Open Road Risk
  • Home
  • Start Here
  • Project
    • Project overview
    • Current model status
    • AI-assisted development
  • Literature
    • Literature overview
    • Literature evidence register
    • AI-assisted literature review
    • Literature-pipeline alignment
    • Crash frequency models
    • Exposure and traffic volume
    • Spatial methods and network risk
    • Junctions and conflict structure
    • Severity modelling
    • Validation and metrics
    • Transferability and open data limits
  • Data Sources
    • Overview
    • STATS19 Collisions
    • OS Open Roads
    • AADF Traffic Counts
    • WebTRIS Sensors
    • Network Model GDB
    • OS Terrain 50 (grade)
    • Deprivation (IoD 2025)
  • Methodology
    • Methodology Overview
    • Joining the Datasets
    • Feature Engineering
    • Empirical Bayes Shrinkage
  • Exploratory Data Analysis
    • Collision EDA
    • Collision-Exposure Behaviour
    • Vehicle Mix Analysis
    • Road Curvature
    • Months and Days of Week
    • Traffic Volume EDA
    • OSM Coverage
  • Models
    • Modelling Approach
    • Stage 1a: Traffic Volume
    • Stage 1b: Time-Zone Profiles
    • Stage 2: Collision Risk Model
    • Facility Family Split
    • Model Inventory
  • Investigations
    • Investigations overview
    • KSI atlas diagnostic
    • Staffordshire data quality
    • Temporal descriptors evaluation
    • AADF counted-only filter
    • Rank stability harness
    • Zero-calibration diagnostic
  • Outputs
    • Key figures
    • Top-risk map
    • QGIS GeoPackage (Kaggle)
  • Tools
    • ukgeo — UK Geocoder
  • Future Work

Road risk at a glance

Plain-language figures summarising where collision risk is highest across the Open Road Risk study area.
Modified

July 7, 2026

Last updated: 2026-07-07 · Full GB output rebuild: 2026-07-04.

Figures summarising the main outputs of the risk model, without the modelling detail. The map is exposure-adjusted: it reflects risk given traffic, not simply where the most collisions happen. Full methodology and diagnostics are on the Stage 2 model page.

For GIS users, the full link-level QGIS-ready GeoPackage is distributed through the public Open Road Risk GB link risk + exposure GIS Kaggle Dataset. It contains one row per scored OS Open Roads link with geometry, estimated traffic exposure, observed injury-collision summaries, and modelled risk ranking fields. Treat it as a screening/research layer, not causal proof or an engineering audit.

The modelled-risk map is not the same as a simple collision-rate map. Traffic volume is a major driver of collision frequency, so modelled risk can still resemble the traffic-volume surface. The diagnostic maps below add two comparators: a crude rate map showing collisions per unit exposure, and an observed-vs-expected map showing where collision occurrence is higher or lower than expected after accounting for exposure and the available road/network context.

The map defaults to equal-size 10 km grid cells across Great Britain. It also includes a simplified GB-wide named reporting-area layer for optional display context. Those named areas are dissolved for legibility only: they are not formal modelling units, validation units, or administrative reporting units. Public-facing values are shown as percentiles or medians rather than calibrated expected collision counts.

Reporting-area map geography

The full GB key-figures map now has a GB-wide reporting-area layer. This is a presentation layer built from local-authority / council-area boundaries and dissolved into larger reporting areas where useful.

This layer is still being refined. The main comparison geography remains the 10 km grid, because it is consistent across Great Britain. The named-area view is intended for readability, not as the primary modelling unit.

Traffic volume

The map below uses the same geography and controls, but colours areas by median estimated AADT rather than modelled risk.

Crude collision rate per million vehicle-km

This is a transparent baseline: observed injury collisions divided by estimated vehicle-km travelled. It is useful because it shows collisions per unit exposure, but it can be noisy where exposure is low.

Higher/lower collision occurrence than expected from the exposure-adjusted model

This map compares observed collisions with model-expected collisions on the same public geography. It is the clearest diagnostic view of where collision occurrence is higher or lower than expected after accounting for exposure and the available road/network context. It should not be read as “true risk”.

Crude rates and observed/expected ratios can be unstable in low-exposure or low-count areas, so these maps should be read as diagnostics rather than definitive site rankings.

Diagnostic correlations

The table below is calculated on the 10 km grid cells shown above. Aggregated traffic volume is represented by median estimated AADT, matching the traffic map.

The 10 km grid is the primary consistent public geography. The named-area view uses a simplified GB-wide display geography generated by scripts/build_reporting_areas_gb.py; compact urban and county-style authorities are dissolved where that makes the map legible. The grid view omits cells with fewer than 20 scored links.

Open Road Risk

 

Built with Quarto