I am slowly trying to learn this DinSAR stuff in my free time. Using this fine forums and some data recipes from ASF, I have been able to make some progress but I continue to have some questions.
I understand that various uncontrollable things introduce errors in the data when processing pairs of images. My current workflow is to take a series of pairs, so date1+date2, date3+date4, etc to a terrain corrected displacement raster. I then add a coherence band, and a projected local incidence angle band, and export the final geotiff into Good Earth Engine as an image Collection. Then lastly, I have made a median image after masking out low coherence pixels. My thoughts were that this would mitigate the errors from atmosphere or whatever. Does this sound correct? Should I be using date1+date2, date 2+date3, etc instead for my processing pairs?
In addition, I see displacement everywhere, and when I create a chart for an area of pixels, I see a general trend in the series of images that is replicated all over the image, suggesting that the indicated movement is questionable.
Is this possible (atmospheric or otherwise) error usually localized to a few pixels, or large areas?
I repeated this process in my area for both ascending and descending passes and got different values. It looks like using the two looks to calculate true motion is not so simple, is there a “close enough” method?
Atmospheric artefacts at quite low frequency, i.e. kilometres to tens of kilometres in size. If two pixels are 100m from each other you can pretty safely assume that the atmospheric contribution is the same.
If you see a general trend over many scenes it suggests something went wrong.
When I say general trend over the scene, I mean that a small polygon on a homogenous area shows similar relative values when charted across all the images. The specific +/- values change between polygon locations, and I see different values for displacement and uplift in different locations that seem to be moderately consistent with changes in surface cover.
Someone on these forums suggests assuming some highly coherent mountain be used as a zero reference point, and all motion corrected to zero based on that areas displacement. Is that a proper way to proceed?
DInSAR in essence is to bring the perpendicular baseline to zero(making both acquisitions same) in order to monitor change in height(not topographic!) or xy . It is also necessary to use a common master among both IFG pairs. Use of Ascending-ascending or descending-descending pass footprint results will be different due to change in geometry and orientation of the objects on ground. Hope this helps.