The bombing of the World Trade Center in 1993 produced a documented record of communications failure inside a very tall building: public address that died with the power, stairwells with no light, fire department radios that could not carry from a lobby command post to companies working forty floors up, and an evacuation of both towers measured in hours. Most of those same failures appear again in the investigations of 2001. This piece, the first of four on the communications record of the response in New York, sets out what the published reviews say went wrong in 1993, what was changed afterward, and why some of those changes held under load while others did not.

The bomb, the casualties, and how the sources count them

A vehicle bomb detonated in the parking structure below the World Trade Center complex on 26 February 1993, shortly after midday. The FBI’s published history of the case puts the explosion at about seventeen minutes past noon, and most chronologies give the time as 12:18 p.m. The charge was carried in a rented truck parked on a below-grade level of the garage, and published weights for that charge differ by several hundred pounds. CNN’s reference summary describes a 1,200 pound bomb, while the Wikipedia article on the bombing gives approximately 1,500 pounds of a urea nitrate device, and fire service retrospectives tend to quote the whole range between those two figures rather than settle on one. The blast tore a crater through several parking levels and brought down the ceiling of the PATH station below the complex, and the published dimensions of that crater also differ, since the FBI describes a crater nearly 100 feet across and several stories deep while Britannica describes one six stories deep and 200 feet wide.

Even the location is reported inconsistently in general reference sources, because Britannica’s summary places the truck in a garage beneath 2 WTC and the Wikipedia article describes the device as below the North Tower, while fire service accounts place it on level B-2 of the shared underground garage. I use the garage description in this article because that is the part every source agrees on, and because nothing in the communications analysis turns on which tower footprint the parking bay sat under.

The casualty figures need the same care. The FBI states that six people were killed almost instantly, and that is the number used in the Britannica entry and in CNN’s reference summary. One of the six was a woman who was seven months pregnant, and some references, including the summary box on the Wikipedia article, state the toll as six people and one unborn child, which is where the figure of seven comes from in the accounts that use it. Injuries are given as more than a thousand by the FBI and by Britannica, while the specific figure of 1,042 appears in many places including a US Army War College vignette on the attack and the Wikipedia entry. That difference is not a rounding error, because an injury count in an event of this kind moves by hundreds according to the counting rule in use, and a total built from everyone treated and released at the scene for smoke inhalation is a different quantity from a count of hospital admissions. A reader who needs a defensible number should take it from the source whose counting rule they can state. CNN’s reference summary gives an estimated 50,000 people evacuated from the buildings, which is an estimate of building population rather than a headcount, and History’s account of the day describes an evacuation that ran through the afternoon.

The documentary record on the 1993 response is better than most people assume. The U.S. Fire Administration published a technical report on the incident through its Fire Investigations Program, titled The World Trade Center Bombing: Report and Analysis and carrying the number USFA-TR-076, prepared by TriData Corporation under contract and catalogued by NIST’s fire research library at 167 pages. The National Fire Protection Association published its own fire investigation report on the explosion and fire, authored by Isner and Klem. Two later federal bodies went back over the 1993 incident as background to 2001, namely the National Commission on Terrorist Attacks Upon the United States in its 2004 report, whose chapter on the emergency response in New York is chapter 9, and the National Institute of Standards and Technology in the NCSTAR 1 series of 2005, which includes a volume on occupant behavior, egress and emergency communication and a separate volume on the emergency response operations. A review of the FDNY response prepared for the department by McKinsey and Company was released in 2002 and belongs in the same set, and I will use it in part two rather than here. I looked for a separately published Port Authority review of its own 1993 response and could not confirm one that is publicly available, so what I say below about Port Authority actions comes from the 9/11 Commission’s account and from the federal fire reports rather than from the agency’s own document.

Power, public address, and stairwells that went dark

The 9/11 Commission’s chapter on the New York response summarizes the 1993 building failures in a few flat sentences, and they are worth reading as an equipment list rather than as narrative. The public address system and the emergency lighting systems failed. The unlit stairwells filled with smoke and, in the Commission’s words, were “so dark as to be impassable”. The NFPA investigation record describes extensive damage across several basement levels and an intense fire that pushed thick black smoke quickly into the upper levels of many of the complex’s seven buildings, which is the mechanism that turned a garage explosion into a smoke event a hundred floors up.

Take the public address failure apart, because the same architecture is sitting in buildings in your jurisdiction right now. A building communication system is an amplifier rack, a console, a distribution network of speaker circuits, and a power feed, and the announcement an occupant hears is the output of all four working at once. Put the head end in a basement equipment room, feed it from building power with a transfer arrangement that was never tested against the loss of that particular room, and the system is available for every scenario except the one where the basement is the problem. In 1993 the damage and the power loss were in the same place as a good deal of the equipment that was supposed to tell tens of thousands of people what to do, and the result recorded by the Commission is that the system did not speak.

Emergency lighting failed in the same way and for the same reason, which matters more in a stairwell than anywhere else in a high rise, because a stair with no light and rising smoke is not merely uncomfortable to descend but genuinely impassable for people who cannot see the next tread. The Commission’s language on this point is unusually blunt for that document, and it is the finding that produced the most visible physical change in the towers afterward.

How long the evacuation took is a figure I would ask you to handle carefully, because it circulates as a round number that the primary sources do not support in that form. The 9/11 Commission Report states that the general evacuation of the towers’ occupants via the stairwells took more than four hours. Reporting on the Commission’s hearings, including a CBS News piece in 2004, attributes a figure of ten hours to the Commission for the 1993 evacuation, which almost certainly counts the very end of the operation rather than the emptying of the stairwells. NIST stated the comparison as a multiple rather than a duration, writing that the escape times available on 11 September were shorter by a factor of three or four than the time needed to clear the tenant spaces of WTC 1 after the 1993 bombing, and shorter by another factor of two than the time needed to get the last person out of the elevators. Those statements are consistent with one another once you see that they are counting different endpoints, and the honest way to use them is to say that clearing the occupied floors took several hours and that finishing the elevator rescues took several hours more.

Steel floors and one crowded channel

The single most useful sentence in the 9/11 Commission’s treatment of 1993, for a communications reader, is its statement of why the fire department radios performed poorly. The Commission gives two reasons and separates them cleanly. The radio signals often failed to penetrate the many steel and concrete floors between companies that were trying to talk to each other, and so many different companies were trying to use the same point-to-point channel that, in the Commission’s phrasing, communications became unintelligible. Those are two distinct engineering problems with two different fixes, and the fact that the Commission wrote them down as a pair in 2004 tells you that the pair was understood at the time.

The first problem is propagation. A portable radio transmitting a handful of watts from a stairwell landing on the eightieth floor is trying to reach a lobby command post through a stack of composite floor slabs on metal deck, through core walls, elevator machinery, ductwork and structural steel, and through whatever the building’s own cabling adds to the mess. Every one of those layers takes signal out of the path, and the loss is not a gentle taper but a series of steep steps that vary by exactly where the operator is standing. I have written separately on this site about why buildings eat radio and what the fixes look like, so I will not repeat the physics here beyond the point that matters for 1993, which is that simplex portable to portable operation across sixty or eighty vertical floors of a steel high rise has never been a reliable design and was not one then.

The second problem is loading, and it is the one people forget when they go shopping for a fix. A single point-to-point talk path can carry one transmission at a time. Put twenty or thirty companies on it at a multiple-alarm incident spread over a hundred floors and two towers, and the channel is stepped on continuously, so that even the transmissions with enough signal to be received are not intelligible when they arrive. Coverage and capacity fail in ways that sound identical to the person holding the radio, because in both cases the answer never comes back, and that similarity is exactly why a fix aimed at one of them gets credited with solving both.

Keep that separation in mind through the rest of this article, because the improvement that came out of 1993 for radio was a repeater, and a repeater is a coverage device. It puts a better antenna and more transmit power into the middle of the path so two portables that cannot hear each other directly can both hear the repeater. It does not create a second talk path, and it does not stop thirty companies from keying over each other on the one path it does provide. Anyone who has run a fireground on a single repeated channel with a working repeater and too many units has heard exactly what the Commission described from 1993.

Climbing without a link, deciding without information

Consider what the two radio failures meant for the people doing the work. A lobby command post in a high rise exists to hold the accountability picture, to assign companies to floors, to receive reports of what those companies found, and to move resources based on those reports. All of that depends on a talk path to the upper floors that works in both directions. When the path is unreliable, the command post’s picture of the building degrades to whatever somebody physically carried back down, and companies operating above the reach of the lobby are working on the assignment they left with rather than on the situation as it developed. The published reviews I can check establish that the radios did not reach and that the channel was unusable, and they do not give me a transmission-level account of how the FDNY worked around it that afternoon, so I am not going to reconstruct one.

On the occupant side the consequence was more direct. With the public address system silent, the building had no way to tell tens of thousands of people anything at all, so the decision about whether to stay, to descend, or to go up was made by each group of occupants on whatever information they had, in a stairwell with no lighting. That is worth stating plainly, because the failure mode of an unavailable notification system shows up as a large number of people making individually reasonable decisions from different starting information, and those decisions diverge in a way that a working announcement would have prevented.

One of those decisions has a long tail, and it needs to be labeled as a myth in the same breath as the telling. The belief that the roof of a World Trade Center tower was an available exit was a myth, and it was untrue in 2001 on the record the 9/11 Commission set out: the doors to the roof were locked, there was no rooftop evacuation plan, both roofs were sloped and cluttered with radiation hazards that made them impractical for helicopter landings or as staging areas, and the South Tower helipad did not meet the 1994 Federal Aviation Administration guidelines. What gave the myth its grip is that something like it had happened once, because in 1993 New York Police Department helicopters did lift several small groups of people who were physically unable to descend from the South Tower roof, and the Commission records that too. CBS News, reporting in 2004 on the Commission’s work, quoted the head of the Port Authority’s World Trade Department acknowledging that some workers may have gone up with a false sense of security based on the 1993 rescues, which is a report of an acknowledgement rather than an investigative finding, and I offer it as such.

What was changed before 2001

A great deal was done, and the popular accounts tend to merge work performed at different times into a single tidy program, so I will be specific about what I can date and explicit about what I cannot. The 9/11 Commission’s account describes a substantial Port Authority effort covering physical, structural and technological changes to the complex along with an enhanced fire safety plan. Summaries of the Commission’s findings published at the time, including the Gotham Gazette’s account, put a figure of about 100 million dollars on that program and list battery-powered emergency lighting added to the stairwells, backup power added to the alarm system, glow-in-the-dark signs and markings, a redesign of each tower’s lobby command board so that it could monitor all of the elevators, and a repeater system installed for the use of the fire department. I have not been able to verify the dollar figure in a primary document, so treat it as a reported program total rather than an audited one.

The U.S. Fire Administration technical report carries its own list of post-bombing tower safety and security improvements, which is reproduced in later engineering presentations that cite it, and that list includes six satellite communications command and control stations staffed by deputy fire safety directors, new addressable fire alarm systems with redundant command stations, radiating cable and antenna installed so that FDNY portables would work inside the towers, cellular telephones issued to each fire floor warden as a backup to the public address system, and battery power added in elevator cars both for the car position indicator and for communication with trapped passengers. Read that list as a communications engineer and the pattern is clear enough, because almost every item is either a redundant path for information or a local power source for a device that previously depended on the building.

Evacuation drills are the item that gets the loosest treatment in general accounts. Fire drills for civilian occupants were held after 1993, and the Commission’s staff work found that people who evacuated in both 1993 and 2001 reported being better prepared the second time. The same staff findings, as reported by CBS News in 2004, record that neither full nor even partial evacuation drills were conducted, and that civilians were never instructed not to evacuate upward. A drill that assembles people at their floor warden’s station and tells them where the stairs are is a real improvement over no drill at all, and it is a different thing from an occupant who has walked down twenty floors of that specific stairwell and knows what the reentry doors and the transfer hallways look like.

What I cannot give you is a clean year-by-year installation schedule. The Commission dates the repeater installation to 1994. For the photoluminescent markings, the stairwell lighting, the alarm system upgrades and the command stations, the sources I can check describe the program without giving me an installation year I would be willing to print for each item, and the record is unclear on that point. If you need those dates for a citation, the place to look is the NIST NCSTAR 1 volumes and the Commission’s endnotes rather than any secondary summary, including this one.

What a repeater does, and what was installed in the towers

A repeater receives on one frequency and simultaneously retransmits what it hears on another, usually from a better antenna position and at higher power than a portable can manage. Two handheld radios that cannot hear each other directly can both work through the repeater, so the practical coverage of a portable stops being defined by the path between two people and becomes the path between each person and one fixed installation. In a building, that installation is often paired with distributed antennas or a radiating cable run through the structure, so the signal is generated inside the shielded volume instead of trying to punch in from a rooftop site several blocks away. The whole point of the arrangement is to remove the floor-to-floor path loss from the problem by giving everyone a short path to a common point.

The 9/11 Commission’s account of what existed at the World Trade Center is specific, and it is the source for everything in this paragraph. The Port Authority installed a repeater system in 1994, at its own expense, to improve FDNY radio communications in the high rise environment of the towers. The Port Authority recommended that the system be left on at all times. The FDNY asked instead that the repeater be turned on only when it was actually needed, because the channel could cause interference with other FDNY operations in Lower Manhattan. The system was installed at the Port Authority police desk in 5 WTC and was to be activated by Port Authority police officers when FDNY units responded to the complex. Contemporary summaries of the Commission’s findings describe the antenna as being on top of 5 WTC.

Two design consequences follow from that description, and both of them are ordinary engineering observations rather than criticism of anyone. First, availability of the system depended on a procedure being executed by a different agency’s desk officer at the beginning of an incident, which is a human step inserted into the middle of a radio path. Second, the interference concern that led to the on-demand arrangement was a real radio frequency problem and not an excuse, because a repeater that covers a building well also puts a signal into the street outside it, and a busy department operating on the same channel in the surrounding blocks has a legitimate objection to that. Those two facts pulled in opposite directions, and the arrangement that resulted was a compromise between them.

The other thing to notice is what the repeater did not address. It was a coverage device, and the Commission’s own statement of the 1993 failures lists channel loading as the second and independent cause of unintelligible communications. Nothing in the description of the installed system suggests it added talk paths. The USFA-attributed improvement list mentions radiating cable and antenna for FDNY radio use, and I cannot determine from the sources available to me whether that cable was part of the same 1994 installation the Commission describes or a separate element of the same broad program, so I am leaving that open rather than guessing. What the repeater was for, how it was to be activated, and what happened when it was used on 11 September is the subject of the next article in this series.

Which improvements held, and which needed a system to work

Several of the post-1993 changes were genuine improvements and the investigations credit them, though it is worth being precise about the kind of credit each one actually received. The 9/11 Commission’s narrative of stairwell A in the North Tower describes an occupant leaving the 84th floor immediately after the building was hit, finding the stairway dark, smoky and difficult to navigate, and the report states that glow strips on the stairs and handrails “were a significant help”. That is an investigative body crediting the improvement in the course of its own narrative, which is stronger than commentary and weaker than a measurement. The Commission’s staff work also recorded that people who went through both evacuations said they were better prepared in 2001.

For anything quantitative about how much time the markings or the drills saved, the place to look is the NIST occupant behavior and egress volume in the NCSTAR 1 series, which is the only study of the 2001 evacuation built on systematic interviews and modeling rather than on published accounts. I have not been able to verify a NIST statement that assigns a specific time saving to the photoluminescent markings, and popular retellings credit them considerably more confidently than the investigations I can check, so if you need that claim for a grant application or a code argument, read the NIST volume itself and quote what it actually says.

The pattern underneath the outcomes is simple enough to state once. Improvements that removed a dependency kept working under conditions nobody had tested, because photoluminescent markings charge from ambient light and then glow whether or not the building has power, and an occupant who has already walked a stairwell does not need an announcement to find it. Improvements that inserted or preserved a dependency only worked if that dependency held, and the dependencies in question were a power source, a piece of equipment in a survivable location, a channel with capacity available, or a person completing a procedure at the start of an incident. Neither category is intrinsically better engineering, since a radio system by its nature depends on equipment and power and cannot be turned into a glow strip, but the distinction tells you where to concentrate your testing and where to accept that you must have a second way of doing it.

There is a third category that gets missed, which is the improvement that was made correctly and then left exposed to an assumption nobody restated. Drills for civilian occupants improved preparedness, and the same drill program apparently never told people that upward movement was not a route out, which left an assumption in place that the 1993 helicopter rescues had encouraged. A correction that addresses part of a failure and quietly leaves the rest is harder to catch than a correction that was never made, because the file shows action taken.

Closed on paper: why a finding is not a correction

The sequence that produced the 1993 improvements is the same sequence that runs in every agency I have worked with. An incident happens, a review body writes findings, the responsible organization accepts them, a purchase or a policy change follows, the finding is marked closed, and the question of whether the underlying failure mode still exists is never asked again in that form. Nothing in that sequence is dishonest, and most of the time the purchase was the right purchase. The gap is that closure is recorded against the action rather than against the condition, so the file eventually shows a list of completed actions with no statement anywhere of what would have to be true for the original failure to be impossible.

This recurs across decades and across hazards. The interoperability guide on this site works the same ground from a different direction, tracing how the same finding about agencies being unable to talk to each other keeps reappearing in after action reports across many years and many incidents, and I will not repeat that argument here beyond noting that the mechanism is identical. A finding gets converted into a procurement, the procurement is completed, and the condition that generated the finding is still there because it was never the thing being tracked.

The practical way out is to insist that every communications finding carry two separate closure statements. One states what was done, which is the purchase, the policy paragraph, the installation or the training block. The other states the test that demonstrates the condition is gone, who ran it, when, and what the result was. Where only the action statement exists, the finding was purchased against rather than corrected, and where both exist the correction holds only for as long as somebody repeats the test on a schedule, because equipment ages, buildings get renovated, codeplugs change and the people who knew how the system worked retire.

Very few agencies have a 110-story tower in their district, and that does not get anyone out of this. Every jurisdiction has a structure where portable coverage is poor, usually a hospital basement, a jail housing unit, a parking deck, a school with a below-grade mechanical space, or a water treatment plant with thick walls and no in-building system, and alongside it there is generally at least one notification or alerting system that depends on power or on a network path which will not be available during the event it was bought for. Somewhere in the same building there is a folder of findings from a previous incident that were closed out administratively. The 1993 record is useful to a working agency mainly because it shows, with dated published documents, what the interval between a written finding and a tested correction actually costs, and because it shows that the corrections which survive contact with an untested condition are the ones that stopped needing something else to work first. Part two of this series takes up what the radios did and did not do inside the towers on 11 September 2001, using the same set of investigations.

What to do at your agency

• Ask your training officer or whoever maintains the incident file for the last three after action reports your agency produced or participated in, and list every communications finding from them on one page with the date each was written.
• For each of those findings, have the person who owns the file write down what action was taken, what it cost, and what test has been run since to show the condition is gone, leaving the test column blank where nothing has been run rather than filling it with the purchase.
• Pick the single worst in-building coverage problem on that list, send two members with currently issued portables to the location named in the finding, run a radio check to dispatch and to a second portable at the command location you would actually use, and record the result with the date and the radio serial numbers.
• Add one item to the agenda of your next officers meeting or communications committee meeting that already exists, asking who activates any in-building radio enhancement system in your response area, how the responding chief confirms it is active, and what happens if the person who normally activates it is not there.
• Ask your fire marshal or building official for the current list of buildings in your jurisdiction that are required to have an in-building emergency responder radio system, together with the date of the most recent acceptance or annual test for each, and file that list with your preincident plans.
• Write one paragraph into your existing high rise or target hazard SOP stating what the lobby command post does when it loses contact with companies above a given floor, naming the fallback channel or relay position by name rather than describing it generally.
• Have someone in your communications unit confirm, for one building notification or mass alerting system your agency relies on, what its power source is, how long it runs without utility power, and when that runtime was last verified under load.

Takeaways

• A vehicle bomb detonated in the parking structure beneath the World Trade Center complex on 26 February 1993, shortly after midday, killing six people according to the FBI, with some references stating the toll as six people and one unborn child because one of those killed was seven months pregnant, and injuring more than a thousand, with the frequently quoted precise figure of 1,042 injuries coming from secondary references rather than from a single authoritative count.
• The 9/11 Commission Report records that in 1993 the public address system and the emergency lighting failed and that the unlit stairwells filled with smoke, and it states that the general evacuation of the towers via the stairwells took more than four hours, while NIST expressed the same comparison as multiples of the escape time available in 2001 and longer figures in circulation are counting the end of the elevator rescues.
• The Commission gives two distinct reasons for the poor FDNY radio performance in 1993, which are that signals did not penetrate the steel and concrete floors between companies and that too many companies shared one point-to-point channel, and those are separate problems requiring separate fixes.
• A repeater is a coverage device that gives every portable a short path to one fixed installation, frequently paired with radiating cable or distributed antennas inside a building, and it does not add talk paths or relieve a channel that is being stepped on by too many users.
• The Port Authority installed a repeater system in the towers in 1994 at its own expense, according to the 9/11 Commission, located at the Port Authority police desk in 5 WTC and activated by Port Authority police when FDNY units responded, with the FDNY having asked that it be on only when needed because of interference with other operations in Lower Manhattan.
• Improvements that removed a dependency, such as photoluminescent stairway markings that need no power, held up in conditions nobody had tested, while improvements that required equipment, power, channel capacity or a procedure at the start of an incident only worked when those dependencies held.
• The 9/11 Commission credits glow strips on the stairs and handrails as a significant help to an occupant descending from the 84th floor of the North Tower, and anyone needing a quantitative claim about time saved should read the NIST occupant behavior and egress volume and quote what it actually says.
• Any communications finding in your own files needs two closure statements, one describing the action taken and one describing the test that shows the condition is gone, because a finding with only the first was purchased against rather than corrected.