Zhang et al. (2026) Long-Term Evolution of Permafrost across the Qinghai-Tibet Plateau: Perspectives from Multi-Model Ensembles and Machine Learning

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Journal of Climate
• Year: 2026
• Date: 2026-03-03
• Authors: Ting Zhang, Guihua Yi, Xiaobing Zhou, Guihua Yi, Xiaojuan Bie, Guoyan Wang, Bo Wen, Mingqi Liu
• DOI: 10.1175/jcli-d-25-0473.1

Research Groups

[Information not available in the provided abstract.]

Short Summary

This study combined CMIP6 data with machine learning models to project permafrost extent and maximum seasonal soil freeze depth (SFD) across the Qinghai-Tibet Plateau (QTP) from 2025 to 2100 under various SSP scenarios. Results indicate continuous permafrost degradation into seasonally frozen ground, with SFD declining significantly, and specific high-risk zones identified, with the Deep Neural Network (DNN) model demonstrating superior performance.

Objective

• To construct optimal machine learning models for predicting permafrost extent and maximum seasonal soil freeze depth (SFD) across the Qinghai-Tibet Plateau (QTP).
• To project permafrost evolution from 2025 to 2100 under four Shared Socioeconomic Pathway scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5).
• To evaluate the relative importance and marginal effects of predictors on permafrost dynamics.

Study Configuration

• Spatial Scale: Qinghai-Tibet Plateau (QTP).
• Temporal Scale: Projections from 2025 to 2100.

Methodology and Data

• Models used: Support Vector Regression (SVR), Categorical Boosting (CatBoost), Light Gradient Boosting Machine (LightGBM), and Deep Neural Network (DNN) for permafrost and SFD prediction; Random Forest (RF) for relative importance evaluation; Shapley Additive exPlanations (SHAP) for marginal effects evaluation.
• Data sources: Coupled Model Intercomparison Project Phase 6 (CMIP6) data.

Main Results

• The Deep Neural Network (DNN) model outperformed other models in capturing frozen-ground dynamics, achieving an R² of 0.961.
• Key drivers for maximum seasonal soil freeze depth (SFD) were identified as summer-autumn precipitation, temperature, and non-frozen season soil moisture.
• Permafrost is projected to continuously degrade into seasonally frozen ground, with transition zones expanding from plateau margins toward the interior, leading to increased spatial fragmentation.
• SFD is expected to decline at rates strongly correlated with emission scenarios, with the most pronounced reductions along current permafrost margins and gradually extending inland.
• Permafrost stability in high-altitude regions will face increasing challenges.
• High-risk zones for frozen-ground degradation under continued CO₂ emission trajectories include the Northern Tibetan Plateau, the margins of the Qaidam Basin, and the foothills of the Gangdise Mountains.

Contributions

• Development and application of optimal machine learning models, particularly DNN, for high-accuracy prediction of permafrost extent and seasonal soil freeze depth on the QTP.
• Comprehensive projections of permafrost degradation patterns and SFD decline across the QTP under multiple future climate scenarios (SSPs).
• Identification and quantification of key climatic and hydrological drivers influencing SFD dynamics using advanced interpretability methods (RF and SHAP).
• Identification of specific high-risk zones for frozen-ground degradation on the QTP, providing critical information for regional climate change adaptation strategies.

Funding

[Information not available in the provided abstract.]

Citation

@article{Zhang2026LongTerm,
  author = {Zhang, Ting and Yi, Guihua and Zhou, Xiaobing and Yi, Guihua and Bie, Xiaojuan and Wang, Guoyan and Wen, Bo and Liu, Mingqi},
  title = {Long-Term Evolution of Permafrost across the Qinghai-Tibet Plateau: Perspectives from Multi-Model Ensembles and Machine Learning},
  journal = {Journal of Climate},
  year = {2026},
  doi = {10.1175/jcli-d-25-0473.1},
  url = {https://doi.org/10.1175/jcli-d-25-0473.1}
}