Subsidence in Qinghai—Tibet Plateau Peatlands Driven by Drainage Disturbance and Climatic Variability

Abstract

Peatlands are globally important carbon sinks, yet these are increasingly threatened by climate change and human disturbances. Among degradation indicators, surface subsidence is gradual and challenging to monitor, particularly in alpine peatlands. This study applied SBAS-InSAR techniques to analyze surface deformation in the Zoige peatland of the eastern Qinghai—Tibet Plateau using Sentinel-1 SAR data from 2017 to 2023. The results showed a maximum interannual subsidence of −167.92 mm and a peak seasonal deformation of −144.11 mm, with a cumulative average of −23.99 mm (−3.43 mm·yr⁻¹). Approximately 80.9% of peatlands within the protected area exhibited subsidence. Drainage ditch construction emerged as the dominant driver, while climatic factors such as precipitation and temperature exhibited seasonal effects. Subsidence was more pronounced in drier years and during winter months. These findings highlight the spatial heterogeneity and seasonal dynamics of peatland subsidence and underscore the urgent need for hydrological restoration and long-term monitoring to mitigate degradation in alpine peatland ecosystems.

1. Introduction

Peatlands are formed through the slow accumulation of partially decomposed organic matter and represent one of the most effective long-term carbon sinks on Earth [1]. Although they occupy only 2.8–3.0% of the global land surface, peatlands store approximately 50–70 billion tons of carbon—about one-quarter of the global soil carbon pool—making them vital for climate regulation [2,3,4]. In addition to their role in carbon storage, peatlands contribute to water regulation, flood mitigation, and biodiversity support [5,6]. However, extensive peatland degradation has been observed globally due to land conversion, drainage, and climate change. Since 1700, an estimated 10–15% of global peatlands have been disturbed, with substantial losses reported in Southeast Asia, Europe, and North America [7,8].

With the growing concern about peatland degradation, remote sensing—particularly Interferometric Synthetic Aperture Radar (InSAR)—is increasingly recognized as a reliable tool for quantifying vertical deformation related to drainage and carbon loss. Recent studies across Southeast Asia, northern Europe, and boreal regions demonstrate the capability of InSAR time series to detect peat subsidence associated with water-table decline, land-use change, and restoration efforts [9,10,11]. These advances provide valuable global references for large-scale peatland monitoring and form a basis for assessments in alpine systems such as the Zoige Plateau.

Peatland degradation is typically characterized by declining water tables, altered vegetation composition, increased carbon emissions, and surface subsidence [9,12]. Among these, subsidence reflects cumulative hydrological and structural degradation but is often rarely addressed due to its gradual nature and the technical challenges of monitoring it accurately over large, remote areas [10]. Recent studies have reported high subsidence rates in lowland agricultural peatlands—for instance, up to 50 cm·yr⁻¹ in Indonesia and 9.5 cm·yr⁻¹ in drained peat forests of Europe [13,14,15,16]. However, high-altitude peatlands remain insufficiently characterized, especially concerning long-term deformation and its climatic or anthropogenic drivers [11,17,18,19].

The Zoige Plateau, located on the eastern Qinghai—Tibet Plateau, contains one of the world’s largest high-altitude peatland complexes [20]. It plays a key role in carbon sequestration and alpine hydrological regulation [21]. Since the 1960s, widespread drainage, peat reclamation, and overgrazing have led to significant hydrological alteration and ecosystem degradation in the region [22]. Drainage-induced drying and compaction have caused observable surface subsidence, yet its magnitude remains poorly quantified [23,24,25,26,27,28]. Given the vulnerability of this ecosystem to both climatic extremes and human disturbance, reliable, large-scale monitoring tools are urgently needed to assess ongoing changes and inform conservation strategies.

Synthetic Aperture Radar (SAR), particularly Sentinel-1 data combined with InSAR techniques, offers a cost-effective and weather-independent approach for subsidence monitoring [29,30,31,32,33,34]. In recent years, InSAR has been increasingly applied to peatland ecosystems worldwide. For example, one study [35] used the APSIS-DInSAR technique to map spatial patterns of tropical peatland subsidence in Malaysia, while another [36] compared subsidence between drained and restored peatlands in Indonesia using Sentinel-1 data. In northern temperate zones, validation of InSAR-derived deformation against field observations in UK peatlands demonstrated high consistency and potential for long-term ecological monitoring [11]. Similar approaches have also been applied to cold-region peatlands, where InSAR revealed thaw-related deformation in permafrost-affected bogs [37]. Together, these studies highlight the versatility of InSAR for quantifying peatland degradation under diverse climatic and management conditions, providing valuable methodological insight for alpine ecosystems such as the Zoige Plateau.

Among InSAR methods, the SBAS-InSAR technique is particularly effective for capturing gradual, spatially distributed subsidence in vegetated and hydrologically dynamic environments such as peatlands [38,39,40,41]. This study advances the current knowledge by providing the first long-term SBAS-InSAR assessment of alpine peatland deformation on the Qinghai—Tibet Plateau and by integrating climatic and anthropogenic factors through a GeoDetector framework to quantitatively disentangle their relative influences. The approach offers a transferable framework for monitoring and attributing peatland subsidence in other high-altitude ecosystems. Accordingly, this study aims to: (1) quantify the magnitude and spatial–temporal patterns of peatland subsidence in the Zoige region using SBAS-InSAR; (2) identify and quantify the relative influence of climatic and anthropogenic factors through GeoDetector analysis; and (3) evaluate the implications of these findings for hydrological restoration and peatland management in alpine ecosystems.

2. Materials and Methods

2.1. Study Area

This study was conducted in the Zoige Plateau Basin, located in Sichuan Province on the eastern margin of the Qinghai—Tibet Plateau. This region contains the world’s largest and best-preserved alpine peatland system, shaped by unique geological, hydrological, and climatic conditions (Figure 1). A national survey in the 1980s identified 220 large and medium-sized peatlands, covering approximately 2228 km² [42,43]. The climate is cold and humid, with long winters and a short growing season. The mean temperature in January is −10.6 °C, in July it is 10.8 °C, and the annual mean is 1.4 °C. Annual precipitation averages 720 mm [44].

2.2. SAR Data

A total of 574 Sentinel-1A SAR images (IW mode, VV (vertical–vertical) polarization) acquired between January 2017 and December 2023 were used for surface deformation analysis. The data were obtained from the Copernicus Open Access Hub (https://www.copernicus.eu/en/access-data/conventional-data-access-hubs, accessed on 25 May 2024) and processed at Level-1 Single Look Complex (SLC) format. Each scene has a spatial resolution of 5 m (range) × 20 m (azimuth). Approximately 10–12 scenes per year were selected based on coverage and coherence.

2.3. Peatland Classification Using Multi-Source Remote Sensing

To extract peatland distribution and support subsidence interpretation, a land cover classification was developed on the Google Earth Engine (GEE) platform using Sentinel-2 multispectral imagery and Sentinel-1 SAR backscatter data. Seventeen input features, including spectral bands and indices, were used (e.g., NDVI, NDWI, ARI, SDWI, VV, VH;

Table 1. Remote sensing feature index image data.

Data	Source	Formula	Description
B2	Sentinel-2	-	Blue (492.4 nm)
B3	Sentinel-2	-	Green (559.8 nm)
B4	Sentinel-2	-	Red (664.6 nm)
B5	Sentinel-2	-	Vegetation Red Edge (703.9 nm)
B6	Sentinel-2	-	Vegetation Red Edge (740.2 nm)
B7	Sentinel-2	-	Vegetation Red Edge (782.5 nm)
B8	Sentinel-2	-	NIR (832.8 nm)
B8A	Sentinel-2	-	Narrow NIR (865 nm)
B11	Sentinel-2	-	SWIR (1613.7 nm)
NDVI	Sentinel-2	$\left({\displaystyle \frac{Band8-Band4}{Band8+Band4}}\right)$	Normalized Vegetation Index. Detecting Vegetation Growth Status, Vegetation Cover and Eliminating Partial Radiation Errors, etc. [45]
NDWI	Sentinel-2	$\left({\displaystyle \frac{Band3-Band8}{Band3+Band8}}\right)$	Normalized water index. Extraction of vegetation water information based on water absorption characteristics in the near-infrared and short-wave infrared ranges, and light penetration in the near-infrared range [46]
ARI	Sentinel-2	$\left({\displaystyle \frac{Band8}{Band2}}\right)-\left({\displaystyle \frac{Band8}{Band3}}\right)$	Anthocyanin Reflectance Index. An index of sensitivity to anthocyanin pigments in plant leaves, often associated with plant stress or senescence [47]
PSRI	Sentinel-2	$\left({\displaystyle \frac{Band4-Band2}{Band5}}\right)$	Vegetation Decay Index. Detecting plant initiation and senescence using carotenoid to chlorophyll ratios [48]
REIP	Sentinel-2	$702+40 \left({\displaystyle \frac{\left(\frac{Band2+Band7}{2}\right)-Band5}{Band6-Band5}}\right)$	Red edge tilt points. Approximation of the hyperspectral index used to estimate the position of the NIR/red inflection point in the vegetation spectrum [49]
SDWI	Sentinel-1	In (10 × VV × VH) − 8	Sentinel-1 Dual-Polarized Water Index. The proposed SDWI (Sentinel-1 Dual-Polarized Water Index) water body information extraction method based on Sentinel-1 satellite data containing C-band synthetic aperture radar for the identification of water body information on a large scale [50]
VV	Sentinel-1	-	Polarized radar backward scattering coefficient. Vertical polarization sends decibel backward scattering from vertically polarized receiving SAR. The polarization of electromagnetic waves is sensitive to the dielectric constant, physical properties, geometry and orientation of the target, and thus polarization measurements can greatly improve the acquisition of various information about the target by imaging radar
VH	Sentinel-1	-	Polarized radar backward scattering coefficient. Decibel backscatter from vertically polarized transmitter and horizontally polarized receiver SARs

), and training samples were collected from field surveys and high-resolution imagery. A Random Forest classifier was used for supervised classification, and accuracy was assessed using a confusion matrix with a 70/30 train-validation split. Overall accuracy reached 88.75%, with a Kappa coefficient of 0.82. Peatland and water bodies showed the highest classification accuracy due to their distinct spectral characteristics, while impervious surfaces had relatively lower accuracy.

2.4. InSAR Processing and Deformation Retrieval

Surface deformation was retrieved using SBAS-InSAR method [40,41,51]. SBAS-InSAR processing was conducted in ENVI SARscape 5.6 following a standard four-step workflow: (1) Interferogram Network Construction: Image pairs with spatial baseline < 200 m and temporal baseline < 180 days were selected. A super-master image was chosen to maximize coherence. (2) Interferometric Processing: Co-registration, interferogram generation, topographic phase removal (using the 30 m SRTM DEM), coherence estimation, and phase unwrapping were performed. Multi-looking was set to 8 × 2 (range × azimuth). (3) Time Series Inversion: Atmospheric noise was suppressed using spatial low-pass filtering (radius: 1600 m) and temporal high-pass filtering (365-day window). Displacement time series were resolved using Singular Value Decomposition (SVD). (4) Geocoding and LOS-to-Vertical Conversion: Deformation was geocoded and converted from line-of-sight to vertical displacement assuming near-vertical radar incidence.

2.5. Accuracy Assessment

The accuracy of SBAS-InSAR–derived deformation was evaluated using both internal indicators and external reference data. Interferometric coherence $\left(\gamma \right)$ was used as the primary internal quality metric to assess the reliability of phase information across image pairs.

Coherence was calculated within a local moving window using the following expression:

$$\gamma ={\displaystyle \frac{\left|〈S_1·S_2^{*}〉\right|}{\sqrt{〈{\left|S_1\right|}^2〉·〈{\left|S_2\right|}^2〉}}}$$

(1)

where $S_1$ and $S_2^{*}$ denote the complex SAR images from the master and slave acquisitions, and $⟨·⟩$ represents spatial averaging over a defined kernel (5 × 5 pixels). Pixels with persistently low coherence were excluded from time-series analysis to reduce phase noise and improve result stability.

External validation was conducted using GNSS time series data from the GSMA station (longitude 102.06° E, latitude 34.02° N; elevation 3562.84 m), obtained from the Eastern Deformation Data Center (https://www.eqdsc.com/, accessed on 25 March 2025). Due to the absence of GNSS stations within the study area, this station—located approximately 35 km from the monitoring area—was used as the nearest available reference. The SBAS-InSAR pixel closest to the GNSS location was extracted, and the two displacement time series were compared using root mean square error (RMSE) to evaluate consistency.

2.6. Analysis of Subsidence Drivers

To explore the drivers of peatland subsidence, two types of explanatory variables were analyzed: climatic and anthropogenic. Climate variables included mean annual temperature (MAT) and mean annual precipitation (MAP), derived from the China 1 km gridded dataset (National Earth System Science Data Center, http://www.geodata.cn, accessed on 20 June 2025). Changes were calculated by comparing 1961–1990 and 1991–2023 averages. Anthropogenic variables included livestock density (2012–2022), distance to drainage ditches (DTD), and distance to roads (DTR).

Spatial relationships between subsidence and these variables were assessed using the GeoDetector model [52,53], which quantifies the explanatory power (q) of each factor based on spatial stratified heterogeneity. The underlying assumption of the model is that if an explanatory factor significantly influences the spatial distribution of a dependent variable, the spatial patterns of the two should be similar. Continuous variables were stratified using the natural breaks (Jenks) classification method to minimize within-group variance and maximize between-group variance. The q-statistic is calculated as follows:

$$q=1-{\displaystyle \frac{{\sum }_{h=1}^L{N}_h{\sigma }_h^2}{N{\sigma }^2}}$$

(2)

where $N_h$ is the number of samples in the hth stratum, ${\sigma }_h^2$ is the variance within the hth stratum, $N$ is the total number of samples, and ${\sigma }^2$ is the overall variance. All variables were resampled to a common spatial resolution and aligned with the InSAR-derived subsidence maps for pixel-wise comparison. The GeoDetector analysis was conducted using GeoDetector v2.0 (available at http://www.geodetector.cn, accessed on 20 June 2025) implemented in Arcgis 10.8. In addition, we incorporated long-term daily snow depth data from the China long-term snow depth dataset (1978–2024, 0.25° resolution; National Tibetan Plateau Data Center, http://data.tpdc.ac.cn, accessed on 10 August 2025) [54], which were aggregated to winter means for comparison with subsidence. Detailed procedures and results are provided in Supplementary Table S1.

3. Results

3.1. Peatland Classification Accuracy

Land cover classification based on Sentinel-2 and Sentinel-1 data using the Random Forest algorithm successfully delineated major land cover types within the Zoige region (Figure 2). The classified categories included water bodies, peatlands, grasslands, forests, and impervious surfaces. Among these, peatlands accounted for 953 km² within the protected area, followed by grassland (539 km²), water surface (392 km²), built-up land (122 km²), and forest (48 km²).

The classification model achieved an overall accuracy of 88.75% and a Kappa coefficient of 0.82. Peatland and water surfaces exhibited the highest classification accuracy, with producer and user accuracies exceeding 90%. Impervious surfaces had the lowest accuracy, likely due to their small area and spectral confusion with other land types.

3.2. SBAS-InSAR Accuracy Evaluation

Temporal coherence (γ) varied across both time and space throughout the monitoring period. Based on interferometric stack averaging, annual γ declined from 0.41 ± 0.09 in 2017 to 0.24 ± 0.12 in 2021, with a partial recovery to 0.27 ± 0.11 in 2023. Spatially, coherence was higher in flat and unvegetated areas, while peat-rich zones with dense vegetation and complex topography exhibited lower coherence (Supplementary Figure S1).

Due to the absence of GNSS stations within the study area, validation of SBAS-InSAR displacement time series was performed using the GSMA GNSS station (102.06° E, 34.02° N; 3562.84 m a.s.l.), located approximately 35 km from the monitoring region. After spatial alignment and projection of InSAR line-of-sight (LOS) deformation into vertical displacement, the SBAS-InSAR and GNSS time series showed strong temporal consistency, with a correlation coefficient (R) of 0.61, a mean bias of −0.30 mm, and a root mean square error (RMSE) of 4.43 mm (Figure 3).

Although the general deformation trends were consistent, the GNSS data exhibited a larger seasonal amplitude (up to ±10 mm) than the SBAS-InSAR results. This difference reflects the contrasting characteristics of the two observation techniques: GNSS provides point-based daily measurements that record localized frost-heave and soil-moisture variations, while SBAS-InSAR represents deformation averaged over 30 m pixels and 6–12-day intervals, which smooths short-term changes. Small residual errors introduced during LOS-to-vertical projection may also contribute to the difference in amplitude. Despite these discrepancies, the consistent overall patterns and the low RMSE indicate that the SBAS-InSAR results are reliable for long-term monitoring of peatland surface deformation in the eastern Qinghai—Tibet Plateau.

3.3. Seasonal Patterns of Subsidence

Seasonal deformation patterns derived from SBAS-InSAR revealed substantial intra-annual variability across the Zoige peatland. The most pronounced seasonal deformation occurred in December 2018, reaching −144.11 mm. In contrast, 2023 exhibited the smallest annual mean seasonal deformation (−5.05 mm), whereas minor surface uplift was observed in 2017 (2.18 mm) (see Figure 4 for 2023 and Supplementary Figure S2 for other years).

Winter subsidence was consistently more pronounced, particularly in central and northern regions. For instance, in 2023, the average seasonal deformation reached −14.88 mm in the central zone, compared to −1.68 mm in the south. Spatial variation also showed that western and northern regions experienced moderate seasonal deformation (−7.33 mm and −7.14 mm, respectively), while eastern regions had the lowest seasonal rates (−3.40 mm). These seasonal patterns reflected interactions between freeze–thaw processes, vegetation growth cycles, and anthropogenic stress. A comparison between winter mean snow depth and InSAR-derived subsidence during 2017–2023 showed no significant correlation, suggesting that the pronounced winter subsidence was not directly controlled by interannual snow cover variations (see Supplementary Table S1 and Figure S3).

3.4. Interannual Patterns of Subsidence

Interannual deformation from 2017 to 2023 revealed persistent and heterogeneous subsidence across the peatland (Figure 5). Maximum deformation occurred in 2020 (−167.92 mm), followed by −163.68 mm in 2018 and −158.83 mm in 2017. The annual average subsidence over the study period was −3.43 mm·yr⁻¹, with a cumulative displacement of −23.99 mm.

Regionally, 74.2% of peatlands showed measurable subsidence, with 80.9% of those within the protected area also exhibiting vertical deformation. The central region displayed the greatest average annual subsidence (−28.92 mm·yr⁻¹), followed by the west (−11.79 mm), north (−10.9 mm), and east (−6.82 mm). The southern region showed the least deformation (−4.25 mm·yr⁻¹), indicating spatially uneven vulnerability to subsidence. Comparison across land cover types revealed the highest subsidence rate in peatlands (−3.43 mm·yr⁻¹), followed by water bodies (−2.53 mm), impervious surfaces (−1.86 mm), grasslands (−0.90 mm), and forests (−0.77 mm). Notably, subsidence within protected areas was more severe than outside, suggesting persistent degradation despite formal conservation status.

3.5. Drivers of Peatland Subsidence

The GeoDetector analysis quantified the relative contributions of anthropogenic and climatic variables to peatland subsidence across the Zoige region. Among all factors, drainage-ditch proximity (DTD) exhibited the highest explanatory power (q = 0.89) (Figure 6), indicating a strong spatial correspondence between artificial drainage networks and deformation intensity. Road proximity (DTR) showed a lower but measurable influence (q = 0.08), suggesting localized deformation around infrastructure corridors. Climatic variables—mean annual precipitation (MAP) and mean annual temperature (MAT)—displayed comparatively weak explanatory power, confirming that human-induced disturbance is the dominant control on the spatial pattern of subsidence.

These quantitative results demonstrate that anthropogenic disturbances exert stronger spatial control on peatland subsidence than climatic variability. These findings provide a quantitative foundation for subsequent discussion of anthropogenic and climatic mechanisms.

4. Discussion

4.1. Anthropogenic Drivers and Mechanisms of Subsidence

Anthropogenic disturbances—particularly drainage—are the dominant drivers of peatland subsidence in the Zoige region. Proximity to drainage ditches revealed a strong spatial alignment between artificial hydrological structures and vertical deformation patterns (Figure 6). Drainage networks, which exceeded 1700 km by 2000, have disrupted natural hydrological conditions and lowered the regional groundwater table [55,56]. As oxygen exposure increases, aerobic decomposition of organic matter accelerates, promoting irreversible peat compaction and volume loss [57]. Although road influence was weaker than that of drainage, it still contributed to localized deformation. Road construction can alter surface water flow, fragment habitats, and cause long-term compaction. This is consistent with field observations suggesting that infrastructure impairs surface resilience to environmental stress. Overgrazing further exacerbates subsidence by degrading vegetation cover and increasing soil bulk density. The livestock population in Zoige County has grown from 950,700 in 1958 to 2.27 million sheep units in 2022, exerting sustained pressure on the peatland surface. Livestock trampling reduces soil porosity, limits water infiltration, and compromises peat integrity [58,59]. This chronic stress leads to a gradual decline in peatland structure and increases susceptibility to deformation.

4.2. Climatic Modulation and Seasonal Feedback

In addition to human-induced factors, climatic variability modulates seasonal patterns of peatland subsidence. The Geo-Detector results showed that precipitation and temperature exerted relatively weak but seasonally dependent effects (Figure 7). Subsidence was more pronounced during dry periods or drought years, reflecting the role of low water content in reducing the stability of peat structures [60]. Reduced precipitation lowers soil moisture, causing shrinkage and weakening the capillary forces that maintain peat matrix cohesion. Conversely, elevated temperatures can promote evapotranspiration and accelerate microbial activity, facilitating further organic matter loss and subsurface volume reduction. These processes reflect a positive feedback mechanism: while climate factors may not dominate spatial variation, they significantly influence temporal sensitivity. Notably, the influence of natural factors was most pronounced in winter, suggesting a possible role for freeze–thaw dynamics. Cyclical freezing and thawing destabilize the peat surface, alter pore structure, and may induce delayed compaction. These effects are well documented in permafrost and alpine peatlands and may contribute to the enhanced deformation observed in cold-season periods. Consistent with this interpretation, the analysis of AMSR2 (Advanced Microwave Scanning Radiometer 2)-derived snow depth data indicated that winter snow depth across Zoige remained shallow (mostly <1 cm), and its interannual variation did not significantly correlate with winter subsidence (Pearson r = −0.04, p = 0.93; Spearman r = 0.36, p = 0.43; see Supplementary Table S1). This implies that freeze–thaw processes strongly modulate seasonal deformation, but snow depth itself is not a dominant explanatory factor at the resolution of available datasets. In contrast, anthropogenic factors such as ditch networks and grazing exert relatively stable influence across seasons. This suggests that while climate introduces seasonal pulses of stress, land-use patterns determine the long-term trajectory and cumulative magnitude of subsidence. Recent InSAR studies reveal clear climatic contrasts in peatland subsidence. In tropical regions, drainage-driven compaction dominates, with deformation rates often exceeding tens of centimeters per year in Malaysia and Indonesia [61,62]. In temperate Europe, deformation is slower but seasonally dynamic, typically below 2 cm yr⁻¹ and partly reversible after rewetting [63,64]. Compared with these systems, the Zoige Plateau exhibits an intermediate pattern—spatially extensive but moderate in magnitude—where anthropogenic drainage and grazing outweigh climatic forcing. This comparison highlights drainage-induced hydrological alteration as the most consistent global driver of peatland subsidence.

Although the InSAR analysis provides spatially consistent measurements, several uncertainties should be acknowledged. Residual atmospheric phase delays and temporal decorrelation over vegetated surfaces may introduce local noise in the deformation estimates. The AMSR2-derived snow-depth dataset, with its coarse 0.25° resolution, may obscure small-scale variability in snow accumulation and melt [65]. In addition, the lack of ground-based leveling or GNSS observations limits the validation of InSAR-derived results. Future integration of in situ and higher-resolution datasets will help reduce these uncertainties.

4.3. Management Implications and Regional Strategies

Our findings reveal that subsidence occurs not only in degraded areas but also within formally protected peatland zones. This suggests that current conservation measures—primarily passive protection—are insufficient to curb hydrological degradation and surface instability. The persistence of drainage-induced impacts within reserves indicates a pressing need for proactive management. Hydrological restoration should be prioritized, especially in areas heavily affected by ditch construction. Rewetting via ditch blocking, water table recovery, and soil moisture retention can significantly reduce peat degradation and limit further subsidence [42,66]. These interventions have shown success in northern temperate peatlands and can be adapted to the high-altitude Zoige landscape. Grazing management is equally essential. To reduce trampling-induced compaction, it is advisable to regulate stocking densities, implement rotational grazing systems, and establish exclusion zones. Considering the overlap between grazing hotspots and areas of severe subsidence, targeted management would likely yield measurable ecological benefits. Finally, long-term monitoring combining SBAS-InSAR and ground-based GNSS data should be institutionalized to track recovery and detect new risks. SBAS-InSAR provides the temporal resolution and spatial coverage needed to evaluate restoration outcomes. Nevertheless, our study is constrained by the lack of in situ measurements within the peatland, and the weak correlations observed between snow depth and subsidence highlight the limitations of current coarse-resolution climatic datasets. Future work should integrate higher-resolution snow and permafrost observations, continuous hydrological monitoring, and process-based ecohydrological models to complement InSAR analysis. Adaptive management informed by multi-source observations and scenario modeling can better enhance the resilience of peatlands under intensifying climate and anthropogenic stress.

5. Conclusions

This study employed SBAS-InSAR techniques to quantify surface deformation in the Zoige peatland from 2017 to 2023. The results revealed persistent and spatially variable subsidence, with peak seasonal and interannual values reaching −144.11 mm and −167.92 mm, respectively. Cumulative average subsidence was −23.99 mm, equivalent to an annual rate of −3.43 mm·yr⁻¹. Approximately 80.9% of peatlands within the protected area exhibited detectable subsidence. Drainage ditch proximity emerged as the dominant driver, followed by road construction and livestock density. Climatic factors such as reduced precipitation and elevated temperature contributed to seasonal variability but had weaker spatial explanatory power. Subsidence was most severe in central peatland zones and during winter periods. These findings demonstrate that even protected peatlands remain vulnerable to subsurface instability driven by hydrological alteration and land cover intensification. Long-term monitoring, targeted hydrological restoration, and regulation of anthropogenic disturbances are essential to mitigate degradation and preserve the ecological integrity of high-altitude peatland ecosystems. Specifically, reducing artificial drainage density, controlling grazing intensity, and maintaining higher groundwater levels are recommended management actions to alleviate subsidence and enhance the resilience of alpine peatland systems.