Minimize Radiometric Scaling and Single-pixel Tests

"Radiometric Scaling" aims to interpolate F1 nadir noise and radiances noise estimates to calculate the composite S7/F1 variables and S8/F2 variables.

SLSTR has two channels operating in the 3.7 µm band: the S7 channel, optimised for measurements near to ambient temperature, and the F1 channel optimised for the high temperatures associated with fires. Owing to the superior resolution and noise performance of the S7 detectors, these are preferred wherever they return valid measurements, otherwise F1 detector measurements are substituted, typically when the S7 channel has saturated.

Similarly, SLSTR has two channels operating in the 10.8 µm band: the S8 and F2 channels. Owing to the superior resolution and noise performance of the S8 detectors, these are preferred wherever they return valid measurements, otherwise F2 detector measurements are substituted, typically when the S8 channel has saturated.

The single-pixel test calculates all the tests associated with the fire radiative power algorithm.

  • Test for pixel exceptions. Only fire exceptions are included here.
  • Test for non-land types in Level-1B surface classification.
  • Test for dark water surfaces. When analysing potential land fire pixels it is important to accurately exclude water and mixed water pixels. Such pixels are usually cooler than adjacent land pixels during the day and can cause false alarms. Note that it is assumed S2 and S3 reflectances are always positive. This may not always be the case for the low reflectances typical of water, as measurement noise may be sufficient to give negative values, so there is a possibility for both false positives and false negatives.
  • Test preferred Level-1B cloud flags.
  • Test for cold or bright clouds. Cloud masking is essential to the fire product due to the fact that optically thick clouds make it impossible to identify active fires through passive remote sensing, and solar reflected Mid Infra-Red (MIR) radiation from certain clouds can appear similar to fire signals. Some cloud-contaminated pixels will likely be falsely classified as fires if they are not masked out prior to fire detection. In the frame of coming evolutions, the quality, with regards to FRP detection, of improvements brought to the SLSTR cloud masking will be investigated and verified.
    Note that use of S9 is probably safe as fire ought not to be visible through a cloudy pixel; therefore S9 saturation is not considered to be a problem.
  • Test for sun-glint. By day, sun glint over small, unmasked water bodies or cloud, or even from areas of wet or sometimes bare soil, can increase the MIR pixel signal considerably above the TIR signal and lead to false alarms.