Towards a Deep Learning-based Spatio-temporal Fusion Approach for Accurately Improving Snow Cover Mapping: A Case Study in the Moroccan Atlas Mountains with Performance Evaluation
Bibliographic record
Abstract
Remote sensing technologies provide continuous and detailed observations of various land surface parameters, including snow cover, vegetation, land surface temperature, soil moisture, and evapotranspiration, offering invaluable information at various scales and contexts. One of the major uses is the precise mapping and monitoring of seasonal snow cover dynamics, which are essential for water management and global water balance modeling. Since an intelligent ecosystem based on accurate snow cover estimation requires a collection of high-resolution satellite images, both temporally and spatially, to capture snow dynamics, particularly in semi-arid areas where snowfall is extremely variable. These requirements can be difficult to achieve based on a single sensor, mainly due to the trade-offs between the temporal, spectral, and spatial resolutions of the available satellites. In addition, atmospheric conditions and cloud contamination can increase the number of missing satellite observations. However, there is a promising solution to these limitations. Exploiting the complementary capabilities of the new-generation multispectral sensors aboard Landsat-8 (L8) and Sentinel-2 (S2), with spatial resolutions ranging from 10 to 30 meters, offers an unprecedented opportunity to significantly advance the accuracy of snow cover mapping. Hence, this study aims to investigate the effectiveness of the combined use of optical sensors through deep learning-based spatiotemporal image fusion to capture snow dynamics and produce detailed and dense Normalized Snow Difference Index (NDSI) time series in a semi-arid context. Three distinct deep learning models, namely Very Deep Super Resolution (VDSR), Super Resolution Unet (SR-Unet), and Residual Convolutional Neural Network (RCNN), were evaluated and compared to fuse L8 and S2 data. The findings indicate that all three approaches can provide accurate estimates for a coarse-resolution image at a given fusion date, although there are notable disparities in prediction quality between the different approaches. Specifically, R-squared values were measured at 0.94, 0.92, and 0.96 for RCNN, SR-Unet, and VDSR, respectively, with corresponding root mean square error (RMSE) values of 0.09, 0.11, and 0.08. Our results suggest that the VDSR model is particularly effective in producing high-resolution merged snow time series and can compensate for the absence of ground snow cover data.
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 | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| 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 | 0.000 |
| 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".