Direct Tropospheric Ozone Retrieval from GOME Xiong Liu

Direct Tropospheric Ozone Retrieval from GOME Xiong Liu Harvard-Smithsonian Center for Astrophysics xliu@cfa. harvard. edu December 20, 2004 1

Outline n Introduction n Algorithm description n Retrieval characterization n Intercomparison with Ozonesonde, TOMS, and Dobson n Global distribution of tropospheric ozone and comparison with GEOS-CHEM model results n Summary and future work 2

Introduction n Current tropospheric ozone retrievals are mainly based on the residual approaches: limited to certain latitude ranges and to monthly level n GOME: first nadir-viewing satellite instrument that allows direct tropospheric ozone retrieval from the space. n Several groups [Munro et al. , 1998; Hoogen et al. , 1999; Hasekamp et al. , 2001; van der A et al, 2002; Muller et al. , 2003; Liu et al. , 2004] have developed ozone profile retrieval algorithms from GOME: each of them demonstrates that limited tropospheric ozone information can be derived. n However, tropospheric ozone retrieval remains very challenging from GOME: u Require accurate and consistent calibrations. u Need to fit the Huggins bands to high precision. u Tropospheric ozone is only ~10% of total column ozone. 3

Algorithm Description n Inversion technique: Optimal Estimation n Measurements: 289 -307 nm, 326 -338 nm; Spatial resolution: 960× 80 km 2 n Perform detailed wavelength and radiometric calibrations: ØDerive variable slit widths and shifts between radiances/irradiances ØFit shifts between trace gas absorption cross-sections and radiances ØCo-add adjacent pixels from 289 -307 nm to reduce noise ØImprove polarization correction using GOMECAL (www. knmi. nl/gome_fd/) ØPerform undersampling correction with a high-resolution solar reference ØFit degradation for 289 -307 nm on line in the retrieval n Use LIDORT to simulate radiances and weighting functions n Improve forward model simulation: ØOn-line correction of Ring filling in of the solar and telluric absorption feature with first-order single scattering RRS model [Sioris and Evans, 2002] ØLook-up table correction of polarization errors [van Oss, personal comm. ] ØMonthly-mean SAGE stratospheric aerosols [Bauman et al. , 2003] ØGEOS-CHEM tropospheric aerosols [Martin et al. , 2002] 4

Variable Slit Widths and Shifts 5

Algorithm Description n Improve forward model simulation (continue): ØBrion’s ozone absorption cross-sections [Brion et al. , 1993] ØDaily ECMWF temperature profiles (www. ecmwf. int) ØDaily NCEP/NCAR surface pressure (www. cdc. noaa. gov) ØCloud-top height from GOMECAT [Kurosu et al. , 1999] ØCloud fraction derived at 370. 2 nm with albedo database [Kolemeijer et ØWavelength dependent albedo (2 -order polynomial) from 326 -338 nm al. , 2003] A priori: latitude and monthly dependent TOMS V 8 climatology (a priori and its variance) [Mc. Peters et al. , 2003, AGU] n n Retrieval Grid: 11 layers, almost the same as the Umkehr grid Ø Ø n Bottom 2 -3 layers are modified by tropopause/surface pressure Tropospheric column ozone is directly retrieved State Vector: 47 parameters 11 O 3 + 4 albedo (1 for ch 1 a & 3 for ch 2 b) + 4 Ring (1 for ch 1 a & 3 for ch 2 b) + 8 O 3 shift + 8 rad. /irrad. shift + 3 degradation correction (ch 1 a only) + 2 undersampling + 2 NO 2 + 2 Br. O + 2 SO 2 + 1 internal scattering Ø n Fitting residual: 0. 40% for band 1 a, 0. 17% for band 2 b, 0. 3% for both n Speed : ~17 hours on a 2 GHz processor for one day, could be operational 6

Retrieval Characterization n Averaging Kernel: characterize the retrieval n DFS: diagonal elements of averaging kernels n A priori influence: 7

Examples of Averaging Kernels 8

9

A Priori Influence (06/7 -9/1997) TOMS V 8 A Priori GEOS-CHEM A Priori Retrieval with TOMS V 8 A Priori Retrieval with GEOS-CHEM A Priori 10

Retrieval Errors 11

Validation and Intercomparison GOME data are collocated at 25 ozonesonde stations during 96 -99. n Validate retrievals against TOMS V 8, Dobson/Brewer total ozone, and ozonesonde. n Ozonesonde data mostly from WOUDC, and some from CMDL, SHADOZ, and NDSC. n Collocation criteria: n Ø Within ~8 hours, 1. 5° latitude and ~500 km in longitude Ø Average all TOMS points within GOME footprint Number of comparisons: 4429, 952, and 1937 with TOMS, Dobson, and ozonesonde, respectively. n http: //www. woudc. org; http: //croc. gsfc. nasa. giv/shadoz http: //ndsc. ncep. noaa. gov; http: //toms. gsfc. nasa. gov/ http: //www. cmdl. noaa. gov/infodata/ftpdata. html 12

Total Column Ozone Comparison GOME-TOMS: within retrieval uncertainties and saptiotemporal variability. n ØBiases: <3 DU except 38 DU at a few highlatitude stations Ø 1 : 2 -4 DU in the tropics, 4 -11 DU at higher latitudes. A Priori Retrieval n. GOME-Dobson: Dobson TOMS within retrieval uncertainties and ozone variability. ØBiases: <5 DU, and <8 DU at two high-latitude stations Ø 1 : 3 -6 DU in the tropics, 6 -19 DU at higher latitudes. 13

Stratospheric Column Ozone Comparison Column ozone between tropopause to~30 -35 km 1%-KI buffered 2%-KI unbuffered A Priori Retrieval Ozonesonde n. GOME-Ozonesonde: ØSystematic differences exists at Carbon Iodine, CMDL, SHADOZ stations ØBias: <3 DU (2%), except at Ny Ålesund and Neumayer (-3. 3% and 4. 5%) Ø 1 : 4 -9 DU (4 -6%) in the tropics and 10 -22 DU (5 -10%) at higher latitudes. 14

Tropospheric Column Ozone Comparison nn. GOME-SONDE withinretrieval uncertainties. ØØBiases: <4 <4 DU DU(15%) except– 5. 5, 4. 4, 5. 6 DU DU (16 -33%)atat. NyÅlesund, Naha, Tahiti Naha, NyÅlesund, ØTahiti 1 : 3 -7 Ø 1 A Priori Retrieval Ozonesonde DU (13 -28%) : 3 -7 DU (13 -28%) 15

Profile: Hohenpeißenberg (48 N, 11 E), 1996 -1999 The degradation is well handled GOME retrievals agree well with ozonesonde n n ØBiases: Ø 1 A Priori Retrieval Ozonesonde <2 DU (10%) : <10 DU (25%) 16

Profile: America Samoa (14 S, 171 W), 1996 -1997 Positive bias in the middle, negative bias at two ends, probably due to some systematic bias in radiance spectra and the wavelength dependent correction is not perfect. n ØBiases: < 4 DU (40%) Ø 1 : <4 DU (30%) Bias in tropospheric/stratospheric column ozone is reduced due to 17 canceling errors. n A Priori Retrieval Ozonesonde

Examples of Daily Global Tropospheric Ozone Low tropospheric ozone in tropical Pacific Bands of high ozone at mid-latitudes High ozone over biomass burning South Atlantic Paradox High ozone at high-latitudes during late winter and early spring 18

Monthly Mean Tropospheric Ozone (09/96 -11/97) 19

GOME vs. GEOS-CHEM Tropospheric Ozone SON, 96 R=0. 67 1. 8± 6. 8 DU DJF, 96 -97 R=0. 83 0. 0± 5. 3 DU MAM, 97 R=0. 82 2. 2 ± 4. 5 DU JJA, 97 R=0. 64 2. 5 ± 5. 7 DU 20

Summary n Ozone profiles and tropospheric column ozone are derived from GOME using the optimal estimation approach after detailed treatments of wavelength and radiometric calibrations and improvement of forward model inputs. n Retrieved total ozone compares very well with TOMS and Dobson/Brewer total ozone. n The profiles, stratospheric ozone, and tropospheric ozone compare well with ozonesonde observations except some stratospheric bias at Carbon Iodine stations, CMDL, and SHADOZ stations. n Global distribution of tropospheric ozone is presented. It clearly shows the signals due to biomass burning, air pollution, stratospherictroposphere exchange, transport and convection. n The overall structures of retrieved tropospheric ozone are similar to those of GEOS-CHEM, but significant differences exist. 21

Future Work n Complete tropospheric ozone retrieval for the 8 -year GOME data record and apply the algorithm to SCIMACHY and OMI data n With the aid of GEOS-CHEM, other observations, or model fields, understand the GOME/GEOS-CHEM similarities and differences, and investigate global/regional distribution of tropospheric ozone. n Tropospheric ozone radiative forcing n Tropospheric/stratospheric ozone variability 22

GOME vs. GEOS-CHEM Tropospheric Ozone 23

GOME vs. GEOS-CHEM Tropospheric Ozone 24

GOME vs. GEOS-CHEM Tropospheric Ozone 25

GOME vs. GEOS-CHEM Tropospheric Ozone 26

GOME vs. GEOS-CHEM Tropospheric Ozone 27