Questions about the processing steps for multi-temporal sentinel-1 SLC data


I have three SLC datas :

In terms of local incident angles and elevations,I want to study:
1.The variation of backscattering coefficients in different periods
2.The variation of polarization parameters obtained by h/a decomposition in different periods

My questions:
1.Coregistration and Multi-tempol speckle filtering.Are they necessary?
2.Multilook.Is there a difference between the looks 1:4(calculated) and 1:5(ESA)?
3.Terrain Correction.
a.output bands:To get the local incident Angle, which one should I choose?Incidence angle from ellipsoid,Local incidence angle or Projected local incidence angle?
b.radiometric normalization:Is it necessary for my research?And when I select “save Sigma0 band” ,it does not work?

The processing steps I am using is as follow:

  1. Read
    calibration(sigma0 band)
    Multilook(GR Square Pixel,1:4)
    speckle filtering (Lee sigma)
    Terrain Correction (DEM : SRTM 3Sec,output LIA/DEM,radiometric normalization)
  2. Read
    calibration(complex output)
    Multilook(GR Square Pixel,1:4)
    Polarimetric speckle filtering (refined LEE 7x7)
    Polarimetric Decomposition (H-A dual pol)
    Terrain Correction (DEM : SRTM 3 Sec)


Especially if you want to compare images of different dates, multi-temporal speckle filtering is not a good idea. It wouldn’t be effective on three images either.

Coregistration only makes sense to me if you want to compare single pixels for their properties. For general scatter plots (e.g. incidence angle vs. Sigma0), this is not required.

If you want to compare backscattering coefficients (your first workflow), the normalization is no longer required. It is kind of a relic from earlier versions (see lveci’s answer here)
If you want to correct radiometric errors caused by topography, you should run Terrain Flattening (as suggested here), but its results strongly depends on the used DEM.

I would select the SRTM 1 Sec for the Terrain Correction because of its higher spatial resolution.

Thank you :smiley:

It is best to derive the polarimetric parameters from undisturbed (non-interpolated) values and to co-register at a later stage if necessary. Multilooking destroys information so it should be done later also if necessary.

1 Like

Thank you for your answer! I am new to this field, and I still have a question.Since my research object is snow cover in mountainous areas, I want to study the effect of local incident Angle and elevation on its backscattering coefficient.If I run Terrain Flattening, is that the same as removing the terrain?And I can’t study the effect of elevation on it?

Thank you for your advice!!

if you want to study the effect of the local incidence angle (topography), you skip the terrain flattening and simply convert to Sigma0 (corrects for the global incidence angle).

This is the rgb image.And there are water, farmland, bare land, mountains, towns…

Such as water,the main scattering mechanism should be surface scattering,so it should have a smaller α value.But it’s actually over 40°.Exactly the α value of the whole image is relatively large.Maybe there’s something wrong with my process?

The processing steps I am using is as follow:

  1. calibration(complex output)
  2. Deburst
  3. Multilook(GR Square Pixel,1:4)
  4. Polarimetric speckle filtering (refined LEE 7x7)
  5. Polarimetric Decomposition (H-A dual pol)
  6. Terrain Correction (DEM : SRTM 3 Sec)
1 Like

the derivation of scattering mechanisms is not very reliable from the dual-polarized Sentinel-1 data (VV/VH). To extract precise proportions of surface, dihedral, and volume scattering, you need quad-pol data.

You can still use Sentinel-1 to get first indications and detect differences between surfaces, but I would not refer to value thresholds which were empirically tested on quad-pol data.

1 Like

So the results of H/α decomposition in SNAP and Polsarpro are same?and if I want to determine the scattering mechanism based on the α value, I can only use quad-pol data?

SNAP and PolSARpro should lead to the same results. The question is how much they help you. It was discussed here, to what extent dual-pol data can be used as input for Ha decompositions.

1 Like