斜坡安全：由一場雨到一套制度

1. Rainstorms are not new: from 1889 to 1948

The earliest recorded case is 29 May 1889, when the Observatory measured 581 millimetres of rain and landslides damaged part of the Peak Tram alignment. There were no geotechnical engineers and no slope register; after the rain the work was road repair and clearance, not asking who was responsible for a slope.

From 14 July 1925, days of rain set off major landslides across the Mid-Levels. Seven houses at Po Hing Fong were destroyed and 75 people died, most of them prominent Chinese merchants. The review that followed still treated the problem as excessive rain; retaining structures were privately built and no one checked the drawings.

On 19 July 1926 a single day of rain set the highest daily total recorded up to 2020. The lesson was not that rainstorms happen but where they land. Official figures after the war make the same point: from 1948 to 2017, landslides killed close to five hundred people in Hong Kong.

2. 1966 and 1972: two disasters that changed direction

On 11 June 1966 three days of heavy rain brought landslides and collapsed buildings across the territory: 64 dead, 48 missing, 29 injured and 6,183 people affected. The disaster showed that the risk was no longer confined to Mid-Levels houses but covered squatter areas and temporary structures.

On 13 June 1968 a slope above Ma Shan Tsuen in Shau Kei Wan gave way, crushing several wooden huts and killing 16 people with 9 injured. Two comparable disasters within two years turned slopes from a site-specific issue into a territory-wide housing risk.

On the evening of 18 June 1972, landslides struck the Po Shan Road area of the Mid-Levels. A debris flow reached Kotewall Court, the building collapsed, an unoccupied block below was also affected, and 67 people died with 20 injured. The chronicle records the consequence directly: the 1972 disaster pushed the Government to take up slope safety.

3. 1977: from relief to regulation

In 1977 the Government established the Geotechnical Engineering Office to regulate geotechnical work, reduce landslide damage and control slope development. The point was not another rescue unit but a single office holding plan checking, inspection and accountability: slopes now had drawings, standards and an owner.

That same year the Observatory and the new office introduced the landslide warning. Its significance was to link rainfall data with slope condition: instead of mobilising after a collapse, the system warns the public when rainfall crosses a threshold.

In 1978 a remote-recording tipping-bucket rain gauge network began operating, sending digital readings to the office control centre over telephone lines every 15 minutes. In November 1979 the first edition of the Geotechnical Manual for Slopes appeared, turning approval and design practice into a common standard.

4. Monitoring and the geological base

Once the system existed, the first task was to watch. Between 1979 and 1990 the office monitored groundwater levels in more than 300 boreholes in the Mid-Levels as part of the Po Shan Road landslide study. Groundwater is a trigger for many failures; without long readings there is no way to tell a stable slope from one that is changing.

In May 1982 the office set up a Hong Kong geological survey unit to produce 1:20,000 geological maps with data and advice. This moved slope work from individual sites to the territory’s geology: the same rock and the same joint sets tend to fail in similar ways wherever they occur.

In 1984 the Government published the Geotechnical Manual for Slopes, which notes that groundwater in Hong Kong is generally perched. Once an internal guide became a public standard, developers, consultants and contractors shared one vocabulary and approvals had a common reference.

5. The 1990s: warnings, monitoring and two large failures

In 1990 a natural slope at Castle Peak in Tuen Mun failed, sending roughly 22,000 cubic metres of debris over about a kilometre. Natural hillsides sit outside building plans, and such failures reminded the authorities that regulating man-made slopes does not cover every risk.

Official records state that from the 1990s the office has continuously carried out landslide prevention measures, and Hong Kong has since avoided landslides causing mass casualties. A large reduction is not zero: on 23 July 1994 a retaining wall below Block D of Kwun Lung Lau in Kennedy Town collapsed, killing 5 people and injuring 3.

From 22 to 26 July 1994 the landslide warning stayed in force for 103 hours, the longest on record. On 13 August 1995 a landslide at Sham Wan Road in Aberdeen reached 26,000 cubic metres, the largest of its kind since Po Shan Road in 1972; in August 1999, 380 households at Shek Kip Mei Estate had to move out because of the slope above them. The risk had shifted from hillside houses to old estates and private slopes.

6. After 2000: natural hillsides and extreme rain

On 7 June 2008 rain triggered more than 2,400 landslides on natural hillsides in remote areas, cutting the North Lantau Highway. The numbers show that in extreme rain the hardest problem is not a register of engineered slopes but thousands of natural hillsides that have never been treated.

In 2009 the office completed the Po Shan drainage tunnel scheme: 2 tunnels and 172 raking drains to lower the groundwater table. The approach shifted from holding debris back to controlling water pressure, so that the trigger does not arise in the first place.

On 21 May 2016 natural hillsides along Sai Wan Road inside Sai Kung East Country Park recorded 34 landslides in total. A 2018 study supplied the basis for prioritisation: for natural hillsides, earthquakes matter markedly less than rainstorms, so resources stay concentrated on the wet season.

7. What remains unresolved

Slope safety today is less about new construction than about maintenance and responsibility: who repairs a slope on private land, when, and to what standard is often a negotiation between owners’ corporations and departments. Such disputes never appear in casualty statistics, yet they are where risk accumulates.

The other open question is climate: whether the same warning thresholds still fit denser extreme rainfall needs case-by-case official review. This dossier stops at the years covered by the chronicle; later updates and reviews are left blank rather than guessed.