On January 18, 2002, a Canadian Pacific Railway freight train derailed just west of Minot, North Dakota, and five tank cars of anhydrous ammonia ruptured near homes where people were asleep. Within a year or so, the night had become a standard example in the national argument over radio ownership. It was retold as the story of a town that could not get an emergency warning on the air because nobody was working at the radio station. Some of that account is documented, and some was added in the retelling. This piece checks both against the NTSB record and explains how a local alert was supposed to reach the public in 2002.

January 18, 2002: the derailment and what the NTSB established

At about 1:37 a.m. local time, a Canadian Pacific Railway freight train derailed 31 cars just west of the Minot city limits, according to the National Transportation Safety Board’s final report on the accident. Five of the derailed tank cars were carrying anhydrous ammonia, and they ruptured catastrophically. The term means the tank shell itself failed and released its load in a very short time, as opposed to leaking from a valve or fitting. By the NTSB’s estimate, the release came to nearly 150,000 gallons of ammonia. Part of the liquid flashed to vapor at once, and a cloud formed over the derailment site and spread into the surrounding area, including a residential neighborhood close to the tracks.

The NTSB report records one death, a resident of a home near the derailment. It also records 11 serious injuries and a little over 300 minor injuries, and the minor-injury figure includes the two members of the train crew. I use those figures because they come from the agency that investigated the accident and published its counting basis. You may see a different number in news coverage or later writing. If so, check whether it counts the same thing. A figure for people who went to an emergency department, or who were examined at a shelter, answers a different question from the NTSB’s injury classification. For training material, take the number from the injury table in the report and say which count it is.

The NTSB’s final report, adopted in 2004 and posted on the board’s website among its railroad accident reports, contains the board’s formal probable cause determination along with its findings and safety recommendations. I am deliberately not paraphrasing that determination here. The board worded it carefully, a compressed version tends to drift each time it is repeated, and anyone using Minot in training should copy the statement word for word from the report rather than from an article, including this one. The findings section runs to several pages and is the place to read the board’s own account of the track, the tank cars and the oversight questions it examined.

For this piece, what matters is what the probable cause statement leaves out. It does not attribute the death or the injuries to the public alerting system, the radio stations, or the timing of the warning. That is relevant to everything that follows, because the most widely repeated version of the Minot story treats the alerting problem as the reason people were hurt, and the board, in its formal finding on this accident, made no such attribution.

Anhydrous ammonia, not chlorine: getting the cargo right

Some retellings of Minot describe the cargo as chlorine. That is wrong. The cargo was anhydrous ammonia, and the NTSB report names the material in its title. I suspect the confusion comes from a separate accident, the Norfolk Southern derailment at Graniteville, South Carolina, in January 2005. That release was chlorine, and it also hit a small community in the early morning. Graniteville has its own NTSB report and its own record, and the two cases should not be blended in slides or articles.

Anhydrous ammonia is shipped and stored as a liquefied compressed gas under pressure. Across the northern plains it is a common agricultural fertilizer, which is why it moves by rail through farm country in volume. When a pressurized tank fails suddenly, part of the liquid boils off at once and part is carried out as a cold aerosol mist. As a result, the early cloud can behave as denser than air and follow low ground, even though pure ammonia gas at room temperature is lighter than air. Ammonia attacks moist tissue, which means the eyes, the airway and the lungs, and at high concentration it can kill quickly. Minot sits in the valley of the Souris River. A dense release on a winter night near that valley put a cold, heavy cloud into terrain where it could settle.

Getting the material right is more than pedantry. Chlorine and ammonia both force the same protective action decision between sheltering in place and evacuating, but they differ in density, odor, and behavior on release. A trainer who teaches Minot as a chlorine case will eventually be corrected by someone in the room who knows the record, and the rest of the lesson loses standing with that audience. For initial isolation and protective action distances for either material, use the current edition of the Emergency Response Guidebook and your local hazmat guidance rather than any number in an article, including this one.

The popular account and where it came from

What follows is the popular account, not established fact. When the ammonia cloud spread, local authorities tried to warn residents over the radio. Six of the seven commercial radio stations in Minot were owned by one company, Clear Channel Communications. The station designated to carry emergency information, usually named as KCJB, was running automated programming with nobody answering the phone. In that telling, the warning could not get out, and residents were left uninformed while the cloud moved over them. No official investigation reached that conclusion, and the full version should be treated as a retelling.

The most influential early source was a 2003 New York Times article by Jennifer 8. Lee about corporate ownership of radio in Minot. The story reached a much larger audience through Eric Klinenberg’s 2007 book Fighting for Air, which used Minot as a central example of what media consolidation could cost a community. Both works were about ownership and localism, and both drew on interviews and contemporary reporting. The six-of-seven ownership figure was widely reported at the time, and I have no basis to dispute it. Ownership counts, however, are a different matter from what went wrong with the alert that night.

As the account moved into advocacy writing, it often went further than the original reporting:

• That the failed alert caused the death and injuries. Some versions imply or state this outright. No official review found it.
• That no warning reached the public at all. Accounts of the response describe responders reaching residents of the nearest neighborhood directly, which the NTSB describes as rescues. They also describe broadcast coverage later that night, on a timeline whose details I treat below as disputed.
• That the whole failure sat inside the radio station. The company contested this in the public argument that followed, and the structure of the alerting system makes the question more complicated than the story allows.

Each of these additions should be labeled as an addition when it appears in a presentation, because none of them came from the investigating agency.

How a local EAS activation was supposed to work in 2002

The Emergency Alert System replaced the older Emergency Broadcast System in the 1990s. Broadcast stations were required to have EAS equipment installed by January 1, 1997, and the rules sit in Part 11 of the FCC’s regulations. Broadcasters were required to carry the national-level Emergency Action Notification, the presidential alert, and to run the required tests. Carrying state and local alerts was voluntary. In practice it ran under a state EAS plan written by a State Emergency Communications Committee and filed with the FCC.

A state plan divides the state into local areas and names stations to serve as Local Primary sources, usually written LP-1 and LP-2. These stations monitor designated sources and relay alerts into the broadcast chain. Every station’s EAS equipment monitors at least two assigned sources. When it receives a coded alert, it can be set either to forward the alert to air automatically or to hold it for an operator to approve. A local emergency official had two general ways into that chain:

• Coded alert. The official sends a coded alert from an EAS encoder over whatever link the local plan provides to the LP station.
• Phone call. The official calls the station and asks the staff to originate the alert there.

The two paths fail in different ways, and this is the mechanical point the popular story skips. An unattended station set to forward local alerts automatically will put a properly coded alert on the air with nobody in the building. Staffing matters only when the official relies on a phone call, or when the station’s equipment is set to hold local alerts for manual approval. A call to an empty control room fails whoever owns the station. A coded alert fails if the encoder, the link or the configuration does not work, again whoever owns the station. Ownership can drive staffing decisions, and staffing affects the phone path, so the ownership argument does connect to the mechanics, but only indirectly.

Testing raises a related problem. Stations run Required Weekly Tests and Required Monthly Tests, and those tests show that a station’s equipment receives and relays alerts. They do not show whether a county dispatcher at 2 a.m. can originate a local alert. That depends on working encoder equipment and on a dispatcher who has actually used it. Where the plan relies on a call, it also depends on a current after-hours number for a person at the LP station. The path a local official actually uses is the one most likely to go untested, because nothing in a station’s routine requires anyone to try it.

What the official record supports about the alert, and what it leaves open

The NTSB investigated the emergency response as part of its accident investigation, and its report discusses notification of the public in that context. I am not going to reconstruct the minute-by-minute account of the notification attempts from memory. Those details are exactly where retellings have gone wrong, and the report is publicly available. What I can say with confidence is narrower. The probable cause statement in the report does not attribute the death or the injuries to the alerting system. In the board’s account, the casualties resulted from exposure to ammonia released by the ruptured cars.

On what went wrong with the broadcast alert, early press reports said authorities could not reach anyone at the designated station. Clear Channel contested that account of its station in the argument that followed, and some later writing places at least part of the failure in alerting equipment or procedures on the government side. I am reporting that these claims exist, not endorsing them. I have not been able to confirm from a source I trust which part of the chain failed first, or how many minutes the delay ran. The claim that the government side failed deserves the same caution as the original story that the station failed, because both have circulated mainly in writing aimed at the ownership debate.

One point does not depend on settling that dispute. The homes closest to the derailment were near enough that the cloud plausibly reached them within minutes. That was before a dispatcher-originated broadcast alert could have woken anyone, and before most people would have turned on a radio at that hour. This is my own judgment from the geography and the time of night, not a finding I can cite. It is consistent with the NTSB’s account of responders reaching residents of the nearest neighborhood directly. A faster radio alert would have mattered most for people farther out. They had more time, and their risk was waking, smelling ammonia, and going outside or into a car.

My summary is this. A real problem with getting a prompt broadcast warning out existed that night. Where that problem sat is contested. No official review I know of connects it to the death or the injuries. The NTSB report and its public docket are the record. Anyone teaching the case should read the emergency response discussion there directly rather than relying on the version told in the ownership debate, or on mine.

What the warning was for: sheltering in place under an ammonia cloud

For a large toxic release near a city at night, emergency managers choose between telling people to shelter in place and telling them to evacuate. For a dense cloud already over a neighborhood, the usual answer is to shelter, because people walking or driving out through the cloud are more exposed than people behind closed doors and windows. Accounts of the Minot response describe sheltering in place as the main protective action for the city, while responders removed residents from the area nearest the tracks. Whether that was the right call is a question the NTSB report addresses in its own terms, and readers should take the board’s assessment from the report.

That choice determines what the alert has to accomplish. A shelter-in-place message has to reach people who are indoors and asleep. It has to reach them before they wake on their own, smell ammonia, and do what comes naturally, which is to open the door or start the car. In 2002, broadcast radio and television reached only people who already had a set on, a small share of any town at 2 a.m. in January, and neither one woke anyone up. Wake-up layers existed for that reason. Tone-alert receivers such as NOAA Weather Radio units with the alarm feature, and outdoor sirens despite their limits indoors in winter, filled that role, and local governments were adopting telephone notification systems in the same period.

This is where the popular story and the operational lesson separate. Suppose every radio station in Minot had been locally owned and staffed around the clock. A radio alert would still have reached only the people awake and listening, and the people asleep nearest the tracks would still have depended on responders at the door. The fix for that problem is a layered warning plan with at least one layer that wakes people. That is a question for the local emergency manager and the 911 center as much as for any broadcaster.

Shelter-in-place guidance for a chemical release generally tells people to:

• close windows and doors,
• shut off heating and ventilation systems that pull in outside air, and
• move to an interior room.

The specific wording belongs in the community’s own plan, informed by current hazmat guidance. On a January night in North Dakota, shutting off the furnace carries a real cost. That is one reason the message should say how long people can expect to stay put and where updates will come from, so they do not abandon shelter because the house is getting cold.

How Minot became an exhibit in the ownership debate

Most people outside the railroad and hazmat world know Minot because of its radio, and the reason is timing. In 2003 the FCC revised its broadcast ownership rules. Opponents in Congress and in public interest groups used Minot as a concrete example of what consolidated ownership could mean on a bad night. The 2003 rules were challenged in federal court, and the Commission opened an inquiry into broadcast localism in the same period. Minot came up repeatedly in the public argument around both.

An example used in a policy argument tends to get compressed. The ownership advocates were making a claim about local staffing and local service. Minot was useful to them because it was vivid and because the one-owner fact was true. The details that complicated it did not help the argument, and they dropped out as the story moved from reporting into commentary. Those details included how local alerts entered the broadcast chain and whether a staffed station would have changed anything for the people nearest the tracks. This pattern is common with disaster stories that become political evidence. It does not mean the underlying concern about local staffing was wrong, only that Minot was a weaker illustration of it than the retellings suggested.

Some writers on the other side of the ownership debate pushed back on the Minot story and argued that it put the failure in the wrong place. I count neither camp as an official review. The ownership question is a policy matter the FCC and the courts have argued over for two decades. The alerting question is an operational matter that the NTSB examined and that every local emergency manager can check in their own jurisdiction. Mixing the two made Minot a stronger talking point. It also left responders with a version of the case that pointed them toward broadcaster staffing instead of their own origination path and their own wake-up layers.

Clear Channel later became iHeartMedia, and the radio ownership landscape in many markets has shifted more than once since 2003. For a working agency, none of that changes the operational question. The question is whether the people on duty tonight can originate a warning through the path your plan names, and whether that warning will reach people who are asleep.

The record applied to alerting today

Federal alerting changed substantially after 2002. I am not aware of any official document that credits Minot as the reason, so I describe the changes without claiming a causal link. Executive Order 13407 in 2006 directed the creation of an integrated public alert and warning system, which became FEMA’s Integrated Public Alert and Warning System. Wireless Emergency Alerts began reaching phones in 2012. A local alerting authority today typically originates through software that sends a Common Alerting Protocol message to IPAWS. IPAWS can then distribute it to EAS participants, to phones through WEA, and to other channels, depending on the authority’s permissions and the state plan.

Two Minot-type failure modes survive the change in technology:

• The untried origination path. The path passes its tests but has never been used by the person who will actually be on duty at 2 a.m. That might be an encoder in a rack, or a set of alerting software credentials that only one person in the county knows how to use.
• The hidden phone call. The plan quietly depends on calling someone outside the agency, such as a broadcaster, a neighboring county or a state warning point, using a contact number nobody has dialed in a year.

Both failures are invisible in daytime testing run by the person who set up the system, and both show up at night when someone else is on the console.

WEA gives most agencies a layer that can sound on a phone at the bedside, which did not exist in 2002. It still depends on phones being on, on the alert’s geographic targeting, and on how the alert is categorized. NOAA Weather Radio can also carry non-weather emergency messages at the request of officials, through arrangements with the local National Weather Service office. Readers should verify their own county’s arrangement with that office and with the state emergency management agency rather than assume it exists.

The version of Minot that holds up in a training room is short:

• A derailment released nearly 150,000 gallons of anhydrous ammonia near homes at night.
• The NTSB’s probable cause determination, which should be quoted from the final report rather than paraphrased, does not attribute the casualties to the alerting system.
• Local officials had trouble getting a prompt broadcast warning out, and the details of that trouble are disputed.
• The story that blames the death on the radio station’s staffing is not a finding of any official review.

The lesson worth taking home concerns the agency’s own origination path and its wake-up layer, which are under local control in a way broadcast ownership never was.

What to do at your agency

• The emergency management director pulls the current state EAS plan from the state emergency management agency or the State Emergency Communications Committee this month. The director finds the local area that covers the county and writes the LP-1 and LP-2 stations, and the plan’s stated method for local origination, into the warning annex of the existing emergency operations plan.
• The 911 center manager has an overnight-shift dispatcher, not the day-shift alerting lead, send a test message through the agency’s alerting software in the test or training environment its authorization provides. The manager records in the training file who sent it and how many minutes passed from the decision to the confirmed send.
• The communications center supervisor calls every after-hours number for outside contacts listed in the warning annex, including broadcasters, the state warning point and neighboring centers, between midnight and 5 a.m. The supervisor corrects any number that reaches voicemail, a disconnected line, or a person who does not know what to do with the call.
• The emergency management coordinator drafts one pre-scripted shelter-in-place message for a toxic gas release near the rail line, short enough to fit the 360-character WEA limit, and saves it in the alerting software’s template library with the approving official’s name on it.
• The training officer adds an inject to the next scheduled hazmat tabletop in which the primary alerting path fails at 2 a.m. The exercise notes record how the group gets a wake-up warning, by another route, to the neighborhood nearest the tracks.
• The training officer reviews any agency training material that cites Minot and corrects the cargo to anhydrous ammonia, the casualty figures to the NTSB report, and the alerting story to what the record supports.

Takeaways

• The Minot derailment happened at about 1:37 a.m. on January 18, 2002, when a Canadian Pacific freight train derailed just west of the city and five tank cars of anhydrous ammonia ruptured catastrophically.
• The NTSB report records one death, 11 serious injuries, and a little over 300 minor injuries, a count that includes the two members of the train crew.
• The cargo was anhydrous ammonia, not chlorine, and the chlorine derailment sometimes blended with Minot is the separate 2005 Graniteville, South Carolina, accident.
• The NTSB’s probable cause determination, in its final report adopted in 2004, should be quoted word for word from that report, and it does not attribute the casualties to the alerting system.
• In 2002, local EAS alerts were voluntary under state plans, and station staffing mattered mainly when officials relied on a phone call or the station held local alerts for manual approval.
• Early reports that nobody could be reached at the designated station were contested, and later claims placing the failure on the government side should be treated with the same caution until checked against the NTSB record.
• A broadcast alert at 2 a.m. reaches few people. For any agency, the durable lesson is a tested origination path used by the people actually on duty, plus at least one warning layer that wakes people up.