MAXDOAS formaldehyde slant column measurements during CINDI intercomparison

MAXDOAS formaldehyde slant column measurements during CINDI: intercomparison and analysis improvement G. Pinardi, M. Van Roozendael, N. Abuhassan, C. Adams, A. Cede, K. Clémer, C. Fayt, U. Frieß, M. Gil, J. Herman, C. Hermans, F. Hendrick, H. Irie, A. Merlaud, M. Navarro Comas, E. Peters, A. J. M. Piters, O. Puentedura, A. Richter, A. Schönhardt, R. Shaiganfar, E. Spinei, K. Strong, H. Takashima, M. Vrekoussis, T. Wagner, F. Wittrock, S. Yilmaz

Overview § HCHO measurements during CINDI and DSCD intercomparison § Sensitivity study Analysis improvement § Error budget § Conclusions NDACC/NORS meeting – July 2012

HCHO Measurements Overview § HCHO measurements during CINDI: Instruments BIRA Measurement Period From 13/6 to 22/7 INTA From 7/7 to 24/7 RASASII Bremen From 8/6 to 21/7 Heidelberg From 17/6 to 24/7 Instrument 2 JAMSTEC From 8/6 to 24/7 NASA From 22/6 to 20/7 WSU From 21/6 to 5/7 Toronto From 30/6 to 4/7 Mainz From 21/6 to 10/7 Detector Total Integration Time UV Resolution characteristics (nm) 2048× 512 pixels 60 s ~0. 4 (-30°C) 1024 x 256 pixels 50 s ~0. 39 (-40°C) 2048× 512 pixels 40 s 0. 4 (-35°C) 2048× 256 pixels 60 s 0. 5 (-30°C) uncooled CCD, spectra average for 5 min 0. 7 3648 pixels uncooled CCD, 16 s 0. 6 2048 x 14 pixels 2048 x 512 pixels 45 s 0. 83 (-70°C) 2048 x 512 pixels ~2 min 0. 2 -0. 8 (-72°C) Stabilized CCD 60 s 0. 6 2048 pixels (4°C) All the instruments pointing in the same direction, with common set of elevation angles NDACC/NORS meeting – July 2012

DSCD Comparison § HCHO common DOAS retrieval settings: 336. 5 -359 BIRA-IASB instrument, 30/6/2009, ~14 h 30 UT, 4° elevation angle (Daily ref. ~11 h 40 ) NDACC/NORS meeting – July 2012

HCHO DSCD [10 x 15 molec/cm²] DSCD Comparison § Creation of a reference dataset: BIRA, Bremen, INTA § HCHO DSCD comparisons of every instrument vs reference: scatter plots and histograms of abs. difference (example at 4° elevation) JAMSTEC and NASA: more scatter and larger gaussian distribution linear arrays detectors and not CCD and not cooled TORONTO: slight under-estimation BUT only 5 days of measurement (and instrumental problems) HEIDELBERG and MAINZ: slight underestimation (Mainz, at 20 m on the tower: possibly affecting the comparison ) NDACC/NORS meeting – July 2012

DSCD Comparison § Overview of scatter plot statistical results (each instrument vs reference for all off-axis elevation) 1. 15 within 15% 0. 85 NDACC/NORS meeting – July 2012

Sensitivity study § In order to evaluate the sensitivity of HCHO results to possible changes in the retrieval settings, sensitivity tests are performed on BIRA data of 4 July 2009. § Tested parameters: § Degree of polynomial and Ring effect § The O 4 absorption cross-section § DOAS fitting interval (and minimizing the impact of Br. O) Optimizations that lead to new recommended analysis settings § Different absorption cross-sections § Calibration and slit function Error budget (random and systematic contributions) NDACC/NORS meeting – July 2012

Sensitivity study § While testing the polynomial degree, we found very large differences in the DSCD behavior during the day Why and which one is « the best » ? investigate the consistency of VCD estimates 2 VCD estimations: geometrical approximation and from direct conversion of the zenith-sky observations using appropriate AMFs Only the 5 th order case leads to geophysical consistent results NDACC/NORS meeting – July 2012

Sensitivity study § Why? - Baseline: Chance and Spurr (1997) - Case A: Wagner et al. (2009) - Case B: from SCIATRAN RTM in a Rayleigh atmosphere - Case C: Principal Component Analysis of a range of SCIATRAN calculations in an ozone containing atm. (Vountas et al. (1998)) § Large sensitivity of HCHO Δd. DSCD to changes in the Ring crosssection, especially for 3 rd order polynomial. § HCHO DSCD changes (d. DSCDs) are linearly related to changes in the Ring fit coefficients. NDACC/NORS meeting – July 2012

Sensitivity study § Test the stability of the retrievals: Various combinations of polynomials (degree 3, 4 and 5) and different Ring cross-sections Root-mean-square of HCHO VCD differences obtained using two alternative methods for the calculation of VCD. Optimal stability (corresponding to smallest HCHO VCD differences), for the different sets of Ring xs, is obtained for cases using a polynomial of degree 5. NDACC/NORS meeting – July 2012

Sensitivity study § O 4 cross-section: Hermans et al. (2003) and Greenblatt et al. (1990) § Greenblatt xs is needed to retrieve coherent Br. O! (in all tested windows) § Misfit due to O 4 correlate both to HCHO and Br. O NDACC/NORS meeting – July 2012

Sensitivity study § O 4 cross-section Change in retrieved HCHO and Br. O DSCDs when exchanging the Hermans et al. (2003) O 4 absorption cross-section by the Greenblatt et al. (1990) data set, expressed as a function of the O 4 DSCD values. a misfit to the O 4 absorption (larger in this case using the Hermans et al. data set) activates a correlation between HCHO and Br. O DSCDs NDACC/NORS meeting – July 2012

Sensitivity study § DOAS fitting interval: search for an optimized wavelength window by minimizing the correlations between the xs of the different absorbers smaller correlations for fitting intervals starting at short wavelengths; the 333 -358 nm wavelength range presents a local minimum Start wavelength Correlation matrix of the absorption xs in 336. 5 -359 nm Overall correlation (RMS of nondiagonal elements of the matrix)

Sensitivity study § DOAS fitting interval: test of theoretical results on real data large instabilities in 333 -358 nm region with respect to the Ring effect interference to be used with care! NDACC/NORS meeting – July 2012

Sensitivity study § Others tests (xs uncertainties, slit fct and calibration): are discussed as part of the systematic uncertainties § New recommended settings: 1. 5 th degree polynomial, in order to minimize the interference with the Ring effect 2. O 4 Greenblatt cross-section, that leads to more consistent Br. O DSCD 3. Fitting window: a new candidate (333 -358 nm) is highlighted from theoretical study, but leads to instabilities related to the Ring effect when applied to real data use with care! NDACC/NORS meeting – July 2012

Error budget § DSCD random error comparison: When normalizing wrt the integration time, 2 group of instruments show up: § Scientific instruments with large cooled detectors § mini. DOAS like devices

Error budget § Summary assessment of the error budget on optimized HCHO d. DSCD (systematic and random contributions) Typical d. DSCD (4°elev) = 3. 8 x 1016 molec/cm² § Scientific instruments: dominated by the systematic contribution (around 20% and slightly increasing with SZA) § mini. DOAS like instruments: both contributions are similar NDACC/NORS meeting – July 2012

Conclusion § Very good comparisons of HCHO DSCD retrieved by 9 groups using harmonized retrieval settings. The scatter plot slopes are close to one, within ~15%. § Sensitivity study has been performed leading to new recommended DOAS settings (a 5 th order polynomial and the O 4 Greenblatt cross-section are needed in order to reduce interference and misfits with Ring and Br. O). § An error budget has been obtained for HCHO DSCD, with total errors around 20 -30% (8 -15 x 1015 molec/cm²). Larger systematic contributions: Ring effect and HCHO and O 3 xs uncertainties. § The paper is ready to be sent to co-authors for last checks and then submitted! NDACC/NORS meeting – July 2012