Mapping Wildfire Burn Scars with NBR: From Raw Imagery to Damage Assessment
Quick Answer: NBR uses the contrast between NIR (which healthy vegetation reflects strongly) and SWIR (which burned areas reflect strongly) to map fire damage. Pre-fire NBR minus post-fire NBR gives dNBR — a measure of burn severity. Values above 0.27 indicate moderate-high severity burns. Sentinel-2's 20m SWIR bands and 5-day revisit make it ideal for rapid fire damage assessment.
Why Fire Mapping Needs a Specialized Index
After a wildfire, the affected area is visually obvious in true-color imagery — charred landscapes appear dark. But visual inspection doesn't scale, and it can't distinguish between lightly scorched vegetation and completely destroyed canopy.
The Normalized Burn Ratio (NBR) solves this by quantifying fire damage on a continuous scale, enabling:
- Automated mapping of burn extent
- Classification of burn severity (low, moderate, high)
- Monitoring of post-fire vegetation recovery
- Comparison across different fire events
The Spectral Logic
NBR exploits a simple physical fact: fire changes how the surface interacts with NIR and SWIR light in opposite directions.
Healthy vegetation:
- High NIR reflectance (strong cellular scattering in leaf structure)
- Low SWIR reflectance (water absorption in healthy leaves)
Burned surface:
- Low NIR reflectance (destroyed leaf structure)
- High SWIR reflectance (exposed soil, charcoal, ash — all dry materials)
This opposite response in two bands creates maximum contrast:
NBR = (NIR - SWIR) / (NIR + SWIR)
For Sentinel-2: (B8 - B12) / (B8 + B12)
| Surface | NIR | SWIR | NBR |
|---|---|---|---|
| Healthy forest | High | Low | +0.6 to +0.9 |
| Sparse vegetation | Moderate | Moderate | +0.1 to +0.3 |
| Bare soil | Low | Moderate | -0.1 to +0.1 |
| Burned area | Low | High | -0.5 to -0.1 |
| Active fire/charcoal | Very low | High | -0.6 to -0.3 |
Calculating Burn Severity with dNBR
A single NBR image shows the current state. To quantify how much the fire changed the landscape, calculate the differenced NBR (dNBR):
dNBR = NBR_pre-fire - NBR_post-fire
Because healthy vegetation has high NBR and burned areas have low NBR, the difference is positive where fire occurred. The magnitude indicates severity:
| dNBR Range | Burn Severity |
|---|---|
| < 0.10 | Unburned / Very low |
| 0.10 – 0.27 | Low severity |
| 0.27 – 0.44 | Moderate-low severity |
| 0.44 – 0.66 | Moderate-high severity |
| > 0.66 | High severity |
These thresholds (from the USGS FIREMON program) are widely used but should be calibrated for your specific ecosystem and fire conditions.
Step-by-Step Workflow
1. Find Pre-Fire Imagery
Search for a cloud-free Sentinel-2 scene from shortly before the fire — ideally within 1-2 months, during the same season. This baseline should represent normal vegetation conditions.
2. Find Post-Fire Imagery
Search for the first cloud-free scene after the fire. For rapid damage assessment, aim for imagery within 1-2 weeks post-fire. For full extent mapping, wait until fire containment.
3. Calculate NBR for Both Dates
Apply NBR visualization to both scenes. The pre-fire scene should show high NBR (green vegetation), while the post-fire scene will show dramatically reduced NBR in burned areas.
4. Interpret the Results
- Sharp NBR boundaries indicate fire perimeters
- Gradients within the burn scar show variable severity
- Islands of high NBR within the scar are unburned patches (potential wildlife refugia)
- Very low post-fire NBR near water features may be false positives — check against true-color imagery
Common Pitfalls
Cloud and Shadow Confusion
Cloud shadows produce low NIR reflectance, mimicking burn signatures. Always verify your post-fire scene is cloud-free over the fire area. Even thin cirrus clouds can affect SWIR bands.
Seasonal Vegetation Changes
If your pre-fire and post-fire images span different seasons (e.g., summer baseline, winter post-fire), normal phenological changes will contaminate the dNBR. The brown winter landscape may produce false "burn" signals.
Agricultural Fields
Harvested cropland has low NBR values similar to burned areas. If your fire area is near agriculture, check the true-color image to distinguish harvested fields from actual burn scars.
Water Bodies
Water absorbs NIR strongly, producing negative NBR values that can look like severe burns. Mask water bodies before interpreting burn severity.
NBR vs Visual Assessment
Why not just use true-color imagery? Four reasons:
- Severity quantification — True-color can show "burned" but can't distinguish moderate from severe burns
- Automation — NBR allows threshold-based mapping over large areas
- Subtle burns — Low-severity surface fires may barely change the color appearance but significantly alter the NIR/SWIR balance
- Recovery tracking — As vegetation regrows, NBR quantifies the recovery rate objectively. Track recovery with NDVI time-series alongside NBR
Combining NBR with SAR
For fires in cloud-prone regions (Southeast Asia, Pacific Northwest), SAR imagery can complement NBR analysis. Burned areas typically show changed SAR backscatter due to:
- Destroyed canopy structure (reduced volume scattering)
- Exposed rough ground surface (increased surface scattering)
- Changed moisture conditions
SAR can confirm fire extent during cloudy conditions when optical NBR analysis isn't possible.
Try It in Off-Nadir Delta
Off-Nadir Delta includes NBR as a built-in visualization for Sentinel-2 imagery:
- Search for pre-fire and post-fire Sentinel-2 scenes
- Add both to the map and select NBR visualization
- Toggle between dates to see the burn scar emerge
- Use change detection to systematically compare the two dates
For recent wildfire events, the 5-day Sentinel-2 revisit cycle typically provides usable post-fire imagery within a week.
