Glacial lakes in permafrost terrain and downstream hazards

A permafrost probability index (PPI) based on rock glacier inventory and machine learning models, including random forest, support vector machine, artificial neural network, and logistic regression, was generated for Kinnaur district, Himachal Pradesh, India. Intact rock glaciers were considered the dependent variable, and elevation, slope, aspect, and potential incoming solar radiation were used as independent variables to generate a spatially distributed, high-resolution permafrost probability index. Daily weather station data and daily multitemporal MODIS satellite data were used to train a linear regression model to predict the annual 0℃ isotherm in the region for the period of 2023-24, aiming to understand potential degradation by overlaying the isotherm on permafrost distribution. The random forest technique produced the best results with an overall accuracy of 89.43%. Seven glacial lakes were identified as located in potentially permafrost-degraded slopes, and the Kashang glacial lake was selected for detailed downstream glacial lake outburst flood process chain modeling based on its size, moraine-dammed proglacial setting, and potential downstream impact. The volume of the lake 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 degrees. Subsequently, three scenario-based process chains for glacial lake outburst floods were modeled. We simulate avalanche initialization, displacement wave generation, overtopping, moraine erosion, and downstream flooding. The modelling results revealed that the potential GLOF can cause a peak discharge of 16,167 ms⁻¹, and floodwater can reach the Kashang, where a hydropower is located, within 16 minutes  in the high-magnitude scenario. The findings can give important insights into GLOF hazard mitigation in the valley and can aid as preliminary data for various stakeholders working towards mitigating glacier-related hazards.

Keywords: Permafrost, GLOF, machine learning, r.avaflow, Himalaya