Pre-processing of multi-temporal Sentinel-1A GRDH level

Dear Step Forum Users,
First of all, many thanks for your endless efforts. I have a question regarding pre-processing of multi-temporal Sentina-1 GRDH images.
Prior to sending my message, I have already checked out the step forum and Sentinel 1 Pre-processing Handbook. However, I could not find any reliable answer.
I have selected 12 Sentinel 1A image in descening phase. My study region is covered by 80% mountains. I have done following steps:
1)Subset image(To speed up pre-processing)
2) Apply Precise Orbit file
3)Thermal Noise Removal
4)Calibration to Beta0
5)Terrain flattering(Unchecked grid and bilinear interpolation)
7)Multi-temporal Speckle filtering
8)Range Doppler Terrain correction
9)Linear to db
My questions are as follows:
a) Is this chain is right for time series of Sentinel GRDH data?
b)Can we do subset prior to the Apply Precise Orbite file?
c) Should Terrain flattering be carried out after coregisteration?
d)Is it better to apply multi-temporal speckle filtering after range doppler or before that?
(It is noteworthy that a user in Stepforum has mentioned the speckle filtering(single) should be used before Range Doppler Terrain Correctiong. Right now I have read a paper in which speckle filtering was used after terrain correctiong. The paper is:
Veloso A. Mermoz S, Bouvet A., Toan B, Planells M, Dejoux J, Ceschia E. 2017.Understanding the temporal behavior of crops using Sentinel-1 and Sentinel-2-like data for agricultural applications. Remote Sensing of Environment, 199,415-456.
Please see page 417, 2.3 Remote Sensing Data, 2.3.1 SAR data, The third Paragraph).
I appreciate it your answer.Many thanks

1 Like

I would say yes. Depends a bit on what you want to do with your data. If your area is mountainous, the regins affected by foreshortening, shadow and layover cannot be used for later image analysis.

c) Should Terrain flattering be carried out after coregisteration?

I don’t think that is technically possible. I would expect a message like: “Input should not be a co-registered product”. But if not, I would be interested in the results as well. Maybe you can compare two images with flattening before and after goregistration.

d)Is it better to apply multi-temporal speckle filtering after range doppler or before that?

This point is often discussed. Originally, speckle filters were designed for data in slant geometry, but they surely also work on GRD data. You just have to consider: Data you performed Terrain Correction on was already resampled to a new projection and pixel size. So this is a first kind of reduction of image information. If you then apply a speckle filter on top it may not be as efficient as performed on data before resampling. But in any case it is better than no filter.

Good to see that you informed in the forum and other topics - keep going! As for the preprocessing, many steps have no distinct order, sometimes trying and comparing what is best for your data and study site is the best advise :slight_smile:

Thank you very much.Wish you best!

I have the same question about the Sentinel-1 pre-processing chain.

Because for many applications, we don’t want to process the whole Sentinel-1 tile. But I haven’t found a good way to do this. Say how did you do the “subeset” before geocoding the GRDH images?