Intercomparison methods for satellite sensors: application to tropospheric ozone and CO measurements from Aura

Daniel J. Jacob, Lin Zhang, Monika Kopacz

and funding from NASA ACMAP

with Xiong Liu (NASA/GSFC), Jennifer A. Logan (Harvard), Kelly V. Chance (SAO), and the TES, OMI, AIRS, MOPITT, and SCIAMACHY Science Teams

Zhang, L., et al., Intercomparison methods for satellite measurements of atmospheric composition: application to tropospheric ozone from TES and OMI to be submitted

Kopacz, M., et al., Global estimates of CO sources with high resolution by adjoint inversion of multiple satellite datasets (MOPITT, AIRS, SCIAMACHY, TES) submitted to Atmos. Chem. Phys.

Tropospheric ozone measurements from TES and OMI

TES (V003) • Thermal IR (3.3-15.4 mm)• Retrieve log mixing ratio at 67 levels• Along-track 5x8 km2 pixels every 1.6o latitude

OMI (Xiong Liu, GSFC)• UV (0.27-0.5 mm)• Retrieve partial columns in 24 layers• 13x24 km2 pixels (nadir), global daily coverage

Convert TES averaging kernels to OMI grid and partial columns

Ozonesonde validation of TES and OMIUse global ensemble of coincident ozonesonde profles for 2005-2007:

528 for TES, 2568 for OMI

Global mean biases at 500 hPa: +4 ± 7 ppbv for TES, +3 ± 6 ppbv for OMI…but data are very sparse : validation space is inadequately sampled

500 hPa ozone from TES, OMI, and the GEOS-Chem CTMYear 2006 data reprocessed with fixed a priori; GEOS-Chem smoothed by the averaging kernels of each instrument

GEOS-Chem simulation with TES vs. OMI averaging kernels shows thatdifferences between the two instruments partly reflect differences in sensitivity;Can we use the residual as measure of the bias between the two instruments?

Intercomparing satellite instruments

TES OMI

O3sondes

1. In situ method:true validation but sparse

3. Averaging kernel smoothing method (Rodgers and Connor, 2003):smooth retrieval of instrument 1 with the averaging kernels of instrument 2

2. CTM method:Compare instruments independently to CTM

GEOS-Chem

What does each method actually intercompare?

TES TES TES a TES TES TES

OMI OMI OMI a OMI OMI OMI

TES OMI TES OMI TES OMI a

ˆ

ˆ

ˆ ˆDifference = (

x = A x + (I - A )x +b +G ε

x = A x+ (I - A )x +b +G ε

Δ x x = b -b (A - A ) x - x )

1, In situ method: directly measure x (ozonesondes)

TES sonde_TES OMI sonde_OMI TES OMIˆ ˆ ˆ ˆ x x x x b b2. CTM method: reference retrievals to local CTM values

TES CTM_TES OMI CTM_OMI

TES OMI TES OMI CTM

ˆ ˆ ˆ ˆ

x x x x

b b A A x x

Start from the retrievals of ozone concentrations x:

3. Averaging kernel smoothing method: process TES retrieval through OMI avker

TES_OMI OMI OMI TES OMI OMI TESˆ ˆ ax x A b b A A I x x

noise!

reduced noise

Intercomparison by the CTM and avker smoothing methodsreferenced to the in situ method

Differences D at 500 and 800 hPa for 180 sonde/TES/OMI coincidences in 2006

• The CTM method closely approximates the in situ method• The avker smoothing method dampens differences and has large noise

Averaging kernelsmoothing method

CTM method

Global intercomparison of TES and OMI by the CTM method

Seasonal mean TES-OMI differences (D) at 500 hPa for year 2006

Differences generally < 10 ppbv except for northern mid-latitudes in summer,some tropical continental regions

GEOS-Chem evaluation using TES and OMIOzone at 500 hPa; TES and OMI have been corrected for their global mean biases

Black areas are where TES and OMI are inconsistent (D > 10 ppbv)

We find that GEOS-Chem is too low in tropics, too high at southern mid-latitudes

Application of the GEOS-Chem model adjoint to optimize CO sources using multi-sensor data

Annual mean CO columnMay 2004- April2005

observed CO

Earth surface

4-D Var sensitivity of observed concentrationsto emissions upwind

sensi

tivity

time

transport

chemistry

transport

chemistry

emission

AIRS

MOPITT

TES

SCIAMACHY (Bremen)

CONSISTENCY BETWEEN SATELLITE INSTRUMENTS FOR CO

Results show good consistency between instruments and with in situ “truth”

Global (2ox2.5o) correlation of daily data with GEOS-Chem, May 2004 –April 2005;GEOS-Chem fields processed by averaging kernels of each instrument

in situ

1. Use AIRS, MOPITT, SCIAMACHY-Bremen in adjoint inversion;

Best prior estimate from current inventories

Annual CO emissions2004-2005

Annual correction factors from adjoint inversion

General underestimate of emissions, but with large seasonal variation

2. Use TES, NOAA/GMD, MOZAIC for evaluation of inversion results

EMEP

Streets

GFED2EDGAR

NEI99x0.4

INVERSE MODEL RESULTS

• Best prior estimate (EPA NEI99 reduced by 60% on basis of ICARTT) is OK in summer when ICARTT was flown but not in other seasons

• Underestimate of emissions from cold vehicle starts in winter?

MOZAIC data observeda prioria posteriori

CORRECTION FACTOR IN US:SEASONAL VARIATION

GMDdata

Cross-instrument bias revealed by common reference to CTM

Ozone retrievals at 500 hPa for year 2006x̂

TESx̂ˆ ˆTES OMIx - x ˆ ˆCTM_TES CTM_OMIx - x Residual