I did the InSAR processing by SNAP and get results in a study area related to an earthquake case study. Now, I need to set the displacement from SNAP to absolute and then decompose to two Z and X directions using both Ascending and Descending orbits of S1. Can I do this one using SNAP or I need to program and do by myself?
Thank you so much. For setting the reference area also? Do you know how I can set the reference area through programming? Because before that, I did it only by StaMPS time series software for land subsidence case study.
you simply add or subtract some value in the band maths. If an area with no change has 15 mms in the relative displacement, you can make all pixels 15 units smaller and have the absolute displacement.
The definition is based on your knowledge on the study area. Ideally a point far from the displacement patten (not too close to the image border) with high coherence. I don’t think it’s a good idea to have it defined automatically because you lose control over the final result.
Yes, I totally agree with your point, best knowledge about the study area and geological map in addition to the coherence map highly assist rp selection, but I got from our colleague question to identify the rp spatially. But still the question is raised on our table, How we could select the rp in SNAP spatially in the step of unwrapping to be sure that the unwrapping is done according to the known point?
Thanks. I know how I can set a reference point in StaMPS but I would like to do it in SNAP software. I checked the unwrapping step in SNAP but I couldn’t find any setting in SNAPHU or SNAP related to set a reference point or area.
So, based on your suggestion, I can use band math option in SNAP to set the value of a reference point or area on displacement output and convert it to absolute one. Yes?
I know that It is not very good solution but only way to apply it.
at least this corrects of an overall shift in the displacement raster and sets it to zero where deformation was unlikely. It of course doesn’t correct unwrapping errors which can result in unrealisticly high deformation values. So extreme values shouldn’t be trusted too much and it could be good to remove outliers based on the histogram.
Thank you so much for your nice description. Oh, your notification was very good because I faced like you said in a research about monitoring subsidence caused by volcano activity, when I used a series of ALOS-1 images in a area with very dense vegetated land cover. So, could you please tell me in more details how you use the histogram to detect the outliers and then remove them?
simply have a look at the distribution of values in your displacement map. The minimum and maximum values are probably outside the reasonable range.
I just pointed that out because many people are irritated by the range of the resulting displacement values because it is unrealistic (examples here, here and here).
Take this as an example:
It is clear that the actual displacement is not reaching the extremes, although these values were computed by unwrapping. You can use the mask manager to check which areas are affected by setting thresholds along this gradient.
I wouldn’t necessarily remove them but keep in mind that outliers shouldn’t be trusted too much. What matters is if the overall displacement leads to a clear pattern that makes sense.
If you let me, I would like to give here an example from my problem in that area. You can see them in the following, which the first one is original displacement without any revision and the second one was taken after some edition. You can see the unreliable range of subsidence in that area but after correction, it changed so much.
Sorry, I wasn’t aware that you are talking about PS InSAR, because your initial images referred to classical interferograms. My comments on outliers are not very helpful then.
If you work with persistent scatterers, the unreliable ones are mostly filtered out at several steps (not considered as a PS at all, removal of atmospheric patterns, weeding of the points…)
Hard to tell from the images you posted because the second has no color scale bar and I don’t know what you mean by “some edition”.
No, this is one of my research I did in this field. I only represented it as sample of using histogram to detect the outliers because in this area, I thought that we cannot have a displacement with a value of 13 cm or uplift as a value of 10 cm. So, after drawing its histogram plot, I found that there is only some random points that I couldn’t remove by filtering and the extreme values are related to them.
Yes, I’m so sorry that the second one is not clear. I mean as you said, using the histogram way to detect the outliers.
I have another question. So, after applying a reference point or area, we can have two absolute displacement outputs from both ascending and descending orbits in tiff format. To get the decomposition, we need to find the common points or pixels between both orbits. Yes? In your view, is it true to consider any pixel like a point to find the common pixels?
Yes I read it. In subsidence monitoring, It is easy to convert SNAP outputs in the input format of StaMPS and then, do the time series by that software because after that, we can have a vector of Ps points with lat, lon, and displacement values. I already did it but about DInSAR using SNAP, I don’t know how I can do decomposition