain
PaprM x DcmpFr
Packing Wood
PackM, CFracMain
PackM x DcmpFr
Furnit ure
FurnM, CFracMain
FurnM x DcmpFr
Fibreboard
FibrM, CFracMain
FibrM x DcmpFra
Const ruction
ConsM, CFracMain
ConsM x DcmpF
Mill Residue
ResiM, CFracMain
ResiM x DcmpFra
Deadwood
DwdM, CFracMain
Decomposit ion
DwdM x DcmpFracDwd
Soil
Humus
HumM, CFracHum
FuelM x DcmpFr
Turnover
Root M x Root TurnFrac
Debris
Humif icat ion
DcyM
x Humf FracDcy
x Xf erFracDcyHum
Decomposit ion
HumM x DcmpFracHum
Bio-Fuel
FuelM, CFracMain
Decomposit ion
Lit M x DcmpFracLit
CAI m3ha-1yr-1
Modelled growth
Assumptions about model
coefficients
Localised biases in output
(weather cycles)
Model domain
Bias and precision of input
– Site Index
Model imprecision
Localised bias
Modelled risk in CAI
Allocation to other biomass
pools
 Proportional
allocation
 Annual movement
between pools
 Multipliers to
original fractions
to ensure pool
ratios (expansion
factors) reasonable
 Simple
correlations
assumed
Simulation of growth change
 Nth Coast NSW
Eucalyptus
plantation
 Sequestration
from 2008-2012
(tree carbon t/ha)
 Plantation
established in
1990
 No harvest or fire
Dominating risks
 Localised
weather
biases
 Density
Carbon
content
 Site Index
 Allocation of
annual
growth
Weather/2011
Weather/2006
Weather/1996
Weather/2001
Model/2011
Density
Model/2006
Model/2001
Model/1996
Expansion/1996
Expansion/2001
Weather/1991
Expansion/2011
Root/1996
Model/1991
Root/2001
-0.5
-0.25
0
0.25
Standard error of regression coefficient
0.5
0.75
Simulation of different
establishment years
 Sequestration from 2008-2012
 Maximum sequestration for 2002 - 2006
 Maximum imprecision in same period
 Unequal variations
Simulation of management
impacts
 Partial thinning at age 12 years
 Plantations established between 1990
and 2000 (harvest before end of Kyoto
Commitment Period)
Simulation of full estate
 Soil carbon
– 100 - 300 t(C)ha at establishment
– Decrease 0.97year-1 for 5 years (0.94 - 1.0)
Simulation of full estate
Mapping error
– Boundaries within 5 or 10 m of true
– Error in area can exceed 40% for
small plantations with systematic 10
m boundary error
Management
– Estate of 500 ha planted each year
from 1990 - 2010 (area boundary
within 5 m)
– Thinned at age 12 years
Carbon (t/ha) sequested
Conclusions
Predicting change is different to
predicting standing stock
Variability in the change for a given
period is influenced by:
– Actual growing conditions in that
period
– Relative location on the CAI curve
Management options
Risk and Uncertainty in
Forest Carbon
Sequestration Projects
Cris Brack
Forestry, ANU