多重災害模擬-崩塌誘發及土石流

Abstract

急促或延時較長的強降雨常誘發崩塌，而坡地土壤中孔隙水壓的舉升及土壤水對土壤材料強度的弱化，為導致邊坡不穩定的主因，若要正確地進行崩塌的預測，則需要一個能有效進行坡地水文收支計算的模式。除了崩塌外，崩落土石集合雨水後所發生之土石流，亦是山區災害肇生的主因，崩塌與土石流往往相伴而生。然而即便現今已有相當多的經驗模式與數值模式可分別對此二類災害進行預測，但卻鮮有一個整合型的模式可同時對此兩類災害進行預測評估，尤其現有之土石流數值模擬多以單一邊坡的土石流事件為對象，少有模式能對整個集水區之土石流災害進行評估，模式的應用也往往受限於資料的品質、代表性的問題，例如土壤資料與降雨分布的資訊。本研究提出一多重災害評估模式ISOLAD (Integrated SOil moisture-LAndslide- Debris flow model)，此為一動態模式，透過整合土壤水模式、崩塌模式以及土石流模式，可對崩塌發生之位置、時機，以及土石流之路徑、影響範圍以及堆積之材積進行估測。此模式可結合雷達降雨資料，增加模式的預測效能，並結合蒙地卡羅模擬方式來進行序率式的模擬，以機率的方式表現災害發生之預測，可更有益於山區集水區的防災決策。 為驗證ISOLAD的有效性，本研究分別對各組成模式一一進行評估。首先，土壤水模式的評估藉由監測一個小型試驗區的土壤水動態與土壤水模式的計算結果來進行比較；其次，於一小區域、單一土石流事件，針對ISOLAD中的崩塌及土石流模式預測能力進行檢驗；前兩階段試驗後，最後則進行集水區尺度的預測試驗。本研究以石門水庫上游之白石溪集水區進行集水區做為ISOLAD之試驗對象，針對過去二場颱風事件所誘發之崩塌及土石流進行預測評估。以2004年艾利颱風來率定模式參數，隨後以2005年海棠颱風來進行驗證。結果顯示，ISOLAD針對崩塌及土石流之影響面積進行預測後，二場颱風事件之正確率均可達80%以上。然而ISOLAD之應用受限於使用資料的特性，例如以不同的DEM網格大小 進行測試後(5公尺、10公尺以及20公尺)，發現當網格大小增加時，模式預測的正確性則降低，尤以對崩塌時機的預測影響為最。ISOLAD針對崩塌及土石流災害提供了一個多重災害之預測評估方式，但若要增進此模式之可應用性，本研究建議必須於不同的地理環境條近下進行更多的試驗。

Landslides are triggered by different causes including intense or prolonged rainfall. An efficient model for evaluating the hillslope hydrology is crucial to landslide prediction. Besides, not only landslides but also rainfall-induced debris flows also lead to devastations. Although many numerical models have been proposed to simulate the physical behavior of a specific debris flow event, current models have rarely incorporated regional landslides and debris flows into a single assessment framework. In addition, applicability of landslide model to watersheds is often impeded by insufficient data quality and data representativity, such as rainfall information and topographic data. ISOLAD (Integrated SOil moisture-LAndslide-Debris flow model) is therefore proposed to predict the development of soil wetness, landslide initiation and debris flow into an integrated modeling framework. Besides, radar data for estimating rainfall patterns and Monte Carlo simulation for evaluating parameter uncertainties are embedded into ISOLAD to assess the coupled landslide-debris flow hazard in mountainous area. ISOLAD involves three major modeling components: the soil moisture model, the landslide model and debris flow model. This study first verified the soil moisture model by a small plot experiment. The landslide model and debris flow model were verified by simulating a past landslide-debris event. Further, ISOLAD was applied to assess the landslide-debris flow hazard for a 120 km2 Baichi watershed located in the northern Taiwan, where landslides and debris flows occurred frequently during typhoon seasons. For watershed scale application, two typhoon events, Typhoon Aere (2004) and Typhoon Haitang (2005), were use to calibrate and validate the ISOLAD respectively. Both two events obtained accuracy higher than 80%. However, the performance of ISOLAD can be affected by input data and DEM resolution (5 m, 10 m and 20 m). It is that the increase of grid size, the decrease of prediction accuracy, especially the prediction of failure timing. The result implies an inherent limit for ISOLAD application using different qualities of input data. ISOLAD is a novel approach for modeling the multi-hazard: landslide initiation and debris flow. However, test over various regions of climate, geology and topography are suggested to improve the applicability of ISOLAD.