SYN Level-2 SDR processing employs an advanced surface reflectance and aerosol retrieval algorithm, which makes use of the combined information from the OLCI and SLSTR instruments on Sentinel-3 to provide synergistic data for land surface and vegetation analysis. The processing aims to produce self-consistent product packages with atmosphere-corrected surface reflectance for OLCI and SLSTR channels and corresponding information on aerosol properties.

The algorithm takes as input TOA reflectance data for five solar-reflective SLSTR channels at both nadir and oblique views (resulting in a total of 10 channels, the fourth channel of SLSTR is discarded from processing as this channel is dedicated to the cloud detection), and the 16 OLCI bands at all non-absorbing channels, avoiding O2 absorption bands 13, 14 and 15, and water vapour band 19 and 20.

The output is aerosol optical thickness at a reference waveband of 550 nm, an estimate of aerosol model and Angstrom coefficient, and atmosphere-corrected surface reflectance for all input channels. The problem of surface reflectance and aerosol retrieval can essentially be formulated as one of multivariate optimisation subject to multiple constraints.

• Given a set of satellite TOA reflectances and an initial estimate of atmospheric profile, the corresponding set of surface reflectances is estimated.
• Application of the observed set to the estimated set of reflectances results in an error metric, where a lower value of the metric corresponds to a set of surface reflectances (and hence atmospheric profile) that is more realistic.

These two steps are repeated with refined atmospheric profiles until convergence at an optimal solution. SYN atmospheric correction consists of the following steps (Note that collocation of TOA radiance measurements onto the grid of the OLCI reference channel and the pixel classification (cloud, snow, land, etc.) is performed in SYN Level-1 processing):

1. Averaging of collocated TOA radiance over 15 pixels squared. The averaging is appropriate for minimizing the effect of errors in image collocation, whilst retrieving aerosol within the spatial scale of aerosol variability.
2. Pixel-wise retrieval of aerosol properties.
3. Interpolation of aerosol properties retrieved for the averaged grid onto the collocated grid.
4. Atmospheric correction of collocated TOA radiance using the retrieved aerosol properties.