Remote Sensing of Environment 268 (2022) 112778

Available online 3 November 20210034-4257/© 2021 Elsevier Inc. All rights reserved.

Magnitudes and patterns of large-scale permafrost ground deformation revealed by Sentinel-1 InSAR on the central Qinghai-Tibet Plateau

Jie Chen a, Tonghua Wu a,b,*, Defu Zou a, Lin Liu c, Xiaodong Wu a, Wenyu Gong d, Xiaofan Zhu a, Ren Li a, Junming Hao e, Guojie Hu a, Qiangqiang Pang a, Jing Zhang a, Sizhong Yang a

a Cryosphere Research Station on the Qinghai-Tibet Plateau, State Key Laboratory of Cryospheric Science, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China b Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou), 511458, China c Earth System Science Programme, Faculty of Science, The Chinese University of Hong Kong, Hong Kong SAR, China d State Key Laboratory of Earthquake Dynamics, Institute of Geology, China Earthquake Administration, Beijing 100029, China e School of Civil Engineering, Lanzhou University of Technology, Lanzhou 730050, China

A R T I C L E I N F O

Edited by Dr. Menghua Wang

Keywords: Permafrost ground deformation InSAR Ground ice loss Large scale Qinghai-Tibet Plateau

A B S T R A C T

Permafrost on the Qinghai-Tibet Plateau (QTP) undergoes significant thawing and degradation, which affects the hydrological processes, ecosystems and infrastructure stability. The ground deformation, a key indicator of permafrost degradation, can be quantified via geodetic observations, especially using multi-temporal InSAR techniques. The previous InSAR studies, however, either rely on data-driven models or Stefan-equation-based models, which are both lacking of consideration of the spatial-temporal variations of freeze-thaw processes. Furthermore, the magnitudes and patterns of the permafrost-related ground deformation over large scales (e.g., 1 × 105 km2 or larger) is still insufficiently quantified or poorly understood. In this study, to account for the spatial heterogeneity of freeze-thaw processes, we develop a permafrost-tailored InSAR approach by incorpo-rating a MODIS-land-surface-temperature-integrated ground deformation model to reconstruct the seasonal and long-term deformation. Utilizing the approach to Sentinel-1 SAR images on the vast regions of about 140,000 km2 of the central QTP during 2014–2019, we observe widespread seasonal deformation up to about 80 mm with a mean value of about 10 mm and linear subsidence up to 20 mm/year. We apply the geographical detector to determine the controlling factors on the permafrost-related deformation. We find that the slope angle is the primary controller on the seasonal deformation: strong magnitudes and variations of seasonal deformation are most pronounced in flat or gentle-slope regions. The aspect angle, vegetation and soil bulk density exhibit a certain correlation with seasonal deformation as well. Meanwhile, we find that a linear subsidence is higher in the regions with high ground ice content and warm permafrost. It indicates that warm and ice-rich permafrost regions are more vulnerable to extensive long-term subsidence. We also observe that the cold permafrost regions experience lower linear subsidence even with high ground ice content, which indicate ice loss is limited. Thus, we infer that under continuously warming, the transition from cold permafrost to warm permafrost may lead to more extensive ground ice melting. Moreover, the strong subsidence/uplift signals surrounding some lakes suggesting that the change of local hydrological conditions may induce localized permafrost degradation/ aggradation. Our study demonstrates the capability of the permafrost-tailored InSAR approach to quantify the permafrost freeze-thaw dynamics as well as their spatial-temporal patterns over large scales in vast permafrost areas.

1. Introduction

Permafrost-on the Qinghai-Tibet Plateau (QTP)-is vulnerable to

thawing and degrading under climate warming and extreme warm events. The total extent of underlying permafrost on the QTP is approximately 1.06 × 106 km2, i.e., 40% of the entire QTP(Zou et al.,

* Corresponding author at: Cryosphere Research Station on the Qinghai-Tibet Plateau, State Key Laboratory of Cryospheric Science, Northwest Institute of Eco- Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China.

E-mail addresses: chenjie@lzb.ac.cn (J. Chen), thuawu@lzb.ac.cn (T. Wu).

Contents lists available at ScienceDirect

Remote Sensing of Environment

journal homepage: www.elsevier.com/locate/rse

https://doi.org/10.1016/j.rse.2021.112778 Received 1 May 2021; Received in revised form 26 October 2021; Accepted 27 October 2021

Remote Sensing of Environment 268 (2022) 112778

2

2017; Cao et al., 2019). The permafrost extent on the QTP had shrunk by around 0.24 × 106 km2 from 1975 to 2006 (Jin et al., 2011). The permafrost temperature on the QTP was within about 4 ∘C below the freezing point of water and it increased by 0.1–0.5 ∘C since the 1980s (Wu and Zhang, 2008; Zhao et al., 2020). Thawing permafrost exposes a large amount of frozen soil carbon vulnerable to microbial breakdown, which consequently contributes to climatic warming through positive feedback (Koven et al., 2011; Schuur et al., 2015; Mu et al., 2020). Permafrost thaw processes have strong impacts on the local hydrological condition (Liljedahl et al., 2016; Walvoord and Kurylyk, 2016; Zongxing et al., 2019), the ecosystems (Jorgenson and Osterkamp, 2005; Nauta et al., 2015; Wu et al., 2018), and thermokarst processes (Farquharson et al., 2019; Nitze et al., 2020). In addition, irregularly or unevenly thawing permafrost is threatening the infrastructure (such as buildings, railways, and airports), which is unprecedentedly challenging the local economic and social sustainability, climate adaptation, and human well-being (Nelson et al., 2001; Hjort et al., 2018; Wu et al., 2020). However, owing to the subsurface characteristics of permafrost thawing processes, it is difficult to directly observe or quantify them over extensive spatial coverage and therefore reveal the spatial-temporal dynamics of the active layer and underlying permafrost, especially in the remote and high-elevation permafrost regions on the QTP.

The gradual ground subsidence and uplift is attributed to progressive freeze and thaw in permafrost environment. The ground subsides seasonally in association with the thawing and volumetric contraction of the active layer due to the phase transition of frozen water to a liquid state in summer. Conversely, ground uplift occurs due to the opposite water phase transition from liquid to solid state. Under warming or local disturbance, the excess ice or ice-rich sediment upon thawing in the uppermost permafrost results in additional long-term ground subsi-dence, which could be regarded as an indicator of permafrost degradation.

Field-based instruments have been designed to measure the thaw subsidence and/or freeze uplift in permafrost environments, including but not limited to frost tube (Mackay, 1973a), leveling (Mackay, 1973b; Mackay and Burn, 2002), and differential Global Positioning System (GPS) (Little et al., 2003; Streletskiy et al., 2017). More recently, GPS interferometric reflectometry (GPS-IR) has been used to measure the decadal seasonal deformation over permafrost regions (Liu and Larson, 2018; Hu et al., 2018). A new tilt-arm-based apparatus with a logging inclinometer is designed to monitor hourly thaw subsidence and frost heave (Gruber, 2020). Due to the harsh environment, the field-based observations are mostly with limited spatial or temporal resolution and thus inadequate for revealing the spatial patterns of ground defor-mation in local or regional scale. In addition, differential digital eleva-tion models (DEMs), obtained from repeat-pass optical and or radar remote sensing techniques, have widely utilized to measure the ground deformation mostly with large magnitude (Kaab, 2002; Günther et al., 2015; Zwieback et al., 2018; Bernhard et al., 2020), but it is difficult to detect the gradual ground deformation with millimeter to centimeter order.

Repeat-pass Interferometric Synthetic Aperture Radar (InSAR) has been successfully used to capture the spatial patterns of permafrost- related ground elevation changes in local (Liu et al., 2010; Chen et al., 2020) and regional scales (Daout et al., 2017, 2020). To isolate the permafrost-related ground deformation, Li et al. (2015) and Daout et al. (2017) applied a sinusoidal deformation model without considering the dynamic processes of freezing and thawing. However, the data-driven models are vulnerable to measurement noises and any deficiencies, such as strong atmospheric errors and strong freeze-thaw-induced decorrelation, especially when there are limited SAR acquisitions. Liu et al. (2010) and Chen et al. (2018b) adapted the one-dimensional Stefan equation to account for thaw subsidence and freeze uplift. However, the Stefan-based studies do not consider the spatial variations on either the temperatures or the freeze-thaw onsets, which are the primary factors controlling the timing and magnitude of freeze-thaw processes.

Moreover, large-scale ground deformation and its controls on the vast permafrost regions of the central QTP are still poorly quantified or understood.

The objective of this study is to decompose the permafrost-related ground deformation from InSAR measurements and reveal the controls of the observed ground deformation in a regional scale over the high- altitude permafrost regions of the central QTP. To account for the spatial heterogeneity of freeze-thaw processes, a permafrost-tailored InSAR ground deformation approach had been developed by incorpo-rating MODIS land surface temperature (LST), soil properties and Stefan equation. We also revealed the spatial-temporal characteristics of both seasonal and long-term deformation over large scales.

2. Study area and field observations

Our study area is located on the central QTP from northern Kunlun Mountain to the southern boundary of permafrost zone and seasonally frozen ground, along the Qinghai-Tibet Highway (QTH) (Fig. 1a). The length and width of our study area are about 580 km and 250 km, respectively, covering around 1.45 × 105 km2, i.e., 10% of the total permafrost areas on the QTP. The elevation varies from around 4300 m to 6100 m. The alpine meadow, steppe and desert occupy around 90% of the study area. The mean annual air temperature ranges from − 5.8 ∘C to − 2.4 ∘C, where the average warming rate is around 0.55 ∘C per decade. The annual precipitation ranges from 210 mm to 580 mm, with an increasing rate of 4–10 mm/year during 2006–2016 (Zhao et al., 2020). Active layer thickness ranges from 1.2 m to more than 5 m in our study area (Zhao et al., 2020). The downward ground thawing usually starts in mid or late May and ends in late September or early November. Overall, the climate of the QTP has become warmer and wetter, whereas the wetter and warmer trends on the central QTP is larger than the average trend of the entire QTP (Zhao et al., 2020; Sun et al., 2020).

3. Datasets and methods

3.1. Land surface temperature (LST)

MODIS LST is one of the best LST products especially on the QTP, which has been used to map permafrost occurrence and active layer thickness (Zou et al., 2017). The 1 km gridded clear-sky MOD11A2 (Terra MODIS) and MYD11A2 (Aqua MODIS), obtained every 8 days from 2014 to 2019, are used to calculate the mean daily LST. The missing values of MOD11A2/MYD11A2 are identified and filled by using pixel-wise harmonic analysis time-series approaches (Roerink et al., 2000; Xu et al., 2013). As the 8-day MODIS LSTs are instantaneous observations, an empirical multi-linear regression model is used to generate the mean daily LST (Zou et al., 2014). More details about how to handle the MODIS LST product can be found from Zou et al. (2014, 2017). The mean daily LST before and after the SAR acquisition day are then linearly interpolated to generate the mean daily LST on each SAR acquisition day.

3.2. Freeze and thaw onsets

The freeze and thaw onsets, which vary significantly cross the QTP, determine the timing of ground deformation occurring over the permafrost regions. However, there are very limited freeze-thaw onset products with high spatial resolution over the study area. Despite the 8- day MODIS LSTs encounter low temporal resolution, we choose to use them in determining the freeze and thaw onset due to its high spatial resolution (1 km). We assume that at any given locations the freeze and thaw onsets remain constant during the study period. The thaw and freeze onsets are then determined by averaging the onsets in each year.

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

3

3.3. Soil thermal conductivity

In this study, we adopt the empirical Johansen parameterization of soil thermal conductivity (κ) (Johansen, 1977; Du et al., 2020):

κ =

{(κsat − κdry)Ke + κdry if Sr > 1 × 10− 5

κdry if Sr ≤ 1 × 10− 5 (1)

where Sr is the degree of saturation; Ke is the “Kersten number”, which is a function of Sr; κsat and κdry are the saturated and dry soil thermal conductivities, respectively. Soil thermal conductivity is mainly deter-mined by soil properties, including soil bulk density, porosity, quartz content. The soil properties can be acquired from gridded soil products at 250 m resolution (Hengl et al., 2017). We determine the averaged soil water content in active layer for both freeze and thaw seasons by using vegetation type and soil information. We resample the soil bulk density and sand fraction as well as the soil water content to the geocoded SAR geometry. Then, the soil thermal conductivity in freeze and thaw sea-sons are calculated pixel-by-pixel based on Eq. (1), respectively. More details about how to determine the soil thermal conductivity can be found in Du et al. (2020).

3.4. Ground ice content map

A ground ice content map over the permafrost regions of the QTP is developed (Cheng et al., 2019), which is mainly determined by borehole records of ground ice content, Quaternary sedimentary type, the permafrost distribution map (Zou et al., 2017) and the permafrost thickness map (Zhao et al., 2020). The total ground ice content is around 12.7 × 103 km3. As the ALT on the QTP is generally 2–4 m, we choose to use the 0–5 m volumetric ground ice content (VIC) without including the seasonal ice content in the active layer.

3.5. Vegetation type and NDVI

Wang et al. (2016) produced a 1 km × 1 km vegetation map of the entire QTP based on MODIS and field investigation. Except for the barren land, water bodies and glaciers, the vegetation-covered ground is classified into four groups: alpine swamp meadow, alpine meadow, alpine steppe and alpine desert. The 8-day 250 m MODIS-NDVI dataset is used. To mitigate the effects from different growing time and possible artifacts, we average NDVI acquired between 1 August and 15 August in 2019, when most of the vegetation covers reach their maximum in our study area.

Fig. 1. Study area and field photos. (a) Esri maps of our study area. The red rectangle outlines the coverage of Sentinel-1 SAR images used in our study. The dark red star shows the position of the 2001 Mw 7.8 Hoh Xil earthquake. The red triangles denote the locations of active layer monitoring stations. The black lines are the Kunlun and Yushu–Fenghuoshan (YF) fault. The thicker black line denotes part of the QTH within our study area; (b) The red rectangle shows the position of our study area on the QTP. The background figure is the permafrost distribution map (Zou et al., 2017); (c) Ground photos show different vegetation covers (alpine swamp meadow, alpine meadow and alpine steppe) on the active layer observational stations. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

4

3.6. Ground elevation change model

Here we adapt a piecewise elevation change model (PECM), devel-oped by Chen et al. (2018b), to combine the long-term and seasonal elevation changes. Stefan-based equation is used to account for the seasonal elevation changes by integrating the freezing and thawing processes. Without loss of generality, we define the whole freeze-thaw cycle as a consecutive permafrost year from one thaw onset until the next thaw onset. In the PECM, Chen et al. (2018b) used the square root of the accumulated degree days of thaw/freeze (ADDT/F) to account for the seasonal thaw subsidence and freeze uplift as well as a linear sub-sidence to represent permafrost thaw subsidence, respectively. ADDT and ADDF are derived based on the Stefan equation (Nelson et al., 1997), which is widely used to estimate the depth of the freeze and thaw front. There are mainly two underlying assumptions: (1) within one freeze-thaw cycle, the seasonal thaw subsidence is equal to freeze uplift; (2) the freezing front only moves from the top downwards without considering the down upwards freezing processes. The piecewise elevation change D at any given time τ can be expressed as (Chen et al., 2018b):

D(τ) = vtrend⋅L(τ) + dseason⋅I(τ) (2)

where the coefficient of linear L(τ) and seasonal deformation I(τ) are:

L(τ) ={

T − 1 + (τ − τt) if τt ≤ τ ≤ τfT if τ > τf

(3)

I(τ) =At(τ)

√−

κf

κt

√

Af (τ)√

(4)

where T is the number of consecutive permafrost years (in units of years); κt and κf are the soil thermal conductivities in the thaw and freeze seasons, respectively; At and Af are the ADDT and ADDF (∘C days), respectively. I(τ) is the composite index to account for the seasonal thaw subsidence and freeze uplift. The composite index is assigned to zero when Iτ is positive, i.e. when the active layer is totally frozen. Moreover, if Iτ,max is larger than zero, κf/κt is adjusted to set Iτ,max as zero. Other-wise, there would be an offset between the end of freezing season and the next onset of thawing season, which is implausible. To account for the spatial heterogeneity of freeze-thaw processes, the ADDT and ADDT are calculated pixel-by-pixel by using the oversampled MODIS LST from intervals of 8 days to 1 day.

3.7. Small baseline subset InSAR

The small baseline subset (SBAS) approach is based on an optimal combination of small-baseline interferograms by selecting image pairs with small temporal and spatial baselines, which are therefore less affected by neither geometric or temporal decorrelation (Berardino et al., 2002). In this study, the SBAS approach is used to retrieve the phase time series. Then, the linear and seasonal ground elevation change maps are decomposed by integrating the MODIS-LST-integrated eleva-tion change model.

The Sentinel-1 SAR images have been acquired every 24 days since October 2014 and every 12 days since around March 2017 in our study area. The SAR dataset includes 4 consistent frames along the same descending orbital track. We choose the SAR image acquired on 31 August 2018 as the common master image, whereas all other images are coregistered to the master geometry. We construct interferograms with temporal baseline not exceeding 40 days (Fig. 2). We also construct the inter-annual image pairs with temporal baseline thresholds in between 350 to 380 days and 725 to 735 days without considering spatial baseline threshold, respectively (Fig. 2). This is because the spatial baseline of Sentinel-1 is on the order of 150 m, making geometric decorrelation ignorable (Yagüe-Martínez et al., 2016). The inter-annual interferogram pairs are aimed to mitigate the effects of low coherence due to strong surface freeze-thaw processes as well as the shallow snow cover in winter (Chen et al., 2018a). Then, we generate the interfero-grams using the InSAR Scientific Computing Environment (ISCE) (Rosen et al., 2012). In addition, the water bodies and the glaciers are masked out. We use the spatial complex multilooking, i.e., averaging a number of complex pixels within a boxcar window, to mitigate the phase noise. The range and azimuth look numbers are 20 and 6, respectively, generating the ground pixels of approximately 90 × 90 m. The 3-arc-sec-ond Shuttle Radar Topography Mission (SRTM) DEM product is used to remove the topographic phase of each interferogram. We further exclude the interferograms with extremely low coherence via manually checking. The perpendicular and temporal baselines of the remaining 466 interferograms are presented in Fig. 2. After applying a power-spectrum adaptive filter (Goldstein and Werner, 1998), we finally unwrap all the interferograms by using the minimum cost flow approach (Chen and Zebker, 2001). Since we only focus on the permafrost regions in this study, the non-permafrost areas are masked out based on the permafrost distribution map (Zou et al., 2017).

The tropospheric artifacts have been regarded as one of the main error sources especially in large-scale InSAR measurements. This is because that the tropospheric artifacts may contaminate the ground

Fig. 2. The connections between Sentinel-1 SAR acquisitions with the perpendicular and temporal baselines. The light blue triangle denotes the Sentinel-1 SAR acquisitions. The gray lines denote the interferogram pairs. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

5

deformation over large scales. In general, tropospheric artifacts include the stratified and turbulent components. Different strategies and ap-proaches can be used to mitigate: (1) the turbulent components by using the spatial-temporal filtering without external information (Ferretti et al., 2001); (2) the stratified or topographic-related components by using the correlation approach between InSAR phase and topography (Delacourt et al., 1998; Zebker, 2021); (3) the stratified and turbulent components by integrating the external data, such as numerical weather models (e.g. WAF, (Jolivet et al., 2011)), remote sensing observations (e. g. MODIS), and geodetic measurements (e.g. GPS). In this study, we mitigate the tropospheric decays by integrating ERA5 reanalysis data (Jolivet et al., 2011; Zhang et al., 2019). Then, the residual tropospheric artifacts are mitigated by using a linear deramping approach. The SB interferograms (φm,n) can be expressed by phase time-series relative to the specific master image (s):

φm,n = φs,m − φs,n ∀m, n = 1,…,N + 1 (5)

For each pixel, let φT = [φ1, …, φM] be the vector of M SB interfer-ometric phases and ϕT = [ϕ1, …, ϕN] be the vector of the N equivalent single-master phases:

φTM×1

= BM×N

ϕTN×1

(6)

where B is the linear design matrix related to the SB construction. The solution of Eq. (6) can be obtained by minimizing the L2-norm of the residual vector (Berardino et al., 2002). In the case of all the SAR images are connected, B is in full column rank. The equivalent single-master phase can be estimated by simple least square methods. To mitigate the possible errors from spatial phase unwrapping, the more robust L1-norm minimization is used to retrieve phase time series without any assumption on ground subsidence in this study (Lauknes et al., 2011; Goel and Adam, 2012). For full-rank B, the L1-norm problem can be solved based on iteratively reweighted least squares method. To avoid the disconnection between interferogram pairs, we use the coherence-weighted approach to make full usage of all the interfero-grams (Tong and Schmidt, 2016).

InSAR-derived ground deformation may contain permafrost-related deformation, groundwater extraction, slope movement, tectonic move-ment and so on. It is difficult to decompose these deformation sources using InSAR measurements alone. However, in permafrost environment at large scales, we expect that the seasonal deformation is dominated by the freezing and thawing of the active layer as the temporal evolution of other deformation sources has no obvious seasonality. Given the PECM, the pixel-by-pixel linear and seasonal deformation are decomposed by incorporating Eq. (2) into Eq. (6).

ϕTN×1

= AN×2

[vtrend, dseason]T

where : A = −4πλ

[L(τ1) L(τ2) ⋯ L(τN)

I(τ1) I(τ2) ⋯ I(τN)

]T (7)

where A is the coefficient matrix corresponding to the linear and sea-sonal deformation, which can be calculated via Eq. (3) and (4).

The phase residuals are calculated by subtracting the linear and seasonal deformation from the InSAR deformation time series for all pixels. The uncertainty of linear and seasonal deformation are then calculated through the variance-covariance matrix (Chen et al., 2018b).

3.8. Determine potential controls for the ground deformation

The magnitudes and patterns of ground deformation on the perma-frost regions is closely correlated to the freeze-thaw processes. Previous studies to determine the controlling factors on ground deformation in permafrost regions are mainly based on correlation analysis (Daout et al., 2020; Chen et al., 2020), which can hardly reveal the spatial patterns of ground deformation and determine the relative importance

of each environmental factors. The geographical detector is a spatial variance analysis tool to assess the relative importance of different components controlling a geographic phenomenon, which is intrinsi-cally immune to collinearity (Wang et al., 2010; Fang et al., 2021). This method has been successfully adapted to determine factors controlling for mapping landslide susceptibility (Luo and Liu, 2018; Yang et al., 2019). For the first time we apply the geographical detector to assess the contributions of each environmental factor on permafrost-related ground deformation. The core idea behind the geographical detector is that the similarity of spatial patterns between ground deformation and relevant environmental factors are determined by the q-statistic, which is determined by the local and global variances (Wang et al., 2010). The value of q-statistic is between 0 to 1, whereas 1 means the factor totally controls the spatial patterns of ground deformation and vise versa.

Seasonal deformation in permafrost regions is mainly controlled by the variations of soil consolidation attributed to total water content and soil properties in the active layer. As the subsurface total water content is challenging to estimate over large scales, there are many proxies relating to the total water content, such as terrain factors, vegetation cover, precipitation and so on. Moreover, permafrost-related linear deformation are relying on how much ground ice is melted and drained. The proxies factors corresponding to linear deformation include the uppermost ground ice content, the magnitude of energy flux, the ALT as well as the terrain factors. Thus, we identify 14 potential environmental factors including longitude, latitude, elevation, slope angle, aspect, maximum degree days of thawing (DDT), maximum degree days of freezing (DDF), DDT to DDF ratio, mean annual ground temperature (MAGT), normalized difference vegetation index (NDVI), vegetation type, 5-m VIC, soil bulk density, and sand fraction. As seasonal and linear deformation may be correlated with each other, we include the seasonal deformation as the 15th factor when conducting geographical detector to linear deformation and vice versa.

The terrain factors, including elevation, slope angle and aspect, are calculated by using 3-arc-second SRTM DEM. The calculation of other factors are mentioned in Sections 3.1 and 3.5. All the factors are reprojected and resampled to the same grid and geometry with InSAR deformation maps.

4. Results

4.1. Ground deformation maps

The Line-Of-Sight (LOS) linear deformation and amplitude of sea-sonal deformation are shown in Fig. 3. All the ground deformation mentioned throughout the paper are the ground deformation in the LOS direction. The linear deformation ranges from around − 15 mm/year to 15 mm/year. The uncertainty of the linear deformation is mostly less than 0.5 mm/year (Fig. S1a). A negative value means that the ground moves away from the satellite along the LOS direction, i.e. ground subsidence; the positive values are moving toward the satellite, i.e. ground uplift. We observe obvious nonlinear patterns along the lat-itudinal direction. We also observed significant inter-annual ground uplift surrounding the Zonag, Cuodarima and Cuorendejia lakes. Meanwhile, many clusters of significant ground subsidence are pro-nounced mostly in the southern and eastern regions.

The maximum amplitude of seasonal deformation reaches up to around 80 mm, whereas the average value is around 10 mm. The un-certainty of the seasonal deformation is mostly less than 5 mm (Fig. S1b). Overall, we do not observe obvious spatial correlation neither in the latitudinal or longitudinal direction. Compared to the linear deformation, the amplitudes of seasonal deformation are larger and more widespread.

4.2. Magnitudes and patterns of ground deformation

By conducting the geographical detector analysis on the selected

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

6

Fig. 3. Maps of (a) the LOS linear deformation; (b) the amplitude of LOS seasonal deformation. The red triangle marks the position of our InSAR reference point. The red rectangles R1, R2 and R3 denote three local regions, which will be used for localized deformation analysis in Section 4.2.4. The black dashed line PP’ and QQ’ marks the profile shown in Fig. 7 and Fig. 5c, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. The q-statistic of environmental factors on controlling (a) the linear deformation; (b) the amplitude of seasonal deformation. For the abbreviations in x-axis labels, longitude (lon), latitude (lat), 0–5 m ground ice content (ice_content), seasonal deformation (defo_season), maximum DDT (ddt_max), maximum DDF (ddf_max), DDT to DDF ratio (ddt_ddf_ratio), linear deformation(defo_rate).

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

7

environmental factors, we obtain the relative contributions of each factors to the total spatial variance of the linear and seasonal deforma-tion, respectively (Fig. 4). We find that slope angle and latitude are mainly responsible for driving the spatial patterns of seasonal defor-mation within our study area. Latitude is the most significant factor for driving the spatial patterns of the linear deformation, which resemble the latitudinal patterns observed in the linear deformation map.

4.2.1. Seasonal deformation To further exploit the relationship between the seasonal deformation

and slope angle, we average the amplitude of seasonal deformation in each binning with slope angle of 1∘, respectively (Fig. 5a). The box plot highlights a strong correlation between seasonal deformation and slope angle, i.e., decreasing seasonal deformation with increasing slope angle between 0∘ and 5∘. It is followed by an unchanging pattern of seasonal deformation with slope angle between 9∘ and 15∘. Between the slope angle of 16∘ to 30∘ or higher, slightly increasing seasonal deformation is observed. The weak correlation denotes that slope angle may not be the primary controls on the seasonal deformation when the slope angle is larger than 5∘. The maximum magnitude of seasonal deformation is around 30–50 mm with slope angle smaller than 2∘, 20–30 mm with slope angles between 2∘–5∘ and 15–20 mm with slope angle larger than 5∘. More specifically, Fig. 5c shows that the valleys or basins between mountain ranges tend to experience larger seasonal deformation, consistent with the spatial patterns of seasonal deformation in Qilian mountain reported by Daout et al. (2020). This phenomenon is common and widespread in our study area.

The amplitude of seasonal deformation exhibits a positive relation-ship with the aspect angle (Fig. 5b). Due to the high variability of aspect angles in flat regions, we exclude the pixels with slope angle smaller than 5∘. The seasonal deformation on the north-facing slopes is pro-nounced about 1 mm larger than that on the west- and east-facing slopes, and 2 mm larger than that on the south-facing slopes.

Fig. 6a represents the correlation between seasonal deformation and

NDVI. The box plot denotes decreasing seasonal deformation with NDVI increasing from 0.1 to 0.35, whereas the amplitude of seasonal defor-mation is in a wider range up to around 40 mm. There is a slightly increasing trend associated with seasonal deformation when NDVI is larger than 0.4, while the amplitude of seasonal deformation is in a narrower range up to 20 mm. A decreasing trend of seasonal deforma-tion is observed with increasing soil bulk density (Fig. 6b). The median value of seasonal deformation with soil bulk density of 1250 kg/m3 is around 11 mm and is about 7 mm larger than that of soil bulk density of 1500 kg/m3. More specifically, the seasonal deformation shows a weak correlation with soil bulk density from 1250 to 1350 kg/m3 when NDVI is smaller than 0.3, i.e., no or sparse vegetation cover (Fig. 6c). A strong negative correlation between seasonal deformation and bulk density is observed when NDVI is larger than 0.3. In addition, we do not find obvious controls of ground ice content on the seasonal deformation (Fig. S2).

4.2.2. Linear deformation The profile of linear deformation along the latitudinal direction

shows an obvious pattern across the Kunlun fault (Fig. 7). The ground surface moves away from the SAR satellite in the northern region of the Kunlun fault, followed by a rapidly increased trend and then a slightly decreasing trend from the Kunlun fault to the southern part of the study area. Our results are consistent with the deformation patterns derived from Envisat InSAR, which is more likely related to the post-seismic ground movement after the 2001 Mw 7.8 Hoh Xil earthquake (Zhao et al., 2018, 2021). For convenience, we use “post-seismic-related pat-terns” to represent the post-seismic ground movement along the lat-itudinal direction throughout the paper. In contrast, the seasonal deformation has no obvious patterns along the latitudinal direction.

We also observe that linear deformation correlates to the maximum of ADDF, NDVI, and the sand fraction. However, all these factors have similar patterns with the post-seismic-related patterns along the lat-itudinal direction. It is difficult to isolate the contributions of these

Fig. 5. Box plot (2% quartile, first quartile, median, third quartile, 98% quartile) showing the distribution of InSAR-derived LOS seasonal deformation under (a) different terrain slope angle; (b) different aspect (x-axis is the cosine values of aspect angle ranging from 0 to 360); (c) The LOS seasonal deformation and elevation along the profile QQ’ shown in Fig. 3.

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

8

factors on the linear deformation from the post-seismic-related patterns over large scales. Thus, we choose to conduct analysis only focusing on the south of the Yushu-Fenghuoshan fault, where the linear deformation is less affected by the post-seismic-related patterns.

The linear deformation exhibits a strong correlation with the 0–5 m VIC and the MAGT. Stronger linear deformation occurs in the regions with high VIC and weaker linear deformation is found on regions with low VIC (Fig. 8a). The median value of the linear deformation is about 2.5 mm/year with VIC larger than 0.6. In the regions with VIC smaller than 0.2, the median value is around 0 mm/year. Similar patterns are found between the linear deformation and the MAGT as the regions with higher MAGT show stronger linear deformation (Fig. 8b). The linear deformation with MAGT larger than − 2 ∘C is about 2.2 mm/year larger than that of MAGT smaller than − 2 ∘C. Moreover, the linear deformation exhibits stronger relationship with VIC when the MAGT is higher than − 2 ∘C, whereas the differences of the median linear deformation be-tween VIC smaller than 0.2 and larger than 0.6 reach to around 4 mm/

year (Fig. 8c). On the other hand, the linear deformation shows no or weak relationship with VIC when the MAGT is lower than − 2 ∘C.

4.2.3. Correlation between seasonal and linear deformation Interestingly, we find that linear deformation is one of the controls

on the amplitude of seasonal deformation, i.e., stronger seasonal deformation are observed in the regions with stronger linear subsidence (Fig. 9a). However, seasonal deformation is not correlated to linear deformation (Fig. 9b). In other words, the regions suffering from strong linear subsidence are most likely to undergo strong seasonal deforma-tion, but the regions suffering from strong seasonal subsidence do not necessarily undergo strong linear subsidence.

4.2.4. Localized ground deformation In addition to the strong patterns along the latitudinal direction in

the linear deformation, there are numerous deformation clusters showing strong subsidence or uplift signals. To mitigate the effects from

Fig. 6. Box graph representing the distribution of LOS seasonal deformation under (a) different NDVI; (b) different bulk density; (c) different NDVI and bulk density.

Fig. 7. The LOS ground deformation across the Kunlun fault along profile PP’ shown in Fig. 3. (a) The linear deformation; (b) the amplitude of seasonal deformation. The red vertical lines denote the fault position. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

9

long-wavelength trends and better understand the controlling factors of ground deformation in the permafrost environment, we choose three local regions, named by R1, R2, and R3 (the red rectangles in Fig. 3), by calibrating them using local references. Noting that since the seasonal deformation is not or less correlated with the latitudinal patterns, we only re-calibrate the linear deformation.

We choose the Hoh Xil region comprising Zonag and Yanhu Lakes (the red rectangle R1 in Fig. 3), where strong lake processes occurred (Liu et al., 2019). Ice-wedge (or sand-wedge) polygonal networks are well-developed on the drained lake basin, which was exposed after the outburst of Zonag Lake in 2011 (Fig. 10e). The drained lake basin ex-hibits obvious uplift up to 10 mm/year and very strong seasonal defor-mation up to 85 mm (Fig. 10a and c). The water level and area of Yanhu lake had continuously increased since 2010 till the outburst at the end of 2019 (Liu et al., 2019). On the contrary, both strong linear subsidence up to 15 mm/year and strong seasonal subsidence up to 50 mm are observed on the flat regions surrounding Yanhu Lake. Overall, most of the high seasonal ground deformation are shown in the flat regions based on the background of hill-shaded map in Fig. 10b. In addition, the time-series comparisons in Fig. 10c and d show good agreements

between the InSAR-derived and the fitted deformation time-series based on our PECM and thus further demonstrate the capability of our MODIS-LST-integrated InSAR approaches.

Another basin surrounding by the Tanggula mountains and glaciers (the red rectangle R2 in Fig. 3), where the QTH passed by, is charac-terized by many thermokarst lakes. We observe both widespread linear subsidence up to 15 mm/year and seasonal deformation up to 45 mm within the basin (Fig. 11a and b). The linear subsidence shows similar spatial patterns to the seasonal deformation. Along the QTH, extensive linear subsidence larger than 5 mm/year seasonal subsidence larger than 30 mm are observed (Fig. 11c), suggesting that the QTH may be threatened by strong thaw settlement and freeze heave.

We identify a flat basin underlying by ice-rich permafrost in the source of the Tuotuo River (the red rectangle R3 in Fig. 3), where extensive glacier lakes and thermokarst lakes are found. We observe significant linear subsidence up to 18 mm/year and moderate seasonal deformation up to 40 mm in the flat basin, whereas most of the moun-tain regions undergo no or very limited linear subsidence and/or sea-sonal deformation. On the western side of the Tuotuo River, relatively homogeneous linear subsidence is observed. Our method captures the

Fig. 8. Box graph representing the distribution of LOS linear deformation under (a) different 0–5 m ground volumetric ice content (VIC); (b) different MAGT; (c) different VIC and MAGT.

Fig. 9. Box plot for (a) seasonal and linear deformation; (b) linear and seasonal deformation.

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

10

major temporal patterns of ground deformation (Fig. 12c and d). How-ever, due to the fixed temporal assumption on permafrost-related ground deformation in our PECM, our model cannot capture inter- annual variations of the ground deformation (Fig. 12c).

4.3. Comparison with active layer thickness

We indirectly validate the seasonal deformation derived from InSAR against the active layer thickness (ALT). To mitigate the impacts of spatial heterogeneity on the validation, we calculate the weighted block average to represent the site-specific InSAR-derived seasonal deforma-tion within 200m × 200m blocks. More details on how to compare the seasonal deformation with ALTs can be found in Chen et al. (2018b). The ALTs are estimated by using the continuously ground temperature at different depths. The seasonal deformation shows well correlations with the ALT in the region covered by the alpine swamp meadow, where the VWC is expected to be high (Fig. 13). However, the VWC in the alpine meadow or alpine steppe may vary in a large range, resulting in a poor correlation between the seasonal deformation and ALTs.

To account for the effect of different water content, we calculate the mean VWC in the active layer by summing the VWC in each layer weighted by depth. The values of the mean VWC is ranging from 0.2 to 0.4. For comparison, the calibrated ALT are all divided by 0.3:

ALTcal =ALT × VWCmean

0.3(8)

The calibrated ALT shows a better correlation with seasonal defor-mation with the root-mean-square error drops from 9 mm to 3 mm and

R2 increases from 0.43 to 0.82.

5. Discussion

5.1. Active layer freeze-thaw dynamics revealed from InSAR measurements

Based on our MODIS-LST-integrated ground deformation model tailored for permafrost environment, we reveal the magnitudes and spatial patterns of seasonal deformation at a high spatial resolution over a vast permafrost region. By comparison with local environmental fac-tors, we determine the main controlling factors of seasonal deformation over large scales on the permafrost regions of the central QTP. Seasonal deformation mainly depends on the volumetric change of water in solid and liquid states in the active layer, which is consequently determined by thaw consolidation. In other words, the InSAR-derived seasonal deformation is proportional to the total water content in the active layer (Chen et al., 2020). The amplitude of InSAR-derived seasonal defor-mation shows strong spatial variabilities, representing the spatial pat-terns of total water content in the active layer.

We find that the terrain slope angle is the dominant factors deter-mining the variations of seasonal deformation. To be more specific, the slope control on seasonal deformation is the strongest when slope angles are smaller than 5∘. The magnitudes of seasonal deformation drop dramatically with increased slope angle. This is mainly because the flat regions bear a high capacity to hold water and their capacity quickly decreases with the increase of slope angle. The large variations of sea-sonal deformation indicate that there are other factors controlling the

Fig. 10. The LOS ground deformation surrounding Zonag and Yanhu Lakes. (a) The linear deformation. The red triangle shows the position of local reference point; (b) The amplitude of seasonal deformation; LOS deformation time series of (c) Pixel P1 (exposed after the outburst of Zonag Lake in 2011 marked in (a)); and (d) Pixel P2; (e) The Esri map showing the ground features of polygonal networks located in P1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

11

amplitude of seasonal deformation with same slope angles, such as the availability of water and the ALT. But in general, flat regions (smaller than 5∘), such as basins or mountain valleys, are more vulnerable to stronger seasonal deformation. In contrast, the slope angle control on seasonal deformation becomes very weak with slope angles larger than

5∘. This suggests that the variations of slope angle have little impact on the water capacity and thus not the primary factors controlling the seasonal deformation when slope angles larger than 5∘. The mountain ridges or steep slope angles can hardly hold water and consequently experience smaller seasonal deformation. Similar spatial patterns of

Fig. 11. The LOS ground deformation in the Tanggula region. (a) The linear deformation; (b) The amplitude of seasonal deformation; LOS deformation time series of (c) Pixel P3; and (d) Pixel P4.

Fig. 12. The LOS ground deformation surrounding the source region of Tuotuohe river. (a) The linear deformation; (b) The amplitude of seasonal deformation; LOS deformation time series of (c) Pixel P5; and (d) Pixel P6.

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

12

seasonal deformation are also observed on the Qilian Mountain of the QTP (Daout et al., 2020) and on the North slope angle of Alaska (Chen et al., 2020). However, the magnitudes of seasonal deformation we observed (mean values is around 10 mm) on the central QTP are about three times larger compared to the western Qilian Mountain (mean values is around 3 mm) (Daout et al., 2020). This may be mainly because that the ALT on the Qilian Mountain (Cao et al., 2017) is observed to be shallower than that on the central QTP (Zhao et al., 2020). The western Qilian Mountain is near the Qaidam basin and characterized with arid climate, where the active layer tends to be limited in water content. In addition, based on the inventory of lakes on the QTP (Zhang et al., 2020), our study area contains more lakes than the western Qilian Mountain, indicating extensive surface water activity and availability.

Aspect angle exhibits control on the seasonal deformation with slope angle larger than 5∘. Our results show that the seasonal deformation on the north-facing slopes is slightly larger than that on the east- and west- facing slopes, which is slightly larger than that of the south-facing slopes. In general, the north-facing slope angle receives less solar radi-ation and thus experience weak evapotranspiration compared to the south-facing slope angle, leading to shallower ALT but higher soil water content in the north-facing slope angle.

Vegetation cover may be another key factor controlling the ampli-tude of seasonal deformation. Most larger seasonal deformation is pro-nounced in the regions with no or very limited vegetation cover, e.g., NDVI is smaller than 0.25. Vegetation layer can slow down the pene-tration of air-to-ground energy flux, acting as a protective layer to decay the thawing processes (Hu et al., 2020). Thus, the active layer is generally deeper in the regions with bare soil. While the regions with deeper ALT fed by abundant water from glaciers and/or lakes, the ground surface would undergo significant seasonal deformation, such as the flat regions surrounding Zonag and Yanhu Lakes (see Fig. 10). Meanwhile, in the regions with moderate or high vegetation cover, we observe a positive correlation between seasonal deformation and NDVI. In this case, the active layer tends to be shallower. The increasing NDVI is likely correlated with the increasing soil water content in the upper-most active layer, which may result in larger seasonal deformation with denser vegetation cover. Soil bulk density is a proxy of soil porosity and determines its water holding capacity, which may have a strong corre-lation with seasonal deformation (Chen et al., 2020). Given the same vegetation conditions, larger soil bulk density denotes smaller soil porosity and thus has a lower water holding capacity, resulting in weaker seasonal deformation.

Our results suggested that in regions with saturated active layer or very dense vegetation cover, InSAR-derived seasonal deformation have good potentials to determine the ALT with prior information of soil

properties, such as on the Alaska North slope angle (Liu et al., 2012). However, more attention should be paid to the regions with unsaturated active layer or limited vegetation cover. Without the knowledge of the saturation degree of soil water content, the ALT estimation from InSAR-derived ground deformation might fail or suffer from large uncertainties.

5.2. Implications for permafrost degradation/aggradation

Extensive long-term ground subsidence has been observed on the central QTP, which may be attributed to the thawing of the ice-rich permafrost due to climatic warming. We find that the regions with high VIC have suffered from stronger linear subsidence, highlighting that ice-rich permafrost is more vulnerable to suffering from ground lowering due to massive ground ice loss. We also find that the regions underlying by warm permafrost (MAGT > − 2∘C) (Cheng and Wu, 2007) and high VIC exhibit the strongest linear subsidence. Limited variations and small magnitude of linear subsidence even with high VIC are pro-nounced on the regions with cold permafrost (MAGT ≤ − 2 ∘C). Evi-denced by borehole measurements, permafrost temperature on the QTP are increasing in recent decades (Zhao et al., 2020). Under continuous warming, the ground is in transition from cold permafrost to warm permafrost. Once the temperature of the uppermost permafrost reaches 0 ∘C or close 0 ∘C, the heat flux would potentially thaw the uppermost permafrost and consequently result in the loss of the uppermost ground ice. Our finding highlights that the warm permafrost with high VIC is experiencing extensive ground ice loss. On the contrary, the cold permafrost even with high VIC is still undergoing limited ground ice loss. However, once the cold permafrost transformed to warm perma-frost, the ground ice near the permafrost table would probably thaw under continuous warming. Noting that extremely warm summer with abundant heat flux may trigger extensive loss of ground ice even in the cold permafrost regions (Zwieback and Meyer, 2021). In addition, the changes of local hydrological conditions may induce permafrost degra-dation. Strong linear and seasonal subsidence are observed surrounding Yanhu Lake, which may be related to the water erosion and the melting of massive ground ice (Lu et al., 2020). Despite the ground ice devel-opment may relate to the slope and/or aspect, the slope and aspect angle exhibit no obvious impacts on the linear deformation in our study area (Fig. 4a and Fig. S3). This may be because the slope movements are mostly local-scale and their occurrences may not only relates to the climate warming but also the local disturbances.

Significant and continuous ground uplift is observed during 2014 and 2019 on the drained lake basin of Zonag Lake, suggesting that the formation of excess ground ice has been persisting since the outburst of Zonag Lake in September 2011. The seasonal deformation in the drained lake basin is very strong and homogeneous with amplitudes of about 60–85 mm, indicating that the active layer is likely saturated. The sea-sonal refreezing of saturated soil promotes the formation of excess ground ice at the bottom of the active layer. We notice that except for the fitted linear deformation, 5–10 mm offset is observed between the original and fitted deformation time series during the late freeze season of 2017–2019 (Fig. 10). The noticeable uplift signal may be related to the lateral water injection due to extra lake water pressure, which also promotes the ice forming (Liu et al., 2014). In this study, the magnitude of uplift trends and late-season uplift are comparable, indicating the two ice-forming mechanisms may occur simultaneously.

5.3. The applicability of MODIS-LST-integrated InSAR ground deformation approach

We incorporate MODIS-LST-integrated deformation model and Ste-fan equation into SBAS InSAR measurements to measure permafrost ground deformation on the QTP. We have successfully depicted the magnitudes and patterns of both linear and seasonal deformation over large scales. With the aid of MODIS LST and soil information (soil bulk

Fig. 13. The comparison of InSAR-derived seasonal deformation and the ALT in different vegetation covers: alpine swamp meadow (ASM), alpine meadow (AM) and alpine steppe (AS). The squares show the calibrated ALT by Eq. (8).

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

13

density and sand fraction), it is viable to generalize our method into the permafrost regions over the entire QTP. All the external temperature and soil parameters are easily exchangeable if more accurate data are available in the future. In the Arctic, due to the effect of heave snow cover, only the SAR images acquired in summer seasons are available. Chen et al. (2018a) and Strozzi et al. (2018) had demonstrated the capability of Sentinel-1 InSAR measurement to measure the inter-annual ground deformation in ice-rich permafrost regions of the Arctic. In addition, MODIS LST with 1 km resolution (Obu et al., 2019) and soil information with 250 m resolution (Hengl et al., 2017) are also available in the Arctic. Thus, we believe that our methods can also be applied over large scales in the Arctic.

More attention should be paid to the tectonic movement when decomposing the permafrost-related ground deformation over large scales. In our permafrost deformation model, we assume linear defor-mation which is purely related to the thawing of the underlying permafrost. However, as seen from Fig. 3, the linear deformation is probably mixed with the post-seismic ground movement after the 2001 Mw 7.8 Hoh Xil earthquake (Zhao et al., 2021) and linear deformation due to the loss of ground ice. Due to linear or nearly linear patterns of both tectonic movement and permafrost-related deformation, it is difficult to distinguish them in the scope of pixel-by-pixel InSAR inver-sion (Daout et al., 2020). Optimistically, due to the yearly periodic characteristic of seasonal deformation, we are still able to isolate the seasonal deformation with high accuracy. However, to quantify the ground ice loss via InSAR measurements over large scales, better un-derstanding of the spatial and temporal patterns of both permafrost-related deformation and tectonic movement is needed.

6. Conclusion

We have developed a permafrost-tailored InSAR approach to isolate permafrost-related ground deformation over large scales by incorpo-rating MODIS LST and Stefan equation. By applying the method to the permafrost regions of about 140,000 km2 on the central QTP, we suc-cessfully decompose the seasonal and linear deformation from the multi- temporal Sentinel-1 InSAR measurements. We observe widespread sea-sonal deformation up to 80 mm and linear subsidence up to 20 mm/ year.

For the first time we apply the geographical detector to determine the controlling factors on the seasonal and linear deformation in permafrost environment. The slope angle is closely related to the water holding capacity of the active layer, which is the primary factor con-trolling the seasonal deformation. We observe stronger magnitude and variations of seasonal deformation in the flat or gentle-slope regions. In the mountain regions, the seasonal deformation on the north-facing slopes is slightly larger than that on the south-facing slopes. In the re-gions with no or sparse vegetation cover (e.g., NDVI < 0.3), the ground surface experience stronger seasonal deformation compared to that with moderate to high vegetation cover (e.g., NDVI > 0.3). However, the seasonal deformation is increased with denser vegetation cover in re-gions with moderate to high vegetation cover. Meanwhile, we find that the linear subsidence are higher in the regions with high VIC and warm permafrost, attesting that warm permafrost regions with high VIC are more vulnerable to extensive long-term ground subsidence. In addition, we observe significant uplift signals in the drained lake basins, illus-trating the permafrost aggradation after lake outburst.

Our study demonstrates the capability of our permafrost-tailored InSAR approaches to quantify the magnitudes and spatial variations of the freeze-thaw processes and the melting of ground ice under different surface conditions (such as different terrains, vegetation types, ice-rich or ice-poor permafrost) at a high resolution over large scales. In the remote or inaccessible permafrost regions, our InSAR-derived ground deformation are capable of quantifying the freeze-thaw processes and the degradation of permafrost, which is valuable for understanding the permafrost response to climate warming and local disturbance.

Author contributions

T. Wu, J. Chen and L. Liu designed the project; J. Chen and T. Wu wrote the manuscript with support from D. Zou, L. Liu, X. Wu, W. Gong, X. Zhu, R. Li, J. Hao, G. Hu, Q. Pang, S. Yang; All the authors reviewed and edited the manuscript; J. Chen and L. Liu conducted the InSAR data processing; D. Zou, X. Zhu collected and analyzed the in-situ measure-ments; Q. Pang provided the ground ice data; R. Li provided the soil conductivity; J. Zhang helped the visualization.

Declaration of Competing Interest

None.

Acknowledgments

This research was supported by the National Natural Science Foun-dation of China (Grant No. 41690142, 42001072 and 41771076), the National Cryosphere Desert Data Center Program (Grant no. E01Z790208), the CAS “Special Research Assistant program” (Jie Chen) and the CAS “Hundred Talents program” (Sizhong Yang).

References

Berardino, P., Fornaro, G., Lanari, R., Sansosti, E., 2002. A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. Geosci. Rem. Sens. IEEE Trans. 40, 2375–2383. https://doi.org/10.1109/ TGRS.2002.803792.

Bernhard, P., Zwieback, S., Leinss, S., Hajinsek, I., 2020. Mapping retrogressive thaw slumps using single-pass TanDEM-X observations. IEEE J. Sel. Top. Appl. Earth Observ. Rem. Sens. 13, 3263–3280. https://doi.org/10.1109/jstars.2020.3000648.

Cao, B., Gruber, S., Zhang, T., Li, L., Peng, X., Wang, K., Zheng, L., Shao, W., Guo, H., 2017. Spatial variability of active layer thickness detected by ground-penetrating radar in the Qilian mountains, Western China. J. Geophys. Res.: Earth Surf. 122, 574–591. https://doi.org/10.1002/2016JF004018.

Cao, B., Zhang, T., Wu, Q., Sheng, Y., Zhao, L., Zou, D., 2019. Brief communication: evaluation and inter-comparisons of Qinghai-Tibet plateau permafrost maps based on a new inventory of field evidence. Cryosphere 13, 511–519. https://doi.org/ 10.5194/tc-13-511-2019.

Chen, C.W., Zebker, H.A., 2001. Two-dimensional phase unwrapping with use of statistical models for cost functions in nonlinear optimization. J. Opt. Soc. Am. A 18, 338–351. https://doi.org/10.1364/JOSAA.18.000338.

Chen, J., Günther, F., Grosse, G., Liu, L., Lin, H., 2018a. Sentinel-1 InSAR measurements of elevation changes over yedoma uplands on sobo-sise island, lena delta. Rem. Sens. 10, 1152. https://doi.org/10.3390/rs10071152.

Chen, J., Liu, L., Zhang, T., Cao, B., Lin, H., 2018b. Using persistent scatterer interferometry to map and quantify permafrost thaw subsidence: a case study of eboling mountain on the Qinghai-Tibet plateau. J. Geophys. Res.: Earth Surf. https:// doi.org/10.1029/2018JF004618.

Chen, J., Wu, Y., O’Connor, M., Cardenas, M.B., Schaefer, K., Michaelides, R., Kling, G., 2020. Active layer freeze-thaw and water storage dynamics in permafrost environments inferred from InSAR. Rem. Sens. Environ. 248, 112007. https://doi. org/10.1016/j.rse.2020.112007.

Cheng, G., Wu, T., 2007. Responses of permafrost to climate change and their environmental significance, Qinghai-Tibet plateau. J. Geophys. Res.: Earth Surf. 112, F02S03. https://doi.org/10.1029/2006JF000631.

Cheng, G., Zhao, L., Li, R., Wu, X., Sheng, Y., Hu, G., Zou, D., Jin, H., Li, X., Qingbai, W., 2019. Characteristic, changes and impacts of permafrost on Qinghai-Tibet plateau (in Chinese). Chin. Sci. Bull. 64, 2783–2795.

Daout, S., Dini, B., Haeberli, W., Doin, M.P., Parsons, B., 2020. Ice loss in the Northeastern Tibetan plateau permafrost as seen by 16 yr of ESA SAR missions. Earth Planet. Sci. Lett. 545, 116404. https://doi.org/10.1016/j.epsl.2020.116404.

Daout, S., Doin, M.P., Peltzer, G., Socquet, A., Lasserre, C., 2017. Large-scale InSAR monitoring of permafrost freeze-thaw cycles on the Tibetan plateau. Geophys. Res. Lett. 44, 901–909. https://doi.org/10.1002/2016GL070781.

Delacourt, C., Briole, P., Achache, J.A., 1998. Tropospheric corrections of SAR interferograms with strong topography. Appl. Etna. Geophys. Res. Lett. 25, 2849–2852. https://doi.org/10.1029/98GL02112.

Du, Y., Li, R., Zhao, L., Yang, C., Wu, T., Hu, G., Xiao, Y., Zhu, X., Yang, S., Ni, J., Ma, J., 2020. Evaluation of 11 soil thermal conductivity schemes for the permafrost region of the central Qinghai-Tibet plateau. Catena 193, 104608. https://doi.org/10.1016/ j.catena.2020.104608.

Fang, L., Wang, L., Chen, W., Sun, J., Cao, Q., Wang, S., Wang, L., 2021. Identifying the impacts of natural and human factors on ecosystem service in the Yangtze and Yellow river Basins. J. Clean. Prod. 314, 127995. https://doi.org/10.1016/j. jclepro.2021.127995.

Farquharson, L.M., Romanovsky, V.E., Cable, W.L., Walker, D.A., Kokelj, S.V., Nicolsky, D.D., 2019. Climate change drives widespread and rapid thermokarst

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

14

development in very cold permafrost in the canadian high arctic. Geophys. Res. Lett. 46, 6681–6689. https://doi.org/10.1029/2019GL082187.

Ferretti, A., Prati, C., Rocca, F., 2001. Permanent scatterers in SAR interferometry. Geosci. Remote Sens. IEEE Trans. 39, 8–20. https://doi.org/10.1109/36.898661.

Goel, K., Adam, N., 2012. An advanced algorithm for deformation estimation in non- urban areas. ISPRS J. Photogram. Rem. Sens. 73, 100–110. https://doi.org/10.1016/ j.isprsjprs.2012.06.001.

Goldstein, R.M., Werner, C.L., 1998. Radar interferogram filtering for geophysical applications. Geophys. Res. Lett. 25, 4035–4038. https://doi.org/10.1029/ 1998GL900033.

Gruber, S., 2020. Ground subsidence and heave over permafrost: hourly time series reveal interannual, seasonal and shorter-term movement caused by freezing, thawing and water movement. Cryosphere 14, 1437–1447. https://doi.org/ 10.5194/tc-14-1437-2020.

Günther, F., Overduin, P.P., Yakshina, I.A., Opel, T., Baranskaya, A.V., Grigoriev, M.N., 2015. Observing muostakh disappear: permafrost thaw subsidence and erosion of a ground-ice-rich island in response to arctic summer warming and sea ice reduction. Cryosphere 9, 151–178. https://doi.org/10.5194/tc-9-151-2015.

Hengl, T., De Jesus, J.M., Heuvelink, G.B., Gonzalez, M.R., Kilibarda, M., Blagotic, A., Shangguan, W., Wright, M.N., Geng, X., Bauer-Marschallinger, B., Guevara, M.A., Vargas, R., MacMillan, R.A., Batjes, N.H., Leenaars, J.G., Ribeiro, E., Wheeler, I., Mantel, S., Kempen, B., 2017. SoilGrids250m: global gridded soil information based on machine learning. PLoS ONE 12, e0169748. https://doi.org/10.1371/journal. pone.0169748.

Hjort, J., Karjalainen, O., Aalto, J., Westermann, S., Romanovsky, V.E., Nelson, F.E., Etzelmüller, B., Luoto, M., 2018. Degrading permafrost puts arctic infrastructure at risk by mid-century. Nat. Commun. 9, 5147. https://doi.org/10.1038/s41467-018- 07557-4.

Hu, G., Zhao, L., Li, R., Wu, X., Wu, T., Xie, C., Zhu, X., Hao, J., 2020. Thermal properties of active layer in permafrost regions with different vegetation types on the Qinghai- Tibetan plateau. Theor. Appl. Clim. 139, 983–993. https://doi.org/10.1007/s00704- 019-03008-2.

Hu, Y., Liu, L., Larson, K.M., Schaefer, K.M., Zhang, J., Yao, Y., 2018. GPS Interferometric reflectometry reveals cyclic elevation changes in thaw and freezing seasons in a permafrost area (Barrow, Alaska). Geophys. Res. Lett. 45, 5581–5589. https://doi. org/10.1029/2018GL077960.

Jin, H., Luo, D., Wang, S., Lü, L., Wu, J., 2011. Spatiotemporal variability of permafrost degradation on the Qinghai-Tibet plateau. Sci. Cold Arid Reg. 3, 281–305. https:// doi.org/10.3724/SP.J.1226.2011.00281.

Johansen, O., 1977. Thermal Conductivity of Soils. Technical Report. Cold Regions Research and Engineering Lab, Hanover NH.

Jolivet, R., Grandin, R., Lasserre, C., Doin, M.P., Peltzer, G., 2011. Systematic InSAR tropospheric phase delay corrections from global meteorological reanalysis data. Geophys. Res. Lett. 38 https://doi.org/10.1029/2011GL048757.

Jorgenson, M.T., Osterkamp, T.E., 2005. Response of boreal ecosystems to varying modes of permafrost degradation. Can. J. Forest Res. 35, 2100–2111. https://doi.org/ 10.1139/x05-153.

Kaab, A., 2002. Monitoring high-mountain terrain deformation from repeated air- and spaceborne optical data: examples using digital aerial imagery and ASTER data. ISPRS J. Photogram. Rem. Sens. 57, 39–52. https://doi.org/10.1016/S0924-2716 (02)00114-4.

Koven, C.D., Ringeval, B., Friedlingstein, P., Ciais, P., Cadule, P., Khvorostyanov, D., Krinner, G., Tarnocai, C., 2011. Permafrost carbon-climate feedbacks accelerate global warming. Proc. Nat. Acad. Sci. 108, 14769–14774. https://doi.org/10.1073/ pnas.1103910108.

Lauknes, T.R., Zebker, H.A., Larsen, Y., 2011. InSAR deformation time series using l1- norm small-baseline approach. Geoscience and remote sensing. IEEE Trans. 49, 536–546. https://doi.org/10.1109/TGRS.2010.2051951.

Li, Z., Tang, P., Zhou, J., Tian, B., Chen, Q., Fu, S., 2015. Permafrost environment monitoring on the Qinghai-Tibet plateau using time series ASAR images. Int. J. Digital Earth 8, 840–860. https://doi.org/10.1080/17538947.2014.923943.

Liljedahl, A.K., Boike, J., Daanen, R.P., Fedorov, A.N., Frost, G.V., Grosse, G., Hinzman, L.D., Iijma, Y., Jorgenson, J.C., Matveyeva, N., Necsoiu, M., Raynolds, M. K., Romanovsky, V.E., Schulla, J., Tape, K.D., Walker, D.A., Wilson, C.J., Yabuki, H., Zona, D., 2016. Pan–Arctic ice-wedge degradation in warming permafrost and its influence on tundra hydrology. Nat. Geosci. 9, 312–318. https://doi.org/10.1038/ ngeo2674.

Little, J.D., Sandall, H., Walegur, M.T., Nelson, F.E., 2003. Application of differential global positioning systems to monitor frost heave and thaw settlement in tundra environments. Permafrost Periglacial Proc. 14, 349–357. https://doi.org/10.1002/ ppp.466.

Liu, L., Larson, K.M., 2018. Decadal changes of surface elevation over permafrost area estimated using reflected GPS signals. Cryosphere 12, 477–489. https://doi.org/ 10.5194/tc-12-477-2018.

Liu, L., Schaefer, K., Gusmeroli, A., Grosse, G., Jones, B.M., Zhang, T., Parsekian, A.D., Zebker, H.A., 2014. Seasonal thaw settlement at drained thermokarst lake basins, Arctic Alaska. Cryosphere 8, 815–826. https://doi.org/10.5194/tc-8-815-2014.

Liu, L., Schaefer, K., Zhang, T., Wahr, J., 2012. Estimating 1992–2000 average active layer thickness on the Alaskan North slope from remotely sensed surface subsidence. J. Geophys. Res.: Earth Surf. 117 https://doi.org/10.1029/2011JF002041.

Liu, L., Zhang, T., Wahr, J., 2010. InSAR measurements of surface deformation over permafrost on the North slope of Alaska. J. Geophys. Res.: Earth Surf. 115 https:// doi.org/10.1029/2009JF001547.

Liu, W., Xie, C., Zhao, L., Wu, T., Wang, W., Zhang, Y., Yang, G., Zhu, X., Yue, G., 2019. Dynamic changes in lakes in the Hoh Xil region before and after the 2011 outburst of

Zonag lake. J. Mount. Sci. 16, 1098–1110. https://doi.org/10.1007/s11629-018- 5085-0.

Lu, P., Han, J., Li, Z., Xu, R., Li, R., Hao, T., Qiao, G., 2020. Lake outburst accelerated permafrost degradation on Qinghai-Tibet plateau. Rem. Sens. Environ. 249, 112011. https://doi.org/10.1016/j.rse.2020.112011.

Luo, W., Liu, C.C., 2018. Innovative landslide susceptibility mapping supported by Geomorphon and geographical detector methods. Landslides 15, 465–474. https:// doi.org/10.1007/s10346-017-0893-9.

Mackay, J.R., 1973a. A frost tube for the determination of freezing in the active layer above permafrost. Can. Geotech. J. 10, 392–396. https://doi.org/10.1139/t73-033.

Mackay, J.R., 1973b. The growth of pingos, Western arctic coast, Canada. Can. J. Earth Sci. 10, 979–1004. https://doi.org/10.1139/e73-086.

Mackay, J.R., Burn, C.R., 2002. The first 20 years (1978–1979 to 1998–1999) of active- layer development, Illisarvik experimental drained lake site, Western arctic coast, Canada. Can. J. Earth Sci. 39, 1657–1674. https://doi.org/10.1139/e02-068.

Mu, C., Abbott, B.W., Norris, A.J., Mu, M., Fan, C., Chen, X., Jia, L., Yang, R., Zhang, T., Wang, K., Peng, X., Wu, Q., Guggenberger, G., Wu, X., 2020. The status and stability of permafrost carbon on the Tibetan plateau. Earth-Sci. Rev. 211, 103433. https:// doi.org/10.1016/j.earscirev.2020.103433.

Nauta, A.L., Heijmans, M.M.P.D., Blok, D., Limpens, J., Elberling, B., Gallagher, A., Li, B., Petrov, R.E., Maximov, T.C., van Huissteden, J., Berendse, F., 2015. Permafrost collapse after shrub removal shifts tundra ecosystem to a methane source. Nat. Cli. Change 5, 67–70. https://doi.org/10.1038/nclimate2446.

Nelson, F.E., Anisimov, O.A., Shiklomanov, N.I., 2001. Subsidence risk from thawing permafrost. Nature 410, 889–890. https://doi.org/10.1038/35073746.

Nelson, F.E., Shiklomanov, N.I., Mueller, G.R., Hinkel, K.M., Walker, D.A., Bockheim, J. G., 1997. Estimating active-layer thickness over a large region: kuparuk river Basin, Alaska, U.S.A. Artic Alpine Res. 29, 367–378. https://doi.org/10.2307/1551985.

Nitze, I., W. Cooley, S., R. Duguay, C., M. Jones, B., Grosse, G., 2020. The catastrophic thermokarst lake drainage events of 2018 in northwestern Alaska: fast-forward into the future. Cryosphere 14, 4279–4297. https://doi.org/10.5194/tc-14-4279-2020.

Obu, J., Westermann, S., Bartsch, A., Berdnikov, N., Christiansen, H.H., Dashtseren, A., Delaloye, R., Elberling, B., Etzelmüller, B., Kholodov, A., Khomutov, A., K&rdquo; a&ldquo;ab, A., Leibman, M.O., Lewkowicz, A.G., Panda, S.K., Romanovsky, V., Way, R.G., Westergaard-Nielsen, A., Wu, T., Yamkhin, J., Zou, D., 2019. Northern Hemisphere Permafrost Map Based on Ttop Modelling for 2000–2016 at 1?.km2 Scale. https://doi.org/10.1016/j.earscirev.2019.04.023.

Roerink, G.J., Menenti, M., Verhoef, W., 2000. Reconstructing cloudfree NDVI composites using fourier analysis of time series. Int. J. Remote Sens. 21, 1911–1917. https://doi.org/10.1080/014311600209814.

Rosen, P.A., Gurrola, E., Sacco, G.F., Zebker, H., 2012. The InSAR scientific computing environment. In: Synthetic Aperture Radar, 2012. EUSAR. 9th European Conference on. VDE, pp. 730–733.

Schuur, E.A.G., McGuire, A.D., Schadel, C., Grosse, G., Harden, J.W., Hayes, D.J., Hugelius, G., Koven, C.D., Kuhry, P., Lawrence, D.M., Natali, S.M., Olefeldt, D., Romanovsky, V.E., Schaefer, K., Turetsky, M.R., Treat, C.C., Vonk, J.E., 2015. Climate change and the permafrost carbon feedback. Nature 520, 171–179. https:// doi.org/10.1038/nature14338.

Streletskiy, D.A., Shiklomanov, N.I., Little, J.D., Nelson, F.E., Brown, J., Nyland, K.E., Klene, A.E., 2017. Thaw subsidence in undisturbed tundra landscapes, barrow, Alaska, 1962–2015. Permafrost Periglacial Proc. 28, 566–572. https://doi.org/ 10.1002/ppp.1918.

Strozzi, T., Antonova, S., Günther, F., M&ldquo;atzler, E., Vieira, G., Wegmüller, U., Westermann, S., Bartsch, A., 2018. Sentinel-1 SAR interferometry for surface deformation monitoring in low-land permafrost areas. Rem. Sens. 10, 1360. https:// doi.org/10.3390/rs10091360.

Sun, J., Yang, K., Guo, W., Wang, Y., He, J., Lu, H., 2020. Why has the inner Tibetan plateau become wetter since the mid-1990s? J. Cli. 33, 1–45. https://doi.org/ 10.1175/jcli-d-19-0471.1.

Tong, X., Schmidt, D., 2016. Active movement of the cascade landslide complex in Washington from a coherence-based InSAR time series method. Rem. Sens. Environ. 186, 405–415. https://doi.org/10.1016/J.RSE.2016.09.008.

Walvoord, M.A., Kurylyk, B.L., 2016. Hydrologic impacts of thawing permafrost-a review. Vadose Zone J. 15 https://doi.org/10.2136/vzj2016.01.0010.

Wang, J.F., Li, X.H., Christakos, G., Liao, Y.L., Zhang, T., Gu, X., Zheng, X.Y., 2010. Geographical detectors-based health risk assessment and its application in the neural tube defects study of the heshun region, China. Int. J. Geograph. Info. Sci. 24, 107–127. https://doi.org/10.1080/13658810802443457.

Wang, Z., Wang, Q., Zhao, L., Wu, X., Yue, G., Zou, D., Nan, Z., Liu, G., qiang Pang, Q., bing Fang, H., Wu, T., Shi, J., Jiao, K., Zhao, Y., Zhang, L., 2016. Mapping the vegetation distribution of the permafrost zone on the Qinghai-Tibet plateau. J. Mount. Sci. 13, 1035–1046. https://doi.org/10.1007/s11629-015-3485-y.

Wu, Q., Sheng, Y., Yu, Q., Chen, J., Ma, W., 2020. Engineering in the rugged permafrost terrain on the roof of the world under a warming climate. Permafrost Periglacial Proc. https://doi.org/10.1002/ppp.2059 ppp.2059.

Wu, Q., Zhang, T., 2008. Recent permafrost warming on the Qinghai-Tibetan plateau. J. Geophys. Res.: Atmos. 113, D13108. https://doi.org/10.1029/2007JD009539.

Wu, X., Zhao, L., Liu, G., Xu, H., Zhang, X., Ding, Y., 2018. Effects of permafrost thaw- subsidence on soil bacterial communities in the southern Qinghai-Tibetan plateau. Appl. Soil Ecol. 128, 81–88. https://doi.org/10.1016/J.APSOIL.2018.04.007.

Xu, Y., Shen, Y., Wu, Z., 2013. Spatial and temporal variations of land surface temperature over the tibetan plateau based on harmonic analysis. Mount. Res. Dev. 33, 85–94. https://doi.org/10.1659/MRD-JOURNAL-D-12-00090.1.

Yagüe-Martínez, N., Prats-Iraola, P., Gonzalez, F.R., Brcic, R., Shau, R., Geudtner, D., Eineder, M., Bamler, R., 2016. Interferometric processing of sentinel-1 tops data.

J. Chen et al.

Remote Sensing of Environment 268 (2022) 112778

15

IEEE Trans. Geosci. Rem. Sens. 54, 2220–2234. https://doi.org/10.1109/ TGRS.2015.2497902.

Yang, J., Song, C., Yang, Y., Xu, C., Guo, F., Xie, L., 2019. New method for landslide susceptibility mapping supported by spatial logistic regression and GeoDetector: a case study of Duwen highway Basin, sichuan province, China. Gemorphology 324, 62–71. https://doi.org/10.1016/j.geomorph.2018.09.019.

Zebker, H., 2021. Accuracy of a model-free algorithm for temporal insar tropospheric correction. Rem. Sens. 13, 1–9. https://doi.org/10.3390/rs13030409.

Zhang, G., Yao, T., Xie, H., Yang, K., Zhu, L., Shum, C.K., Bolch, T., Yi, S., Allen, S., Jiang, L., Chen, W., Ke, C., 2020. Response of Tibetan plateau lakes to climate change: trends, patterns, and mechanisms. Earth-Sci. Rev. 208, 103269. https://doi. org/10.1016/j.earscirev.2020.103269.

Zhang, X., Zhang, H., Wang, C., Tang, Y., Zhang, B., Wu, F., Wang, J., Zhang, Z., Zhang, X., Zhang, H., Wang, C., Tang, Y., Zhang, B., Wu, F., Wang, J., Zhang, Z., 2019. Time-Series InSAR monitoring of permafrost freeze-thaw seasonal displacement over Qinghai-Tibetan plateau using sentinel-1 data. Rem. Sens. 11, 1000. https://doi.org/10.3390/rs11091000.

Zhao, D., Qu, C., Bürgmann, R., Gong, W., Shan, X., 2021. Relaxation of Tibetan lower crust and afterslip driven by the 2001 Mw7.8 kokoxili, China, earthquake constrained by a decade of geodetic measurements. J. Geophys. Res.: Solid Earth. https://doi.org/10.1029/2020JB021314 e2020JB021314.

Zhao, D., Qu, C., Shan, X., Bürgmann, R., Gong, W., Zhang, G., 2018. Spatiotemporal evolution of postseismic deformation following the 2001 Mw7.8 kokoxili, China, earthquake from 7 years of insar observations. Rem. Sens. 10, 1988. https://doi.org/ 10.3390/rs10121988.

Zhao, L., Zou, D., Hu, G., Du, E., Pang, Q., Xiao, Y., Li, R., Sheng, Y., Wu, X., Sun, Z., Wang, L., Wang, C., Ma, L., Zhou, H., Liu, S., 2020. Changing climate and the permafrost environment on the Qinghai-Tibet (Xizang) plateau. Permafrost Periglacial Proc. 31, 396–405. https://doi.org/10.1002/ppp.2056.

Zongxing, L., Qi, F., Zongjie, L., Ruifeng, Y., Juan, G., Yuemin, L., 2019. Climate background, fact and hydrological effect of multiphase water transformation in cold regions of the Western China: a review. Earth-Sci. Rev. 190, 33–57. https://doi.org/ 10.1016/j.earscirev.2018.12.004.

Zou, D., Zhao, L., Sheng, Y., Chen, J., Hu, G., Wu, T., Wu, J., Xie, C., Wu, X., Pang, Q., Wang, W., Du, E., Li, W., Liu, G., Li, J., Qin, Y., Qiao, Y., Wang, Z., Shi, J., Cheng, G., 2017. A new map of permafrost distribution on the Tibetan plateau. Cryosphere 11, 2527–2542. https://doi.org/10.5194/tc-11-2527-2017.

Zou, D., Zhao, L., Wu, T., Wu, X., Pang, Q., Wang, Z., 2014. Modeling ground surface temperature by means of remote sensing data in high-altitude areas: test in the central Tibetan plateau with application of moderate-resolution imaging spectroradiometer Terra/Aqua land surface temperature and ground-based infrared. J. Appl. Rem. Sens. 8, 083516. https://doi.org/10.1117/1.jrs.8.083516.

Zwieback, S., Kokelj, S.V., Günther, F., Boike, J., Grosse, G., Hajnsek, I., 2018. Sub- seasonal thaw slump mass wasting is not consistently energy limited at the landscape scale. Cryosphere 12, 549–564. https://doi.org/10.5194/tc-12-549-2018.

Zwieback, S., Meyer, F.J., 2021. Top-of-permafrost ground ice indicated by remotely sensed late-season subsidence. Cryosphere 15, 2041–2055. https://doi.org/ 10.5194/tc-15-2041-2021.

J. Chen et al.