Yup. I’ve read from your paper that you’ve used a linear approach. I tried to apply a linear correction for my trial test but seems to me there is no significant improvement.
Then I will just stick for now with the classic StaMPS/MTI processing and use bulk of SAR data.
I am using StaMPS PSInSAR to examine a volcanic eruption. I have a question concerning the use of TRAIN.
Firstly, I did run TRAIN in StaMPS, but there was hardly any difference to the results. I am wondering if this may be because of my volcano’s very small size?
Here is my thinking:
I note atmospheric effects correlated with topography were previously observed on large stratovolcanoes (e.g. Mt Etna and Mt Llaima), and the effects are visible in interferograms as concentric fringes centered on the volcano (Parker et al., 2015). There are no such fringes visible on my volcano’s interferograms.
I understand stratigraphic atmospheric effects are caused by vertical stratification of moisture in the atmosphere in areas of significant topographic relief (e.g. Parker et al., 2015). However, my volcano is only a small island - (300 m peak elevation / 4 km2 area), so very small compared to Mt Etna (3000 m / 1000 km2) or Mt Llaima (2000 m / 300 km2).
My question: Is my volcano too small an area for stratigraphic atmospheric effects to be a problem, or too small to be corrected by TRAIN? is this discussed anywhere in the literature? Also, I assume that turbulent atmospheric effects will not be a problem because I am using multi-temporal StaMPS with large stacks (40 pairs). Is this assumption OK?
I’m no expert in atmospheric corretion but the footprint of your volcano is so small that the non-stratigraphic atmospheric effects are likely to be small to negligible. I guess that in some extreme foggy situations the stratigraphic effect could make a difference(?)
Thanks @mengdahl Do you mean the stratigraphic effects are likely to be small? If so, then I agree. I understand turbulent atmospheric effects are greatly reduced by using long stacks. @mdelgado can you comment on my original post above?
The typical answer is : ‘it depends!’
The fact that you are using StaMPS does not imply directly that you used all the tools StaMPS provides to mitigate them.
Are you using Step 8 in StaMPS? Otherwise, it may be still part of such turbulence in the results.
Still, long term deformation should be ok as far as turbulence does not happen in all acquisition images. Right?
I personally suggest to check the results and not taken that software does all for granted.
I agree with @mengdahl. If the size is about 300m I do not expect to have topographic linked fringes in the APS. I have seen though for high volcanoes.
There is a quite extensive user manual of TRAIN (Toolbox for Reduce the Atmospheric InSAR Noise) which can work together with StaMPS developed by the Dr. Bekaert, which can bring light on your question. (https://github.com/dbekaert/TRAIN )
One of the possible choices of that toolbox is exactly that. ‘a_linear’
Trying to explain it here will take me much time that for you to read the user manual.
Let us know.
@mdelgado , I read Page 15 of the manual, but did not understand what to do? Should I do this linear method after performing 8 steps of Stumps or before those?I do performed stamps(1,8), then setpath to application/pi/software/TRIAN/ directory and called aps_linear
I feel that those white parts should not be present, and there should be more fixed ps points that seem to have been removed, and this has prevented them from showing subsidence.
the white areas simply mean that they do not contain persistent scatterers. They were removed from the analysis based on their amplitude dispersion (defined in the mt_prep command), their temporal coherence (step 3), the threshold for random phase pixels, and the standard deviation during the weeding (step 4).
This is not really related to atmospheric correction, because atmospheric correction is only applied to the pixels remaining after these steps.
Probably the white areas contain vegetation or water areas?
if the time-series covers events of massive disruptions, you can no longer observe it with PS DInSAR, because these changes exceed what is measurable based on phase differences.
Is it an option to split the time-series into two periods - one before and one after the event (if there was one)?