This study highlights the importance of seasonality in permafrost dynamics. I think that this is a good paper, including clear logic and professional writing. The results will be important for permafrost research. My comments are the following.

L12, some readers may be interested in how annual mean and seasonal amplitudes changes, here.

L104, what is the deepest soil column of simulation for CryoGridLite permafrost model?

L37, “Brierley et al. (2020); Otto-Bliesner et al. (2021)” should be (Brierley et al., 2020; Otto-Bliesner et al., 2021).

L170, the authors say that speleothem growth suggests absence of permafrost here, but in L180 the authors say that locations with speleothem growth is agreement with the model if permafrost probability <90%. The former and the latter appear to be contradictory.

Figure 4, remove a “on the”, it is repeated.

Figure 5, I suggest to add the significance of these correlations.

Lines 240-245, the phenomenon is clear, i.e., mean temperature control the permafrost area rather than seasonal temperature amplitude. However, what are the physical explanations?

L320 and L324, what is the difference between “a mean global surface warming above recent conditions” and “MH-PI global temperature anomaly”?

L467-471, here, I suggest to state that whilst the most recent interglacial climates providing less analogues for the future with respective to permafrost dynamics, this do not exclude that the past older warming period may be appropriate. For instance, the most simulation study (https://doi.org/10.1073/pnas.2301954120) on the mid- Pliocene warm period (mPWP, ~3.264 to 3.025 Ma) permafrost. During the mPWP, the temperature increase significantly in both winter and summer. The period also evolves regional differences in warming, in particular in the high latitudes. These are similar to the future warming. So, mPWP permafrost remains one of the analogs for the future permafrost dynamics, and likely the results (highly restricted extent) has implications for the future.

Summary 

This study explores how both seasonal and annual mean temperature variations influence permafrost dynamics during past interglacial periods, specifically the Mid Holocene (MH) and Last Interglacial (LIG). The authors employ simulations from the AWI-ESM-2.5 Earth System Model in conjunction with the CryoGridLite permafrost model to analyze ground thermal characteristics such as permafrost extent, thaw depth, and thermal contraction cracking. The key finding is that permafrost extent is primarily controlled by annual mean temperatures, whereas thaw depth and thermal contraction cracking are predominantly influenced by the amplitude of seasonal temperature variations. Importantly, the study concludes that past interglacial climates are not suitable analogues for projecting future permafrost thaw, as modern climate change trends involve rising mean temperatures coupled with decreasing seasonality, a pattern not observed in the past.



Major Comments:

1. Missing Literature on Borehole Temperature Reconstructions: The manuscript omits an important body of work related to reconstructing past climate from borehole temperature data. Borehole climatology offers critical insights into ground surface temperature (GST) changes over the past centuries and millennia. For instance:

• Liu et al. (2021) applied Tikhonov regularization to infer ~1.8°C of ground warming over the past ~308 years, aligning with instrumental records.
• Mareschal & Beltrami (1992) and Beltrami & Mareschal (1991) pioneered SVD-based inversions of borehole data, identifying historical warming in regions like eastern Canada.
• Shen & Beck (1991) used least-squares inversion with regularization for stable GST reconstructions.
• Pollack & Huang (2000) presented a global borehole climatology, documenting multi-century surface warming. These studies offer a complementary line of evidence that should be discussed to provide a more complete context for past permafrost dynamics.

2. Clarification of Paleo Proxy Records: A paragraph introducing and defining speleothem and pollen records is needed for readers unfamiliar with these proxies. While valuable, these records are spatially limited and prone to interpretive uncertainty, which affects their robustness in validating model outputs.

3. Model Bias and Alternatives: The AWI-ESM-2.5 model has a documented cold bias in high-latitude regions, which likely leads to an overestimation of permafrost extent. The authors should discuss whether other Earth System Models (e.g., CESM, MPI-ESM) have been considered for comparison, especially those with better temperature performance at high latitudes.

Minor Comments:

P2, Line 50: The introduction notes that few models address permafrost conditions in past climates. The conclusion should explicitly state how this study advances that field and relates to previous work.

Figure 3d: Why is there no simulated permafrost in the southwestern part of Russia? A similar issue arises in Quebec, Canada—please clarify.

Figure 4: A consistent colorbar across all panels would improve clarity. What does the current colorbar represent?

Figures: Expand abbreviations like PI (Pre-Industrial), LIG (Last Interglacial), etc. Also, labels like "ΔTa" need more descriptive titles.

P3, Line 60: Clarify why focusing on both the current and last interglacial periods is important, particularly in relation to temperature amplitude.

Cracking Quantification: Explain in more detail how thermal contraction cracking is quantified in the model.

P3, Line 80: Are glacier masks, topography, and related inputs dynamic in the model, or are they static? How might this affect the results?

P5, Line 108: If dynamic hydrology were included in CryoGridLite, how might this change the study’s findings?

P6, Line 132: The text mentions subsidence. Was uniform subsidence assumed across the region, or was spatial variability considered?

P8, Line 192: I suggest that important locations discussed in the results, such as the Nahanni Plateau, be marked directly on the maps for clarity.

Figure 7: Needs a more detailed and explanatory caption.

Comparison Figures (multiple subplots): Consider presenting differences between simulations for a clearer quantitative comparison. Matching areas could be quantified in percentage terms.

Conclusions: The conclusion should be restructured to directly reflect and address the key objectives and issues raised in the introduction, creating a more cohesive narrative.



References: 

1. Liu et al. (2021). Application of Tikhonov regularization to reconstruct past climate record from borehole temperature. Inverse Problems in Science and Engineering, 29(13), 3167–3189. https://doi.org/10.1080/17415977.2021.1975700
2. Mareschal, J.-C., & Beltrami, H. (1992). Evidence for recent warming from perturbed geothermal gradients: examples from eastern Canada. Climate Dynamics, 6, 135–143.
3. Beltrami, H., & Mareschal, J.-C. (1991). Recent warming in eastern Canada inferred from geothermal measurements. Geophysical Research Letters, 18, 605–608.
4. Shen, P.Y., & Beck, A.E. (1991). Least squares inversion of borehole temperature measurements in functional space. Journal of Geophysical Research: Solid Earth, 96(B12), 19965–19979.
5. Pollack, H.N., & Huang, S. (2000). Climate reconstruction from subsurface temperatures. Annual Review of Earth and Planetary Sciences, 28(1), 339–365.
6. Bodri, L., & Čermák, V. (2011). Borehole climatology: A new method how to reconstruct climate. Amsterdam: Elsevier.