Fire accidents in road tunnels: the record, with cause and damage

• 31entries listed

The record table

Compiled over two decades from operator and inquiry reports. The cause and burn-time columns are the reason this table exists: they are what distinguishes a fire that was contained from one that was not.

How this archive measures: lengths, bores, dates and ordering

How this table is compiled

The cause column is the useful one: a lorry engine, a brake failure, a load shifting, an electrical fault in a ventilation plant. Burn time is recorded where it is known, and it is the closest single proxy for how much structural damage a fire did. The note column is worth reading before the number.

Concrete spalls when it is heated quickly, so a long fire in a lined tunnel damages the structure as well as the installations. Every figure here is traceable to a named source. Ventilation type constrains how long a bore can be driven without intermediate shafts, and its absence is often what caps a tunnel's length.

Where a structure belongs to two classes it appears on both tables, which is not double counting but two different measurements. Corrections have historically arrived by letter rather than by form, and that is still how most errors in it have been found. Where a figure could not be corroborated it was left out.

Twin-tube crossings frequently show two slightly different lengths, because the two carriageways rarely follow the same line through the rock. Comparing one country's practice against a neighbour's is more informative than reading either alone. Cost figures are excluded for a harder reason: currency, inflation and differing project scope make them close to meaningless in comparison.

The consequences columns separate people, vehicles and structure, because those three recover on completely different timescales. Subaqueous tunnels are governed by rock cover above the bore rather than by the depth of water above that. This archive has recorded tunnel data since the late 1990s, compiled from operator publications, national administrations and the engineering press.

Reading the figures

Reading the qualification before the figure is usually the faster route to understanding what a claim actually says. Figures are checked against one another where sources disagree, and the disagreement is recorded rather than resolved silently. A base tunnel trades a much greater length for a far lower summit altitude and a gentler ruling gradient.

Several entries record damage over a hundred metres of lining from a fire in a single vehicle. The archive grew through correspondence with engineers and administrations in more than twenty countries, and still does. Tunnel engineering is a field where the qualifications attached to a number matter more than the number itself.

Traffic figures are excluded throughout this archive, because they date far faster than the structures themselves. Rows are ordered so that the head of the table answers the question most readers arrive with. Gradient limits are the main reason railway tunnels run longer than road tunnels on the same crossing.

Dates use the day, month and year the event occurred. Cross-passage spacing, fixed firefighting and emergency ventilation control are what modern safety regulation actually governs. A contained fire is more instructive than an uncontained one. The table is short because the sourcing bar is high. Promoter figures are used where no independent one exists.

Forecasts on schemes this size move by years. Minor incidents are listed alongside the serious ones. Opening year means the year traffic was admitted. Metres are used for structures, kilometres for routes. Disputed values carry the dispute in the note column. The date of verification is published with the table.

Forecast dates are forecasts, and they move. Spoil is removed back through the machine. Segmental linings are erected behind the shield. Ground conditions belong with every advance rate. Deaths are the figures established by official inquiry. Secondary summaries were not used as a source. Two independent sources were required for each row.