Create Stack with different image parts

Hey Guys,
I am currently working on a time-series of Sentinel 1-GRD data.
This is my workflow:
Read -> Thermal Noise Removal -> Apply Orbit file -> Subset -> Calibration -> Speckle-Filtering (Lee 5x5) -> Terrain-Correction (Range-Doppler).

Afterwards I tried to stack those subsets, but somehow this does not work properly. The images are not perfectly overlapping, since some parts of the Lake I am looking at are not on every image:

The red part is the whole lake, whereas the other colours should represent other images.

If I just stack them like I did, the images that are not fullfilling the whole subset are really strange in terms of form etc. Any possibility to solve this?

With best regards,

Do all your GRD, have the same orbit, and slice frame?

Please take a look at this post, when selecting your data,

Source of the post

Then your subset should match each other,

No they are varying, since I want as many images where my lake is covered as possible in a specific time-frame.

I will read the threads, hopefully I will find some help there. Thought I could use the images since I am subsetting to get only the area where they (partly) overlap.

Did you maybe use different UTM zones for the Terrain Correction?
If all are projected to WGS84 the stacking should be able to correctly align them. Can you maybe upload a screenshot to clarify how they are wrong after the stacking?

If you use the coregistration for the terrain corrected products (this makes sense because they are of different tracks and orbits), you have to select “product geolocation” in the offset method instead of “orbit”.

did you have a look at the images where the NaNs occur?
You should check if the pins are located correctly in the stack and if all rasters of the stack contain data.

I used a Graph and Batch-Processing for it, so probably I did not change the UTM zone. I will check this but I do not think it is possible the way I did it.

I will check where I saved them before closing and provide a screenshot as soon as I can find them. But probably I already deleted them since my hard drive is really small :frowning:

I used Orbit so far, since I did not pay attention. I will try this ASAP with product geolocation (which makes a lot of sense tbh).

I will check after the ‘product geolocation’-stacking if everything is on the right spot! Thank you very much!

By the way, Is there a limit of images that can be cross-corelated? I don’t think there should be one besides the obviously hard time on low-RAM machines.

I used geolocation, Master-extent (my master is a subset with the full extent in this case). Hopefully this will work now :slight_smile:

Hey Guys,
Good news! (As far as I can say).

It worked perfectly for the three Scenes/four Pins I used this time. I will now try to stack all my images and more Pins.
Hopefully, you will not here something from me in the near future, which means everything works (I do not really believe it though :smiley:)

With best regards, and big thank you!

One last question: Would you rather use the Values in the bands after the preprocessing or convert it to dB? (Saw this in another presentation)

This is explained well in here,

Source of the post

So I should not use it, since I will lose “comparability” since each image is adjusted based on its own properties? (If I got that right)

Would you please to share this presentation,

it is part of a course I guess, but they did it.

conversion to dB simply changes the distribution of your histogram. The values stay calibrated and therefore comparable. It really depends if you need high contrasts in the low value regions or in the high value regions.

Since I am using the absolute Values to compare them over time, it will neither help nor destroy my results, so I will just take the no-dB values :slight_smile: