I am bit confused regarding the range doppler TC normalization for sentinel 1 SLC data. If I do NOT check the “apply radiometric normalization” box I get the sigma nought refered to the area of the elipsoid (considering I performed the calibration of the S1 at the beginning of my pre-processing which btw does not allow me to calculate beta nought but only sigma nougth). Is it correct? Meanwhile If I check the box I get is the sigma nought terrain corrected using the projected local incidence angle (since i cannot select anything from the options)? If so, this means I cannot obtain sigma nought terreain corrected with local incidence agnle?
What it is really strange is that if I perform the calibration at the beginning of my preprocessing of SLC data then when I perform Terrain correction there is not difference in the results if i check or not “applied radiometric normalization”. So in the end I did not get if I get sigma ellipsoid or sigma norlim. @lveci can you help me?
Moreover in the TC tool there are two strange behaviors of the software:
when one inputs the pixel spacing in m and then click the box with the pixel spacing in deg the pixel spacing just input in meter changes (and viceversa).
when the dem is smaller than the image you correct the final output has the size of the input image even though you have no values where the dem is not available
I’ve an additional question regarding the generation of the theta ellipsoid. By opening it, it is possible to recognize some features that look like topographic features. This makes me a bit confused since I though only the ellipsoid should be used for its calculations. So how the theta ellipsoid is calculated?
I assume that your source sentinel-1 SLC product has been calibrated and debursted. If the product is not debursted, the data are still in overlapped bursts and the terrain corrected result is useless. By the way, your calibration result can be output in gamma naught or beta naught as virtual bands in the S-1 Calibration operator.
Since your source product has already been radiometric calibrated, you should not select the “apply radiometric normalization” option in Terrain Correction. In this case, you will get terrain corrected sigma naught.
If you source product has not been calibrated, then you can select the “apply radiometric normalization” option in Terrain Correction. It should produce the same result as in the above case. The options for incidence angle selections is not applicable for S-1 product because incidence angle is not used in S-1 calibration. This option should be disabled in this case.
Hi Jun, thank you very much for the answer. Yes the products are of course debursted before Terrain Correction.
As a matter of fact I was doing some tests to understand it better. The idea is I need calibrated images still in complex form. So i calibrated using the the option “save as complex output”. This calibration is using the sigma nought LUT (this wasn’t clear to me, since any of the LUTs could have been used, right? I had to check the metadata and compared the result with an “intensity” sigma calibration). Once obtained the calibrated image from complex is not possible anymore to use the tool for converting from sigma to beta nought or gamma (they are unavailable options). Moreover, from the complex calibration is not possible to remove the thermal noise. This forced me to add another calibration block for going with the intensity calibration. Maybe this can be easily fixed and make the entire procedure smoother?
So in both the cases (calibration before TC and radiometric normalization) the local incidence angle information is not used for the radiometric correction?
Hi Carlo, now I understand your question better. For the “save as complex output” option, currently sigma naught LUT is used, meaning that only sigma naught can be saved in complex. You are right that other LUT’s could be used and beta naught or gamma naught should also be allowed to save in complex. To convert sigma naught to gamma naught, you can use Band Maths by simply dividing sigma naught by cos(incidence_angle). But this is not accurate. For S-1 thermal noise removal, the current operator does not handle complex input properly. This should be fixed. But what you want is complex out of the de-noised product, right?
For S-1 calibration (calibration before TC and radiometric normalization), the LUT’s provided in the metadata are used, no local incidence angle information is used.
I don’t think thermal noise can be removed before detection (complex -> intensity). Thermal noise has a known intensity but the phase of the noise is random it cannot be corrected for in the complex domain.
Thanks for the answers. I hope this discussion can improve the quality of the software or at least clarify some concepts
For what concern the calibration&TC I have still some doubts. Let’s assume I apply the calibration and ask to output in complex (that means inphase and quadrature). This would be useful for performing polarimetric analysis. Theoretically, I can decide to use the beta, sigma or gamma nought LUT. Let’s see the various options:
If I go for the beta nought LUT then during the geocoding in order to obtain a calibrated sigma nougth image I need to multiply by sin(theta_loc) during the TC if the option “radiometric correction” is check. If I do not require the radiometric normalization I obtain the beta nought geocoded. Is this making sense to you?
At this point what is not clear to me is what is happening when I go for the sigma or gamma LUT. I expect that sigma or nought LUT depend on beta LUT/sin(theta) and gamma nougth LUT depends on sigma LUT/cos(theta). If this is correct the question is to what reference is refered theta. I would say it is theta ellipsoid. Can you confirm this? In this case during TC the radiometric normalization should be performed by multiply for sin(theta_loc). Is this correct in your opinion?
For what concern the thermal noise I am not really an expert of thermal noise correction but I think is possible to apply the factor to the inphase and quadrature terms (here we are not speaking about phase and intensity) in the same way the calibration factors for beta and sigma theta are applied. From this correction polarimetric analysis should also benefit I suppose. Is this right?
For what concern the thermal noise I am not really an expert of thermal noise correction but I think is possible to apply the factor to the inphase and quadrature terms (here we are not speaking about phase and intensity) in the same way the calibration factors for beta and sigma theta are applied. From this correction polarimetric analysis should also benefit I suppose. Is this right?
No, since the phase of the thermal noise term is random and unknown, it is not possible to apply a correction to I and Q for the thermal noise. The thermal noise correction can only be applied to detected products after the phase (I and Q) has been discarded.
Thanks mengdahl for the answer. It seems than in SLC pre-proc it is better to perform 2 calibrations:
“complex” calibration for going then with polarimetric analysis
Intensity calibration followed by a thermal noise removal to go for purely intensity analysis.
A thing that would help is the possibility to decide the type of calibration for “complex” data ie. beta sigma or gamma. Would it be possible to implement in a next release?
A complex SAR image Z can be presented in two equivalent forms:
The in-phase and quadrature form Z(I,Q) or the amplitude and phase form Z(A,ϕ)
Z(I,Q) = Z(A,ϕ) = Ae^{iϕ}
I = A cos{ϕ}
Q = A sin{ϕ}
amplitude A=sqrt{I^2+Q^2} and
phase ϕ = arctan(Q/I)
and
intensity = A^2 = I^2+Q^2
The bottom-line is that you can calibrate and terrain-correct the intensity that you extract from I and Q in any way you like using the available tools.
