Permafrost Distribution, Degradation, And Potential Mass Movement Cascades In The Western Himalaya Using Machine Learning And Numerical Models
--
https://doi.org/10.1038/s44304-026-00217-4 <-- shared paper
--
https://doi.org/10.1038/s41598-025-22051-w <-- shared (earlier) paper
--
https://doi.org/10.1080/2150704X.2025.2488532 <-- shared (earlier) paper
--
H/T @Abhinav Alangadan
“Can we develop a first-order understanding of permafrost degradation and glacial lakes exposed to degradation-induced mass movements in the Himalaya?
[The authors] tried to address this question. The study [first link above] integrates machine learning, statistical modeling, and numerical modeling to investigate high-resolution permafrost distribution, potential degradation, and associated mass-movement hazards in the Kinnaur district of Himachal Pradesh, India.
Using rock glaciers as proxies, [they] generated a high-resolution permafrost distribution using machine learning, while potential degradation zones were delineated using the 0°C isotherm as a first-order indicator. [They] further identified glacial lakes located near potentially degrading permafrost zones and reconstructed their bathymetry. A detailed scenario-based GLOF process-chain simulation was then carried out for Kashang Lake using r.avaflow and HEC-RAS.
[Their] results indicate that seven glacial lakes in #Kinnaur are located close to potentially degrading permafrost zones. The simulations further show that a potential GLOF from Kashang Lake could inundate critical downstream infrastructure, including nearly 11 km of National Highway 5…”
--
“Permafrost degradation poses a serious threat to communities located in the Himalayan region. It can cause land subsidence and mass movements, damaging or destroying infrastructure downstream. In the current study, a high-resolution permafrost probability index based on machine learning was generated for the Kinnaur district in Himachal Pradesh, located in the western Himalayan region. Machine learning methods, including random forest, support vector machine, logistic regression, and artificial neural network, were trained on spatially distributed mean atmospheric air temperature, potential incoming solar radiation, elevation, slope, and aspect as predictive variables and intact rock glaciers as the target variable. Degrading permafrost zones and the overlaying glacial lakes were identified using the current zero-degree isotherm. The random forest technique provided the best results with an accuracy score of 89.43%. Scenario-based process chain models of glacial lake outburst floods (GLOFs) were modeled using r.avaflow and HEC-RAS for the Kashang glacial lake, which was identified as very highly hazardous. The lake volume was estimated to be 8.6 × 106 m³ by extrapolating the contours from overdeepening of the main glacier. Three sources of avalanches were identified based on permafrost degradation and slopes greater than 30°. The modeling results revealed that the potential GLOF can cause a peak discharge of 16,167 m³/s, and floodwater can reach Kashang, where a hydropower is located, within 16 min in the high-magnitude scenario. The findings can provide important insights into GLOF hazard mitigation in the valley and serve as preliminary data for various stakeholders working to mitigate glacier-related hazards...”
#permafrost #distribution #GIS #spatial #mapping #Himalayas #India #Kinnaur #HimachalPradesh #KashangLake #massmovement #engineeringgeology #machinelearning #AI #model #modeling #numericalmodel #glaciallakes #glaciet #glacial #glaciallakeoutburstflood #GLOF #cryosphere #geostatistics #rockglaciers #GeoAI #bathymetry #processchainsimulation #HECRAS #avaflow #risk #hazard #mitigation #riskassessment #infrastructure #HEP #publicsafety #downstream #avalanche