Wildfire case studies · California, United States · 2020
Castle Fire 2020: How Giant Sequoia Mortality Estimates Changed
Evidence review: Historical documentary analysis. Sources checked 2 November 2026. This page is not a current incident alert.
The Castle Fire is a case study in how a mortality estimate is constructed. Its legacy cannot be described responsibly by quoting a single percentage without the population, observation period and method behind it. This analysis follows the transition from preliminary loss estimates to later research and restoration decisions.
The central finding—and its denominator
The 2021 preliminary assessment by Nathan Stephenson and Christy Brigham estimated that the 2020 fire killed 31–42% of large sequoias within its footprint, equivalent to 10–14% across the species’ natural Sierra Nevada range. These are different denominators, not competing estimates of the same proportion. “Large” refers to trees at least approximately four feet in trunk diameter, not every seedling or tree. Source: preliminary mortality assessment.
The assessment combined mapped burn severity and mortality information from sampled groves. It also explained why earlier estimates were lower: older grove maps included non-sequoia areas. Some observation sets were less than a year post-fire, while others incorporated three years of delayed mortality. Consequently, a mapped severity class was an input to estimation, not a complete census of dead trees.
A practical reading test
| Question | What to record | Why it matters |
|---|---|---|
| Which population? | Large trees within the fire versus large trees range-wide | A percentage changes when the denominator changes |
| Which boundary? | Actual sequoia distribution versus administrative grove polygons | Non-sequoia land can distort area-based inference |
| Which observation window? | Immediate damage versus later mortality | A surviving crown at one visit does not establish long-term survival |
| Which method? | Field observations, remote sensing and transferred mortality relationships | An estimate should not be presented as a tree-by-tree count |
What newer research adds
A 2026 study listed by USGS estimated mortality for 26,403 sequoias in 19 national-park groves affected by Castle and KNP combined. It integrated satellite imagery, lidar and field observations in a Bayesian analysis. The study estimated lower mortality odds where prescribed burns had occurred within the preceding ten years. Its counterfactual outcomes are model estimates, and its sampled population is not identical to the earlier Castle-only, range-wide calculation. Source: Dixon and colleagues, 2026.
Tecnobosque interpretation: The newer work should inform understanding of treatment effects, but its pooled results cannot simply replace the Castle figure. First reconcile incident coverage, tree inventory, geographic scope and observation dates. Otherwise an apparent revision may merely be a comparison of different populations.
Mortality and regeneration are separate evidence questions
NPS later used field surveys to decide where planting was necessary after Castle and KNP. Its project record reports planting in Board Camp in October 2023 and April 2025, following evidence of insufficient natural regeneration. The decision process concerns future forest establishment, not just the number of mature trees lost. Source: NPS seedling re-establishment project.
A dead-tree inventory and a seedling survey measure different outcomes. Even a planting record does not demonstrate successful establishment: that requires survival monitoring. A useful recovery record would preserve the survey area, species, planting date and later survival observations, rather than labeling the grove “restored” as soon as seedlings are placed.
Limits of this case study
This page reviews published evidence. It does not reclassify imagery, calculate a new mortality estimate or infer loss for individual named trees. It does not treat all burned forest as high severity. The preliminary estimate remains identified by its publication context rather than being relabeled a current census.
Related evidence: Chuckegg Creek: early area estimates and final extent. Both cases require keeping the measured population and reporting date attached to a number.
Cite this analysis: Tecnobosque, “Castle Fire 2020: How Giant Sequoia Mortality Estimates Changed.” Documentary analysis, 2 November 2026. Cite the linked original records for their findings.