Dear Reviewer,

Thank you very much for taking the time to review our manuscript entitled: “Field-validated imaging of decadal and seasonal changes in permafrost bedrock using quantitative electrical resistivity tomography (Zugspitze, Germany/Austria)”.

We greatly appreciate your constructive feedback and the detailed list of improvements provided. We will address all points raised in the revised version of the manuscript.

Best regards,

Riccardo Scandroglio, on behalf of all authors.

Dear Reviewer,

Thank you very much for taking the time to review our manuscript entitled: “Field-validated imaging of decadal and seasonal changes in permafrost bedrock using quantitative electrical resistivity tomography (Zugspitze, Germany/Austria)”.

We greatly appreciate your constructive feedback and the detailed list of improvements provided. We will address all points raised in the revised version of the manuscript.

Best regards,

Riccardo Scandroglio, on behalf of all authors.

Dear Reviewer,

Thank you very much for taking the time to review our manuscript entitled: “Field-validated imaging of decadal and seasonal changes in permafrost bedrock using quantitative electrical resistivity tomography (Zugspitze, Germany/Austria)”.

We genuinely appreciate your detailed and constructive feedback. We are confident that addressing your comments will significantly improve the manuscript's quality.

The core objective of this paper remains the detection and quantification of permafrost degradation. This study is tightly connected to the previous work of Krautblatter et al. (2010 – surely well known to the reviewer) for most methodological details, that are therefore not repeated here. Still, we agree that adding specific information regarding processing, inversion, and clustering will enhance the transparency and reproducibility of our study.

Major Comments:

1. Lack of ERT Data Processing Details: For a manuscript whose primary findings rely almost entirely on ERT measurements, the description of the data processing is extremely brief. The authors state only that data were filtered using the variance of resistance to remove systematic errors. There is no information provided regarding data exclusion thresholds, contact resistance issues, or the percentage of data retained per dataset. These details are critical for assessing the quality of the raw data before inversion.

Answer 1) We aimed at maintaining brief data processing, since most of the data processing is consistent with the previous study (Krautblatter et al, 2010), but we will be happy to improve this section. The required information (e.g. data exclusion threshold and contact resistances) can be easily provided. The percentage of retained data per dataset can be obtained from the lower graph in Fig. 1d, where we will provide a proper axis label.

2. Missing Time-Lapse Inversion Strategy: The authors note that an updated version of CRTomo was used, but they completely omit their time-lapse inversion strategy. Time-lapse ERT is highly sensitive to the chosen regularization approach (e.g., independent inversions, difference inversion, or fully 4D spatio-temporal regularization). Without knowing how the temporal constraints were applied, it is impossible to determine whether the reported 25% decrease in resistivity is a robust physical trend or an artifact of the inversion process.

Answer 2) Very good point, indeed the time-lapse strategy can have strong effects, as shown in Scandroglio et al, 2021. Similar to the previous point, we kept the strategy of Krautblatter et al (2010) and used independent inversions. Therefore, this information is missing but will be added in the revised manuscript.

3. Ambiguity in the Temperature-Resistivity (T-rho) Comparison: The methodology linking ERT-derived resistivities to measured borehole temperatures is unclear. The text states: "Values from the side tunnel (Transect A-A') were used for the field calibration together with values of rho extracted from raw data, as normally done in laboratory calibrations".

Does "raw data" refer to apparent resistivity or inverted cell resistivity? 

If inverted data were used, how were the cells extracted relative to the temperature sensors? Was a nearest-neighbour approach used, or an average of surrounding cells? The spatial scaling between a 1D point sensor and a 2D ERT grid should be explicitly defined.

Answer 3) This technical information can be provided in a short additional paragraph.

4. K-means Clustering and Missing Uncertainty bounds: The authors apply a k-means clustering algorithm to segment the tomograms. However, there is no justification for why k-means was selected over other clustering methods. Furthermore, the authors acknowledge that "clustering is not unique, since multiple solutions are possible". Despite this non-uniqueness, the clustered results are used to generate highly specific, deterministic volume estimates (e.g., estimating 40,000 to 60,000 m³ of frozen rock, and projecting the loss of 39,000 m³ in the next decade). I find deriving these volumetric metrics without assigning any confidence intervals or uncertainty bounds quite dangerous.

Answer 4) The aim of this paper is not to discuss clustering strategies and different algorithms; therefore, a standard algorithm was selected (included in the Matlab basic packages). Confidence intervals are not computable with the “hard” k-means algorithms but could be obtained with other algorithms (e.g. “soft” k-means). An implementation seems easily feasible in a short time.

5. Overemphasis on the Error Model: One of the main focal points of the paper is the application of an updated error model. While proper error weighting is important for image resolution, it distracts from the physical findings of the paper. A range of existing literature has already demonstrated the impact of error models on inversion results. Elevating this to a core research question fragments the narrative, especially given the lack of detail on the broader inversion strategy. This discussion would be better suited as a methodological footnote or moved to the supplementary material.

Answer 5) Emphasis on the error model can be reduced or moved to supplementary material.

6. Unclear Manuscript Goals: The paper lacks a clear "red line." The introduction lists four distinct research questions, but the manuscript struggles to synthesize them into a unified conclusion. The authors should decide what the primary message of this paper is, is it a methodological paper about field calibration and error models, or is it a geomorphological paper about climate-driven permafrost degradation? Streamlining the narrative will greatly improve the manuscript's impact.

Answer 6) We believe that with the suggested improvements, the “red line” will be more visible. The focus is and will remain the climate-driven permafrost degradation, but – as the reviewer also underlined – it is also important to provide precise details on data analysis.

Based on these answers, we kindly ask the editor for the opportunity to submit a revised manuscript addressing all points raised. The suggested revisions can be implemented promptly, as many of these points have already been discussed and proved during our internal review phase. 

Best regards,

Riccardo Scandroglio, on behalf of all authors.