Tecnobosque Wildfire Knowledge · Technical explainer · Research note · Prepared 5 November 2026 · Reviewed 6 November 2026
Three different scientific questions
Fire intensity concerns the rate of energy release during combustion. Burn severity describes effects of fire on vegetation, soil or other environmental components, with definitions varying across ecological and remote-sensing applications. Spectral change describes how remotely sensed reflectance differs between images. These are related but not interchangeable quantities: a post-fire satellite pixel does not directly record flame length, radiant heat flux or the intensity experienced by responders.
MTBS describes its burn-severity product as mapped patterns of fire effects detectable in satellite imagery, including changes to vegetation and charred material, and stresses the limitations of the interpretation. MTBS also distinguishes its burn scar products from operational incident perimeters. MTBS FAQ
What NBR measures
The Normalized Burn Ratio (NBR) contrasts near-infrared (NIR) and short-wave infrared (SWIR) reflectance:
NBR = (NIR − SWIR) / (NIR + SWIR)
For Landsat 8/9, the USGS NBR product uses Band 5 (NIR) and Band 7 (SWIR). The product may be distributed as scaled integer data, and users must apply the published scale factor before interpreting the physical index. Use surface reflectance products appropriate to the instrument. USGS Landsat NBR
The differenced NBR (dNBR) is commonly calculated from paired images:
dNBR = NBR(pre-fire) − NBR(post-fire)
The subtraction makes a change product. An unburned pixel that changed because of drought, harvest, shadows, snow or phenology may also generate spectral differences. A fire perimeter or independent validation is required before attributing every changed pixel to burning. USGS burn severity glossary
A transparent, synthetic pixel calculation
Imagine a valid pre-fire pixel with NIR reflectance = 0.42 and SWIR = 0.18. Its NBR is (0.42 − 0.18) / (0.42 + 0.18) = 0.40.
Now imagine a later pixel at the same location, correctly co-registered and masked, with NIR = 0.19 and SWIR = 0.31. Its NBR is (0.19 − 0.31) / (0.19 + 0.31) = −0.24. Therefore dNBR = 0.40 − (−0.24) = 0.64. All inputs are hypothetical; no real wildfire or image acquisition is implied.
What can be concluded? The pixel shows substantial spectral change in bands relevant to fire-effects mapping. What cannot be concluded from this calculation alone? The precise flame intensity, a universally valid ecological severity class, how long the fire burned, or whether a specific structure was exposed.
Why threshold tables need caution
Some workflows classify dNBR into low, moderate and high categories. Those labels are conditional on sensor preprocessing, the pre- and post-fire image dates, vegetation type, environmental background and the chosen classification calibration. MTBS explains that severity classification depends partly on analyst interpretation and that locally informed thresholds may be appropriate. Never present an arbitrary universal dNBR boundary as an incident-specific measured truth. MTBS methods
A robust publication should record the imagery product, acquisition dates, cloud/shadow mask, pixel size, band definitions, NBR scaling, seasonal comparability, dNBR thresholds, field-validation evidence where available, and which areas were excluded from analysis.
Image timing changes what the map says
Satellite mapping shortly after the incident may capture initial visible effects, while images months later may also capture regrowth, delayed vegetation mortality, erosion or seasonal changes. MTBS documents how previously severely burned areas that burn again can be categorized in ways that do not represent ecological resilience, because image-based indices track changes from the immediate pre-fire condition. Thus the same numerical dNBR does not necessarily indicate the same ecological consequences across ecosystems or fire histories. MTBS FAQ
Resolution, exclusions and interpretation
At a nominal 30 m resolution, one pixel represents 900 m² = 0.09 ha before considering projection geometry and raster alignment. A satellite-based perimeter or classification is not a centimetre-accurate boundary. Clouds, smoke, sensor artifacts, sub-canopy burning and tiny unburned islands can cause the mapped footprint to diverge from actual heterogeneous fire effects. Historical datasets often mix imagery of different dates and mapping objectives.
To report a result responsibly, state: “X hectares were assigned to class Y by this named sensor-product-method at date Z,” rather than “X hectares burned at precisely severity Y” without qualifications. In a reburn, include prior burn history and baseline condition.
Tecnobosque minimum evidence standard
A publishable case-level severity analysis should supply a reproducible method; source links and rights; the actual measured pixel counts or polygon areas; quality-control masks; sensitivity checks around thresholds when possible; uncertainty discussion; and a clear distinction between satellite change, ground ecological condition and real-time fire behaviour. This page supplies methodology and an illustrative calculation, not a new severity measurement of any particular wildfire.
Related Tecnobosque research: Perimeter discrepancies; QGIS perimeter comparison; evidence timeline; Riverside Fire 2020 case.