This is the accepted version of the journal article: Li, Jialing; Wu, Chaoyang; Peñuelas, Josep; [et al.]. «The start of frozen dates over northern permafrost regions with the changing climate». Global change biology, Vol. 29, issue 16 (Aug. 2023), p. 4556-4568. DOI 10.1111/gcb.16752 This version is available at https://ddd.uab.cat/record/284324 under the terms of the license 1 The start of frozen dates over northern permafrost regions with 2 the changing climate 3 Jialing Li1, Chaoyang Wu2,3*, Josep Peñuelas4,5, Youhua Ran6, Yongguang 4 Zhang1,7 5 1. International Institute for Earth System Sciences, Jiangsu Center for Collaborative 6 Innovation in Geographical Information Resource Development and Application, Nanjing 7 University, Nanjing, 210023, China; 8 2. The Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographical 9 Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, 10 100101, China; 11 3. University of the Chinese Academy of Sciences, Beijing, 100049, China; 12 4. CSIC, Global Ecology Unit CREAF-CSIC-UAB, Bellaterra, Barcelona 08193, Catalonia, 13 Spain; 14 5. CREAF, Cerdanyola del Valles, Barcelona 08193, Catalonia, Spain; 15 6. Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, 16 Lanzhou 730000, China 17 7. Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, 18 Key Laboratory for Land Satellite Remote Sensing Applications of Ministry of Natural 19 Resources, School of Geography and Ocean Science, Nanjing University, Nanjing, 20 210023, China; 21 *Corresponding author: C Wu (wucy@igsnrr.ac.cn) 22 1 23 Abstract: The soil freeze-thaw cycle in the permafrost regions has a significant impact on 24 regional surface energy and water balance. Although increasing efforts have been made 25 to understand the responses of spring thawing to climate change, the mechanisms 26 controlling the global interannual variability of the start date of permafrost freezing (SOF) 27 remain unclear. Using long-term satellite observations of SOF between 1979–2020, and 28 analytical techniques, including partial correlation, ridge regression, structural equation 29 modelling (SEM), and machine learning, we explored the responses of SOF to multiple 30 climate change factors, including warming (skin and air temperature), start date of 31 permafrost thawing (SOT), soil properties (soil temperature and volume of water), and the 32 snow depth water equivalent (SDWE). Overall, climate warming exhibited the maximum 33 control on SOF, but SOT in spring was also an important driver of SOF variability; among 34 the 65.9% significant SOT and SOF correlations, 79.3% were positive, indicating an overall 35 earlier thawing. The machine learning analysis also suggested that apart from warming, 36 SOT ranked as the second most important determinant of SOF. Therefore, we identified 37 the mechanism responsible for the SOT-SOF relationship, and SEM analysis revealed that 38 soil temperature change exhibited the maximum effect on this relationship, irrespective of 39 the permafrost type. Finally, we analysed the temporal changes in these responses using 40 the moving window approach and found increased effect of soil warming on SOF. Therefore, 41 these results provide important insights into understanding and predicting SOF variations 42 with future climate change. 43 Key words: start date of freezing; permafrost; climate warming; start date of thawing 44 1. Introduction 2 45 Permafrost occupies 25% of land in the northern hemisphere and is highly sensitive 46 to climate change (Schuur et al., 2015; Obu et al., 2019; Smith et al., 2022). With warming 47 three times the global mean, significant changes have been observed in the permafrost 48 regions, which have affected the global carbon cycle and other environmental factors 49 because the permafrost regions could release approximately half of the global soil carbon 50 stock in the form of methane, carbon dioxide (Hugelius, et al., 2014; Vikhamar-Schuler et 51 al., 2016; McGuire et al., 2018; Biskaborn et al., 2019; Irrgang et al., 2022), and other 52 poisonous elements, such as mercury (Schaefer et al., 2020; Miner et al., 2021), that may 53 threaten the infrastructure and ecosystems of the region (Nielsen et al., 2022; Ran et al., 54 2022a; Wang et al., 2022). 55 56 Several studies have reported the degradation of permafrost based on soil warming, 57 deepened active layers, reduced permafrost extent, and decreased surface freezing 58 duration (Biskaborn et al., 2019; Zhang et al., 2005; Peng et al., 2018; Li et al., 2022a; Ran 59 et al., 2022b). The soil freeze-thaw cycle has significantly changed due to permafrost 60 degradation, which has impacted the surface energy and water balance by influencing the 61 hydrothermal properties of the soil; therefore, it has been considered as an indicator of the 62 interactions between permafrost and climate change (Li et al., 2012; Smith et al., 2022). 63 Permafrost thawing triggers emission of soil organic carbon (SOC), which accelerates 64 global warming and impacts soil nutrients, river flow, and vegetation productivity (Grosse 65 et al., 2011; Olefeldt et al., 2016; Mu et al., 2020; Liu et al., 2022; Rößger et al., 2022). 66 Previous studies have investigated the impact of climate change on the start date of 3 67 permafrost thawing (SOT) in spring (Chen et al., 2022), based on the importance of SOT 68 in characterising land surface change and its effect on climate (Miner et al., 2021; Domine 69 et al., 2022). For example, permafrost thawing causes a significant decline in water level, 70 which could restructure the regional energy balance (Webb, et al., 2022). However, 71 compared to the data on available on spring thawing, the start date of permafrost freezing 72 (SOF) has not been substantially investigated. Therefore, research on SOF as an 73 important permafrost-related factor will help in the elucidation of the underlying 74 mechanisms and factors causing permafrost degradation. Moreover, the link between 75 vegetation growth and SOF suggests that SOF plays an important role in bridging climate 76 change and carbon uptake by the permafrost regions (Li et al., 2022b). 77 78 Therefore, to understand the impact of climate change on SOF on the permafrost 79 regions of the northern hemisphere, we used long-term satellite-observed SOF data 80 between 1979–2020 and data pertaining to various climate change variables, including 81 skin temperature (Tskin), air temperature (Tair), SOT, soil temperature (Tsoil), volume of water 82 in soil (SWV), and the snow depth water equivalent (SDWE), to comprehensively analyse 83 the response of SOF to climate change over the past four decades. The key objectives of 84 our study were (1) to understand the impacts of climate change on the interannual 85 variability of SOF; (2) to explore potential carry-over effects between thawing and freezing 86 and their mechanisms; and (3) to analyse the temporal evolution of the strength of these 87 climate change factors on SOF variability. 88 3. Results 4 89 3.1. Spatial and temporal patterns of SOF 90 We observed spatial variations in SOF over northern lands, with 75% SOFs ranging 91 between day of year (DOY) 260 and 290 (Figure 4a). SOFs earlier than DOY 260 were 92 observed in the northern Tibetan Plateau, whereas SOFs later than DOY 290 were mostly 93 found in the boreal regions of Canada. Overall, the regions with earlier SOFs exhibited 94 smaller standard deviations (SD). For example, large areas in Eurasia with an SOF 95 between DOY 260–280 exhibited a lower SD of <15 days (Figure 4b). Moreover, 7.7% of 96 the region exhibited significantly delayed SOF dates, whereas 8.6% of the region exhibited 97 earlier SOFs (Figure 4c). Nevertheless, high-latitude permafrost regions exhibited delayed 98 SOFs over the past four decades. 99 100 We also analysed climate change factors from to 1979–2020 and found warming in all 101 permafrost regions analysed in this study based on significant increase in both Tskin (59.2%) 102 and Tair (67.4%) (Figure 4d–e). However, 91.8% of all regions exhibited earlier SOT, among 103 which data for 66.1% regions were significant (Figure 4f). Moreover, Tsoil significantly 104 increased in 70.5% of the regions that underwent significant soil warming (Figure 4g). The 105 SWV exhibited an overall declining trend, with significant proportions for negative and 106 positive trends of 16.5% and 8.2%, respectively. Similar results were observed for SDWE; 107 19.2% of the regions exhibited significant decrease in SDWE, whereas 4.7% of the regions 108 exhibited an increasing trend (Figure 4i). 5 109 110 Figure 4 Spatial and temporal patterns of the start date of permafrost freezing (SOF) 111 and trends in climate factors. a and b are SOF and its standard deviation (SD), 112 whereas c represents its temporal trends between 1979–2020. d–f represent trends 113 of skin temperature (Tskin), air temperature (Tair), start date of permafrost thawing 114 (SOT), soil temperature (Tsoil), volume of water in soil (SWV), and the snow depth 115 water equivalent (SDWE). N and P indicate negative and positive correlations, 116 respectively. ** and * represent significance at p=0.01 and p=0.05, respectively. 117 118 3.2. Relationship between SOF and climate change 119 We elucidated the relationships between SOF and climate factors using partial 6 120 correlation analysis and found that Tskin was negatively correlated with SOF, and among 121 the 89.3% negative correlations, 77.8% were significant (Figure 5a). However, Tair was 122 positively correlated with SOF (Figure 5b). An overall positive correlation was also 123 observed between SOT and SOF; 52.3% (out of 72.3%) of the regions exhibited a 124 significant positive relationship between SOT and SOF (Figure 5c). Significant positive 125 correlations were also found between SOF and Tsoil, but both SWV and SDWE were 126 negatively correlated with SOF (Figure 5d–f). We also determined the sensitivity of SOF to 127 these factors using ridge regression analysis (Figure 5 g–l). The signs of the sensitivity 128 values were consistent with those obtained through partial correlation analysis. For 129 example, the slopes of the SOT-Tskin correlations were mostly negative, whereas an 130 increase in Tair implied a later SOF, as shown by a positive sensitivity value. Moreover, we 131 used machine learning techniques and found that Tskin-dominated SOF changed in 42% of 132 the areas (Figure 5m). However, SOT was the second-most important determinant of SOF; 133 SWV exhibited the least effect on SOF among all variables. 7 134 135 Figure 5 Partial correlations and sensitivities of the start date of permafrost freezing 136 (SOF) and climate change. a–f are partial correlations between SOF and skin 137 temperature (Tskin), air temperature (Tair), start date of permafrost thawing (SOT), soil 138 temperature (Tsoil), volume of water in soil (SWV), and the snow depth water 8 139 equivalent (SDWE); g–l represent their sensitivities. m shows the dominant factors 140 for each grid as per the machine learning technique. Significance was set at p <0.05. 141 142 3.3. Underlying mechanisms of the correlation between SOT and SOF 143 We used structural equation modelling (SEM) path analysis to elucidate the underlying 144 reasons for the correlation between SOT and SOF in different permafrost types to elucidate 145 the changes expected in SOF in the future (Figure 6). We found that among the four paths 146 through which SOT can be connected to SOF, the most important path was the positive 147 correlation of SOT with Tsoil, and because Tsoil was positively correlated with SOF, SOT 148 was positively correlated with SOF. These linkages were dominant in all paths and were 149 independent of the permafrost type. Although changes in SWV explained changes in SOF, 150 the overall impact of this path was minor. 9 151 152 Figure 6 Structural equation models for the correlation between the start date of 153 permafrost thawing (SOT) and start date of permafrost freezing (SOF). a–d represent 154 results for (e) continuous, discontinuous, isolated, and sporadic permafrost types, 155 respectively. Tsoil, SWV, and SDWE are soil temperature, volume of water in soil, and 156 snow depth water equivalent, respectively. *** represent p <0.001. 157 158 3.4. Temporal changes in the response of SOF to climate change 159 We further analysed the temporal trends in the effect of these factors on SOF and 160 changes in their sensitivity towards SOF (Figure 7). Overall, we found divergent patterns 10 161 among (66.7%) regions where Tskin had a significant impact on SOF and 33.9% regions 162 exhibited increased effect of Tskin; this proportion is comparable with that of the regions 163 with reduced effect of Tskin (Figure 7a). Most of the regions exhibiting enhanced control of 164 warming were located in the highlands of Eurasia (e.g. Siberia), whereas mid-latitude 165 regions exhibited weakened control. Similar results were observed for Tair, SOT, SWV, and 166 SDWE, but the proportion of the regions where these factors exhibited enhanced and 167 weakened control varied. For example, Tair exhibited enhanced control of SOF in 34.8% of 168 permafrost regions and 31.7% of counterparts. However, we found that the overall Tsoil 169 exhibited increased control over SOF, and among the 67.2% of regions where Tsoil trends 170 were significant, 42.8% exhibited enhanced control (Figure 7d). We further grouped the 171 enhanced/weakened trends by permafrost type (Figure 8). We found that most of these 172 factors, including Tskin, Tair, Tsoil, and SWV, exhibited enhanced control on SOF for 173 continuous permafrost, whereas SOT and SDWE exhibited the opposite effect. For the 174 other three types (discontinuous, isolated, and sporadic), weakened control was more 175 evident. However, Tsoil was an exception, as it exhibited enhanced control in case of all 176 permafrost types. 11 177 178 Figure 7 Spatial distribution and sensitivity of the temporal changes in climate 179 change factors on the start date of permafrost freezing (SOF). a–f are spatial 180 distributions of the enhanced and weakened regions and their slopes for skin 181 temperature (Tskin), air temperature (Tair), start date of permafrost thawing (SOT), soil 182 temperature (Tsoil), volume of water in soil (SWV), and the snow depth water 183 equivalent (SDWE). Significance was set at p <0.05. 12 184 185 Figure 8 Effect of temporal changes in climate change factors on the start of date of 186 permafrost freezing (SOF) according to permafrost types. a–f are spatial 187 distributions of the enhanced and weakened regions and their slopes for skin 188 temperature (Tskin), air temperature (Tair), start date of permafrost thawing (SOT), soil 189 temperature (Tsoil), volume of water in soil (SWV), and the snow depth water 190 equivalent (SDWE). 13 191 192 4. Discussion 193 The timing of freezing represents a significant transition of the status of permafrost 194 soil, and changes in the frozen/thawing processes significantly influence the hydrothermal 195 properties of the soil, resulting in significant changes in the surface energy and water 196 balance (Guo et al., 2011; Li et al., 2012; Smith et al., 2022; Webb et al., 2023). Previous 197 studies have reported that permafrost thawing is highly sensitive to climate change, with 198 far-reaching implications on regional climate and vegetation that can mitigate or amplify 199 permafrost thawing (Jafarov et al., 2018; Polvani et al., 2020; Heijmans et al., 2022). Here, 200 we elucidated the impacts of climate change factors on SOF in late autumn and winter, 201 which has received much less attention than spring thawing in previous studies. We found 202 that that the proportion of regions with significantly delayed and advanced SOF was 203 comparable (8.6% vs. 7.7%, Figure 4c), but with clear spatial distribution. Delayed SOF 204 was mostly observed in the highlands and the Tibetan Plateau, i.e. in areas with more 205 pronounced warming, whereas earlier SOF was observed in the western boreal regions of 206 Canada. Several climate change indicators were analysed in this study, which exhibited 207 substantial impacts on SOF variability. These results contribute to a better understanding 208 of permafrost change, given that the timing of freezing could impact snow-related soil 209 warming in winter through biogeophysical feedbacks, such as changes in albedo and soil 210 moisture (Li et al., 2022b); this could cause uncertainty in the evaluation of consequences 211 of thaw-related disturbances (Heijmans et al., 2022). Moreover, SOF responded differently 212 to these climate factors, and complicated patterns were observed; besides warming, other 14 213 environmental cues (e.g. SWC, soil temperature) could also be responsible for SOF 214 variability in the changing climate (Guo et al., 2012; Koven et al., 2013). 215 Global warming is the main driver of permafrost change, as it deepens the thickness 216 of the active layer (Peng et al., 2018) and collapses the ground or thermokarst 217 (Farquharson et al., 2019; Turetsky et al., 2020). Our results confirm the importance of 218 global warming, as per the sensitivity of SOF to Tskin. Moreover, we identified the 219 importance of SOT in regulating SOF variability. The overall positive correlation between 220 SOF and SOT implies that an earlier thawing in spring could be associated with an earlier 221 timing of freezing. These carry-over effects have been reported in previous phenological 222 analyses (Fu et al., 2014; Xu et al., 2020; Shen et al., 2022); in accordance with these 223 results, we observed a linkage between the thawing/frozen phenology of permafrost, which 224 will be useful to obtain a comprehensive view of permafrost degradation. Using machine 225 learning analysis, we found that SOT in spring was ranked as the second most important 226 driver of SOF in the later years (Figure 5m), and this linkage can be explained by the 227 changes in soil temperature, using the SEM path analysis for all permafrost types. Although 228 the specific mechanism underlying this correlation is unclear, these results support 229 previous reports on the importance of complex soil hydrothermal processes underlying 230 permafrost degradation (Lawrence et al., 2005). Moreover, the positive correlation between 231 SOT and SOF established in this study demonstrates that freezing/thawing cycle of a 232 permafrost region can be understood based on soil warming and moisture changes 233 throughout the year (Figure 6). Moreover, we observed increased effect of SOT on SOF at 234 mid-low latitudes, where warming might not be that strong compared with highlands. This 15 235 is important because mid-low latitude ecosystems have larger productivity, and including 236 the SOT-SOF correlation could be useful in understanding the interaction between 237 permafrost degradation and vegetation change. 238 We further identified the temporal evolution of the effects of different climate change 239 factors on SOF variability; this will provide insights into the sensitivity of SOF to future 240 climate change, although our current prediction of future scenarios regarding permafrost 241 degradation could be biased (Webb and Liljedahl, 2023). We found differential changes in 242 the sensitivity of SOF to most factors, but the differences were spatially distributed. 243 Moreover, we observed a consistently increased control of Tsoil on SOF, indicating an 244 urgent need to investigate soil warming trends with respect to future climate change. Since 245 warmer soil delays SOF (Figure 5d), the enhanced effect of Tsoil on SOF could aggravate 246 later freezing and accelerate permafrost degradation in the future. Our results are in line 247 with the reported fast response of cold ice-rich permafrost to global warming (Nitzbon et 248 al., 2020), considering that the enhanced regions are more in continuous permafrost 249 (Figure 8d). Moreover, future studies should characterise soil types (Fisher et al., 2016), 250 soil nutrition (Mao et al., 2020), and soil age (Tanentzap et al., 2021) to understand 251 permafrost degradation under a changing climate. 252 ACKNOWLEDGMENTS: This work was funded by the National Natural Science 253 Foundation of China (42125101) and the CAS Interdisciplinary Innovation Team (JCTD- 254 2020-05). JP acknowledges support from the Spanish Government Grant TED2021- 255 132627B-I00 funded by MCIN, AEI/10.13039/501100011033 European Union Next 256 GenerationEU/PRTR, the Catalan Government Grant SGR2021-1333, and the 257 Fundación Ramón Areces Grant CIVP20A6621. 16 258 259 References 260 Biskaborn, B.K., Smith, S.L., Noetzli, J. et al. (2019). Permafrost is warming at a global 261 scale. Nature Communications, 10, 264. https://doi.org/10.1038/s41467-018-08240-4 262 Chen, X., Jeong, S., Park, C. E., et al. (2022). Different responses of surface freeze and 263 thaw phenology changes to warming among Arctic permafrost types. Remote Sensing 264 of Environment, 272, 112956. 265 Domine, F., Fourteau, K., Picard, G. et al. Permafrost cooled in winter by thermal bridging 266 through snow-covered shrub branches. Nature Geoscience, 15, 554–560 (2022). 267 https://doi.org/10.1038/s41561-022-00979-2. 268 Farquharson, L. M., Romanovsky, V. E., Cable, W. L., Walker, D. A., Kokelj, S., & Nicolsky, 269 D. (2019). Climate change drives widespread and rapid thermokarst development in 270 very cold permafrost in the Canadian High Arctic. Geophysical Research Letters, 271 2019GL082187. https://doi.org/10.1029/2019GL082187. 272 Fisher, J. P. , C Estop-Aragonés, Thierry, A. , Charman, D. J. , Wolfe, S. A. , & Hartley, I. 273 P. , et al. (2016). The influence of vegetation and soil characteristics on active-layer 274 thickness of permafrost soils in boreal forest. Global Change Biology, 22, 3127-3140. 275 Fu, Y., Campioli, M., Vitasse, Y., Boeck, H., Berge, J., Abdelgawad, H., et al. (2014). 276 Variation in leaf flushing date influences autumnal senescence and next year's flushing 277 date in two temperate tree species. Proceedings of the National Academy of Sciences, 278 111, 7355-7360. 279 Grosse, G. et al. (2021). Vulnerability of high-latitude soil organic carbon in North America 17 280 to disturbance. Journal of Geophysical Research, Biogeosciences 116, 1–23. 281 Guo, D., Yang, M., & Wang, H. (2011). Characteristics of land surface heat and water 282 exchange under different soil freeze/thaw conditions over the central Tibetan Plateau. 283 Hydrological Processes, 25(16), 2531-2541. 284 Guo, D.,Wang, H., Li, D. (2012). A projection of permafrost degradation on the Tibetan 285 Plateau during the 21st century, Journal of Geophysical Research Atmospheres, 117. 286 Heijmans, M.M.P.D., Magnússon, R.Í., Lara, M.J. et al. (2022). Tundra vegetation change 287 and impacts on permafrost. Nature Reviews on Earth and Environment, 3, 68–84. 288 https://doi.org/10.1038/s43017-021-00233-0 289 Hugelius, G., Strauss, J., Zubrzycki, S., Harden, J. W., Schuur, E. A. G., Ping, C.-L., 290 Schirrmeister, L., Grosse, G., Michaelson, G. J., Koven, C. D., O’Donnell, J. A., 291 Elberling, B., Mishra, U., Camill, P., Yu, Z., Palmtag, J., & Kuhry, P. (2014). Estimated 292 stocks of circumpolar permafrost carbon with quantified uncertainty ranges and 293 identified data gaps. Biogeosciences, 11(23), 6573-6593. https://doi.org/10.5194/bg- 294 11-6573-2014. 295 Irrgang, A.M., Bendixen, M., Farquharson, L.M. et al. Drivers, dynamics and impacts of 296 changing Arctic coasts. Nature Reviews on Earth and Environment, 3, 39–54 (2022). 297 https://doi.org/10.1038/s43017-021-00232-1 298 Jafarov, E., MacDougall, A. H., Marchenko, S., and Nicolsky, D.: Dependence of the 299 evolution of carbon dynamics in the northern permafrost region on the trajectory of 300 climate change, Proceedings of the National Academy of Sciences, 115, 3882-3887, 301 2018. 18 302 Koven, C. D., Riley,W. J., and Stern, A. (2013). Analysis of permafrost thermal dynamics 303 and response to climate change in the CMIP5 Earth System Models. Journal of 304 Climate, 26, 1877-1900. 305 Lawrence, D. M. Slater, A. G. (2005). A projection of severe near-surface permafrost 306 degradation during the 21st century, Geophysical Research Letters, 32. 307 Li, X., Jin, R., Pan, X., et al. (2012). Changes in the near-surface soil freeze–thaw cycle on 308 the Qinghai-Tibetan Plateau. International Journal of Applied Earth Observation and 309 Geoinformation, 17, 33-42. 310 Mao, C., Kou, D., Chen, L., Qin, S., Zhang, D., & Peng, Y., et al. (2020). Permafrost 311 nitrogen status and its determinants on the tibetan plateau. Global Change Biology, 312 26(9), 5290-5302. 313 McGuire, A. D., Lawrence, D. M., Koven, C., Clein, J. S., Burke, E., Chen, G., et al. (2018). 314 Dependence of the evolution of carbon dynamics in the northern permafrost region on 315 the trajectory of climate change. Proceedings of the National Academy of Sciences, 316 115(15), 3882-3887. https://doi.org/10.1073/pnas.1719903115. 317 Miner, K.R., D’Andrilli, J., Mackelprang, R. et al. Emergent biogeochemical risks from Arctic 318 permafrost degradation. Nature Climate Change, 11, 809–819 (2021). 319 https://doi.org/10.1038/s41558-021-01162-y 320 Mu, C., Abbott, B. W., Norris, A. J., et al. (2020). The status and stability of permafrost 321 carbon on the Tibetan Plateau. Earth-Science Reviews, 211, 103433. 322 Nielsen, D.M., Pieper, P., Barkhordarian, A. et al. Increase in Arctic coastal erosion and its 323 sensitivity to warming in the twenty-first century. Nature Climate Change, 12, 263–270 19 324 (2022). https://doi.org/10.1038/s41558-022-01281-0. 325 Nitzbon, J., Westermann, S., Langer, M. et al. Fast response of cold ice-rich permafrost in 326 northeast Siberia to a warming climate. Nature Communications, 11, 2201 (2020). 327 https://doi.org/10.1038/s41467-020-15725-8. 328 Obu, J., Westermann, S., Bartsch, A., Berdnikov, N., Christiansen, H. H., Dashtseren, A., 329 Delaloye, R., Elberling, B., Etzelmüller, B., Kholodov, A., Khomutov, A., Kääb, A., 330 Leibman, M.O., Lewkowicz, A.G., Panda, S.K., Romanovsky, V., Way, R. G., 331 Westergaard-Nielsen, A., Wu, T., et al. (2019). Northern Hemisphere permafrost map 332 based on TTOP modelling for 2000–2016 at 1 km2 scale. Earth-Science Reviews, 193, 333 299–316. https://doi.org/10.1016/j.earscirev.2019.04.023. 334 Olefeldt, D. et al. Circumpolar distribution and carbon storage of thermokarst landscapes. 335 Nature Communications, 7, 13043 (2016). 336 Peng, X., Zhang, T., Frauenfeld, O. W., et al. (2018). Spatiotemporal changes in active 337 layer thickness under contemporary and projected climate in the Northern Hemisphere. 338 Journal of Climate, 31(1), 251-266. 339 Polvani, L. M., Previdi, M., England, M. R., Chiodo, G., & Smith, K. L. (2020). Substantial 340 twentieth-century Arctic warming caused by ozone-depleting substances. Nature 341 Climate Change, 1-4. https://doi.org/10.1038/s41558-019-0677-4. 342 Ran, Y., Cheng, G., Dong, Y. et al. Permafrost degradation increases risk and large future 343 costs of infrastructure on the Third Pole. Communications Earth and Environment, 3, 344 238 (2022a). https://doi.org/10.1038/s43247-022-00568-6 345 Rößger, N., Sachs, T., Wille, C. et al. Seasonal increase of methane emissions linked to 20 346 warming in Siberian tundra. Nature Climate Change, 12, 1031–1036 (2022). 347 Schaefer, K., Elshorbany, Y., Jafarov, E. et al. (2020). Potential impacts of mercury 348 released from thawing permafrost. Nature Communications, 11, 4650. 349 https://doi.org/10.1038/s41467-020-18398-5. 350 Schuur, E. A. G. et al. Climate change and the permafrost carbon feedback. (2015), Nature, 351 520, 171-179. 352 Shen, M., Wang, S., Jiang, N. et al. (2022). Plant phenology changes and drivers on the 353 Qinghai–Tibetan Plateau. Nature Reviews on Earth and Environment, 3, 633–651. 354 https://doi.org/10.1038/s43017-022-00317-5 355 Smith, S.L., O’Neill, H.B., Isaksen, K. et al. (2022). The changing thermal state of 356 permafrost. Nature Reviews on Earth and Environment, 3, 10–23. 357 https://doi.org/10.1038/s43017-021-00240-1 358 Tanentzap, A. J., Burd, K., Kuhn, M.K., C Estop-Aragonés, Tank, S.E. , Olefeldt, D. (2021). 359 Aged soils contribute little to contemporary carbon cycling downstream of thawing 360 permafrost peatlands. Global Change Biology, 27, 5368-5382. 361 Turetsky, M. R., Abbott, B. W., Jones, M. C., Anthony, K. W., Olefeldt, D., Schuur, E. A. 362 G., Grosse, G., Kuhry, P., Hugelius, G., Koven, C., Lawrence, D. M., Gibson, C., 363 Sannel, A. B. K., & McGuire, A. D. (2020). Carbon release through abrupt permafrost 364 thaw. Nature Geoscience, 13(2), 138-143. https://doi.org/10.1038/s41561-019-0526- 365 0. 366 Vikhamar-Schuler, D., Isaksen, K., Haugen, J.E., 2016. Changes in winter warming events 367 in the nordic arctic region. Journal of Climate, 29, 6223–6244. 21 368 Wang, J., Liu, D. (2022). Vegetation green-up date is more sensitive to permafrost 369 degradation than climate change in spring across the northern permafrost region. 370 Global Change Biology, 28, 1569-1582. 371 Webb, E.E., Liljedahl, A.K. (2023). Diminishing lake area across the northern permafrost 372 zone. Nature Geoscience, https://doi.org/10.1038/s41561-023-01128-z. 373 Webb, E.E., Liljedahl, A.K., Cordeiro, J.A. et al. (2022). Permafrost thaw drives surface 374 water decline across lake-rich regions of the Arctic. Nature Climate Change, 12, 841– 375 846 (2022). https://doi.org/10.1038/s41558-022-01455-w 376 Xu, X., Riley, W.J., Koven, C.D. et al. Earlier leaf-out warms air in the north. Nature Climate 377 Change, 10, 370–375 (2020). https://doi.org/10.1038/s41558-020-0713-4 22