I appreciate the authors’ effort in editing the manuscript and answering my comments.
I still had some issues with understanding the goals of this work. However, based on the original submission, the revised submission, and the authors’ comments to my first review, I now suspect that some of it has to do with some persisting lack of clarity in communicating the authors’ goals and findings - especially early in the manuscript - in a way that would – in my opinion - best serve their work. It is now my impression that the actual intents are:
1) to make the remote sensing community aware that their satellite-derived product and whichever validation dataset they use to assess the performance of these products, are inadequate, as long as they only categorize soils into two states – thawed and frozen – based on the 0 degC temperature. Because - as previously known and again showed here - soils in cold regions are in fact in an intermediate (partially frozen) state for an extended period of time;
2) To show that the ‘problematic’ (i.e. requiring soil-specific calibration) conversion of permittivities into water content is, in fact, unnecessary if qualitative evaluation of a soil is sought in terms of frozen - partially frozen – unfrozen categories;
3) To show that quantitative evaluation of soil freezing state is possible in terms of freezing probability, based on the permittivity data. Additionally, the authors confirm that rather than air and soil temperatures, it is the soil type and vegetation/ecozone factors that are controlling the soil freezing probability.
Assuming that my understanding as described above is now correct, I would have the following suggestions to the authors:
Regarding point 1) above: As I’ve already commented earlier, the fact that soil doesn’t turn into fully frozen at 0 degC is a fact well known to (not only) cold regions scientists working in some capacity with soils. Even if common remote sensing practices consider ground below a temperature of 0 degC as frozen, it is obvious, even just from physics and chemistry, that this cannot be the actual case. However, this may not be reflected in the current state-of-the-art of remote sensing products. This being the shortcoming is suggested in the references in the authors’ response to my comments. If this is the case, I can see how the wide climatic and geographic spread of the presented in-situ dataset would provide possibly more compelling evidence for the remote sensing practitioners than a dry reference to soil/cold region science literature. But in the interest of clarity and readability, this should be explained more clearly, and earlier in the paper. The scope and applicability of the presented work only become clear in the Conclusions, which I think provide a much better summary of the relevance and application of the presented work than the Introduction currently does.
I’d suggest to the authors to focus more attention in the Introduction (and the Abstract) on current practices and limitations in remote sensing (which this contribution claimed to be targeting) rather than re-stating known facts of soil science. The authors’ own response to my first review could be used as a base for this: “Many remote-sensing freeze-thaw studies—from early work (Kim et al., 2011; Zhang and Armstrong, 2001) to recent applications (Taghipourjavi et al., 2024; Gao et al., 2020; Kou et al., 2017; Roy et al., 2020; Derksen et al., 2017)—rely on 0°C soil or air temperature thresholds for training and evaluation, without accounting for partially frozen states. Only recently have researchers begun integrating soil moisture, temperature through SFCCs into freeze-thaw model evaluation (Rautiainen et al., 2025).” I think that if the Introduction section was developed around this key message - and perhaps around additional clearly outlined points, such as the mine ones above (if my understanding is correct now) - then the purpose, scope and contributions of the paper would be clearer from the start.
I apologize to the authors for referring to permafrost in my previous round of comments. I have now understood that my confusion as to which soil state they were analyzing has come from an interchangeable use of words like frozen ground, seasonally frozen ground, and permafrost. From the Line 23, the authors write: “Frozen ground is defined as a condition in which pore water (the water inside the soil) turns into ice (Williams and Smith, 1989). Recognized as a key climate change indicator by the Environmental Protection Agency (EPA), this phenomenon is widespread and affects most land areas above 45 degN latitude (Zhang et al., 2003)” I find this confusing, because the reference Zhang et al, 2003, discusses “distribution of seasonally and perennially frozen ground” in the northern hemisphere. But seasonally frozen ground is not a “key climate change indicator by the EPA”. According to Climate Change Indicators in the United States report (Fifth Edition, July 2024 EPA 430-R-24-00), it is permafrost i.e. the permanently, not seasonally, frozen ground, that EPA considers a climate change indicator (specifically, the temperature of permafrost). I am (in Europe) used to permafrost being defined as permanently frozen ground, or ground frozen for a minimum of 2 consecutive years, not just frozen ground which, by authors’ definition and references also encompasses seasonally frozen ground, which is not permafrost, which is not a climate change indicator. I would suggest to the authors to edit and align definitions of key concepts they’re using/referring to.
Specific comments:
Line 46: “In this zone, the pore water content remains almost constant, regardless of temperature changes.” I’d suggest an alternative formulation “In this zone, the pore water content is independent of the soil temperature.” If the authors’ current formulations was correct, then they wouldn’t have had all the issues with constructing the STCC where they argue that the variable water content (and water input) during thawing makes it problematic to reliably construct the STCC.
Line 63: as I understand the reference, the hysteresis discussed was observed in the active layer (seasonally frozen portion of the ground above permafrost), not in the permafrost portion of the ground.
Fig 5: I am not super familiar with forest plots; however, the confidence intervals (plotted in the vertical direction) don’t seem to have a corresponding scale on the y-axis. Should the confidence intervals be horizontal instead? Or should there be a secondary y-axis added ? Or are the confidence intervals only relative?
Figure 6: The subfigures could benefit from a legend.
Line 300-302: Our results were generally consistent with the physical factors affecting SFCC shape, including soil mineral composition, particle size, plasticity, initial water content, dry density, solute concentration, freezing rate, confining stress, and freeze-thaw history (references). – Were all these soil parameters known, to say that the results were consistent with the values in the references? In general, an overview of the vegetation cover and soil type at each site could be relevant for drawing conclusions, as this information should be available from the mositure probes installation.
Line 302-305: “Networks with wetter soils or limited insulating organic layers—such as BJ and LR FM and CP (wet), and GR, TV, (with thin organic layers)—exhibited higher b values (b > 3), indicating sharper freezing transitions driven by abundant capillary water and/or faster freezing rates… In contrast, finer textured or drier networks, such as KN and BT, showed smaller b values (b < 1.5), reflecting more gradual phase transitions”. - I’d expect the wetter soils to freeze relatively slower, due to more water that has to undergo phase change. Such soil would also typically be more fine grained. Also “abundant capillary water” would in my opinion be typical of a finer soil, and causing slower, not faster, freezing rates. I understand if the findings don’t align with all the previous references, but there seems to be contradictions in the conclusions drawn by the authors. These apparent contradictions lead me to question whether one can say, as in line 300-302, that the results are indeed consistent with references?
Line 320: The volume mismatch and/or the sensor geometry should only affect the measurements in a narrow range of temperatures close to the freezing point, and thus unlikely to explain the consistent hysteresis observed at lower temperatures, such as -5 C in Overduin et al. (2006), and -4C in Tomaskovicova & Ingeman-Nielsen (2024). In any case though, it appears that it is mostly the thawing point that is affected, and the temperature at which the soil begins to turn into ice is more physically plausible in all of the Pardo-Lara, Overduin and Tomaskovicova… datasets. Since the authors of the present manuscript excluded the thawing curves from the analysis, shouldn’t it help constrain the Tf estimation to physically more plausible (below 0 degC) values? And, in Line 321: I wonder if some of the positive Tf could be remedied by temperature sensor calibration? The temperature sensor accuracy of only 0.5 degC in what corresponds to the transitional zone is relatively coarse (best practices for permafrost monitoring recommend 0.1 degC or better; here I am citing permafrost practices, because it is in monitoring of frozen ground where accuracy of soil temperature measurement below 0 degC is very important). Moreover, the temperature offset is non-linear, so the error can be different through a range of negative temperatures experienced at a site – even more so for what looks like a temperature sensor with relatively low accuracy.
Line 362: “We did not perform site or sensor-specific calibrations for permittivity or temperature, but the sensors used—TEROS12, HydraProbe, and CS616—have been extensively validated and shown to perform reliably across diverse soil conditions”. Temperature sensor accuracy is not directly linked to soil conditions, so I don’t think that part of the argument adds credibility to the accuracy of the soil temperature measurements.
Line 328: the two sentences starting in this line are repeated twice, delete the repetition.
Line 374: “This [underestimation of permittivity] is unlikely to affect our analysis, as saturated soils are rare during freezing periods—except at FM, where soil rarely freezes.” Yet the E boreal forest was described as a “wet”. Do we have estimates of the actual water contents ?
Line 343: “The thawing cycles, however, were not analyzed in this study because constructing the STCC from in situ measurements is not reliably feasible.” – In my experience, before the temperature at the depth of the moisture sensor first time goes above the freezing point in the thawing season, the thawing curve should be possible to construct, because the liquid moisture content dependance primarily on temperature holds, assuming that a frozen ground doesn’t accept much additional moisture from rainfall or thawing. Perhaps hysteresis, or biased Tf during thawing, are more valid reasons as to why the STCC is unreliable, or why it is different from the SFCC.
Line 351-353: For this kind of application, could another solution be to monitor the soil moisture at a slightly larger depth, where only the sustained freezing /thawing cycles penetrate, ie. at least below the depth of daily temperature fluctuations (20cm?). |
Please find my comments in the attached pdf document.