A parametric and process-oriented view of the carbon system

1
A parametric and process-oriented view of the
carbon system
2
The challenge explain the controls over the
systems response
3
Carbon emissions and uptakes since 1800 (Gt C)
4
Expanding the model
A model for (Fba-Fab) Fab ?G(Di, pi, S i)
photosynthesis Fba ?G(Di, pi, S i)
respiration and fire
5
A Hierarchical view of the carbon system
Causation goes in this direction

• Drivers (weather, nutrients, fires)

Fluxes
Concentrations
Inverse models do something is this direction
6
A-R A key feature of the system

• What we measure Net Ecosystem Exchange
• (the flux of CO2 across an imaginary plane above
  the canopy)
• But NEE cannot be directly parameterized
• NEE Photosynthesis - Respiration
• The model (or observation equation) must
  transform the observation (NEE) into physically
  modeling components.
• This is neglecting complex but different
  processes such as fire and forest harvest.

7
Ecosystem Model Structure
Photosynthesis (Phenology,Soil Moisture, Tair,
VPD, PAR)
Plant Respiration (Plant C, Tair)
Plant Carbon
Precip.
Transpiration
Litterfall (Plant C, Phenology)
Soil Respiration (Soil C, Soil Moisture, Tsoil)
Soil Carbon
Soil Moisture
Drainage
8
Some key model equations

• NEE Ra Rh - GPP
• GPPmax AamaxAdRleaf
• GPPpot GPPmaxDtempDvpdDlight
• Rh CsKhQ10sTsoil/10(W/Wc)
• GPP canopy photosynthesis, R denotes
  respiration, Amax max leaf-level carbon
  assimilation, Ds are scalars for environmental
  factors, Ad, a scaling factor over time, Cs
  substrate, K, rate constant, Q10 the temperature
  scalar and W, water scalars.

9
Estimation

• (zj - H(Fapj,Fpaj))tR-j1 (zj - H(Fapj,Fpaj))/2
  (pj - Pj)tR-j1 (pj - Pj) /2
• The rubber bands are the prior estimates of
  parameters

10
Assimilation of fluxes provides consistency
between prior knowledge and observed carbon
exchange
11
Control variables

• Temperature
• Soil moisture
• Nutrient availability
• Fire regime
• Light interception
• Land management
• Atmospheric CO2
• etc

12
Concentrations have less information about
processes and parameters than do fluxes

• Why?
• They are one step more removed (by transport)
• That step includes invertible (advective)
  processes and irreversible (diffusive) processes
• There is information loss along the chain of
  causation

13
Get closer to the answer measure fluxes
Tower-based measurements
14
FLUXNET
15
(No Transcript)
16
More gadgets
My little flux tower.
17
More gadgets

• CO2, H2O T, u,v,w

w
18
(No Transcript)
19
Time-scale character of carbon modeling

1. Variability is at a maximum on the strongly
   forced time scales
2. They have an annual sum of 0
3. Modeling the carbon storage time scales (years)
   is the goal

Diurnal
Seasonal
20
Observed variability of fluxes
21
Analyzed variability of processes
22
Analysis of controls
Warm springs accelerate growth but also
evaporation. Despite the overall positive
response shown earlier, the annual relationship
of flux to temperature is negative
23
Self-consistent parameter sets
Fit to the diurnal cycle (12 hour time steps)
Fit to daily data 24 hour time steps
24
Assimilating water and carbon
Just water
Carbon only or carbon plus water
25
Adding water doesnt help carbon, but it helps
water
Carbon only Carbon and water
26

• Evaluation against an independent water flux
  measurement

27
Normal Model Parameterization Method
28
Step 2..
29
Self-consistent parameter sets
Range from prior knowledge
Validate-tune
Second parameter dictated
First parameter
30
Analysis of controls
The emergent Relationship of temperature and
carbon uptake. Note the multiple Regimes. The
lower lines are the water-limited response
Realized T response wet
Realized T response, dry
31
What does this type of local study contribute to
global modeling?

• We can use this to understand the information in
  different types of observation

32
Carbon from space
OCO uses reflected sunlight to make measurements
during the day
33
Day and Night

• Remember, weve shown a huge loss of process
  information without diurnal information

34
Future active CO2 experiments make day and night
observations
LIDAR
35
Process priors for global models
Tower-based estimates of parameters can be used
as priors to invert global concentration data to
estimate parameters controlling fluxes instead of
fluxes (Knorr, Wofsy, Rayner)
36
The global scale is very distant from processes

• Distributed local measurements and innovative
  measurement approaches can bridge the gap

37
(No Transcript)
38
(No Transcript)
39
ACME prepares for its first flight
40
Vertical profiles and CO2 lakes
41

• Carbon data assimilation

Carbon data assimilation and parametric
estimation are fast-moving fields
42
A few references

• Vukicevic, T., B.H. Braswell and D.S. Schimel.
  2001. A diagnostic study of temperature controls
  on global terrestrial carbon exchange. Tellus (B)
  53150-170. (variational)
• Braswell, B.H., W.J. Sacks, E. Linder and D.S.
  Schimel. 2004. Estimating ecosystem process
  parameters by assimilation of eddy flux
  observations of NEE. Global Change Biol.
  11335-355 (MCMC)
• Williams, M. Schwarz, B.E. Law, J. Irvine, and
  M.R. Kurpius. 2005. An improved analysis of
  forest carbon dynamics using data assimilation.
  Glovbal Change Biol. 1185-105 (EKF)
• Wang, Y-P. and D Barrett. 2003. stimating
  regional terrestrial carbon fluxes for the
  Australian continent using a multiple-constraint
  approach. I. Using remotely sensed data and
  ecological observations of net primary
  production. Tellus (B) 55270-289 (Synthesis
  inversion)