search

yumorishita / LiCSBAS2

GNU General Public License v3.0
28 stars 5 forks source link

Geotiff cumulative displacement map #41

Closed ditafaith closed 4 months ago

ditafaith commented 1 year ago

Hi @yumorishita

I've read your latest paper (2023) and it would be good to put the paper in your GitHub (LiCSBAS2) as a reference. Anyway, how to generate a cumulative displacement map (geotiff) ? may be I missed this step.

yumorishita commented 1 year ago

You can use LiCSBAS_cum2flt.py and LiCSBAS_flt2geotiff.py.

ditafaith commented 1 year ago

I completely get the desired results. Now, when I try to another frame, I get following error,

  File "/Users/saka/LiCSBAS2/bin/LiCSBAS12_loop_closure.py", line 397, in main
    refy1 = refyx[0][0] # start from 0, not 1
IndexError: index 0 is out of bounds for axis 0 with size 0

If my understanding is correct, this message is related to lack of coherence in the whole area. Is it correct ?

yumorishita commented 1 year ago

Yes.

ditafaith commented 12 months ago

Since I didn't find the parameter in your recent paper, did you use the same parameter to process palsar data ?

I have set up the environment to LiCSBAS2 and I can see the different e.g step 1-6. But I get following error, AttributeError: module 'matplotlib' has no attribute 'colormaps' The existing matplotlib is 3.4.2 and upgraded to 3.8.2 However, when I check through the check install, it showed the old one version. Even I uninstalled, it showed the old one. What is the problem, ?

yumorishita commented 12 months ago

Most of the used parameters are written in section 2.3 of the paper. Please read it again.

I have no idea about the matplotlib issue because it is because of your local environment setting.

ditafaith commented 12 months ago

Most of the used parameters are written in section 2.3 of the paper. Please read it again.

I have no idea about the matplotlib issue because it is because of your local environment setting.

Noted it. I got the ALOS results. However, I still curious why the magnitude relatively too small compare to the Sentinel one. ALOS 1 cm, meanwhile Sentinel 8 cm. Based on existing GNSS 9 cm. I will check it again if there is missing parameters e.g 1-5.

yumorishita commented 12 months ago

Maybe radar_frequency in slc.mli.par is wrong? It must be 1.27e9 for ALOS specified here while it is 5405000000.0 for Sentinel-1.

ditafaith commented 12 months ago

It is solved as follows your comment. Using batch file for ALOS, may not produce calculate dem error, since there is no option to enable. However, it looks LiCSBAS-derived ALOS results (~5 cm) smaller magnitude than previous publications using ALOS and Sentinel which both of them reported ~10 cm/yr.

ditafaith commented 12 months ago

Look forward to getting your insight @yumorishita As mentioned in the paper, dem error ~2 cm, but I want to know (mm/m), is it in millimeter per meter ? demerr

yumorishita commented 12 months ago

It looks quite consistent with Chaussard et al. (2013) for me. mm/m means displacement(mm)/Bperp(m). For example, if the DEM error is 0.02mm/m and Bperp is 1000m, the displacement error caused by the DEM error is 0.02mm/m*1000m=2cm.

image

ditafaith commented 12 months ago

You can obtained -7 cm/yr, meanwhile mine -4 cm/yr. Is there any parameter that may I missed or wrong ? If we use a default one, then it should be similar, I guess. Should I clipped ? rather than whole image ?

yumorishita commented 12 months ago

Use --auto_crange 100 option in LiCSBAS_plot_ts.py and select the point with max displacement for the time series plot.

ditafaith commented 12 months ago

Regarding the DEM error, the area is flat but the error is not evenly distributed. What are the chances of this happening? error_dem1

After applied --auto_crange 100 , the result still has difference ~10 mm with yours. How you can set up the range value 0-100 mm ?

Screenshot 2023-12-04 at 19 44 12
yumorishita commented 12 months ago

I have no idea about the DEM error. I used LiCSBAS_plot_ts.py -c jet_r --vmin -100 --vmax 0. I also manually selected the reference area (see the figure above).

ditafaith commented 11 months ago

Actually, I'm still figuring out the cause, why this result occurred (see the figure below). The number of asc ifgs images is almost 2 times of desc. However, there are many rectangular holes. Both of them use same parameters in licsbas version, to obtain ts and velocities. errors

yumorishita commented 11 months ago

I see the difference even in the sea area. The ascending result has no data in the sea but the descending has. Are these Sentinel-1 or ALOS?

ditafaith commented 11 months ago

Both of them Sentinel-1. Do you have suggestion, may be I missed to check some parameter.

yumorishita commented 11 months ago

I see. The LiCSAR products may have different mask settings of the sea area depending on the data creation time. As far as I know, the areas where the elevation is <0m are masked as the sea in some (or most) data. I suggest checking the input unwrapped data rather than LiCSBAS parameters.

ditafaith commented 11 months ago

Noted, thanks. Let me to check later. However, with the same area and same period, it should be same parameter for data from LiCSAR products. How about the rectangular holes in ascending images ?

yumorishita commented 11 months ago

Do you mean each small white pixel by the rectangular holes? I mean they may correspond to the pixels that have <0m elevation and are already masked in the LiCSAR products in the ascending data. I don't know why they are not masked in the descending data.

ditafaith commented 11 months ago

I am not sure if it is related to <0, since some locations not located near the sea. Also, different results from descending one. I assume it is related to bad pixel in unw data from LiCSAR products since I found some unw data in 2022 mostly same pattern.

ditafaith commented 11 months ago

Hi, recently, I got an error as follows, where I should find the bperp ?

 File "h5py/_objects.pyx", line 54, in h5py._objects.with_phil.wrapper
  File "h5py/_objects.pyx", line 55, in h5py._objects.with_phil.wrapper
  File "h5py/h5o.pyx", line 190, in h5py.h5o.open
KeyError: "Unable to open object (object 'bperp' doesn't exist)"
yumorishita commented 11 months ago

What command did you use?

ditafaith commented 11 months ago

I ran command on step 1-6, LiCSBAS16.py --nomask -t TS -y 0.15, then got that error.

yumorishita commented 11 months ago

Did you use LiCSBAS2 from the start, not LiCSBAS? Also please upload the full log.

ditafaith commented 11 months ago

I use LiCSBAS2 and ran it manual step by step. However, I can send you the log file tomorrow after I run the batch file.

ditafaith commented 11 months ago

Did you use LiCSBAS2 from the start, not LiCSBAS? Also please upload the full log.

Hi @yumorishita, I use LiCSBAS2, and then when I disable to estimate dem error, it can produce the result.

ditafaith commented 11 months ago

This paper looks similar to I want to do. I would like to know if there is any hint to use results from LiCSBAS for further stages as input in GBIS package, is it possible to convert to *.map format ?

yumorishita commented 11 months ago

I have no experience with GBIS. What is *.map format?

ditafaith commented 11 months ago

Actually *.mat file is matlab format. If my understanding correct, it is converted from a txt file which consist of lat/lon/value/inc/heading. I was thinking you know it, since some paper mentioned LiCSBAS.

ditafaith commented 10 months ago

Hi @yumorishita These are the first time I got the following errors (using licsbas2) after run command LiCSBAS_cum2vel.py

Calc vstd... Running 100/100th bootstrap... Traceback (most recent call last): File "/home/saka/LiCSBAS2/bin/LiCSBAS_cum2vel.py", line 303, in sys.exit(main()) File "/home/saka/LiCSBAS2/bin/LiCSBAS_cum2vel.py", line 278, in main plot_lib.make_im_png(delta_t, dtfile+'.png', cmap_dt, title) NameError: name 'cmap_dt' is not defined

Another error after run command LiCSBAS_flt2geotiff.py

Traceback (most recent call last): File "/home/saka/LiCSBAS2/bin/LiCSBAS_flt2geotiff.py", line 213, in sys.exit(main()) File "/home/saka/LiCSBAS2/bin/LiCSBAS_flt2geotiff.py", line 190, in main data = io_lib.read_img(infile, length, width, endian=endian) File "/home/saka/LiCSBAS2/LiCSBAS_lib/LiCSBAS_io_lib.py", line 199, in read_img data = np.fromfile(file, dtype=dtype).reshape((length, width)) ValueError: cannot reshape array of size 1600 into shape (3077,2842)

Is it related to my results or any other things?

yumorishita commented 10 months ago

The error in LiCSBAS_cum2vel.py is a bug. I have fixed it (commit). Please update the package. Thank you for reporting the bug.

Regarding the error in LiCSBAS_flt2geotiff.py, I have no idea without the real data. Probably your input files have inconsistent file size.

ditafaith commented 10 months ago

I can get the results after updating the package. Thanks for your help.

as1234554321 commented 2 months ago

Actually, I'm still figuring out the cause, why this result occurred (see the figure below). The number of asc ifgs images is almost 2 times of desc. However, there are many rectangular holes. Both of them use same parameters in licsbas version, to obtain ts and velocities. errors

Hii @ditafaith I think this area is on the island of Java, Indonesia, can I discuss it with you personally? if allowed please tell me your email or something else. I really hope you will allow me to discuss with you. Thank You