Repeated tests showed consistent model behaviour: sharper curvature, smaller radius, lower friction, reduced visibility, and slower response assumptions generally increased the comparative output.
The Results page summarises saved road-analysis cases from the Road Risk model. These records show road geometry, scenario settings, derived physics values, comparative outputs, and confidence information.
These are saved records from the model. They show how selected roads behaved under the chosen assumptions. They are not official safety ratings or crash predictions.
This dashboard combines reviewed public cases loaded from the committed site dataset with local cases saved explicitly in this browser. Local browser cases are stored only on this device/browser. They will not appear on another phone, computer, or browser unless exported, reviewed, and committed to the public dataset.
Charts update with the active filters and describe saved model-output cases only. They support interpretation of the saved dataset; they do not establish observed collision risk.
Graphing visible filtered cases.
No graph rendered yet.
These cards define the exact fields a documented result should contain. They are shown only as guidance when no cases are loaded. Real public cases must come from exported app JSON that has been reviewed and committed to the site dataset.
No measured case data has been committed yet. This slot is ready for a CSV/JSON/GeoJSON-backed selected-road example.
This slot should compare the same road under two defined assumption sets, such as dry baseline versus rain, fog, or fatigue.
This slot should include route distance, duration, average output, highest segment, hotspot count, and supporting route export data.
| Variable | Category | Interpretation Role | Boundary |
|---|---|---|---|
| Heading change, radius, and curvature | Derived geometry | Describe how sharply the selected road changes direction. | Sparse or noisy public geometry can reduce precision. |
| Safe speed | Physics-informed output | Shows a friction-limited curve-speed check under the active assumptions. | Not a posted speed limit or legal recommendation. |
| Stopping distance | Physics-informed output | Connects speed, reaction time, friction, and scenario settings. | Simplifies gradients, tyres, ABS, surface wear, and visibility obstructions. |
| Annualised Comparative Model Output and Daily Interpretive View | Comparative model output | Provide a consistent reporting basis for comparing roads and scenarios. | Not observed crash rates. |
| Percentile position and tail markers | Comparative statistics | Place one road inside a sampled network distribution. | Only meaningful within the sampled comparison set. |
| Scenario Factor | Expected Model Direction | Primary Mechanism | What Not to Conclude |
|---|---|---|---|
| Rain / wet surface | Usually higher output | Lower effective friction, reduced safe-speed margin, longer stopping distance. | It does not prove a specific wet-road crash risk. |
| Fog / reduced visibility | Higher interpretive concern | Visibility and reaction context become more restrictive. | It does not measure actual visibility at the site. |
| Fatigue / distraction / BAC | Usually higher output | Longer reaction component and higher behavioural multiplier. | It does not observe any real driver. |
| Overspeed | Higher output, often non-linear | Higher lateral acceleration and velocity-squared stopping-distance demand. | It is a scenario test, not evidence of actual speeding. |
| Vehicle type | Changes stopping and handling interpretation | Vehicle profile assumptions affect braking, mass/context, and visibility interpretation. | It is not a full vehicle-dynamics simulation. |
The route mean provides a compact summary of sampled model output along a route. It is useful for comparison, but it can obscure a single demanding bend or junction context.
The highest sampled route segment is often more useful for review because it identifies where local geometry or scenario demand is concentrated.
The current results are not calibrated against complete national collision-history datasets.
A high output may indicate a segment worth closer review, but it does not certify that the road is dangerous.
A low output does not prove a road is safe; it only means the selected model context produced a lower comparative value.
Missing OSM tags do not prove missing infrastructure. Field verification would be required for operational claims.
Use the live app's case-evidence export to create reviewed JSON records for the static public dataset.