Development of the SEVIRI Aerosol Retrieval Algorithm (SARA) Aerosol retrievals using MSG-SEVIRI images Y.S. Bennouna and G. de Leeuw
Dec 21, 2015
Development of the SEVIRI Aerosol Retrieval Algorithm (SARA)
Aerosol retrievals usingMSG-SEVIRI images
Y.S. Bennouna and G. de Leeuw
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images2
Outline
• MSG-SEVIRI: Instrument characteristics • Aerosol retrievals over the ocean: algorithm principles
• Aerosol models and radiative transfer simulations
• Cloud screening process
• AOD results over the ocean for 2 case studies
• Conclusions and on-going work
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images3
Instrument characteristics of MSG-SEVIRI
Photo: ESA
Altitude: ~36 km
Nominal position: Equator, 0° Longitude
Geographic coverage: a full Earth hemisphere over Europe and Africa
Spatial resolution at the sub-satellite point: narrow band channels ~3 km, HRV ~1 km
Time sampling : 15 minutes
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images4
Retrievals over the ocean: Equations
• Assuming a plane parallele homogeneous atmosphere
- Over a sea target with directional effect [Tanré, 1979]
- Specular reflection of a wavy surface [Cox and Munk, 1954]
- Reflectance of oceanic whitecaps [Monahan and Muircheartaigh, 1980]
- Subsurface scattering
tttTTtTT isoseaseasea
dirseaatmvs
'mod ),,,(
)cos(
),,,(),,,(
0 s
vsmesvs
mes
F
L
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images5
• Atmosphere path reflectance splitted into individual path reflectance due to aerosol and and gas constituents:
• Hypothesis on reflectance for an external mixture of aerosols:[Wang and Gordon, 1994]
• Linear dependence approximation between TOA reflectance and AOD:[Durkee, 1986]
• Best mixture determined by a least square fitting method
21, )1( aeraermixaer ww
AODC 0
aerrayatm
2
mod0
mod
mes
mes
w
AOD retrievals: Equations
• Aerosol is considered to be an external mixture of a coarse (natural) and a fine (anthropogenic) aerosol type with monomodal lognormal size distribution
• Expansion coefficients of the scattering matrix and single scattering albedo calculated using:
- MIE code for spherical particles [De Haan et al., 1987]
- TMATRIX code for randomly oriented spheroids [Mischenko and Travis, 1998]
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images6
Aerosol models
NAM
OPAC
MODIS
[Hess et al., 1998]
[de Leeuw et al., 1989]
[Levy et al., 2006]
• LUT data generated using DAK radiative transfere model [Stammes,2001]
- Midlatitude Winter/Summer profile for gases [McClatchey et al., 1972]
- aerosol layer height: 2 km
- exponentially decreasing concentrations with height
- 15 solar and viewing zenith angles, 37 azimuth angles
- 11 aerosol optical depths at the reference wavelength of 500 nm (ranging between 0.02 and 6).
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images7
Radiative transfer simulations
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images8
Cloud Screening - Method
• Basic principle: “Clouds are generally colder and brighter than most surfaces”
• Method based on the cloud screening technique used in AATSR aerosol retrieval algorithm [Robles-Gonzalez, 2003]
Treshold technique based on image histogram analysis
• Detection tests:
- Infrared Gross Temperature Test (IGTT) – (over land and sea (T12.0)) - Dynamic Visible Test (DVT) – (over land (R0.6) and sea (R0.8))
- Ratio Test (DRT) – (over sea only (R0.8/R0.6))- Dynamic Spatial Coherence Test (SCT) – (over sea only (T10.8, R0.6))
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images9
Cloud Screening -Cloud mask comparisons over Europe
SEVIRI and SAFNWC/MFSEVIRI and SAFNWC/MF
SEVIRI and MODISSEVIRI and MODIS
4 week periods in (Jan/May/Aug/Oct) 2006:Most differences observed on cloud edges, and broken cloud fields.In general ~85% agreement with both products
Case study 1 (FF):Forest-Fire Smoke from the Iberian peninsula advected
over the Atlantic - 7 August 2006
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images10
Cabo da Roca True color scene from Terra-MODIS 11:10 UTC
SARA results for FF -7 August 2006, 11:15 UTC
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images11
AOD at 635 nmAOD at 635 nm
Angstrom 635-870 nmAngstrom 635-870 nm
Fine Mode Weight at 500 nmFine Mode Weight at 500 nm
Comparison with MODIS (MOD04) for FF
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images12
Comparison with MODIS for FFComparison with MODIS for FF
• Spatial variations well represented
• SEVIRI-AOD635 < < MODIS-AOD660
underestimation by ~ 50 %
MODIS - AOD at 660 nmMODIS - AOD at 660 nm
Validation with AERONET for FF
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images13
Validation with AERONET for FFValidation with AERONET for FF
• Diurnal variations well reproduced• Best results obtained for the visible channels• MODIS-AOD860 and SEVIRI-AOD810 overestimated
Case study 2 (DD):Dust Storm across the Eastern Mediterranean Sea
25 February 2006
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images14
Forth Crete
True color scene from Terra-MODIS 09:00 UTC
SARA results for DD -25 February 2006, 09:00 UTC
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images15
AOD at 1640 nmAOD at 1640 nm
Angstrom 870-1640 nmAngstrom 870-1640 nm
Fine Mode Weight at 500 nmFine Mode Weight at 500 nm
Comparison with MODIS (MOD04) for DD
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images16
Comparison with MODIS for DDComparison with MODIS for DD
•Very good agreement between SEVIRI-AOD1630 and MODIS-AOD1630
MODIS - AOD at 1630 nmMODIS - AOD at 1630 nm
Validation with AERONET for DD
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images17
Validation with AERONET for DDValidation with AERONET for DD
• Time series cloud contaminated• General trends well represented• Best results obtained for AOD in the near-infrared
Helsinki, Tuesday 30 september 2008Aerosol retrievals using MSG-SEVRI images18
Conclusion and on-going work
• The stand-alone algorithm for cloud detection with MSG-SEVIRI provided satisfying cloud masks over Europe
• Aerosol retrievals with SARA over the ocean have been validated for 2 case studies (FF and DD): favorable comparison with MODIS and AERONET data.
• Aerosol retrievals over land is an ongoing work. Preliminary results expected in winter 2008...
Thanks for your attention !