15% isn't bad, check systematic, potentially scattering, differing materials, different geometries, calibration. You need to check all these margins for error, polarisation misalignment, other assumptions about alignment. You need to put a number on all of these and see what they add to. Good luck.

15% low equates to 0.7 dB. For airborne systems this is pretty good, especially at long ranges and shallow grazing angles. If we assume the radar itself is properly calibrated, then what is left is the transmission path and target itself. If as you say that the target's RCS is correct for the viewing angle, and we rule out multipath effects, then all that is left is the atmosphere itself. The atmosphere is quite capable of rendering such an error. The next question is "How consistent is this error over time, weather conditions, geographic area, etc.?" This might be able to shed some light on the source. You might peek at the attached reports for some other ideas.

Thanks Armin and Neil for the answers and Happy new year to everyone. I am not sure if I understand what Armin meant by the 0.7dB of the 15%? The Theoretical RCS is about 30 dB and what I get is ~26dB. I've analysed the radiometric stability of the CRs for over a year and I want to make sure that I did that correclty. I have to explain why the observed RCS is like this and your answers are very helpful. I've attached a figure for the radiometric stability analysis for 7 corner reflectors. One of these reflectors is accidentally damaged throughout the experiment and is no longer usable. I appreciate any commnets/feedback that could help me in my research.

Many thanks

The factors that can act to reduce the RCS of a CR at boresight compared to the theoretical value are misalignment of the reflector (which is not the case here if we assume orientation to be optimum), inter-plate orthogonality, plate curvature, size deviation, and surface irregularities. For Sentinel-1 (C-band) , atmospheric effect might not be so significant.

The details about the reduction in RCS due to various factors can be found in Zink and Kietzmann (1995), Ulander et al. (1991), and Döring, Schwerdt, and Bauer (2007).

Hi Ahmed,

Just wondering how you compute the RCS. In the plot you mention backscatter coefficient, which is dimensionless, and needs to be scaled by the resolution area in order to obtain the RCS (in m^2, or dBsm for logarithmic scale). The resolution area varies between 10*log10(22.5*2.7)=17.8355 dBsm and 10*log10(22.5*3.5)= 18.9625 dBsm. These number need to be added to the (calibrated) slant range backscatter coefficient (Beta0) [dB] to obtain the RCS. The resolution area should not be confused with the pixel area, which is smaller 10*log10(14.5*2.3)=15.2310 dBsm, and should not be used in the RCS computation. Using pixel area instead of resolution area might explain the difference.

Hans

Hi Hans,

Thanks for the comments. With respect to the graph, the y-axis represents the RCS of the CRs and not the backscatter coefficient (that was a mistake in the figure and sorry for that).These values were ccalcualted using the procedrue of point target calibration for Sentinel-1 SLC products ( Miranda, N., Meadows, P.J., TYPE, D. and NOTE, T., 2015. Radiometric Calibration of S-1 Level-1 Products Generated by the S-1 IPF).

Hi Hans Van Der Marel ,

I think in one of your presentations I have seen a different formula for calculation of RCS from Beta0. That formula is: RCS=10*log(Beta0(db)*rg*az)). This formula gives a slightly different (~2 dbsm more) RCS compared to what you suggested above, i.e. : RCS=Beta0+10*log(rg*az)). For my CR the result from the first formula is closer to the theoretical RCS value. Which formula is correct?

Just to check the RCS value in a quick way, I tried the GRD image in addition to SLC, and calibrated and found Beta0 in SNAP. I used 20*22m for the resolution area (https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1-sar/resolutions/level-1-ground-range-detected), and got about ~2 dbsm difference in RCS value (RCS from GRD was ~2 dbsm different from RCS from SLC). Maybe using GRD images for such analysis are not that accurate (because of multilooking) but maybe good for quick analysis. What do you think?

Thanks

Faramarz