- severity_non_qa.ipynb
- Results: non_qa/
- Note: This version shows inflated "High" severity (~40%) due to border artifacts (NIR/SWIR2 = 0 outside AOI not masked → false low NBR values).
This was shared on LinkedIn as a "non-curated" example to highlight QA importance.
- severity_with_qa.ipynb
- Results: with_qa/
- Key fix: Valid pixel mask
(nir > 0) & (swir2 > 0)before NBR calculation → removes edge effects.
→ High severity significantly reduced, more realistic distribution aligned with Copernicus EMS reports.
- qa_impact_comparison.ipynb
Side-by-side stats, area tables (ha per class), and visual difference maps.
See LinkedIn post for context: [https://www.linkedin.com/posts/augusto-salonio-442621132_wildfire-burnseverity-dnbr-activity-7421506656993210368-TSqu?utm_source=share&utm_medium=member_desktop&rcm=ACoAACB70zIBpz1rXQlFlhXLOTC7w4-pCdCXzw0]