Detecting mass wasting of Retrogressive Thaw Slumps in spaceborne elevation models using deep learning
Bibliographic record
Abstract
Climate change has led to stronger warming in the Arctic, causing higher ground temperatures and extensive permafrost thaw. Retrogressive Thaw Slumps (RTSs) represent one of the most rapid and considerable geomorphological changes in permafrost regions, occurring when ice-rich permafrost is exposed and thaws. However, large-scale quantification of RTS-related mass wasting in Arctic permafrost landscapes is currently lacking, despite its importance to understand impacts on local environments and the global permafrost carbon cycle. Generating differential digital elevation models (dDEMs) from TanDEM-X single-pass Interferometric SAR (InSAR) observations enables us to quantify volume changes induced by rapid permafrost thaw. To extend this capability across the entire Arctic permafrost region, automation in data processing and RTS detection is essential. This study introduces a method that employs deep learning on InSAR-derived dDEMs to map RTSs and quantify volume changes from RTS activity. We chose eleven study sites with a total area of 71 400 km 2 to reflect the diverse character of Arctic environments for model training, testing, and inference. Our trained UNet++ model delivers a scalable solution for mapping RTSs and quantifying mass wasting towards a pan-Arctic scale, achieving segmentation accuracies of 0.58 (Intersection over Union) and classification accuracies of 0.75 (F1) on previously unseen test sites, with volume change estimates from model predictions being within ± 20% of the actual values. We found a total of almost 5000 RTSs active between 2010 and 2021 with volume change rates between 40.75 m 3 yr −1 km 2 for sites in the Siberian to 1164.11 m 3 yr −1 km 2 in the Canadian Arctic. • We developed a deep learning method to detect RTS activity on differential DEMs. • The method achieved a segmentation IoU of 0.58, mean IoU of 0.73, and F1 of 0.75. • The automatic segmentation method matches domain experts’ RTS labelling accuracy. • Model predictions of RTS area and volume change have an accuracy of ±20%. • We identified 4882 active Arctic RTSs between 2010 and 2021 across 71 400 km 2 .
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".