Assessing multi-hazard susceptibility to cryospheric hazards: Lesson learnt from an Alaskan example.

Area of Science:

• Geosciences
• Environmental Science
• Climate Change Research

Background:

• Geomorphological process susceptibility is standard in low to mid-latitudes but underexplored in periglacial regions.
• Global warming is accelerating changes in periglacial environments, necessitating understanding of geomorphological dynamics.
• Permafrost degradation induces cryospheric hazards like retrogressive thaw slumps (RTSs) and active layer detachments (ALDs), impacting infrastructure and releasing greenhouse gases.

Purpose of the Study:

• To explore data-driven models for identifying locations prone to RTSs and ALDs in periglacial environments.
• To estimate the probability of RTS and ALD occurrences in the North Alaskan territory using a binomial Generalized Additive Model.
• To develop an automated, open-source Python tool for replicating the spatial predictive experiment.

Main Methods:

• Utilized a binomial Generalized Additive Modeling structure to classify landscape susceptibility.
• Tested model accuracy using goodness-of-fit (AUC), random cross-validation, and spatial cross-validation routines.
• Developed an automated Python-based analytical protocol for data pre-processing and spatial prediction.

Main Results:

• The binary classifiers demonstrated high accuracy in recognizing locations prone to RTSs (AUC=0.83) and ALDs (AUC=0.86).
• Cross-validation results confirmed the models' robustness, with mean AUC values for random cross-validation at 0.82 (RTS) and 0.86 (ALD), and spatial cross-validation at 0.74 (RTS) and 0.80 (ALD).
• An open-source Python tool was created to automate the entire analytical protocol, enabling replication and application.

Conclusions:

• Data-driven models, specifically Generalized Additive Models, are effective for predicting cryospheric hazards in periglacial regions.
• The developed analytical protocol and open-source tool facilitate automated spatial prediction of RTSs and ALDs.
• Understanding these geomorphological dynamics is crucial for decision-making in rapidly changing Arctic environments and for anticipating future changes at lower latitudes.