| 16 Jun 2026
Simulating the permafrost thermal regime in the Northwestern Antarctic Peninsula from 1950 to 2100
Abstract. Permafrost underlies most of the Antarctic ice-free areas, being crucial for terrestrial ecosystems, influencing pedogenesis, hydrology, geomorphic dynamics, and the carbon biogeochemical cycle. However, uncertainty in the evolution of permafrost temperature is particularly relevant in the Antarctic Peninsula, a climatic hotspot of the continent where an increase in mean annual air temperature of 3.4 ± 1.2 °C has been recorded, together with a rise in both the frequency and intensity of warm-weather episodes. The impact on permafrost is difficult to foresee, due to a scarce monitoring network implemented following the International Polar Year in 2007–2008, with limited temporal coverage that constrains trends evaluation. In this study, we simulate past and future permafrost temperature evolution in the Northwestern Antarctic Peninsula using the CryoGrid Community Model at five permafrost observatories from the University of Lisbon’s network (PERMANTAR). Simulations forced with ERA5 reanalysis reconstruct ground temperature evolution since 1950, revealing a warming trend at all depths, with mean annual ground surface temperature, temperature at the top of permafrost, and mean annual ground temperature warming at rates between 0.19 and 0.29 °C dec⁻¹. Four distinct periods are identified since 1950: early sustained warming (1950–1975), highly variable warming and cooling (1975–2000), a short cooling period with increased snowfall (2000–2015), and intense warming after 2015 that accelerates permafrost temperature increases at 20 m depth, resulting in mean annual ground temperatures above -2 °C. Future projections using the ACCESS‑CM2 model under SSP1‑2.6, SSP2‑4.5, and SSP5‑8.5 reveal progressive permafrost degradation and a widespread transition to positive mean annual ground temperatures at 20 m depth during the 21st century, with timing and magnitude strongly influenced by elevation, snow cover duration, and local changes in ocean–atmosphere interactions. In the SSP1-2.6, warming rates ranging from 0.21 ± 0.01 to 0.23 ± 0.01 °C dec⁻¹ are predicted for the sites, resulting in extensive permafrost degradation at low altitude. Under SSP2‑4.5, all sites are projected to develop positive mean annual ground temperatures at 20 m between 2049 and 2087, with warming rates ranging from 0.31 ± 0.01 to 0.43 ± 0.01 °C dec⁻¹, implying increased permafrost degradation and increased susceptibility of coastal terrestrial ecosystems in the Northwestern Antarctic Peninsula to climate-driven change. Under the SSP5-8.5, with warming rates between 0.46 ± 0.01 and 0.62 ± 0.01 °C dec⁻¹, the degradation of permafrost is predicted to occur more rapidly, with positive mean annual ground temperature values at 20 m reached between 2045 and 2071. The projected warming and associated permafrost loss will change surface and subsurface hydrology, biochemical fluxes and geomorphological processes, impacting the sensitive terrestrial ecosystems.
This is an important and timely contribution on the recent past and future ground temperature dynamics in the Northwestern Antarctic Peninsula. Not only is there no similar study to date, but the numerical modelling approach used is novel in Antarctica. However, I have several comments and recommendations that should be addressed before the manuscript can be considered for publication.
(1) The simulation of recent past conditions was based on the bias-corrected ERA5 for 1950–2020 (with 10 spin-up loops in 1940–1949). By contrast, the simulation of future conditions was based on the bias-corrected CMIP6 models for 1850–2100, with 10 spin-up loops in 1850–1860 and bias-corrected ERA5 used in 1940–2022. Why did not you apply the modelling strategy used for future to recent past as well? The 100-year model initialization phase running since 1850 would likely be better than the 10 spin-up loops over 1940–1949. Moreover, this would mean better consistency and less work since you would have single time series covering the whole period 1850–2100 and both recent past and future scenarios.
Additionally, was the CMIP6 bias correction based on the bias-corrected ERA5? This is unclear to me.
(2) Considering that you did the bias correction only in the case of air temperature, while the rest of the variables retained their uncorrected values, is the resulting dataset internally consistent? For instance, if the air temperature was corrected, while the dew point temperature remained uncorrected, this may affect outputs calculated from these two variables. Likewise, precipitation retained its uncorrected value, but solid and liquid precipitation usually occurs at some specific temperatures, which can also be affected if the temperature is corrected. I have no idea how this might affect the simulation outputs, but the issue is here obvious.
(3) In the methods, explain why you modelled ground temperatures specifically at a depth of 20 m. Frequently, you also present ground temperatures from a depth of 10 m, which is much closer to a maximum depth of the boreholes and might therefore have a higher validity. Was this because that is at or below the depth of zero annual amplitude?
(4) For consistency, descriptions of both recent past and future simulations should present the evolution of mean annual air temperatures, mean annual ground surface temperatures, temperatures at the base of the freeze-layer layer (TTOP under permafrost conditions), and mean annual ground temperatures at a depth of 20 m. Some of these temperature metrics are frequently missing, which makes a little sense in terms of consistency and (not only) mutual context. If mean annual ground temperatures at a depth of 10 m are shown as well, I would present them before those from 20 m. Because of heat diffusion, the temperature patterns at 10 m precede those at 20 m, so it would have more logic to proceed in this order.
(5) The discussion is relatively short and in principle deals with results of this study alone, and the context of other studies is provided to a very limited extent. While I am aware that there is currently no analogous study published for Antarctica, there are several studies that assessed short-term trends in permafrost temperatures or active-layer thickness (e.g., Guglielmin and Cannone, 2012; Ramos et al., 2017; Biskaborn et al., 2019; Hrbáček and Uxa, 2020; Hrbáček et al., 2021, 2023, 2025; Baptista et al., 2025) or some several long-term meteorological series (e.g., Oliva et al., 2017). I do not understand why you do not mention your previous study in the discussion (Baptista et al., 2025).
What I also miss in the discussion is a section devoted to the evaluation of the ERA5 and CMIP6 models used as model forcings. Consequently, I would move the Appendix C into the discussion. You may also refer to Baptista et al. (2025) or Kaplan Pastíriková et al. (2025) when evaluating the suitability of ERA5 or reananalyses for permafrost studies in Antarctica in general.
(6) Consider carefully whether the content of the appendices really needs to placed in them. I think that most of the appendices could be involved in the main body of the manuscript without making it fragmented.
Specific comments:
P1L13: State the period over which the mean annual air temperature increase of 3.4±1.2 °C was recorded.
P1L18: Specify the depth for the mean annual ground temperature. Is it 20 m?
P1L20: In contrast to the above, it is unnecessary to specify the depth of 20 m here because it is a general statement.
P2L42–43: Again, state the period over which the mean annual air temperature increase of 3.4±1.2 °C was recorded. Considering that mean annual air temperature can exhibit substantial variations from year to year, is it appropriate to use a single year as a baseline value?
P2L50: Change (Obu et al., 2020) to Obu et al. (2020).
P2L57–58: In the aims of the manuscript, be specific and clearly define the time frames for both the recent past and the future modelling periods.
P3L61–62: This previous study (Babtista et al., 2025) should be briefly described before the aims of the manuscript, so that the objectives follow logically from it. I think that, in this case, it disrupts the structure of the paragraph devoted to the aims.
P3L76: Change (Thomas e Tetzner 2019) to (Thomas and Tetzner 2019).
P4L96: Considering the substantial warming trend, the mean annual air temperature could be reported for a shorter, more recent period to better represent current climatic conditions. Of course, this depends on your intentions.
P5L100–101: Rewrite the sentence “… the highest areas of Hurd Peninsula were deglaciated at 20 ka with most high interfluves becoming ice-free around 14-16 ka …” so that it is clear when these locations were actually deglaciated. Was it at 20 ka or around 16-14 ka?
P5L101: Change 14-16 ka to 16-14 ka.
P5L106: State the period for which the mean annual air temperature and mean annual precipitation are reported.
P5L115–116: State the period for which the mean annual air temperature and mean annual precipitation are reported.
P5L124–125: State the period for which the mean annual air temperature and mean annual precipitation are reported.
P6L127–140: Move the whole section 3.1 to the section 2.2. The details on permafrost temperature and active-layer thickness at individual sites can be presented in the sections 2.2.1 to 2.2.4.
P7L151–152. This sentence on the calibration strategy is difficult to understand. Rewrite it even though it is partly obvious from the following text.
P7L156: What is the geothermal heat flux of 50 mWm^-2 based on? Please cite the source.
P7L157: What does the phrase “With the structure set, …” mean?
P7L160: Is the supplement “… using a probe.” necessary?
P7L164: What do you mean by “relative volume fraction” here?
P7L171: Indicate that Fub(t) is the energy flux into the uppermost grid cell. The rest of the variables (radiation and heat fluxes) would be better to describe after Eq. (1).
P7L175: “The heat conduction based on Fourier’s law is the main mode of heat transport in the subsurface.” is too brief description of the subsurface heat transfer.
P8L183: Change “phenomenologically” to “phenomenological”.
P8L186: State also that you calculated p-value.
P9L219: I guess there should be “SPPs” in “The RCPs are used to represent 4 pathways…”
P10L230: Note that the SPP5-8.5 scenario is now considered an extreme case that is rather implausible. You should point this out in the manuscript (here and in the discussion).
P10L235: Remove the word “models”.
P10L243–245: Place this paragraph at the end of the whole section 3.3.2.
P10L253: I would remove “…in the selected PERMANTAR monitoring sites.” because it is redundant here.
P11L258: The vertical resolution of the model is unclear from this: “… from 0.05 to 0.5 m, below 5 m depth.” Revise it.
P11L259: Specify the depths at which you calculated the mean annual ground temperature.
P11L266–269: Rewrite this so it is more clear that you bias-corrected the period 1850–2100 from CMIP6 using the statistical relationships with ERA5 for the sub-period 1940–2020.
P12L300: Remove the sentence “In the figures, the x-axis spans the period 2000-2100.”
P13L326: In Table 3, you should indicate that the thermal conductivity relates to the rock.
P14L354: In Fig. 2, the range seems to be somewhat smaller than -5 to 5 °C.
P16L364–380: In addition to the trends, present also the temperatures or their ranges. This is only done at three stations for TTOP and at all of them for MAGT. Report also the values and trends from the mean annual air temperature (and maybe snowfall if both are shown in Fig. 3).
P16L381: In Table 4, specify the depth for the mean annual ground temperature.
P17L385: Fig. 3 has a misleading caption because snowfall and air temperature cannot be referred to as ground temperatures.
P18L388: In Fig. 4, I think it is a bit misleading to present the temperature for a single year 1950 because it was exceptionally cold. Is it appropriate to use a single year here, at least for the mean annual ground surface temperature, which has an instantaneous response to air temperature?
P18–19L391–416: Check that it is clearly stated everywhere what temperature the reported trends refer to.
P18L409: Rewrite the sentence “Despite continued interannual air temperature variability, MAGST fluctuated between -4 to -1 °C.” which does not make complete sense.
P19L412: It is unclear what temperature the trend of 2.10±0.60 °C refers to.
P19L423: You should rather write for 2020–2100. Trends should always refer to a period, not a single year. Additionally, it is unclear what the temperature trend of 0.59±0.22 °C refers to.
P20L427: Change “… being the warming more pronounced at the …” to “… with more pronounced warming at the…”
P20L429: Remove “For most of the observatories, “ Then, the sentence will make more sense.
P20L434–437: Maybe move this paragraph before the previous one. Definitely, it is better to start with the statement that all stations host permafrost at present, and then describe that it will warm and likely degrade or disappear at some of them.
P20L439: Again, it is unclear what the first trend of 2.59±0.20 °C refers to.
P21L469–470: Change “At Cierva despite a warming rate of 0.51 ± 0.01 °C.dec-1 positive MAGT is projected…” to “At Cierva, despite a warming rate of 0.51 ± 0.01 °C.dec-1, positive MAGT is projected…”
P21L476: Table 5 has a misleading caption because it presents other temperature variables in addition to the mean annual ground temperature at 20 m.
P22L480: Fig. 5 has a misleading caption because air temperature cannot be referred to as ground temperature.
P24L505–509: The interpretation that ground temperature warming in the periods 1950–1975 and 2015–2020 was favoured by decreased snowfall and shorter duration of the snow cover contradicts former explanations that thicker and longer snow cover tends to warm the ground. Likewise, cooling simulated for 2000–2015 is difficult to explain by reduced snowfall.
P24L513–514: Remove the first sentence in the section 5.2; it is redundant.
P24L516: Why are the periods 2000–2015 and 2016–2022 discussed in the section entitled “Future evolution of permafrost temperature (2022 - 2100)”?
P25L532: Change “Under the more optimistic SSP1-2.6 scenario, …” to “Under the SSP1-2.6 scenario, …”
P25L539: Add “with values ranging from 1 to 2 °C in 2100.”
P26L557: Specify to what depth the bias refers to.
P26L574: Change “In the worst-case scenario (SSP5-8.5), …” to “Under the SSP5-8.5 scenario, …”
P27L582: Change “thermokarst and landslide activity should increase” to “thermokarst and landslide activity will likely increase”
P31L666: Change “An historical period …” to “A historical period …”
P31L674: Change “improved coupled between” to “improved coupling between”
P31L679: Change “a RMSE of 1.33 °C. and a standard deviation” to “a RMSE of 1.33 °C and a standard deviation”
References
Baptista, J., Brito Guapo Teles Vieira, G., Manuel De Carvalho Soares Correia, A., Lee, H., Westermann, S., 2025. Modelling the evolution of permafrost temperatures and active layer thickness in King George Island, Antarctica, since 1950, The Cryosphere, 19, 3459–3476, https://doi.org/10.5194/tc-19-3459-2025
Biskaborn, B.K., Smith, S.L., Noetzli, J., Matthes, H., Vieira, G., Streletskiy, D.A., Schoeneich, P., Romanovsky, V.E., Lewkowicz, A.G., Abramov, A., Allard, M., Boike, J., Cable, W.L., Christiansen, H.H., Delaloye, R., Diekmann, B., Drozdov, D., Etzelmüller, B., Grosse, G., Guglielmin, M., Ingeman-Nielsen, T., Isaksen, K., Ishikawa, M., Johansson, M., Johansson, H., Joo, A., Kaverin, D., Kholodov, A., Konstantinov, P., Kröger, T., Lambiel, Ch., Lanckman, J.-P., Luo, D., Malkova, G., Meiklejohn, I., Moskalenko, N., Oliva, M., Phillips, M., Ramos, M., Sannel, A.B.K., Sergeev, D., Seybold, C., Skryabin, P., Vasiliev, A., Wu, Q., Yoshikawa, K., Zheleznyak, M., Lantuit, H., 2019. Permafrost is warming at a global scale. Nature Communications, 10, 264. https://doi.org/10.1038/s41467-018-08240-4
Guglielmin, M., Cannone, N., 2012. A permafrost warming in a cooling Antarctica? Climate Changem, 111, 177–195. https://doi.org/10.1007/s10584-011-0137-2
Hrbáček, F., Uxa, T., 2020. The evolution of a near-surface ground thermal regime and modeled active-layer thickness on James Ross Island, eastern Antarctic Peninsula, in 2006–2016. Permafrost and Periglacial Processes, 31, 141–155. https://doi.org/10.1002/ppp.2018
Hrbáček, F., Vieira, G., Oliva, M., Balks, M., Guglielmin, M., de Pablo, M.Á., Molina, A., Ramos, M., Goyanes, G., Meiklejohn, I., Abramov, A., Demidov, N., Fedorov-Davydov, D., Lupachev, A., Rivkina, E., Láska, K., Kňažková, M., Nývlt, D., Raffi, R., Strelin, J., Sone, T., Fukui, K., Dolgikh, A., Zazovskaya, E., Mergelov, N., Osokin, N., Miamin, V., 2021. Active layer monitoring in Antarctica: an overview of results from 2006 to 2015. Polar Geography, 44, 217–231. https://doi.org/10.1080/1088937X.2017.1420105
Hrbáček, F., Oliva, M., Hansen, C., Balks, M., O'Neill, T. A., de Pablo, M. A., Ponti, S., Ramos, M., Vieira, G., Abramov, A., Kaplan Pastíriková, L., Guglielmin, M., Goyanes, G., Rocha Francelino, M., Schaefer, C., Lacelle, D., 2023. Active layer and permafrost thermal regimes in the ice-free areas of Antarctica, Earth-Science Reviews, 242, 104458, https://doi.org/10.1016/j.earscirev.2023.104458
Hrbáček, F., Kňažková, M., Láska, K., Kaplan Pastíriková, L., 2025. Active Layer Warming and Thickening on CALM‐S JGM, James Ross Island, in the Period 2013/14–2022/23, Permafrost and Periglacial Processes, 36, 378–389, https://doi.org/10.1002/ppp.2274
Kaplan Pastíriková, L., Hrbáček, F., Matějka, M., 2025. Validation of ERA5-Land-based reconstructed air temperature and near-surface ground temperature on James Ross Island. Polar Geography, 48, 95–115, https://doi.org/10.1080/1088937X.2024.2434744
Oliva, M., Navarro, F., Hrbáček, F., Hernández, A., Nývlt, D., Pereira, P., Ruiz-Fernández, J., Trigo, R., 2017. Recent regional climate cooling on the Antarctic Peninsula and associated impacts on the cryosphere. Science of the Total Environment, 580, 210–223. https://doi.org/10.1016/j.scitotenv.2016.12.030
Ramos, M., Vieira, G., de Pablo, M.A., Molina, A., Abramov, A., Goyanes, G., 2017. Recent shallowing of the thaw depth at Crater Lake, Deception Island, Antarctica (2006–2014). Catena, 149, 519–528. https://doi.org/10.1016/j.catena.2016.07.019
All the best,
Tomáš Uxa