[Help]: Can GMTSAR do pixel offset？

[Xiongwei789]

Hi,
Pixel offset images can be used to determine fault surface traces, especially for areas with severe decoherence, such as the 2023 Turkey Mw7.8 and Mw7.5 doublet. Pixel offset images are clear (See the figure below).

So, Can GMTSAR do pixel offset？I have tried with "make_a_offset.csh", for exmaple:
make_a_offset.csh S1_20230129_033427_F1.PRM S1_20230210_033427_F1.PRM 618 1200 16 16 1
make_a_offset.csh S1_20230129_033428_F2.PRM S1_20230210_033427_F2.PRM 725 1125 16 16 1
make_a_offset.csh S1_20230129_033427_F3.PRM S1_20230210_033426_F3.PRM 698 1200 16 16 1
But the results is strange. Can you give me some suggestions?

[Xiaohua-Eric-Xu]

You probably need to increase nx and ny so the merging could work properly. I used 512x3072 for every subswath. To get more stable solution, you'll have to use larger search window size. It may take quite a while to run but I think someone published a SRL paper on paralleling GMTSAR's xcorr and implemented GPU version too.

[Xiongwei789]

You probably need to increase nx and ny so the merging could work properly. I used 512x3072 for every subswath. To get more stable solution, you'll have to use larger search window size. It may take quite a while to run but I think someone published a SRL paper on paralleling GMTSAR's xcorr and implemented GPU version too.

Yeah, it took a lot of time. There is another problem. There is only one output file named "azi_off.grd". Whether it means shift along the azimuth. How can I get the range shift?

[Xiaohua-Eric-Xu]

You'll neeed to use the output freq_xcorr.dat to create a range offset grid. Similar to what's in make_a_offset.csh. Then use c_speed/rng_samp_rate/2 to calculate rng pixel size and convert the shift in pixels to shift in meters.

[Xiaohua-Eric-Xu]

We didn't implement this to the script as usually people would use phase rather than range pixel tracking, as the latter is noisier though less affected by decorrelation.

[Xiongwei789]

Ok, Thank you. I will try it.

[Xiongwei789]

You'll neeed to use the output freq_xcorr.dat to create a range offset grid. Similar to what's in make_a_offset.csh. Then use c_speed/rng_samp_rate/2 to calculate rng pixel size and convert the shift in pixels to shift in meters.

Hi professor，
Can you provide me with your code? I lack knowledge in this area and would like to refer to it.
Thank you

[Xiaohua-Eric-Xu]

Here's some script to start with. Modify the parameters at the beginning and then convert freq_xcorr.dat to a grid.
make_xcorr_plot.csh.txt

[Xiongwei789]

Thank you. I'll try.

[Xiongwei789]

Thank you. I'll try.

[Xiongwei789]

Dear Xiaohua,
Based on your code, I obtained pixel offsets for the 2023 Turkey earthquake. The maximum value along the range are over 10 m (Figure1). However, your and some results show the maximum values are ~4m (Figure 2).

[Xiaohua-Eric-Xu]

Could you add a colorbar for your plot? Or simply share your grids? The results look quite similar to each other.

[Xiongwei789]

Yes, the colorbar is below the figure. It may be too close. The unit is m.

[Xiongwei789]

Here is scripts and modules I used.
make_a_offset.csh S1_20230129_033427_F3.PRM S1_20230210_033426_F3.PRM 512 3072 64 64 1
make_range.txt

[Xiaohua-Eric-Xu]

problem might be from this line?

gmt grdmath roff_ll.grd $azi_size MUL = rg_offset_ll.grd

[Xiongwei789]

Yes. This is indeed the problem. Sorry for my careless.

[Elcappo2019]

Can you share me the full steps that enable me to create the Pixel-Offset Tracking method with scripts that you used. I stucked many times about this mission ? @Xiongwei789