Investigation of catastrophic deep-seated landslides and landslide dams in Taiwan ~ Lessons from Matai'an landslide dam disaster, 23rd Sep. 2025

On 3 April 2024, M_w 7.4 earthquake struck Hualien County. A total of 779 landslides were recorded following the earthquake, affecting 433.93 hectares. So, when excessive rain courtesy of Typhoon Wipha arrived in July, which triggered a landslide blockading a flushed Matai’an creek. The resulting 200-meter-tall barrier lake could hold up to 86 million cubic meters of water, and intense rainfall brought by super typhoon Ragasa triggered the overtopping of the landslide dam at Matai’an in Taiwan on 23 September 2025. The flood resulted in 19 deaths. From the disaster zone, 717 were rescued, 157 injured, and 5 remain missing. Over 8,000 people encompassing 3 villages were directly impacted by this incident. In certain parts of Guangfu, the receding waters left behind sludge up to 2 meters deep. What happened in Guangfu Town is a significant compound disaster example. It challenges the present warning, forecasting and response system of debris flow. New concepts and new procedures are necessary to cope with the compound disasters triggered by extreme heavy rainfall.
The remote location of the landslide dam and the lack of road access, as well as the soft soil and rocks of the dam body, and ongoing landslides in the surrounding area all pose difficult restrictions. In addition, the risks of typhoons, heavy rainfall, and earthquakes all need to be taken into consideration. All of these factors make machinery access to the construction site very difficult and highly hazardous. The government had installed rain gauges, water gauges, and CCTV surveillance cameras around the dam area. These are linked to the downstream water level station and surveillance images of the river to keep abreast of the latest conditions at all times. However, due to the steep terrain of the dam crest, the lack of access roads, and mountainous climate conditions, the installation of the instruments has proved challenging. The rainfall hydrograph shows long-duration, high-intensity, high-accumulation and large-extent characteristics. It suggests the correlation of each disaster type with the rainfall characteristics by reference to the report of eyewitness memory. The causality between those in Guangfu Town occurred disasters could then be deduced. In order to characterize the disaster and suggest a strategy, it is necessary to try to rebuild the temporal order and spatial distribution of the disaster processes. 
This study describes briefly each single disaster, the relationship among those disasters and the approach to rebuild the disaster process by field investigation, topographic survey and satellite image processing. The results serve to intensify the disaster prevention system of debris flow and shallow landslide which then could be applied to the warning system of deep-seated landslide and landslide dam. The derivative issues and the approach to compound disaster prevention are suggested. The related discussions, evaluation and assessment are also summarized as the reference of further tasks.