2 Issues: Reference Point Step & Overall Low Coherence

[Ozzy Tirmizi]

Good morning everyone,

As the title suggests, I am having 2 issues with my MintPy processing.

Firstly, I seem to have overall low coherence in my stack of Sentinel-1 images (highest the avg goes is about 0.6). Initially, everything was extremely low so I re-created my stack using stackSentinel and topsStack processing in ISCE and set the coherence threshold to 0.2 instead of the default based on some recommendations. Although I have seen an improvement, the stack still has relatively low coherence (0.4-0.6) and I was wondering if anyone had any advice on that.

Secondly, when running the Reference Point step, I get the following error: TypeError: Indexing arrays must have integer dtypes. I have tried it on all auto settings and by manually setting the minCoherence value to 0.6. Not sure how to fix this issue. I appreciate any input y'all might have.

Thanks!
Ozzy

[Zhang Yunjun]

Hi Ozzy,

If you could post your pic/coherence/unwrapPhase/connectCompoonent*.png files here, I could take a look and maybe give you some advice.

The Indexing error is due to a bug in h5py version 3. We have add a constrain recently in the repo to use h5py version 2 only. Updating your mintpy code and re-install should fix this error.

Cheers,

Yunjun

[Ozzy Tirmizi]

Hi Yunjun,

Thank you for the quick response. Unfortunately I have been using the Jupyter Notebook from the ISCE short course so my pic folder is not populated. I will re-run the steps so far in the terminal and upload the pngs as requested.

The first thing I did when I ran into the error is to do a git pull to update the mintpy code before trying again. Any other suggestions regarding this issue? In the meantime, I will make a new environment and re-install mintpy to see if that helps.

Thanks again!
Ozzy

[Ozzy Tirmizi]

Good morning Yunjun,

Just wanted to give you an update. Re-installing MintPy fixed the h5py error and the reference point step run just fine. I am currently on the network inversion step. Although I have mintpy.plot set to auto, it has not been plotting and saving to the pic folder on each step. Once the processing is complete, I will run the bash script to generate the plots and share the requested ones.

I really appreciate you taking the time to give your input.

Thanks!
Ozzy

On Tuesday, November 24, 2020 at 12:18:10 PM UTC-6 Zhang Yunjun wrote:

[Ozzy Tirmizi]

Good evening Yunjun,

Before I downsample the data and re-process as advised in the other thread, I wanted to go ahead and share the requested png files to see if there is any way to improve it this time around. One thing I think will help is probably applying a watermask and I tried to make one using mask.py but ended up with the entire scene being a constant value of 1. I am also using a TanDEM-X dem so I'm not sure if using the regular SRTM routine will work. I tried the create watermask jupyter notebook as a guide but I keep getting a syntax error whenever I try to specify my coordinates to download the SWBD files for my area.

Thanks for all your help!

Ozzy

On Tuesday, November 24, 2020 at 12:18:10 PM UTC-6 Zhang Yunjun wrote:

[Ozzy Tirmizi]

update: got the watermask to work when running wbd.py from ISCE in Jupyter Notebooks instead of terminal and followed this notebook (https://nbviewer.jupyter.org/github/insarlab/MintPy-tutorial/blob/main/applications/water_mask.ipynb) to make it compatible with MintPy.

[Zhang Yunjun]

Hi Ozzy,

From your connectComponent_*.png files, I see there are some interferograms with sparse connected components coverage, e.g. 13, 14, 26, 27, 31 etc., which should be caused by low spatial coherence for those interferograms. I would suggest using mintpy.network.excludeIfgIndex to exclude all those interferograms; or using mintpy.network.coherenceBased = yes to exclude interferograms and low coherence on the area of interest defined by mintpy.network.aoiYX/LALO, such as Fig. 10 in Yunjun et al. (2019).

Yunjun

[Ozzy Tirmizi]

Good morning Yunjun,

Thank you once again for your valuable feedback. With the downsampled dataset, I have been using coherence based network modification and got some good results with setting the minimum coherence to 0.2 and 0.5. Currently trying with 0.7 and running into errors and slowly working through them (the higher threshold is more so for my own curiosity as I already have some usable results so no big deal). I also wanted to mention your paper has been an incredible resource and I really appreciate the citations next to certain steps/parameters in the config file as it makes it significantly easier to understand how they will affect the processing.

Thanks!

Ozzy

[Zhang Yunjun]

I am glad to hear!

Yunjun