Embed Size (px)
Citation preview
Evaluation of the GMI Combo Model
Yasuko Yoshida and Bryan Duncan
Goddard Earth Science & Technology (GEST) Center, UMBC
Introduction
We are evaluating Combo simulation results using surface, aircraft and satellite measurements.
Three simulations:
1 ) GEOS-4-DAS (2004-5)
2) FVGCM (1994-1998 SSTs)
3) GEOS-4-Forecast (July 2004-June 2005) –preliminary results
Projects
Impact of European pollution on surface ozone in northern Africa and Middle East
Impact of 2004 boreal fires on US air quality
Differences in transport between GEOS-4-DAS, GEOS-4-Forecast, and FVGCM meteorological fields
ObservationsAircraft missions:• Japan Airlines (JAL) - CO• Measurements of OZone and water vapor by Airbus In-service airCraft
(MOZAIC) – CO, O3
Surface measurements: • NOAA Global Monitoring Division (GMD) - CO• U.S. Environmental Protection Agency (EPA) – O3
• Cooperative Programme for Monitoring and Evaluation of Long-range Transmission of Air Pollutants in Europe (EMEP) – O3
Satellite instruments:• The Aura Microwave Limb Sounder (MLS) – CO• The Ozone Monitoring Instrument (OMI) – NO2
• Atmospheric Infrared Sounder (AIRS) – CO ; see Joanna’s presentation• OMI-MLS – Tropospheric Column O3 ; see Mark Schoeberl’s
presentation ; Jerry Ziemke’s product
GMD CO (1994-98) vs
GMI FVGCM (1994-98 SSTs)
⇒ Simulated CO OK in local summer/fall : photochemistry Dominates
⇒ Simulated CO too low in local winter/spring : transport dominates
A potential partial fix?Jennifer estimates that seasonality may be too low for Chinese emissions.
GMD CO (2004-2005) vs
GMI DASvs
Forecast (12 and 24 hr)
⇒ Models OK in summer/fall but too low in winter/spring
⇒ DAS and Forecasts similar in the higher NH
GMD CO (2004-2005) vs
GMI DASvs
Forecast (12 and 24 hr)
⇒ DAS and Forecast different in the tropics: due to different convections?
0 4 8 12 16 20
lon_dim
0 60 120 180 240 300-90
-70
-50
-30
-10
10
30
50
70
90
0
2
4
6
8
10
12
14
16
18
20
22
24
lon_dim
0 60 120 180 240 300-90
-70
-50
-30
-10
10
30
50
70
90
Precipitation Jul 2004
GMI DASvs
Forecast
[mm/day]
Forecast
DAS
⇒ DAS shows much more (about x2) precipitation than Forecast over ITCZ and SPCZ
⇒ DAS has excessive convection
ITCZ
SPCZ
⇒ DAS is higher than Forecast at the higher SH : Transport from tropics dominates
GMD CO (2004-2005) vs
GMI DASvs
Forecast (12 and 24 hr)
⇒ DAS and Forecast different in the tropics: due to different convections?
EMEP Day-time O3 (2004) vs
GMI (GEOS-4-DAS 2004)
⇒ Simulated surface O3 in Europe is generally OK, especially from winter to spring.
⇒ Simulated O3 is too high in summer – grid resolution
A potential partial fix?Turn back on HO2 uptake onaerosols.
EPA Day-time O3 (2004) vs
GMI (GEOS-4-DAS 2004)
⇒ East coast of US - similar story as in Europe, but worse over-prediction in summer:• grid resolution• isoprene nitrate formation uncertainties (EPA is working on issue).
O3 Difference (DAS Model – EPA) of Monthly Mean (2004)
CO Latitudinal Distribution ComparisonGEOS-4-DAS: 2004 and 2005
JAL : Tokyo ⇔ Sydney
Jan
50
100
150
200
CO
[p
pb
]
JALModel
Apr
50
100
150
200
Jul
30oS 20oS 10oS 0o 10oN 20oN 30oNLatitude
50
100
150
200
CO
[p
pb
]
Oct
30oS 20oS 10oS 0o 10oN 20oN 30oNLatitude
50
100
150
200
Too low Chinese Emissions?
20
50
100
150
CO
[p
pb
]
DJF JAL (94-98)Model (94-98)
MAM
30oS 20oS 10oS 0o 10oN 20oN 30oNLatitude
20
50
100
150
CO
[p
pb
]
JJA
30oS 20oS 10oS 0o 10oN 20oN 30oNLatitude
SON
CO Latitudinal Distribution ComparisonFVGCM : 1994-1998 SSTs
JAL : Tokyo ⇔ Sydney
Climatological BB emissions misses boreal fire?
CO Cruise Plots MOZAIC vs GMI (GEOS-4-DAS)
N = 623R2 = 0.77|bias| = 24%
N = 523R2 = 0.70|bias| = 12%
N = 433R2 = 0.16|bias| = 23%
N = 585R2 = 0.66|bias| = 24%
N = 406R2 = 0.59|bias| = 20%
N = 381R2 = 0.19|bias| = 18%
OSAKA->VIENNA jul 5 2004
0 100 200 300 400 500 600 7000
0
50
100
150
200
CO
[p
pb
]
35
40
45
50
55
60
65
Lat
itu
de
[N]
1000
800
600
400
200
Pre
ssu
re [
hP
a]
MOZAICGMI_DASLatitudePressure
PEKIN->MUNCHEN jul 19 2004
0 100 200 300 400 500 600 7000
0
50
100
150
200
CO
[p
pb
]
40
42
44
46
48
Lat
itu
de
[N]
1000
800
600
400
200
Pre
ssu
re [
hP
a]
WASHINGTON->VIENNA jul 14 2004
0 100 200 300 400 5000
0
50
100
150
200
CO
[p
pb
]
40
45
50
55
Lat
itu
de
[N]
1000
800
600
400
200
Pre
ssu
re [
hP
a]
MUNCHEN->LOS ANGELES jul 15 2004
0 100 200 300 400 500 600 700 8000
0
50
100
150
200
CO
[p
pb
]
35
40
45
50
55
60
65
Lat
itu
de
[N]
1000
800
600
400
200
Pre
ssu
re [
hP
a]
DELHI->VIENNA jul 3 2004
0 100 200 300 400 5000Time [min]
0
50
100
150
200
CO
[p
pb
]
30
35
40
45
Lat
itu
de
[N]
1000
800
600
400
200
Pre
ssu
re [
hP
a]
MUNCHEN->DELHI jul 29 2004
0 100 200 300 400 5000Time [min]
0
50
100
150
200C
O [
pp
b]
30
35
40
45
Lat
itu
de
[N]
1000
800
600
400
200
Pre
ssu
re [
hP
a]
Latit
ude
[N]
Latit
ude
[N]
Latit
ude
[N]
Latit
ude
[N]
Latit
ude
[N]
Latit
ude
[N]
Time [min]Time [min]
jan
0 50 100 150
0
1000
2000
3000
4000
nu
mb
er o
f d
ata
po
ints O
3 threshold = 120 ppbv
below thrabove thr
apr
0 50 100 150
0
500
1000
1500
2000
2500
jul
0 50 100 150
0
500
1000
1500
2000
oct
0 50 100 150
0
500
1000
1500
2000
0 50 100 150
0
1000
2000
3000
4000
nu
mb
er o
f d
ata
po
ints O
3 threshold = 150 ppbv
0 50 100 150
0
500
1000
1500
2000
2500
0 50 100 150
0
500
1000
1500
2000
0 50 100 150
0
500
1000
1500
2000
0 50 100 150CO [ppb]
0
1000
2000
3000
4000
nu
mb
er o
f d
ata
po
ints O
3 threshold = 180 ppbv
0 50 100 150CO [ppb]
0
500
1000
1500
2000
2500
0 50 100 150CO [ppb]
0
500
1000
1500
2000
0 50 100 150CO [ppb]
0
500
1000
1500
2000
CO Histogram Comparison: MOZAIC vs GMIEurope ⇔ North America (2005) Stratosphere
MOZAIC CO (Tropospheric)
MOZAIC CO (Stratospheric)
DAS model CO (Tropospheric)
DAS model CO (Stratospheric)
Troposphere
MLS CO vs GEOS-4-DAS for March 2005
MLS CO (March 2005) vs FVGCM (March 1994-98 Mean)
MLS (March 2005) vs GMI FVGCM (March 1994-8 Mean)
20°N-20°S at 100 hPa
Probability Distribution of CO:
⇒ Distributions are generally OK. (MLS CO is known too high)
Probability distributions (20°S-20°N) of MLS CO (March 2005), DAS (March 2005) and FVGCM (March 1994-98 Mean)
• 15-20 ppbv CO is observed by ER-2 STEP in LS, so FVGCM is reasonable• In LS, DAS is closer to MLS due to biases of both model and observations.• DAS causes more vertical mixing
Probability distributions (20°S-20°N) of MLS CO, DAS, FVGCM and Forecast (12- and 24- hour)
• Both Forecast runs are close to FVGCM in UT.• In LS, Forecasts are between FVGCM and DAS.• There is no difference between the two Forecast runs in UT.
Tropospheric O3: OMI-MLS* vs DAS (March 2005)
0.0 11. 22. 33. 44. 56. 67. 78. 89. 100.
0.0 3.4 6.9 10.3 13.8 17.2 20.7 24.1 27.6 31.
• Model captures the general features well, but overestimates the observations over the tropical Atlantic, northern Africa, near the Himalayas.
*Jerry Ziemke’s product.
[days][ppbv]
Tropospheric O3: OMI-MLS vs DAS (August 2005)
0.0 11. 22. 33. 44. 56. 67. 78. 89. 100.
0.0 3.4 6.9 10.3 13.8 17.2 20.7 24.1 27.6 31.
• Model overestimates the observations from west Asia to entire China, and also over northeastern US.
•Middle East ozone maximum not obvious in observations.
[days][ppbv]
Surface CO comparisons using GMD observations show that our model CO is too low during local winter and spring indicating possible problems with transport and/or seasonality of emissions.
Latitudinal distribution comparisons of CO using JAL data show the GMI DAS and FVGCM both underestimate the CO especially close to emission regions near Tokyo.
For surface O3 comparisons, the GMI model show high biases during summer over western Europe and the northeastern US, partly reflecting uncertainties in our isoprene chemistry and problems of grid resolution.
Based on histogram comparisons of CO using MOZAIC data, our model overestimates stratospheric CO, but underestimates tropospheric CO especially in winter and spring.
Conclusions
The model Comparison to MLS CO is reasonable, if we take into account the known high bias of current MLS CO concentrations.
The model reproduces very well the seasonal cycles and latitudinal distributions of tropospheric O3 columns except high emission regions such as China.
The intermodel comparisons using CO histograms with DAS, FVGCM and Forecast fields show that DAS gives the fastest vertical transport to the LS, and Forecast yields similar results as FVGCM in the UT, but higher in the LS.
Conclusions
OMI NO2 product vs GMI
AIRS CO vs GMI
OMI-MLS O3 vs GMI
On going evaluation
