Request for pointers for how to extract SLC data

Hi, I’ve been reading the SNAP help trying to work out how to do what I want, and I find it’s generating as many questions as answers for me. I thought I would explain my requirements here and see if anyone could give me some tips about how to achieve them.

I have some data, which I am happy with, that lets me find a land use category from a latitude and longitude. I also have a sequence of about ten level 1 SLC 1x1 Sentinel 1 IW images of almost exactly the same area downloaded from the Copernicus open access hub (and I intend to download more such sequences).

I wish to access my required data in C or Matlab, but I’m not sure of the extraction strategy, or how much of the extraction process I should do in C or Matlab, and how much in SNAP, so I’m just going to describe the data I want and hope that someone can advise me. I want to select about ten small patches (each perhaps 10 x 10 pixels) for each represented land use category within the intersection of the coverage area of my image sequence, and for each patch I want access to i) the complex amplitudes for each pixel in the patch calibrated to \sigma_0, ii) a look direction for each patch (or at least an incidence angle).

From reading the help, My guess is that I should use the Graph-Builder, and that I need at least the following steps, but I’m not sure of the order: a) Read, b) Apply-Orbit file, c) back geocoding to a single master, d) range shift, e) azimuth shift, f) Split for polarisation and swath (maybe just take one for simplicity) g) Split for burst (maybe just take one for simplicity) h) calibrate to \sigma_0 i) write out complex data to geoTiff?

After that, I guess I need to find a library to read geoTiff, and write out the complex amplitudes and the lat-longs and the look directions, but maybe there is a way that lets me deal with smaller files.

Anyway, I hope this is clear. Some of it is guesswork. Thanks for any guidance you can give me.

I wonder why you want to use SLC:s - to maximise resolution?

Since you are not doing interferometry a complex end-product is of no extra value to you, as the phase-tem is (deterministic) noise. You should get rid of it as early as possible to save in processing time and storage.

I’m not doing interferometry exactly, but I need to capture the phase relationships between the images.

Well, that is interferometry :slight_smile:

1 Like

…and if you are interested in phase-relations for land cover classification for example, you need to compute coherence. This is certainly possible but limits data-selection and is quite a complex subject. Therefore it would be good to start with intensity and move to InSAR once you are comfortable with intensity-only data.

Thank you. Why does computing coherence limit data selection? I suppose if I initially aim for intensity only, then that reduces my question to a single image, and I won’t need back-geocoding. My initial question still applies though, in simplified form.

Perhaps you should use GRD-data as it makes you life a lot easier (no bursts as in SLC). WIth InSAR you need to select coherent data to benefit from the technique, otherwise all you get is noise.

Thanks again. You might be right that I should switch to GRD, the reasons I’m using SLC are that i) it is presumably more directly and simply related to the ground scattering properties, which I want to characterise; ii) I understand that relationship a little better, iii) extra processing can’t gain information, but can lose it; and iv) I’ve already loaded some SLC data from UAVSAR into my model, so I’d like other data that I add to match this as closely as possible. But perhaps by doing de-bursting, I am largely discarding these benefits anyway.

Because I am only trying to characterise the scattering properties, I think it is fine if sometimes my data has low or even zero coherence.