camels - FastOpt

A global Carbon Cycle Data Assimilation
System (CCDAS) and its link to CAMELS
Marko Scholze1, Peter Rayner2, Wolfgang Knorr3,
Thomas Kaminski4, Ralf Giering4 & Heinrich Widmann3
1st CarboEurope Integration Workshop, Potsdam, 2004
1
2
3
4
FastOpt
QUEST
c
• QUEST is a newly, NERC funded directed programme (5 years).
• QUEST aims to achieve a better qualitative and quantitative understanding
of large-scale processes and interactions in the Earth System, especially the
interactions among biological, physical and chemical processes in the
atmosphere, ocean and land and their implications for human activities.
• QUEST mainly focuses on: (1) the contemporary carbon cycle and its
interactions with climate and atmospheric chemistry; (2) the natural
regulation of atmospheric composition on glacial-interglacial and longer time
scales; and (3) the implications of global environmental changes for the
sustainable use of resources.
• QUEST consists of a core team, strategic activities, fellowships, and
collaborative grants.
• QUEST website: http://quest.bris.ac.uk
CAMELS
Carbon Assimilation and Modelling of the
European Land Surface
CAMELS
an EU Framework V Project
(Part of the CarboEurope Cluster)
CAMELS
c
CAMELS
CAMELS PARTICIPANTS
(the “Jockeys”)









Hadley Centre, Met Office, UK – Coordinator: Peter Cox
LSCE, France
MPI-BGC, Germany
UNITUS, Italy
ALTERRA, Netherlands
European Forestry Institute, Finland
CEH, UK
IES/JRC, EC
FastOpt, Germany
CAMELS
CAMELS AND INVERSE MODELLING
• CAMELS Goals and General Strategy: Combining Inverse
and Forward Model Strategies (material by Peter Cox,
Hadley Centre)
• Carbon Cycle Data Assimilation and Calculation of
Uncertainties (CCDAS consortium)
CAMELS
CAMELS Goals
• Best estimates and uncertainty bounds for the
contemporary and historical land carbon sinks in Europe
and elsewhere, isolating the effects of direct landmanagement.
• A prototype carbon cycle data assimilation system
(CCDAS) exploiting existing data sources (e.g. flux
measurements, carbon inventory data, satellite products)
and the latest terrestrial ecosystem models (TEMs), in
order to produce operational estimates of “Kyoto sinks“.
CAMELS
CAMELS Motivating Science Questions
• Where are the current carbon sources and sinks located
on the land and how do European sinks compare with
other large continental areas?
• Why do these sources and sinks exist, i.e. what are the
relative contributions of CO2 fertilisation, nitrogen
deposition, climate variability, land management and landuse change?
• How could we make optimal use of existing data sources
and the latest models to produce operational estimates of
the European land carbon sink?
CAMELS
Inverse Modelling
Method : Use atmospheric transport model to infer CO2
sources and sinks most consistent with atmospheric CO2
measurements.
Pros : a) Large-scale; b) Data based (transparency).
Cons : a) Uncertain (network too sparse); b) not constrained
by ecophysiological understanding; c) net CO2 flux only
(cannot isolate land management).
CAMELS
Forward Modelling
Method : Build “bottom-up” process-based models of land
and ocean carbon uptake.
Advantages : a) Include physical and ecophysiological
constraints; b) Can isolate land-management effects; c) can
be used predictively (not just monitoring).
Disadvantages : a) Uncertain (gaps in process
understanding); b) Do not make optimal use of large-scale
observational constraints.
CAMELS
The Case for Data-Model Fusion
• Mechanistic Models are needed to separate contributions to the
land carbon sink (e.g. as required by KP).
• Large-scale data constraints (from CO2 and remote-sensing) are
required to provide best estimates and error bars at regional
and national scales.
• Data-Model Fusion = ecophysiological constraints from forward
modelling
+ large-scale CO2 constraints from inverse
modelling
CAMELS
CAMELS Flow
Diagram
Combined ‘top-down’/’bottom-up’ Method
CCDAS – Carbon Cycle Data Assimilation System
Misfit 1
Misfit to
observations
Forward Modeling:
Parameters –> Misfit
CO2 station
concentration
Atmospheric Transport
Model: TM2
Fluxes
Biosphere Model:
BETHY
Model parameter
Inverse Modeling:
Parameter optimization
CCDAS set-up
2-stage-assimilation:
1. AVHRR data
(Knorr, 2000)
2. Atm. CO2 data
Background fluxes:
1. Fossil emissions (Marland et al., 2001 und Andres et al., 1996)
2. Ocean CO2 (Takahashi et al., 1999 und Le Quéré et al., 2000)
3. Land-use (Houghton et al., 1990)
Transport Model TM2 (Heimann, 1995)
BETHY
(Biosphere Energy-Transfer-Hydrology Scheme)
lat, lon = 2 deg
•
•
•
•
GPP:
C3 photosynthesis – Farquhar et al. (1980)
C4 photosynthesis – Collatz et al. (1992)
stomata – Knorr (1997)
Plant respiration:
maintenance resp. = f(Nleaf, T) – Farquhar, Ryan (1991)
growth resp. ~ NPP – Ryan (1991)
Soil respiration:
fast/slow pool resp., temperature (Q10 formulation) and
moisture dependent
Carbon balance:
average NPP = b average soil resp. (at each grid point)
t=1h
t=1h
t=1day
soil
b<1: source
b>1: sink
Calibration Step
Flow of information in CCDAS. Oval boxes represent the various
quantities. Rectangular boxes denote mappings between these fields.
Methodology
Minimize cost function such as (Bayesian form):


1   T -1  
1  
J ( p )  p  p 0  C p 0 p  p 0   M ( p )  D
2
2


T

 

CD M ( p )  D
-1

where


- M is a model mapping parameters p to observable quantities
- D is a set of observations
- C error covariance matrix
 need of  p J (adjoint of the model)
Calculation of uncertainties
• Error covariance of parameters
  J
C p    2
 p i, j
2



1
= inverse Hessian
• Covariance (uncertainties) of prognostic quantities
 
  T
X(p)  X(p)

CX 
 Cp

p
p
• Adjoint, Hessian, and Jacobian code generated
automatically from model code by TAF
Gradient Method
1st derivative (gradient) of


J (p) to model
p:
 parameters

 J (p ) p
yields direction of steepest
descent.
2nd derivative (Hessian)

of J (p):

2
2
 J (p ) p
yields curvature of J.
Approximates covariance of
parameters.

cost function J (p)

Model parameter space (pp)
Figure from Tarantola, 1987
Data fit
Seasonal cycle
Barrow
Niwot Ridge
observed seasonal cycle
optimised modeled seasonal cycle
Global Growth Rate
Atmospheric CO2 growth rate
Calculated as:
C GLO B  0.25C SPO  0.75C MLO
observed growth rate
optimised modeled growth rate
Parameters I
• 3 PFT specific parameters (Jmax, Jmax/Vmax and b)
• 18 global parameters
• 57 parameters in all plus 1 initial value (offset)
Param
Initial
Predicted
Prior unc. (%)
Unc. Reduction (%)
fautleaf
c-cost
Q10 (slow)
t (fast)
0.4
1.25
1.5
1.5
0.24
1.27
1.35
1.62
2.5
0.5
70
75
39
1
72
78
b (TrEv)
b (TrDec)
b (TmpDec)
b (EvCn)
b (DecCn)
b (C4Gr)
b (Crop)
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.44
0.35
2.48
0.92
0.73
1.56
3.36
25
25
25
25
25
25
25
78
95
62
95
91
90
1
Parameters II
Relative Error Reduction
Carbon Balance
Euroflux (1-26) and other
eddy covariance sites*
net carbon flux 1980-2000
gC / (m2 year)
*from Valentini et al. (2000) and others
latitude N
Uncertainty in net flux
Uncertainty in net carbon flux 1980-200
gC / (m2 year)
Uncertainty in prior net flux
Uncertainty in net carbon flux from prior values 1980-2000
gC / (m2 year)
NEP anomalies: global and tropical
global flux anomalies
tropical (20S to 20N) flux anomalies
IAV and processes
Major El Niño events
Major La Niña event
Post Pinatubo period
Interannual Variability I
Normalized CO2 flux and ENSO
Lag correlation
(low-pass filtered)
ENSO and terr. biosph. CO2:
Correlations seems strong with
a maximum at ~4 months lag,
for both El Niño and La Niña
states.
Interannual Variabiliy II
Lagged correlation on grid-cell basis at 99% significance
correlation coefficient
Low-resolution CCDAS
• A fully functional low resolution version of CCDAS, BETHY runs
on the TM2 grid (appr. 10° x 7.8°)
• 506 vegetation points compared to 8776 (high-res.)
• About a factor of 20 faster than high-res. Version -> ideal for
developing, testing and debugging
• On a global scale results are comparable (can be used for preoptimising)
Conclusions
• CCDAS with 58 parameters can fit 20 years of
CO2 concentration data; ~15 directions can be
resolved
• Terr. biosphere response to climate fluctuations
dominated by El Nino.
• A tool to test model with uncertain parameters
and to deliver a posterior uncertainties on
parameters and prognostics.
Future
•
•
•
•
Explore more parameter configurations.
Include missing processes (e.g. fire).
Upgrade transport model and extend data.
Include more data constraints (eddy fluxes, isotopes,
high frequency data, satellites) -> scaling issue.
• Projections of prognostics and uncertainties into
future.
• Extend approach to a prognostic ocean carbon cycle
model.