In this paper, the authors discuss the use of the coherence method on multi-polarization ice penetrating radar data as a tool to investigate properties of ice fabric. Specifically, they demonstrate that narrowing the radar signal's bandwidth helps provide information in deeper areas. They validate their approach from experimental data.

The manuscript is in the form of a brief communication. It is concise, well written, and easy to follow. The paper should be of interest of the EGUsphere readership.  I only have the following comments:

1) On Page 3, line 80, the authors state  "all applied at the respective minimum available center frequencies (Table 1)".  I wonder, however, if it is better to keep the same center frequency and reduce the bandwidth of the signal. Rather than change both simultaneously.  Can the authors provide a comparison of the results obtained when keeping the center frequency fixed? This would be useful to other radar systems, in which data is collected in ultra-wideband mode, and then sub-banded in post-processing.

2) Minor comments:

a) In equation (5), c_0 is defined as the speed of light in vacuum, but that variable is already defined for equation (4). Please consider eliminating redundant variable definitions

b) Page 3, ~line 64-65 "and on the radar bandwidth" consider changing to " and on the bandwidth of the radar signal"

This paper presents a method for increasing the depth at which polarimetric coherence between HH and VV radar measurements drops to zero due to lack of coregistration. The method simply reduces the bandwidth which coarsens (increases) the range resolution so that the two images stay co-registered for longer two-way travel time offsets. Since the HH/VVtravel time offset increases with depth, this implies that the images will stay co-registered and coherent into greater depths. Since only a subband of the full bandwidth is used, the method allows a range of centre frequencies to be chosen. The authors report that the lowest centre frequency produces the best results reporting that higher frequencies resulted in “a less accurate estimate of the horizontal fabric asymmetry”; their preferred explanation is that the longer wavelength reduces the phase variations within a single range resolution cell. I think this explanation should be expanded in the following ways:

1. Please include plots as a function of center frequency along with an explanation of an evaluation metric.
2. I think higher frequencies should perform better as long as everything else (e.g. signal to noise ratio) is held constant. I suggest this because the predicted phase noise will be identical as a function of frequency under constant SNR, but the higher frequencies will cause the signal to generate greater phase variations that are therefore easier to see. In any case, I think a more thorough investigation is needed and recommend including:
   1. A short time Fourier transform (STFT) result with some multilooking/snapshot averaging of the HH and VV signals to estimate frequency dependence of the signal to noise ratio as a function of depth. (Regarding suggestion of STFT: I think any power spectral density estimation as a function of depth would probably be fine.)
   2. A plot of the coherence as a function of depth for several different frequencies some some subband (probably 75 MHz or whatever was producing the best results).
   3. Consider three other explanations that I think are slightly more specific than the one suggested at this point: 1) reduced SNR at higher frequencies; 2) depolarization of the HH or VV waves due to slight misalignments of the crystals would happen faster at higher frequencies/shorter wavelengths; and 3) the worsening of the specular layer assumption at higher frequencies in a way that increases decorrelation due to a difference in the HH and VV scattering from the layer roughness. Plots from (2a) and (2b) above may help support this argument and more specifically point towards the cause(s) of the reduced higher frequency performance.

Also could the reference for equation (2) be more explicitly stated? I suggest adding (Zeising et al., 2023) just before the declaration of (2).

The paper is well written and I was able to prove all results presented from the descriptions and citations provided. The result is interesting, but given that coregistration is always possible and straightforward to apply, I think the most interesting result is the frequency dependence result that the authors refer to but do not show.

--

Two typographical corrections:

Our tests of the bandwidth limitations of ApRES measurements show that the best results were achieved with bandwidths of 75 and 100MHz for centre frequencies of 237.5 and 250MHz, respectively.