A subway bore can carry a radio signal farther than free space would and still leave a crew cut off two hundred feet in. The tunnel is a waveguide with a temper, favoring some frequencies, punishing curves, and going dark the moment a train fills the cross section. The engineered answer is radiating cable, the operational answer is a plan for the portal and the entry control point, and the failure that kills people is the one where the interior crew hears everything and nobody hears them.

The tunnel as a waveguide with a hostile personality

A running tunnel is a long conducting tube with lossy walls, and above a cutoff frequency set by its cross section it supports guided propagation the way an oversized waveguide does. For a bore several meters across, that cutoff sits well down in the tens of megahertz, so VHF, UHF, and 700/800 MHz all travel along the bore as overlapping modes rather than radiating away into space. In a long straight section with smooth walls the attenuation per hundred meters can be surprisingly modest, and the higher bands frequently do better along a bore than the lower ones, which is the reverse of the intuition most of us built on the surface.

The personality problem is that tunnels are not smooth straight tubes. Every curve strips energy off the guided modes and dumps it into the wall, station boxes change the cross section abruptly, and cross passages, sump rooms, ventilation shafts, fan plants, and back of house corridors sit off the waveguide entirely. Multimode propagation also produces deep standing wave nulls, so a member who shifts three feet recovers a signal that a member standing still cannot get, and a digital system holds clean audio right to the edge of that null before dropping with no scratchy warning.

The item that surprises people most is the train. A revenue vehicle fills most of the cross section with grounded steel and aluminum, and a train stopped between your position and the source is effectively a short circuit across the guide. Preplans that assume a clear shot down a bore need to account for the fact that on a real incident there is usually a train in the tunnel, sometimes more than one, and it is parked exactly where the smoke is.

Leaky feeder and radiating cable

Because you cannot rely on the bore itself, tunnels get radiating cable, still called leaky feeder in most firehouses. It is coaxial cable with apertures cut in the outer conductor so that a controlled fraction of the energy escapes along the entire length, turning the cable into a continuous distributed antenna. Portables couple to it in the near field at a few meters, so coverage follows the cable rather than the geometry, and the cable does not care about curves.

Two design details drive everything else. Aperture design is chosen against the bands the cable has to serve, so a product engineered for one band plan is not freely re-taskable to another later, and that decision is locked in when the tunnel is built. The link budget also carries two separate loss terms, longitudinal loss along the cable itself, which rises with frequency and distance, and coupling loss between the cable and a handheld, which manufacturers publish at a stated reference distance and stated probability rather than as one number. Line amplifiers spaced along the run make up the longitudinal loss, and each one adds noise and a dependency on power.

Modern installations rarely carry one service. The same cable is commonly combined to carry transit operations radio, public safety 700/800 MHz, commercial cellular, Wi-Fi, and in some systems a train control link. That is efficient, and it also means shared fate, because work done for one occupant takes the plant away from all of them, and an intermodulation product created by one tenant’s equipment lands in everybody else’s receiver. The fire department is usually the occupant with the least visibility into change control.

How a radiating cable system fails

A radiating cable is a linear asset, which gives it failure modes a building distributed antenna system does not have. One cut, crush, or burn-through takes out everything downstream of the break. The standard mitigations are feeding the run from both ends so that a mid-run break leaves two live segments, running redundant cables on opposite sides of the bore, and specifying fire performance for the jacket and the supports. Whether any of that exists in your tunnel is a question with a specific answer that somebody in the authority’s communications shop can give you.

The environment degrades things slowly and quietly. Brake dust is conductive and coats everything, water intrudes at joints and connectors, and de-icing chemicals migrate down from the street. Traction return current and shoe or pantograph arcing raise the noise floor, worst in the lower bands. Standoff distance from the tunnel wall affects coupling performance, so hangers that loosen or a cable pushed against steel during track work will cost you coverage without anything appearing broken, and none of that trips an alarm.

The failure that matters operationally is the one inside the fire area. Cable in the immediate fire zone will eventually fail, and it will fail at the worst point in the incident, which is the argument for dual-end feed and for never treating the fixed system as your only path. Treat radiating cable the way you treat a standpipe, as an excellent asset you plan around, verify on arrival, and are prepared to work without.

Handoff at the portal

At the portal, one of three things happens. The tunnel system may be a bidirectional amplifier fed off the air from a donor site, in which case there is no handoff at all because it is the same site and the same channels, with the donor path and the amplifier’s noise contribution as the weak links. The tunnel may be its own site in your trunked system, in which case handoff is automatic when roaming, site adjacency, and signal thresholds are set correctly, and radios ping-pong at the boundary when they are not. Or the tunnel may be an entirely separate system, in which case handoff is a human being changing zones while walking into smoke.

That third case is common and it is where accountability breaks first. If the crew has to change channels at the boundary, the boundary needs to be a marked, briefed, physical place, and confirming the change belongs to the entry control officer as a recorded task rather than to each member’s memory. Automatic site hunting does not save you either, because the classic failure is a portable that clings to a marginal affiliation with the outdoor site instead of dropping onto the tunnel site.

Two other things quit at the portal. GPS stops, so automatic vehicle location, mapping, and any location-based dispatch feature go stale at the entrance. Commercial cellular stops unless the carriers are on the cable, which takes push to talk over cellular, phones, and most tablet applications out of play exactly where you wanted a backup. Below ground, plan on the assumption that the only paths are the fixed tunnel system, hardwired telephones, and people walking.

The transit authority interface

The transit operations control center holds the things that keep your people alive, including traction power removal and its confirmation, train movement authority, and the ventilation fans that decide which way the smoke goes. None of that runs on your radio system. It runs on the authority’s own radio and telephone network, often in a different band and frequently on different technology, so the interface is not optional and it is not something to invent at the incident.

The interface that works in practice has three parts. A transit liaison with a transit radio stands physically at the command post, and for anything protracted a fire officer or communications unit member goes to the rail control center. Radio caches are exchanged in advance so the battalion vehicle carries programmed transit portables and the control center holds yours. A console patch or gateway between a transit channel and an incident talkgroup exists as a preconfigured and tested option dispatch can bring up on request, understood as a bridge for coordination rather than a primary path, since a patch buys you one talk path with added latency and degraded audio.

Use the fixed plant the authority already owns. Trackside and cross-passage emergency telephones, station operator booths, tunnel telephone circuits, and the public address system are hardwired, fail for different reasons than radio does, and are usually forgotten because nobody has picked up a handset since the last drill. NFPA 130 covers fixed guideway transit and passenger rail systems and NFPA 502 covers road tunnels, both of which address emergency communications provisions, and the editions and local amendments in force should be confirmed with your fire marshal rather than assumed.

What the record shows

The King’s Cross Underground fire in London on 18 November 1987 killed 31 people, and the public inquiry report published under Desmond Fennell in 1988 documented that emergency service radio communication below ground was ineffective and that messages moved by runner. The London bombings of 7 July 2005, in which 52 victims were killed in addition to the four bombers, produced much the same finding almost two decades later. The London Assembly’s 7 July Review Committee report and the later coroner’s inquests both record that responder radios did not work underground and that command depended on people carrying messages up and down. Underground coverage for the emergency services in London was extended after that, which is the point of citing it.

The severity end of the subway fire record is worth knowing without pretending the numbers are settled. The Daegu subway fire in South Korea on 18 February 2003 is widely reported as having killed 192 people. The Baku metro fire in Azerbaijan in October 1995 is generally reported in the high 200s, and published accounts disagree on the exact toll, so cite it as a contested range rather than a single figure. In both cases the driving factors were smoke movement and egress rather than radio, and in both cases responders worked long distances from a portal in zero visibility.

Closer to home, the smoke event at L’Enfant Plaza in Washington on 12 January 2015 killed one passenger and sent dozens of people to hospitals, and the National Transportation Safety Board investigated it and issued an accident report. That report documents communication problems between the transit authority and the responding fire department, including difficulty communicating in the tunnel environment. Read the NTSB report and the docket directly rather than relying on summaries, including this one, because the interoperability findings are specific and they map cleanly onto systems in other cities.

Accountability when nobody can hear the interior crew

Underground, the asymmetry that plagues high-rise operations gets worse. The interior crew hears the platform, the control center, and command through a system transmitting with power and gain, while their five watt portable at chest height, in smoke, behind a train, may reach nobody. A member who receives clearly and assumes the path is two way will transmit a mayday, hear no acknowledgement, and key up again in the same spot. The only proof of a two way path is an acknowledgement by name, and that has to be taught long before the incident.

The response is to move accountability off the radio and onto a fixed point. Set an entry control point at the portal or the station end, run a tally of names, entry times, and cylinder pressures, and manage it with a clock rather than with roll calls, because a PAR by radio across twenty people in a bore is a request for garble. Travel distances underground are long enough that air management rather than the fire is usually the limiting factor, so the entry control officer’s timekeeping is the accountability system when the radio is not.

Then staff the gap. A relay chain of members posted at cross passages, at the platform end, and at the portal, assigned in the initial alarm instead of improvised later, is the oldest fix in the trade and the only one that reliably works when the fixed system is compromised. It costs people, which is exactly why it belongs in the assignment. Wire it where you can, using trackside telephones, the transit system’s own handsets, and hard line reels, all of which pass traffic when radio does not.

What you do before the call

Walk the tunnel with a radio. Transit authorities grant track access windows, and one night spent walking each bore with a portable and a signal meter, taking readings at the authority’s own chainage markers, produces a coverage map worth more than any vendor prediction. Note the readings against those markers, and add the trackside telephones, the standpipe connections, and any spot where a relay would have to be posted, then put all of it on the preplan officers actually carry.

Get the written answers on the fixed system while things are calm. Which bands and talkgroups the radiating cable carries, whether it is fed from both ends, what the standby power runtime is, who tests it and how often, whether the fire department receives the results, and who notifies you when a section is taken out of service for track work. Any rebanding, site relocation, or system migration you undertake also affects that tunnel system, so put it on the inventory of dependent elements alongside every in-building amplifier in the jurisdiction.

Then drill it jointly, at night, with the actual equipment, including the cache exchange, the patch, the liaison positions, the entry control point, and the relay chain. A tunnel incident is a low frequency, high consequence event for both organizations, and the parts that fail are the parts nobody has rehearsed: who calls for traction power removal and how the confirmation comes back, who decides fan direction, and who is tracking the interior crews when the radio quits.

What to do at your agency

• Ask the transit authority, in writing, which bands, talkgroups and portable to portable simplex channels the in-tunnel radiating cable actually carries, and file the reply with the pre-incident plan rather than in an inbox.
• Walk one bore on foot this month with a portable and a partner, and write down the chainage or the nearest marker where the surface system drops out and where the tunnel system picks up.
• Put the portal channel change into the next company drill, so that switching is a briefed and verified action announced aloud rather than something each member is expected to remember.
• Name the person who sits at command with a transit radio, by shift and not by title, and confirm that the position is actually filled on nights and weekends.
• Ask the transit engineering group whether the radiating cable is fed from both ends and what its fire performance rating is, and record the answer with the date you received it.
• Test the dispatch patch between your system and the transit system once this month on an ordinary weekday, and log who tested it and what failed.

Takeaways

• A tunnel bore propagates radio as a lossy oversized waveguide, so straight sections can perform well while curves, station boxes, cross passages, and a train parked in the bore kill the path entirely.
• Radiating cable, also called leaky feeder, distributes signal along the whole tunnel through apertures in the outer conductor, and its performance is set by longitudinal loss along the run plus coupling loss to the handheld.
• Because the cable is a linear asset, one cut or burn-through takes out everything downstream, which is why dual-end feed, redundant runs, and fire performance specifications matter and should be verified rather than assumed.
• A tunnel with radiating cable does not necessarily carry your frequencies, your talkgroups, or portable to portable simplex, so get that answer in writing during design review.
• At the portal, GPS and commercial cellular stop, and if the tunnel is a separate system the channel change has to be a marked, briefed, and verified action rather than something each member remembers.
• The transit operations control center holds traction power, train movement, and ventilation, so a liaison at command with a transit radio, a pre-tested patch, and an exchanged radio cache are the interface that actually works.
• King’s Cross in 1987, the London bombings in 2005, and the NTSB report on the 2015 L’Enfant Plaza smoke event all document responders unable to communicate below ground, and those primary sources are worth reading in full.
• When the interior crew cannot be heard, accountability moves to a physical entry control point with a tally and a clock, backed by a staffed relay chain assigned in the initial alarm rather than improvised.