amospheric inversions: investigating the recent inter-annual flux variations ! p. peylin, c....
TRANSCRIPT
Amospheric inversions:
Investigating the recent
inter-annual flux variations !
P. Peylin, C. Rödenbeck, P. Rayner, experimentalists, …
• Inverse models • Data signal• Net carbon fluxes ?• inter-annual flux time series• 2003 summer flux annomaly
2 independent inversions (Similarities / differences) :
LSCE MPI
Time-dependent Bayesian Inversion / Solve for “pixel” fluxes
LMDz (2.5° x 3.7°) TM3 (4° x 5°)
ORCHIDEE mean fluxes & GFED priors
Distance / Biome correlation
“generic” No IAV prior& GFED priors
distance correlation
Observations :
transport model :
Prior information
Inverse approach :
Monthly mean conc. Individual flask / hourly data-> Monthly fluxes -> ~ Weekly fluxes
“CSIRO”
Time – independent140 regions
Montlhy CO2+ 13CO2
Match(4° x 4°)
“CASA model”No IAV prior
Raw data / Fit (use in LSCE inversion) Deviation from a linear fit of winter (DJF) / summer (JAS) mean values
CMN
SCH
CMN
SCH
(Ppm
)
European scale: raw fluxes
Pixgro
Mod17
ANN
JENA_ref
ORCHIDEE
LPJ
JULES
BIOME-BGC
bottum-up range
European scale: raw fluxes
Pixgro
Mod17
ANN
JENA_ref
LSCE_ref
LSCE_ObsJenaORCHIDEE
LPJ
JULES
BIOME-BGC
bottum-up range
European scale: raw fluxes
Pixgro
Mod17
ANN
JENA_ref
LSCE_ref
LSCE_ObsJena
“CSIRO”_ Peter
T3 mean
ORCHIDEE
LPJ
JULES
BIOME-BGC
bottum-up range
Annual land fluxes :
LSCE LSCE(Jena obs)
JENA CSIRO
Europe -0.41
-0.15
-0.36
-0.01
-0.96
-0.87
-0.18
-0.22
N. Asia -0.52
-0.38
0.00
0.28
-0,27
-0,18
-0.67
-0.38
Mean over1997-2003
2003
In GtC / year
-> Impact of fossil fuel emissions : Differences between “Edgar” and “IER” up to ~ 0.2 Gt / year
Net annual fluxes not “robust” yet !
Flux anomalies
filtered fluxes : 120 days
Continental scale: 120 days filtered
Agreement
for the major
anomalies !
JENA_ref
JENA_s99
LSCE_ref
LSCE_ObsJena
European scale: « flux anomalies »
Pixgro
Mod17
ANN
ORCHIDEE
LPJ
JULES
BIOME-BGC
bottum-up range
De-seasonnalised + zero mean+ filtering high freq. (< 120 days)
JENA_ref
European scale: « flux anomalies »
Pixgro
Mod17
ANN
ORCHIDEE
LPJ
JULES
BIOME-BGC
bottum-up range
De-seasonnalised + zero mean+ filtering high freq. (< 120 days)
JENA_ref
LSCE_ref
LSCE_ObsJena
European scale: « flux anomalies »
Pixgro
Mod17
ANN
ORCHIDEE
LPJ
JULES
BIOME-BGC
bottum-up range
De-seasonnalised + zero mean+ filtering high freq. (< 120 days)
JENA_ref
LSCE_ref
LSCE_ObsJena
“CSIRO”_ Peter
T3 mean
European sub-region: (120 days filtering)
North Europe
West Europe
Central Europe
MPI_ref
MPI_s99
LSCE_ObsJena
LSCE_new
bottum-up range
MPI_ref
MPI_s99
LSCE_ObsJena
LSCE_new
bottum-up range
European sub-region: summer anomalies
(Jul-Aug-Sep)North Europe
West Europe
Central Europe
LSCE ref ORCHIDEE
gC/m2/mth
Biome BGC
June – July – August anomalies
LPJ
MPI Ref
JULES
MP
IAnnual anomalies
BIO
ME
LS
CE
2001 2002 2003
LP
JO
RC
HID
EE
JUL
ES
Annual anomalies2001 2002 2003
Robustness is scale dependant
Uncertainties increase with decreasing spatial scale
Prior fluxes & errors / correlations are critical !
Summary
Major flux anomalies are seen by
two completely independent inverse approaches Net annual fluxes : remain “uncertain” at European scale
But
Future• Synthesis under preparation !
• Use additional data CCDAS approach !
• Using regional/better Models & more data will reduce the uncertainties
• Pixel based inversion
• LMDz zoomed over Europe
(0.5 x 0.5 degres over Europe)
• Daily fluxes
• Using “Pseudo-data”
• 10 sites (continuous)
• Prior fluxes from TURC model + random noise
• TRUE fluxes from ORCHIDEE + random noise
Potential of the current network :
perfect transport experiment !
“Correlation” &
“Normalized standard
deviation”
between
“True fluxes” and
“Estimated fluxes” Sp
atia
l ag
gre
gat
ion
(km
)
temporal aggregation (days)
Correlation prior NSD prior
Correlation posterior NSD posterior
LSCE ref LSCE (ObsMPI )
MPI Ref
gC/m2/mth
MPI old case
June – July – August anomalies
Error reduction on estimated COError reduction on estimated CO22 fluxesfluxes
2001 surface network Future surface network
% of error reduction
Carouge, phd, 2006.
Case with large noise (equivalent to real data inversion)
compute “Correlation” &“Normalized standard deviation” between
“True fluxes” & “Estimated fluxes”
Raw data / Fit (use in LSCE inversion) Deviation from a linear fit of summer (JAS) mean values
Monte Cimone
Schauinsland
(Pp
m)
Atmospheric data :
Annual land fluxes :
LSCE LSCE(Jena obs)
JENA JENA
(old ref)
Europe -0.40
-0.05
-0.26
+0.21
-1.24
-0.97
-0.96
-0.87
N. Asia -0.52
-0.38
0.00
0.28
-0.87
-0.86
-0,27
-0,18
Mean over1997-2003
2003
In GtC / year
-> Impact of fossil fuel emissions : Differences between “Edgar” and “IER” up to ~ 0.2 Gt / year
Net annual fluxes not “robust” yet !