[seppi-santiago]

Hello,

I’ve been carrying out tests of PS-processing with a stack of 9 CSK images, selecting always the Sparse Points according to their amplitude Stability Index.

What I would like to know is whether with this approach the obtained results are reliable (the coherence I get both before and after APS removal is high), or if I should increase the number of used scenes , which is something I would like to avoid since I’m only interested in a certain period of time. Making PS-processing on these scenes in another software (Envi Sarscape) did not lead to good results with this number of scenes.

Thanks,
Santiago

[periz]

If you have more images, use them.
Using more images makes all statistical parameters more stable, and estimates are much more accurate.
If you are interested only in a given period of time, you will focus your attention there. But you have to use all images you can get.
You can e.g. use the ampl. stab. index estimated on many more images, get the height retrieved from the whole stack and estimate the velocity (or non-parametric time series) only in a shorter interval. You can do this in Sarproz by playing with read/estimate/neglect each single parameter.

However, if the number of available images is limited by external factors (by no means you can get more images) you have to get results with what you have. In such case, there is no minimum number. With 2 images you’ll just get a differential interferogram. With more than 3 images, you will try to reduce the atmospheric bias and/or to get the most reliable results that you can.

When you have a small number of images, the most difficult thing is that you cannot blindly trust statistical indexes (as ampl. stab index or temporal coherence) any more. So, you have to proceed step by step verifying each single output. Sarproz has been designed right for that. But it takes a good amount of time to become experienced enough to be able to judge when you can trust the outputs or not…

Please consider that there are many options that you could need to change when passing from processing long time series to processing short ones.
If you have few images you should ask yourself: should I remove residual fringes from interferograms? which interferograms (images graph) to use? redundant or star graph? should I filter interferograms? should I unwrap them? which model should I use for the time series? what range of parameters? can I estimate the APS or should I simply estimate average parameters? and so on…

[Author]

Posts