dNBR: Turning Two NBR Maps Into a Burn-Severity Map
A burn scar tells you where fire passed; dNBR tells you how hard it hit. By subtracting post-fire NBR from pre-fire NBR, it isolates the change caused by the fire itself and scales it to recognised severity classes.
The physical basis
Differencing cancels the stable background and leaves the fire-driven drop in NBR. The USGS / Key–Benson thresholds then bin the result into unburned, low, moderate and high severity.
The formula
where NIR is near-infrared reflectance, SWIR2 is shortwave-infrared 2 reflectance (~2.2 µm).
Reading the values
0.66 high severity (USGS/Key & Benson thresholds).
Which bands to use
| Sentinel-2 | NIR B8 · SWIR2 B12 |
| Landsat 8/9 | NIR B5 · SWIR2 B7 |
In practice
It is the backbone of official burn-severity mapping (such as MTBS in the United States) and of post-fire erosion, recovery and rehabilitation planning.
Where it struggles
dNBR lives or dies on comparability between the two dates — same sensor, similar season, cloud-free. Vegetation green-up or senescence between images can inflate or mask severity, so choose anniversary-date pairs where possible.
Compute dNBR on your own study area
Skip the code. Draw or upload a boundary and Spatial Research Suite runs dNBR on live Sentinel-2 or Landsat imagery — with cloud masking, exports and citations built in.
Run this analysis in GISforus →Frequently asked
What dNBR value counts as high severity?
Under the common USGS/Key–Benson scheme, dNBR above roughly 0.66 is high severity; 0.27–0.66 moderate; 0.1–0.27 low; below 0.1 unburned.
Why must the two dates be comparable?
Because dNBR measures change — differences in season, sun angle or sensor between dates create false severity signal.