Burnt Area Monitoring Using Graph Convolutional Networks Based On Multi-Sensor Satellite Data
Bibliographic record
Abstract
Recent catastrophic wildfire seasons, e.g. in Greece 2023, Canada 2023, and Chile 2023/2024, underscore the critical need for rapid and accurate wildfire data to facilitate emergency response, assess environmental damage, and keep the public informed. Although satellite-based thermal anomaly data is accessible in near real-time (NRT), accurately mapping the areas affected by fires from NRT imagery remains a significant challenge. The proposed approach combines a superpixel segmentation algorithm with both rule-based and deep learning classification techniques to reliably identify burnt areas (BA) in NRT. This method is compatible with a range of optical sensors, from medium to high resolution, and integrates data from diverse sources to continuously refine the detection of burnt areas as active fires unfold. The region of Central Chile, enduring tremendous wildfire events in early 2024, was used as a testing region. An NRT product (DLRBAv2NRT) based on Sentinel-3 OLCI was generated, together with a refined non-time critical product (DLRBAv2NTC). Both products are tested against established global BA products (Copernicus CGLBA31nrt and NASA MCD64A1v061). The DLRBAv2NRT achieved the highest accuracies, outperforming the DLRBAv2NTC product by 5%, the CGLBA31nrt product by 9% and the MCD64A1v061 product by 10% IoU. The DLRBAv2NRT showed the highest sensitivity detecting BA (Recall: 0.78), while MCD64A1v061 produced high number of false negatives (Recall: 0.63). A third variant (DLRBAv2NTCfusion), incorporating results from multiple mid- and high resolution sensors is generated for the Valparaíso focus region. The results are inter-compared with local ground truth data, yielding an IoU of 0.75. The proposed mapping procedure demonstrates a fully-automated, flexible approach to derive burnt area delineations from satellite data in NRT with high accuracy. This allows for high-frequency monitoring of NRT burnt areas on a global scale.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.004 | 0.005 |
| Meta-epidemiology (broad) | 0.003 | 0.001 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.005 | 0.003 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.003 | 0.005 |
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; both teacher heads agree on what is shown here.
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".