Debris covered glacier mapping using newly annotated multisource remote sensing data and geo-foundational model
Bibliographic record
Abstract
The automated mapping of debris covered glaciers remains challenging due to spectral similarity between supraglacial debris (on-glaciers) and periglacial debris (off-glaciers). Deep learning offers promising capabilities, yet the lack of high-quality publicly available datasets and limited exploration of optimal model architecture constrain progress in this domain. To address this, we introduce the Global Supraglacial Debris Cover Dataset (GSDD), consisting of 1,876 images (∼49,000.00 km 2 ) collected globally from diverse glacierized regions, including High Mountain Asia, Andes, Western Canada, Alaska, and Swiss Alps, to incorporate the heterogeneity of glacial features and environments. This multisource remote sensing dataset includes 10 spectral bands—Blue, Green, Red, Near-Infrared, Shortwave Infrared (SWIR1 & SWIR2), Normalized Difference Rock Index (NDRI), Slope, Elevation, and Velocity—providing critical information to distinguish glacier debris. To evaluate the efficacy of deep learning models for mapping glacier debris, we compare Prithvi Geo-Foundational Model (GFM) combined with multiple decoders, CNN-based models (UNet, Attention U-Net, and DeepLabv3+), a Vision Transformer-based model (TransNorm), and variant of the Prithvi GFM (i.e., UViT). Our results show Prithvi GFM with UperNet decoder outperformed all, achieving mIoU = 0.80 and F1-score = 0.91, surpassing DeepLabv3+ (0.71 mIoU), Attention U-Net (0.73), U-Net (0.72), TransNorm (0.71), and UViT (0.70). Our results demonstrate significant methodological advances in accurately mapping glacier termini, along with the identification of glacier snouts. Feature analysis identified the optimal band combination (B-G-NIR-SWIR-Slope-Elevation) for debris mapping. The GSDD dataset enables direct comparison, development, and evaluation of deep learning models, supporting advancement in fast and reliable automated glacier mapping.
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 | 0.001 |
| 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) | 0.002 | 0.001 |
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".