Simulated response of the active layer thickness of permafrost to climate change

Ruiho Li , , Jino Xie , , Zhenghui Xie , , Junqing Go , Bingho Ji , Peihu Qin ,Longhun Wng , , Yn Wng , , Bin Liu , , Si Chen ,

a State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences,Beijing, China

b School of Mathematics and Statistics, Nanjing University of information Science and Technology, Nanjing, China

c College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing, China

Key words:Active layer thickness CAS-LSM Variation trends Relative changes Climate change

ABSTRACT The active layer thickness (ALT) in permafrost regions, which affects water and energy exchange, is a key variable for assessing hydrological processes, cold-region engineering, and climate change. In this study, the authors analyzed the variation trends and relative changes of simulated ALTs using the Chinese Academy of Sciences Land Surface Model (CAS-LSM) and the Chinese Academy of Sciences Flexible Global Ocean-Atmosphere-Land System Model, gridpoint version 3 (CAS-FGOALS-g3). Firstly, the simulated ALTs produced by CAS-LSM were shown to be reasonable by comparing them with Circumpolar Active Layer Monitoring observations. Then, the authors simulated the ALTs from 1979 to 2014, and their relative changes across the entire Northern Hemisphere from 2015 to 2100. It is shown that the ALTs have an increasing trend. From 1979 to 2014, the average ALTs and their variation trends over all permafrost regions were 1.08 m and 0.33 cm yr − 1 , respectively. The relative changes of the ALTs ranged from 1% to 58%, and the average relative change was 10.9%. The variation trends of the ALTs were basically consistent with the variation trends of the 2-m air temperature. By 2100, the relative changes of ALTs are predicted to be 10.3%, 14.6%, 30.1%, and 51%, respectively, under the four considered hypothetical climate scenarios (SSP-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5). This study indicates that climate change has a substantial impact on ALTs, and our results can help in understanding the responses of the ALTs of permafrost due to climate change.

1. Introduction

Permafrost is widespread in the Northern Hemisphere, accounting for about 25% of its total land area ( Muller, 1947 ; Zhang et al., 1999 ).Permafrost has degraded significantly in recent years due to climate warming ( Koven et al., 2013 ), where rising air temperatures have led to ground warming and thickening of the active layer. Indeed, changes of the active layer thickness (ALT) of permafrost have a significant influence on water resources, engineering applications, and the terrestrial carbon cycle ( Zimov et al., 2006 ; Schuur et al., 2008 , 2009 ;Cuo et al., 2015 ). The active layer of permafrost contains vast amounts of carbon ( Mu et al., 2015 ); thus, as the ALT increases, a large amount of greenhouse gas stored in the Arctic permafrost will be released into the atmosphere and exacerbate global warming ( Schuur et al., 2015 ).Melting permafrost can increase the amount of hazards in cold regions,which in turn can have an important impact on hydrological processes,ecosystem carbon cycle processes, climate change, and cold-zone engineering and infrastructure ( Wu and Zhang, 2010 ; Guo and Wang, 2016 ;Hjort et al., 2018 ). The World Climate Research Programme has included studies of the changing ALTs of permafrost as one of the main observational goals of the Climate and Cryosphere project ( Brown et al.,2000 ; Nelson et al., 2004 ).

Early studies of Arctic permafrost regions showed that ALTs have been increasing over the past few decades ( Harris et al., 2003 ). In recent decades, the ALT on the Tibetan Plateau has also shown an obvious increasing trend ( Zhao and Ping, 2004 ; Wu et al., 2010 ). ALTs are an important indicator of climate change in permafrost regions( Peng et al., 2018 ). Thus, studying variations of the ALT of permafrost and their change over time is of great importance in the context of global climate change. However, permafrost is mainly distributed in the Northern Hemisphere at high latitudes and high altitudes, where observational data of ALTs across large spatial and temporal scales are scarce.Thus, we need to use numerical simulations to study how the ALTs of permafrost change across large spatial and long-term temporal scales.

Simulations performed via land surface models (LSMs) combined with various observations are an important tool to study the variation of ALTs at large spatial and temporal scales ( Lawrence et al., 2008 ,Lawrence et al., 2012 ). Guo and Wang (2017) obtained ALTs by using soil temperature interpolation simulated with version 4.5 of the Community Land Model (CLM4.5; Oleson et al., 2013 ), and used the results to study the distribution of permafrost and the variation of the ALTs.Gao et al. (2019) coupled the Stefan algorithm to CLM4.5 to simulate the dynamic changes in the depth of soil freeze-thaw fronts, as well as ALTs. However, there has been little research on the future variation trends and relative changes of ALTs in land surface models.

In this study, we adopted the Chinese Academy of Sciences Land Surface Model (CAS-LSM; Xie et al., 2018 ) and the Chinese Academy of Sciences Flexible Global Ocean-Atmosphere-Land System Model, gridpoint version 3 (CAS-FGOALS-g3) climate system model ( Li et al., 2020 ;Zhou et al., 2018 ) to simulate and study the variation trends and relative changes of ALTs due to climate warming.

2. Data and methods

2.1. Observational data

To evaluate the performance of the simulated ALTs in this work, we collected ALT observational data from the Circumpolar Active Layer Monitoring (CALM) network ( Brown et al., 2000 ;http://www.gwu.edu/～calm/ ). We selected 76 sites over the period 1991-2000. These observational data were used to validate the simulated ALTs over the same period. We also selected 54 sites over the period 1996-2010. These data were also used to validate the variation trends of the simulated ALTs. These observational data have been demonstrated to be reliable ( Guo et al., 2017 ; Gao et al., 2019 ).

2.2. CAS-LSM and CAS-FGOALS-g3

CAS-LSM is based on CLM4.5, which couples groundwater lateral flow and human water regulation ( Zeng et al., 2016 , 2017 ), freezing and thawing front schemes ( Gao et al., 2019 ), and river nitrogen transport processes ( Liu et al., 2019 ). CAS-LSM has been previously applied to study China’s Black River basin ( Xie et al., 2018 ). The distribution of groundwater, evapotranspiration, and frozen soil is well reproduced by CAS-LSM, which is coupled with the Stefan algorithm to simulate the dynamic changes of soil freeze-thaw front depth, as well as ALTs. Detailed descriptions of CAS-LSM can be found in the works of Xie et al. (2018) and Gao et al. (2019) .

The CAS-FGOALS-g3 climate system model consists of atmosphere, land, ocean, and sea-ice components linked by a coupler( Wang et al., 2004 ; Liu, 2010 ; Zhou et al., 2018 ; Yu et al., 2018 ;Li et al., 2020 ). The land component is simulated using CAS-LSM( Xie et al., 2018 ).

2.3. Experimental design

As described previously, we used CAS-LSM to simulate the ALTs.The atmospheric forcing dataset GSWP3 ( Kim et al., 2017 ) was used to drive CAS-LSM, which has a resolution of 0.9°×1.25°. GSWP3 is the default atmospheric forcing dataset for the offline LS3MIP ( van den Hurk et al., 2016 ) land simulations, which consists of a three-hourly global forcing product with a 0.5° longitude-latitude grid. The GSWP3 data were generated through the dynamical downscaling of the 20Century Reanalysis, version 2 ( Compo et al., 2011 ), using a spectral nudging technique. Bias corrections for temperature, precipitation, and longwave and shortwave radiation were conducted using the Climate Research Unit TS v3.21, Global Precipitation Climatology Center v7, and Surface Radiation Budget datasets, respectively. To balance the model,a 20-year atmospheric forcing data cycle was used, which was spun up for 100 years. The simulation period was 1850-2014.

We also used CAS-FGAOLS-g3 to simulate the future variation trends and relative changes of the ALTs. We adopted the CMIP6 SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5 climate change scenarios( Eyring et al., 2016 ). The atmosphere and land resolution for these simulations was 2° in both latitude and longitude. The simulation period was 2015-2100.

2.4. Methods

It should be noted that the definition of permafrost used in this paper is near-surface permafrost. Near-surface permafrost is identified as ground where the monthly soil temperature is less than 0°C for 24 consecutive months in at least one layer of the upper 10 soil layers (10 layers equates to a depth of 3.8 m in LSMs) ( Lawrence et al., 2012 ; Guo and Wang, 2017 ). As near-surface permafrost is more sensitive to climate change than deep permafrost ( Lawrence et al., 2008 ), near-surface permafrost has been used to study the sensitivity of permafrost to climate change ( Lawrence et al., 2012 ; Guo and Wang, 2017 ). The ALTs considered here are the maximum thaw depths of permafrost over the course of a year ( Lawrence et al., 2012 ), which were simulated based on CASLSM coupled with the Stefan algorithm ( Gao et al., 2019 ) in this paper.In general, ALTs are calculated based on soil temperature interpolation simulated with LSMs ( Lawrence et al., 2012 ; Guo and Wang, 2017 ). We also calculated ALTs based on soil temperature interpolation and compared the ALTs derived from both methods.

3. Results

3.1. Model validation

Firstly, the simulated ALTs were validated against observations made at the CALM sites. Notably, the ALT values of the model grid cells were interpolated to each station’s location using bilinear interpolation.

As described previously, we obtained ALTs via CAS-LSM simulations and via soil-temperature interpolation ( Guo and Wang, 2017 ). Fig. 1 shows the simulated (blue points), soil-temperature-interpolated (red points), and observed (black lines) ALTs. From 1991 to 2000, the CASLSM-simulated and soil-temperature-interpolated ALTs are consistent with the observed values, although they do show larger thicknesses compared with the observed values ( Fig. 1 ). In Fig. 1 (a), the mean bias (MB)between the simulated and observed ALTs is 0.49 m, the root-meansquare error (RMSE) is 0.88 m, and the correlation coefficient (CC) is 0.53. The MB between the interpolated and observed ALTs is 0.42 m,the RMSE is 0.90 m, and the CC is 0.51. The MB between the ALTs obtained by soil temperature interpolation and model simulation is 0.11 m,the RMSE is 0.32 m, and the CC is 0.85. It can be seen from the MBs,RMSEs, and CCs that the simulated ALTs from these two methods are similar, which indicates that the simulated ALTs from CAS-LSM are reasonable. In Fig. 1 (b), the CAS-LSM-simulated ALTs, soil-temperatureinterpolated ALTs, and observed ALTs show increasing trends over the period 1996-2010, with interannual change rates of 0.4, 0.2, and 0.6 cm yr, respectively.

Fig. 1. (a) Scatterplot of observed and simulated ALTs from 1991 to 2000. (b) Annual variation of observed and simulated ALT from 1996 to 2010, including simulated values from CAS-LSM (blue), soil temperature interpolation (red), and observed ALTs (black).

3.2. Historical variation trends and relative changes of ALT

We next analyzed the spatial distributions, relative changes, and variation trends of the ALTs over the period 1979-2014. The extent of permafrost was found to decrease due to climate change, although not by much over the period 1979-2014. Therefore, we fixed the extent of permafrost to study historical ALTs, as had been done in a previous study( Guo and Wang, 2017 ). The variation trends and relative changes of the ALTs from 1979 to 2014 over the simulated permafrost region (in 1979),where permafrost remained in each year during the period 1979-2014,are shown in Fig. 2 . As can be seen from Fig. 2 (a), in addition to the Tibetan Plateau, the simulated ALTs decreased with an increase in latitude, and the ALTs were deeper in regions near the boundaries between permafrost and seasonal frozen soil. The average ALT was 1.08 m and the largest ALT was 3.8 m in near-surface permafrost. Fig. 2 (b) shows the spatial distributions of 2 m air temperature. It should be noted that the 2 m air temperature is that of summer because the depth of thawing reaches its maximum in summer. The ALTs in Fig. 2 (a) and the 2 m air temperatures in Fig. 2 (b) show similar spatial distributions.

Figure 2 (c, d) show the simulated ALT variation trends (cm yr) and 2-m temperature variation trends (°C yr) from 1979 to 2014. The ALT variation trends are basically positive, showing an increasing trend yearby-year. The regionally averaged variation trend is 0.33 cm yr, while the trends for the Tibetan Plateau, North America, and northern Eurasia are 0.39, 0.31, and 0.35 cm yr, respectively. As can also be seen from Fig. 2 (d), the variation trend of the 2 m air temperature is basically positive, i.e., it has an increasing trend of 0.036°C yr. As the 2 m air temperature increases, the ALTs also increase. Next, Fig. 2 (e) shows the changes of ALTs in 2014 relative to 1979, which were calculated as variation trends ×(2014-1979 + 1) / 1979 in values of ×100%. The averaged relative change over the entire permafrost region is 10.9%.Fig. 2 (f) shows the spatial distribution of the CC between the simulated ALTs and 2 m air temperature, for which the regional average is 0.61.This indicates that climate change has a substantial impact on the ALT of permafrost.

3.3. Future variation trends and changes of ALTs under different scenarios

Next, we analyzed the variation trends and relative changes of ALTs over an 86-year future period (2015-2100). As described in sections 2.2 and 2.3 , CAS-FGOALS-g3 was used to simulate and predict the future variation trends and changes of ALTs under different climate scenarios (SSP-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) during this period.

The extent of permafrost was found to change considerably over the period 2015-2100. Thus, we predicted future ALTs in the simulated permafrost region. Based on the previous definition of near-surface permafrost, we chose two consecutive years (2099-2100) to simulate the permafrost region. The near-surface permafrost was found to undergo significant degradation at the end of this century due to climate change. The area of permafrost was reduced by 11%, 24%, 52%, and 61%, respectively, for the four hypothetical scenarios, in 2100 relative to 2015. The relative changes of ALTs are in 2100 relative to 2015,which were calculated as variation trends ×(2100-2015 + 1) / 2015 in values of ×100%. Fig. 3 (a-d) show the changes of the ALTs in 2100 relative to 2015. The averaged relative changes are 10.3%, 14.6%, 30.1%,and 51%, respectively, for the four hypothetical scenarios. The relative changes of ALT under the SSP8.5 scenario were the largest, while under SSP2.6 they changed the least.

Next, Fig. 3 (e-h) shows the simulated ALT variation trends (cm yr)under the different scenarios from 2015 to 2100. The variation trends are generally positive, showing increasing trends year-by-year. The averaged variation trends are 0.11, 0.31, 0.42, and 0.65 cm yr, respectively, for the four hypothetical scenarios. The variation trends under SSP5-8.5 are the largest, while under SSP1-2.6 they are the smallest.These results indicate that climate change may seriously influence the ALTs of permafrost in the future.

4. Discussion and conclusion

In this study, we used CAS-LSM to study the historical and future variation trends and changes of the ALTs of near-surface permafrost due to climate change. We were chiefly concerned with the variation trends and changes of the ALTs of near-surface permafrost, which is highly sensitive to climate change. CAS-LSM was coupled with the Stefan algorithm to simulate the ALTs in this paper. However, the CAS-LSMsimulated ALTs showed larger thicknesses compared with the observed values from CALM. The primary reason for this difference may be that the Stefan equation assumes that all absorbed or released energy by the soil is used in the transformation of soil water, and it ignores the sensible heat energy arising from temperature changes in the soil ( Jumikis, 1977 ;Woo et al., 2004 ). Thus, we still need to improve the permafrost parameterization schemes in LSMs to provide more accurate simulations of ALTs.

Fig. 2. (a) Spatial distribution of the simulated mean ALTs. (b) Spatial distribution of the 2 m air temperature. (c) Simulated variation trends of the ALTs (units: cm yr − 1 ). (d) Simulated variation trends of the 2 m air temperature (°C yr − 1 ). (e) Changes in the simulated ALTs in 2014 relative to 1979. (f) Spatial distribution of the correlation coefficients between the simulated ALTs and 2 m air temperature. The period shown here is from 1979 to 2014. The shaded areas indicate regions where the difference passed the 95% confidence level of the Student’s t -test.

Fig. 3. Predicted changes of the ALTs in 2100 relative to 2015, and variation trends (cm yr − 1 ) of the ALTs under the climate change scenarios SSP-2.6, SSP2-4.5,SSP3-7.0, and SSP5-8.5. The period is from 2015 to 2100. The shaded areas indicate regions where the difference passed the 95% confidence level of the Student’s t -test.

From our analysis, we have drawn the following conclusions:

(1) The simulated ALTs from CAS-LSM and the soil-temperatureinterpolated values were similar, and were close to the observed values, which indicates that the simulated ALTs produced by CAS-LSM are reasonable.

(2) We obtained the spatial distributions, relative changes, and variation trends of the ALTs from 1979 to 2014. The regionally averaged ALT was 1.08 m, the regionally averaged relative change was 10.9%, and the regionally averaged variation trend was 0.33 cm yrover the entire permafrost region. As the 2 m air temperature increased, the ALTs also increased.(3) We also predicted the relative changes and variation trends of the ALTs of permafrost over an 86-year future period (2015-2100).The regionally averaged variation trends were 0.11, 0.31, 0.42, and 0.65 cm yr, respectively, for the four hypothetical scenarios (SSP-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5). The regional averaged relative changes in 2100 relative to 2015 were 10.3%, 14.6%, 30.1%,and 51%, respectively, for the four hypothetical scenarios. The relative changes of the ALTs under SSP5-8.5 were the largest, while under SSP1-2.6 they were the smallest.

Our results indicate that climate change has a significant impact on the ALTs of permafrost, and this has the potential to further our understanding of the responses of ALT to climate change.

Funding

This research was supported by the National Key R&D Program of China [grant number 2018YFC1506602 ], the Key Research Program of Frontier Sciences, CAS [grant number QYZDY-SSW-DQC012 ],and the National Natural Science Foundation of China [grant number 41830967 ].