FOREST PRODUCTION MODELS

1
FOREST PRODUCTION MODELS Towards improved
interactions between experimentalists and
modellers

• Ross McMurtrie,
• Belinda Medlyn,
• Dave Pepper
• School of Biological, Earth and Environmental
  Sciences,
• University of New South Wales, Sydney, NSW,
  Australia

2
Themes in this talk

• Contrast models applied to Eddy-covariance data,
  versus, Forest-stands experiments

Eddy-covariance data Forest-stand data
Data availability Lots, hourly Less plentiful, More diverse, Infrequent, Longer timestep
Model-data links Good Weaker
Processes Up-scale leaf to canopy Radiation, Photosynthesis, Respiration, Stomata Integrate over time Carbon allocation Acclimation
3
Themes in this talk

• Contrast models applied to Eddy-covariance data,
  versus, Forest-stands experiments
• Plant ecosystem feedbacks are crucial.
• Can models be simplified?
• Model parameterisation
• What parameters can be estimated from given data
  set?
• If wish to estimate particular set of parameters,
  what data are required ?

4
Timescales

• Physiological Minutes to Hours
• Experimental Months to Years
• Human Decades

5
Forest-growth models
Models of growth but not biogeochemical cycling
Forest-field Experiments
Organisational
Scale
Water balance
Nutrient availability
Competition
Community /
Competition
(short
-
lived
Ecosystem
lived species)
(long-
-
species)
Growth
Allocation
Adaptation
Plant
Turnover
Nutrient uptake
Photosynthesis
Adaptation
Stomatal
conductance
Acclimation
Leaf / Organ
Respiration
Timescale
(10
yrs)
x
Physiological
Experimental
Human
Evolutionary
(Minutes to Hours)
(Months to Years)
(Decades)
(Centuries )
6
Eddy-covariance data versus Forest-stand
experiments
EC
Region
Water balance
Nutrient availability
Competition
Community /
Competition
(short
-
lived
Ecosystem
(long
-
lived species)
species)
Growth
Allocation
Adaptation
Plant
Turnover
Nutrient uptake
Photosynthesis
Adaptation
Stomatal
conductance
Acclimation
Leaf / Organ
Respiration
Timescale
(10
yrs)
x
Physiological
Experimental
Human
Evolutionary
(Minutes to Hours)
(Months to Years)
(Decades)
(Centuries )
7
Eddy-covariance data versus Forest-stand
experiments
EC
Region
EC
Water balance
Nutrient availability
Competition
Community /
Competition
(short
-
lived
Ecosystem
(long
-
lived species)
species)
Growth
Allocation
Adaptation
Plant
Turnover
Nutrient uptake
Photosynthesis
Adaptation
Stomatal
conductance
Acclimation
Leaf / Organ
Respiration
Timescale
(10
yrs)
x
Physiological
Experimental
Human
Evolutionary
(Minutes to Hours)
(Months to Years)
(Decades)
(Centuries )
8
What do we learn from successes failures in
flux modelling?
An example Data set Hourly Canopy CO2-flux
data Timescale 1 year Site Sitka spruce,
Griffin, Scotland Model BEWDY sun-shade
model Farquhar photosynthesis model
autotrophic respiration
Medlyn, Robinson, Clement, McMurtrie (2005) On
the validation of models of forest CO2 exchange
using eddy covariance data. Tree Physiol. (in
press).
9
Site Griffin
Sitka spruce forest, Scotland. 18 y.o.
plantation.
10
Illustrative Model
Respiration (1) Simple Q10 dependence on soil
T R R0 Q10 (T/10)
11
Model 2 Carbon substrate dynamics Acclimation
of respiration
Feedback Timescale 3-10 days
12
Current Conditions vs. T 2
Model 1
Model 2
Data
Medlyn et al (in press) Tree Physiology
13
Issues raised by respiration example

• We ignore feedbacks at our peril.
• What extra data would be required to discriminate
  
• between the 2 respiration models?
• Potential for model simplification
• At acclimation NPP/GPP is conservative.

14
Do similar ideas apply to other processes?
Respiration Y Photosynthesis ? Radiation
interception ? Stomatal conductance ? Allocation
?
15
Parameter Sensitivity for BEWDY
First ask Is model sensitive to parameters for
photosynthesis, radiation, stomata, allocation?
Rank Name djmsqr 1 aJ 0.42 2 LAI
0.40 3 Jmax 0.24 4 Vcmax 0.22 5 Q10
0.21 8 g1 0.11 12 kext 0.032
1. Which of the sensitive parameters are
identifiable?
Calculations UNCSIM www.uncsim.eawag.ch
2. Which of the sensitive parameters undergo
acclimation?
Medlyn et al (in press) Tree Physiology
16
Do similar scaling issues apply to other
processes?
Process Model 1 Model 2 Simplest model
Respiration Q10 Substrate-dynamics Constant NPP/GPP
Leaf photosynthesis Farquhar Photosynthetic acclimation Light-use efficiency
Radiation interception Sun-shade model Big-leaf model LUE (1-e-k LAI)
Stomatal conductance Response to VPD (Ball Berry) Response to ABA (Tardieu Davies) Water-use efficiency
Allocation Hydraulic architecture (Comstock Sperry) Simple hydraulic arch (Magnani) Constant allocation coefficients
17
A model experiment with GDAY ecosystem model
- Response to 2 X CO2
Norway spruce, Flakaliden, Sweden
Is simulated response to 2 X CO2 sensitive to
following parameters? Respiration (Ratio
NPP/GPP) Photosyn (Jmax, Vcmax) Radiation
(kext, beam fraction) Stomata (Ci/Ca) Allocation
(RootShoot allocn) Soil N feedbacks (Soil NC,
Ndep)
CO2 chambers
18
What is NPP response to 2 X CO2 sensitive to ?
Relative sensitivity
NPP/GPP ratio
Vcmax, Jmax
Ci/Ca
kext
beam fraction
RootShoot allocation
Ndeposition
Soil NC
Calculations UNCSIM package
19
Conclusions

• Applying ecosystem models to forest experiments
  is
• a challenge because of complex interactions
  feedbacks on various timescales.
• Statistical methods are available to determine
• What parameters can and cannot be estimated from
  experimental data?
• What extra data would make it possible to
  estimate specific parameters?
• Potential for model simplification
• Provided NPP/GPP, LUE, WUE are well-behaved on
  longer timescales.
• Links between models experiments are improving.

20
The End
21
An illustration with the GDAY model Generic
Decomposition And Yield
22
Sune Linder (2004)
23
Summary
Processes MODELS FEEDBACK INSIGHT FROM
Canopy C H2O fluxes MAESTRA CANOAK C substrate dynamics respiration Theoretical C substrate model
Allocation MATE HYDRALL(Magnani) ECOPHYS CABALA 3-PG C allocation Simple model of hydraulic constraints
Soil N GDAY CENTURY Soil N feedbacks Conservation of ecosystem C N
Less complex formulations of canopy processes
feedbacks. Longer time-step
More comprehensive. Longer time-scale
24
Illustrative Model
Photosynthesis BEWDY model (sun-shade
type) Leaf-level photosynthesis Pl (a Il
Pmax,l )/ (a Il Pmax,l) where a and Pmax,l
determined from Farquhar model. Parameters Jmax,
Vcmax (T-dependent), aJ, Ci Ca
Canopy photosynthesis assume Pmax,l varies with
average irradiance level in canopy. Radiation
incident on sunlit foliage Isun(L) a k Ib
Id exp(-kL) Radiation incident on shaded
foliage Ishade(L) a k Id exp(-kL) Sunlit
fraction at depth L exp(-kL). Integrate from 0
to Lmax.