[bridgetwang]

Hello,

When I use the non-linear deformation processing, I got some questions even after reading the related discussions on the forum. Hope I can get your answers.

1、In sarproz, if “smart” is used, then linear deformation won’t be estimated. I want to know, is the linear deformation phase firstly removed from the interferometricc phase, and nonlinear deformation is estimated from the residual phase which contains atmosphere phase? or sarproz directly extract the nonlinear deformation from the whole interferometric phase ?

2、This question is related with question 1. The nonlinear is assumed to be temperal correlated,so a low pass filter is used. I’d like to ask, when compared with linear deformation, is nonlinear deformation acts as high frequency in time domain, and linear part act as low frequency ?

3、I suppose any deformation contains some nonlinear part, even if the linear part dominates the deformation. Since sarproz can’t estimated linear&nonlinear at the same time, should I always run a linear processing and a nonlinear processing, and then add the 2 results to get my final deformation? And could you tell me what’s the name of the deformation result files?

Thank you!

Best regards

[periz]

1. the atmosphere here has nothing to do. if you use the smart function, the linear phase won’t be removed. However, you still have the option to remove it, by choosing “read” on the corresponding line. Clearly, to read it from the disk, you must have previously processed and saved it.
2. if the non-linear model is obtained with a low pass filter, it will be a low frequency signal, not high frequency one.
3. this question reveals your confusion. the non-linear model will include also the cases of signals which are linear. you cannot estimate linear + non-linear together exactly because you would do twice the same thing. you can export non-linear time-series via csv
best

[bridgetwang]

Thank you so much for your kindly answers. Now I know clearly that SarProz will give linear+nonlinear estimation at the same time when smart is choosen.

I have a further question here: For most cases, I have no prior knowledge of whether the deformation is linear or not. Then should I choose linear model first or nonlinear model first?

As I saw your recommendation to use linear model in other answers and said that it will work fine,signals appearing when something non linear happens. What the nonlinear signal will be like when linear model is used ?

I got these question because for StaMPS, it will not seperate nonlinear&linear signal. But when using SarProz, I always have the confusion which model I should choose and which result is more reliable, as for some cases I don’t have validation data to tell.

So could you please explain a little bit more and the strategy to choose proper model in Sarproz?

Thank you!

Best regards

[periz]

if you have a very good dataset, many images, regular time sampling, good normal baselines dispersion, stable and robust estimation, then go directly with the non-linear model.
if your dataset is not ideal, not many images, not so robust and so on, go with the linear model. then only if you notice problems, you can run a non-linear estimation, maybe only on a sub-selection of points.
to learn reading the signs of possible non-linearity problems, you should take a course. you are welcome to organize one in your institution, then we can send some of us to you to teach the classes. we can discuss this via email.
alternatively, depending on the case you are working on, you could unwrap your interferograms. then the non-linearities would not be a problem
best

[Author]

Posts