Transcript Study on the transport and inverse modeling of CO2

Study on the transport and
inverse modeling of CO2
Yosuke Niwa
Ryoichi Imasu, Masaki Satoh
Center for Climate System Research (CCSR),
The University of Tokyo
1
Study on the transport and
inverse modeling of CO2
Yosuke Niwa
Ryoichi Imasu, Masaki Satoh
Center for Climate System Research (CCSR),
The University of Tokyo
1
Overview
1. Uncertainty of CO2 fluxes
2. CO2 flux estimation methods
3. Inverse modeling
4. Flux estimation
5. Comparison with other study
6. Summary
2
Uncertain Surface CO2 Fluxes
global CO2 growth rate
Deforestation
Tierras Bajas
Deforestation,
Bolivia
from NASA
from WDCGG site
Biomass Burning
Global CO2 concentration is
determined almost by CO2 flux at the
Earth surface.
Our understanding of the surface CO2
flux is insufficient.
Bush Fires in
Southern
Mozambique
from NASA
3
Surface CO2 Flux Estimation
 Bottom-Up Approach
• Direct measurement at flux towers or above oceans
• Biosphere model
precise
very few measurement sites.
hard to cover globe
 Top-Down Approach
• Inversion modeling :
derive flux information from atmospheric observation data
relatively more measurement sites
Easy to cover globe
Estimates of CO2 fluxes from several studies show considerable
disagreement.
4
Inverse Modeling
1. Forward Simulation
Atmospheric Tracer Transport Model
a priori data
Observation Data
Surface CO2 flux
a posteriori data
2. Inversion
Bayesian Statistics
5
Inversion Studies
Bousquet et al., 2000 : 19 regions, 1980-1998
Rodenbeck et al 2003 : 8deg. X 10deg., 1982-2001
Patra et al., 2005, 2006: 64 regions
Baker et al, 2006: 22 regions, 1991-2000, TransCom experiment (13 models)
Estimated fluxes are quantitatively very different by inversion set ups,
especially due to transport models
Expanding Measurement Network
WDCGG surface measurement net work
spatial coverage broaden
by air-craft and satellite
measurements
commercial air-line
( JAL Foundation)
more frequent measurement
monthly → hourly
GOSAT
OCO
(JAXA)
(NASA)
A highly sophisticated transport model is needed to use many kinds of data
6
Tracer Transport Model
Nonhydrostatic Icosahedral Atmospheric Model (NICAM)
• Next Generation GCM
• Consistent With Continuity (CWC)
– Tracer transport is completely consistent with air density change
Both mass conservation and Lagrangian conservation are achieved.
Good property for simulation of long-lived tracers
Horizontal Resolution
glevel-05, (dx~240km)
Vertical Layer
z*, 40 layer (~60km)
Advection
Horizontal:
Upwind-biased scheme (Miura, 2007)
Vertical:
2nd centered difference with limiter (Thuburn, 1996)
Cumulus Convection
Arakawa-Schubert (1974)
Boundary Layer
Improved version of Mellor-Yamada2
(Nakanishi-Niino, 2004)
7
Purpose of our study is…
• to know how much our inversed fluxes are different
from other studies and understand the reason of its
difference.
comparing with TransCom3 models
8
Inverse model
c j  [T ( f )] j   si [T (Vi )] j
observation
i
modeled concentration
d j  c j  Tj ( f )
i : region in which flux is estimated
j : observatio n point
f : background flux
T : transport operator
V : basis flux in each region i
s : scaling factor
1
cost function : J  [(Ms  d) T C(d) 1 (Ms  d)  (s  s 0 )T C(s 0 ) 1 (s  s 0 )]
2
seek s which minimize J
estimated : s  s 0  [M T C(d) 1 M  C(s 0 ) 1 ]1 M T C(d) 1 (d  Ms0 )
error covariance of estimated flux :
C(s)  [MT C(d) 1 M  C(s 0 ) 1 ]1
M : response function made by transpo rt model
s : estimated scaling factor
s 0 : a priori scaling factor
C(*) : error covariance of *
(Baker, 2001)
9
Inversion Setup
• Fluxes to be estimated
22 regions (land 11+ocean 11)
for 1991-2000
Observation site used and 22 regions
• Background fluxes
– Biospheric flux:
NEP flux from CASA model
– Fossil fuel emission: CDIAC
– Air-sea exchange:
Takahashi et al., 1999
• a priori estimate and uncertainty
– The same as Baker et al., 2006
• Observations
– GLOBALVIEW-2006, 78 sites
10
Estimated Interannual
Variability of CO2 Fluxes
land
ocean
Global
• Global interannual variability
is simulated consistent with
other 13 models.
• During ‘97~’98 El Nino, the
amplitude of flux vaiability in
tropical area is smaller, while
in southern area larger.
• No difference in Ocean flux
viability
bold line
2 thin lines
background
Northern
Tropical
: estimated flux
: estimated error
: estimated flux of TransCom
Southern
11
Long Time Mean Flux Estimation
Land Flux
Ocean Flux
blue: this study, green: TransCom models
Relatively large sinks and sources can be seen in some areas.
e.g. Boreal N America, Temp. S America, Tropical Asia, Southern Ocean
12
Aggregated Long Term Mean Flux
• Stronger source in Tropical lands and oceans
• Stronger sinks in Southern areas, especially in Southern Lands
13
Why we got strong source in tropical and strong sink in south?
Simulated annual mean surface zonal
CO2 from background flux data
black : TransCom models
red : NICAM
Simulated
inter-hemispheric
difference (IHD)
Simulated IHD by NICAM is smaller than other models.
14
Why Strong Source in Tropical and Strong Sink in South?
Tropical Ocean
Estimated Flux
Tropical Land
IHD
Southern Land
Southern Ocean
red : this study
black : TransCom3
15
Why Strong Source in Tropical and Strong Sink in South?
• In southern area:
– Observed CO2 concentration in southern area is lower than
simulated one and smaller IHD needs stronger sinks .
– Relatively many observation data at ocean area constrain
ocean fluxes, while land fluxes are not constrained.
• Tropical area:
– Strong upward transport dilutes flux information at the
surface (most measurement sites are located at the surface)
• More observation data are needed to constrain fluxes at those areas
16
Summary
• Our understanding of the surface CO2 flux is insufficient.
• Inversion method is one method for estimating surface CO2
fluxes.
• Estimated temporal and spatial flux variability by using
NICAM are generally similar to those by other models.
• Larger flux variability in southern lands and smaller flux
variability in tropical lands during 97/98 ENSO.
• Strong source in tropical and strong sink in southern region.
• Strong sink in southern oceans is related to small IHD
simulated by NICAM.
17
Comparison with Bottom-up Approaches
There are still much differences…
Bottom-up approaches
This Study
Baker et al, 2006
McGuire et al., 2001
Northern Land
-2.8±0.3
-2.6±0.3
-1.3 ~ -0.3
Tropical Land
3.1±0.8
1.9±0.7
-0.2 ~ 0.5
Southern Land
-2.2±0.7
-1.4±0.6
0.0 ~ 0.2
This Study Baker et al, 2006
Takahashi et al., 2002
Northern Ocean
-0.8±0.2
-1.1±0.2
-1.1
Tropical Ocean
1.1±0.3
0.8±0.3
0.9
Southern Ocean
-1.1±0.3
-0.8±0.3
-1.5
This Study
Baker et al, 2006
IPCC, AR4
Global Biosphere
-1.9±0.6
-2.1±0.5
1.0±0.6
Global Ocean
-0.8±0.5
-1.1±0.5
2.2±0.4
18