The 9/21 Earthquake: A Fault Line Running Through a Campus, Taiwan Leaves the Ruins in Place

30-Second Overview: At 1:47:15.9 AM on September 21, 1999, a magnitude 7.3 earthquake began in the suburbs of Nantou, with a focal depth of only 8 kilometers. ¹ The central region suffered the heaviest casualties: official park records document 2,415 deaths, 11,305 injuries, and 29 missing persons, with 51,000 buildings damaged. ² But this article is not just about a disaster; it is about a choice. Taiwan did not flatten all traces for reconstruction; instead, it left the uplifted playground, the shifted dam structure, and the collapsed school buildings in place, turning them into textbooks for the next generation to understand earthquakes.

A Line Suddenly Appearing in the Night

At 1:47:15.9 AM on September 21, 1999, Taiwan time, the Central Weather Bureau’s earthquake report recorded coordinates 23.85°N, 120.82°E, a depth of 8 kilometers, and a Richter magnitude of 7.3. ¹ This was not merely a point on a distant instrument. Nantou reached a maximum intensity of 7, while Yunlin, Taichung, and Chiayi reached intensity 6. The shaking spread across the central plains and mountains; many people realized their houses were moving only after hearing objects fall in the dark.

• Taiwan Panorama*’s retrospective refers to this earthquake as the 9/21 Earthquake or the Jiji Earthquake, noting that the shaking lasted 102 seconds, causing 2,415 deaths, over 11,000 injuries, and damaging or destroying more than 80,000 buildings. ³ The significance of this earthquake lies not just in its magnitude, but in the fact that surface deformation occurred at a scale close to human life: it turned the fault line from a line on a geological map into a site that could cut through schools, dams, and transportation infrastructure.

📝 Curator’s Note: The most unforgettable image of 9/21 is not the abstract term “the earth shaking,” but the ground actually rising like a table pushed from underneath, causing the playground, dam, and roads to change height independently.

The Playground at Guangfu Junior High Was Not Erased

Guangfu Junior High School in Wufeng, Taichung, is the most visible place along this line. After the Chelungpu Fault cut through the campus, the eastern half of the playground uplifted by 2.6 meters, and multiple school buildings were damaged. ² Public Television’s retrospective recorded this tangible result that remains today: the running track was lifted by the fault, and the damaged school buildings and landforms were left together, later becoming part of the 9/21 Earthquake Educational Park. ⁴

This decision was somewhat counter-intuitive. The most natural thought after a disaster is usually to clear dangerous buildings, fill cracks, and restore life to “how it was.” But Guangfu Junior High’s original site was not completely erased. In 2000, the Ministry of Education selected the campus as the base for an earthquake museum, and in 2001, it was officially named the “9/21 Earthquake Educational Park.” The Fault Preservation Hall opened in 2004, and the entire park began operations in 2007. ²

The park does not simply enclose collapsed buildings. The National Museum of Natural Science’s explanation divides it into three aspects: historical humanities, earth sciences, and earthquake-resistant technology, integrating the exhibition hall with the originally damaged school buildings. ² Visitors do not see a monument removed from its site, but rather how the fault cut through a school, how it changed the height of the track, and how architects arranged circulation between safety setbacks and in-situ preservation.

What is preserved here is evidence that “things happened,” and also a classroom for “why they happened that way.” The former makes people stop; the latter teaches them what to do next time.

Shigang Dam: Earthquakes Shake More Than Just Houses

Downstream of the Dadu River, Shigang Dam demonstrates another scale of disaster. The Water Resources Agency describes it as a key hub for water resource utilization in the Dadu River basin. After 9/21, the fault passed through the dam area, uplifting the right bank by 2.2 meters and the left bank by 9.8 meters. The dam body and energy dissipator cracked, multiple spillway gates were damaged, and the water conveyance tunnel was also destroyed. ⁵

The consequences were more complex than “the dam broke.” Damage to the weir caused water loss, the intake shifted, and the conveyance tunnel collapsed. Fengyuan water purification facilities, main water supply pipes, and local distribution pipelines were also simultaneously damaged. The Water Resources Agency recorded that this left approximately 2 million people in the greater Taichung area facing water shortages. ⁵ An earthquake thus transformed from a geological event into an interruption of daily life: faucets, factories, hospitals, and schools were all tied together by the same invisible line.

Public Television’s retrospective notes that Shigang Dam was repaired in 2000, but some damaged sections of the dam were preserved as witnesses to the disaster. ⁴ This echoes the choice at Guangfu Junior High: restoring function does not mean erasing memory; allowing the dam to continue supplying water also allows people to see what it once endured.

📝 Curator’s Note: 9/21 translated “infrastructure” from an engineer’s term into a problem everyone understands—If a dam cannot hold water, can tomorrow’s city still live normally?

Reconstruction Is Not Returning to the Starting Point

Displaced residents initially had no abstract historical questions, only the question of where to sleep tonight. The central disaster zone saw a large number of homeless people after the quake, and the government implemented temporary prefabricated houses, rent subsidies, and discounted purchases of national housing as安置 (settlement) measures. ⁴ But settlement is only the first layer of reconstruction. Houses can be rebuilt, but community relationships, work, care, and psychological security do not automatically return.

The 9/21 Earthquake Reconstruction Foundation’s original reconstruction experience page retains entry points for data on joint work between civil groups, communities, and the government after the disaster, covering aspects such as housing, community building, social welfare, psychological rehabilitation, and rescue. ⁶ These categories themselves indicate that reconstruction is not just about fixing buildings one by one, but about reconnecting the relationship between people and place.

Data from Public Television also records that the Executive Yuan and civil groups jointly formed the 9/21 Earthquake Reconstruction Foundation. After the foundation dissolved in July 2008, remaining funds and real estate were transferred to the Disaster Relief Foundation. ⁴ This institutional history reminds us that donations are not the end of a disaster narrative. They enter organizations, plans, and long-term responsibilities, ultimately facing the question of how to end and how to hand over.

The foundation’s original website states it was established on October 13, 1999, and over eight and a half years, promoted 32 reconstruction plans, handing over unused resources to the subsequent disaster relief system. ⁶ These numbers are the foundation’s own narrative of achievements and should be read separately from government statistics or academic research; they allow us to see how a civil reconstruction institution describes its work, but cannot replace case-by-case verification of individual plan effectiveness.

Temporary Prefabricated Houses Are Not the Period at the End of the Story

Temporary prefabricated houses are often placed in the background of post-disaster photos, as if they were merely a transition before formal housing appeared. But for those who lost their homes, temporary shelters were simultaneously bedrooms, living rooms, family meeting rooms, and places where they redecided whether to stay in their original location. The government provided prefabricated houses, rent subsidies, and discounted national housing, addressing “first, giving people a place to live.” Communities and civil groups took over “how to live after settling down.” ⁴⁶

This distinction is important. Housing repair has engineering progress; human trust does not. Neighbors may be scattered to different settlement points, children changed schools, and the elderly lost familiar markets and care networks. Within the same village, some may wish to rebuild in place, while others want to leave the dangerous area. Putting everyone into the noun “displaced persons” makes these choices disappear.

The 9/21 Earthquake Reconstruction Foundation’s data page divides reconstruction experience into entry points such as local participation, reconstruction, rescue, and the Kobe experience, leaving articles on community building, psychological rehabilitation, social welfare, and housing work. ⁶ These materials may not represent everyone’s feelings, but they remind later readers: post-disaster recovery is never a single path, nor does it happen automatically after central orders are issued.

Here, the protagonist of reconstruction is not just a safe new house, but whether a place can regenerate communal life. If reconstruction only looks at completion numbers, it easily forgets that some people need to find neighbors, work, schools, medical care, and a night where they can sleep with the door locked in peace.

Three Types of Time Left by an Earthquake

9/21 has three different times. The first is the time of instruments: 1:47:15.9 AM, depth 8 kilometers, magnitude 7.3. ¹ The second is the time of architecture: school buildings were damaged in seconds, the dam lost function after the quake, but repair spanned several years. ²⁵ The third is human time: a family may still discuss that night years later, while the next generation may only hear about it for the first time in a classroom or exhibition hall. ⁴

These three times do not align naturally. News needs to explain casualties on the day; engineers need to interpret cracks afterward; community workers accompany people through grief without a clear endpoint; museums must place these memories of different speeds in the same space. The existence of the 9/21 Earthquake Educational Park is precisely to turn a few dozen seconds of earthquake into a long-term course that can be walked into, understood, and questioned. ²

• Taiwan Panorama*’s retrospective mentions that the disaster prompted Taiwan to rethink building codes, disaster relief systems, and the relationship between economic development and the natural environment. ³ These terms sound grand, but they ultimately fall into details: how a house is designed, how a bridge is reinforced, how a water pipeline is backed up, whether a school should preserve its damaged site.

📝 Curator’s Note: Disaster turns “talk about it later” into a bill; what reconstruction truly pays for is not just engineering costs, but whether we are willing to write experience into rules.

From Fault Lines to Databases

9/21 also changed how people understand engineering data. The National Center for Research on Earthquake Engineering’s technical report database preserves 1999 building damage surveys as report numbers, authors, and outcome records. ⁷ The Bridge Damage Online Historical Museum breaks down bridge history, drop-prevention, isolation, pier reinforcement, and numerical simulation into searchable research entry points. ⁸

For general readers, these pages may not bring immediate emotion like a memorial hall. But it is precisely this calm classification that ensures the next discussion does not start from zero. Which structure failed first, which reinforcement is worth adopting, which road might block rescue—all require comparable records.

Data also has its boundaries. Technical reports are good at describing structures but cannot answer alone how long someone waited in a temporary shelter; official statistics are good at organizing totals but cannot put the name of every lost person back into a complete context. Good post-disaster knowledge does not choose between engineering or people, but allows both to calibrate each other. ⁷⁶

Therefore, preserving the site and preserving the archives are two sides of the same coin. The uplifted track at Guangfu Junior High lets people see surface deformation; technical reports let people know how to analyze damage; the damaged dam body at Shigang Dam lets people understand the water supply crisis; water data fixes the displacement of 2.2 meters on the right bank and 9.8 meters on the left. ⁵ These data complement and limit each other: earthquake reports do not tell us how a family lost its home; oral memory cannot alone replace structural appraisal. Only by placing both in the same article can we simultaneously see the speed of disaster, the length of reconstruction, and the weight institutions must bear.

From Collapse to Resilience, Separated by Many Practices

After 9/21, Taiwan’s changes occurred not only in memorial halls. _ Taiwan Panorama_’s retrospective directly points out that the disaster exposed deficiencies in building standards and disaster relief systems, prompting society to re-examine building regulations, geological faults, and post-disaster reconstruction. ³ These changes were not completed in one declaration, but by breaking post-disaster experience into work that can be repeatedly corrected: building safety, engineering research, community drills, and government coordination.

The National Center for Research on Earthquake Engineering’s technical report database preserves the formal report record of the 1999 “9/21 Jiji Great Earthquake Comprehensive Disaster Survey—Building Damage Investigation,” numbered NCREE-1999-054. ⁷ Engineering reports do not finish the story for displaced persons, but they can organize collapse patterns, structural weaknesses, and repair methods into knowledge usable next time. Insurance cannot prevent houses from falling, nor can it replace lost lives; it addresses another often-ignored problem: when disaster shatters a family’s residence and assets simultaneously, who bears the cost of starting over?

The 9/21 Bridge Damage Historical Museum preserved by the National Center for Research on Earthquake Engineering focuses attention on engineering issues such as bridges, drop-prevention, isolation, pier reinforcement, and numerical simulation. ⁸ These studies seem far from daily life, but actually determine whether rescue vehicles can pass, whether supplies can be delivered, and whether broken transportation networks can be reconnected.

Thus, the “after” of 9/21 is not a smooth curve of progress. New building codes may increase costs; reinforcing old buildings requires resident consent; community reconstruction may also show gaps due to resource allocation and local power dynamics. Describing post-disaster change as a beautiful success story instead obscures places that still need maintenance.

The Scale of Viewing Disaster

Photos, earthquake reports, and damaged buildings are not the same kind of evidence. Photos bring the viewer’s gaze to a specific moment; reports break the moment into time, coordinates, and numbers; buildings fix the result of force into material. Only when placed together do they prevent 9/21 from being shrunk into a shocking disaster picture that cannot be questioned. ¹³

The value of retrospective articles lies not only in recounting casualties. They also allow us to see how society at the time understood disaster: which images were repeatedly published, which questions were placed in special topics, which institutional gaps are still worth checking years later. ³ For those who did not personally experience 9/21 today, these images are entry points, not conclusions; after the entry point, one must still return to official data, in-situ preservation, and local memory.

Images also have areas requiring careful use. They can bring readers closer to the situation of the受灾 (disaster-affected), but cannot treat others’ losses as consumables for viewing. Therefore, this article does not use photos as decoration for casualty numbers, but reads them within context of source, time, and preservation: which institution preserved them, when were they published, whether they record the disaster itself or later retrospectives—these must be explained separately.

Putting Memory in Place

In 2024, Public Television reorganized this earthquake with a 25-year retrospective, opening by reminding the younger generation: many people no longer have direct impressions of that night. ⁴ This sentence is important because time turns disaster from bodily memory into a date in a textbook. As those who personally experienced it gradually decrease, preserving the original site is not just for commemoration, but to allow later people to still compare what “ground uplifted by 2.6 meters” actually means. ²

The track at Guangfu Junior High, the dam body at Shigang Dam, and an official earthquake report each preserve three scales. The track is a scale felt by human height; the dam is a scale the city’s water supply can bear; the report fixes those early morning seconds into repeatable time, coordinates, depth, and magnitude. ¹ Only when placed together does 9/21 not remain just a disaster photo.

📝 Curator’s Note: Commemoration is not polishing the wound, but preserving a shape accurate enough so that those who did not experience it can still approach it.

Taiwan cannot choose whether the next earthquake will come, but it can choose what to leave, what to correct, and what to teach to those not yet born. The most powerful place of the 9/21 Earthquake Educational Park may be precisely that it did not restore the campus to a place where no one could see it had been injured.

It lets a fault line continue to run through memory. This is not an addiction to staring at disaster, but an acknowledgment: true reconstruction is not making everything look like it never happened, but allowing life to continue after knowing it happened.

What This Memory Still Lacks

An article can place official numbers, engineering reports, and preserved sites together, but cannot pretend it represents everyone’s 9/21. Death tolls can be counted; how those who lost homes spent their first winter requires oral history, local archives, and community records to supplement. ⁴ This is why the Reconstruction Foundation’s data page lists not only housing and engineering, but also retains article entry points for social welfare, psychological rehabilitation, and community building. ⁶

Similarly, in-situ preservation is not the only answer. For those living in dangerous buildings, leaving ruins may be education; for those who still need to live, safe, affordable housing and stable income are the starting point of reconstruction. Guangfu Junior High was able to become an educational park because campus functions were arranged elsewhere, and in-situ preservation was jointly borne by educational institutions and local society. ²

Therefore, 9/21 should not be written as the ending “Taiwan overcame the earthquake.” A more accurate statement is: after a massive disaster, Taiwan left several sites that can be repeatedly visited and established some systems that still require continuous maintenance. When the next earthquake comes, the true test is not whether we remember 1999, but whether those memories have become checks, reinforcements, drills, and mutual care. This is why articles cannot just write 9/21 as “Taiwanese resilience.” Resilience is not an innate personality trait, nor a virtue displaced persons should automatically possess; it requires safety codes, accessible information, reliable water and transportation, communities that can be heard, and institutions willing to admit deficiencies. When official reports, engineering data, media retrospectives, and civil reconstruction experiences calibrate each other, readers have the opportunity to distinguish which repairs are completed and which are merely ongoing commitments. ¹⁵³⁶

Further Reading

Readers can continue to consult original data from Central Weather Bureau’s 9/21 Original Earthquake Report, 9/21 Earthquake Educational Park History, and National Center for Research on Earthquake Engineering’s Technical Report Database. These links correspond to earthquake monitoring, in-situ preservation, and engineering investigation respectively; readers can deconstruct the narrative in the article back into verifiable records.

References