I have data from end of march until the beginning of November from S1A and S1B.
The workflow of preprocessing this data is following:
Apply orbit file
Subset
Thermal Noise Removal
Calibration (just beta nought because of the very rough terrain, ignoring the local incidence angle)
Speckle-Filter
Terrain Flattening
Terrain-Correction (Is the resulting distortion in the picture problematic for further data evaluation?)
Afterwards i create stack of both S1A and S1B and apply multi-temporal speckle filter on the stack, but here comes the first problem, opening the first band takes about ten minutes. Even though i have 32 GB RAM.
But the question is, is this the right way of combining S1A and S1B? Or do i have to create it separately and after finishing preprocessing and combine them afterwards, if so, how to combine two stacks?
I started with remote sensing a month ago, so i am relatively new. I appreciate every kind of help.
actually, if the multi-temporal speckle filter has written the result to a new product, opening of each band shouldn’t take longer than opening it before the filtering. Did you save the data as BEAM DIMAP format?
I would skip the speckle filter of single images and only perform the multi-temporal filter on the final stack.
In my opinion, you can put all images into one stack before multi-temporal filtering as long as they were acquired from the same track.
Data is automatically saved as BEAM DIMAP always.
They are always put in one stack but they are from different tracks, am i only allowed to use data from the same track for analytics?
If so, are there any papers for preprocessing multiple track data which can be put in one stack?
Technically, you can still stack data from different tracks, but there might be some topographically induced patterns (despite Terrain Flattening) which can confuse the multi-temporal speckle filter which decides which pixels are affected by speckle and to which degree. So the result has to be checked carefully before you proceed.
As there is no general order for preprocessing, you could maybe share your overall goal do we can give better advice.
Thank you very much for the help and explaination.
I am doing my thesis on mapping irrigated crops on a partly very rough terrain.
I am trying to get soil moisture estimates, and DVI’s that correlate with irrigated crops.
Soil moisture estimate is done with Sentinel 1A&B, and DVI’s are retrieved with Sentinel 2A&B.
Now i am trying to retrieve the dielectric permitivity from the Dubois empirical model.
In some papers there is a workflow process called incidence angle normalization to where they retrieve the sigma nought for 30 degrees only.
My incidence angle from left to right varies at most 1 degree and is around 43°, so i am curious if i have to apply this process and if so, is there a function in SNAP?
(From my knowledge only Range-Doppler terrain correction normalizes some sigma nought in regard to the incidence angle by multiplying the ellipsoid value with the ratio of the DEM incidence angle to the ellipsoid incidence angle)
Actually, Range Doppler Terrain Correction only reduces geometric distortions in the images which result from foreshortening of the signa. In turn, Terrain Flattening reduces radiometric distortions based on the different areas (per pixel) illuminated by the sensor due to topography. This completely removes the impact of incidence angles in your image. I don’t know how to normalize backscatter for a defined incidence angle, but these things have been discussed in these topics: