Concurrent measurements of tropospheric NO2 from OMI and SCIAMACHY Folkert Boersma, Daniel Jacob, Henk Eskes, Rob Pinder, Jun Wang, and Ronald van der A Folkert Boersma

Motivation Examine consistency between two satellite data sets Example: GOME and SCIAMACHY Jan-May 2003: Overlap between GOME and SCIA Average difference: 0.030.16 1015 GOME SCIA van der A et al. (2006)

GOME and SCIAMACHY NO2 Same retrieval: Slant columns from BIRA Stratospheric columns through data assimilation in TM4 Same AMF look-up table and inputs

OMI and SCIAMACHY NO2 GOME and SCIAMACHY overpass at similar times (10:30 and 10:00) OMI and SCIAMACHY overpass at different times (13:30 and 10:00) Consistent retrieval algorithms NO2 has a relatively short lifetime (4-24 hours) NOx has diurnal cycle in source and loss terms Can we observe fast photochemistry and changes in emissions from space? QUESTION:

Common algorithm SCIAMACHY and OMI NO2 with same retrieval Similarities Non-linear least squares fitting for slant columns Stratospheric columns through data assimilation in TM4 Same AMF approach and inputs (albedo, profile shape, temp.) Main differences Slant column fitting details Cloud inputs

Similar NO2 absorption cross sections Relative amplitude SCIAMACHY 2.4% stronger than for OMI SCIA OMI Absorption cross sections are bias-corrected here (mean bias = 2.4%; SCIA higher) Relative difference determined by comparing top-valley distances for the six peaks (left valley-top): 2.6%; right valley top: 2.3% Relative difference determined as standard deviation SCIA Xsec vs standard deviation OMI Xsec = 1.5%

Similar cloud retrievals Clouds -SCIAMACHY: FRESCO+ (760 nm) -OMI: O2-O2 (470 nm) Average cloud fraction agrees within 1.5% Bias between cloud pressures of ~60 hPa SCIAMACHY underestimates the reflectance by ~13% at 442 nm (10% of this is in the irradiance). Since then, improved radiometric calibration has likely solved much of the problem but BIRA fits are still being done with the Indian Ocean spectrum.

Collocated, cloud-free (rcl<50%) measurements at 0.5°0.5°

Differences larger than expected uncertainties

Differences over source regions What is the cause of the differences? - slant columns  information in the spectra - AMF  information in the physics of retrieval Compare slant columns and AMFs correcting for different viewing geometries: - slant columns: Ns/Mg - AMFs: Mtr/Mg

Ratio SCIAMACHY:OMI over source regions

What causes AMF differences? AMF depends on: albedo cloud fraction cloud height a priori profile shape Cloud fractions are the same within 0.05 (bias <0.015) Differences too small to explain AMF differences of up to 20%

What causes AMF differences? Differences in NO2 profile shape Better mixed boundary layer at 13:30 hrs NO2 profile peaks at higher altitudes Sensitivity study with TM gives on average up to 15% higher AMFs at 13:30 compared to 10:30 How frequent is the BL updated in TM? Could this lead to too high AMFs for SCIA (and too low columns).

What causes AMF differences? Differences in cloud pressures (+60 hPa) Sensitivity study with 60 hPa higher cloud pressures gives ~10% higher OMI AMFs

Diurnal variation of NO2 columns NNO2: NO2 column (t): NO2:NOx ratio E(t): NOx emissions k(t) : rate constant NNOx: NOx column  E rate constant k(t) applies to conversion of NOx to HNO3 NO2+OH+M  HNO3+M k

GEOS-Chem Relative decrease in NO2 column from 10:00 to 13:30 pm Observed GEOS-Chem US: -16% -28% EU: -6% -13% China: -26% -22%

2003 2005 2001 2004 2002 Biomass burning mainly in afternoon Jun Wang Relative increase in NO2 column from 10am to 1:30 pm Observed GEOS-Chem Africa: +48% +16% Indon.: +60% +10% Brazil: +54% +13%

Extra SLIDES

tm4no2a-omi, v0.9.3.4

tm4no2a-omi, v0.9.3.4

tm4no2a-omi, v0.9.3.4

Southeastern United States August 2006 tm4no2a_omi, version 0.9 n = 27 (including days with ‘zoom in’)

Geometrically normalized

Other retrieval differences Reference spectra SCIAMACHY: Earth radiance Indian Ocean OMI: solar irradiance spectrum Stripes SCIAMACHY: no stripes OMI: stripe correction (tm4no2a-omi, v0.9) SCIAMACHY underestimates the reflectance by ~13% at 442 nm (10% of this is in the irradiance). Since then, improved radiometric calibration has likely solved much of the problem but BIRA fits are still being done with the Indian Ocean spectrum.

Statistical analysis slant column differences

Average from weighting with cos2()-1 No weighting Average from weighting with cos2()-1 Absorption cross sections are bias-corrected here (mean bias = 2.1%; SCIA higher) Relative difference determined by comparing top-valley distances for the six peaks (left valley-top): 2.6%; right valley top: 2.3% Relative difference determined as standard deviation SCIA Xsec vs standard deviation OMI Xsec = 1.5%

Sampled when both SCIA and OMI took measurements

Nv,omi>1.0 1015 r = 0.994 (n=1.9x105) weighted:unweighted = 1.014

OLD New lv0-1 New – no stripe correction Bas Mijling

Retrieval method What is the air mass factor (AMF)? represents the relative length of the light path, converts the slant into a vertical column M = Ns/Nv Computed with radiative transfer model AMF depends on a priori knowledge of atmospheric state Cloud fraction Cloud height Vertical distribution NO2 Surface albedo

Saturday Sunday GOA is a project in which a.o. IUP and KNMI participate

Slant column of Air Mass Factor effect? Both contribute to higher SCIAMACHY NO2 columns

Does China have less of a rush hour?

Outlook – “Veni proposal” Key issue: relation between NOx emissions and observed NO2 columns NO2 is chemically produced, transported, and removed: need a model Proposal to develop data assimilation system with TM5 (0.25° x 0.25°) Data assimilation system Ensemble of NOx and VOC emissions Observations NO2 & HCHO Ensemble Kalman Filter Improved NOx and VOC emissions + uncertainties TM5 is global CTM with nested grid option to zoom (0.25° x 0.25°) Apply system on U.S., Europe, and China Also incorporate EPA, EMEP surface observations

EPA emission reports Trend 1995-2005 Ronald van der A (KNMI) Studies by Richter et al. (2005) and van der A et al. (2006) also show negative trend in NO2 columns OMI This slide could be cut Mobile Other Industry Power plants

Outlook – “Veni” Innovations Hi-res satellite observations applied in air pollution in a nested-grid CTM High regional resolution, while… …accounting for long-range transport (China!) Data ass. with EnKF used for the first time for estimating NOx and VOC emissions Apply anywhere in the world xb H(xb), P y, R xa = xb + K(y - H(xb)) TM5 Emiss. OMI