FIGURE LEGENDS

Fig. 1 . Schematic of the soil column showing variables used to represent permafrost dynamics, modified after (Dobinski, 2011; Burke et al. , 2020; Elshamyet al. , 2020).
Fig. 2. Location of the study area, temperature boreholes, and permafrost classification – the two selected sites are highlighted in the focused view.
Fig. 3. Permafrost’s annual maximum and minimum temperature profiles for A) 85-12-B borehole, B) 99TC03 borehole.
Fig. 4. Detailed illustration of the model configuration in terms of antecedent water condition. Numbers between brackets above each bar correspond to the portions of liquid and frozen water content for each experiment, respectively.
Fig. 5 . Summary of the required number of spin-up cycles needed for model state variables to reach equilibrium for A) JMC, and B) WH. For each climate condition, the number of spin-up cycles is sorted in ascending order for all the 21 initial moisture scenarios in Fig. 4 . Dashed lines separate the groups of climate experiments. Convergence criteria are 0.1°C for temperature and 0.01 m3 m-3 for liquid and frozen water.
Fig. 6. Temporal progression of spin-up convergence of soil temperature, soil liquid content, and soil frozen content at A) JMC site - Exp. Dry 7 (upper row), B) WH site – Exp. Wet 17 (middle row), and C) WH site – Exp. Avg 10 (lower row). Layers below SDEP have no moisture; only heat can transfer vertically between soil layers.
Fig. 7. Summary of the required numbers of spin-up cycles needed by all state variables to reach equilibrium under different A) climate conditions, and B) initial soil moisture; only experiments that formed permafrost are included; black boxes correspond to the median of each group of experiments in subplot B. The configuration Id label is as inFig. 4 .
Fig. 8 . Comparison of A) cumulative daily precipitation, B) daily air temperature, C) histogram of daily air temperature based on the selected five climate conditions at WH site.
Fig. 9. Temporal variation in A) external forcing, B) soil temperature, C) soil liquid moisture content, and D) soil frozen moisture content for the first spin-up cycle under Average climate (right panel) and cold climate (left panel) for WH site.
Fig. 10. Soil temperature (column A and B), soil liquid content (column C), and soil frozen content (column D) at the end of the 2000 spin-up cycles at JMC site (upper panel) and WH site (lower panel); shading indicates the range of variability; each line represents an individual experiment; SDEP and ODEP indicate depth to the bedrock and depth to the organic matter, respectively. The presented state-variables are plotted at the middle of each soil layer. *: B is zooming into A to the depth shown for C & D.
Fig. 11. Time series of the error in the simulated ALT, and the mean RMSE of the simulated annual temperature envelope (mean RMSE for Tmax and Tmin) among all the configurations at JMC site (column A), WH site (column B). Dashed lines separate the groups of climate experiments presented with the same numeric order as in Fig. 4 ; A: Average, C: Cold, D: Dry, Wr: Warm, and W: Wet. Gaps correspond to unavailable records.
Fig. 12. Time series of the RMSE of the simulated annual temperature envelopes. The lines represent the envelope mean, and the shadings represent the range of total variability because of the initial moisture’s and climate conditions’ uncertainties.
Fig. 13. Time series of the simulated A) MAGTp, B) ALT-DOY, C) DZAA, and D) PB at JMC site (upper panel), and WH site (lower panel), after the spin-up for 2000 cycles. The simulation label is as in Fig. 11 .