Mapping retrogressive thaw slumps using deep neural networks

Yang, Yili; Rogers, Brendan M.; Fiske, Greg; Watts, Jennifer D.; Potter, Stefano; Windholz, Tiffany; Mullen, Andrew; Nitze, Ingmar; Natali, Susan M.

doi:10.1016/j.rse.2023.113495

Cited by 17 publications

(8 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Advances in image processing and classification techniques based on deep learning create new opportunities for automated detection and monitoring of large RTSs using high‐resolution remote sensing imagery. Several recent studies have applied deep learning to map RTSs in permafrost zones 50,74,119,128–130 . Among deep learning methods, Nitze et al used four convolutional neural network models, U‐Net, U‐Net++, DeepLabV3, and Mask‐RCNN, to map RTS in Arctic regions and the Tibetan Plateau with high accuracy 50,119,128–131 .…”

Section: Evaluating the Susceptibility And Stability Of Rtssmentioning

confidence: 99%

“…Several recent studies have applied deep learning to map RTSs in permafrost zones 50,74,119,128–130 . Among deep learning methods, Nitze et al used four convolutional neural network models, U‐Net, U‐Net++, DeepLabV3, and Mask‐RCNN, to map RTS in Arctic regions and the Tibetan Plateau with high accuracy 50,119,128–131 . These studies highlight the importance of training data and the limitations of interregional model transferability.…”

Section: Evaluating the Susceptibility And Stability Of Rtssmentioning

confidence: 99%

See 1 more Smart Citation

Advances in retrogressive thaw slump research in permafrost regions

Li,

Liu,

Chen

et al. 2024

Permafrost & Periglacial

View full text Add to dashboard Cite

A retrogressive thaw slump (RTS) is a slope failure formed by slope thaw settlement and retrogressive slump following the thawing of ice‐rich permafrost or the melting of massive ice. Here, we review recent literature on RTSs, one of the main geomorphological landscapes developed in the process of permafrost degradation. The main topics are as follows: development and temporal evolution, mechanisms and processes, influencing factors, evaluation susceptibility and calculation, and assessment of engineering and environmental impacts. There has been a rapid increase in the number and distribution area of RTSs over permafrost in recent years. Climate warming events, extreme rainfall, forest fires, bank and coast erosion, and anthropogenic activity are the primary factors leading to RTSs in permafrost regions, disrupting the initial hydrothermal equilibrium of permafrost slopes. This causes a rise in ground temperature and the thaw of ice‐rich permafrost. Meltwater seeps down and collects on the ice surface, weakening freeze–thaw interface shear resistance and resulting in soil collapse. The development of RTSs may last several decades or longer. RTSs destabilize infrastructure, destroy vegetation, boost soil erosion and land desertification, alter the environment of nearby waters, and increase emissions of some major greenhouse gases. Numerous methods have been developed and adopted to explore RTSs, including geographic information systems (GIS) and equilibrium, numerical, and reliability analysis methods. However, research on formation mechanisms and processes, quantitative prediction, engineering and environmental influences, and mitigative measures of RTSs under a warming climate are still inadequate. Existing research methods, such as numerical simulations, remote sensing, airborne ground‐based geophysical surveys, investigations and mapping, and hydrothermal and deformation field monitoring, should be systematically integrated. Additionally, equipment for laboratory testing and numerical models for simulating RTSs may need to be timely introduced and better developed.

show abstract

Section: Evaluating the Susceptibility And Stability Of Rtssmentioning

confidence: 99%

Section: Evaluating the Susceptibility And Stability Of Rtssmentioning

confidence: 99%

Advances in retrogressive thaw slump research in permafrost regions

Li,

Liu,

Chen

et al. 2024

Permafrost & Periglacial

View full text Add to dashboard Cite

show abstract

“…However, new developments in automated mapping may constitute the solution. Very recently, deep-learning routines have been develop to map RTS (see, Nitze et al, 2021;Yang et al, 2023), although most of the applications have been placed in Tibet (Huang et al, 2020(Huang et al, , 2021 and only a few are available in high-arctic regions (Witharana et al, 2022). In the case of ALD, their occurrence has been mapped through change-detection (Rudy et al, 2013).…”

Section: Opposing Argumentsmentioning

confidence: 99%

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

Elia¹,

Castellaro²,

Dahal³

et al. 2023

Preprint

View full text Add to dashboard Cite

Classifying a given landscape on the basis of its susceptibility to surface processes is a standard procedure in low to mid-latitudes. Conversely, these procedures have hardly been explored in periglacial regions, primarily because of the limited presence of human settlements and, therefore, the little need for risk assessment. However, global warming is radically changing this situation and will change it even more in the future. For this reason, understanding the spatial and spatiotemporal dynamics of geomorphological processes in peri-arctic environments can be crucial to make informed decisions in such unstable environments and shed light on what changes may follow at lower latitudes. For this reason, here we explored the use of data-driven models capable of recognizing locations prone to develop retrogressive thaw slumps (RTSs) and/or active layer detachments (ALDs). These are cryospheric hazards induced by permafrost degradation, and their development can negatively affect human settlements or infrastructure, change the sediment budget dynamics and release greenhouse gases. Specifically, we test a binomial Generalized Additive Modeling structure to estimate the probability of RST and ALD occurrences in the North sector of the Alaskan territory. The results we obtain show that our binary classifiers can accurately recognize locations prone to RTS and ALD, in a number of goodness-of-fit (AUC_RTS = 0.83; AUC_ALD = 0.86), random cross-validation (mean AUC_RTS = 0.82; mean AUC_ALD = 0.86), and spatial cross-validation (mean AUC_RTS = 0.74; mean AUC_ALD = 0.80) routines.Overall, our analytical protocol has been implemented to build an open-source tool scripted in Python as part of an interactive Jupyter notebook where all the operational steps are automatized for anyone to replicate the same experiment. Our protocol allows one to access cloud-stored information, pre-process it, and download it locally to be integrated for spatial predictive purposes. Data and codes can be accessed at this GitHub repository: https://github.com/zincoblenda/CryoS

show abstract

“…The model achieves predictive performance as good as Mask R-CNN but with a much faster inference speed. Yang et al developed a semantic segmentation framework based on U-Net to segment thaw slumps, leveraging geospatial data from multiple sources [6]. All these studies have fostered the in-depth integration of AI in permafrost mapping.…”

Section: Introductionmentioning

confidence: 99%

Segment Anything Model Can Not Segment Anything: Assessing AI Foundation Model’s Generalizability in Permafrost Mapping

Li,

Hsu,

Wang

et al. 2024

Remote Sensing

Self Cite

View full text Add to dashboard Cite

This paper assesses trending AI foundation models, especially emerging computer vision foundation models and their performance in natural landscape feature segmentation. While the term foundation model has quickly garnered interest from the geospatial domain, its definition remains vague. Hence, this paper will first introduce AI foundation models and their defining characteristics. Built upon the tremendous success achieved by Large Language Models (LLMs) as the foundation models for language tasks, this paper discusses the challenges of building foundation models for geospatial artificial intelligence (GeoAI) vision tasks. To evaluate the performance of large AI vision models, especially Meta’s Segment Anything Model (SAM), we implemented different instance segmentation pipelines that minimize the changes to SAM to leverage its power as a foundation model. A series of prompt strategies were developed to test SAM’s performance regarding its theoretical upper bound of predictive accuracy, zero-shot performance, and domain adaptability through fine-tuning. The analysis used two permafrost feature datasets, ice-wedge polygons and retrogressive thaw slumps because (1) these landform features are more challenging to segment than man-made features due to their complicated formation mechanisms, diverse forms, and vague boundaries; (2) their presence and changes are important indicators for Arctic warming and climate change. The results show that although promising, SAM still has room for improvement to support AI-augmented terrain mapping. The spatial and domain generalizability of this finding is further validated using a more general dataset EuroCrops for agricultural field mapping. Finally, we discuss future research directions that strengthen SAM’s applicability in challenging geospatial domains.

show abstract

Mapping retrogressive thaw slumps using deep neural networks

Cited by 17 publications

References 31 publications

Advances in retrogressive thaw slump research in permafrost regions

Advances in retrogressive thaw slump research in permafrost regions

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

Segment Anything Model Can Not Segment Anything: Assessing AI Foundation Model’s Generalizability in Permafrost Mapping

Contact Info

Product

Resources

About