Nature

Predicting the Storm, Not the Fate: Taiwan and Typhoons Across Four Centuries

On the morning of 2009's Typhoon Morakot, 71-year-old Lo Pan Chun-mei stood on her second-floor balcony and watched Mount Hsientou collapse, swallowing 462 of her relatives in Xiaolin Village. Fifteen years later, Taiwan has used six AI models to compress its 24-hour typhoon track error from 172 kilometers down to 57. But who within those 57 kilometers goes to work in the storm, and who never receives the warning — that, no forecast can predict.

Nature 氣候與災害

Predicting the Storm, Not the Fate: Taiwan and Typhoons Across Four Centuries

30-Second Overview: On August 8, 2009, after three consecutive days of torrential rain, Mount Hsientou collapsed and swallowed the village of Xiaolin — 462 lives gone. Sixteen years later, Taiwan has compressed its 24-hour typhoon track forecast error from 172 kilometers in 2000 to 57 kilometers in 2025. Every day, FORMOSAT-7 transmits over 4,000 atmospheric soundings; six AI models generate a 30-day warning map in four minutes. But the moment Lo Pan Chun-mei stood on her second-floor balcony and watched her relatives disappear — no radar, however precise, could catch that. We can predict the storm. We cannot predict the fate.

"Xiaolin Is Gone!"

On the morning of August 8, 2009, 71-year-old Lo Pan Chun-mei stood on her second-floor balcony.

The mountain that had been there the day before — Hsientou Shan — had been eating rain for three straight days. It collapsed. A yellow torrent of earth and rock rushed down from the summit, swallowing the streets, the houses, and the indigenous community she had lived in all her life. All 462 of her relatives vanished into the mountains.1

"I cried until I couldn't cry anymore. Now the tears don't come so easily. We have to keep moving forward." — Lo Pan Chun-mei, reflecting on Typhoon Morakot ten years later

That cry of grief is the single most devastating instant in four centuries of Taiwan's struggle with typhoons. On this island, a typhoon was the poem a Penghu official wrote in 1705, the colored flags flying above a Takao harbor rooftop in 1865, the morning Xiaolin disappeared in 2009 — and also the rain in which 81% of workers simply go to work every summer.

Metric Figure
Average typhoons hitting Taiwan per year 3.5
Typhoon contribution to annual rainfall nearly 50%
2024 typhoon agricultural damage approx. NT\ billion (98.88% of annual agricultural disaster losses)
24-hour typhoon track forecast error (2000→2025) 172 km → 57 km
FORMOSAT-7 daily atmospheric profiles 4,000–5,000

The Colored Flags of Takao, to 1,094 Millimeters on Alishan

Taiwan's history of confronting typhoons is older than the Republic of China itself.

During the Qing dynasty's Tongzhi period (around 1865), the British hung wind-and-rain colored signal flags on the roof of the MacKay Hospital in Takao (present-day Kaohsiung) — Taiwan's earliest meteorological warning system. The colors indicated wind intensity levels; ships watching the flags would know whether it was safe to put to sea.2

One hundred and sixty years earlier, in 1705 in Penghu, an official named Sun Yuan-heng witnessed a typhoon firsthand and wrote 85 lines of verse titled "Song of the Hurricane": "Autumn winds rise overnight in raging squall — the hurricane's mother comes from the west, proud in fury."3 Three hundred years later, his lines still get shared in some Facebook post every time a typhoon approaches.

Fast-forward to July 31, 1996. Typhoon Herb struck Taiwan. Alishan recorded 1,094.5 millimeters of rainfall in a single day — the equivalent of Taipei's annual rainfall, dumped in 24 hours.4 It was the highest reading since the weather station opened in 1933. PTT users recalled years later: "Herb directly flooded the ground floor of my home." Another: "My father's fish pond and farmland, most of it washed away."4 Herb was nicknamed the "921 of typhoons," with total damages estimated at NT\–30 billion.

Five years later, in September 2001, Typhoon Nari moved through Taiwan on a bizarre trajectory, stalling for 49 hours. Taipei's weather station recorded 425 millimeters in a single day — the highest in the station's 105-year history.5 The Bannan Line of the Taipei Metro shut down entirely; 16 stations and the depot flooded. Bannan Line section manager Hu Tsung-li fled with his keys and a few hundred thousand in petty cash while all 12 of his stations submerged overnight. Beihu Bishan residents had spent a year building up their recycling program — one typhoon wiped out 150,000 tons of recyclables and paralyzed the entire system.5

From a civil servant's poem in 1705 to flooded MRT stations in 2001, the details recorded are different — but they all record the same event: this island never knows where it will break when the typhoon comes.

The "Guardian Mountain" Is a Pressure Cooker, Not a Shield

Whenever a typhoon approaches from the east, Taiwanese instinctively look toward the Central Mountain Range — averaging over 3,000 meters — which locals affectionately call the "Guardian Mountain," grateful for how it weakens typhoons before they reach the western plains.

Meteorological Agency former forecast center director Wu Derong has publicly rebutted this idea many times.6

"If Taiwan were flat, the rainfall from Morakot would have been 'incomparably smaller.' It is precisely because of the towering terrain that warm, moist air is forced to rise — which pours extreme rainfall onto the windward slopes."

Scientific data backs him up. Typhoon Herb's three-day total rainfall on Alishan reached 1,994 millimeters — nearly two thousand millimeters. Morakot's total rainfall on Alishan broke 3,000 millimeters, a historical record.4 How could that possibly be "blocked by the divine mountain"? It was squeezed out by the mountain. The range turns wind into water and dumps it on the windward side.

📝 Editor's note

The "Guardian Mountain" label is fundamentally the perspective of western plains residents. For mountain communities on the windward side, the Central Mountain Range is a pressure cooker — wind gets squeezed into water and pours down on them. In the same typhoon, western plains residents thank the range for blocking the wind; the mountain communities shoulder the two thousand millimeters alone. This geographic question of "who benefits, who bears the cost" foreshadows the same fault line that runs through every typhoon story that follows.

When the typhoon circulation crosses the range and descends on the leeward side, it often triggers the Foehn effect — hot, dry air that spikes temperatures in Taitung and Taichung, causing severe crop damage.4

Fewer but Fiercer: Every Typhoon a Potential Annual Catastrophe

Taiwan's statistics from 1951 to 2023 show six years with no typhoon forming before May. Between 2020 and 2022, the island recorded three consecutive years with no typhoon making landfall.1

But behind this trend of "fewer typhoons" is a more alarming shift.

The Research Center for Environmental Changes (RCEC) at Academia Sinica, collaborating with National Taiwan Normal University and using the HiRAM high-resolution cloud model, projects:1 by the end of this century (2080–2099), Taiwan may be affected by only 1–2 typhoons per year — but the proportion of Category 4 or stronger typhoons will increase by over 150%, typhoon rainfall intensity will increase by 40%, and landfall wind speeds will increase by 10%.

Metric Present (baseline 1979–2015) Mid-century (2040–2059) Late century (2080–2099)
Annual typhoon count 4–5 3–4 1–2
Category 4+ typhoon share baseline +105% +150%+
Typhoon rainfall intensity baseline +20% +40%
Typhoon landfall wind speed baseline +8% +10%

The research also found that 6.5% of Morakot's extreme rainfall was amplified by anthropogenic climate change.1 Without global warming, Mount Hsientou might not have collapsed on that particular morning. This is a specific number connecting a typhoon directly to global temperature rise — accepted by the scientific community.

"Fewer but fiercer" overturns past disaster preparedness logic. The old assumption: "several will come every year, so distribute resources accordingly." The new assumption: "the whole summer might be quiet, but when one does come, it will carry the destruction of a whole year." One typhoon becomes one annual extreme event.

Storm Chasers: Taiwan Flying Into the Eye

Humanity's tools for facing typhoons have evolved from the colored flags of 1865 to AI in 2025. On that trajectory stands National Taiwan University professor Wu Chun-chieh.

Starting in 2002, he led "Dropwindsonde Observations for Typhoon Surveillance near the TAiwan Region" (DOTSTAR) — Asia's first large-scale typhoon research program. The first operational chase was on September 1, 2003, during Typhoon Dujuan. Through the end of 2012, the team completed 49 typhoon observations across 64 flights, deployed 1,051 dropwindsondes, and flew a combined 334 hours. That first-hand data reduced average 24–72 hour typhoon track forecast errors by 20%.7

Wu Chun-chieh described being inside the eye in the first person:

"I grew up in Taitung, and I've chased typhoons. You first feel the north wind — because typhoons rotate counter-clockwise — then suddenly there's no wind, and that means you're in the eye. A few tens of minutes later the south wind picks up, and that means the eye has passed."

But chasing typhoons is only part of Taiwan's typhoon observation picture. In 1998, Taiwan became the first country in the world to use an Aerosonde unmanned aircraft to observe typhoons.8 After the US banned Aerosonde exports in September 2001, Taiwan became the last country in the world to still own and operate the system.

FORMOSAT-7, launched June 25, 2019, elevated the observation angle from the troposphere to space.9 Six small satellites, in partnership with US NOAA, provide 4,000–5,000 atmospheric profile data points daily — roughly 80% penetrating to below 1 kilometer altitude, twice the coverage of the previous FS3/COSMIC system.

By 2025, the Meteorological Agency integrates data from US, Taiwan, and Japan into 18 track predictions using six AI models. The 24-hour track forecast error fell from 172 km in 2000 to 57 km — a 67% improvement in 25 years. AI models can generate 30-day warning data in 4 minutes, 900 times faster than traditional methods.9

Technology Key Data Disaster Preparedness Impact
DOTSTAR (drone surveillance) 64 flights, 1,051 dropwindsondes 24–72h track forecast error reduced 20%
FORMOSAT-7 (satellite) 4,000–5,000 atmospheric profiles/day 80% penetrate below 1km, 2× previous system
NCU unmanned aircraft IP65 waterproof, 3,000m altitude 1,000+ observation sorties
AI weather models (HuaFeng + 5 others) 30-day forecast in 4 minutes 24h track error from 172km to 57km

From colored flags to a 4-minute 30-day forecast — 160 years of accumulated precision is enough to let the government deploy emergency supplies 72 hours ahead, enough to let farmers harvest their bananas a week early. But precision ultimately lives on a map. The map tells you where a typhoon will make landfall; it cannot tell you who on the street where it lands is still expected to show up for work.

NT\.5 Billion a Day — Who Pays?

Typhoons also gave Taiwan a uniquely local institution: the typhoon day off.

That institution was born from a tragedy. On July 30, 2001, during Typhoon Toraji, Xu Bi-lan, a teacher at Beigang Elementary School in Changhua, fell into a drainage ditch while protecting her students and died. Then-president Chen Shui-bian personally visited the memorial hall. Twelve years later in 2013, the relevant regulations were formally renamed the "Measures Governing Work and School Suspension Due to Natural Disasters."10

"Every time school and work suspension is called for one day, the impact exceeds NT.5 billion."

This figure originated in a 2005 op-ed by former Environment Minister Peng Chi-ming, later recalculated by Three-Three Association chairman Lin Po-feng using 2023 GDP data to arrive at a net daily loss of approximately NT.5 billion.11

But this arithmetic has a structural problem it leaves out. A yes123 survey of 1,330 workers found that 81% had gone to work during a typhoon day, with 65% ordered by their supervisors. An FTNN News survey found that 53.5% of workers received their full pay, but 37.7% received nothing at all.12 Civil servants and office workers waited at home for the suspension announcement; workers in wholesale, retail, agriculture, fisheries, and food service went out into the same typhoon.

📝 Editor's note

The typhoon day story and the "Guardian Mountain" story are actually two versions of the same story. The mountain turns wind into water — whose houses does that water fall on? The suspension announcement gives a holiday to whom, and leaves out whom? In the same typhoon, the rainfall on the map is uniform, but the people bearing the cost are anything but.

The full class breakdown of typhoon day inequality, the blind spots in the NT.5 billion arithmetic, and the situation of migrant workers — these are a separate story, covered in the Typhoon Day article.

The Tribal Weather Station: Ancient Wisdom as the Last Safety Net

Technology is not Taiwan's only way of predicting typhoons.

In the Shendao community of Laiyi Township, Pingtung County, Paiwan tribal elders predict weather changes by observing natural phenomena. A rainbow appearing in the direction of the rising sun means the typhoon will weaken; appearing toward the setting sun, it will draw the typhoon in. Large numbers of crabs crawling onto land, ant nests relocating en masse, earthworms surfacing in large numbers — all are signals that a typhoon or earthquake is coming.13

During Typhoon Morakot in 2009, members of the Kakanami (Shendao) community in Taitung County noticed that the river water had turned murky, warning of landslide risk — and evacuated the entire village in time.13 That year, 462 lives in Xiaolin were not saved by any satellite. But the people of Kakanami survived because of a turbid stream.

The Amis people of Gangkou Tribe in Fengbin, Hualien have their own ocean-reading wisdom. North winds blow before a typhoon arrives; south winds prevail after it passes. If black rocks are covered by large waves, the typhoon will sweep through that area. The place names of the Tao people on Orchid Island encode disaster risk knowledge: "Ji-Rako a Poas" marks a zone of major landslides; "Ji-Igang" marks a flood danger zone.13 The Tao's traditional semi-underground architecture outperformed modern concrete buildings during Morakot and Typhoon Tembin.

National Chengchi University professor Guan Da-wei's research notes that indigenous traditional weather prediction, while less precise than modern instruments, reflects centuries of observing nature and coexisting with the environment.14 In scenarios where AI models fail or edge mountain areas have no network coverage, this knowledge may be the last safety net.

Four hundred years ago, Sun Yuan-heng read the sky and wind in Penghu; three hundred years later, Tao elders read the river's color to assess landslide risk; today the Meteorological Agency uses AI to calculate 30-day warnings. Three systems operate in layers. When the map's precision closes in to 57 kilometers, what actually protects people inside those 57 kilometers may still be a turbid stream.

Xiaolin at Fifteen: From Collapse to Ancient Song

  1. The 15th anniversary of Typhoon Morakot.

Pan Yuan-ming, chairman of the Xiaolin Community Development Association, returned to the memorial shrine. He changed the offering flowers from sunflowers to chrysanthemums — symbolizing long life, expressing the wish that ancestors hope their descendants live well.1

"Because it was a typhoon day, everyone had fear in them, afraid to come back here."

Five Li Pu permanent housing — 90 units built by the Red Cross for Xiaolin survivors — has an occupancy rate of only 30–40%. Livelihood is difficult, and the younger generation keeps leaving.15 But some young people have chosen a different path.

Wang Min-liang (A-Liang), director general of the Nji-Gushan Xiaolin Community Development Association, founded the Damajia Dance Troupe in 2011, leading Xiaolin people through ancient songs and dance to heal from trauma. The troupe won recognition at the 2019 Golden Melody Awards for Traditional Arts and Music.1 Five-year-old Pang Si-chi had listened to her mother singing ancient songs constantly: "I just kept listening, kept listening — I learned it by hearing Mom sing!"

In a 2019 interview marking the tenth anniversary of Morakot, survivor Weng Jui-chi had rebuilt a family with neighbor Yang Mei-lu, who had also lost relatives, in the Five Li Pu permanent housing. He said: "Life has to go on."16

📝 Editor's note

The Xiaolin Village story reveals a paradox of the typhoon era: modernization has moved more people out of risk zones, but cultural roots have frayed along with the move. The Damajia Dance Troupe's attempt to "replant home" through ancient song is answering a more fundamental question — when the physical homeland can never be returned to, where should the cultural homeland be built? This connects to the core idea behind Taiwan's 335 community-based disaster prevention programs: save your own community yourself. Promoted by the Water Resources Agency since 2010, by 2015 this initiative had established 335 communities — from Yilan's Meizhou and Keelung's Dingxiang to Yunlin's Huli Liren — each neighborhood organizing its own early warning and rescue network.17

Typhoons will keep coming. AI models will keep improving. But what Xiaolin's fifteen years tells Taiwan is this: no matter how accurate the technology, reconstruction still depends on the relationships washed apart and then pieced back together — between people, between people and land, between the living and the dead.

Predicting the Storm, Not the Fate

The typhoon track forecast error has been compressed from 172 kilometers to 57.

But on that morning in 2009, the instant Lo Pan Chun-mei stood on her second floor and watched Mount Hsientou collapse, watched 462 relatives disappear from her sight — no forecast precise to any scale could have reached her in time.

We can predict the storm. We cannot predict the fate.

Further Reading

  • Typhoon Day — Same typhoon, public sector workers at home, retail and logistics workers still out. The class fault line hidden in the NT.5 billion arithmetic
  • Taiwan's Climate Crisis and Net-Zero Transition — Behind the 40% increase in typhoon rainfall intensity lies the larger context of global warming and Taiwan's energy transition
  • Taiwan's Alpine Ecosystems and Glacial Relicts — The Central Mountain Range doesn't just redirect typhoon paths — it is home to the world's highest-elevation ecosystems
  • Plum Rain Season — Beyond typhoons, the plum rain season is Taiwan's other major rainy season, equally affected by climate change
  • Outlying Islands and Ocean Culture — The Tao people's traditional architecture and place name knowledge carry unique disaster prevention value in the typhoon era

References

  1. Climate report warns: by century's end, Taiwan may see just 1–2 typhoons per year, but they'll be much stronger — CNA 2024 report, Academia Sinica RCEC simulation of typhoon trends over the next century using HiRAM cloud model, including background on Lo Pan Chun-mei and Wang Min-liang's Damajia Dance Troupe.
  2. Typhoon FAQ — Central Weather Administration official science encyclopedia, complete explanation of typhoon formation, classification, forecasting, and history, including the Takao colored signal flag background.
  3. MOA Knowledge Portal — Old Taiwan's legends of wind and rain — Collects classical typhoon poetry including Qing dynasty Sun Yuan-heng's "Song of the Hurricane" and Cheng Yung-hsi's "Hurricane."
  4. When a typhoon comes, is the Central Mountain Range really the 'guardian mountain?'? Expert: absolutely wrong — Storm Media 2016, in-depth exploration of the Guardian Mountain myth and Foehn effect formation, including Herb's Alishan rainfall data.
  5. MRT at 30: Typhoon Nari flooding was horrifying — TVBS report on the 30th anniversary of Taipei's MRT, reviewing Nari typhoon flooding — Bannan Line shut down for 3 months, with Hu Tsung-li's on-site account.
  6. Using Morakot as an example: meteorologist Wu Derong says the Central Mountain Range is absolutely not a guardian protector — Yahoo News 2016, Wu Derong explains how topographic rain amplifies rainfall using Morakot as an example.
  7. "DOTSTAR" at 20! Wu Chun-chieh discusses Asia's first major typhoon research program — MOST Science Panorama, full record from DOTSTAR's first flight (Typhoon Dujuan, 2003) through the program's 20-year history (2023), including Wu Chun-chieh's first-person typhoon eye description.
  8. Song of the Sky and Sea: Flying deep into a typhoon! Taiwan's unmanned drone sounding team — MOST Science Panorama, records Taiwan's full history from introducing Aerosonde in 1998 to next-generation drones in 2014.
  9. FORMOSAT-7 introduction — National Space Organization official page, explaining how FORMOSAT-7 improves weather forecast accuracy, including AI model 24-hour error data.
  10. Storm Media — How 'typhoon day off' came to be: a tragedy 24 years ago that changed Taiwan's disaster prevention thinking — Traces the origin of typhoon day policy to the 2001 death of teacher Xu Bi-lan.
  11. Updated figures: Huang Yang-ming says one typhoon day off costs Taiwan more than this figure — NOWnews Today, recalculating the economic cost of a typhoon day off using 2023 export data.
  12. FTNN News: Survey — typhoon day: half of bosses give full pay — 1111 Job Bank typhoon day pay survey, source of the 37.7% "no pay at all" figure.
  13. Traditional wisdom reads typhoon paths — accounts from Shendao and Gangkou tribal elders — Indigenous Television Network (TITV), records Paiwan, Amis, and Tao traditional meteorological knowledge, including the Kakanami community's 2009 Morakot evacuation case.
  14. NCCU Humanity Island — facing typhoons and extreme climate: NCCU's Guan Da-wei shares indigenous views of nature — NCCU professor Guan Da-wei shares the complementary relationship between indigenous nature views and modern meteorology.
  15. RTI — Morakot at Ten: The ongoing erasure of Xiaolin Village as it awaits its people's return — Occupancy rate and current conditions at Five Li Pu permanent housing, ten years after Morakot.
  16. In raging waters: Xiaolin Village survivors find small consolations amid great loss — CNA 2019 Morakot tenth anniversary report, documenting survivor Weng Jui-chi and others rebuilding their lives at Five Li Pu.
  17. National Taiwan University Climate and Weather Disaster Research Center — Save your own community yourself — NTU disaster research center introduction to community-based flood disaster prevention, source for the 335 community figure.
About this article This article was collaboratively written with AI assistance and community review.
Typhoon Morakot Herb Nari Climate Change Typhoon Holiday Indigenous Knowledge Wu Derong Wu Chun-chieh Lo Pan Chun-mei AI Forecast FORMOSAT-7
Share