I’d like to stack average several (~5) S1 IW SLC images of an area, but am uncertain about where radar calibration could/should fit in.
I’ve used Sentinel-1 TOPS co-registration flow to co-register my images, but am curious if it makes a difference whether I calibrate each band (to beta naught) and then take a stack average, or take a stack average of the intensities and then calibrate the average to radar brightness. Anyone have an opinion?
calibrating before coregistration shouldn’t make any difference for the coregistration itself but if you want to compare pixel values over a longer time or apply stack averaging you should calibrate each band first. Otherwise you would just use intensities which are dependent from incidence angle, sensor modifications and so on which is why they are not directly comparable.
I’ve created a graph mimicking the S1 TOPS co-registration, but for several images. My thought was to co-register and then radiometrically calibrate each band (as per you recommendation, before stack averaging).
Alternatively, I could create a graph that applies the orbit file, calibrates and debursts each image and then uses the product in DEM-assisted or cross-correlation co-registration for all images. Do you think one calibraton-averaging approach makes more sense than another? Do you have an opinion on which of the co-registration approaches (DEM or cross-correlation) is more accurate?
I am not sure as well but I would personally first apply orbit files to each scene, deburst and calibrate and then use an overall coregistration.
If your area is topographically not too complex you could even perform Range Doppler Terrain Correction on each image and then just use the Create Stack module before applying the Stack Averaging.
I believe I tried to calibrate each image originally before co-registration, but got an error as the Backgeocoding step requires complex input.
As co-registration alters each slave image into the geometry of the master, do you think it feasible to apply radiometric calibration then on the co-registered image (based on the orbit information for the master)?
Simply put, my overall question is whether it’s most accurate to try to utilize S1 TOPS co-registration (which would require calibration after co-registration), DEM-assisted or cross-correlation for co-registering S1 IW SLC products?
TOPS co-registration only uses 2 rasters, so you would have to make a pair of each slave image with the same master.
What is tbe most accurate? I don’t know - could be worth a try. Compare the coherence statistics of different methods to see if there are remarkable changes.
If you calibrated the images already, the back geocoding is no longer possible, but you can use the standard coregistration. DEM assisted coregistration only works if you have pronounced topography which is somehow present in the distribution of intensities.
You should not try to calibrate a coregistered stack. The calibration relies on a look up table for each pixel. These pixels in the slave products have been moved around by the coregistration.
Calibrate, Apply orbit, Deburst, DEM assisted coreg if you’re working with Sigma0 intensities.
If you want to use the TOPS Backgeocoding on complex data, calibrate first with the complex output option.