Estimating regional C fluxes by exploiting observed correlations between CO and CO 2 Paul Palmer Division of Engineering and Applied Sciences Harvard University http://www.people.fas.harvard.edu/~ppalmer
Jan 10, 2016
Estimating regional C fluxes by exploiting observed
correlations between CO and CO2
Paul Palmer
Division of Engineering and Applied Sciences Harvard University
http://www.people.fas.harvard.edu/~ppalmer
IPCC
61 60
5.51.6
+ ballpark flux estimates for fast exchange processes (109
tonnes C)
IPCC
Chinese Government Statistics Shown Downward Trend in Chinese CO2 Emissions
(Streets et al., Science, 294, 1835-1837, 2001)
China Energy Databook v6, 2004
Ch
ina G
DP (
Bill
ion
19
95
yu
an
co
nst
an
t)
Year
Large uncertainty
E = A F
Bottom-up Emission Inventories are Very Uncertain
Emissions (Tg C yr-
1)
Activity Rate (Tg fuel yr-
1) (amount of fuel burned)
Emission Factor (TgC / Tg fuel)
Coal-burning cook stoves in Xian, China
RH + OH … CO CO2
1000s km
Direct & indirect emissions
CMDL site
Many 100s km10s km
Increasing model transport error
Remote data have limitations in estimating regional C budgets
Aircraft data can improve level of disaggregation of continental emissions
110 E 120 E 130 E 140 E 150 E 160 E
Longitude
0 N
10 N
20 N
30 N
40 N
50 N
Lat
itu
de
DC-8 FlightsP-3B Flights
cold front
cold air
warm air
Main transport processes:
DEEP CONVECTION
OROGRAPHIC LIFTING
FRONTAL LIFTING
100 E 130 E 160 E 190 E 220 E 250 E 280 E
Longitude
0 N
10 N
20 N
30 N
40 N
50 N
60 N
La
titu
de
DC-8 FlightsP-3B Flights
Feb – April 2001
NASA TRACE-P
Sources of CO from Asia
Main sink is the hydroxyl radical (OH) Lifetime ~1-3 months
Product of incomplete combustion
BB BF
FF
+Oxidation of hydrocarbons
BF BB
FF
Offshore China
Over Japan
Slope (> 840 mb) = 51
R2 = 0.76
Slope (> 840 mb) = 22
R2 = 0.45
Suntharalingam et al, 2004
ATMOSPHERIC CO2:CO CORRELATIONS PROVIDE UNIQUE INFORMATION ON SOURCE REGION AND
TYPE
- CO2:CO emission ratios vary with combustion efficiency
- Range in regional emission ratios reflect mix of sources and variation in fossil fuel combustion ratio
A priori bottom-up
Top-down
CO CO
CO
2
CO
2
Observation vector y
State vector (Emissions x)
Modeling Overview
Inverse model
x = Fluxes of CO and CO2 from Asia (Tg C/yr)
y = TRACE-P CO and CO2 concentration data
Forward model(GEOS-CHEM)
x = xa + (KTSy-1K + Sa
-1)-1 KTSy-1(y – Kxa)^
y = Kxa +
Jacobian describes CTM
http://www-as.harvard.edu/chemistry/trop/geos/index.html
GEOS-CHEM global 3D chemical transport
model
•Driven by NASA GMAO met data (3/6 hr)
•2x2.5o resolution/30 vertical levels
•O3-NOx-VOC-aerosol coupled chemistry
•Evaluated using ground-based, aircraft, and satellite observations
Consistent CO and CO2 Emissions Inventories Biomass
burning: Variability from observed daily firecount data (AVHRR)
Heald et al, 2003
Anthropogenic emissions for 2001: domestic ff, biofuel, transport, industrial ff Streets et al, 2003
Seasonal Cycle of Chinese CO and CO2 Emissions during TRACE-P
TERRESTIAL BIOSPHERE: CASA (Randerson, et al, 1997) OCEAN BIOSPHERE: Takahashi et al, 1999
Gt
C y
r-1
Fra
cti
on
of
an
nu
al
em
issio
ns
CO
Annual Mean
Streets et al, 2003
TOTAL
FOSSIL
BIOSPHERE
BIOBURN
BIOFUEL
TOTAL
TRACE-P
[OH] from full-chemistry model (CH3CCl3 = 6.3 years)
State vector x = emissions from individual countries and individual processes
Estimating the Jacobian [CO]/COemission
China (CH)
Japan (JP)
Southeast Asia (SEA)
Rest of World (ROW)
Global 3D CTM 2x2.5 deg resolution
Korea (KR)
Boreal Asia (BA)
Linear calculation is straightforward:
JCHBB= [CO]CHBBCOCHBB/emissions
0-2 km
Latitude [deg]
CO
[p
pb
]C
O2 [p
pm
]4-6 km
2-4 km
GEOS-CHEMTRACE-P Observations
Remove CO2 bias using 10th
percentile of [CO2]: 4-4.5 ppm
Linear Inverse Model
x = xa + (KTSy-1K + Sa
-1)-1 KTSy-1(y –
Kxa)
S = (KTSy-1K + Sa
-1)-1
Xs = retrieved state vector (the CO sources)Xa = a priori estimate of the CO sourcesSa = error covariance of the a priori K = forward model operatorSy = error covariance of observations = instrument error + model error + representativeness error
Gain matrix
^
^
Sy Measurement accuracy Representation
Model error (most important)
GEOS-CHEM
Error specification for CO and CO2
Sa Anthropogenic (c/o Streets): China (78%), Japan (17%), Southeast Asia (100%), Korea (42%) – uniform 25% Biomass burning: 50% 30% Chemistry (~CH4): 25% Biosphere: 75%
GEOS-CHEM
2x2.5 cell
TRACE-P
All latitudes
(measured-model) /measured
Alt
itu
de [
km
]
Mean bias
RRE
CO
(y*RRE)2 ~38ppb (CO)
~1.87ppm (CO2)
RRE = total observation error
NUMBER OF EIGENVALUES OF PREWHITENED JACOBIAN 1 =
DOF
K = S KS~
-1/2 1/2
aCO: CH ANTH*, KRJP&, SEA, CH BB, BA BB@, ROW
CO2: CH ANTH*, KRJP&, CH BB$, BA BB@, BS, ROW (inc SEA$)*Collocated sources; &coarse resolution forces merging; $observed gradients too weak to resolve source; @not well resolved
Rodgers, 2000
Independent Inversion of CO and CO2 emissionsA priori
A posteriori
CO
2 e
mis
sio
ns
[Tg
Marc
h 2
001]
CO
em
issio
ns
[
Tg
yr-
1]
Biospheric CO2
Anthropogenic CO2
1~ K
Results consistent with [CO2]:[CO] analysis
•Estimated Chinese anthropogenic CO(CO2) sources are currently too low (high).
•Chinese biospheric CO2 fluxes are estimated too high.
CO2 state vector
A posteriori correlation matrix illustrates the ambiguity between anthropogenic and biospheric CO2
emissions
Chinese anthropogenic CO2
Chinese biospheric
CO2
^C
Monte Carlo approach to modeling correlations between
CO and CO2ECO = (A + AA) (FCO + COFCO)
ECO2 = (A + AA) (FCO2 + CO2FCO2)
Carbon Conservation (CO+CO2 ~ 0.9-1.0)
Perturbed F
N
N
Unperturbed F 10.9
r > 0 CO
Emissions
CO
2
Em
issi
on
s
F
A
CO Emissions
CO
2
Em
issi
on
s
F
A
A >> F A << F
r < 1
Interpretation of correlations
VALUES OF UNCERTAINTY FROM STREETS’ INCONSISTENT WITH DATA ANALYSIS AND LEAD
TO SMALL CO2:CO CORRELATIONS
E = A FA: CO 5-25%; CO2 5-20%
F: CO 50 - 200%; CO2 5-10%Correlations: China ~0 Korea/Japan -0.2 Southeast Asia ~0
Correlations within sectors > lumped sectors
Alternative Correlations Tested…
CO
2:C
O C
orr
ela
tion
Chinese anthropogenic
Korea + Japan
Southeast Asia
Streets’Min(A 25%)Min(A 50%)
Also r = 0.5,…,1.0
A correlation of > 0.7 is needed to start decoupling biospheric and anthropogenic
CO2A
poste
riori
Un
cert
ain
ty
[un
it]
Anthropogenic CO2
Biospheric CO2
Anthropogenic CO
Lowest correlations correspond to those calculated using Monte Carlo method
Future satellite missions
The “A Train”
MODIS/ CERES IR Properties of Clouds
AIRS Temperature and H2O Sounding
Aqua
1:30 PM
CloudsatPARASOL
CALPSO- Aerosol and cloud heightsCloudsat - cloud dropletsPARASOL - aerosol and cloud polarizationOCO - CO2
CALIPSOAura
OMI - Cloud heights
OMI & HIRLDS – Aerosols
MLS& TES - H2O & temp profiles
MLS & HIRDLS – Cirrus clouds
1:38 PM
OCO
1:15 PM
OCO - CO2 column
C/o M. Schoeberl
• Launch date in 2007. • Will provide column CO2
measurements• 3 spectrometers that measure CO2 at
1.61 m and 2.05 m and O2 at 0.76
m• Field of view of spectrometers is 1x1.5
km2 • Sun-synchronous orbit with 16-day
repeat cycle and 1:15 pm equator crossing time
Orbiting Carbon Observatory (OCO)
New Concept: Testing science objectives of satellite instruments before launch
Tropospheric Emission Spectrometer (TES)
• Launched in July 2004• An IR, high resolution Fourier
spectrometer • Measures spectral range 3.3 - 15.4 m• Limb and nadir view (footprint is 8x5
km2)• Sun-synchronous orbit with 16-day
repeat cycle Will measurements of CO and CO2 from TES and OCO provide accurate constraints on carbon fluxes from different regions in Asia?
Jones et al, 2004
Simulation: Constraining Asian Carbon Fluxes from Space
Generate pseudo-data from the satellites for March 1-31, 2001
Inverse model with realistic instrument and model errors, and which accounts for data loss due to cloud cover and the vertical sensitivity of the instruments
CO2 column along OCO orbit (1 day)CO (825 mb) along TES orbit (1 day)
ppmppb
Jones et al, 2004
Significant reduction in uncertainty in estimates of the dominant Asian biospheric fluxes (China and Boreal Asia)
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
China JapanKorea
SEAsia
India BorealAsia
Rest ofworld
A priori
A posteriori
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
China JapanKorea
SEAsia
India BorealAsia
Rest ofworld
A priori
A posteriori
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
China JapanKorea
SEAsia
India BorealAsia
Rest ofworld
A priori
A posteriori
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
China JapanKorea
SEAsia
India BorealAsia
Rest ofworld
A priori
A posteriori
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
China JapanKorea
SEAsia
India BorealAsia
Rest ofworld
A priori
A posteriori
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
China JapanKorea
SEAsia
India BorealAsia
Rest ofworld
A priori
A posteriori
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
China JapanKorea
SEAsia
India BorealAsia
Rest ofworld
A priori
A posteriori
CO Sources
CO2 Sources
Biospheric CO2
A P
oste
riori
Err
or
Esti
mate
s
[%]
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
ChinaFuel
JP/KRFuel
SE AsiaFuel
IndiaFuel
ChinaBB
SE AsiaBB
IndiaBB
BorealAsia BB
Chinese biospheric fluxes weakly coupled to anthropogenic emissions
Jones et al, 2004
Closing Remarks
•Estimated Chinese anthropogenic CO(CO2) sources are currently too low (high).
•Chinese biospheric CO2 fluxes are estimated too high but they are coupled to anthropogenic CO2. Correlations between CO2 and CO can decouple these signals.
• Emission correlations summed over sectors are too weak – need r > 0.7, impossible with current inverse model configuration.
•Work in progress – much still to explore.