I have made a sequence of interferograms with a time frame of 12 days and with a spatial distance of less than 90 meters following the ESA manual.
I have sifted the result using only those pixels with coherence value > 0.4. I have taken linear displacement and terrain correction. I export the final file “displacement_TC” from SNAP to take it to R.
It is the temporary record of a very inclined slope where works have been carried out for a hydroelectric project. The civil works area covers an approximate area of 1.5 km. There, I have a population of pixels or points of 30,000 data per interferogram.
My surprise is to find that in several cases, the displacement values of one interferogram are all positive and in the immediately following one, all the values are negative. Since this is repeated several times in a sequence of 148 interferograms, I want to ask for help in understanding the problem. The questions are:
Is there a problem with constructing an interferogram of a wide region, with large decorrelated areas, and selecting a subset where the coherence values are high (>0.7) and where the wrapped and unwrapped interferogram is of average quality?
We have identified that the displacement values change from positive to negative and vice versa as “toad jump” (“salto de sapo” in spanish). What is the cause of this error? I find in a recent document that the negative values are changed to positive because the practical meaning of this does not make sense when talking about landslides on a very steep slope. However, it seems to me that it is a not very solid hypothesis in the study of interferometry.
Rapid changes from positive to negative displacements are likely due to tropospheric water vapor delays on certain dates. If you have a series of dates A,B,C, and there is a big tropospheric delay effect in date B, then the A-B interferogram will have an apparent decrease in range, and the B-C interferogram will have an increase in range around the same amplitude. This is why people run time-series analysis programs on sets of interferograms to better separate the temporary atmospheric effects (tropospheric water vapor in most cases) from ground displacements.
E.J. Fielding: Thank you very much for your response. It guides me towards the idea of eliminating interferograms with very high atmospheric error. But I would like to know if this is appropriate.
To date, there is no standard procedure that can be carried out with free software to correct the atmospheric error in the interferograms made with Sentinel 1. It is a personal assessment based on the level of knowledge I have on the subject. Or does this possibility of correcting the atmospheric error already exist in SNAP?
I have 146 interferograms built for the period December/2017 to July/2022. In them, the transition from negative values to positive values for all pixels is repeated eight (8) times.
In a consecutive sequence of interferograms A, B and C, interferograms A and C have their positive values and B has negative values. The solution would be to eliminate interferogram B from the time series?
I am monitoring a slope that in April 218 suffered surface instability as a result of an internal collapse caused by an underground excavation.
On the surface there are areas that have not moved and I assume that their values, positive or negative, are the result of atmospheric error.
But that all the values of the interferograms are positive is a difficult situation to explain and even if all the values are negative.
Could you please tell me which time series analysis programs would be most appropriate to adequately separate tropospheric error from ground displacements.
The time-series package that I use is MintPy, which is available on GitHub. It has a section on how to use stacks processed with SNAP. There are two repositories, one with the source code and instructions on installation (the easiest is “conda install mintpy -c conda-forge” if you have Anaconda), and another repository with tutorials on using the software.
EJ Fielding: Once again, I appreciate your collaboration. I’m going to read MintPy and we’ll try to apply it. The challenge that we have set for ourselves is to obtain the “terrain correction” of each interferogram and from there, process them with the R statistical program.
We are going to try to work with Mintpy and if we have difficulties, we would be asking you for additional guidance.