The submitted manuscript, "High-resolution inventory and classification of retrogressive thaw slumps in West Siberia," presents an extensive update to the existing RTS inventory by manually mapping over 6,000 features across a vast region (445,226 km²) using multi-source, high-resolution satellite basemaps from 2016 to 2023. The study aims to enhance our understanding of RTS distribution, scale, and environmental controls in the West Siberian Arctic, a region where abrupt permafrost thaw remains poorly characterized.

This is a highly valuable and timely contribution to the field of permafrost research. The study is methodologically robust, clearly structured, and well contextualized within the broader scope of Arctic environmental change. The manual mapping approach, supported by field knowledge and verification, lends high confidence to the dataset, which achieves >90% accuracy in RTS identification. The comprehensive classification scheme—incorporating morphology, terrain position, and associated processes—greatly enhances the utility of the dataset for geomorphological, ecological, and climate-related applications.

Moreover, the dataset's potential as a reference for future remote sensing and machine learning studies is significant. The clarity of the methods, the transparent discussion of limitations, and the open-access nature of the data all reflect strong scientific standards and ensure broad usability. Overall, this is an excellent and much-needed piece of work, and I strongly support its publication in Earth System Science Data.

I have only a few minor observations that may help further improve the manuscript:

• You mention that a preliminary inventory of RTSs was published by Nesterova et al. (2021), but that it likely underestimated RTS distribution due to limited resolution. However, it is not entirely clear whether the RTSs previously mapped in that study were integrated into the current inventory, or if this work represents a completely new and independent mapping effort. If this is a fully new mapping process, it would be helpful to explicitly reference the earlier inventory in the Discussion section and highlight how the present approach offers improved results, particularly in terms of accuracy, resolution, or spatial coverage.
• One notable limitation of the study is that the RTS inventory was compiled using satellite imagery from various years. While it appears that the dataset spans approximately a decade, the exact temporal coverage is not clearly stated. For clarity and transparency, it would be helpful to synthesize all data sources and acquisition years in a concise table. Recent studies have shown that RTS activity in regions like the Yamal Peninsula is highly dynamic. For example, as you mention (line 56), new RTSs have emerged in recent years, and Ardelean et al. (2020) reported that in a small area of the Yamal Peninsula, the number of RTSs increased from 24 to 37 between 2004 and 2012. This raises the question of how reliable and representative a static inventory can be, considering that the number of RTSs may increase by 20% or more after a single warm season in some areas.

While I appreciate that this limitation is acknowledged in the Discussion, it may also be helpful to provide readers with a clearer temporal context for the mapped RTSs. For instance, including the year of observation for each RTS in the database, and possibly summarizing this information in a simple figure (perhaps overlaid in Fig. 4b), could give users a better sense of when most RTSs were identified. This would also strengthen the interpretation of the dataset's temporal relevance and help users better understand its limitations and potential applications.

• Have you observed any particular patterns in the distribution of RTSs in relation to yedoma deposits? The manuscript does not make it clear whether yedoma terrains are more or less favorable to RTS development. While I know that this region may not be particularly representative for extensive yedoma deposits, it could still be helpful to briefly address this aspect in the Discussion. Even a short statement noting the presence or absence of RTS in yedoma areas—or the limitations of assessing this due to their restricted extent—would add value and context for readers interested in the geomorphic controls on RTS formation.
• Lines 44-45, please add a citation.
• Line 48: I couldn`t find Jones et al., 2019 at the Bibliography. I reccomand also to refer to the study by Barth et al. (2023).
• Line 50: Are you sure that all the study area is in continuous permafrost? Acccording to Obu et al (2019) model the southern part of peninsulas are in discontinuous permafrost.
• Line 93: would be good to list the names of peninsula on the map (Figure 3). In addition, would be good to overlap the type of permafrost on Fig. 3 (you can yo use Obu et al., 2019), or at least the limit between continuous and discontinuous permafrost.
• Line 187: coastal is correct here?
• Lines 237-238: these tests should be presented in the Methodology section.
• 13b: in the center of the Gydan Peninsula there is a hexagon with a 20-30 RTS per cell. Can you check if it is correct and there are coastal RTS there?
• In the Introduction you start by saying that RTS are formed due to the thaw of exposed ice-rich permafrost. However, along the study there are no references to the warming of the climate in this area in the last decade or to a temperature increase of permafrost in the region. Would be useful to put the inventory in the climatic context since this phenomenon is climatically controlled and refer to climate evolution in Western Siberia.

References:

Ardelean et al., 2020. doi:10.3390/rs12233999

Barth et al., 2023. https://doi.org/10.1594/PANGAEA.961794

Obu et al., 2019. https://doi.org/10.1016/j.earscirev.2019.04.023

The submitted manuscript, "High-resolution inventory and classification of retrogressive thaw slumps in West Siberia," presents a new and comprehensive inventory of 6,168 retrogressive thaw slumps (RTSs) in the West Siberian Arctic, using high-resolution imagery. The work is scientifically valuable and fills a geographic gap in permafrost monitoring efforts, offering one of the first large-scale, high-resolution inventories from this understudied region. It fills a substantial geographic gap in RTS inventories, especially in a region where abrupt thaw processes are underdocumented.

The classification scheme and the spatial scale elevate the value of this dataset. The overall structure is clear and transparent, and the open-access release will allow broad scientific reuse. Nevertheless, some clarifications and minor improvements would enhance the dataset’s interpretability, reproducibility, and scientific context.

A few minor comments that may help further improve the manuscript:

• Line 84: The inventory provides point data for each RTS, but it is not explained whether this point corresponds to the slump's headwall or geometric center. I would suggest clarifying how point locations were assigned. A consistent rule, such as placing the point at the headwall or initiation zone, would be most relevant for geomorphological or modeling applications.

• Line 103: The dataset uses 3.9 × 3.9 km grid cells for summarizing RTS densities.
  
  Is there a reason this specific resolution was chosen? Does it align with other pan-Arctic products, or was it selected to balance detail and generalization?

• Line 134, you mentioned that the RTS points underwent two visual corrections by the first author. While this allows for consistency, it introduces potential subjectivity in feature recognition and classification thresholds. You mention subjectivity in the limitations, but I wonder if a second opinion would have increased the accuracy. Please acknowledge this as a limitation and suggest the value of inter-observer validation or consensus mapping.

• Line 209: F1 scores vary substantially across validation subsets. While this is valuable transparency, the implications are not clearly discussed. Please add 2–3 sentences interpreting these metrics. For example, what factors drove the lowest scores (e.g., coarser imagery, seasonal effects, older basemaps)? Would you recommend restricting the use of the inventory for automated training to higher-confidence regions?

• Lines 340-345: The manuscript states that some mapped features were likely misclassified RTSs. Are these RTS flagged, removed, or included in the final dataset?

• Lines 346-347: While the manuscript refers to Yedoma, there is no discussion of whether RTSs coincide with yedoma terrain, known to be highly susceptible due to excess ground ice. Maybe add a figure including yedoma distribution and the mapped RTSs.

• Finally, to better contextualize the presented dataset within the growing body of RTS studies, the authors may wish to cite recent advances in RTS.

Ardelean et al., 2020, DOI: https://doi.org/10.3390/rs12233999

Makopoulou et al., 2024, DOI: https://doi.org/10.1002/esp.5890  

Yang et al., 2025 (ARTS Dataset), DOI: https://doi.org/10.1038/s41597-025-04372-7