Lower Manhattan is an island district served by a small number of vehicle crossings and by a rail system that runs under the rivers, and in 2000 its daytime working population was roughly ten times its residential one. On the morning of September 11, 2001, the crossings closed and the subway stopped, and large numbers of people walked toward the shoreline. The fleet that carried them off the island had no tasking and no legal authority to evacuate anyone, and nobody counted who it moved. The event repays study because it worked, and because the reasons it worked are not the ones a planner would expect.

The island stated as a transportation problem

Set the events aside for a moment and look at the district the way a transportation planner would look at it on an ordinary Monday. Lower Manhattan occupies the southern tip of an island, and the crossings nearest to it are the Brooklyn Bridge, the Manhattan Bridge, the Brooklyn-Battery Tunnel (now the Hugh L. Carey Tunnel) and the Holland Tunnel, with everything else lying further north. Rail into and out of the district runs under the rivers through a small number of tubes, so road capacity and rail capacity are constrained by different structures but by equally few of them. The demand side is what makes the district unusual, because during business hours the population at risk is overwhelmingly made up of people who do not live there and who therefore have somewhere else they need to get to.

The Skyscraper Museum, in its Residential Rising exhibition on Lower Manhattan, gives the area a daytime working population of 400,000 in 2000 against fewer than 40,000 residents in 20,300 dwelling units, and puts the resident count at 39,000 for that year. A New York State Comptroller report cited by Financial Advisor magazine gives the 2000 population of Lower Manhattan, which it defines as Tribeca and the Financial District, as just over 32,440. Those two residential figures disagree because the boundaries disagree, which is the usual reason neighborhood census figures disagree, and the defensible way to use them is to say that the resident population was in the low tens of thousands while the weekday population was an order of magnitude larger. The New York City Department of City Planning published a comparison of Lower Manhattan covering 2000 to 2005, and that document is a better starting point than any secondary summary if you need figures you will have to defend.

As for what closed and when, published timelines of the morning, including the one maintained by FOX 5 New York, give 9:21 a.m. as the point at which the bridges and tunnels into Manhattan were closed, which is about thirty-five minutes after the first impact. The tracks and stations under the World Trade Center went out of service within minutes of that impact. The compiled summary of closures on Wikipedia, which cites a United States Department of Transportation report, gives 10:20 a.m. to 12:48 p.m. as the window in which all remaining New York City Subway service was suspended, and if the exact minute matters to your work you should get it from the MTA’s own record and not from me. Gothamist, reporting on photographs released by the MTA, described the loss of the tunnel between Park Place and Cedar Street and the destruction of the Cortlandt Street station on the 1 line, which did not reopen to the public until 2018.

Two qualifications belong here before anyone concludes that the water was the only way out, because that overstatement is common and it damages the argument. Closing the bridges to vehicles did not close them to pedestrians, and the photographic record of that morning shows very large numbers of people crossing the East River bridges on foot, so walking to Brooklyn was a real and heavily used route, as was simply walking north on the island. What the water did that the bridges could not do was carry people to New Jersey, because there is no pedestrian crossing from Lower Manhattan to the west bank of the Hudson, and it moved people who had reached the shoreline in the hour after the collapses and were in no condition to walk several miles.

What is established about the water that morning

The Final Report of the National Commission on Terrorist Attacks Upon the United States, published in 2004 and generally called the 9/11 Commission Report, is the reference for the sequence, and it records the two impacts at 8:46 and 9:03 a.m. and the collapses of the South and North towers at 9:59 and 10:28 a.m. Those four times matter to the maritime account because vessels were working the shoreline before the first collapse and the federal call for help went out after the second one, which fixes the Coast Guard’s contribution as organizing and sustaining a movement that had already started rather than initiating it.

Commuter ferries were mid-run at 8:46 because that is what commuter ferries are doing at that hour. History.com’s account of the boat lift states that a spontaneous flotilla had been in action for close to two hours before the Coast Guard issued its call, and Jessica DuLong, in her book on the event, Saved at the Seawall: Stories from the September 11 Boat Lift, which Smithsonian Magazine describes as originally published in 2017 and reissued in paperback in 2021, collected accounts of mariners who moved toward the smoke without instruction because they could see it from the wheelhouse. The Smithsonian piece by John Hanc, published in September 2021, follows one charter vessel making three runs between roughly 10 a.m. and 1 p.m. from the southern tip of Manhattan to Liberty Landing Marina in Jersey City, carrying about 150 people in total, and then being called back to Chelsea Piers to help distribute water to the crowd waiting there.

The call itself was made by Lieutenant Michael Day of Coast Guard Activities New York. In an account published by U.S. Coast Guard News, Day describes preparing to leave for a meeting in Tower One when the command center alert came in, and states plainly that there was no pre-planned response and no concept of operations for two aircraft striking the towers. Andrew McGovern, pilot of the 100-foot pilot boat New York, arrived for that same meeting, and according to the Coast Guard account the two of them gathered lifejackets, a Coast Guard ensign and the OPSAIL 2000 plan and went to the pilot boat, the OPSAIL plan being taken because Day knew it contained ambulance staging areas. The wording of the broadcast is reported consistently across accounts, including those carried by WWL and 13WMAZ, as a call for all available boats, identifying the United States Coast Guard, asking anyone wanting to help with the evacuation of Lower Manhattan to report to Governors Island, and those accounts give the time as 10:45 a.m.

I could not confirm the radio channel, and I am not going to supply one. gCaptain states that Day made the call on his VHF radio, which is consistent with the rest of the account, but nothing I could verify names a channel number. Above Day, Rear Admiral Richard E. Bennis was Captain of the Port of New York and closed the port that morning; an account published by the Brookings Institution describes Bennis as the Coast Guard captain of the port and the incident commander, and Maritime Reporter, announcing his appointment as Associate Under Secretary of Transportation for Maritime and Land Security at the Transportation Security Administration in April 2002, credited him with directing the evacuation by water of approximately 500,000 people. On fleet size, History.com puts the effort at at least 150 vessels and roughly 800 mariners, and the Smithsonian piece gives about 150 commercial vessels, with the Staten Island Ferry among the participants.

The number of people moved, and why it is the softest figure

The figure in general circulation is half a million people in nine hours, and it is the least well supported thing in the whole account. An excerpt from Kendra and Wachtendorf’s American Dunkirk: The Waterborne Evacuation of Manhattan on 9/11, published by Power & Motoryacht, states that credible estimates range from 300,000 according to the South Street Seaport Museum to as many as 500,000 according to the Coast Guard, and the authors add that “Precise numbers will probably never be fully established.” James Kendra and Tricia Wachtendorf are the director and associate director of the Disaster Research Center at the University of Delaware, they went to New York two days after the event to document the response, and their book is the most thoroughly researched account available. Because the researchers who did that fieldwork decline to settle the number, a briefing that states a single figure without the range is going beyond the evidence.

The duration is treated the same way. History.com describes the evacuation as taking less than nine hours, while a review of American Dunkirk published by gCaptain says around ten hours, and neither of those figures is attached to a stated start and stop, so they are not really in conflict so much as loosely bounded. Any number you publish for this event should carry both the source and the range, in the form that estimates run from 300,000 published by the South Street Seaport Museum to about 500,000 published by the Coast Guard, over a period variously given as under nine to about ten hours.

The reason no count exists is mechanical and it is worth walking through, because the same mechanism will defeat counting in your jurisdiction. No tickets were sold and no manifests were kept, and nobody stood at a gangway counting heads. Vessels made repeated trips, so any count assembled afterwards from crew recollection risks double counting people who transferred between vessels and misses people carried on runs that nobody logged. There is also a definitional problem underneath the arithmetic, since a commuter who boarded a scheduled ferry at 9:30 a.m. and rode it home is counted by some accounts and not by others, and the line between an ordinary departure and an evacuation movement was never drawn at the time because nobody was drawing lines that morning.

The Dunkirk comparison, which is where the title of Kendra and Wachtendorf’s book comes from, is worth handling carefully as well. The gCaptain review puts the two operations in the same order of magnitude while noting that the 1940 evacuation ran over a period of days and not hours, and the Brookings account referenced above recounts a conversation with a curator at the Imperial War Museum on the same comparison. The comparison is useful for conveying scale to an audience and it is useless as evidence of anything, so keep it out of the analytical part of a briefing.

Why it worked, asset by asset

The first asset was a standing professional community that already knew the water. New York Harbor supports docking pilots and Sandy Hook pilots, tug and barge operators, commuter ferry companies, excursion and dinner boat operators, municipal ferry crews and a Coast Guard presence, and those people work the same berths, currents, tides and shoal patches every working day. When a vessel master decided to put a bow against a seawall that was never designed for landing passengers, that decision rested on knowledge of local depths and structures that no plan could have transferred to an outsider on the day. The community also has internal reputation, which matters more than outsiders expect, because a master who says on the radio that a berth will take a certain draft is believed by people who know him.

The second asset was that the vessels were already crewed, fueled and under way. The attack occurred during the morning commute, which is the hour of maximum deployment for a ferry system, so the fleet was not sitting at a dock waiting for crews to be called in from home. Compare that with almost any land-side surge resource, where the first two hours are consumed by notification, travel, and finding somebody with keys and a current medical card, and the difference in response time explains why vessels were alongside the seawalls inside the first hour.

The third asset was the common channel. Marine VHF is a shared regime instead of a collection of separate agency systems, and the Coast Guard’s Navigation Center describes channel 16 as the distress, safety and calling frequency, with FCC rules requiring boaters with VHF radios to keep a watch on channel 9 or channel 16 whenever the radio is on and not in use. That produces something that land responders almost never have, which is a single frequency that an entire industry monitors as a matter of routine and legal habit, so a general call reaches everyone without a patch, a cache radio, a loaned portable or a pre-coordinated talkgroup. The Communications section of this site has a separate piece on what the radios did and did not do inside the towers, and the contrast between the two situations is stark enough that it is worth reading them together.

The fourth asset was that the destinations needed no preparation. A ferry slip on the New Jersey or Staten Island side is already a transportation node connected to roads, buses, parking and rail, so landing people there put them somewhere they could continue from without anyone opening a facility, staffing a shelter or writing an agreement. None of those four items is a plan. Every one of them was a pre-existing asset maintained for commercial or regulatory reasons that had nothing to do with emergency response, and the improvisation on the day consisted of connecting assets that already existed instead of creating capability from nothing.

Accountability, warning and dust

Accountability was effectively absent, and the contested headline figure is a direct consequence of that. No manifest was kept, no destination was recorded against any individual, and no agency could say at the end of the day who had been moved or where they had been put ashore. The operational consequences of that are the ones every emergency manager should be able to name without prompting: families could not be told where somebody had landed, missing person lists could not be reduced by checking them against a movement record, medical follow-up had no denominator, and later health registries had to be built from self-identification instead of from a contemporaneous roster. Kendra and Wachtendorf’s conclusion that precise numbers will probably never be established is a statement about the record, and the record is thin because accountability was never established while the movement was happening.

Receiving communities were not warned in any structured way, which follows directly from the absence of a tasking authority. There was no mechanism by which a vessel deciding to run to a marina in Jersey City could tell that marina how many people were coming or what condition they were in, and the Smithsonian account of a charter vessel delivering people to Liberty Landing Marina describes exactly that kind of arrival. I want to be careful here, because I could not find a published formal review of receiving-side operations on the New Jersey and Brooklyn waterfronts to the standard I would want before asserting what happened there, and if such a review exists it did not surface in my searching. What I will assert is the structural point, which is that an evacuation with no tasking authority has no notification path to the receiving end, and that is a design defect, not something any individual on the water did wrong.

Contamination moved with the people. Evacuees boarding vessels were carrying pulverized building material on clothing, skin and hair, and that material was transferred to decks, seating, ventilation intakes and the crews handling them, and then to the receiving shoreline. The exposure question for boat crews is a live one and it belongs to the responder health literature and not to the maritime literature, so the place to check it is the World Trade Center Health Program administered by the National Institute for Occupational Safety and Health within the Centers for Disease Control and Prevention, which publishes its eligibility criteria and its list of covered conditions. Whether a particular category of mariner is eligible is a question to look up in the current program documentation instead of assuming it in either direction, and that caution applies to anyone writing about this subject.

There is a fourth cost that is easier to reason about than to document, which is passenger capacity. A certificated passenger vessel carries a stated number of persons because that number was derived from stability and lifesaving equipment calculations, and an uncontrolled embarkation from a seawall is exactly the situation in which that number stops governing. I have not verified specific instances, so I will not claim any, but the planning implication holds regardless: if your contingency involves civilian vessels taking on people at unimproved landings, the certificated capacity is not what you should be using for your arithmetic, and the risk that gets accepted in the moment should at least have been discussed in advance with the operators and with the Coast Guard captain of the port.

What got written down afterwards, and what is still assumed

This is the section where I have to be most careful, because describing current doctrine from memory is how articles like this go wrong, and the doctrine in this area is distributed across federal, state and local documents that are revised on their own schedules. What I can say with confidence is that maritime evacuation is now a recognized category in planning in a way it plainly was not in 2001, since the event itself was conducted without any plan that contemplated it. What I am not going to do is tell you what your current authority document says about it.

The places to look are specific and you can work through them in an afternoon. Transportation support in the National Response Framework is organized under Emergency Support Function 1, which covers the maritime mode along with the others, and the current annex text is published by FEMA. At the local level, the Coast Guard captain of the port for your zone chairs an area maritime security committee established under federal maritime security legislation passed in 2002, and I have deliberately not cited that statute by number because I did not re-verify the citation for this piece. Many ports also have a harbor safety committee with pilots, tug operators and terminal operators at the table, and that committee is generally the fastest route to the people who actually operate the vessels. If your jurisdiction has a coastal storm or hurricane evacuation annex, the waterborne section of it, if there is one, will tell you what your own government currently assumes.

The honest limit is more interesting than the doctrinal question. Writing a maritime annex creates a document, and it does not create the community of operators, the crewed vessels, the shared monitored channel or the landing points that made the 2001 movement possible, all of which existed for commercial reasons and continue to exist or not exist for commercial reasons. Where a ferry operator has withdrawn from a route, where a tug company has consolidated, or where a terminal has been redeveloped into housing, the capability assumed by an annex written eight years ago may simply not be there anymore, which is why the asset inventory behind any maritime annex needs a named owner and a review date.

The same harbor community turned out again in January 2009, when commercial ferries reached a US Airways aircraft that had come down in the Hudson River and took people off the wings and out of the water. I am not going to discuss that accident beyond the fact of the maritime response, and the National Transportation Safety Board’s report is the record for anyone who wants the rest of it. What the second event demonstrates is that the capability is a standing property of the harbor and not a one-time act, which is an argument for keeping a current list of operators, berths and landing points instead of retelling the 2001 story.

The same asset in a jurisdiction with no waterfront

Most readers of this site do not have a harbor, and the transferable finding has nothing to do with boats. The thing that saved the day was a standing professional community with routine mutual familiarity, common communications, equipment already in service and destinations that required no preparation, and the useful exercise is to ask which communities in your jurisdiction have those four properties whether or not anyone in your agency has ever spoken to them.

In most American counties the honest answer starts with school transportation, because a public school district operates the largest passenger fleet in the county, employs licensed drivers who are on duty and in their vehicles twice a day, dispatches them through a system the district owns, and runs them to fixed sites with parking and restrooms. After that comes the private motorcoach industry, then the ambulance services your county does not own, then the towing and recovery operators who show up at every serious crash, then the utility line crews and the contractors with heavy equipment, and in agricultural counties the grain haulers and the farm cooperative fleets. Each of those is a community with internal reputation and the habit of helping each other, which is the social substrate that made the harbor work.

The communications question is where the comparison stops being flattering to land-side emergency management, and this is the part I would press hardest. Marine VHF gives the harbor a single monitored channel with a legal watchkeeping expectation behind it, and there is nothing equivalent in most counties for school buses, motorcoaches, tow trucks or utility crews. School districts often run their own conventional VHF or UHF system or a talkgroup on the county trunked system, motorcoach operators frequently dispatch by cellular telephone alone, and towing companies may be on a business radio license, on a paging arrangement or on nothing at all. If you want the 2001 outcome in your own jurisdiction, somebody has to find out what each industry actually monitors, document it, and establish whether reaching all of them at once requires one call or a phone tree that nobody has ever tested.

The planning question that follows is narrower than a full transportation annex, and the doctrine side of evacuation, meaning clearance times, contraflow, paratransit and the households with no vehicle, is covered separately on this site in the piece on evacuation as a transportation problem. What belongs here is the relationship question, which is whether anybody at your agency has the mobile number of the district transportation director, the dispatcher at the largest motorcoach operator and the owner of the towing company that works your interstate, and whether any of those people know who to call at your emergency operations center without looking it up.

Reading an event that had no single authority

There is no consolidated after action report for the waterborne evacuation, and the reason is structural, because after action reports are produced by organizations about their own operations and no single organization ran this one. That absence shapes the whole evidence base, so the sources are a scholarly book by two disaster researchers, a journalist’s book built on interviews with participants, the Coast Guard’s own oral history collection, press accounts of varying quality and the recollections of individual mariners recorded years later. Anyone building a briefing from that material has to say out loud which class of source each claim comes from, because a figure from a museum exhibit, a figure from a service’s press office and a figure from a fieldwork-based book are not the same kind of evidence.

The two canonical investigations are quieter on this than readers expect. The 9/11 Commission Report of 2004 is an investigation of the attacks and of the institutional failures preceding them, and its treatment of the New York response concentrates on what happened inside the complex and among the city emergency services. The National Institute of Standards and Technology investigation of the World Trade Center disaster, published in 2005 as the NCSTAR 1 series and including separate volumes on occupant behavior and egress and on emergency response operations, was scoped to the buildings and the people in them, and NIST’s own index of the series is the correct place to get the volume numbers, and a secondary summary is not. Neither body was tasked with the movement of the district’s daytime population off the island, which is why the waterborne evacuation sits outside the documents most people reach for first.

That gap is itself a finding worth taking to your own program. An investigation sees what its terms of reference tell it to see, so a phenomenon that crosses agency boundaries, involves no regulated failure and produces no litigation can be the largest thing that happened and still leave almost no official trace. The practical version for a working agency is that your after action reviews should include at least one question about who helped that you did not task, because the answer is frequently substantial and it routinely fails to reach the written report.

Two other pieces on this site carry the parts deliberately left out here. The Communications section is running its own sequence on the 1993 bombing and the findings that were already on paper, on what the radios did and did not do inside the towers, on the two command posts and the absence of unified command, and on the policy record afterwards, and none of that is repeated above. Part two of this series turns to the casualty surge that hospitals prepared for and that never arrived, which is a planning failure of a completely different shape and one that recurs in almost every mass casualty event since.

What to do at your agency

• Have your operations or logistics officer write down every passenger-carrying and heavy transport asset in the jurisdiction that your agency does not own, naming the school district fleet, each private ambulance service, each motorcoach or charter operator, the transit authority if there is one, and the largest towing and recovery operators, with the vehicle counts taken from the operators and not estimated.
• Telephone the district transportation director this month and ask two questions, which are what radio system or dispatch method the buses actually use during a normal afternoon run and how many drivers can be reached within ninety minutes on a weekday evening, then write both answers into the transportation section of your emergency operations plan.
• Ask your radio system administrator to produce a one-page list of every non-government entity in the county that holds a talkgroup on your system or a mutual aid channel in its radios, and identify from that list which transport industries are missing from it entirely.
• Find out by asking each of your section chiefs by name whether anybody at your agency has ever spoken to the operators identified above outside of a crash investigation, and record the name of the person who holds each relationship so the relationship survives that person’s retirement.
• Add one item to the agenda of a meeting that already exists, such as your LEPC, your fire chiefs association or your county department head meeting, inviting the largest private transport operator in the county to describe how their dispatch works and what they could move on two hours’ notice.
• Pull your agency’s special event plans for the county fair, the largest race or the annual parade, and check whether the staging areas, access routes and landing zones in them match what your emergency operations plan assumes, since that document is often the only one with real locations in it.
• If your jurisdiction touches navigable water, request the current waterborne section of the state evacuation annex from your state emergency management agency and ask the Coast Guard captain of the port’s office what the area maritime security committee holds on evacuation, and date both answers when you file them.

Takeaways

• Lower Manhattan’s exposure was a demand problem before it was a capacity problem, with the Skyscraper Museum giving a daytime working population of 400,000 in 2000 against a resident population in the low tens of thousands, and published timelines giving 9:21 a.m. for the closure of the bridges and tunnels.
• Estimates of the number moved by water range from 300,000 published by the South Street Seaport Museum to about 500,000 published by the Coast Guard, over a duration variously reported as under nine hours or about ten hours, and the researchers who did the fieldwork state that precise numbers will probably never be established.
• The maritime movement began before the Coast Guard call, which Lieutenant Michael Day made at about 10:45 a.m. directing willing vessels to Governors Island, so the service organized and sustained a response that had already begun.
• The radio channel used for that call is not established in sources I could verify, and anyone teaching this event should say marine VHF and stop there.
• Four pre-existing assets explain the outcome, being a standing professional community that knew the water, vessels already crewed and under way during the morning commute, a single channel the whole industry monitors under FCC watchkeeping rules, and destinations on the far shore that needed no preparation to receive people.
• The cost was accountability, because no manifest existed, no agency could say who had been landed where, and receiving communities had no notification path, which is a structural defect of any untasked evacuation.
• Contamination traveled with the evacuees onto vessels and across the water, and the eligibility of any given category of mariner under the World Trade Center Health Program is something to look up in the current program documentation and not to assume.
• Neither the 9/11 Commission Report of 2004 nor the NIST NCSTAR 1 series of 2005 was scoped to the movement of the district’s daytime population off the island, which is why the largest waterborne evacuation on record leaves so little official trace and why your own after action reviews should ask who helped that you did not task.