Yang et al. (2025) Unraveling three-decade dynamics and drivers of thermokarst lakes on the Tibetan Plateau

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2025
• Date: 2025-12-11
• Authors: Guoqing Yang, Haijun Qiu, Ninglian Wang, Dongdong Yang, Ya Liu, Kailiang Zhao
• DOI: 10.1016/j.jag.2025.105022

Research Groups

• Shaanxi Key Laboratory of Earth Surface and Environmental Carrying Capacity, College of Urban and Environmental Sciences, Northwest University, Xi’an 710127, China
• Institute of Earth Surface System and Hazards, College of Urban and Environmental Sciences, Northwest University, Xi’an 710127, China

Short Summary

This study analyzed the three-decade dynamics (1990-2022) of 58,538 thermokarst lakes on the Tibetan Plateau using remote sensing and machine learning, revealing that over 82% of lakes expanded, primarily driven by topographic and climatic factors, while 15% shrank due to evaporation and soil temperature trends.

Objective

• To unravel the three-decade dynamics (1990-2022) and identify the primary environmental and climatic drivers of thermokarst lakes (TLs) on the Tibetan Plateau, distinguishing factors for lake expansion and shrinkage.

Study Configuration

• Spatial Scale: Entire permafrost region of the Tibetan Plateau, covering an area greater than 1.2 × 10^6 km², focusing on 58,538 thermokarst lakes ranging from 0.01 km² to 3 km² in area.
• Temporal Scale: 33 years, from 1990 to 2022.

Methodology and Data

• Models used: LandTrendr algorithm for detecting disturbance years, CatBoost (optimal machine learning regression model), LightGBM, XGBoost, NGBoost (for comparison), and Shapley additive explanations (SHAP) for variable importance ranking.
• Data sources:
  • Landsat-derived Global Surface Water Dataset (GWSD) from the Joint Research Centre (JRC) of the European Commission, specifically "max_extent" and "transition" products.
  • Time-series Landsat imagery (1990-2022).
  • ERA5-Land monthly reanalysis data (1990-2022) for climatic variables (air temperature, soil temperature, precipitation, solar radiation, evaporation, soil water volume, and runoff).
  • Harmonized World Soil Database (FAO and IIASA, 2023) for soil types.
  • Field surveys in the Hoh Xil region (June 2024 and June 2025) using unmanned aerial vehicle (UAV) orthophoto maps and portable GPS measurements (referenced from Wei et al., 2021).
  • TimeSync tool for temporal validation of disturbance years.
  • High-resolution historical images from ESRI World Wayback for visual validation.

Main Results

• A total of 58,538 thermokarst lakes (0.01–3 km²) were identified on the Tibetan Plateau, primarily distributed in the central inland permafrost regions.
• 82.1% of TLs experienced an increase in area (59.1% expanding, 23.0% emerging), while 15% experienced shrinkage (11.2% shrinking, 3.8% vanishing). Only 2.9% remained unchanged.
• The annual number of lake expansion events peaked in 2016, with approximately 40% occurring between 2014 and 2020. Lake shrinkage was most common in 2019, with peak activity during 2017–2022.
• Spatially, over 92.3% of TLs were located between 82°E and 98°E, and 92.5% between 32°N and 38°N. Most TLs (81.6%) were found at elevations below 5,000 meters and in gentle areas (93.2% with slopes less than 5°).
• 83.6% of TLs were concentrated in continuous permafrost regions, and 60.9% were in endorheic basins, particularly the inner plateau (57.3%).
• Accuracy assessment showed a cumulative area error of ± 2.43 km² for all TLs. Average relative area errors were 32.9% for small lakes, 13.8% for medium, and 6.3% for large lakes. Temporal validation of disturbance years yielded high consistency (R² > 0.87) with mean absolute errors of 1.54 years for gain-type and 1.29 years for loss-type lakes.
• Machine learning models identified topographic and climatic factors as primary drivers. Slope was the most important factor, with TLs more common on slopes less than 2°.
• Lake expansion (gain-type TLs) was positively influenced by ground ice content (optimal between 23–35%), precipitation trend, and soil water volume trend.
• Lake shrinkage (loss-type TLs) was primarily influenced by increased evaporation trend and decreased soil temperature trend.

Contributions

• Provides the first comprehensive, long-term (three-decade) analysis of thermokarst lake dynamics across the entire Tibetan Plateau, addressing a significant knowledge gap.
• Quantitatively identifies and distinguishes the specific environmental and climatic drivers for both thermokarst lake expansion and shrinkage using advanced machine learning techniques.
• Offers valuable insights for understanding permafrost degradation, improving carbon sequestration estimations, and informing infrastructure maintenance strategies on the Tibetan Plateau.

Funding

• National Natural Science Foundation of China (42271078, 42301090, 42471083)
• China Postdoctoral Science Foundation (2022M722564)

Citation

@article{Yang2025Unraveling,
  author = {Yang, Guoqing and Qiu, Haijun and Wang, Ninglian and Yang, Dongdong and Liu, Ya and Zhao, Kailiang},
  title = {Unraveling three-decade dynamics and drivers of thermokarst lakes on the Tibetan Plateau},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2025},
  doi = {10.1016/j.jag.2025.105022},
  url = {https://doi.org/10.1016/j.jag.2025.105022}
}