The paper “The Effects of Late Cenozoic Climate Change on the Global Distribution of Frost Cracking” by Sharma et al. presents the results of three frost cracking models applied on global scale. The authors used published temperature reconstruction in an 80 to 80 km resolution for four time slices:  Pre-Industrial (~1850 CE, PI), Mid-Holocene (~6 ka, MH), Last Glacial Maximum (~21 ka, LGM) and Pliocene (~3 Ma, PLIO) times. These temperatures were used to calculate days spent in the frost cracking window (Anderson, 1998) and two existing frost cracking models (Andersen et al., 2015; Hales and Roering, 2007). The authors analyzed spatial variation of frost cracking intensity (FCI) at individual time slices and observed large deviations of FCI between warmer (PI) and colder (LGM) climate. The paper addresses an important topic about spatial and temporal variation of frost weathering that shapes the Earth’s surface. My major concern is that there is no independent data set that enables the validation of the model results. The authors provide three model results but there is no possibility to review the model results for a location where data exist or for different time slices.

The use of paleodata

The study uses paleo-temperatures, which are air temperatures according to the papers by Mutz & Ehlers (2019) and Mutz et al. (2018) and not land surface temperatures as indicated in this paper. Snow cover and vegetation will result in temperature offsets between air and surface temperatures, which will cause large difference in the frost cracking results and are not addresses in this manuscript at all. The paleo-data is available at 80 to 80 km resolution which is much to coarse to apply these to high-topographic environments as the European Alps, Andes or Tibetan Platea. The coarse resolution is not integrating topographic effects in is not applicable to mountains. The authors should downscale their data, which is a standard procedure in alpine studies (e.g. Fiddes and Gruber, 2014). The data is available at daily time steps and could be used directly to calculated frost cracking. However, the authors calculate a mean annual temperature and half amplitude of annual temperature. They used sinusoidal daily temperatures but it remains unclear if these temperatures are from the paleo-temperatures or assumed values. A more direct use of paleo-temperatures would be better suited.

Frost cracking models

The authors used three proxies or models for frost cracking but only focus on model 3 in their paper. The days spent in the frost cracking window is only a poor proxy for frost cracking  (Anderson et al., 2013). The model by Hales and Roering (2007) is out-dated and not including any lithological differences. Both models are barely used in the results and discussion section, therefore they could be omitted from the manuscript.

The model by Andersen et al. (2015) is applied using soil thickness to constrain a soil layer with a assumed porosity of 30% which is located above a bedrock layer of 2% porosity. The soil thickness is derived from a global database with 5 km resolution and used for every time slice, however, it is unrealistic that soil thickness is a constant over Cenozoic time scales. The substrate classification into soil and bedrock changes water flow in the subsurface within the frost cracking model. For alpine regions the database provides relative high soil depths, however, rockwalls with 30 % porosity are not existing, which highlights the problem of spatial solution and applicability of this model in this way to alpine conditions using a soil map. In addition, the model by Andersen et al. (2015) uses a fixed frost cracking widow between -8 and -3 °C that is not supported by laboratory data (e.g. Murton et al., 2006), field data (Girard et al., 2013) or physical models (Walder and Hallet, 1985). As lithology and rock strength show variations across the Earth, lithology will control weathering, which could be incorporated to include more realistic results.

Glaciation

The authors provide a glacier mask in the supplementary and compare this mask to FCI. The glacier mask is not including any glaciations in the European Alps during LGM or 1850 (Little Ice Age). On which scientific basis is the map derived? Why are the authors comparing the spatial distribution of FCI with their glacier mask? When a glacier is there, then there is no frost cracking as the rock is disconnected to atmospheric processes  (Grämiger et al., 2018). By not including a glacial cover, the authors are overestimating the FCI by far.

Scale issues

The authors use a simple bottom-up approach to model frost cracking for different time slices. They have no independent data that they could use to validate their models. Consequently, they have a problem to discuss their own results and put them into a perspective. They compare a 80 x 80 km model for North America and Alaska for PI, MG and PLIO to a frost cracking studies at Jungfraujoch that measured frost weathering using acoustic emissions on one rockwall at 3500 m for 4 days (Amitrano et al., 2012) or one year (Girard et al., 2013). I cannot see how these studies support the author’s results on much larger scale at different time steps in the past in completely different environments.  Furthermore, the author states that their model results at higher Asia and Alaska during LGM are consistent to periglacial processes observed in Oregon (Marshall et al., 2015; Marshall et al., 2017). I cannot see the context between periglacial conditions and landforms in Oregon and the author’s observed FCI in other areas of the Earth. These are just a few examples but the whole discussion shows no argumentation. Model results will be compared to models from Hales and Roering (2007) or Andersen et al. (2015), which are used to derive the same model results.  

See also detailed comments in the attached pdf.

References

Amitrano, D., Gruber, S., & Girard, L. (2012). Evidence of frost-cracking inferred from acoustic emissions in a high-alpine rock-wall. Earth and Planetary Science Letters, 341, 86-93. doi:10.1016/j.epsl.2012.06.014

Andersen, J. L., Egholm, D. L., Knudsen, M. F., Jansen, J. D., & Nielsen, S. B. (2015). The periglacial engine of mountain erosion - Part 1: Rates of frost cracking and frost creep. Earth Surf. Dynam., 3(4), 447-462. doi:10.5194/esurf-3-447-2015

Anderson, R. S. (1998). Near-surface thermal profiles in alpine bedrock: Implications for the frost weathering of rock. Arctic and Alpine Research, 30(4), 362-372. doi:10.2307/1552008

Anderson, R. S., Anderson, S. P., & Tucker, G. E. (2013). Rock damage and regolith transport by frost: an example of climate modulation of the geomorphology of the critical zone. Earth Surface Processes and Landforms, 38(3), 299-316. doi:10.1002/esp.3330

Fiddes, J., & Gruber, S. (2014). TopoSCALE v.1.0: downscaling gridded climate data in complex terrain. Geosci. Model Dev., 7(1), 387-405. doi:10.5194/gmd-7-387-2014

Girard, L., Gruber, S., Weber, S., & Beutel, J. (2013). Environmental controls of frost cracking revealed through in situ acoustic emission measurements in steep bedrock. Geophysical Research Letters, 40(9), 1748-1753. doi:10.1002/grl.50384

Grämiger, L. M., Moore, J. R., Gischig, V. S., & Loew, S. (2018). Thermo-mechanical stresses drive damage of Alpine valley rock walls during repeat glacial cycles. Journal of Geophysical Research: Earth Surface, 123(10), 2620-2646. doi:10.1029/2018JF004626

Hales, T. C., & Roering, J. J. (2007). Climatic controls on frost cracking and implications for the evolution of bedrock landscapes. Journal of Geophysical Research-Earth Surface, 112, F02033. doi:10.1029/2006jf000616

Marshall, J. A., Roering, J. J., Bartlein, P. J., Gavin, D. G., Granger, D. E., Rempel, A. W., Praskievicz, S. J., & Hales, T. C. (2015). Frost for the trees: Did climate increase erosion in unglaciated landscapes during the late Pleistocene? Science Advances, 1(10). doi:10.1126/sciadv.1500715

Marshall, J. A., Roering, J. J., Gavin, D. G., & Granger, D. E. (2017). Late Quaternary climatic controls on erosion rates and geomorphic processes in western Oregon, USA. Geological Society of America Bulletin, 129(5-6), 715-731. doi:10.1130/B31509.1

Murton, J. B., Peterson, R., & Ozouf, J. C. (2006). Bedrock fracture by ice segregation in cold regions. Science, 314(5802), 1127-1129. doi:10.1126/science.1132127

Walder, J., & Hallet, B. (1985). A Theoretical-Model of the Fracture of Rock During Freezing. Geological Society of America Bulletin, 96(3), 336-346. doi:10.1130/0016-7606(1985)96<336:ATMOTF>2.0.CO;2

This manuscript tackles an interesting and understudied question – where and when frost cracking has influenced bedrock weathering on a global scale during the late Cenozoic. The authors have done an impressive job of implementing and testing different existing models for frost cracking. My main concern is that some of their conclusions seem circular or do not add much to the original publications. I think new knowledge can be gained by critically investigating the predictions of frost cracking intensity arising from different assumptions, and that the authors are well-placed to do so here. In the following I therefore suggest a few suggestions to restructure the manuscript.

Main comments

1. The assumption of soil cover being comparable in the Pliocene is likely most heavily violated in regions that experienced Pleistocene glaciation. I suggest removing these areas with a ‘maximum Pleistocene ice extent mask’ from the Pliocene results, similarly to what is also done for the LGM. Both masks should be highlighted in another color than the background grey on Fig. 6 to make it more apparent. It is important to also show them on the FCI difference maps for the relevant time-slices (Fig. 7-10), so you don’t compare FCI for regions within and outside of ice-sheets for different time slices. The latter would also eliminate the problem with sentences in e.g., line 306-307 and 317-318, where you seem to be unsure about whether FCI-differences result from ice cover during LGM or not.
2. Your soil thickness data seem to saturate/max out at ~1 m (fig. 1). Are these minima estimates? It is not clear to me how you handle soils >= 1 m in the FCI model, or if you exclude these (extensive) regions. I would also like to see a discussion on how the uncertainties and coarse spatial resolution of the soil data may influence the modelled FCI on a sub-grid scale.
3. I fail to see the relevance of the comparison to permafrost extent, and suggest cutting these sections out of the paper (Sec. 3.3, Sec 5.4.2, and Fig. 11+12).
4. The discussion includes a number of ‘predictions’ that ‘confirms’ or ‘agrees with’ the models (e.g., line 287-288, line 429). These statements appear circular since your results are based on the same models, and does not really add anything new compared to reading the original papers. I suggest that you spend more space on comparing the models and testing the effect of the underlying assumptions in the main paper. For example: it is disputed whether the penalty functions that make FCI depend on distance to water give a better representation of the frost cracking process or not. Since you have gone through the trouble of implementing all three models, it would be interesting to use them to evaluate what predictions about global frost cracking the different choices result in. For example, you could test the effect of the penalty function by running models with and without the postulated influence on FCI, but maintaining the influence of porous (wet) soil on the temperature-profiles. Similarly, your section 5.4.1. would be better framed as an evaluation/discussion of the assumptions behind the different models, rather than evaluating your results directly. At present this section does not really add something new that is not in the original papers, which is why everything ends up being in agreement with your results.
5. Section 4 and 5.1-5.3 and Fig. 7-10. These sections are rather long and hard to read. Please try to condense the most important lessons. I suggest referring to Fig. 8-10 as part of your global discussion in replacement of Sec. 5.3. Perhaps you could even consider showing the FCI-difference panels grouped by time-slice (e.g., PI-MH) instead of region, and then showing the regional details as sub-panels (b-d) to each global model (a) in each figure. This would also reduce the number of figures by one.
6. Consider calculating a globally summed FCI to highlight what periods frost cracking globally are more important to surface processes.

Figures, Tables

Table 2: Too many digits?

Fig. 11+12: Stippled black line very hard to see.

Line specific comments

l. 20: ‘In contrast’ – these sentences does not really contrast

l. 26: Consider removing ‘long ()’

l. 29-31: Not so clear why vegetation is considered indirect, but other surface processes are considered direct.

l. 70: ‘Europe’ is not an orogen

l. 74: ‘and soil’ read strange here

l. 96-98: I don’t understand the reference soil depth information – consider to cut it unless it is relevant enough to explain in more detail?

l. 145: I would not say more complete, but certainly more complex. It is disputed whether the penalty functions (or the ad hoc choice of parameter values in different media) give a better representation of the frost cracking process than a simpler model.

l. 181-182: This seems redundant. No need to mention it in each section, and twice in this section (also l. 197).

l. 186: In the case ‘of’ permafrost

l. 187: Fig. 3, not 2

l. 275: discussion ‘of’ regional variations

l. 484: frost cracking ‘occurs’ at lower latitudes