Improvement of the retrieval of aerosol optical properties over oceans using SEVIRI A. Vermeulen 1, C. Moulin 2, F. Thieuleux 3, I. Chiapello 3, J. Descloitres 1, F. Ducos 3, J-M Nicolas 1, F.-M. Bréon 2 1. ICARE/CGTD, CNRS-USTL, Villeneuve d Ascq 2. ISPL/LSCE, CEA-CNRS, Gif-sur-Yvette 3. LOA, CNRS-USTL, Villeneuve d Ascq FRANCE Acknowledgements: NASA/AERONET and LOA/PHOTONS teams Help with POLDER/Parasol products: JL Deuzé
Introduction Tropospheric aerosols represent a major uncertainty in climate change prediction Satellite provide adequate observations of global aerosol characteristics and their temporal changes The SEVIRI instrument (onboard Meteosat-8 and -9): 3 Aerosol bands (0.635, 0.81, 1.6 mic), a spatial resolution of 3 km at nadir, and a high temporal resolution with an image every 15 min. High frequency of observations is a major advantage for monitoring rapid aerosol events at local scale for global climate studies because it increases the number of cloud-free and sun-glint free retrievals
Aerosol from MSG/SEVIRI data Algorithm initially developed by Thieuleux et al. (2005, Annales Geophysicae) and tested on one month of data Retrieved aerosol properties currently available: AOT at 550 nm and Angström exponent (AE, indicator of the particle size) for each image and at full resolution, and the daily average of AOT and AE RGB image, 11 March 2006 12:00 UTC Daily AOT 550nm Example of the disk of the Earth as observed by Meteosat-8 on 11 Mar 2006, 12:00, showing a Saharan dust outflow over the Atlantic Daily AE
Algorithm implemented at the ICARE Data and Services Center to process Meteosat-8 and Meteosat-9 data in near real-time: http://www.icare.univ-lille1.fr Retrieved aerosol properties currently available: AOT at 550 nm and Angström exponent (indicator of the particle size) for each image and at full resolution, and the daily average of AOT and AE
Aerosol from MSG/SEVIRI data: Description of the algorithm Screening stage to identify regions contaminated by cloud or sun glint, or with invalid geometric conditions Retrieval of aerosol AOT and AE: 0.635 and 0.810 µm Based on a LUT of modeled TOA reflectances, using 15 simple aerosol models adapted from Shettle and Fenn (1979) The LUT is used to find the aerosol model that best fits the measured TOA reflectances at both wavelengths
Comparison with ground-based sun photometer measurements from AERONET Systematic comparison with all available data in 2006 in coastal environment AERONET: coastal sites with observations in 2006 ~20 coastal sites in the Atlantic (7), around the Mediterranean perimeter (11), and in the North Sea (4) Sites are influenced by: dust pollution biomass burning maritime aerosols or a combination of them
Comparison at FORTH_CRETE Jan-June 2006 AOT 550nm AERONET (red) SEVIRI (green) Angström Exponent Julian Day (Jan-June)
FORTH_CRETE Dakar
Calibration issue of the solar bands We use the calibration coefficients in L1.5 image data to convert DC into normalized reflectance Calibration in progress for 0.6, 0.8 and 1.6 mic (JM Nicolas) Current results shows an underestimation of 5% of the VIS 0.6 and VIS 0.8 mic, and an overestimation of 6% of the NIR 1.6 (MSG-1) Results to be confirmed
Comparison SEVIRI/AERONET/Parasol - FORTH_CRETE
Comparison SEVIRI/AERONET/Parasol - FORTH_CRETE
Use of Parasol products to improve the SEVIRI aerosol database POLDER/Parasol: in the A-Train constellation, operational since Dec. 2004 A single measurement per day but multidirectional and polarized observations very sensitive to the aerosol properties Parasol output products AOT fine mode AOT coarse modes (spherical and non-spherical) Fine mode: 4 effective radius, 3(+1) refractive indices (prescribed values) Coarse mode: 1 effective radius, 3(+1) refractive indices (prescribed values), and relative contribution of non spherical aerosol AOT (5 prescribed values) For model classification: we added 5 relative contributions of the fine mode AOT fraction at 0.865mic.: 4x4x1x4x5x5=1600 different models Transfer the accurate aerosol characterization obtained from Parasol to the MSG scale Approach: We analyzed the Parasol products over a whole year (2006), around an area of +/- 1 degree around each AERONET validation site to identify the most frequently retrieved aerosol models. We selected retrievals in ideal conditions, i.e. cases with large aerosol loads, and optimal viewing conditions (large and wide range of scattering angles) as defined in Parasol specifications documentation
Selection of aerosol models from Parasol products 76000 Parasol retrievals in 2006
Selection of aerosol models from Parasol products 76000 Parasol retrievals in 2006
Selection of aerosol models from Parasol products 76000 Parasol retrievals in 2006
Conclusion and Future Work Calibration study in progress: Aerosol retrieval based on spectral characteristics, so the calibration is of paramount importance Reprocessing of SEVIRI data with new calibration coefficients Comparison with AERONET data using all available data during 2006 Finish selection of Parasol aerosol models: select ~25 aerosol models and implement into the SEVIRI algorithm Use of the 1.6mic band New algorithm will be implemented at ICARE Data and Services Center in 2008. The SEVIRI will complement the A-Train products