Mapping Burn Severity After Wildfires: NBR and Satellite-Based Assessment

Burn severity — how much a fire has altered vegetation and soil — is mapped from satellites with the Normalized Burn Ratio (NBR), a simple index built from near-infrared and shortwave-infrared reflectance. Comparing NBR before and after a fire (differenced NBR, or dNBR) classifies the landscape from unburned to high severity, and Landsat and Sentinel-2 supply the free 20–30m imagery to do it. It is the difference between a strategic map produced in weeks and a ground survey that would take years.

After the 2020 wildfire season in California — the largest in the state's recorded history — resource managers faced an urgent question: where exactly did the fires burn, and how severely? The answer determines everything from erosion risk assessment to reforestation planning to insurance claims processing. Ground survey of the 1.7 million hectares that burned would have taken years. Satellite-based burn severity mapping provided the answer within weeks.

What is burn severity?

Burn severity describes the degree of ecological change a fire causes, from a surface fire that only singes ground litter to a stand-replacing burn that kills the canopy and exposes mineral soil. It is distinct from fire intensity (the energy released while the fire burns): severity is the lasting effect on the ecosystem, which is what post-fire management responds to. Field crews and satellites both sort it into classes:

Unburned/Very Low: Fire didn't reach this area, or surface fire passed through without significantly affecting vegetation or soil.

Low Severity: Surface fire consumed ground litter and some understory vegetation. Overstory trees survived. Soil is largely intact.

Moderate Severity: Understory consumed, some overstory mortality. Mix of surviving and killed trees. Some soil organic matter consumed.

High Severity: Complete canopy mortality. All surface organic matter consumed. Mineral soil exposed. Root systems killed.

The distinction matters because high-severity burns fundamentally change the ecosystem for decades — destroying seed banks, eliminating soil biota, increasing erosion risk dramatically, and often converting forest to shrubland.

How does the Normalized Burn Ratio (NBR) work?

NBR is the fire-mapping counterpart to NDVI: it contrasts the near-infrared and shortwave-infrared bands, both of which respond strongly to burning but in opposite directions. As the USGS, which documents the index, puts it, "NBR is used to identify burned areas and provide a measure of burn severity." The formula exploits two spectral responses that change dramatically with fire:

NIR (Band 8A in Sentinel-2, ~865 nm): Healthy vegetation reflects strongly in NIR. Burned vegetation absorbs NIR. Dead trees and charred surfaces have very low NIR reflectance.

SWIR (Band 12 in Sentinel-2, ~2190 nm): Healthy vegetation absorbs SWIR moderately. Burned areas, exposed soil, and charred surfaces reflect SWIR more strongly.

NBR = (NIR − SWIR) / (NIR + SWIR)

• Healthy forest: High NBR (0.4-0.8) — high NIR, moderate SWIR
• Burned area: Low or negative NBR (−0.5 to 0.1) — low NIR, high SWIR

Differenced NBR (dNBR)

The change in NBR between pre-fire and post-fire images:

dNBR = NBR_pre − NBR_post

dNBR Range	Severity Class
< 0.1	Unburned
0.1 - 0.27	Low severity
0.27 - 0.44	Moderate-low
0.44 - 0.66	Moderate-high
> 0.66	High severity

These thresholds are approximate — they vary by vegetation type, climate, and pre-fire condition. Local calibration against field plots improves accuracy.

Relativized dNBR (RdNBR)

Standard dNBR is biased by pre-fire vegetation condition: a high-biomass forest shows larger dNBR for the same severity than a sparse woodland simply because it had more to burn. RdNBR normalizes by pre-fire condition:

RdNBR = dNBR / √(|NBR_pre|)

This produces more comparable severity estimates across different vegetation types and pre-fire conditions.

When should burn severity be mapped?

Timing changes what a burn severity map actually measures, so the right window depends on the decision it will inform. Imagery captured within days to weeks of containment records the direct fire effect — charring, canopy consumption, exposed soil — and drives emergency response. Imagery from the first growing season afterward instead records the ecosystem's response, separating scorched-but-surviving trees from those that were killed, which is what long-term planning needs.

Immediate Assessment (Days to Weeks)

Using imagery acquired as soon as clouds and smoke clear after the fire:

• Captures the direct fire effect (charring, canopy consumption, soil exposure)
• May include actively smoldering areas and smoke residuals
• Best for: emergency response, erosion risk assessment, BAER (Burned Area Emergency Response)

Extended Assessment (1 Growing Season Later)

Using imagery from the first growing season after the fire:

• Captures vegetation response (what's regrowing vs. what's dead)
• More accurately reflects long-term ecological impact
• A tree that was scorched but survived will show green regrowth; one that was killed won't
• Best for: reforestation planning, ecological impact assessment, long-term monitoring

The USGS MTBS program produces both initial and extended assessment products for every large US fire.

Sentinel-2 vs. Landsat for Burn Severity

Both sensors are widely used:

Feature	Sentinel-2	Landsat 8/9
Resolution	20m (SWIR bands)	30m
Revisit	5 days	16 days
Archive	2015-present	1984-present
SWIR bands	B11 (1610nm), B12 (2190nm)	B6 (1610nm), B7 (2190nm)

Sentinel-2 advantage: Higher resolution and more frequent imaging increase the chance of a cloud-free post-fire observation. See the Sentinel-2 complete guide for band details.

Landsat advantage: The 40-year archive enables historical burn severity assessment back to 1984. The MTBS program uses this for comprehensive fire history analysis, as covered in the Landsat program guide.

Applications

Burned Area Emergency Response (BAER)

Within days of fire containment, burn severity maps drive emergency stabilization:

• High-severity areas → highest erosion and debris flow risk
• Prioritize erosion control treatments (mulching, log erosion barriers, channel protection)
• Identify infrastructure at risk from post-fire debris flows
• Determine watershed impacts on water supply

Reforestation Planning

Burn severity maps guide planting decisions:

• High severity: Active reforestation often needed (no surviving seed source)
• Moderate severity: Natural regeneration may be sufficient if seed trees survived
• Low severity: Monitoring only; forest will recover without intervention

Fire Effects Monitoring

Long-term monitoring of burn severity patterns reveals:

• Are fires getting more severe over time? (Yes, in many western US forests)
• What proportion of each fire burns at high severity?
• How does severity relate to pre-fire fuel conditions, weather, and topography?

Insurance and Damage Assessment

Satellite burn severity maps provide objective documentation of fire impact for:

• Property damage claims
• Agricultural loss assessment
• Timber loss quantification
• Infrastructure damage documentation

Major Wildfire Burn Severity Mapping Reference

Contextualizing satellite burn severity assessments with documented events helps calibrate expectations:

Fire Event	Location	Year	Total Area	High Severity %	dNBR Accuracy vs. CBI	Mapping Latency
Camp Fire	California, USA	2018	62,000 ha	~50% of forested area	R² = 0.71	< 2 weeks post-containment
Black Summer	SE Australia	2019–20	5.5 million ha	~25–35% (variable)	R² = 0.65–0.75	MTBS-equivalent product
Dixie Fire	California, USA	2021	390,000 ha	~38% high severity	R² = 0.68	MTBS 12–18 months post-fire
Amazon fire season	Brazil	2019	~900,000 ha burned	15–25% (savanna-like)	Moderate	INPE PRODES product
Boreal fires (NW Territories)	Canada	2023	4+ million ha	~30–45% (black spruce)	R² = 0.60–0.70	NRCAN product
Greece wildfires	Evros region	2023	~93,000 ha	~40–55%	Copernicus EMS assessment	< 5 days (crisis mode)

The high-severity threshold matters for management decisions: In California chaparral and conifer forests, areas with dNBR > 0.66 (high severity) are at dramatically elevated risk of post-fire debris flows when rainfall occurs. USGS debris flow hazard assessments use dNBR maps directly to identify watersheds where emergency stabilization is most critical — a direct link from satellite severity classification to emergency resource allocation.

Accuracy by classification level: The ±1 severity class error rate (e.g., classifying moderate as high, or high as moderate-high) is approximately 20–30% in most validation studies. The practical implication is that burn severity maps should be used for strategic prioritization, not treated as ground-truth at individual pixel level. Validation is most important in heterogeneous landscapes with mixed conifer/hardwood or complex topography where spectral variability is high.

Beyond NBR: Additional Indicators

NDVI change: Simpler than NBR but less sensitive to burn severity; more sensitive to vegetation response timing.

Burn Area Index (BAI): Designed to emphasize charcoal signal in the red-NIR spectral space.

SAR-based detection: Sentinel-1 coherence loss and backscatter change detect burned areas through smoke and clouds — useful when optical imagery is unavailable immediately post-fire.

Thermal anomalies: Active fire detection from MODIS and VIIRS maps fire progression in near-real-time, complementing post-fire severity assessment.

How accurate is satellite burn severity mapping?

Accuracy depends on how finely you slice the severity scale: a simple burned/unburned split is highly reliable, while distinguishing moderate from high severity is much harder because those classes overlap spectrally. Errors concentrate at class boundaries and in heterogeneous terrain, so the maps are best used for strategic prioritization rather than pixel-level ground truth. Validation against field-measured Composite Burn Index (CBI) plots breaks down as follows:

• Binary burned/unburned: 90-95% accuracy
• 4-class severity: 70-80% overall accuracy
• Continuous severity (correlation with CBI): R² = 0.60-0.80

Errors are highest at class boundaries (where moderate and high severity overlap spectrally) and in heterogeneous landscapes where small-scale severity variation occurs within single pixels.

Burn severity mapping from satellites has become so routine and reliable that it's embedded in standard post-fire management workflows worldwide. The combination of well-understood physics (NBR), free satellite data (Landsat, Sentinel-2), and established operational programs (MTBS, Copernicus EMS) means that every significant wildfire on Earth can be mapped to a consistent standard — providing the spatial intelligence that post-fire landscape management requires. Burn severity is the post-fire half of a wider picture: pair it with a live active-fire map to track a blaze as it burns, and with forest degradation monitoring to catch the understory fires that leave the canopy standing but the forest changed.