pyro-convective cloud from aircraft ~ 10km (n57, w125) june 27, 2004
DESCRIPTION
High Temporal Resolution Inverse Modeling Analysis of CO Emissions from North American Boreal Fires During the Summer of 2004 Importance of Their Injection Height. - PowerPoint PPT PresentationTRANSCRIPT
Solène Turquety – AGU fall meeting, San Francisco, December 2006
High Temporal Resolution Inverse Modeling Analysis of CO Emissions from North American Boreal Fires During the Summer of 2004
Importance of Their Injection Height
S. Turquety1,2, D. J. Jacob1, J. A. Logan1, C. L. Heald4, D. B. Jones3, R. C. Hudman1, F. Y. Leung1, R. M. Yantosca1, S. Wu1, L.K. Emmons5, D. P. Edwards5, G. W. Sachse6
Pyro-convective cloud from aircraft ~ 10km (N57, W125) June 27, 2004www.cpi.com/remsensing/midatm/smoke.html
1Harvard University, Cambridge, USA2Service d’Aéronomie, IPSL, UPMC, Paris, France3University of Toronto, Canada
4University of California Berkeley, USA5NCAR, Boulder, USA 6NASA Langley Research Center, Hampton, USA
• Uncertainty on the fire emissions (area burned, fuel consumed, etc.)• Importance of injection heights more and more recognized but highly uncertain
Solène Turquety – AGU fall meeting, San Francisco, December 2006
19 Tg
11 Tg
We constructed a daily area burned: •Temporal variability: daily reports from the U.S. National Interagency Fire Center•Location of the fires: MODIS hotspot detection
Fuel consumption and emission factors including the contribution from peat burning
Daily inventory of boreal fire emissions for North America in 2004(Turquety et al., submitted, JGR)
Summer of 2004: Largest fire year on record in terms of area burned in Alaska and western Canada;
Pfister et al., GRL, 2005: Inverse modeling a posteriori estimate 30 ± 5 Tg CO emitted based on MOPITT CO ~ twice their a priori estimate
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Evaluation using the MOPITT CO observations(Turquety et al., submitted, JGR)
Highlights the importance of peat burning
Strong uncertainty remain:→ Areas burned/Timing of fires? → Fuel consumption? → Impact of injection heights?
GEOS-Chem: no peat burning
GEOS-Chem: with peat burning
MOPITTModel with peatModel without peat
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Importance of high altitude injection in 2004
Average vertical distribution of boreal fires emissions in the CTM (F-Y Leung):
• 40% boundary layer• 30% FT ~ [600–400hPa]• 30% UT ~ [400–200hPa]
Variability CO emissions and max TOMS AI Alaska-Yukon [165-125W]
Several studies have shown that pyro-convective events occurred – and could explain some long-range transport events : e.g.
→ Damoah et al., 2006 : event end of June→ DeGouw et al., 2006 : event in mid-July
Peaks in TOMS AI suggest pyro-convection events: end of June, beginning of July, mid-July and mid-August
Solène Turquety – AGU fall meeting, San Francisco, December 2006
a priori sources xaa posteriori estimates
Inversion
x̂
Forward model: Observations:amodel xKy xKy obs
Inverse modeling of boreal fire emissions
GEOS-Chem CO * MOPITT AK
+
MOPITT CO – summer 2004
)(1
ˆ aaaa xKySKSKKSxx
TT
111ˆ
aSKSKS T (MOPITT – MODEL)
Gain matrix
Maximum a posteriori solution (Rodgers, 2000)
With S∑ : observation and model error Sa : a priori error K : Jacobians (∂y/ ∂x)
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Kalman Filter
Kalman Smoother
Analysis
Analysis
Analysis
update
tmodel ,y
tobs ,y
t01tt xxx ˆ...ˆˆ
Kalman smoother: observations from ‘future’ also used to update emissions
Time dependant inversion using a Kalman smoother
Initial conditions = MOPITT CO assimilation(D. Jones, U. Toronto)
Solène Turquety – AGU fall meeting, San Francisco, December 2006
t
1
1t
t
t,11tttttttt ε
x
x
x
KKKεxKy
...
...,,
Observations influenced by emissions for current day but also past emissions!
Separate contribution from different time steps in the model
Jacobian K now time dependant:
Time dependant inversion using a Kalman smoother
t1
t2,t1t2,t1 x
yK
with
t
update
t0
tmodel ,y
tobs ,y
Pt1tt xxx ˆ...ˆˆ
Fixed
Each emission time step update P times, last estimate = best estimate
Emissions during 3 days (1 timestep); P = 5 timesteps updated (5 x 3 = 15 days)
GEOS-Chem CO * MOPITT AK
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Model pulse simulations including vertical distribution of the emissions
State vector including vertical distribution:→ 3 biomass burning regions x 3 vertical regions: BL, MT, UT→ North American FF/BF, Asia, Rest of the world + chemical production
GEOS-Chem model simulation to be compared to the MOPITT observations:
background xKymodel
Decaying background : initial conditions = assimilated MOPITT CO (University of Toronto)
Emissions during 3 days (1 timestep); P = 5 timesteps updated (5 x 3 = 15 days)
Solène Turquety – AGU fall meeting, San Francisco, December 2006
t-1
t-2
Observationsy
Forward modelK x + bckgd
Contribution at t from emissions at t-2
Contribution at t from emissions at t
Contribution at t from emissions at t-1
t(3 days timestep)
BB AK-YK – Boundary layer BB AK-YK – Middle trop. BB AK-YK – Upper trop.
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Initial a priori uncertainty on the emissions Sa
• 50% on biomass burning emissions in our region of interest• 30% on emissions for the rest of the world• 20% uncertainty on chemical production1st adjustment of the emissions at a given timestep => errors uncorrelated 2nd adjustment of a given time step: Sa(t,t) = Sx(t,t-1) => introduce correlations
A priori uncertainty on the observations and model Se
• Determined using the method described by Heald et al., JGR, 2004uncertainty = observation – model
• Assume correlation length scale = 147 km
Total CO
Maximum error over the fire region, reflecting the large uncertainties ~ 30 –
50%
~ 5 – 20 % elsewhere
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Inversion of the emissions in 3 vertical regions: boundary layer (BL), middle troposphere (MT) and upper troposphere (UT)
Pyroconvective event end of June
Still update…
Sensitivity of the inversion to injection height, information seems to be available for the inversion of this parameter in parallel
A priori “vertdis”:40% BL, 30% MT, 30%UT
(preliminary results)
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Inversion of the emissions in 3 vertical regions: boundary layer (BL), middle troposphere (MT) and upper troposphere (UT)
Sensitivity of the inversion to injection height, information seems to be available for the inversion of this parameter in parallel
A priori “vertdis”:40% BL, 30% MT, 30%UT
(preliminary results)
Variability CO emissions and max TOMS AI Alaska-Yukon [165-125W]
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Large event in the beginning of August
Inversion of the emissions in 3 vertical regions: boundary layer (BL), middle troposphere (MT) and upper troposphere (UT)
(preliminary results)
Variability CO emissions and max TOMS AI Central Canada
From Alaska
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Conclusions and future directions
• Bottom-up emissions inventory estimate of 30 Tg CO, incl. 11 Tg CO from peat burning [Turquety et al., subm., 2006]• Including peat burning allows better agreement with first top-down estimates of 30 ± 5 Tg by Pfister et al. [2005] • Injection height is important for specific events – less important on CO averaged over the summer
• Injection heights have an impact on high temporal resolution top-down emissions inversions from MOPITT• Limited information on the vertical distribution in MOPITT• Information in the MOPITT transport pathways on injection height can be used to constrain this parameter
• Data could be used to specify injection height together with inventories: → TOMS AI→ POAM stratospheric aerosols (Fromm et al.)→ MISR : see poster Fok-Yan Leung A51C-0099→ Calipso lidar in space?→ Solar occultation measurements from ACE?
• Efforts currently undertaken to include a physical parameterization of injection heights in models• One focus of the POLARCAT international campaign to be held in 2008
Solène Turquety – AGU fall meeting, San Francisco, December 2006
Detection of vertical distribution over source regions and downwind with CALIPSO
MODIS fire detection20-26 July, 2006
Courtesy J. Pelon, Service d’Aéronomie
Solène Turquety – AGU fall meeting, San Francisco, December 2006
CO
C2H6
HCN
(+) Large variety of species measuredO3, H2O, H2O2, CO, CH4, C2H6, C2H2, HCN, CH3Cl, SF6, OCS, HNO3, PAN,…
(+) Very good vertical resolution
(+) Orbit scheduled sample boreal regions in July
(-) Lack coverage (-) No data at altitudes < ~6km
Solar occultation measurementsfrom the ACE/SCISAT-1 instrument: