A comparative study of data input selection for deep learning-based automated sea ice mapping
Bibliographic record
Abstract
The precise monitoring of sea ice parameters, including sea ice concentration and stage of development, is imperative for tactical navigation. Recent studies have showcased the enhanced mapping accuracy achieved by incorporating multi-source auxiliary data, such as passive microwave data, with Synthetic Aperture Radar (SAR) images. However, there remains a lack of research assessing the impact of individual features on model performance. This paper addresses this knowledge gap through ablation studies and alternate comparisons of data inputs. Building on the success in the AutoIce Challenge, we leverage the AI4Arctic Sea Ice Challenge Dataset to train multitask sea ice mapping models employing a U-Net architecture. Results from cross-validation and testing sets with all season data reveal the significant enhancement in estimation accuracy for all parameters when utilizing most of the AMSR2 channels. Additionally, the incorporation of time and location information as ancillary channels further amplifies the classification accuracy of all major ice types. Furthermore, among the various available ERA5 weather parameters, the inclusion of wind speed data proves effective in mitigating misclassifications in ice regions, particularly under melting scenarios. The paper culminates with a feature importance ranking table encompassing all available features, providing valuable guidance for the selection of pertinent data inputs. This comprehensive comparative study not only contributes to advancing sea ice mapping methodologies but also offers valuable insights into the nuanced impact of individual features on model performance.
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 | 0.005 | 0.011 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| 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".