ESD in GPT doesn't work

Hi,everyone
I am confused why ESD in gpt does not work ,even I just choose a stack of 5 slc with only IW 2-5 bursts.
My configure is -Xmx22680m -Xms2048m.
snap.jai.tileCacheSize=1024
snap.jai.defaultTileSize=512
And my computer with 16cores and 32RAM


and when I run ESD in SNAP, the process seems like stuck in this part for a long time.

Hi,

What do you mean by “a long time”? ESD is quite a heavy process.

Quentin

Hi,sorry for my improper description :sweat:. I mean ESD just is stuck in Azimuth shift,and doesn’t go on for 3h
So it is normal for spending too much time? and could you mind telling whether it is necessary to run ESD?

Hi,

So it is normal for spending too much time?

Not that long. To understand if it comes from computer limitations, you can check your computer’s performance during the process. On Windows, open the Task Manager and look at the performance tab. On Linux, launch the “htop” command in a terminal. Check if you run out of ram. If it does, your computer may swap, meaning using part of your storage as RAM, drastically decreasing your performances.

You may have noticed that between successive bursts, you can have phase jumps. ESD is a technique to remove them. Careful that it assumes that you are working on a “non moving area” (in azimuth). If your area of interest is moving everywhere, don’t even try to do ESD.

Quentin

2 Likes

thanks for your suggestion.so you mean if i use ESD, I will not detect the landslides in any directions of my ares . and If I dont know the specific displacement direction , this step is unnecessary?

DInSAR gets you range displacement. The thing is, using TOPSAR, you will have phase jumps between bursts. One way to remove it is to use ESD. ESD computes the average phase shift between bursts to remove it from the entire scene. It just assumes that it is not moving “too much” in azimuth direction.

The best thing for you is maybe simply to try? Does ESD give you a better interferogram?

So I am reading some researches about using snap-stamps to detect urban subsidence .In their works ,they build graph with ESD in it .and last week I just using back-geocoding , then get the deformation of PS by stamps, I find some PS in river ,I think it is impossible .SO I was thinking that it is something about wrong registration. I will try to using ESD and do my work again .BUT recently I have some troubles in using GPT and SNAP GUI . IT does not work .So I will firstly solve this problem

StaMPS leaves a certain degree of random phase pixels (depending on your settings) to conserve a basic distribution of PS over the entire scene. They do not harm your results as long as your point density is higher over areas which matter. I don’t think this is caused by (not) using ESD.

OK ,Thank you for explaining

There’s no way though that PSI-processing over water with a sufficiently number of scenes will identify PS in water. The changes of that happening must be astronomically small(?)

StaMPS allows to modify the tolerance for random phase pixels. Depending on the general point density, points over water can still remain and I have seen some examples, but as you say, their proportion is really small.