Wildfire Black Box · Colorado · Historical evidence record
A destroyed home and a burned forest measure different outcomes. Fourmile Canyon offers a field investigation and a survey conducted before the fire. Read together, they show why a mitigation claim needs a defined objective, a comparable population and evidence about the home itself.
What happened in the Fourmile Canyon Fire?
The Fourmile Fire Protection District dates the start to 6 September 2010, near Boulder, Colorado. Its current historical page reports 6,181 acres burned, 169 homes lost and 3,000 community members evacuated. Those are the district’s attributed incident figures. [2]
Two research documents address different questions: the Forest Service’s Fourmile Canyon Fire Findings examines the fire and its effects; Brenkert-Smith and Champ’s Living with Wildfire examines an earlier homeowner survey. The later risk-management paper by Calkin and colleagues draws on both. It is a subsequent analysis, not a third independent property inventory. [1] [3] [4]
The district’s 169-home figure and the field report’s 168 should remain attributed. This review has not reconciled the underlying addresses or classification rules. A one-home difference is not a reason to average totals or silently replace one source’s inventory inside its calculations.
How many exposed homes were destroyed? Start with the denominator
The field report assumes significant exposure for 474 homes inside the final perimeter or within 100 feet of it. Its inventory reports 168 destroyed: 29 associated with crown fire and 139 with surface fire. [1]
| Question | Calculation | Meaning |
|---|---|---|
| Destroyed / assumed exposed | 168 ÷ 474 × 100 ≈ 35.4% | Loss share within this inventory |
| Surface-associated / destroyed | 139 ÷ 168 × 100 ≈ 82.7% | Distribution within the destroyed group |
| Surface-associated / assumed exposed | 139 ÷ 474 × 100 ≈ 29.3% | Joint classification within the exposure inventory |
Tecnobosque interpretation: these answer three different questions. The second describes the makeup of the losses. It cannot be read as the probability that a surface-fire-exposed home was destroyed. That probability would require the total number of homes exposed to surface fire, including survivors.
Likewise, the third is a share of the whole inventory, not a surface-fire failure rate. The missing denominator cannot be supplied by the total number of destroyed homes. Before using a percentage in a comparison, write the numerator and denominator in plain language.
“Assumed exposed” also records a study rule. A boundary-based selection does not mean every selected property experienced identical heating, ember deposition or duration. An area-average result cannot supply the exposure history of a particular address.
Surface-fire association does not reconstruct every ignition
Retain the term associated when describing the report’s classes. Classification of the surrounding fire and identification of the first ignition point on a building are different evidence tasks.
The later Calkin paper distinguishes wildfire exposure from a home’s susceptibility to ignition. Its home ignition zone includes building characteristics and immediate surroundings. It discusses embers igniting material on a building or producing nearby fires that reach it. [4]
This provides a useful framework, but a framework is not an address-level reconstruction. A photograph after destruction can document a condition without revealing the complete sequence that produced it. To investigate that sequence, retain observations of the building, nearby materials, possible heat sources, timing and any response activity.
For example, a surviving tree beside a destroyed home is not a thermometer reading. Nor does it identify the first combustible item that ignited. Evidence about vegetation effects and evidence about building ignition can inform each other while remaining separate records.
Why the pre-fire survey is not a mitigation success rate
Brenkert-Smith and Champ conducted their survey in 2007. Across Boulder County, 421 people responded after 1,148 invitations were successfully delivered. The authors later identified 127 respondents within the Fourmile evacuation zone; nine lost their homes. The article reports self-described actions and attitudes, not a controlled treatment trial. [3]
| Dataset | Selected group | Loss calculation | Selection boundary |
|---|---|---|---|
| Field inventory | 474 homes | 168 ÷ 474 ≈ 35.4% | Final perimeter and nearby homes |
| Pre-fire survey subset | 127 respondents | 9 ÷ 127 ≈ 7.1% | Survey response and later evacuation zone |
The survey’s broader response fraction is 421 ÷ 1,148 ≈ 36.7%. This describes response to delivered invitations, not participation among every Boulder County household. The 7.1% loss fraction concerns the evacuation-zone respondents, a different group from the field exposure inventory.
A lower observed loss fraction in the survey subset does not demonstrate a mitigation effect. The groups may differ in location, exposure, building characteristics and selection into the survey. Their membership has not been matched here. Dividing those two percentages would produce a numerical comparison without establishing its causal meaning.
The survey also reports that only 4% of its Fourmile respondents took none of the listed mitigation actions. Reporting at least one action says nothing by itself about its scope, completion, maintenance or adequacy for later conditions. [3]
Self-reported preparation is valuable evidence about behavior. It should not become an independently inspected measure of building resistance. Preserve the date and wording of the survey item, and keep a reported action separate from a verified condition.
What does Fourmile show about fuel-treatment effectiveness?
The field report’s treatment section says varied prescriptions, unfinished work and maintenance issues prevent a general efficacy inference. Post-fire evidence also leaves uncertainty about how a treatment encountered the fire. [1]
Editorial implication: start an evaluation by naming the intended outcome. Reduced vegetation damage, changed spread, better access and reduced home ignition are separate outcomes. A treatment can be evaluated against its specified purpose; evidence of one outcome does not automatically establish the others.
Calkin and colleagues frame risk reduction through both exposure and susceptibility. They also retain landscape values such as watersheds and recreation alongside structures. Their discussion does not make vegetation objectives interchangeable with home-survival objectives. [4]
A stronger comparison would record treatment boundaries, prescription, completion and maintenance dates before the event. It would then compare sufficiently similar sites, document the fire conditions encountered and measure the stated outcome. Differences in terrain or property selection would need to be addressed rather than assigned to the treatment.
This record therefore offers no universal percentage benefit, survival guarantee or homeowner blame. It identifies the measurements needed to test a claim. For an individual property, use current local requirements and a qualified site assessment; this historical synthesis is not a clearance specification.
Known At: the survey came before the disaster
| Checkpoint | Evidence object | Boundary |
|---|---|---|
| 2007 | Homeowner survey | Reported preparation before the fire |
| 6 September 2010 | Incident start | Documented retrospectively |
| 2011 | Survey article published | Earlier respondents identified within the evacuation zone |
| August 2012 | Field report published | Retrospective exposure assumptions and classifications |
| 2013–2014 | Risk-management paper | Online 16 December 2013; journal issue 14 January 2014 |
| 12 October 2026 | Tecnobosque evidence review | Access date; no current incident claim |
Recording attitudes before a disaster reduces reliance on later recollection of those attitudes. It does not remove nonresponse, self-report or exposure differences. Nor does publication of a finding after the fire establish that residents received that finding during the event.
Fire DNA and what remains unknown
- Exposure: the field inventory uses a stated spatial selection rule.
- Loss: use each source’s count with its own population.
- Ignition: surrounding-fire association is not a complete ignition sequence.
- Preparation: survey responses describe reported behavior.
- Mitigation: match the outcome measured to the objective claimed.
No severity ranking or TB-ECI score is assigned. This review does not determine the original ignition cause, reconcile every loss address or estimate the independent effect of a particular measure. Those remain separate research questions.
Primary sources, method and citation
- Graham and colleagues: Fourmile Canyon Fire Findings. Forest Service RMRS-GTR-289, August 2012. Full GovInfo-preserved primary PDF retrieved; this review focuses on printed pp. 55–60 and 64–68.
- Fourmile Fire Protection District: Four Mile Canyon Fire. Current agency historical page, accessed 12 October 2026.
- Brenkert-Smith and Champ: Fourmile Canyon: Living with Wildfire. Fire Management Today 71(2), 2011, pp. 31–37. Survey methods and Fourmile subset.
- Calkin, Cohen, Finney and Thompson: How risk management can prevent future wildfire disasters in the wildland-urban interface. PNAS 111(2), 2014, pp. 746–751. DOI: 10.1073/pnas.1315088111. Uses the earlier field report and survey; not an independent loss inventory.
Method: document review and separately identified arithmetic. Percentages are rounded to one decimal place. No address matching, raw survey reanalysis, fire simulation or original fieldwork was performed. Interpretation and suggested evaluation methods are Tecnobosque’s synthesis.
Cite: Tecnobosque, Fourmile Canyon Fire 2010, record TB-WBB-US-2010-FOURMILE, version 1.0, evidence review 12 October 2026.