REBURN RESEARCH GUIDE · Reviewed 9 October 2026
Does low burn severity after a second wildfire mean the forest recovered? No: the map alone cannot establish that. A reburn assessment compares the vegetation present before the later fire with conditions afterward. If the earlier fire already removed much of the canopy, a small additional spectral change can conceal a large departure from the original forest. Read the baseline, interval and measured outcome together. [1]
A reburn changes the question behind the map
A reburn occurs when a later fire overlaps land burned previously. Record both events and the interval between them: “previously burned” does not say whether the site had one growing season or several decades to develop before burning again.
MTBS explicitly warns that high-severity reburns can receive low mapped severity after herbaceous growth, even where the original forest canopy is absent. The later map describes recent vegetation change; it does not automatically capture accumulated change across both fires or ecological integrity. [1]
Keep three questions separate: what changed during the later fire? what vegetation is present now? and how does that vegetation compare with the forest before either fire? One pair of images does not answer all three. For the underlying products, start with our guide to NBR, dNBR and MTBS maps.
Worked example: the same post-fire NBR, different dNBR
dNBR is pre-fire NBR minus post-fire NBR. RAVG documentation distinguishes this difference from relative indices such as RdNBR, which accounts for pre-fire vegetation density. Derived canopy-loss and basal-area-loss products also involve models calibrated against field observations; they are not the raw index itself. [2]
Invented arithmetic example—not observations from any wildfire. Suppose two image pairs have the following values:
| Illustrative image pair | Pre-fire NBR | Post-fire NBR | dNBR | If stored as dNBR × 1,000 |
|---|---|---|---|---|
| A: higher starting NBR | 0.60 | 0.00 | 0.60 − 0.00 = 0.60 | 600 |
| B: lower starting NBR | 0.20 | 0.00 | 0.20 − 0.00 = 0.20 | 200 |
Pair B has one-third of pair A’s dNBR because its starting value is lower. The arithmetic does not establish one-third of the fire intensity, tree mortality or ecological damage. These numbers are not proposed severity thresholds.
The example demonstrates baseline dependence, not a method for reconstructing forest condition. Identical post-fire NBR values do not establish identical plant communities, and NBR of zero does not mean zero vegetation. Check scaling in the actual dataset rather than assuming every delivered raster uses the illustrated convention.
Nor does choosing a relative index make field context unnecessary. RAVG cautions that its regression models are not calibrated to non-forest vegetation. Before interpreting a canopy-loss estimate in a site that changed from forest to shrubs, check the model’s applicability to the vegetation present before the later fire. [2]
Compare rates within each prior-fire group
A second mistake is comparing hectares without checking how much land each group contains. Define the shared mapped area first, retain only locations with valid classifications for both events, and report exclusions separately.
Hypothetical accounting example: the table below contains 100 hectares with valid classifications for both fires. It is not a reanalysis of Biscuit or another incident. “Other classes” groups all remaining valid severity classes solely to simplify this example.
| Prior-fire classification | Later fire: high severity | Later fire: other classes | Group total |
|---|---|---|---|
| High severity | 25 | 35 | 60 |
| Other classes | 15 | 25 | 40 |
| Total | 40 | 60 | 100 |
Within the prior-high group, the later-high share is 25 ÷ 60 = 41.7%. Within the prior-other group it is 15 ÷ 40 = 37.5%. The difference is about 4.2 percentage points. Reporting only “25 versus 15 hectares” hides the different group sizes.
This table describes an association in invented data. It contains no adjustment for weather, terrain, vegetation, management or spatial dependence. It cannot establish that the first fire caused the difference, and its class transitions do not directly measure forest recovery. Matching class labels across fires also requires checking what each classification actually represents.
What Silver–Biscuit research adds
Thompson, Spies and Ganio examined southwest Oregon land burned in 1987 and again in the 2002 Biscuit Fire, a 15-year interval. Using satellite data, agency records and aerial photographs, they found that previously severe areas tended to reburn at high severity after accounting for several terrain and biophysical covariates. Salvage-logged and planted areas also burned more severely than comparable unmanaged areas. [3]
That is evidence from a particular landscape and management history, not a universal rule that every reburn or treatment has the same effect. Read the Biscuit Fire case study for the wider context.
Vegetation response is a different outcome. Donato and colleagues compared short-interval and long-interval burns in the same region two years after the 2002 fire. Their field study found substantial persistence of native flora in both burn groups; the short-interval group had the highest species richness and total plant cover. Conifer regeneration was present in both groups. [4]
Tecnobosque interpretation: these findings make a blanket “reburn means ecological failure” claim inappropriate. They also do not turn early plant cover or seedling presence into proof that a mature forest canopy has recovered. Map change, species persistence and long-term forest structure need distinct measurements.
What the 2026 USGS release does—and does not—show
On 3 February 2026, USGS published a data release by Nadav Mouallem and Larissa L. Yocom on ponderosa pine regeneration and plant-community composition in high-severity reburn patches of Arizona’s Colorado Plateau. Its stated questions include the ecological significance of dNBR in reburns. [5]
The cited landing page identifies a data release and describes its scope; it does not supply numerical results sufficient to quote a regeneration effect here. This guide does not present the release as a peer-reviewed finding that reburning universally improves or prevents forest recovery.
Six checks before making a reburn claim
Tecnobosque editorial checklist: use these questions to make an interpretation traceable.
- Identify the sequence. Record both incident names or IDs, dates and the interval. State whether the comparison includes only the overlap or a larger landscape.
- Inspect the baseline. Describe the vegetation before the second fire using evidence beyond the severity colour. Was it forest, young regeneration, shrubs or herbaceous cover?
- Keep observation dates visible. Record the pre- and post-fire image dates for each assessment. A map produced soon after burning and one incorporating the next growing season observe different stages.
- Audit the product. Record the index or model, units, scaling, class thresholds and valid-data mask. Do not silently combine a continuous index, a thematic class and a modeled percentage.
- Use explicit denominators. Report hectares and within-group shares. Preserve excluded area and avoid treating masked locations as unchanged or low severity.
- Name the ecological outcome. If the claim concerns recovery, specify the measured feature and observation year: canopy cover, seedling survival, species composition or another defined outcome. A colour alone does not establish the trajectory.
A reporting template: “Within [defined overlap], [area/share] was classified as [class] in the later assessment, using imagery from [dates]. The site had [documented pre-reburn vegetation]. Recovery was evaluated separately using [field measure and date], or remains unmeasured.” This wording exposes the evidence still needed rather than filling the gap with an assumed mechanism.
Primary research and official documentation
- MTBS: Frequently Asked Questions — see the question on interpreting severity in previously high-severity reburns.
- USGS / Forest Service RAVG: Background, Products & Applications — indices, field-calibrated models and vegetation applicability.
- Thompson, Spies and Ganio (2007): Reburn severity in managed and unmanaged vegetation in a large wildfire. DOI: 10.1073/pnas.0700229104.
- Donato and colleagues (2009): Vegetation response to a short interval between high-severity wildfires in a mixed-evergreen forest. DOI: 10.1111/j.1365-2745.2008.01456.x.
- Mouallem and Yocom (2026): Ponderosa Pine Regeneration and Plant Community Composition in High Severity Reburn Patches of Arizona’s Colorado Plateau. USGS data release, 3 February 2026. DOI: 10.5066/P1AT78PN.
Continue with how to read fuel-treatment evidence, or explore the Fire Archive.