Combining SLSTR A + B strips

I’m trying to use the Near Real Time (NR) SLSTR level 1 data. The dataset includes the A and B stripe data but not the combined TDI files. I’m a little new to the Sentinel data and to SNAP so it isn’t clear how I could merge these data. Is there a SNAP tool to merge the A and B stripes to generate a combined dataset scene? If not, can someone suggest a recipe to combine available SNAP tools to generate the combined scene?

I think I may have misunderstood how the stripes should work. I thought I needed to column-interleave the A and B files in order to create the 500m resolution data. After further review I think that they are both 500m resolution views of the same area, so perhaps I should reproject each then average the two in order to create a scene with a higher Signal-to-Noise ratio. Am I understanding this correctly? If not, what is the purpose of having the two versions of the same scene?

Since several month, those c stripe bands are not included in the products anymore.
This is mentioned in the Sentinel-3A-SLSTR-Product-Notice-Level-1B-NRT (

I don’t know how you can work around this situation. I’m not sure if they can easily be calculated.
The general procesude is described in the ATBD:
SLSTR: Algorithm Theoretical Basis Definition Document for Level 1 Observables (

You can find more documentation in the Sentinel-3 SLSTR Document Library - Sentinel Online (

I did find the answer to why there are A and B SWIR sets in the ATBD link. I copied excerpt below. The intended purpose is to average the duplicate samples to reduce the noise although there is a warning about possible distortion at the swath edge.

  1. In order to provide resolutions of 500 m in the SWIR channels, the design of the current FPA for
    these channels incorporates a 2 by 4 array of detectors, the long edge being aligned in the alongtrack direction. As a consequence, the swath is effectively scanned twice in the across-track
    direction, once by each column of the detector array, and so it is appropriate to perform averaging
    on these duplicate samples to reduce noise. Note that at the swath edge the pixels of the
    resultant averaged image will be distorted, but the individual plnes will be retained to support level
    1c processing.

Note that this step follows signal calibration so that the averaging is applied to calibrated
reflectances. This is because the calibrations of the detectors in the two columns may differ.