The main improvements in the GOME Data Processor

The main improvements in the GOME Data Processor Version 5 algorithm and future planned updates for the Ozone_cci project M. Van Roozendael 1 C. Lerot 1, J-C. Lambert 1, J. van Gent 1, D. Loyola 2, W. Zimmer 2, R. Spurr 3, D. Balis 4, M. Koukouli 4 1 Belgian Institute for Space Aeronomy (BIRA-IASB) 2 German Aerospace Center (DLR-IMF) 3 RT-Solutions, Inc. 4 LAP, Aristotle University of Thessaloniki (AUTH) ACC Meeting, Washington, 21 -22 June 2011

Total ozone algorithms TOMS • 317. 5 and 331. 2 nm • Version 8 in 2004 • Nimbus 7: 1978 to 1993 • Earth Probe: 1996 to 2005 • Aura OMI: 2004 GDP 4 • DOAS in 325 to 335 nm • AMF at 325. 5 nm • Version 4 in 2004 • GOME/SCIAMACHY/GOME-2 GODFIT/GDP 5 • 325 to 335 nm • Direct-fitting scheme (no AMF) • GOME: 1995 to now, to be released this summer • SCIAMACHY, GOME-2 BIRA scientific prototype ACC Meeting, Washington, 21 -22 June 2011 CCI baseline

GODFIT-GDP 5 algorithm • Accurate wavelength calibration using Fraunhofer lines • Direct fit of measured reflectivity using on-the-fly MS radiative transport calculations (LIDORT v 3. 3) • TOMS v 8 ozone profile column-resolved climatology • Temperature shift adjustment algorithm • O 3 cross-sections from Malicet, Daumont & Brion • Parameterised Ring correction • Albedo closure • Cloud correction: IPA + OCRA/ROCINN cloud algorithm (DLR) • GDP 5 is the current ESA GOME baseline, to be released in late summer 2011 – For Ozone_cci the same algorithm is to be applied to GOME, SCIAMACHY and GOME-2 ACC Meeting, Washington, 21 -22 June 2011 Van Roozendael et al. , to be submitted to JGR

DOAS against direct-fitting DOAS residuals ACC Meeting, Washington, 21 -22 June 2011 Direct-fitting residuals

Temperature effects ACC Meeting, Washington, 21 -22 June 2011 Hendrick et al. , ACP, 2011

Temperature shift algorithm in GDP 5 (z) is a shape function • The T-shift parameter is part of the state-vector of the retrieved parameters. Its impact on air density and ozone absorption cross-sections is treated rigorously in the algorithm ACC Meeting, Washington, 21 -22 June 2011

Comparison with DOAS Total O 3 Diff. (%) • Effective temperatures retrieved from DOAS have a bias at large SZA (due to ozone misfit effect) ACC Meeting, Washington, 21 -22 June 2011 ∆T=10 K ∆O 3=3% Eff. T° Diff

Averaging Kernels AK = Gy. Kx = xc/ x Kx= y/ x is the weighting function Gy= xc/ y is the contribution function Albedo= 0. 05 ACC Meeting, Washington, 21 -22 June 2011 Albedo=0. 8 Provided for each GOME pixel in GDP 5

Validation GDP 4. 1 GDP 5 • Reduced seasonality w. r. t. to Brewer reference data ACC Meeting, Washington, 21 -22 June 2011

Validation ACC Meeting, Washington, 21 -22 June 2011

Ongoing and planned improvements for the ESA Ozone_cci project • Improved Ring correction • Improved ROCINN algorithm for effective cloud top determination using CAL scheme • Acceleration of the algorithm low-stream interpolation and fast LUT scheme • Transfer to SCIAMACHY and GOME-2 • Transfer to OMI (option in CCI) to allow better link between ESA and NASA algorithms • Improved error characterization ACC Meeting, Washington, 21 -22 June 2011

Improved Ring correction • Current empirical correction: FF( ) = Ring filling factor • New semi-empirical correction: with Semi-empirical formulation of the Ring filling factors, ajusted on simulations using LIDORT_RRS. 0 represents the probability of Raman scattering. ACC Meeting, Washington, 21 -22 June 2011

Semi-empirical filling factors Total O 3: 300 DU SZA= 40° VZA= 30° RAA= 10° Albedo = 0. 06 This formulation includes: • Source and loss terms due to inelastic scattering • Ozone absorption in the incoming and outgoing path • 0, Aout and Ain( ) are adjusted to LIDORT_RRS simulated FF values, and stored in LUTs for use in the ozone algorithm ACC Meeting, Washington, 21 -22 June 2011 Total O 3: 500 DU SZA= 85° VZA= 50° RAA= 10° Albedo = 0. 8

Impact on ozone • Further reduction of the seasonality w. r. t. Brewer reference data (to be further quantified with more data) GDP 5 (old Ring correction) Improved Ring correction ACC Meeting, Washington, 21 -22 June 2011

0 Cloud top height retrieval ? • The RRS probability is a strong function of the altitude of the reflecting surface in principle the Ring effect contains information on cloud top height (cf. OMI) • Does not work easily with polarised instruments, since FFs also depend on the polarisation sensitivity of the instrument (complex level-1 issue) ACC Meeting, Washington, 21 -22 June 2011

OCRA/ROCINN – Overview • • Cloud properties retrieved from GOME measurements: • PMD: cloud fraction • Oxygen A-band: cloud-top height and cloud-top albedo (optical thickness) Loyola D. , Thomas W. , Spurr, R. , B. Mayer, Global patterns in daytime cloud properties derived from GOME backscatter UV-VIS measurements, ACC IJRS Meeting, Washington, 21 -22 June 2011 2010.

Improved cloud correction GOME vs Brewer • Current GDP 5 version overestimates ozone over tropical thick high clouds • Known problem, due to intracloud O 3 absorption new cloud algorithm (CAL) under development at DLR CRB ACC Meeting, Washington, 21 -22 June 2011 CAL

GDP 5 transfer to SCIAMACHY and GOME-2 GODFIT-DOAS GOME 2 - GOME SCIA - GOME ACC Meeting, Washington, 21 -22 June 2011 • Differences between DOAS and GODFIT algorithms are consistent for all sensors • But… systematic differences between sensors remain whatever the algorithm likely related to level-1 issues • Need for vicarious calibrations?

Summary • GOME Data Processor version 5 (GDP 5) implements a direct fitting scheme. New ESA operational baseline for GOME (release late summer 2011). • Validation results show better agreement with reference Brewer network, in comparison to GDP 4 • New developments ongoing within ESA Ozone_cci project transfer of GDP 5 to other sensors, improved Ring effect treatment, and cloud correction. Acceleration of the algorithm to allow application to high resolution sensors (GOME 2, OMI, TROPOMI, …) ACC Meeting, Washington, 21 -22 June 2011