Remote Sensing Labs
Browse all docs

Burn Severity Mapping

Burn severity mapping asks a focused question of the imagery: where a fire has passed, how much did it change the surface, and how confident are we in that estimate unit by unit. The canonical measurement is the differenced Normalized Burn Ratio (dNBR), computed by comparing pre-fire and post-fire imagery, but the number only means something inside a full workflow — from framing the assessment to interpreting the classified map. Severity maps are a recovery-phase product used by rehabilitation teams, land managers, and recovery funders; who needs them and on what clock is the subject of Disaster Response and Forestry, and fire’s place among ecosystem changes is taken up in Environmental Monitoring.

Frame the assessment question first

Start with the assessment itself, because it fixes every later choice. Decide the burn perimeter and area of interest, the pre-fire reference window (a clear, representative period before ignition), and the post-fire window: one opening near containment, usually within a few days of it, for an initial assessment, or the growing season after the fire — ideally matched to the pre-fire image’s time of year — for an extended assessment that captures delayed mortality and early recovery. Name the severity classes the project actually needs — commonly unburned, low, moderate, and high, with some schemes splitting moderate into moderate-low and moderate-high and adding an enhanced-regrowth or increased-greenness class — and who consumes them. A recovery-monitoring baseline can use the dNBR classes directly; a post-fire emergency (BAER) team needs soil burn severity, which in US practice starts from a dNBR-derived Burned Area Reflectance Classification (BARC) and is then adjusted in the field, because dNBR mostly records what happened to the vegetation. Initial and extended severity answer different questions, and a threshold that is honest for one is misleading for the other.

Acquire suitable pre- and post-fire imagery

The Normalized Burn Ratio uses near-infrared and the longer shortwave-infrared band near 2.2 µm (SWIR2), so the first practical step is choosing an optical sensor that carries both, with the spatial detail and revisit your perimeter needs. On Landsat 8 and 9 that pair is band 5 and band 7 (band 4 and band 7 on Landsat 4–7); on Sentinel-2 it is B8A (or B8) and B12. The shorter SWIR1 band near 1.6 µm (Landsat 8/9 band 6, Sentinel-2 B11) is a common mistake: it gives a weaker burn signal and does not match published thresholds. The Resolution concept frames the spatial tradeoff — a finer pixel resolves small stands and edges, a coarser sensor maps a large fire quickly but blurs mixed-severity patches. The severity signal lives in the NIR and SWIR response the Spectral Bands concept describes. Use calibrated surface reflectance, following the Digital Imagery concept, and hold season and sun geometry as close as the archive allows so phenology is not mistaken for fire effect. In practice that means Sentinel-2 for a recent fire, and Landsat when the pre-fire reference has to reach further back than the newer archives go. A perimeter is usually a small part of each scene it falls in, so this is a workflow that rarely needs whole frames: fetching just the burned window is the read pattern cloud-optimized GeoTIFFs exist to support, and it is what keeps re-running the pair for an extended assessment a season later cheap.

Prepare the imagery before you measure

Mask both dates before computing anything, following the Cloud Masking concept: besides cloud, cloud shadow, and snow, a fire scene often carries lingering smoke, and any of them left in will show up as fake severity. Put the pair on one coordinate reference and one grid, as the Coordinate Systems concept requires; a half-pixel misregistration manufactures a bright ring of false change at every sharp boundary. Where terrain is steep, topographic shadow shifts between two dates with different sun angles and can imitate burn signal, and late-season fires with low sun make that worse.

Compute NBR and difference it to dNBR

The Normalized Burn Ratio is NBR = (NIR - SWIR2) / (NIR + SWIR2). It works because fire drives the two bands in opposite directions: combustion removes the NIR-bright canopy, so near-infrared reflectance drops, while exposed char, ash, and dry soil raise SWIR2 reflectance, so NBR falls sharply over burned ground. Compute NBR the same way — same bands, same reflectance product — for the pre-fire and post-fire dates, then take the difference dNBR = pre-fire NBR - post-fire NBR. Higher dNBR indicates greater change. In principle unburned surfaces sit near zero, but in practice two dates rarely match exactly in phenology and moisture, so unburned vegetation outside the fire carries a small systematic difference of its own. Standard practice is to measure that dNBR offset — the mean dNBR of unburned pixels of similar vegetation just outside the perimeter — and subtract it, so class boundaries mean the same thing from fire to fire. This differencing is the multi-temporal reasoning the Change Detection concept develops: severity is not a property of one image but the departure between two.

Classify severity into interpretable bands

A continuous dNBR image is data, not yet an answer. Turn it into the severity classes your question named by applying thresholds, which is the labeling step the Classification Basics concept covers. The widely used starting scheme comes from Key and Benson (2006), and it is expressed in dNBR × 1000: roughly −100 to +99 unburned, +100 to +269 low, +270 to +439 moderate-low, +440 to +659 moderate-high, and +660 and above high severity, with negative values below −100 marking enhanced regrowth. Applying those numbers to unscaled dNBR (which runs from about −2 to +2) labels almost an entire fire unburned. Key and Benson describe the ranges as flexible and scene-pair dependent, and US national mapping (MTBS) sets thresholds fire by fire, so calibrate them against field plots of the Composite Burn Index (CBI) — the 0–3 field rating of severity that dNBR is validated against — wherever you have them.

Absolute dNBR also depends on how much vegetation was there to lose: a severe burn in sparse shrubland produces a smaller difference than a moderate burn in dense forest. The relativized dNBR (RdNBR; Miller and Thode, 2007) divides by a function of pre-fire NBR to compare severity across dense and sparse cover, but it becomes unstable where pre-fire NBR is near zero; the relativized burn ratio (RBR; Parks et al., 2014) was proposed partly to avoid that. In grass and shrub systems every variant weakens quickly, because green-up within weeks can erase the signal. Whatever scheme you choose, record the exact break points and the offset, because the classified map is only reproducible if they travel with it.

Separate real fire effect from confounders

Not every large dNBR is fire. Before believing the map, rule out the ordinary explanations: normal phenology between two seasonally offset dates, agricultural harvest, water and its variable level, exposed soil, snow, and step changes introduced by switching between sensors. Terrain shadow is the classic severity confounder — a slope that fell into shadow on only one date reads as change that has nothing to do with the fire — and residual cloud or smoke the mask missed does the same. The Time Series concept matters here because a thin or seasonally biased set of clear observations can bias the pre-fire baseline on its own. Treat a severity patch as a hypothesis and check whether it survives the obvious confounders before it is escalated for planning.

Produce outputs, interpret, and make it reproducible

A severity run typically delivers a classified severity map, a table of burned area by class, and summaries rolled up to the units the project cares about — management unit, slope class, vegetation type, or watershed. State the uncertainty plainly: which areas were thin on clear pixels, where the cloud or terrain mask was doing heavy work, and how far the local calibration was trusted. Record the perimeter and area of interest, the exact pre-fire and post-fire dates and sensor, the bands used, the masking and reflectance products, the NBR/dNBR (or RdNBR/RBR) definition and scaling, the dNBR offset, the classification thresholds, and the summary units — the parameters that let an extended assessment next season be compared honestly with this one. None of those parameters depends on a platform: the pre- and post-fire scenes can be found through any STAC catalog, and the Earth Engine indices lesson codes two other burn indices, the Burned Area Index and a thermal NBR variant.

Sources