Minimize Derivation pixel-by-pixel

Note: this section is called "Common Processing" in the SLSTR Level-2 documentation.

The Derivation Pixel-by-Pixel process is part of Common Level-2 Processing. Before deriving estimates and uncertainties of BT and radiances, the needed constants are defined and specific arrays are processed and generated. These arrays are contained in one of the four grid sizes defined in the Level-1B product: a 500 m nadir grid, a 1 km nadir grid, a 1 km oblique grid and 16 km tie-point grid. The successive main processing steps are:

  1. Calculate normalised path lengths, i.e. the secants of the satellite zenith angles for each view (proportional to the atmospheric path length in the planar approximation) on the tie-point grid for the nadir and oblique views.
  2. Extract the 1 km nadir image grid time stamp.
  3. Calculate 1 km nadir averaged shortwave radiances and radiance fractions for channels S2, S3 and S6. The 1 km averages are calculated from the two 500 m A-stripe detectors and the two successive integration periods which most closely map to the corresponding 1 km field of view. This also avoids neglecting un-gridded 500 m pixels and duplication of measurements that have been used to cosmetically fill a pixel.
  4. Calculate brightness temperatures and noise estimates. The detector noise performance is characterised separately for each detector in each channel and view with LUTs which map the noise performance of the SLSTR instrument measured during ground testing. The LUTs hold NEAT as a function of the scene brightness temperature and detector temperature for each permutation of detector and, where appropriate, electronic integrator (the "odd" and "even" integrators in channels S8, S9 and F2).

The radiance fraction is the ratio of the measured radiance to the radiance that would be reflected from a white, diffuse surface with no atmosphere and illuminated by an overhead sun (solar zenith angle = 0°). Radiance fraction, divided by the cosine of the solar zenith angle, is equivalent to the diffuse reflectance. Radiance fraction is used in preference to reflectance as it is not vulnerable to divide-by-zero errors near to the terminator.

In the image grid for a particular channel and view, measurements from the detectors are interleaved in a quasi-random fashion, effectively determined by the satellite orbit and the surface topology in the Level-1B re-gridding process.

Radiometric quality files, one for each channel and view, hold measured in-flight values of detector noise over the two black bodies or, for the reflectance channels, one black body and the VISCAL unit. This information is held once per scan line in the measurement frame. For each scan, the nominal NEAT curve is rescaled to fit the measured values with small gain (~1) and offset (~0) adjustments and this adjusted curve is used to look up noise estimates for the brightness temperature pixels.

The following figure describes the succession of the different processing sub-steps.

Logical flow of the Derivation Pixel-by-Pixel section