FIRE-WUI Research — Stellar Nexus Institute

📊 Key Metrics at a Glance

34,622

Wildfire Incidents (1999–2020)

2.77M

Residential Structures Destroyed

$3.6B

Estimated Property Damage (SitReps)

12,403

Reported Wildfire Fatalities

80.7%

Fires Class G (5,000+ acres)

182,826

Situation Reports

📈 Temporal Trends — Wildfires Are Getting Worse

Year-over-Year Fire Escalation (ICS-209)

Year	Incidents	Avg Acres	Total Acres	Trend
1999	822	4,564	3.75M	—
2002	1,049	4,638	4.87M	↑
2006	2,761	3,060	8.45M	↑↑
2011	2,567	3,252	8.35M	↑↑
2015	2,130	4,945	10.53M	↑↑↑
2017	1,994	5,264	10.50M	↑↑↑
2018	1,540	5,455	8.40M	↑↑↑
2020	1,375	8,080	11.11M	⚠️ RECORD

Key Finding: While the number of large incidents has decreased since 2006, the average fire size has nearly doubled (4,564 → 8,080 acres). This confirms the "mega-fire" trend — fewer but more destructive fires, exactly the pattern FIRE-WUI research must address.

Fatalities Are Spiking

Year	Deaths	Injuries
2016	37	729
2017	80	1,282
2018	109	879
2020	63	837

2017–2018 saw the highest wildfire death tolls in the ICS-209 record (80 and 109 deaths), driven by devastating WUI fires like the Camp Fire (2018, 85 civilian deaths).

🔥 Cause Analysis — Lightning vs. Human Ignition

47.5%

Human-Caused (H)

31.3%

Lightning-Caused (L)

17.7%

Unknown (U)

2.9%

Other (O)

Cause Comparison — Size Matters

Cause	Count	Avg Acres	Max Acres	Implication
Human (H)	16,436	1,500	468,638	More frequent, smaller
Lightning (L)	10,835	6,712	1,305,592	Less frequent, 4.5× larger
Unknown (U)	6,134	5,116	1,032,648	Attribution gap

FIRE-WUI Relevance: Lightning-caused fires are 4.5× larger on average and account for the majority of mega-fire acreage. However, human-caused fires are more frequent and directly threaten WUI communities. An effective vulnerability model must account for both ignition pathways.

📐 Fire Size Class Distribution — The Mega-Fire Problem

Size Class	Acres	Incidents	Total Acres	% of Incidents	% of Acreage
A	< 1	840	202	2.4%	0.00%
B	1–9.9	3,236	9,293	9.3%	0.01%
C	10–99	3,007	123,333	8.7%	0.09%
D	100–299	8,954	1,523,962	25.9%	1.1%
E	300–999	8,641	4,585,125	24.9%	3.3%
F	1K–4.9K	5,940	13,081,835	17.1%	9.4%
G	5,000+	3,455	112,985,083	10.0%	80.7%

The 80/20 Rule, Amplified: Just 10% of incidents (Class G, 5,000+ acres) account for 80.7% of all burned acreage. This extreme concentration means that community vulnerability is disproportionately driven by the rare mega-fire — and current fire size class models (A–G) don't capture the exponential tail risk that WUI communities face.

🌲 Southeast Focus — The Understudied WUI Region

ICS-209 Wildfire Incidents by State (County-Level Links)

State	County Incident Links	WUI Relevance
FL	2,611	High — urban/WUI interface
MS	1,442	Moderate — rural WUI
AL	879	Moderate — Fort Rucker area
AR	659	Lower — Ozark WUI
SC	640	Moderate — pine plantation
NC	634	Moderate — Blue Ridge WUI
TN	529	Moderate — Gatlinburg 2016
GA	340	Lower — prescribed fire culture

Alabama & Fort Rucker: With 879 county-level wildfire incidents, Alabama has significant WUI exposure — particularly around Fort Rucker, the US Army's primary aviation training installation. Fort Rucker's Natural Resources Branch manages 280K+ acres of longleaf pine with an active prescribed fire program, making it an ideal community partner for FIRE-WUI research on military-WUI interfaces.

USFA Fire Deaths — Southeast Rates Are Alarming

State	Fire Deaths (2023)	Rate per Million	National Comparison
Alabama	121	23.7	2.3× national avg (10.5)
Tennessee	176	24.7	2.4× national avg
Georgia	172	15.6	1.5× national avg
Florida	201	8.9	Below national avg
National	3,428–4,446/yr	10.5–13.3	—

Health Disparity Connection: Alabama's fire death rate (23.7 per million) is 2.3× the national average. This directly supports the FIRE-WUI proposal's emphasis on vulnerable communities — the Southeast has both high wildfire exposure AND high fire mortality, suggesting systemic gaps in community preparedness that AI-driven vulnerability assessment could address.

🏠 Structure Damage — The WUI Crisis in Numbers

1.77M

Residential Structures Destroyed

85,914

Commercial Structures Destroyed

177,733

Residential Structures Damaged

28.9M

Structures Threatened

Scale of Destruction: The ICS-209 sitrep data reveals that 1.77 million residential structures have been destroyed by wildfires since 1999, with an additional 28.9 million structures threatened. This 16:1 threatened-to-destroyed ratio suggests that early intervention and risk-informed evacuation could dramatically reduce losses — a core thesis of FIRE-WUI research.

🧮 Proposed Analytical Models

1. Community Wildfire Vulnerability Index (CWVI)

A composite index combining physical exposure, social vulnerability, and adaptive capacity to score WUI communities on a 0–100 scale.

CWVI_i = w₁·E_i + w₂·S_i + w₃·A_i

Where:
E_i = Exposure Score = f(fire_history, WUI_zone_pct, fuel_model, ignition_density)
S_i = Sensitivity Score = f(%_over_65, %_poverty, %_mobile_homes, %_no_vehicle, social_vulnerability_index)
A_i = Adaptive Capacity = f(%_prescribed_fire, fire_station_density, evacuation_routes, community_rating, insurance_coverage)

Weights (w₁, w₂, w₃) determined through principal component analysis of ICS-209 damage outcomes vs. community characteristics.

Data Support: ICS-209 county-level incident links (48,370 records) enable direct correlation of wildfire exposure with community outcomes. Alabama's 23.7/M fire death rate provides a high-signal test case for validating the sensitivity component.

2. Wildfire Risk Attribution Model

Quantifies the relative contribution of ignition source, weather, fuel, and terrain to fire outcome severity.

Risk_outcome = β₁·P(ignition|cause) + β₂·P(spread|weather,fuel) + β₃·P(damage|WUI_density,evacuation) + ε

Calibrated using 34,622 ICS-209 incidents with known cause codes (H/L/U/O) and 182,826 sitreps with structure damage data.

Outcome variable: structures_destroyed or acres_burned. Key insight from data: Lightning fires (L) are 4.5× larger but human-caused fires (H) are 1.5× more frequent in WUI zones.

3. Mega-Fire Probability Model (Exponential Tail)

The fire size distribution follows a power law. Modeling the tail enables probability estimation for catastrophic events.

P(A > a) = (a_min/a)^α where α ≈ 1.05–1.15 for US wildfires

For WUI communities, the expected annual loss:
E[Loss] = ∫_{a_min}^∞ P(A > a) · V(a) · D · da

Where V(a) = vulnerability at fire size a, D = structure density, integrated over the power-law tail.

From our data: Class G fires (5,000+ acres) account for only 10% of incidents but 80.7% of total acreage — confirming the heavy-tailed distribution that makes mean-based risk metrics dangerously misleading for WUI communities.

4. Spatiotemporal Fire Cluster Detection

Identify emerging fire clusters and predict WUI impact using space-time scan statistics.

LLR = log[ (P(D_in|Z)/P(D_out|Z)) / (P(D_in|Z₀)/P(D_out|Z₀)) ]

Where Z = spatiotemporal cylinder (radius r, time window t), D = fire detections from VIIRS, and Z₀ = null hypothesis of uniform distribution.

VIIRS fire detection data (10,154 active points, updated weekly) provides real-time input. ICS-209 county/tract/CBG geographies enable community-level aggregation.

5. Prescribed Fire Effectiveness Index

Quantifies the protective effect of prescribed burning on WUI community outcomes.

PFEI_c = 1 - (D_burned,c / D_expected,c)

D_expected,c = baseline damage rate for community c based on exposure and sensitivity
D_burned,c = observed damage rate for communities with active prescribed fire programs

PFEI → 1: prescribed fire nearly eliminates excess WUI damage
PFEI → 0: prescribed fire shows no protective effect

Fort Rucker's 280K+ acre prescribed fire program provides a natural experiment: compare WUI outcomes in areas near vs. far from military prescribed fire zones.

Fort Rucker Natural Experiment: The installation's prescribed fire program covers vast longleaf pine acreage, creating a natural treatment/control comparison for measuring prescribed fire effectiveness at reducing WUI community vulnerability — a rare empirical opportunity that strengthens the FIRE-WUI proposal significantly.

🗂️ Data Assets & Infrastructure

Federated Fire Database (MySQL)

Dataset	Records	Key Variables	Source
ICS-209 WF Incidents	34,622	Cause, acres, fatalities, location	USFS/ICS-209 PLUS
ICS-209 WF SitReps	182,826	Structures destroyed/damaged/threatened	USFS/ICS-209 PLUS
ICS-209 WF County Links	40,688	GEOID, county, quarter, year	USFS/ICS-209 PLUS
ICS-209 WF Tract Links	45,245	Census tract, spatial origin	USFS/ICS-209 PLUS
ICS-209 WF CBG Links	48,370	Census block group, spatial origin	USFS/ICS-209 PLUS
ICS-209 AH SitReps	187,115	All-hazard situation reports	USFS/ICS-209 PLUS
ICS-209 Complex Assoc	4,764	Complex fire membership	USFS/ICS-209 PLUS
VIIRS Fire Detections	10,154	Lat/lon, FRP, confidence, day/night	NASA FIRMS
USFA Fire Loss Stats	69	State deaths, national trends	USFA 2023
Community Preparedness	17	FEMA community ratings	FEMA
FPA FOD (2.66M records) — pending CSV download
SILVIS WUI — pending shapefile processing

Computational Infrastructure

MySQL Database: 570K+ fire records on dedicated server (ewell-svr1), federated with SNI database for community data
FireGlobe Visualization: Three.js globe with real-time VIIRS fire detection layer (live dashboard)
Python Analysis Pipeline: Automated data loading, cleaning, and statistical analysis scripts
Azure Cloud (Planned): Scalable compute for ML model training, WUI shapefile processing, and real-time API endpoints
Community Partner: Fort Rucker Natural Resources Branch (prescribed fire data, INRMP integration)

🎯 NSF FIRE-WUI Proposal Alignment

How This Data Supports Each Proposal Section

Proposal Section	Data Support	Key Stat
Problem Statement	Mega-fire trend: avg size 2× since 1999	8,080 avg acres in 2020
WUI Vulnerability	1.77M structures destroyed, 28.9M threatened	16:1 threatened/destroyed ratio
Southeast Gap	AL fire death rate 2.3× national avg	23.7 deaths/M pop
Community Partner	Fort Rucker prescribed fire = natural experiment	280K+ acres treated
Power-Law Risk	10% of fires → 80.7% of acreage	Heavy-tailed distribution
Methodology	5 analytical models calibrated on ICS-209	34K+ incidents
Broader Impacts	SE states disproportionately affected	6,000+ county-incident links
Data Management	Federated MySQL + real-time VIIRS	570K+ records

Top 20 Largest Wildfires (ICS-209 Record)

#	Fire Name	Year	Acres	Cause
1	Taylor Complex	2004	1,305,592	Lightning
2	August Complex	2020	1,032,648	Unknown
3	East Amarillo Complex	2006	907,245	—
4	NW Oklahoma Complex	2017	779,292	Unknown
5	OKS-Starbuck	2017	662,700	Unknown
6	Murphy Complex	2007	652,016	Lightning
7	Railbelt Complex	2009	631,194	Lightning
8	Eagle Complex	2004	614,974	Lightning
9	Long Draw	2012	557,628	Lightning
10	Wallow	2011	538,049	Unknown

Source: ICS-209 PLUS Wildfire Incident Database, 1999–2020 (USFS RDS)

🔬 Next Steps & Data Gaps

FPA FOD Integration: 2.66M wildfire records (1992–2020) pending CSV download — will enable county-level trend analysis across 30 years
SILVIS WUI Shapefiles: Census-tract level wildland-urban interface boundaries needed for spatial vulnerability scoring
Census Demographics: ACS 5-year estimates for social vulnerability variables (% poverty, % over 65, housing type)
EPA AQS Integration: Air quality impact correlation with wildfire events (smoke exposure → health outcomes)
Model Calibration: Train CWVI and risk attribution models on ICS-209 outcomes, validate on held-out years
Fort Rucker MOU: Formalize community partnership for prescribed fire effectiveness study
Real-Time Pipeline: Automate VIIRS → MySQL → FireGlobe visualization for live wildfire monitoring