At 7:17 on the morning of 19 September 1985, by the US Geological Survey’s origin time, a great earthquake ruptured the subduction zone off the Pacific coast of Michoacan. The epicenter was roughly 350 km from Mexico City, yet most of the deaths were in the capital, because much of its center sits on the soft clay of a former lake that amplified the shaking. The government’s response was overwhelmed in the first days, and residents did much of the early rescue themselves. That experience led to Mexico’s national civil protection system. This guide covers the geology, the collapses, the response and a death toll that is still a range.

19 September 1985: a subduction earthquake off the Michoacan coast

The earthquake is usually named for Michoacan, the state whose coastline lies above the rupture, although in Mexico it is more often called simply the earthquake of 85. The US Geological Survey catalogs it at moment magnitude 8.0, and some sources give 8.1. It happened where the Cocos plate descends beneath the North American plate along the Middle America Trench, on a stretch of the margin near the Michoacan and Guerrero border that seismologists had already flagged as a seismic gap. A seismic gap is a segment of a plate boundary that has gone a long time without a large earthquake compared with the segments on either side of it, and a gap is treated as a likely place for the next one.

The coastal region nearest the rupture did take damage, including in and around the port and industrial city of Lazaro Cardenas, and some inland communities further from the coast also reported damage. The concentrated destruction, though, was in Mexico City. The capital was at the start of a working day when the shaking arrived. Many offices, schools and government buildings were not yet full, and it is often observed in accounts of the disaster that a later hour would probably have produced a higher toll. Hospitals, hotels and apartment blocks were occupied, and some garment workshops in the city center already had workers inside on early shifts.

A second large earthquake struck on the evening of 20 September, given at about magnitude 7.5 in most catalogs. It brought down or further weakened buildings that had survived the first shock, and it hit while rescue work was going on in and around the rubble of the first day’s collapses. Anyone reconstructing the response has to keep both events in view, because a city already short of shelter, communications and organized rescue capacity was struck again before it had finished counting what it had lost.

About 350 km away: why the capital took the worst of it

Published accounts give the distance from the epicenter to Mexico City as about 350 km, and some say closer to 400 km, depending on the point measured to. At that range most earthquakes do little structural harm, because seismic energy spreads out and is absorbed as it travels. The absorption is not even across frequencies. High-frequency shaking, the rapid jolting that does most of the damage near an epicenter to short, stiff buildings, dies away faster with distance than low-frequency motion does. The waves that reached the Valley of Mexico had lost much of their high-frequency content and were carrying relatively more of their energy at longer periods, in slow swaying measured in seconds per cycle.

On firm ground in the city, including the volcanic rock of the hill zone where the main campus of the National Autonomous University of Mexico (UNAM) stands, the shaking was moderate and damage was light. On the old lake bed a few kilometers away, the same incoming waves produced violent, long-lasting motion. The difference was in the ground beneath each site, since the earthquake that reached both was the same. Engineers compared the strong-motion record from the university campus with the record from an instrument at the compound of the Secretaria de Comunicaciones y Transportes (SCT), which sits on lake-bed clay, and the comparison became one of the most cited pairs of records in earthquake engineering.

This was not a new lesson for Mexico City. The 1957 earthquake, centered in Guerrero, also caused its worst damage in the lake zone and toppled the statue from the Angel of Independence monument. The city’s building regulations were revised after 1957 and treated the lake zone as a separate, more demanding case. In 1985 the shaking on the lake bed exceeded what those provisions had anticipated, which is why the revision of the code became one of the first technical tasks after the disaster.

The old lake bed: soft clay, a two-second period and long shaking

The Valley of Mexico is a basin with no natural outlet, and before the Spanish conquest much of it was covered by a system of shallow lakes. Tenochtitlan, the Aztec capital, was built on an island in Lake Texcoco. Over the following centuries the lakes were drained, and the city spread across the exposed bed. Beneath that part of the city lies a thick layer of lake clay with an extremely high water content and very little stiffness, tens of meters thick in places. Pumping of groundwater through the twentieth century made the clay consolidate, and the city has been sinking unevenly for decades as a result.

A layer of soft soil over stiffer material has a natural period of vibration, much as a bowl of gelatin does when it is tapped, and that period depends mainly on how thick the layer is and how stiff it is. In the lake zone the clay’s natural period near the SCT instrument was about two seconds, and it varied across the city with clay thickness. When incoming waves carried energy near that period, the clay layer responded by amplifying it. The SCT record shows many cycles of strong, nearly regular motion at about two seconds, and the shaking there went on far longer than on firm ground, because the soft layer trapped energy and kept moving. The reconnaissance and research that followed, including work by UNAM’s Instituto de Ingenieria and by teams associated with the Earthquake Engineering Research Institute in the United States, treated the resonance of the lake clay as the central explanation for the damage pattern.

Long duration matters as much as peak strength. A building that can survive a few cycles of strong motion may not survive dozens, because each cycle cracks concrete further, loosens connections and lowers the structure’s resistance to the next one. On the lake bed the shaking supplied exactly that kind of repeated loading.

Which buildings fell: mid-rise frames, pounding and lengthening periods

The collapses clustered in the lake zone, in central districts that include the historic center, Colonia Roma, Colonia Doctores and the Nonoalco-Tlatelolco housing complex. The buildings that failed in the greatest numbers were mid-rise structures of roughly six to fifteen stories, while many low colonial-era buildings and some much taller towers came through with less damage. The Torre Latinoamericana, which had already survived the 1957 earthquake, came through again.

The explanation engineers gave for the concentration in mid-rise buildings follows from the soil. Every building has its own natural period, and taller, more flexible buildings have longer ones. A building whose period sat near the two-second period of the clay was being pushed at its own resonant frequency. Damage made the problem worse, because as a concrete frame cracks it becomes more flexible and its period lengthens. A building that began somewhat stiffer than the ground period could soften during the shaking until it was moving in step with the clay, and the long duration gave it time to do so.

Several other features showed up repeatedly in the reconnaissance reports. Adjacent buildings of different heights swayed at different periods and hammered against each other, which engineers call pounding. Buildings on corner lots, with solid walls on the two sides facing neighbors and open frames on the street sides, twisted as they shook. Flat-slab floors, in which the slab rests directly on columns without beams, failed where the columns punched through. Some buildings were carrying floor loads well above their design assumptions, from storage, files or machinery, and some had an open ground story with much less stiffness than the floors above.

Among the best-known collapses were the Nuevo Leon building at Tlatelolco, the Hotel Regis, the Hospital Juarez and parts of the Hospital General, a tower of the Pino Suarez complex that fell onto its neighbor, and garment workshops along the San Antonio Abad corridor. The loss of hospital buildings meant that part of the medical capacity the city needed for its injured was itself under the rubble.

The official response: a military plan, a declined offer and a cut-off city

Mexico had no national civil protection system in 1985. The main standing federal instrument for disasters was the armed forces’ Plan DN-III-E, under which the army and other military units support the civilian population. Troops deployed in the capital, and much of what residents saw them do was cordoning collapse sites and keeping order. Accounts from the period describe residents’ anger at cordons that kept volunteers away from buildings where people were believed to be alive, and I report that as the widely published perception rather than as a finding about every site, because practice varied from one location to another. The government of the Federal District, then administered by an appointed regent rather than an elected mayor, did not have a structure designed to run rescue at hundreds of sites at once.

The federal government under President Miguel de la Madrid initially signaled that Mexico could handle the disaster with its own resources and did not need foreign aid, and it reversed that position within days. International teams arrived, some with search dogs and specialist equipment, though much of the earliest and most time-critical rescue had already been done by then, largely by residents. I am stating the initial position and the reversal at the level the general histories support, and readers who need the exact sequence of statements should go to the contemporary press record.

Communications compounded the problem. Damage to the telephone system, including the central facility that handled long-distance traffic, cut much of the capital’s connection with the rest of Mexico and with the world. For a period, a large share of the information leaving the city, including messages from families trying to learn whether relatives were alive, went out by amateur radio. Inside the city, officials had limited means to build a picture of where collapses were and who was working them, which made coordination by any agency harder.

Citizen brigades: bucket lines, tunnels and the groups that stayed organized

Within hours of the first shock, residents across the affected districts were working the collapses. Accounts describe long lines of people passing rubble by hand and in buckets, students from the city’s universities organizing themselves into crews, neighbors setting up kitchens and collection points for food and blood donations, and volunteers directing traffic where signals had failed. Much of this was organized by the people doing it, at the scale of a block or a building, with no direction from any agency and sometimes in open conflict with the officials at the scene.

At Tlatelolco, volunteers crawled into the voids of the collapsed Nuevo Leon building to reach trapped residents. Groups known as the Topos, the moles, trace their origin to those volunteers, and in the years since they have formalized, trained and deployed to disasters outside Mexico. Rescues continued for days after the first shock, and the widely published accounts include newborn infants brought out alive from hospital rubble long after most people had stopped expecting survivors. That kind of rescue was extraordinary, but the work behind it carried real risk, since volunteers were entering unshored voids in damaged buildings during an aftershock sequence that included the magnitude 7.5 event of 20 September.

The garment workers of the San Antonio Abad district became one of the most remembered stories of the disaster. Accounts from the time, including testimony collected by the writer Elena Poniatowska in her book on the earthquake, describe workers trapped in collapsed workshops and survivors’ anger at owners who were said to have tried to recover machinery and records ahead of bodies. The surviving seamstresses organized an independent union, the 19 de Septiembre union, which became a symbol of the wider civic organizing that followed.

The essayist Carlos Monsivais wrote about the earthquake as a moment when society in the capital organized itself in the space the state had left empty, and that interpretation, which I paraphrase rather than quote, has become a common reading of 1985 in Mexico. Groups of displaced residents formed coordinating organizations to press for housing, and the government expropriated several thousand inner-city properties in the weeks after the earthquake for a reconstruction housing program. The political consequences ran for years, and the general histories of Mexican civil society treat them as part of the story of the earthquake.

The death toll: roughly 5,000 officially, 10,000 or more by other counts

Nobody knows how many people died. The figure usually reported as the Mexican government’s official count is in the region of 5,000, although the exact official number varies between the sources that cite it. The US Geological Survey’s historical summary of the earthquake gives at least 9,500 deaths. Some Mexican journalists, researchers and victims’ organizations have argued for 10,000 or more, and I have seen much higher figures in popular retellings that I cannot trace to any count. These sources disagree, and I am not going to average them or pick one. The honest statement is that published estimates run from roughly 5,000 to 10,000 or more, with the low end coming from the official count and the higher end coming from the USGS and from Mexican critics of that count.

The reasons for the spread are mostly practical. Bodies were recovered from hundreds of sites by many different groups, including volunteers who kept no records, and were taken to improvised morgues, among them a baseball stadium by widely published accounts. Identification was difficult, missing-persons reports came through many channels at once, and no single list reconciled the people recovered with the people reported missing. Many victims lived in crowded rental housing in the central districts, worked informally or had come from other states, and people in those circumstances are easily missed by an official registry.

Critics have also suggested that the official count was kept low for political reasons. That is a suspicion held and published by those critics, and I report it as such rather than as an established fact. The practical failures described above would have produced an undercount on their own, whatever anyone intended.

What followed: SINAPROC, CENAPRED, the seismic alert and the 19 September drill

The institutional response came over several years. In May 1986 a presidential decree established the bases for a National Civil Protection System, the Sistema Nacional de Proteccion Civil, usually shortened to SINAPROC. It set out a framework linking federal, state and municipal governments with private and social organizations, so that disaster response would no longer rest on the armed forces’ plan and on whatever a given city could improvise. A national center for disaster prevention, CENAPRED, was set up at the end of the 1980s with technical cooperation from Japan to carry out research, monitoring and training. Federal civil protection law followed in 2000 and was replaced by a general civil protection law in 2012, and anyone who needs the current legal framework should read the law as it now stands rather than rely on a summary.

Mexico City revised its building regulations, first with emergency provisions soon after the earthquake and then with a new code later in the decade, with particular attention to the lake zone. In the early 1990s a seismic early warning system began operating, built around sensors on the Pacific coast that detect an earthquake and send an alert that can reach the capital before the slower damaging waves arrive. The same distance that let the 1985 waves lose their high-frequency energy gives the system a useful margin for coastal earthquakes, often measured in tens of seconds, although the margin is much shorter or absent for earthquakes closer to the city.

Mexico observes 19 September as a civil protection day, and a national drill has been held on or around that date. On 19 September 2017, an earthquake centered in the state of Puebla struck about two hours after that year’s drill and caused deadly collapses in Mexico City, again concentrated in buildings on soft ground. The coincidence of date is widely remarked on. The 2017 earthquake is a separate event with its own record, and it is the place to look for how the post-1985 system performed in practice.

What to do at your agency

• The emergency management coordinator requests the soil or site-class map for the jurisdiction from the state geological survey or the county GIS office, overlays it on the critical facilities list, and writes into the existing hazard mitigation plan which hospitals, shelters and dispatch centers sit on soft or filled ground.
• The emergency management coordinator checks whether the earthquake annex of the emergency operations plan starts its first-morning damage estimate from distance to the epicenter, and if it does, rewrites that paragraph to start from USGS ShakeMap and the local soil map instead.
• The communications center supervisor schedules a test this month in which the emergency operations center reaches the state warning point and one neighboring county with commercial phone and internet treated as unavailable, and records in writing which paths worked and how long each one took.
• The emergency management coordinator asks the local amateur radio emergency group’s leader for the current roster and the list of served-agency locations the group can staff, and adds an amateur radio activation item to the agenda of the next regular emergency management meeting.
• The fire department’s special operations or technical rescue officer adds one paragraph to the existing structural collapse procedure covering spontaneous volunteers already working a collapse when the first company arrives, stating who briefs them, where they are moved, and which tasks they may keep doing.
• The coroner or medical examiner’s office, with the emergency management coordinator, names in the existing mass fatality plan the single role that owns the reconciled missing-persons list, and confirms that one person or office holds that list from the first operational period.

Takeaways

• The Michoacan earthquake of 19 September 1985 was a subduction earthquake of magnitude 8.0 in the USGS catalog, centered off the Pacific coast roughly 350 km from Mexico City.
• Most of the deaths were in the capital because the old lake-bed clay under its center resonated at a period near two seconds and shook far longer than firm ground did.
• Collapses concentrated in mid-rise buildings of roughly six to fifteen stories, whose own periods were near the ground period or lengthened toward it as they were damaged.
• Pounding between neighbors, twisting of corner buildings, flat-slab punching failures and overloaded floors appeared repeatedly in the damage surveys.
• Mexico had no national civil protection system in 1985, the federal response leaned on the military’s Plan DN-III-E, and telephone damage cut much of the city’s long-distance communication.
• Residents, students and neighborhood groups did much of the early rescue, and some of those groups, including the Topos, stayed organized afterward.
• The death toll is a range, with roughly 5,000 by the official count, at least 9,500 by the USGS summary, and 10,000 or more by some Mexican estimates.
• The disaster led to SINAPROC in 1986, CENAPRED, a revised building code, a seismic early warning system and an annual national drill on or around 19 September.