How Fast Does a Car-Deck Fire Spread?

A single vehicle fire on a car deck does not stay single. Full-scale calorimeter tests put fire spread to the adjacent car at roughly 20 minutes, with two cars fully involved by about 29 — and vehicles on a car deck sit far closer than the test spacing. That spread time is the detection-lead budget: flag the first vehicle inside it, or fight several across several decks.

The spread clock, from the fire-test data

Peer-reviewed full-scale car-fire tests give the timeline directly. A single burning passenger car peaks at roughly 1.5–6 MW depending on model and fuel load, and takes on the order of 10–14 minutes to reach that peak. Left next to a neighbour, the fire spreads: large-scale calorimeter work measured ignition of a second car at about 20 minutes, with both cars fully developed near 29 minutes, and two-to-three cars burning together reaching around 8 MW. Those are the numbers that bound how long a deck fire stays a one-vehicle problem.

Why a car deck is faster than the test

The laboratory spacing is generous next to a real stow. Ro-ro and PCTC loading packs vehicles with roughly 10–30 cm between them to maximise the car-equivalent-unit count, while the tests that measured a ~20-minute car-to-car spread used wider parking-lot gaps. Closer spacing shortens the radiant-ignition time to the neighbour, and an enclosed deck traps the heat and smoke that an open car park vents away. Treat the ~20-minute figure as an upper bound on the budget, not the deck reality — the real vehicle-to-vehicle window on a packed deck is shorter, and the tighter the stow, the shorter it gets.

The EV compression

An EV in thermal runaway compresses the budget further. A cell going into runaway vents and jets burning electrolyte, so it can ignite an adjacent vehicle by direct flame impingement rather than the slower radiant build-up that sets the ~20-minute figure for ICE cars, and it re-ignites after a knockdown because the reaction is self-heating from inside the pack. The clean radiant-spread picture from conventional car-fire tests is the optimistic case; a jetting lithium-ion pack on a tightly stowed deck can beat it. The spread clock a detection layer has to race is therefore shortest exactly where the cargo is battery-electric.

Spread time is the detection-lead budget

This reframes what detection has to deliver. The task is not to detect the fire — by the time a deck-level smoke panel trips, spread has usually already begun. The task is to flag the originating vehicle before the second one ignites, inside that single-car window. A per-vehicle localisation that names the unit early turns a one-car problem into a one-car response; a deck-level alarm that reports only 'smoke on deck 8' after the fire has moved is the start of a multi-car, multi-deck fight. The useful measure of a detection layer is its lead over the spread clock, not its raw sensitivity in isolation.

What the Höegh Xiamen timeline shows

The Höegh Xiamen is the case study for spending the budget blind. Per the NTSB Marine Accident Report (MAR-21/04, Jacksonville, 4 June 2020), the fire started in the aft of deck 8 — loaded with used vehicles — and spread to decks 7, 9 and 10/11; the ship and its 2,420 vehicles were a roughly $40 million total loss, and the fire burned for eight days. The NTSB found the operator had no procedures to minimise the time the fixed fire-detection system was deactivated, so detection was effectively down during loading. The early window the spread clock allows was never used. The failure was not a sensor that missed a fire; it was a budget spent before anyone was watching the deck.

Designing to the budget

For a class engineer or operator, the number that matters is the layer's lead over the spread clock, and two properties produce it. First, per-vehicle localisation: the alarm has to name the originating unit in minutes, not consume the window while a patrol walks the deck to find the source. Second, continuous coverage through loading and discharge, when vehicles are moving, hot and being connected, and when the Höegh Xiamen shows the detection system is most likely to be off. A detection layer that flags the first vehicle inside the single-car window, and stays live when the risk is highest, is what converts the spread-time budget from a countdown to a total loss into a window for action.

Sources

• 1. Peer-reviewed full-scale car-fire calorimeter tests — e.g. Energies (MDPI), 2019, 12(8):1465, 'Experimental Study on the Fire-Spreading Characteristics and Heat Release Rates of Burning Vehicles Using a Large-Scale Calorimeter', and related full-scale car-fire studies: single-car peak HRR ~1.5–6 MW, spread to the adjacent car ~20 min, two cars fully developed ~29 min, two–three cars ~8 MW.
• 2. Fire Technology, 2023 — 'Design Fire Methodology for Vehicle Spaces Onboard Ships': representative HRR curves for passenger cars and HGVs and FDS fire-propagation modelling on a ro-ro vehicle deck (cited for the ro-ro design-fire context).
• 3. EMSA — FIRESAFE and FIRESAFE II studies on ro-ro fire safety and detection on open ro-ro and weather decks: early detection and rapid response identified as the decisive factors in vehicle-space fires.
• 4. NTSB — Marine Accident Report MAR-21/04, 'Fire aboard Roll-on/Roll-off Vehicle Carrier Höegh Xiamen' (Jacksonville, 4 June 2020): origin on deck 8, spread to decks 7, 9 and 10/11; 2,420 vehicles, ~$40M total loss, eight-day burn; no procedures to minimise fire-detection-system deactivation time.
• [VERIFY: the car-to-car spread time (~20 min), two-car full-development time (~29 min) and HRR peaks (~1.5–6 MW single / ~8 MW two–three) are drawn from peer-reviewed full-scale car-fire tests via search summaries; confirm the exact figures against the open Energies 12(8):1465 text before publish, and note the test spacing is wider than car-deck stow.]