Discover real Sentinel-2 scenes, compare pre- and post-fire spectral measurements, and map an observed burn index around a location. Built for exploratory research—not a prescription for ecological intervention.
PUBLIC SATELLITE DATA · EXPERIMENTAL WORKFLOWCheck readiness before starting an analysis. External data endpoints may be unavailable due to connectivity, CORS or service limits.
A short, reproducible route through public satellite data. Demonstration coordinates are not a verified burn perimeter.
OpenStreetMap basemap · the green square is the approximate analysis extent; click anywhere to reposition.
Compare real RGB rasters, inspect observed dNBR on the same sampling grid, or overlay spectral change on post-fire imagery.
Select real pre- and post-fire scenes, then load RGB imagery.
RGB images use a shared display stretch. dNBR is derived separately from NIR and SWIR2 with two-date SCL screening. These experimental layers show sampled spectral change—not a verified fire perimeter, calibrated severity class, or ecological recovery forecast.
dNBR = NBR before − NBR after, where NBR = (NIR − SWIR2) / (NIR + SWIR2). Pixel sample uses Sentinel-2 L2A Cloud-Optimized GeoTIFF assets. Results can be affected by cloud, smoke, shadows, misregistration, vegetation phenology and processing offsets.
Only locally screened pixels are used. Satellite scene-level cloud percentage is not site-level cloud coverage.
Ranges are descriptive bins, not validated burn severity thresholds. Fractions are calculated only from usable pixels in this sample.
Measure the mean NBR of selected summer satellite scenes at the same coordinates over successive years. NBR can reflect vegetation changes, but is not a direct measurement of ecological recovery or a forecast.
Run the post-fire analysis above to unlock a multi-year comparison.
Each year uses the lowest overall cloud-cover candidate found between 12 August and 30 September. Comparisons may still be biased by phenology, differing scene footprints, radiometric offsets and cloud artefacts. Yearly acquisition date, valid pixel coverage and scene IDs are shown for traceability. A failed year is shown as missing, never interpolated.
Scene search uses Element 84 Earth Search STAC v1, collection sentinel-2-l2a. The workbench accesses the scenes' NIR (B8, 10 m), SWIR2 (B12, 20 m), and scene classification (SCL, 20 m) Cloud-Optimized GeoTIFFs, resampling to a common 96×96 rectangular grid. Values are digital-number-scale reflectance proxies; normalized indices are calculated from each pair, with checks for zero, missing and invalid pixels. Sentinel-2 L2A processing baseline and offsets may change reflectance interpretation, so cross-date harmonization should be confirmed before publishing scientific results. Both-date SCL pixels classified as cloud, shadow, snow or no-data are excluded; a full two-date atmospheric and quality-screening pipeline is NOT implemented. Global scene cloud percentages do not ensure this individual site is cloud-free. Mean and median summarize this experimental sample only. For any operational decision, use validated cloud/shadow masking on both dates, radiometric harmonization, site-ground-truth calibration, spatially appropriate dNBR thresholds, local ecological expertise and uncertainty estimates. Maps are not burn perimeters or recovery forecasts.
Data: © Copernicus Sentinel data, distributed via AWS Open Data / Element 84 Earth Search. Basemap © OpenStreetMap contributors. User-uploaded GeoJSON boundaries are clipped to the current analysis square and sampled at pixel centers. Their geometry and stated provenance are included in exported metadata but are not independently authenticated. No claim is made that the demonstration coordinates were burned.