| 07 Apr 2026
Evolution of L-band SAR Response for Soil Freeze/Thaw Monitoring: A Case Study Over Snow-Covered Canadian Mid-latitude Agricultural Region
Abstract. Soil freeze/thaw (F/T) cycles are critical regulators of global hydrological and biogeochemical processes, yet monitoring these subsurface dynamics beneath snow cover remains a significant observational challenge. There is a corresponding need for physically-based retrieval frameworks to support upcoming spaceborne Earth observation missions, such as the NASA-ISRO Synthetic Aperture Radar (NISAR) mission. Despite the importance of these cycles, there remains a critical lack of understanding regarding L-band Synthetic Aperture Radar (SAR) response beneath snow cover, primarily resulting from a reliance on coarse-resolution data and a lack of coincident, season-long ground validation. To address this, we introduce an integrated physical framework that couples high-resolution (1 m) airborne L-band (1.3 GHz) observations with coincident in situ measurements of soil temperature and permittivity. This approach utilizes analysis of backscatter responses, Freeman-Durden polarimetric decomposition, and the Improved Integral Equation Model (I2EM) to physically interpret microwave scattering and characterize subnivean F/T transitions under frozen and thawed conditions. VV-polarized backscatter exhibited the strongest sensitivity to F/T transitions, increasing during thaw and decreasing under frozen soil. Decomposition analysis revealed dominant surface scattering under frozen conditions, increased surface scattering during thaw, and enhanced volume scattering associated with melt–refreeze cycles. The I2EM simulations captured the VV and HV backscatter trends within an acceptable range across most soil stations, while significantly underestimating the HH backscatter. Overall, these results advance process-level understanding of the L-band SAR response to subnivean soil F/T transitions and demonstrate the potential of high-resolution observations for improving retrieval algorithms and calibrating forthcoming global L-band satellite missions.
The authors analyzed data of the airborne CryoSAR L-band system over an agricultural site in Ontario, Canada, during the two winters of 2022–2024. Backscatter responses, Freeman-Durden polarimetric decomposition, and I2EM simulations were utilized to physically interpret microwave scattering and characterize subnivean F/T transitions under frozen and thawed conditions. Six soil monitoring stations recording temperature and dielectric permittivity as well as a meteorological station measuring air temperature and snow depth were used for supporting the study.
Below are a few general comments followed by more specific comments concerning specific places in the text.
General comments
The CryoSAR mission provides is a unique dataset with high potential for improving the understanding on L-band SAR responses following soil freeze/thaw (F/T) transitions. Moreover, the supporting in situ soil measurement setting is a valuable source for the validation of these observations. Although the potential of this study is promising, the quality of the analysis and presentation is in my view lacking. The results are not clearly analyzed and presented, and it is hard for a reader to evaluate whether the SAR data are useful for soil F/T retrieval or not. If properly analyzed and presented, the analysis can, in my opinion, provide more insight whether these data can or cannot well explain the F/T transitions. Currently the reader remains in a feeling that although the authors mention a few times in the manuscript that the VV-pol backscatter and the polarization decompositions enable F/T detection, the results do not necessarily reflect this statement. Quantitative validation based on the in situ observations will increase the reliability of the study. If this is not possible, at least a clearer qualitative assessment that is backed with some concrete statistics or validation approach should be added. The authors can consider changing the plots or images to a format that will allow the reader to directly evaluate the relationships between in situ observations indicating freeze and SAR observations indicating freeze and vice versa, without the need to switch between figures.
The results are also not properly discussed and compared with previous research in the field. Hence, it is difficult for the reader to know whether these results are in line with previous ones. There are also many small inaccuracies and mistakes in the text, indicating that the authors have not made the required effort to carefully re-read and check small details. The authors justly conclude that the I2EM is unable to simulate the observations. Inclusion of this analysis in the paper is therefore in my view optional. The analysis can be left in the manuscript, but the authors can consider removing it completely, while adding more robust analysis and better presentation with regards to the backscatter signal and polarimetric decomposition parts, which are more relevant.
Therefore , in my view, the paper can be considered suitable for publication, but only after major revisions have been applied. Without the uniqueness and value of the available airborne data combined with the soil station and snow temperature data, I would have leaned more towards a rejection of the manuscript.
Specific comments
Line 40-53: C-band is not mentioned at all even when it is one of the most common frequencies used in SAR. Please add it to the frequency considerations.
Line 54-66: Please add background information on how soil state transitions influence backscatter and polarimetric scattering mechanisms. No information is now provided about whether there is an increase or decrease of these SAR responses after soil freezing and relative to snow conditions.
Figure 2: Increase text size of legend and scalebar.
Table 2: What is the imaging angle? How uniform was it?
Line 108: Airborne data collection did not start in November 2022, but in January 2023.
Line 144-146: Sentence not clear, please check and rephrase.
Line 148: Add figure numbers as references to “time series plots”.
Table 4: Why is 14.2.2023 added here if the precipitation was 0 mm? Also, in case snow depth did not decrease, how do you conclude that the precipitation was rain?
Line 155: The sentence “For our study … snow layer” sounds to me too trivial to be added here. It is obvious that the main scope of this study is soil F/T beneath the snow.
Figure 5: Add the red and blue dotted lines of the top plot to the legend. For a better interpretation of the bottom plot, I would recommend masking out the snow temperature measurements which are during no snow conditions, because they represent air temperature.
Line 249-252: From Figure 6 I understand that there were two temperature and two permittivity sensors in each station. What is the classification logic? Here you explain the logic as if there were just one temperature and one permittivity measurement.
Figure 6: Add vertical lines marking the CryoSAR flight dates, as in Figure 5. Also, increase image resolution.
Line 256-261: Please use same names or numbers in text and figures. Now in the text only the names of the stations are mentioned while in the figures only numbers (Fig 6) or nothing (A1). Also, add a reference to Table 1 when listing the stations in the text.
Line 287: I understand that in theory VV-pol is optimal for F/T detection of soil. However, analyzing the HH-pol and cross-pol is still interesting for finding possible differences, speculating the reasons for them, and for validating the theoretical assumption that VV-pol is optimal.
Line 291-295: The comparison of scattering properties for the different polarizations is interesting and useful, however, there are a few things that should be further clarified. 1) Is it that VV and HH can similarly penetrate the snow layer. If yes, begin the section by saying that co-pol penetrates the snow better than cross-pol (not VV penetrates better than cross-pol). 2) The reduced sensitivity of co-pol to vegetation is well known, but is co-pol (or specifically VV-pol) less sensitive to terrain roughness than cross-pol? If yes, add a reference justifying this claim. 3) Complete “Zeng and Chen” references.
Line 296-299: The combination of these two sentences is not clear. The first sentence “As expected, frozen … content” as itself does not justify what is said in the following sentence “This confirms … channel”. Nevertheless, the first sentence can be moved to the introduction (see comment regarding Lines 54-66).
Line 300-306: There seems to be errors in some of the CryoSAR flight dates. Please correct.
Line 303: “even under dry snow cover” is an odd expression here. Dry snow is almost invisible for L-band and therefore should not be a problem.
Figure 7: The legend ranges between -20 dB and 10 dB. Is this true? If yes, can there be a problem in data calibration? 10 dB is very high for SAR. It seems like the forests show very high backscatter compared to fields. The forests are anyway not relevant in this study, so should they be masked off? Narrower color scale would then allow better separation between pixel values around the different soil stations. Currently it is very hard to observe any backscatter differences between the sites based on the colors of the images. This comment applies to Figures 9, 10 and A2 as well.
Line 316: Add station numbers (see comment regarding Lines 256-261).
Line 322-327: For winter 2023-2024 no in situ observations are shown. At least the evolution of air temperature and snow depth should be available for this time period I guess. If not, it is hard to draw any strong conclusions.
Figure 8: The plot is not clear, too many lines on top of each other, some with similar colors. There seems to be a mismatch between values of this figure and Figure 7. The plot shows stronger backscatter than the colors in Figure 7, at least for stations 4, 5 and 6. But as said in the comment regarding Figure 7, it is hard to estimate the backscatter values from the images.
Line 333: Should it say Figure 9?
Line 340-341: This sentence seems to be contradictory. If higher soil moisture levels cause more surface-like scattering, then how can the surface-like scattering be more originating from the snow surface? Please check the sentence but also confirm the physical interaction with the layers. Is it that wet snow causes high surface scattering from the air-snow interface, more than the snow-soil interface? See comment regarding Line 54-66 about adding more theoretical information on SAR responses over snow-soil combinations in variable conditions.
Line 344: Please be precise, do you mean from the snow-air interface or from the snow medium. I assume the first is true.
Line 351: Are you still referring to Mid-January 2023? 16.9 mm rain happened in mid-February. Be precise about the time periods you are explaining in the text.
Line 352-353: Is it still in mid-January? The snow was dry and thus did not have much influence on the variability of surface scattering. Or was the snow wet here? On what date did this happen?
Line 355: What about surface scattering in dry snow conditions? Is it weak as well?
Line 359: What about the presence of ice layers during other dates? Are you sure that the high volume scattering in the beginning of March is due to these ice layers? What happened to the ice layers later in spring when volume scattering decreased? For example in March 15 the temperatures were negative (most likely dry snow), but volume scattering was much lower than in March 1 and 3.
Line 362: On March 1 temperatures seem to be negative.
Line 364: Please explain more about the influence of field orientation relative to the sun. It is the first time in the manuscript where this phenomenon is brought up, and only as a side comment. What was the orientation usually?
Line 364: On March 1 the snow depth was at its peak. Was the snow cover still heterogeneous?
Figure 10: Change surface scattering to volume scattering in the figure caption.
Figure 11: Please add one plot showing at least air temperature and snow depth as in Figure 6. This would make it much easier for the reader to analyze the results with respect to the in situ conditions. Note one of my general comments about presenting the results in a better format that would help to directly evaluate the capability of the SAR responses to detect soil F/T.
Line 377-383: More information is needed for winter 2023-2024. It is not clear what the conditions were in Nov and Jan. According to Figure 8 the soil state was thaw in Nov and freeze in Jan. However, in the polarimetric analysis it sounds like the soil state in Jan was thaw. There is no in situ data to support the study, and there seems to be a contradiction between the VV-pol soil state interpretation and the polarimetric one.
Chapter 3.5: This chapter should in my opinion be located straight after Chapter 3.2, because the soil F/T state retrieval is based on the VV-pol backscatter as explained in Chapter 2.6. Generally, the soil state retrieval results should be analyzed and presented more thoroughly. Quantitative validation of all flights against in situ sensors is also important, and I don’t see a reason why this cannot be done. The current Figure 11 only shows four flights and is anyway not really helpful when evaluating the capability of VV-pol to detect F/T. Please see general comments concerning better presentation of results.
Line 470: January which year?
Chapter 4.5: This chapter is currently more suitable to be merged with the results. Instead, a discussion concerning the F/T state estimation accuracy with respect to previous research should be added.
Line 544: The freezing mechanism of tree canopy and its contribution to SAR response is quite complex. I would just omit this sentence and say that forests were neglected in this study.
Line 553: Wasn’t it the VV-pol backscatter that was employed for mapping the soil state?
Line 557-559: As said in the general comments, the reader is currently not fully convinced that this statement is true. How well do the VV-pol backscatter and polarimetric scattering components capture the soil F/T conditions?