Wildfire Black Box Methodology & Evidence Standard

TECNOBOSQUE · EVIDENCE STANDARD · METHODOLOGY v1.0

Wildfire Black Box
Methodology & Evidence Standard

Permanent wildfire intelligence built to preserve uncertainty

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.

METHOD · WILDFIRE BLACK BOX v1.0 EVIDENCE INDEX · TB-ECI beta 0.1 RELEASE · SEPTEMBER 2026 5 PILOT RECORDS
Scope. Wildfire Black Box is an independent evidence-preservation and technical-analysis method. It is not an official investigation, an incident-command product, a legal finding or a probability model of individual loss.
METHOD ARCHITECTURE

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.

01
CAPTURE
02
CLASSIFY
03
SCORE
04
INTERPRET
05
VERSION
CLAIM CLASSIFICATION

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.

OBSERVED

Directly measured or visible in an identified observation or dataset.

REPORTED

Stated by an identified source without Tecnobosque reproducing the underlying observation.

DERIVED

Calculated from disclosed inputs using a documented transformation or formula.

INTERPRETED

Technical interpretation that goes beyond the literal source material.

DISPUTED

Credible sources conflict and the difference has not been reconciled.

SUPERSEDED

Retained in the history but replaced by later evidence for current use.

UNKNOWN

The available evidence is insufficient for a responsible claim.

CORE PRINCIPLE

Uncertainty is recorded as information. It is never hidden simply to make a record appear complete.

TECNOBOSQUE EVIDENCE CONFIDENCE INDEX

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.

PUBLIC SCORING MODEL 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?
85–100 VERY HIGH
70–84 HIGH
50–69 MODERATE
30–49 LOW
0–29 VERY LOW

The bands describe evidence quality and auditability. They are not statistical confidence intervals.

ANTI-FALSE-CONFIDENCE RULES

Evidence strength has hard limits

Certain evidence conditions place an explicit ceiling on the confidence that can be assigned to a claim.

CAP 69 · MODERATE
An unresolved conflict between credible sources prevents a claim from exceeding 69.
CAP 49 · LOW
A single unarchived social-media post or unattributed media report prevents a claim from exceeding 49.
CAP 29 · VERY LOW
An anonymous screenshot, unknown-origin map or unverifiable quotation prevents a claim from exceeding 29.
SUPERSEDED ≠ CURRENT
A superseded or retracted value cannot be presented as the current value.
REPUBLICATION ≠ CORROBORATION
Multiple outlets repeating one upstream statement count as one evidence line.
CAUSE ≠ LEGAL RESPONSIBILITY
A cause label without released supporting investigation cannot establish individual responsibility or a legal finding.
DERIVED VALUES MUST BE REPRODUCIBLE
Inputs, units and transformation must be disclosed for a derived value.
UNKNOWN IS A VALID RESULT
Missing evidence is retained as uncertainty rather than replaced by speculation.
CLAIM PROFILE

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.

REVISION CONTROL

Versioning and correction

A permanent record must be able to change without rewriting history.

R1
EDITORIAL

Metadata or editorial clarification that does not change a claim. Public action: patch note.

R2
SMALL CORRECTION

Factual correction without a material interpretation change. Public action: increment minor version.

R3
MATERIAL EVIDENCE

New evidence, chronology or a material score/finding change. Public action: increment major version.

R4
SCHEMA CHANGE

Processing or schema change that may affect automated use. Public action: new major version and compatibility notice.

R5
FUNDAMENTAL ERROR

Withdrawal or replacement required. Public action: preserve a tombstone and issue a replacement.

PERMANENT-RECORD RULE

No material value disappears without a dated revision note. Superseded information remains visible so readers can reconstruct what was publicly known at each stage.

ONE METHOD · TWO CONTROLLED PROJECTIONS

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.

PUBLIC EVIDENCE LAYER

Wildfire Black Box

Curated claims, public sources, evidence states, final TB-ECI results, limitations, visible conflicts, permanent URLs and material revision history.

PRIVATE INTELLIGENCE LAYER

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.

METHOD VALIDATION SET

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.

PILOT 01 · UNITED STATES · 2025

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
WHY THE IMPACT CLAIM IS MODERATE: official Los Angeles County and CAL FIRE structure totals currently differ. Both values remain visible until their definitions or revision history can be reconciled.
PILOT 02 · UNITED STATES · 2023

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
WHY SOME CLAIMS REMAIN OPEN: official sources differ by minutes on phase timing, while building totals cannot be treated as interchangeable without controlled definitions.
PILOT 03 · CANADA · 2016

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
WHY THE EVACUATION CLAIM IS MODERATE: authoritative sources use different population scopes. The mass evacuation is well supported, but the exact total remains definition-dependent.
PILOT 04 · AUSTRALIA · 2009

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
WHY SCOPE IS PART OF EVIDENCE QUALITY: homes, buildings, hectares and individual fire perimeters describe different objects and aggregation levels. They must remain distinct.
PILOT 05 · UNITED STATES · 2025

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
WHY REVIEW SCOPE IS EVIDENCE: a system-wide weakness finding and a narrower conclusion about particular decisions can both be true when their geography, time window and remit differ.
GOVERNANCE

Rules above the score

The methodology is governed by controls that apply regardless of incident, source or commercial context.

Commercial relationships never increase an evidence score or alter a technical conclusion.
Automated extraction may propose claims; a human approves public evidence states and scores during beta.
Material causal and legal claims receive heightened review and restrained wording.
Method revisions are versioned and any resulting recalculations are disclosed.
The beta methodology will continue to be tested on historical fires before any claim of external validation.
Uncertainty, disagreement and missing data remain visible components of the record.
TECHNICAL FOUNDATIONS

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.

W3C PROV-O

General provenance relationships for entities, activities, sources and derivation.

OPEN REFERENCE →
NASA EARTHDATA

Data-product quality, documentation and uncertainty principles used as a scientific reference framework.

NASA EARTHDATA →
USGS DATA QUALITY

Recommended practices for scientific data-quality assessment and review.

OPEN REFERENCE →
USGS REVISION GUIDANCE

Scientific-data revision principles informing permanent-record change management.

OPEN REFERENCE →
THE METHOD IN PRACTICE

From evidence standard to wildfire intelligence

Explore the permanent incident records or compare how the same evidence architecture behaves across different wildfire disasters.