1.School of Energy Resources,China University of Geosciences (Beijing),Beijing 100083,China
2.School of Artificial Intelligence,China University of Geosciences (Beijing),Beijing 100083,China
3.Oil and Gas Survey,China Geological Survey,Beijing 100083,China
4.Observation and Research Station of Gas Hydrate and Permafrost Environment in Muli Town (Qinghai Province),Ministry of Natural Resources,Beijing 100083,China
The accelerating degradation of permafrost on the Qinghai-Tibet Plateau poses severe challenges to regional ecological security and infrastructure construction, urgently requiring the development of permafrost temperature profile prediction methods that combine high accuracy with interpretability. This study proposes a prediction method that combines an Enhanced Temporal Convolutional Network (TCN-Plus) deep learning architecture with symbolic regression, balancing prediction accuracy and physical interpretability. Validation using measured data from the Muli field observation station in 2021 demonstrates that TCN-Plus achieves a Mean Absolute Error (MAE) of 0.187 ℃ at 0.5 m depth, representing reductions of 67.8%, 35.7%, and 24.3% compared to Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Transformer, respectively. The study reveals differentiated dominant mechanisms at various depths: shallow layers are controlled by thermal inertia, middle layers are influenced by hydrothermal coupling, and deep layers exhibit interannual memory effects; all models show maximum prediction errors in the 2.5 - 3.0 m depth interval, indicating that deep-layer predictions require consideration of moisture migration and latent heat of phase change. The proposed method achieves dual improvements in prediction accuracy and physical interpretability; the research findings can provide technical support for thermal stability assessment of engineering construction in permafrost regions of the Qinghai-Tibet Plateau, offer new insights for interpretable applications of deep learning in geophysical fields, and hold practical value for permafrost monitoring and climate change impact assessment on the Qinghai-Tibet Plateau.
SCHUURE A G, MCGUIREA D, SCHÄDELC, et al. Climate change and the permafrost carbon feedback[J]. Nature, 2015, 520(7546): 171 - 179.
[2]
HUGELIUSG, STRAUSSJ, ZUBRZYCKIS, et al. Estimated stocks of circumpolar permafrost carbon with quantified uncertainty ranges and identified data gaps[J]. Biogeosciences, 2014, 11(23): 6573 - 6593.
[3]
BISKABORNB K, SMITHS L, NOETZLIJ, et al. Permafrost is warming at a global scale[J]. Nature communications, 2019, 10(1): 264.
[4]
ZOUD, ZHAOL, SHENGY, et al. A new map of permafrost distribution on the Tibetan Plateau[J]. The Cryosphere, 2017, 11(6): 2527 - 2542.
WUQ, ZHANGT. Changes in active layer thickness over the Qinghai‐Tibetan Plateau from 1995 to 2007[J]. Journal of Geophysical Research: Atmospheres, 2010, 115(D9):d09107.
[8]
PENGX, ZHANGT, FRAUENFELDO W, et al. Response of seasonal soil freeze depth to climate change across China[J]. The Cryosphere, 2017, 11(3): 1059 - 1073.
NATALIS M, WATTSJ D, ROGERSB M, et al. Large loss of CO2 in winter observed across the northern permafrost region[J]. Nature Climate Change, 2019, 9(11): 852 - 857.
[12]
MAW, MUY, WUQ, et al. Characteristics and mechanisms of embankment deformation along the Qinghai–Tibet Railway in permafrost regions[J]. Cold Regions Science and Technology, 2011, 67(3): 178 - 186.
[13]
YANGY, FANGJ, TANGY, et al. Storage, patterns and controls of soil organic carbon in the Tibetan grasslands[J]. Global change biology, 2008, 14(7): 1592 - 1599.
[14]
NIUF, LUOJ, LINZ, et al. Development and thermal regime of a thaw slump in the Qinghai-Tibet plateau[J]. Cold Regions Science and Technology, 2012, 83: 131 - 138.
[15]
RISEBOROUGHD, SHIKLOMANOVN, ETZELMÜLLERB, et al. Recent advances in permafrost modelling[J]. Permafrost and Periglacial Processes, 2008, 19(2): 137 - 156.
[16]
WESTERMANNS, LANGERM, BOIKEJ, et al. Simulating the thermal regime and thaw processes of ice-rich permafrost ground with the land-surface model CryoGrid 3[J]. Geoscientific Model Development, 2016, 9(2): 523 - 546.
[17]
BAIR, LAIY, ZHANGM, et al. Study on the coupled heat-water-vapor-mechanics process of unsaturated soils[J]. Journal of Hydrology, 2020, 585: 124784.
[18]
WALVOORDM A, KURYLYKB L. Hydrologic impacts of thawing permafrost—A review[J]. Vadose Zone Journal, 2016, 15(6): vzj2016. 01.0010.
CHANCER, AHAJJAMA, PUTKONENJ, et al. Machine Learning for Prediction of Permafrost and Active Layer Temperatures along the North Slope of Alaska[C]//AGU Fall Meeting Abstracts. 2023(1055): C53C - 1055.
[23]
LIR, ZHAOL, WUT, et al. Soil thermal conductivity and its influencing factors at the Tanggula permafrost region on the Qinghai–Tibet Plateau[J]. Agricultural and Forest Meteorology, 2019, 264: 235 - 246.
[24]
WANGT, YANGD, FANGB, et al. Data-driven mapping of the spatial distribution and potential changes of frozen ground over the Tibetan Plateau[J]. Science of the Total Environment, 2019, 649: 515 - 525.
[25]
NIJ, WUT, ZHUX, et al. Simulation of the present and future projection of permafrost on the Qinghai‐Tibet Plateau with statistical and machine learning models[J]. Journal of Geophysical Research: Atmospheres, 2021, 126(2): e2020JD033402.
[26]
LIUQ, NIUJ, LUP, et al. Interannual and seasonal variations of permafrost thaw depth on the Qinghai-Tibetan plateau: A comparative study using long short-term memory, convolutional neural networks, and random forest[J]. Science of The Total Environment, 2022, 838: 155886.
CHANCER, AHAJJAMA, PUTKONENJ, et al. Artificial intelligence for predicting arctic permafrost and active layer temperatures along the Alaskan North Slope[J]. Earth Science Informatics, 2024, 17(6): 6055 - 6073.
[29]
REICHSTEINM, CAMPS-VALLSG, STEVENSB, et al. Deep learning and process understanding for data-driven Earth system science[J]. Nature, 2019, 566(7743): 195 - 204.
[30]
JOSEPHL P, DEOR C, PRASADR, et al. Near real-time wind speed forecast model with bidirectional LSTM networks[J]. Renewable Energy, 2023, 204: 39 - 58.
[31]
GUOD, WANGH. CMIP5 permafrost degradation projection: A comparison among different regions[J]. Journal of Geophysical Research: Atmospheres, 2016, 121(9): 4499 - 4517.
[32]
LIUY, RANY, LIX, et al. Multisite evaluation of physics-informed deep learning for permafrost prediction in the Qinghai-Tibet Plateau[J]. Cold Regions Science and Technology, 2023, 216: 104009.
[33]
BAIS. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling[J]. arxiv preprint arxiv:2018.
[34]
LEAC, FLYNNM D, VIDALR, et al. Temporal convolutional networks for action segmentation and detection[C]//proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 156 - 165.
[35]
CRANMERM. Interpretable machine learning for science with PySR and SymbolicRegression[J]. arxiv preprint arxiv:2023.
HOWARDA G, ZHUM, CHENB, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[J]. arxiv preprint arxiv:2017.
[41]
HUJ, SHENL, SUNG. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132 - 7141.
[42]
ZHANGT, BARRYR G, KNOWLESK, et al. Statistics and characteristics of permafrost and ground‐ice distribution in the Northern Hemisphere[J]. Polar Geography, 1999, 23(2): 132 - 154.
[43]
JORGENSONM T, ROMANOVSKYV, HARDENJ, et al. Resilience and vulnerability of permafrost to climate change[J]. Canadian Journal of Forest Research, 2010, 40(7): 1219 - 1236.
[44]
CLARKM P, BIERKENSM F P, SAMANIEGOL, et al. The evolution of process-based hydrologic models: historical challenges and the collective quest for physical realism[J]. Hydrology and Earth System Sciences, 2017, 21(7): 3427 - 3440.
[45]
ZHANGT. Influence of the seasonal snow cover on the ground thermal regime: An overview[J]. Reviews of Geophysics, 2005, 43(4): RG4002.
[46]
HEK, ZHANGX, RENS, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770 - 778.
[47]
MAJ, LIR, LIUH, et al. The surface energy budget and its impact on the freeze-thaw processes of active layer in permafrost regions of the Qinghai-Tibetan Plateau[J]. Advances in Atmospheric Sciences, 2022, 39(1): 189 - 200.
[48]
ZHAOL, ZOUD, HUG, et al. Changing climate and the permafrost environment on the Qinghai-Tibet plateau[J]. Permafrost and Periglacial Processes, 2020, 31(3): 396 - 405.
[49]
DINGK, JIANGP, NIJ, et al. Machine learning uncovers a multi-year climate memory in permafrost degradation on the Qinghai-Tibet Plateau: the critical roles of precipitation and lagged temperature[J]. Journal of Hydrology, 2025,663: 134272.