Methodology
Four layers, in order
The method is layered deliberately: a global framework first, then the European compliance layer, then the national rules that apply where the tunnel is, and finally the data sources that feed it. Each layer sits on the one above it.
- 1
Global framework
PIARC — World Road Association
The methodological foundation for road tunnel risk assessment: scenario-based analysis, barrier modelling, and the separation of frequency from consequence. PIARC is international, which is why the model is not tied to any one country's practice.
- — PIARC Road Tunnels Manual
- — PIARC quantitative risk analysis and evaluation reports
- — PIARC/OECD dangerous-goods quantitative risk model
- 2
European compliance layer
EU Directive 2004/54/EC
The minimum safety requirements for tunnels in the trans-European road network, and the reference most national regulations in Europe are written against. Compliance is evaluated requirement by requirement.
- 3
National regulations
The rules that apply where the tunnel is
National regulations and technical standards are added as their own layer on top of the directive, per country. They are never mixed into the global framework — a national threshold is not a method.
- — Norway — see the country page
- — Further countries on request
- 4
Data layer
National road databases
Where a country maintains a national road database, tunnel geometry, traffic and equipment can be imported from it rather than typed in. Norway's NVDB is the first supported source; the model does not depend on any single one.
Quantitative and qualitative — both, and labelled
Tunnel risk work is not one or the other. Some questions have enough data behind them to be counted; others have to be judged by people who know tunnels. A model that only does the first is quietly narrow. A model that does both without saying which is which is worse.
Quantitative
Where the data carries a number, the model calculates one.
- Scenario frequencies and consequences computed from registered tunnel, traffic and equipment data.
- Barrier effects applied per scenario, so a measure is credited where it actually works.
- Sensitivity and uncertainty carried alongside the result, not stripped off it.
- Comparable across a portfolio, because the same method produced every figure.
Qualitative
Where a number would be false precision, the model asks for judgement — and structures it.
- Risk-indicating characteristics assessed as their own question: indicated, ruled out, or must be assessed — never averaged into a score.
- Expert and owner judgement recorded as a decision with its reasoning, not as an anonymous input.
- Hazards that are real but not countable are carried as findings rather than dropped because they resist quantification.
- Some requirements are organisational rather than physical. They are evaluated as such.
Every value carries the class of evidence behind it: verified source, owner estimate, expert judgement, or not quantifiable. That label travels with the number into the report. It is what lets a reader tell a measurement from an assessment — and it is the difference between a model you can defend and one you can only present.
How a risk score is composed
Every score can be opened up: which inputs drove it, which barriers reduced it, which values are registered and which are assumed. A number you cannot take apart is a number you cannot defend.
The composition below is the shape of the calculation, not a specification of it. The values behind each term, and the sources they come from, are shown in the product for every result.
risk = exposure × frequency × consequence × barrier effect × uncertainty