Interannual variability in CO and ozone as seen by TES and MLS and the GMI Combo model.

1
Interannual variability in CO and ozone as seen
by TES and MLS and the GMI Combo model.

• Jennifer A. Logan, Inna Megretskaia, Lin Zhang,
  and the GMI, TES, and MLS teams
• Harvard University
• NASA/Goddard
• JPL

TES meeting, Feb. 24, 2009.
2
The Global Modeling Initiative (GMI) Combo Model

• Combo tropospheric stratospheric chemical
  mechanism
• GEOS-4 meteorological fields
• 2 x 2.5 resolution
• Aura4 simulation, 2004-2007
• Model output is available to the community for
  Aura science
• Output saved along satellite track at overpass
  times
• GFED2 biomass burning emissions for 2004-2007
• MEGAN inventory for biogenics
• Lightning NOx regional scaling to average
  OTD/LIS lightning spatial patterns (Allen and
  Pickering).
• GEOS-Chem similar, without stratospheric
  chemistry.

3
Satellite data

• TES V003
• Validated with ozonesondes by Lin Zhang, similar
  high bias to V002, 3-10 ppb.
• Filters applied to remove C-shaped ozone
  profiles (Lin Zhang)
• Omit data with cloud optical depth gt2 for
  pressure lt750 hPa
• TES AKs and prior applied to model output
• Uniform prior used for 30N-30S
• MLS data V2.2
• MLS AKs applied to model (makes very little
  difference)

4
Outline

• Use TES CO to evaluate model performance in lower
  troposphere to gain insight into reasons for
  discrepancies with MLS in the upper troposphere
• Can the model match the interannual variability
  in tropical CO and ozone, and if not, why not?

5
GMI model compared to CO at NOAA/GMD surface
sites in the tropics, 2005-2006
Data Model
6
GMI model and MOZAIC aircraft data for CO in the
tropics
Cairo Abidjan Delhi
Caracas
Data Model 2005 Model - 2006
Locations with enough aircraft data for model
evaluation
7
CO in the tropics at 700 hPa, July-Nov. 2005SH
biomass burning season.
TES model
model-TES
Too much export in easterlies in lower trop.,
convection N. of equator similar problems with
MOPITT comparisons (Junhua Liu, GEOS-Chem runs)
Fire emissions too low over Africa
8
CO in the tropics at 700 hPa, July-Nov. 2005SH
biomass burning season.
TES model
model-TES
Too much export in easterlies in lower trop.,
convection N. of equator similar problems with
MOPITT comparisons (Junhua Liu, GEOS-Chem runs)
Fire emissions too low over Africa
9
CO in the tropics at 150 hPa, July-Nov. 2005
MLS model
model-MLS
Observed max. over India is missing in model.
Model is biased low everywhere, except where is
it too high in equatorial band same problem in
LT
Model maximum from convection is one month late,
implying not enough convection in October.
10
Time series over region of Asian maximum in UT CO
The UT maximum at 150 hPa in June-August is too
late in the model. Same problem at 215 hPa.
2005 2006 2007
Papers on high CO seen by MLS over the Himalayas,
and effect of Asian monsoon Li et al., 2005, Fu
et al., Randel et al., Park et al. 2008, 2009
11
CO in the tropics at 700 hPa, July-Nov. 2006
TES model
model-TES
Similar features to 2005 too much export in
equatorial easterlies too low BB emissions over
Africa too high CO near Andes (this appeared
with switch to MEGAN biogenic emissions) Huge
difference in BB emissions from Indonesia (Logan
et al. 2008, Nassar et al. 2009)
12
CO over South America in LT (TES) and UT (MLS)
Model lower than MLS in UT, peaks one month late
in 2005 and 2006. Suggest a problem with timing
of convection, since LT looks good.
2005 2006 2007
Good match with TES in LT in 2005-2006, GFED too
high in 2007 Model w/AK suggests lower CO in
Aug. and Sept. 2005 caused by lack of
sensitivity.
Bench warm-up
GFED CO emissions
TES Model w. AK Model w/out AK
13
S. America - CO from MOPITT and TES in the lower
trop.
2005 2006 2007
MOPITT data also show 2006 had lowest CO over S.
America in BB season.
14
CO over Southern Africa in LT and UT
2005 2006 2007
GFED emissions too low over S. Africa, but timing
looks OK. UT max. is a month too late over S.
Africa also. GFED emissions similar each year
GFED CO emissions
15
CO in the tropics at 700 hPa, Dec. 2005-April
2006NH biomass burning season.
TES model
model-TES
High CO near the Andes Largest differences
related to BB emissions in N. Africa in March
April.
16
CO in the tropics at 700 hPa, Dec. 2005-April
2006
MLS model
model-MLS
17
CO in the tropics at 700 hPa, Dec. 2006-April
2007
BB emissions from N. Africa appear to be too
high, or transport out of source region too
strong. BB CO is transported south and west
18
CO time series over Equatorial Africa
2005 2006 2007
Model CO decreases a month too soon in LT,
implying emissions in February are too
low. Model UT maximum in Feb.-April similar to
timing in MLS data. But since CO decreases too
soon in the LT, caution is needed in
interpreting the MLS comparison.
x
GFED CO emissions (N. Africa)
19
CO time series over Indonesia
2005 2006 2007
See Nassar et al. (2009) for detailed discussion
of Indonesia in late 2005 and 2006 (El Nino).
20
CO tape recorder, 10ºN- 10ºS
MLS
Means for 2005 subtracted from time series
GMI Combo
CO from Indonesian fires
The GMI Combo model looks pretty good.
Interannual variability driven by CO fire
emissions, especially from Indonesia.
Interannual variability in emissions in NH fire
season apparent (Jan.-April).
Update of Schoeberl et al. (GRL, 2006), see also
Combo model study of Duncan et al. (JGR, 2007),
with GCM met. fields.
21
Issues with V003 ozone data (and V002)
Some retrievals had C-shaped profiles,
identified in V002 by Helen Worden, in validation
with IONS data over N. America. Test devised to
remove them. The original C-test removed some
valid looking profiles in the tropics over e.g.,
North Africa. Lin Zhang devised a better test,
based on validation of V003 data. See example to
left.
22
Ozone in the tropics, July-November, 2005
TES
Model - TES
model
The worst model agreement globally is in the S.
Atlantic in Sept.-Nov. (sonde data shows the same)
The problem is confined to Atlantic
sector. Outflow to Indian Ocean is OK
23
Ozone in the tropics, July-November, 2006
TES
Model - TES
model
Discrepancies are much smaller in 2006. TES is
lower in 2006, and model is higher.
24
July-November, interannual variability in TES data
2005
2006
2007
25
Ozone in Oct 2006, 6º-14ºS
TES vertical resolution 6 km!
GMI
S. Amer.
Africa
GMI
Ascension Island sondes, 8S
Problem is in LT not UT. Ozone too low over
Africa, so outflow from Africa does not supply S.
Atlantic with enough ozone in easterlies
26
Ozone over South America, Nov. 2004-Dec. 2008
In the model, ozone is related to lightning
NOx (but not so simple) LIS data show more
lightning in 2006 than 2005. TES data and
OMI/MLS data show higher ozone in the South
Atlantic region in 2005 and 2007. Independent
data from OMI/MLS confirms the IAV in the TES
ozone data.
200 hPa
500 hPa
Lightning NOx
Sept. in red
Sauvage et al., Martin et al. lightning NOx is
main source of ozone in the tropics
27
Ozone over Southern Africa, Nov. 2004-Dec. 2008
Lightning NOx
Oct.
28
Ozone in the tropics, Dec. 2005-April 2006
Discrepancies smaller than in Sept. Nov.
29
North Africa
30
Indonesia
For a detailed analysis of this region using
GEOS-Chem, and the effects of the El Nino in late
2006 (and the huge fires in Borneo), see Nassar
et al. (2009).
31
Tropospheric ozone column from TES and OMI/MLS
OMI/MLS products OMI total O3 column - MLS
strat. ozone OMI scans, so has better global
cover than MLS. Variability in TES and OMI/MLS
products is essentially the same.
TES column (integrated profile) Schoeberl product
(uses trajectories to fill in MLS) Ziemke/Chandra
product
32
Conclusions

• Interpretation of MLS CO in upper trop. requires
  careful analysis of CO data in the lower trop.,
  as errors in LT propagate to the UT.
• Over S. America, S. Africa GEOS-4 max. convection
  appears to be a month too late, but hard to tell
  for N. Africa as errors in LT CO.
• Need to look at convective mass fluxes in model
• TES reveals interannual variability in tropical
  ozone, but model has problems matching this in S.
  Atlantic, likely due to lightning NOx.
• See poster by Junhua Liu for analysis of model
  meteorology and NOx in 2005/2006
• TES and OMI/MLS products show similar variability