Methodology

We classify mapped paths by surface, split the network at intersections and entrances, then traverse only firm, stable, step-free segments that do not have a measured slope blocker.

Elevation is sampled about every 5 m from USGS 3DEP through a configurable Open Topo Data service. Adjacent center samples produce running grade; perpendicular left/right samples produce cross grade. Accessible routes are limited to 5% running grade unless treated as a ramp (5–8.33%, shown as limited), and 2% cross grade. Outdoor trails use the Access Board’s 8.33%/200 ft, 10%/30 ft, and 12%/10 ft bands, a 5% cross-grade ceiling, and the 30% steep-length limit. Known widths below 36 inches are blockers; passing-space compliance remains unverified because OSM and DEM data do not establish it.

The slope component is the measured share of firm, entrance-reachable path within those thresholds. Unknown elevation contributes no compliance credit. A slope blocker breaks reachability and is pinned like a surface or step blocker.

Score weights

• Accessible path ratio: 35%
• Reachable network: 25%
• Slope: 15% (measured evidence only; unknown receives no credit)
• Entrance and parking: 10%
• Amenities: 10%
• Community adjustment: up to ±15 points

Coverage gates the tier. Sparse data remains Unverified even when known paths look good.

Community evidence

Confirm and flag confidence uses the Wilson lower bound, with 12-month half-life decay and extra emphasis on the latest 90 days. Independent mobility-device types add a bounded diversity bonus; trusted verifiers carry the highest weight. Community evidence may lower or dispute a result, but a report can raise a tier only with photo evidence and two verified visits.

AI suggestions

For untagged paths, a CPU-only aerial-imagery classifier may suggest paved or unpaved. These records are labeled AI-suggested, low trust, and inactive. They cannot affect a score until a person reviews and confirms the evidence.