Wildfire Black Box
Methodology & Evidence Standard
A wildfire record should preserve not only what is known, but where every material claim came from, how it changed, what evidence supports it and what remains uncertain.
From an event to an auditable record
Each reference fire is assembled through the same evidence path. The permanent incident URL remains stable while claims, source registers and revision history evolve.
The five-stage evidence path
Every material statement passes through the same controlled sequence.
Seven evidence states
A material statement cannot enter the Wildfire Black Box as an unqualified fact. It receives an explicit evidence state that can change when new documentation appears.
Directly measured or visible in an identified observation or dataset.
Stated by an identified source without Tecnobosque reproducing the underlying observation.
Calculated from disclosed inputs using a documented transformation or formula.
Technical interpretation that goes beyond the literal source material.
Credible sources conflict and the difference has not been reconciled.
Retained in the history but replaced by later evidence for current use.
The available evidence is insufficient for a responsible claim.
Uncertainty is recorded as information. It is never hidden simply to make a record appear complete.
TB-ECI: a claim-level evidence index
TB-ECI scores the strength and auditability of the evidence supporting one material claim. It does not estimate the statistical probability that the claim is true.
TB-ECI = T + D + C + X + M + U
Maximum score: 100
| DIMENSION | MAXIMUM | QUESTION |
|---|---|---|
| T — Traceability | 25 | Can the exact artifact, publisher, date and version be retrieved? |
| D — Directness | 20 | How close is the evidence to the underlying observation, measurement or responsible authority? |
| C — Independent corroboration | 20 | Do genuinely independent evidence lines agree rather than repeat one upstream source? |
| X — Spatial and temporal fit | 15 | Does the evidence match the correct incident, location, time and resolution? |
| M — Maturity and stability | 10 | Is the information preliminary, consolidated, final, superseded or still changing? |
| U — Uncertainty disclosure | 10 | Are gaps, processing, limitations and contradictions explicitly stated? |
The bands describe evidence quality and auditability. They are not statistical confidence intervals.
Evidence strength has hard limits
Certain evidence conditions place an explicit ceiling on the confidence that can be assigned to a claim.
An unresolved conflict between credible sources prevents a claim from exceeding 69.
A single unarchived social-media post or unattributed media report prevents a claim from exceeding 49.
An anonymous screenshot, unknown-origin map or unverifiable quotation prevents a claim from exceeding 29.
A superseded or retracted value cannot be presented as the current value.
Multiple outlets repeating one upstream statement count as one evidence line.
A cause label without released supporting investigation cannot establish individual responsibility or a legal finding.
Inputs, units and transformation must be disclosed for a derived value.
Missing evidence is retained as uncertainty rather than replaced by speculation.
No incident-wide score can hide weak evidence
Wildfire records contain claims of very different quality. Tecnobosque therefore publishes a confidence profile rather than a simple incident average: the distribution of core claims across the evidence bands, unresolved conflicts, superseded values, missing modules, review date and record version.
Versioning and correction
A permanent record must be able to change without rewriting history.
Metadata or editorial clarification that does not change a claim. Public action: patch note.
Factual correction without a material interpretation change. Public action: increment minor version.
New evidence, chronology or a material score/finding change. Public action: increment major version.
Processing or schema change that may affect automated use. Public action: new major version and compatibility notice.
Withdrawal or replacement required. Public action: preserve a tombstone and issue a replacement.
No material value disappears without a dated revision note. Superseded information remains visible so readers can reconstruct what was publicly known at each stage.
Transparency without surrendering the engine
Each incident has one canonical record. The public layer exposes the information required for auditability; Tecnobosque Core retains the deeper structured intelligence used for internal processing and future data products.
Wildfire Black Box
Curated claims, public sources, evidence states, final TB-ECI results, limitations, visible conflicts, permanent URLs and material revision history.
Tecnobosque Core
Detailed source lineage, exact per-claim component allocations, dependency relationships, internal transformations, derived variables, processing history, review intelligence and future licensed products.
Advertising, sponsorship and affiliate relationships never influence either evidence states or TB-ECI conclusions.
Records using this method
Each Wildfire Black Box pilot applies the same claim states, scoring rubric, confidence caps and revision principles so that evidence can be compared without hiding uncertainty.
Palisades Fire
The first provisional manual application tests conflicting official impact totals and evolving cause information.
| CLAIM | STATE | TB-ECI | BAND |
|---|---|---|---|
| Start, final area, containment and jurisdiction | Reported | 88 | Very high |
| Structure damage and destruction totals | Disputed | 58 | Moderate |
| Santa Ana wind and satellite smoke context | Observed / Reported | 88 | Very high |
| Current agency cause label: arson | Reported | 63 | Moderate |
| Detailed spread, evacuation and causal reconstruction | Developing | 42 | Low |
Lahaina Fire
Tests a globally significant urban-interface disaster with a documented two-phase ignition sequence and later systems-level findings.
| CLAIM | STATE | TB-ECI | BAND |
|---|---|---|---|
| Morning origin and official accidental-cause determination | Reported | 87 | Very high |
| One fire with two phases | Reported | 85 | Very high |
| Exact phase timestamps across official sources | Disputed | 64 | Moderate |
| Wind and dry-fuels context | Observed / Reported | 91 | Very high |
| Systems-level contributing factors | Interpreted | 82 | High |
| Exact Lahaina building totals | Unknown / Developing | 36 | Low |
Fort McMurray / Horse River Fire
Tests urban-scale evacuation, multiple dated satellite observations, systems reviews, built-environment losses and environmental monitoring.
| CLAIM | STATE | TB-ECI | BAND |
|---|---|---|---|
| May 1 initial report, size and location | Reported | 90 | Very high |
| May 3 Landsat observation and urban transition | Observed / Reported | 93 | Very high |
| Exact evacuation total across authoritative sources | Disputed | 64 | Moderate |
| May 24 observation-time footprint of 5,229 km² | Observed / Reported | 90 | Very high |
| Structures and dwelling units lost | Reported | 88 | Very high |
| Final official ignition cause in processed source set | Unknown | 40 | Low |
Black Saturday Bushfires
Tests a multi-fire disaster model where statewide impacts remain at disaster level while ignition and chronology remain attached to individual fires.
| CLAIM | STATE | TB-ECI | BAND |
|---|---|---|---|
| 173 fatalities across the disaster | Reported | 97 | Very high |
| 2,029 homes destroyed | Reported | 95 | Very high |
| Melbourne 46.4 °C and Hopetoun 48.8 °C | Observed / Reported | 96 | Very high |
| Wind change broadened active fire fronts | Observed / Interpreted | 94 | Very high |
| No single event-wide ignition cause | Derived / Interpreted | 88 | Very high |
| Approximately 400 fires and 78 communities | Reported | 82 | High |
Eaton Fire
Tests a versioned single-fire record where cause information changed after finalization and different official reviews answer different questions.
| CLAIM | STATE | TB-ECI | BAND |
|---|---|---|---|
| Final area: 14,021 acres | Reported | 97 | Very high |
| Aircraft grounded at about 6:45 p.m. | Interpreted | 84 | High |
| 3:25 a.m. West Altadena order preceded documented main-front crossing | Interpreted | 84 | High |
| AAR: no single failure; multiple systemic weaknesses | Reported | 92 | Very high |
| Citygate: no misconduct or delay in scoped decisions | Reported | 84 | High |
| Cause state: Under Investigation → Undetermined → Electrical | Reported | 96 | Very high |
Rules above the score
The methodology is governed by controls that apply regardless of incident, source or commercial context.
Built on established data-quality principles
Wildfire Black Box is an independent Tecnobosque methodology. It draws on established principles for provenance, scientific data quality, uncertainty and revision control without claiming that TB-ECI is an adaptation of any external operational standard.
General provenance relationships for entities, activities, sources and derivation.
OPEN REFERENCE →Data-product quality, documentation and uncertainty principles used as a scientific reference framework.
NASA EARTHDATA →Recommended practices for scientific data-quality assessment and review.
OPEN REFERENCE →Scientific-data revision principles informing permanent-record change management.
OPEN REFERENCE →From evidence standard to wildfire intelligence
Explore the permanent incident records or compare how the same evidence architecture behaves across different wildfire disasters.