On the night of 13 November 1985 Nevado del Ruiz erupted, and roughly two and a half hours later a wall of hot mud and rock came down the Lagunillas valley and buried the town of Armero, Colombia. Commonly published death tolls run from about 21,000 to 25,000, with the United States Geological Survey usually giving about 23,000. A hazard map placing Armero inside the lahar zone had been completed and circulated roughly a month earlier, and the volcano had been restless for a year. The technical warning was correct and early, and it still did not produce a protective action.

A year of the mountain clearing its throat

Nevado del Ruiz is a glacier-capped volcano in the Colombian Andes, and the hazard that makes it dangerous is the ice sitting on top of it as much as the eruption itself. Hot material erupted onto snow and ice produces meltwater, the meltwater picks up loose volcanic debris, and the result is a lahar, which behaves like fast-moving wet concrete and follows river valleys for tens of kilometers. That mechanism was not new in 1985, since historical accounts of eruptions in 1595 and 1845 describe lahars sweeping the same drainages, and the 1845 event is commonly reported to have killed on the order of a thousand people in the Lagunillas valley. Armero was built on the deposit it left behind.

Unrest began around November 1984, with increased fumarolic activity at the summit and earthquakes felt in the region. Colombia had very little volcano monitoring capability at the time, and published accounts generally agree that the first seismographs went onto the mountain only months before the eruption, with thin instrument coverage and improvised data handling. Colombian geologists and university staff were joined over the following year by visiting scientists from several countries and by United Nations technical assistance, and the people doing the technical work understood what they were looking at.

On 11 September 1985 the volcano produced a phreatic eruption that threw ash over nearby towns and generated a small lahar down one of the summit drainages. That was the mechanism demonstrating itself in miniature, two months in advance and in front of witnesses, and it removed any remaining argument about whether ice melt and channelized mudflows were a realistic scenario. It is also what pushed the hazard mapping work to completion.

The hazard map, and what it actually said

A preliminary volcanic hazard map for Nevado del Ruiz was completed in early October 1985, roughly a month before the eruption, produced by Colombian geological authorities with input from the international scientists then working on the mountain. It identified the river valleys draining the ice cap as lahar corridors and showed inundation reaching well downstream, with Armero squarely inside the zone. Accounts of the period describe the map being reproduced in the national press in October, so it had circulated well beyond the geologists who drew it.

Be precise about what a document like that can do. The map told you where the mud would go if it came, and it said nothing about timing, about magnitude, or about what any particular official was supposed to do on any particular evening. That is normal for hazard mapping, and it is exactly the gap a warning program is supposed to fill with triggers, thresholds, and pre-assigned decisions. In Armero that gap was never filled.

There is a second and uglier part of the record. Several published accounts, including the peer-reviewed retrospective literature, describe officials and local interests treating the map and the scientists’ warnings as alarmist, and describe pressure not to frighten the population or damage the local economy. I would characterize that as a well-supported pattern in the published accounts rather than as a single documented order from a single named person, and readers who want the specifics should go to those analyses rather than take my summary as the record.

The warning chain, and the link that had no owner

Trace the chain that existed on paper. Scientists on the mountain observed the volcano and passed their observations to national technical and civil defense bodies in Bogota, which communicated with departmental authorities in Tolima and Caldas, who in turn communicated with municipal officials including the mayor of Armero, the local Red Cross, and the civil defense volunteers. The municipal level was where somebody could actually tell nearly thirty thousand people to walk uphill. Every one of those links existed, and what did not exist was a defined trigger obligating anyone in the chain to act, or a designated person at the bottom of it holding both the information and the authority at the same moment.

Add the context of that week. On 6 November 1985 guerrillas of the M-19 seized the Palace of Justice in Bogota, and the military assault to retake it left about a hundred people dead, including Supreme Court justices, after which the national government spent the following week consumed by that crisis and its aftermath. A functioning warning system should not depend on the undivided attention of a national cabinet. The people at the top are always busy with something else, which is the ordinary condition of government rather than a Colombian peculiarity, and any warning architecture that requires them to notice will eventually fail on a busy week.

The technical message also degraded as it traveled. Scientific caution is appropriate in a scientific document and it is poison in a public message, and a statement that lahars are possible in the event of a significant eruption, passed down four levels by people who are not volcanologists, arrives at a municipal official as something that might happen someday. Nothing in that sentence tells a mayor to do anything tonight.

The night of 13 November

The volcano produced a phreatic explosion in the afternoon of 13 November 1985, and ash began falling on Armero in the late afternoon. This is the part of the record where accounts differ in detail and agree in substance. Local authorities and Red Cross personnel reacted, some form of evacuation was considered or begun, and when the ashfall eased the urgency was withdrawn, with published accounts describing residents being told over local radio and from the church to remain calm and stay in their homes. I cannot reconstruct exactly who said what to whom, and anyone who claims to with confidence is reading one source.

The main eruption began at about 9:08 pm local time. Hot material moved across the summit ice cap and melted part of it, lahars formed in several drainages and began moving downstream, and heavy pumice and ash fell on Armero in the hours that followed. The evening also brought a thunderstorm, and accounts describe power failures and degraded telephone and radio communication in the affected area during exactly the window when a warning needed to move.

The first lahar reached Armero at roughly 11:30 pm after traveling tens of kilometers down the Lagunillas channel, and successive pulses followed, covering most of the town in mud several meters deep within minutes. On the western flank the town of Chinchina was struck by a lahar down its own river, with about 1,800 deaths commonly reported there. The interval between the start of the main eruption and the arrival of the mud in Armero was on the order of two and a half hours, and higher ground lay close enough that people moving on foot could have reached it, so the time existed even though the decision did not.

The false alarm nobody wanted to own

Evacuating a town of nearly thirty thousand people is expensive and disruptive, and in November 1985 it also meant pulling an agricultural community away from its land and its stored crop at a bad time of year. The official who signs an unnecessary evacuation order gets the bill the next morning with a name attached to it, since every merchant and farmer in the municipality knows exactly who signed. Declining to sign carries a cost as well, and that cost stays diffuse and deferred and easy to discount in the room where the decision is being made, until the night it lands all at once. That asymmetry is the standing structural incentive in every warning decision anyone reading this will ever make.

The way out is to move the decision off the individual and into the plan before the event. If the trigger is written down, agreed in advance by the elected body, and tied to an observable condition, then the person who pulls it is executing policy rather than gambling a career. Communities that do this well also say out loud, in advance and in public, that a certain number of evacuations will turn out to have been unnecessary, and that this is the designed and accepted cost of the system.

The research literature on public response is worth knowing here, because the folk belief runs the wrong way. The persistent claim that people who evacuate for a false alarm will refuse the next warning is not well supported by the evidence, and it functions mostly as cover for officials who did not want to make the call. What does erode credibility is a warning that is vague, that comes from a source people do not recognize, or that tells people to stay calm when the correct instruction is to leave.

The toll, and where reputable sources disagree

Armero’s population is usually given as around 28,700 to 29,000. Deaths in the town are commonly reported at roughly 20,000 to 23,000, with about 1,800 more at Chinchina, and total figures published by reputable sources range from about 21,000 to 25,000. The United States Geological Survey and the Smithsonian Global Volcanism Program generally give approximately 23,000, and that is the figure I would use if I had to use one, while other reputable accounts, including some Colombian sources, give up to 25,000. Those numbers disagree, they are estimates rather than counts, and an exact count was never possible because most of the dead were buried in place under meters of mud.

Injuries are commonly reported at around 5,000, with more than 5,000 homes destroyed across the affected municipalities, and economic loss estimates near one billion United States dollars in 1985 terms are widely cited. Treat all of these as period estimates compiled under chaotic conditions. Nevado del Ruiz 1985 is generally described as the deadliest volcanic disaster of the twentieth century after the 1902 destruction of Saint-Pierre on Martinique, and as the deadliest lahar event on record.

The response failure deserves its own line. Rescue capability arrived slowly into a landscape of deep mud with no roads and no landing areas, and people who survived the flow died over the following days because nobody could reach them or extract them. The photograph most people associate with Armero shows Omayra Sanchez, a thirteen-year-old girl trapped in debris and standing water, who was reached by rescuers but could not be freed with the equipment available and who died on 16 November 1985 after roughly three days. That image won World Press Photo of the Year and did more to force a national reckoning than any technical report did.

Cumulative human error, and what the retrospective found

The most cited technical retrospective is by the volcanologist Barry Voight, published in the peer-reviewed volcanology literature around 1990. Its central conclusion, paraphrased, is that the catastrophe came from an accumulation of human and institutional errors rather than from any failure of the science or the available technology, and it names bureaucratic inertia, unclear responsibility, poor communication of hazard information, and reluctance to act on ambiguous evidence. Voight concluded that the deaths were avoidable. That is his analysis and not a legal finding, and readers who want the argument should go to the paper rather than to my summary of it.

One technical point inside that analysis matters more than any other for planning purposes. Predicting the timing and size of a volcanic eruption in 1985 was genuinely hard, and by some standards it still is, while predicting where the mud would go was not hard at all. Lahars follow river channels, the channels were mapped, the historical deposits were visible in the ground under the town, and the September event had already demonstrated the process. You did not need to know when the eruption would come in order to know that Armero was in the path.

Colombia did respond institutionally. A national system for disaster prevention and response was created in 1988, in the direct aftermath, and the legal framework was substantially rewritten in 2012 into a national disaster risk management system with a dedicated national unit. If you need the current structure, statutory basis, or terminology, get it from Colombia’s national disaster risk management authority rather than from a historical article, because these frameworks change.

What this means for a warning program today

Start by naming the person. For every hazard in your plan there should be a named position that issues the protective action order, a named alternate, and a documented delegation of authority that survives the primary being unreachable at 9 pm on a Wednesday. If the answer to who orders the evacuation is a committee, a department, or the state, then the answer at the moment of decision will be nobody. Armero’s chain had many participants and no owner at the point where authority and information had to meet.

Then write the triggers against observable conditions rather than against forecasts. A trigger reads like a gauge reading, a confirmed report from a specific source, a National Weather Service product, or an activity level declared by a monitoring agency, and it is paired with a pre-scripted message and a planned route. Get the elected body to adopt the triggers in advance and in public, and have the conversation about false alarms then, when it is cheap, instead of at the moment somebody has to spend political capital they do not have.

Build the delivery path assuming the easy parts will be gone, because on the night of 13 November there was ash, rain, darkness, and power failure, which is a fair description of the environment in which most warnings have to be delivered. That means redundant paths, at least one of which does not depend on commercial power or the internet, and it means the message itself is unambiguous, since telling people to stay calm and remain in their homes is defensible for some hazards and fatal for a lahar. Exercise that last mile rather than the conference room: choose a specific valley at a specific hour with the power out, and see how far the message actually gets. That is the test Armero failed, and it is the only one that counts.

Takeaways

• A preliminary hazard map completed about a month before the eruption placed Armero inside the lahar zone, so the failure was in acting on the information rather than in producing it.
• The 11 September 1985 phreatic eruption demonstrated the ice melt and lahar mechanism two months in advance, which removed any legitimate argument that the scenario was speculative.
• Roughly two and a half hours elapsed between the start of the main eruption at about 9:08 pm and the arrival of the lahar in Armero at about 11:30 pm, and higher ground was close enough to reach on foot.
• Commonly published total death tolls run from about 21,000 to 25,000, with the USGS and the Smithsonian Global Volcanism Program generally giving about 23,000, and an exact count was never possible.
• Predicting the eruption was genuinely difficult, but predicting where the mud would go was not, which means the evacuation decision could have run on an observable trigger instead of a forecast.
• The published retrospective analysis by Barry Voight concluded that the disaster resulted from cumulative human and institutional error rather than technological failure, and that the deaths were avoidable.
• A warning that does not reach a person holding both the information and the legal authority to order protective action, at the moment it matters, is not a warning.
• Decide your triggers, name your decision maker, and have the false alarm conversation with elected officials in advance, because the incentives at the moment of decision run against ordering the evacuation.