運用流槽實驗探討坡度與降水對淺層崩塌機制之影響

Abstract

臺灣受到活躍的板塊作用影響，地質狀態脆弱，且位處季風氣候帶，使得坡地崩塌之災害事件頻傳。隨著氣候變遷所導致之極端氣候日漸加劇，崩塌的發生頻率以及規模都將隨之增加，因此相關的防災、減災及預警之研究顯得更加重要。在過往的研究中，對於坡地崩塌的預警，主要關注在降水與崩塌間的相關性。然而，大多數的研究都為仰賴降水的被動研究，存在變因控制上的問題。另外，受限於監測設備的架設成本，部分研究無法獲得大量數據。若是在實驗室中進行模擬實驗，則可以通過控制坡度、土壤性質、降水強度等變因來主動獲取研究數據，更能夠釐清崩塌發生的機制。 本研究在實驗室中架設了流槽坡地模型，將坡度設定為25°、35°、40°、45°，降水強度設定為50 mm/h、75 mm/h、100 mm/h，試圖以實驗模擬的方式觀察崩塌的發生及相關機制。在模擬過程中，以深度攝影機觀察坡體表面的變化，並且透過埋藏於距土壤表面10–30公分深的振弦式水壓計，測量土壤層中孔隙水壓的變化。最後，藉由降水的控制以及不同坡度的模擬，建立崩塌發生的降水強度降水延時之閾值，以期建立一針對淺層土壤崩塌的預警模型。 在本研究實驗當中出現的崩塌可分為面狀崩塌與槽狀崩塌兩種型態；前者出現於低坡度與低降雨強度條件，而較陡坡度與高強度降雨則更容易形成表面逕流，進而導致槽狀崩塌的發生。在孔隙水壓的觀測結果中發現，在崩塌發生前常伴隨土壤孔隙水壓於10分鐘內上升至少0.5 kPa之現象，顯示此現象具作為崩塌預警前兆的潛力。此外，崩塌區土壤含水量普遍高於其液性限度，顯示達到液性限度為誘發崩塌的重要條件。 降雨強度–延時（I–D）閾值分析指出，坡度越緩或土壤相對密度越高，則需更高強度或更長時間的降雨才能引發崩塌。與統計實測閾值相比，本研究所推導之實驗閾值普遍偏高，這可能與土壤性質、尺度效應或初始含水量等因素相關。 綜上，本研究認為，土壤局部含水不均、孔隙水壓的快速變化與表面逕流的產生，皆為造成坡地發生破壞以及決定崩塌方式的重要機制。此外，以流槽實驗得到之降雨閾值可以和實際統計結果互相參照，惟須注意其可能具有對降雨閾值高估之可能性。這些結果預期將可對淺層崩塌預警系統之建立提供關鍵參考依據。

Taiwan’s geotectonic setting, characterized by active plate interactions and compounded by its location within the East Asian monsoon zone, results in inherently fragile geological conditions. This makes the island particularly susceptible to frequent landslides. As extreme weather events become more severe under ongoing climate change, both the frequency and magnitude of landslides are expected to increase. Consequently, research on disaster prevention, mitigation, and early warning has become increasingly important. Previous studies on landslide early warning have primarily explored the relationship between precipitation and landslide occurrence. However, most of these studies adopt a passive observational approach focused on rainfall data, which limits the ability to clearly understand how different factors interact. Moreover, financial and logistical constraints often hinder the acquisition of sufficient field data. Laboratory-based physical modeling, allows researchers to control variables, such as slope angle, soil properties, and rainfall intensity, and thereby facilitate a mechanistic understanding of landslides. This study employed a laboratory flume model to simulate slope failure under varying conditions. Slope gradients were set at 25°, 35°, 40°, and 45°, while rainfall intensities of 50 mm/h, 75 mm/h, and 100 mm/h were imposed to examine their respective impacts on slope instability. A depth camera was used to monitor real-time deformation on the slope surface, and vibrating-wire piezometers, installed at depths ranging from 10 to 30 cm below the soil surface, were used to record pore water pressure dynamics during rainfall infiltration. The landslides observed during the experiments could be categorized into two types: planar failures and channelized failures. Planar failures predominantly occurred under conditions of relatively low slope angles and low rainfall intensities. In contrast, steeper slopes combined with higher rainfall intensities were more likely to induce surface runoff, which subsequently led to the formation of channelized failures. A recurring phenomenon was observed where a rapid increase of at least 0.5 kPa in soil pore water pressure within 10 minutes frequently preceded the occurrence of slope failure, suggesting its potential use as a precursor for early warning. Additionally, soil moisture content in failure zones generally exceeded the liquid limit, indicating that reaching this threshold is an important condition for triggering landslides. The intensity and duration (I–D) threshold analysis indicated that gentler slopes or higher relative density of soil require greater rainfall intensity or longer durations to trigger landslides. Compared to thresholds derived from actual landslide events, the I–D thresholds determined in this study were generally higher, likely due to variations in soil properties, scale effects, or initial moisture conditions. This study identifies localized moisture heterogeneity, rapid changes in pore water pressure, and the generation of surface runoff as key mechanisms determining the type of landslides. While flume-derived rainfall thresholds can complement field-based statistical results, their potential tendency to overestimate should be considered. These findings offer valuable insights for enhancing the accuracy and reliability of early warnings for shallow landslides.