Permafrost thaw poses significant risks to Arctic environments and livelihoods, impacting societies, cultures, health, ecosystems, and economies. However, existing studies lack comprehensive frameworks for understanding such impacts on coupled socio-ecological systems from a transdisciplinary perspective. To address this gap, we introduce a holistic, comparative, and transdisciplinary risk analysis based on multidirectional knowledge exchanges and thematic network analysis (Gartler, Scheer and Meyer et al. 2025). Local stakeholders and scientists’ perceptions shape our understanding of risks that we holistically define as dynamic socio-natural phenomena involving physical processes related to permafrost thaw, ensuing key hazards, and societal consequences. Consolidating findings from four regions (Longyearbyen, Svalbard, Norway; Avannaata Municipality, Greenland; Beaufort Sea region and Mackenzie River Delta, Canada; Bulunskiy District, Sakha Republic, Russia), we identify five key hazards of permafrost thaw: infrastructure failure, disruption of mobility and supplies, decreased water quality, challenges for food security, and exposure to diseases and contaminants. The novelty of this study resides in the comparative approach spanning different environmental and societal contexts, and transdisciplinary synthesis considering various risk perceptions. We present the main local risks from permafrost thaw faced by diverse Arctic communities, seen through the lenses of multiple scientific disciplines and local experts. Developing such an understanding is vital for informed policy-making and adaptation efforts.

References

Susanna Gartler, Johanna Scheer, Alexandra Meyer, with Khaled Abass, Annett Bartsch, Natalia Doloisio, Jade Falardeau, Gustaf Hugelius, Anna Irrgang, Jón Haukur Ingimundarson, Leneisja Jungsberg, Hugues Lantuit, Joan Nymand Larsen, Rachele Lodi, Victoria Sophie Martin, Louise Mercer, David Nielsen, Paul Overduin, Olga Povoroznyuk, Arja Rautio, Peter Schweitzer, Niek Jesse Speetjen, Soňa Tomaškovičová, Ulla Timlin, Jean-Paul Vanderlinden, Jorien Vonk, Levi Westerveld, and Thomas Ingeman-Nielsen. (2024) A transdisciplinary, comparative analysis reveals key risks from Arctic permafrost thaw. Communications Earth & Environment, 6(1), 21–20. https://doi.org/10.1038/s43247-024-01883-w

Permafrost degradation is accelerating in ice-rich Arctic lowlands, reshaping thermokarst landscapes and mobilising formerly frozen organic carbon. As permafrost thaws, microbial decomposition speeds up, producing greenhouse gases (GHG) and altering carbon cycling. GHG emissions are closely linked to the microbial degradability of organic matter (OM), which depends on both physical accessibility and substrate quality. This study investigates OM characteristics in five landscape units on the Baldwin Peninsula, Alaska, representing successive stages of thermokarst-driven transformation: undisturbed Yedoma upland, thermokarst lake, semi-drained lake basin (SDL), drained lake basin (DLB), and a near-shore marine deposit as an end-member. Focusing on the uppermost meter of sediment, we applied a multi-proxy approach including n-alkane biomarker analysis, elemental composition (TOC, C/N), δ¹³C stable isotopes, radiocarbon dating, and sedimentological characterisation. We addressed two main questions: (1) How does OM quality vary across these landforms? (2) What is the potential for future OM decomposition and mobilisation? Our findings show that OM in thermokarst-influenced units (lake, SDL, DLB) is more strongly degraded than in upland deposits. The DLB represents a refrozen talik, while the SDL retained an unfrozen talik at the time of coring; OM degradation was highest in the SDL. These results highlight the role of talik dynamics in shaping OM decomposition potential and underscore the vulnerability of carbon stored in yedoma uplands under future thermokarst activity. Our study contributes to understanding permafrost carbon dynamics and the climate sensitivity of Arctic lowland systems.

Debris flows occur as one of the most significant natural hazards in high mountain regions worldwide, and play a substantial role within the sediment budget. Despite their importance, it remains unclear how their frequency may evolve under changing climatic conditions. Long-term and comprehensive records are essential to detect significant changes in their activity over time. However, such multi-decadal datasets are scarce. As debris flows are predominantly triggered by high-intensity rainfall, the identification of (potential) trigger events from precipitation data presents an alternative, at least for transport-limited debris flow systems with sufficient sediment availability. Due to the lack of long-term precipitation records with sufficient temporal resolution, we employed dynamical downscaling of a Regional Climate Model (RCM) based on the Advanced Weather Research and Forecasting model (WRF). This approach produced a high-resolution dataset with 2x2 km spatial and 15-minute temporal resolution, enabling the analysis of intense rainfall events back to 1850.

We applied this approach to the Horlachtal catchment in Tyrol, Austria, where we have elaborated a debris flow inventory from remote sensing and lichenometric dating that spans the past 170 years. This integrated approach allows us to assess changes in debris flow frequency and their climatic drivers since the end of the Little Ice Age. As the RCM dataset covers much of the Central Alps, the analyses can be extended to other regions, offering broader insights into the development of debris flow frequencies in a changing climate.

The geological surveys of the federal states play a vital role in mitigating the impacts of landslides and safeguarding communities in Germany.
However, certain decisions require a cross-border approach to geohazard assessment, like the landslide susceptibility assessment (1:250 000 scale) that the Federal Company for Radioactive Waste Disposal commissioned. They requested it for areas, in which favorable geological conditions for the safe final disposal of high-level radioactive waste can be expected. To facilitate this we will homogenize state inventories and data from publications supported by state geological surveys (SGDs).
We will distinguish rockfall, translational, and rotational landslides in rule-based simplification steps. We will select specific methodology for each landslide type based on its causes, available information and best practices developed on the state level. For rockfall, we will estimate runout zones with the shadow angle method and identify source areas based on a critical slope angle, both derived from a digital elevation model (DEM) and calibrated against the landslide inventory. Translational landslides will be assessed using the infinite slope method, incorporating a nationwide soil map and a DEM. Geotechnical paramaters will be inferred via a pedotransfer function from the soil data, acknowledging associated uncertainties.

Rotational landslides will be evaluated with the Analytical Hierarchy Process to quantify regional expert knowledge from multiple parameters, because they are the least generalizable type. The available data will be used to define trivial areas. We aim to collaborate with SGDs and local experts to ensure technically sound and regionally adapted methodologies are used.

This study presented a hybrid machine learning framework for high-resolution landslide susceptibility mapping in the Svaneti region of the Georgian Greater Caucasus, a steep alpine environment prone to frequent mass movements. The objective was to enhance predictive accuracy and spatial generalization of susceptibility maps by integrating geomorphological knowledge with data-driven modeling.

We combined geostatistical, geomorphometric, and spectral terrain features, including slope derivatives and multi-scale gray-level co-occurrence matrix texture extracted from diverse raster sources. These were systematically preprocessed and complemented by spatial clustering, which segmented the landscape into zones with internally coherent terrain characteristics using tree-based projections, dimensionality reduction, and spatially-aware k-means.

Supervised classification leveraged a binary landslide inventory aligned to the feature grid. Class imbalance was addressed through stratified undersampling. Base classifiers, including extreme gradient boosting, random forests, k-nearest neighbors, and a multi-layer perceptron neural network, were tuned using Bayesian optimization and combined into a stacked ensemble with extreme gradient boosting as the meta-learner. The final model yielded high cross-validated performance (AUC = 0.98, balanced accuracy = 0.95).

The resulting susceptibility map showed good spatial agreement with known events and provided a detailed zonation of hazard potential. Despite a high recall (0.98) for landslide prediction, the low precision (0.13) and moderate Cohen’s Kappa (0.21) reflected challenges posed by extreme class imbalance and spatial noise in event inventories. Overall, the proposed framework demonstrated strong potential for integrating spatially-informed clustering with ensemble learning to enhance landslide susceptibility mapping in complex mountainous regions and supported scalable, region-specific hazard assessment.

The Gutenberg-Richter law, relating earthquake frequency to magnitude via a power-law, remains a cornerstone of seismology. It captures the scaling of brittle failure from microfracturing to plate-boundary rupture. However, earthquake catalogs are inherently discrete, and this discreteness is often neglected in modeling and hazard estimation.

Here, we construct idealized earthquake populations as finite sets of discrete events drawn from a bounded power-law distribution, defined by upper and lower magnitude limits and a fixed total number of events. This formulation introduces a well-defined event density and direct control over seismic moment release, enabling a more physically grounded and reproducible basis for synthetic catalog generation and hazard analysis.

We extend this with a three-segment, piecewise power-law model constrained by a Bayesian prior on the largest observed earthquake. The steep upper segment, covering only Mw > 9 events, is interpreted as representing the largest possible ruptures—likely limited to subduction megathrusts. The meaning of the intermediate and lower segments remains open, potentially reflecting geometric constraints, energy saturation, or deeper tectonic structure.

Even in the absence of full physical interpretation, the segmented model offers a tighter constraint on recurrence rates, particularly for large, rare events. This has direct implications for seismic hazard modeling and estimates of maximum credible earthquake size, making the framework both conceptually and practically valuable.