At 5:46 a.m. on January 17, 1995, a fault rupture under the Akashi Strait and northern Awaji Island shook Kobe and the cities along Osaka Bay for roughly twenty seconds. Older wooden houses collapsed onto sleeping families, and a section of elevated expressway fell onto its side. Fires started in the first minutes and burned through whole blocks of Nagata Ward over the following day. Japan’s Fire and Disaster Management Agency gives the toll as 6,434 dead, a figure that includes deaths in the aftermath. This guide covers how the buildings failed, why the fires spread, who actually pulled survivors out, how slowly the national response formed, and how a surge of volunteers changed Japanese law.

5:46 a.m. on January 17: the fault, the shaking and the belt of damage

The Japan Meteorological Agency named the event the 1995 Southern Hyogo Prefecture Earthquake. The Japanese government named the resulting disaster the Great Hanshin-Awaji Earthquake, and most English sources call it the Kobe earthquake. JMA gives the magnitude on its own scale as 7.3, revised from the 7.2 it reported in 1995, while the U.S. Geological Survey gives a moment magnitude of 6.9, so readers comparing it with other earthquakes should know which scale a source is using. The hypocenter was shallow, about 16 kilometers by JMA’s figure, and the rupture broke the surface along the Nojima fault on Awaji Island. A stretch of that scarp is preserved today at a memorial museum on the island.

The earthquake struck on a Tuesday morning in midwinter. Most people were asleep, and the first scheduled trains had not yet started. In the worst-hit parts of Kobe, Ashiya and Nishinomiya, the shaking was later assessed at intensity 7, the top of the JMA scale as it stood then. It was the first time that level had been assigned. JMA did not measure it on instruments in the moment. It assigned it after field surveys of damage, and the surveys traced a narrow band of severe destruction running roughly east to west through the urban strip. Japanese writers often call this band the earthquake’s belt of damage. Its location reflected the ground conditions and the basin geometry beneath the cities as well as distance from the fault.

Kobe, with about 1.5 million people, sits on a narrow strip of land between the Rokko mountains and Osaka Bay. Road, rail and utility lines all run parallel through that strip, so damage to a few corridors cut most of the region’s east-west movement at once. The port had been built out onto reclaimed islands, and Port Island and Rokko Island liquefied, with quay walls pushed seaward. Kobe had been one of the busiest container ports in the world before the earthquake. Most accounts say it never recovered that rank, because shipping lines moved to other Asian ports while it was being rebuilt.

Older wooden houses: heavy roofs, weak ground floors and the 1981 line

Most of the people who died were killed inside houses that collapsed in the first seconds. The typical failure was in an older two-story wooden house of post-and-beam construction. Many of these houses carried heavy clay-tile roofs bedded on a layer of mud. That roof suited a typhoon climate, because the weight resisted wind, but it put a large mass at the top of a light frame. Many ground floors had few solid walls, particularly where the front of the house had been opened up as a shop. Older timber had often been weakened by rot or termites. When the ground floor racked and gave way, the upper floor and roof came straight down. People sleeping on the ground floor had almost no space left around them.

Japan’s Building Standard Law had revised its seismic provisions in 1981, and those provisions are commonly called the new seismic standard. Post-earthquake surveys, including work by the Architectural Institute of Japan, found that buildings put up before that revision failed at much higher rates than those built after it. The pattern held for older mid-rise concrete buildings as well as for wooden houses. Several of those concrete buildings lost a single middle story, which collapsed while the floors above and below stayed largely intact. The Fire and Disaster Management Agency counts more than 100,000 residential buildings as totally destroyed. Many more were classed as half destroyed or partially damaged.

The housing that failed was concentrated in older inner-city wards such as Nagata, Hyogo and Higashinada, where a large share of the residents were renters, older people and people on low incomes living in housing stock that predated the modern code. Japan passed a law in late 1995 to promote seismic retrofitting of existing buildings. Retrofit of privately owned older houses has moved slowly since, because the owners least able to pay are often the ones living in the weakest buildings.

The Hanshin Expressway at Fukae and the other structures that failed

The best-known image of the earthquake shows the elevated Kobe Route of the Hanshin Expressway lying on its side at Fukae in Higashinada Ward. A section of about 635 meters toppled over. That section was carried on single reinforced concrete columns. The viaduct dated from the late 1960s and had been designed to the standards of that period. Post-earthquake investigations, including committee work for the Ministry of Construction and by the Japan Society of Civil Engineers, generally attributed the failure to the column detailing. The columns had too little transverse reinforcement to confine the concrete, and some longitudinal bars were cut off partway up the column, which left a weak zone where the columns failed.

The expressway was one failure among many. A span of the Bay Route fell near Nishinomiya, and the Sanyo Shinkansen viaducts were damaged or collapsed at several points. The earthquake struck before the first bullet trains of the day were running, which is the main reason that damage killed no passengers. Daikai Station on a Kobe underground rail line collapsed, and the road above it subsided. Engineers cite it as one of the first major collapses of a modern subway structure in an earthquake. Conventional rail lines through the corridor were cut in many places.

The transportation damage mattered to the response as much as to the economy. With the elevated routes down and rail lines cut, residents leaving the area, relatives coming in, relief convoys and emergency vehicles all crowded onto the surface roads through the strip, and those roads jammed for days. Fire engines and ambulances were stuck in the same traffic as everyone else. Japan revised its highway bridge design specifications in 1996, and older column-supported viaducts across the country were retrofitted, much of that work with steel jacketing of the piers. The Kobe Route itself did not fully reopen until 1996.

Fires after the shaking: Nagata, dry hydrants and the power coming back

Fires broke out within minutes of the shaking. Kobe’s fire bureau, by its own published records, faced more than a hundred fires on the first day, far more than it had engines to send. The ignition sources were the ordinary ones of a winter morning in a dense residential area. Kerosene and gas heaters overturned, gas leaked from broken lines, and damaged electrical equipment sparked. A later category of fire drew particular attention in Japan. When utilities restored power to districts that had lost it, some damaged houses caught fire from wiring or appliances that came back on in a broken building, and some of these fires started hours or days after the earthquake.

The fire service’s water supply failed with the ground. Broken mains left hydrants without pressure, and the underground fire cisterns that some neighborhoods had were emptied quickly. Crews drafted from rivers and the sea where they could and laid long relay lines. Collapsed houses, downed utility poles and gridlocked traffic slowed every one of those operations. Firefighters were also being stopped in the street by residents asking for help with people trapped under houses nearby. The crews had to choose between rescue and fire attack with no way of doing both. Winds were light that morning, and that is a large part of why the fires did not grow into the kind of conflagration that followed the 1923 Great Kanto earthquake.

The worst burning was in Nagata Ward, where small factories, including the synthetic-leather shoe industry, stood among closely packed older wooden houses. The Fire and Disaster Management Agency’s final figures count about 7,000 buildings as totally burned. Japan’s fire services set up Emergency Fire Response Teams in 1995 so that units from other prefectures could be dispatched into a disaster area under a standing national arrangement. Many cities also added earthquake-resistant water tanks and cisterns, so that fire suppression would not depend entirely on a pressurized main.

The death toll: 6,434 by the official count and what that count includes

The figure most often published is 6,434 dead. It comes from the Fire and Disaster Management Agency’s final damage report on the disaster, issued in 2006. The same report lists three people missing and more than 43,000 injured. The 6,434 figure includes what Japan classifies as disaster-related deaths. These are people who survived the earthquake itself and then died in the following weeks and months from causes the authorities linked to the disaster. Anyone quoting the number should say that it includes those deaths, because other published figures do not.

Counts limited to people killed directly by the collapses and fires are usually given as about 5,500. The difference, roughly 900 deaths, consists of disaster-related deaths certified by municipalities in the months that followed. Those certifications were tied to the national system of condolence payments to bereaved families, so the counting rule is administrative and depends on local review. Commonly cited causes include pneumonia among older evacuees in cold, crowded shelters, cardiovascular deaths and the effects of exhaustion and stress. Hyogo Prefecture’s own figure for deaths within the prefecture is slightly lower than the national total, at about 6,400, because the national count includes a small number of deaths in neighboring prefectures such as Osaka. Counts published in 1995 and 1996 were lower than the final figure, because certification of disaster-related deaths went on after the first tallies.

For a briefing, a defensible wording is “6,434 dead by the Fire and Disaster Management Agency’s final count, including disaster-related deaths, of whom about 5,500 died directly in the earthquake.” Medical examiner data reported by Hyogo Prefecture and widely summarized since attribute most of the direct deaths to crushing and suffocation in collapsed buildings. A smaller share, most often given as around a tenth, was attributed to burns or fire. Most published breakdowns also show that older residents made up a disproportionate share of the dead.

Who got people out: neighbors, the first hours and the myth of the late rescue

A persistent retelling holds that most survivors of the Kobe collapses were dug out alive by professional teams over the following days, and that is a myth. The record points the other way. Japanese government disaster management white papers state that about eight in ten people rescued from collapsed buildings were freed by family members and neighbors. A survey by the Japan Association for Fire Science and Engineering, reproduced in later government materials, asked people who had been trapped or buried how they got out. Most said they freed themselves or were freed by family, neighbors or passersby, and only a very small share credited rescue teams.

Timing explains most of that pattern. Rescue data from Kobe’s fire bureau, which has been widely reproduced, show the share of people pulled out alive dropping steeply after the first day. Most of the live rescues happened in the first hours. Fire companies were fighting fires and were stuck in traffic. Police were dealing with their own losses, and Self-Defense Force units arrived in numbers only later. The people standing on the block when the shaking stopped were the residents, and they worked with their hands, house jacks, bars and whatever tools the neighborhood had. That outcome does not reflect badly on the professionals. Distance, roads and the sheer number of collapses meant that organized teams reached most sites after the period in which most survivors were found, and much of their work then shifted to recovering the dead.

Japan drew a policy conclusion from this, usually expressed as a division between self-help, mutual help and public help. Government encouragement of neighborhood-level voluntary disaster prevention organizations grew sharply after 1995, with training in light rescue, fire suppression and support for vulnerable neighbors. The U.S. Community Emergency Response Team program, whose origins in Los Angeles predate Kobe, rests on the same reasoning. The Kobe rescue data are among the strongest evidence for building that capacity at the block level rather than counting on outside teams to arrive in time.

The government response: a slow picture in Tokyo and a late request for troops

The local governments that would normally have driven the response were themselves casualties. City and prefectural employees were dead, injured or cut off at home, and many could not reach their offices that morning. A middle floor of one of Kobe’s city hall buildings collapsed. Telephone networks were damaged and saturated. The hardest-hit districts were the least able to report what had happened to them, so the information that reached Tokyo in the first hours understated the scale badly. Accounts of that morning describe national officials relying heavily on television coverage. Prime Minister Tomiichi Murayama’s government set up an emergency headquarters on the first day, and it was criticized afterward for how long it took to grasp the size of the disaster and act on it.

Under the law and practice of the time, the Self-Defense Forces generally deployed for disaster relief when a prefectural governor formally requested it. By most accounts, Hyogo Governor Toshitami Kaihara’s formal request came at about 10 a.m., more than four hours after the earthquake. Some nearby units had begun moving on their own initiative before that. The full deployment built up over the following days rather than arriving in force on the first morning. Offers of help from abroad, including search teams with dogs, were widely reported to have run into delays over acceptance and entry procedures. Accounts of those delays vary in detail, and the government inquiries and histories of the period are the place to check the specifics.

The reforms that followed were aimed at these gaps. The Disaster Countermeasures Basic Act was amended in 1995. Changes to Self-Defense Force dispatch rules allowed units to deploy on their own judgment in urgent cases where waiting for a request was not possible, and they gave municipal mayors a route to press for a request. A 24-hour Cabinet Information Collection Center opened in 1996. The post of Deputy Chief Cabinet Secretary for Crisis Management was created in 1998. JMA moved to instrument-measured seismic intensity in 1996 and expanded its network of intensity meters, so that the strength of shaking would be known within minutes. That network gives the central government an automatic signal of where the damage is likely to be, without waiting for reports from the places hurt worst.

The volunteer surge and the laws that followed

Volunteers came to Kobe from across Japan, many of them university students who had never done relief work before. Hyogo Prefecture’s estimate, the figure most often cited, is about 1.4 million volunteer participations in the first year. That is a cumulative count, so one person who worked ten days counts ten times, and it should not be read as the number of different people. Japanese media called 1995 the first year of the volunteer era. That label stuck because organized citizen relief on this scale had not been part of the country’s disaster system before.

Nobody had planned to receive these volunteers. Many arrived without food, lodging or a task, and shelters run by schoolteachers and municipal staff absorbed them as best they could. Volunteer groups and some shelter residents improvised coordination, matching people to work such as food distribution, moving the elderly and clearing debris, with uneven results. Civic groups also faced a legal obstacle. Before 1998 it was difficult for a citizen group to obtain legal status in Japan, since incorporation largely depended on discretionary approval by a government ministry. A group without legal status had trouble signing a lease, opening a bank account or accepting donations in its own name.

The Cabinet designated January 17 as Disaster Prevention and Volunteer Day in late 1995, along with a disaster prevention and volunteer week around it. The Act on Promotion of Specified Nonprofit Activities, usually called the NPO Act, passed in 1998 and gave citizen groups a simpler route to incorporation. In later Japanese disasters it became common practice for local social welfare councils to open disaster volunteer centers that register volunteers and assign them work. Kobe also left a lesson about recovery. Temporary housing was allocated largely by lottery, with priority for older and vulnerable residents. That allocation scattered neighborhoods, and deaths of older residents living alone in the temporary units became a recognized problem. Later Japanese recovery planning has tried harder to move communities together.

What to do at your agency

• The emergency management coordinator asks the building department or the county assessor for construction-year data on residential buildings and produces a map layer that marks neighborhoods dominated by buildings that predate the jurisdiction’s modern seismic or wind code.
• The communications center supervisor writes one paragraph into the existing 911 surge plan stating what call-takers tell callers when structural collapse calls exceed available units, including neighbor rescue guidance and gas shutoff instructions, and gets it signed by the center director.
• The fire chief asks the water utility for its list of mains and pump stations considered most vulnerable to ground failure, and the department marks static water sources and drafting sites on the same map before the next quarterly update.
• The emergency management coordinator writes a paragraph into the emergency operations plan that names a trigger for requesting state support based on observed conditions, such as loss of contact with a jurisdiction for a set period, so that the request does not wait for a formal damage assessment.
• The CERT or volunteer program manager pulls the roster of trained members by neighborhood and identifies, in writing, every neighborhood on the pre-code building map that has no trained member living in it.
• The emergency management coordinator adds an item to the next scheduled utility coordination meeting asking the electric utility to describe how it checks damaged structures before re-energizing a district, and records the answer in the meeting notes.

Takeaways

• The Great Hanshin-Awaji Earthquake struck at 5:46 a.m. on January 17, 1995, with a JMA magnitude of 7.3 and a USGS moment magnitude of 6.9, and the worst shaking followed a narrow belt through Kobe, Ashiya and Nishinomiya.
• The Fire and Disaster Management Agency’s final count is 6,434 dead, including roughly 900 disaster-related deaths certified later, and counts of direct deaths are usually given as about 5,500.
• Most direct deaths came from crushing and suffocation in older wooden houses with heavy tile roofs and weak ground floors, and buildings put up before Japan’s 1981 seismic revision failed at much higher rates.
• The Hanshin Expressway’s Kobe Route toppled over a stretch of about 635 meters at Fukae, and investigators generally attributed the failure to inadequate confinement and reinforcement detailing in its single concrete columns.
• Fires spread in Nagata Ward because broken mains left hydrants dry and roads were jammed, and some fires started later when power was restored to damaged buildings.
• The idea that most survivors were pulled out alive late by professional teams is a myth, because government data show about eight in ten rescued people were freed by family and neighbors, mostly in the first hours.
• The national response was slow because the hardest-hit governments could not report, and reforms followed in Self-Defense Force dispatch rules, Cabinet crisis management, interprefectural fire support and seismic intensity monitoring.
• About 1.4 million volunteer participations in the first year, by Hyogo Prefecture’s cumulative estimate, led to a national volunteer day and to the 1998 NPO Act that eased incorporation for citizen groups.