Risk Assessment Methodology

All risk calculations are grounded in publicly available, authoritative data sources. The table below lists every data source used, its refresh cadence, and what it provides.

The overall risk score for each jurisdiction is calculated using the Public Health Risk Assessment Tool (PHRAT) quadratic mean formula. This approach prevents a high risk score in one domain from being diluted by low scores in others.

Why quadratic mean? Unlike a simple weighted average, the quadratic mean (root mean square) gives greater emphasis to higher-risk domains. A jurisdiction with one domain at 0.9 risk will score meaningfully higher than one where all domains sit at 0.4, even if the weighted averages would be similar.

Individual risk domains are calculated using the Exposure-Vulnerability-Resilience (EVR) framework. This formula is applied consistently across natural hazards, utilities, and extreme heat domains.

Risk Level Categories

The natural hazards score is the equal-weighted average of four hazard types, each calculated using the EVR framework described above.

EVR Components by Hazard

• Exposure: FEMA NRI hazard scores with climate change projection multipliers from data/climate/natural_hazard_climate_projections.json
• Vulnerability: All 4 CDC SVI themes with hazard-specific calibrated weights, plus Census ACS demographics (population aged 65+ (%), mobile home %, population density)
• Resilience: Socioeconomic capacity and infrastructure factors

SVI Vulnerability Weights by Hazard Type

Climate Change Projections

Wisconsin is divided into 3 climate zones (Southern, Central, Northern) with RCP4.5/SSP2-4.5 scenario projections applied as exposure multipliers:

Health metrics risk is based on respiratory disease surveillance data from Wisconsin DHS and vaccination coverage from County Health Rankings.

Calculation Method

• Disease activity: Risk scores are calculated from cases_per_100k using sigmoid normalization, producing a 0–1 score that responds proportionally at low case counts and saturates at high counts.
• Vaccination coverage: Uses a strategic framework weighting flu vaccination (30%), COVID vaccination (30%), MMR (25%), and pediatric vaccination (15%). Lower coverage yields higher risk.

Wisconsin DHS PDFs are extracted weekly via pypdf, providing statewide activity levels that inform county-level risk estimates.

Vector-borne disease risk covers Lyme disease and West Nile Virus (WNV) using real county-level incidence data from Wisconsin DHS, supplemented by environmental and climate range expansion factors.

Data Refresh

WI DHS EPHT CSV files (lyme-county.csv, west-nile-data-county.csv) are downloaded automatically each week by the background scheduler. These files cover all 72 Wisconsin counties with confirmed and probable case counts sourced from WEDSS (Wisconsin Electronic Disease Surveillance System) as published on the DHS Environmental Public Health Tracking portal. All 72 counties are represented; counties with zero reported cases receive a rate of 0.0 rather than a fallback estimate.

The active shooter domain evaluates population-level environmental and social indicators correlated with elevated risk conditions. This framework is intended for public health preparedness planning, not individual threat assessment.

Data Sources: Gun Violence Archive (GVA), FBI Crime Data API, NCES SSOCS 2019-2020, Census ACS, CDC SVI.

Extreme heat risk combines the EVR residual risk calculation with wet bulb temperature and climate trend adjustments.

Five Components

1. Exposure: Historical heat event frequency and geographic heat island factors
2. Vulnerability: CDC SVI themes, population aged 65+ (%), outdoor worker density
3. Resilience: Cooling center access, healthcare capacity, community preparedness
4. Wet Bulb Temperature: Humidity-adjusted temperature danger thresholds
5. Climate Trend: Long-term warming trajectory for the jurisdiction

Climate Adjustments (NOAA/IPCC)

• 40% increase in heat event frequency by 2050 (RCP4.5)
• 2–4°F warming projected for Wisconsin by 2050
• 60% increase in heat wave duration (IPCC AR6)

Data Sources: NOAA climate summaries, EPA heat island reports, CDC SVI, WICCI 2021 Assessment.

Four components are assessed, each using the EVR framework:

Exposure, vulnerability, and resilience for each component are modeled based on county infrastructure characteristics, population density, geographic factors, and SVI vulnerability themes.

The cybersecurity risk score models the relative exposure of public health infrastructure to cyber threats based on three components:

• Threat Landscape (35%): County population size and institutional presence as proxy for target visibility
• Vulnerability (40%): County fiscal resources as proxy for IT infrastructure age, adjusted by CDC SVI socioeconomic factor (up to 25% increase for more vulnerable jurisdictions)
• Capability (25%): County staffing and urbanization as proxy for organizational security readiness (inverted: higher capability = lower risk)

This domain is included to support comprehensive preparedness planning. Jurisdictions seeking detailed cyber risk assessments should consult CISA resources and conduct dedicated cybersecurity evaluations.

Air quality risk combines current EPA AirNow data with strategic climate projections to assess long-term environmental health planning needs.

Components

• Current AQI Baseline: EPA AirNow API data, retrieved via background scheduler and served from cache
• Climate Projections: 40% ozone increase and 110% wildfire smoke increase projected by 2050
• Strategic Assessment: 5-year historical baseline patterns integrated with forward-looking projections

Dam failure risk is a standalone EVR domain assessing the probability and downstream consequence of dam failure for each Wisconsin county.

EVR Components

• Exposure: Dam count, hazard class (High/Significant/Low per USACE/WI DNR classification), and dam condition ratings for dams within or upstream of the jurisdiction
• Vulnerability: Downstream population exposure within modeled inundation zones; population density (Census ACS); CDC SVI housing/transportation theme
• Resilience: Emergency Action Plan (EAP) availability, dam inspection recency, and watershed management capacity

Data Sources

The CDC/ATSDR Social Vulnerability Index is integrated throughout CARA as a core component of vulnerability assessment.

Four SVI Themes

How SVI Is Used

• Natural Hazards: Integrated directly into the EVR vulnerability component with hazard-specific calibrated weights (see Natural Hazards section)
• Air Quality: SVI housing/transportation and socioeconomic themes applied as a combined multiplier (capped at 1.5x) to the base score before PHRAT
• Dam Failure: SVI housing/transportation theme applied as a multiplier (up to 1.2x) to the base EVR score before PHRAT
• Vector-Borne Disease: SVI socioeconomic theme applied as a multiplier (up to 1.15x) to the base EVR score before PHRAT
• Active Shooter and Extreme Heat: SVI themes integrated as weighted inputs within sub-component calculations

SVI data is sourced from the CDC/ATSDR SVI API with a 365-day cache, providing county-level percentile rankings for each theme.

CARA provides a Baseline-Seasonal-Trend-Acute (BSTA) temporal framework as a complementary analytical lens for each risk domain. BSTA decomposes domain risk into four time-horizon components to support strategic planning context. It is displayed alongside the primary PHRAT score but does not modify the PHRAT calculation directly. Temporal factors such as long-term storm event trends, climate projections, and seasonal patterns are embedded within each domain's underlying data inputs. The BSTA display shows how those temporal dimensions break down for each hazard type.

In strategic planning mode, long-term structural risks receive the highest emphasis.

Seasonal Patterns by Hazard

CARA aligns with the CDC's Public Health Emergency Preparedness (PHEP) Cooperative Agreement 2024-2028. All risk assessment results, action plans, and preparedness recommendations support the following PHEP capabilities:

PHEP Capability Implementation in Risk Assessment and Action Planning

For each risk type, specific PHEP capabilities are identified and integrated into a three-phase implementation approach:

CARA supports all 11 federally-recognized tribal nations in Wisconsin as distinct health jurisdictions, each with its own dashboard, risk scores, and action plans. Because tribal land bases span multiple counties and include both reservation land and off-reservation trust land, a simple single-county mapping would produce inaccurate risk scores. CARA uses a geometric area-weighted approach to resolve this.

County Weight Computation

County weights for each tribal jurisdiction are derived from a spatial area overlay (using GeoPandas in EPSG:3174, Great Lakes Albers equal-area projection) between two authoritative GeoJSON datasets:

• Wisconsin tribal boundary polygons sourced from the U.S. Census Bureau TIGER/Line files, which include both reservation polygons and off-reservation trust land polygons as separate features.
• Wisconsin county boundary polygons (72 counties) from the same TIGER/Line source.

For each tribe, the intersection area between tribal features and each county polygon is computed. Weights are assigned proportionally to intersection area, normalized so all county weights for a tribe sum to 1.0. Counties that contribute less than 0.5 percent of a tribe's total land area are excluded to avoid incorporating boundary slivers or mapping artifacts.

Trust Land Inclusion

Off-reservation trust lands are included in the area-weighted calculation. This has a material effect on several jurisdictions:

• Ho-Chunk Nation: approximately 78.5 percent of total tribal land area is off-reservation trust land distributed across 12 counties (Adams, Clark, Crawford, Eau Claire, Jackson, Juneau, Marathon, Monroe, Shawano, Trempealeau, Vernon, Wood). Omitting trust lands would have reduced the county count from 13 to 5 and would have systematically overweighted Sauk County (which contains the main reservation) to 74 percent.
• Forest County Potawatomi Community: approximately 15.9 percent trust land, including parcels in Fond du Lac and Milwaukee counties that are absent from reservation-only mappings.
• St. Croix Chippewa Indians of Wisconsin: approximately 16.2 percent trust land, concentrated in Burnett County and adding a small Polk County component.
• Bad River Band: trust land in Forest County (1.1 percent of total area) adds to the reservation's primary Ashland County base.

Tribes with no trust land features in the source GeoJSON (Lac du Flambeau, Sokaogon) are represented by reservation-only weights.

Risk Score Aggregation

For each risk domain, county-level risk scores are multiplied by the corresponding county area weight and summed to produce the tribal composite score. This is equivalent to a weighted mean where the weights reflect the geographic distribution of the tribe's land base.

Known Limitations