The Level 2 Aerosol products over ocean are generated from the SEVIRI Level 1B data from a science code provided by IPSL/LSCE and adapted for operational processing. The processing code uses the calibrated reflectances in both SEVIRI solar channels (635 and 810 nm) as input, thus allowing derivation of the spectral signature of aerosol optical thickness. The ouptput product are the AOT given at one wavelength (550 nm) and the Angström exponent. The Angström exponent provides information about the aerosol size and thus about the dominant aerosol type: the smaller the AE, the larger the particle radius. The products are provided at native resolution for each slot. The L2 aerosols data are obtained in clear (i.e. not cloudy) atmosphere over ocean.

Level 3 Aerosol Statistics (mean, standard deviation, maximum and minimum, and number of valid satellite retrievals) are computed from Level 2 products for each pixel and at different time scales:

  • The Level 3 Daily Aerosol product is generated from all individual slots of the day, i.e. from 04:00 UTC to 19:45 UTC.
  • The Level 3 Monthly Aerosol and the Level 3 Yearly Aerosol products are generated from all Level 3 Daily Aerosol files.

The following table lists the Level 2 SEV_AER-OC products and the Level 3 statistics. The table provides links to the product description and algorithm description when they are available.

SEVIRI Level-2 and Level-3 aerosol products
Retrieved parameter Product Name Spatial Resolution Temporal Resolution Period covered
Level-2 Near Real Time aerosol over ocean – Product description N/A Resp.: IPSL/LSCE
Aerosol Optical Thickness (550 nm)
Aerosol Angstrom Exponent
SEV_AER-OC-L2 Native
(3 km at nadir)
15 min
and daily
Level-2 aerosol over land – Product description () Resp.: HYGEOS
AOT (550 nm)
Aerosol Angstrom Exponent
(3 km at nadir)
15 min
and daily
2009 and
since Jan. 2012
Level-3 aerosol statistics over ocean Resp.: IPSL/LSCE/ICARE
AOT Mean
AOT Standard Deviation
AOT minimum
AOT maximum
Number of valid retrievals
Native daily

Summary of the processing scheme

For further details, see the article by Thieuleux et al. (2005) [1].
The first step is the conversion of the digital counts (Level 1.5) to Top of the Atmosphere (TOA) reflectances (Level 2) using the calibration coefficients provided by EUMETSAT in the Level 1.5 data file.
The second step is a screening stage to identify regions contaminated by clouds or Sun glint, or with out-of-range geometric conditions . Over oceans, clouds are more heterogeneous than aerosols and than the underlying surface. Their detection is thus based on their spatial variability of the TOA reflectance at 0.81 µm: an area of 3×3 pixels is considered cloudy if the spatial variability is larger than 0.0045 (empirically determined). The pixels adjacent to cloudy pixels are also rejected. In order to avoid any Sun glint contamination, all pixels whose line-of-sight makes an angle less than 40° with the specular direction are discarded. Finally, pixels with extreme Sun or view zenith angles (>75°) are also discarded because they are outside the valid range of the radiative transfer code (6S*) used for the look-up tables (LUT) computation (see final step).
The final step is the actual retrieval of AOT and AE. The algorithm relies on a set of 15 aerosol models, based on the work of Shettle and Fenn (1979). These 15 aerosol models are used in the 6S radiative transfer code to generate beforehand LUTs of modelled TOA reflectances. Each LUT is used to estimate the AOT that corresponds to the 0.81 µm TOA reflectance for each of the 15 aerosol models, accounting for the actual observation geometry by linearly interpolating between the tabulated values of the Sun and viewing angles, and relative azimuth angle. The LUT is further used to interpolate the corresponding theoretical TOA reflectance at 0.63 µm for each of the 15 aerosol models. The model that best fits the observed TOA reflectance at 0.63 µm is selected.
The final output of the inversion algorithm is the retrieved aerosol model with its associated AE, and the AOT extrapolated at 0.55 µm. When the aerosol load is low (typically lower than 0.07), a oceanic model representative of remote ocean is assumed, and a new AOT is derived.


A first validation exercise of both optical thickness and Angström exponent showed a rather good agreement with AERONET/PHOTONS measurements for June 2003 (Thieuleux et al., 2005) [1]. A more extensive comparison to ground-based measurements was performed in the framework of the GEOmon project.

SEV_AER-OC Product File Naming Convention

The Level-2 and Level 3 SEVIRI products over oceans are made available in HDF4 format.


SEV = SEVIRI mission
<PRODUCT> = Level 2 and Level 3 products name
Examples :

  • SEV_AER-OC-L2: Level 2 Aerosol products derived from the Level 1B for each slot.
  • SEV_AER-OC-D3: Daily Level 3 Aerosol statistics products derived from the Level 2 Aerosol products.
  • SEV_AER-OC-M3: Monthly Level 3 Aerosol statistics products derived from the Daily Level 3 Aerosol statistics.
  • SEV_AER-OC-Y3: Yearly Level 3 Aerosol statistics products derived from the Daily Level 3 Aerosol statistics.

<YYYY-MM-DDThh-mm-ss> = Date and Time (Year,Month,Day,hour,minute,second)

V<X-XX> = Product version

.hdf = HDF file extension

[1] F. Thieuleux, C. Moulin, F. M. Bréon, F. Maignan, J. Poitou and D. Tanré : Remote Sensing of Aerosols over the oceans using MSG/SEVIRI Imagery, Ann. Geophys., 23, 3561-3568, doi:10.5194/angeo-23-3561-2005

back to the top