In this study, Könönen and colleagues build a dataset on peatland palsa and peat plateau occurrence, then extract environmental variables from gridded datasets to determine the climate space conducive to palsas. Then they use statistical modeling to predict the area of peatland palsas, and how this area will change in the future under changing climate. The question is interesting, if not particularly novel (see Halsey et al., 1995, Fewster et al. 2020, 2022). The main distinction here is that the authors expand on this from a regional analysis to a pan-Arctic analysis, which is a factor that dooms this manuscript in it’s current form because of un-even spatial coverage in their dataset. The results that peat plateaus and palsas will disappear with climate change isn’t especially new (see Halsey et al., 1995, Beilman et al., 2001, Tarnocai 2006, Camill & Clark 1995, Camill 2005), but the scope is interesting and alarming.

I think in it’s current form, the manuscript is unfocused as well as missing key information and methodology that prohibit the evaluation of the results and conclusions. The discussion is superficial and doesn’t address key uncertainties and other factors that would affect the results. For the focus, the majority of the methods and half of the conclusions are focused on present day conditions. Only one short section in the results section is and the discussion about this is superficial. There is no section explaining how the forward projections were done and what data was used beyond the scenario names (and they are from AR5).

The most crucial information, the dataset about the palsas, is missing (cited as Appendix A). The information is not in the paper, is not already openly or provisionally available in an online repository (e.g. Zenodo or Pangaea or a Uni Helsinki repository), or in the supplementary materials. This is problematic because from what I see, the model results for North America seem to be particularly biased towards where there are samples or not. I think the data coverage is exceptional for Fennoscandia but really limited for North America. The authors don’t explore the representativeness of the dataset that they’ve collected towards earlier described or known inventories of permafrost peatland areas or palsa areas or peatland areas. This could be normalized and some confidence assigned based on the number of samples per peatland area. I thought that the results look really biased towards where there are samples, and quite limited outside that as I would expect a lot more coverage in Northwestern Canada (Alberta). Other areas that don’t look right to me, in particular a high chance for peat plateaus and palsas on the North Slope of Alaska and on the Seward Peninsula where the dataset is quite limited. To my knowledge, the peatland area is relatively limited on the North Slope and often limited to riverine systems and not so frequently peat occurence. Then further north, mostly polygonal tundra peatlands are found or no peatlands at all. As mentioned in the discussion, interior Alaska would make sense. I am also aware of some other large peatland datasets that could either be incorporated and referenced or used for validation (Treat et al., 2016, Table S1) as a presence/absence marker or a newer dataset from Olefeldt (BAWLD) that tackles this more directly (see “Permafrost Bog”). They show much more extensive permafrost bog coverage in northwestern Canada, for example.

My other major concern in this study was the approach for validation. Not much space or effort was dedicated towards convincing me as a reviewer that this approach worked (and from my visual inspection of Figure 2 above, I’m not convinced) and what the results would actually represent. Is this potential permafrost palsa are or actual permafrost palsa area? The model validation isn’t presented in the results section, or really at all. The closest we come is Table 1, giving the areas by region and the change over time. Comparisons about the spatial distributions are made to Fewster, but the areas are not put into context of peatland areas, permafrost peatland areas, regional permafrost peatland areas, or other independent datasets so it is difficult to glean if these area estimates are even reasonable. I would expect to see a table comparing the areas predicted in this study to other estimates from Webster 2018, Hugelius et al. 2020, Olefeldt et al., and Fewster. Finally, the real trick in these discontinuous permafrost environments is to separate permafrost that shouldn’t be there in today’s climate envelope and is only there because it formed under colder climate conditions and non-permafrost, as would happen at the exact southern edge of the permafrost environment (e.g. when there are paired cores at a site, one with permafrost, one without and commonly found in palsa regions of Fennoscandia and Canada). My understanding in this is that these samples would be excluded because they are too close in location, which might limit the accuracy of this approach in the most sensitive regions. If this is only change in potential permafrost palsa area with no distinction between where there is permafrost and not, this really limits the utility of the whole analysis for future predictions.

Specific comments:

15-16: why -98.2 and 89.2 loss? Signs don’t match?

25: Wang 2022 is missing from refs or wrong reference.

34: see additional refs above.

33: Hugelius has additional peatlands included in permafrost peatlands, not necessarily plateaus and palsas.

49: see additional refs above

51: yes, Siberia is a big unknown!

57: what about snow? Wind? Trees?

65: why not some areas where palsa thaw is observed?

84: appendix not found

90: Check Treat et al. 2016

103: but this is the crucial distinction.

Figure 1: coordinates for Kiruna leave me in a lake. Would be helpfult to see the traditional permafrost map from Brown also for reference.

110: this is all present day, good to acknowledge role of past climate and the limitations this presents.

136: why? what was the goal or motivation of this?

Section 2.3: for all these paragraphs, WHY? What specifically was the purpose or goal of this analysis, what did you actually do? Where are the forward projections?

139: add citation for biomod2 package

3.1 model evaluation of what?  Why don’t discuss suitable environments already? Where is the evaluation of the representativeness of the dataset? Or independent evalution? What about normalizing for areas? Also I’ve never heard of palsas in Iceland.

Figure 2. The results looks really biased towards where there are samples.

Figure 3: y-axis labels

Section 3.4 doesn’t provide any real (independent) model evaluation, it only compares the different techniques used.

382: Yes, it is good that the model found palsas where the input data indicated there should be palsas. But it also found them where there is little evidence for palsa.

Data availability: The dataset should be provided with DOI not upon contact to author, especially since it is listed as Appendix A..

References

Beilman, D. W., D. H. Vitt and L. A. Halsey (2001). "Localized permafrost Peatlands in Western Canada: Definition, distributions, and degradation." Arctic Antarctic and Alpine Research 33(1): 70-77.

Camill, P. (2005). "Permafrost thaw accelerates in boreal peatlands during late-20th century climate warming." Climatic Change 68(1-2): 135-152.

Camill, P. and J. S. Clark (1998). "Climate change disequilibrium of boreal permafrost peatlands caused by local processes." American Naturalist 151(3): 207-222.

Fewster, R. E., Morris, P. J., Swindles, G. T., Gregoire, L. J., Ivanovic, R. F., Valdes, P. J., & Mullan, D. (2020). Drivers of Holocene palsa distribution in North America. Quaternary Science Reviews, 240, 106337.

Fewster, R. E., Morris, P. J., Ivanovic, R. F., Swindles, G. T., Peregon, A. M., & Smith, C. J. (2022). Imminent loss of climate space for permafrost peatlands in Europe and Western Siberia. Nature Climate Change, 12(4), 373-379.

Halsey, L. A., D. H. Vitt and S. C. Zoltai (1995). "Disequilibrium Response of Permafrost in Boreal Continental Western Canada to Climate-Change." Climatic Change 30(1): 57-73.

Hugelius, G., J. Loisel, S. Chadburn, R. B. Jackson, M. Jones, G. MacDonald, M. Marushchak, D. Olefeldt, M. Packalen, M. B. Siewert, C. Treat, M. Turetsky, C. Voigt and Z. Yu (2020). "Large stocks of peatland carbon and nitrogen are vulnerable to permafrost thaw." Proceedings of the National Academy of Sciences 117(34): 20438-20446.

Olefeldt, D., M. Hovemyr, M. A. Kuhn, D. Bastviken, T. J. Bohn, J. Connolly, P. Crill, E. S. Euskirchen, S. A. Finkelstein, H. Genet, G. Grosse, L. I. Harris, L. Heffernan, M. Helbig, G. Hugelius, R. Hutchins, S. Juutinen, M. J. Lara, A. Malhotra, K. Manies, A. D. McGuire, S. M. Natali, J. A. O'Donnell, F. J. W. Parmentier, A. Räsänen, C. Schädel, O. Sonnentag, M. Strack, S. E. Tank, C. Treat, R. K. Varner, T. Virtanen, R. K. Warren and J. D. Watts (2021). "The Boreal–Arctic Wetland and Lake Dataset (BAWLD)." Earth Syst. Sci. Data 13(11): 5127-5149.

Tarnocai, C. (2006). "The effect of climate change on carbon in Canadian peatlands." Global and Planetary Change 53(4): 222-232.

Tarnocai, C. and S. Zoltai (1988). Wetlands of Arctic Canada. Wetlands of Canada. N. W. W. Group. Ottawa, Sustainable Development Branch, Environment Canada: 28-53.

Treat, C. C., M. C. Jones, P. Camill, A. Gallego-Sala, M. Garneau, J. W. Harden, G. Hugelius, E. S. Klein, U. Kokfelt, P. Kuhry, J. Loisel, P. J. H. Mathijssen, J. A. O'Donnell, P. O. Oksanen, T. M. Ronkainen, A. B. K. Sannel, J. Talbot, C. Tarnocai and M. Väliranta (2016). "Effects of permafrost aggradation on peat properties as determined from a pan-Arctic synthesis of plant macrofossils." Journal of Geophysical Research: Biogeosciences 121(1): 78-94.

Treat, Claire C; Jones, Miriam C; Camill, Philip; Gallego-Sala, Angela V; Garneau, Michelle; Harden, Jennifer W; Hugelius, Gustaf; Klein, Eric S; Kokfelt, Ulla; Kuhry, Peter; Loisel, Julie; Mathijssen, Paul J H; O'Donnell, Jonathan A; Oksanen, Pirita O; Ronkainen, Tiina M; Sannel, A Britta K; Talbot, Julie; Tarnocai, Charles; Väliranta, Minna (2016): Synthesis dataset of physical and ecosystem properties from pan-arctic wetland sites using peat core analysis. PANGAEA, https://doi.org/10.1594/PANGAEA.863697,

Webster, K. L., J. S. Bhatti, D. K. Thompson, S. A. Nelson, C. H. Shaw, K. A. Bona, S. L. Hayne and W. A. Kurz (2018). "Spatially-integrated estimates of net ecosystem exchange and methane fluxes from Canadian peatlands." Carbon Balance and Management 13(1): 16.

Zoltai, S. C., C. Tarnocai, G. F. Mills and H. Veldhuis (1988). Wetlands of Subarctic Canada. Wetlands of Canada. N. W. W. Group and C. C. o. E. L. Classification. Montreal, Polyscience Publication Inc.: 54-96.

Zoltai, S. C., R. M. Siltanen and J. D. Johnson (2000). A Wetland Data Base for the Western Boreal, Subarctic, and Arctic Regions of Canada. Edmonton, AB, Canada, Northern Forestry Centre, Canadian Forest Service: 30 pp.

- Claire Treat, Alfred Wegener Institute for Polar and Marine Research

The manuscript "Environmental spaces for palsas and peat plateaus are disappearing at a circumpolar scale" by Könönen et al. presents statistical modeling of the future distribution of the palsa/peat plateau landform, with the key result that the largest part of todays palsas/peat plateaus will move outside the suitable climate space for some of the future scenarios considered. The study is well-designed and well-written, so I recommend it for publication after revisions.

-Introduction: the authors should make an attempt to clarify the terms "peat plateau" and "palsa", or state that they made no clear distinction between the two (since there are many transitional types, for example). In some parts of the text, it appears that the authors do distinguish between them. They for example only refer to palsas in Scandinavia (l.48, and actually continue only using palsa to l. 65), but there are many plateau-like structures in Scandinavia, which published studies have referred to as peat plateaus. They also cite studies claiming that “peat plateaus are only 1m high” (l. 30), but there are much higher plateau-like features in many peatlands with permafrost.

-L. 84: same here, how do the authors distinguish a “true” palsa?

-L. 87: were high enough resolution images available on Google Earth for all of the sites?

-L. 123: These data sets can have a poor quality, especially at the resolution needed for the study setup. Please comment on this already in the Methods.

-Eq. 1: shuffled

-L. 188: it is easier to read if the authors spell out Random Forest (RF) once more at the beginning of the Results section.

-Fig. 3: spell out the abbreviations in the figure caption.

-L. 237: MAJOR COMMENT: Maybe I have overlooked it, but the authors should clarify if they only consider areas that are in the suitable palsa space now AND in the future as “remaining areas”, or if they also count areas that are not suitable now (e.g. too cold in summer), but will become suitable in the future. As shown e.g. in the works of Seppälä, one needs permafrost-free conditions in the vicinity of emerging palsas for this process to work. So areas which today are too cold for palsas to exist may not develop palsa landforms for a long time (or never), even if they move to the suitable space (for example continuous permafrost first needs to thaw, etc.) On the other hand, areas with palsas today will be preserved if they still fall in the suitable space in the future. I think it is worth to make this distinction and possibly present the numbers for both cases. The wording in this section is incoherent, the authors use “remain” and “persist” in some cases, and “could be found” in others.

-Sect. 3.3 The comparison with the thermokarst map is the weakest part of the study, e.g. it is unclear to what extent some of the data sets used by Olefeld et al. may have been similar to the data sets used for this palsa mapping. Furthermore, it is unclear if one should expect a close match or not, given the methods used by Olefeld (i.e. were the specific thermokarst conditions of palsas/peat plateaus accounted for in this work?). I leave it to the authors to decide, but I am not learning anything from this comparison, it more dilutes the very nice results from the previous section with a poorly motivated add-on.

-L. 308: Extrapolate instead of present?

-L. 360: MAJOR COMMENT: this is an extremely important point that needs to be discussed in much more detail. Some of the studies cited, e.g Borge, provide indications that peat plateaus were already degrading more than 50 years ago, so they may have left the suitable climate space already after the end of the Little Ice Age (or in the 1990s, when exactly is unclear and certainly depends on the exact location), but the degradation is slow so that it takes decades or even centuries to complete. However, these areas were still used by the authors for training their model. With the simple analysis and no means of telling which palsa areas are stable and which are degrading already now, I don’t think that this can be taken into account, but it is a limitation that should be stated clearly. Furthermore, given this complexity, I strongly disagree with the statement “support the rapid degradation of the landforms.“ First, with rates of 1% per year, it will take longer than the 2080 timeframe considered by the authors for the palsas/peat plateaus to actually disappear, in any case several decades. Second, it is very likely that palsas degrading rapidly already now, e.g. in the ones in Scandinavia, will indeed disappear until 2080, but palsas in areas that are still largely stable today might only be pushed just outside the suitable climate envelope and only then start to degrade slowly (similar to the ones in Scandinavia after the LIA). So we might expect palsas and peat plateaus to exist there much longer than 2080, although they are outside the suitable climate envelope. 

I very much like that the authors use the wording “inside/outside the suitable climate envelope/environmental space” throughout the manuscript and do not refer to their study as a model for palsa degradation. But it is important to clarify and discuss this relationship in much more detail, so this section of the discussion should be extended. In particular, the authors should point out that not all palsas inside the suitable climate envelope are equal, but palsas on the “warm side” of the envelope likely degrade earlier and more rapidly, while the ones on the "cold side" might persist for many more decades.