氣候變遷對霧臺潛在大規模崩塌之影響

近年來全球受到氣候變遷之影響，使得強降雨等極端氣候事件頻傳，往往引發嚴重之土砂災害，且大規模崩塌相關議題逐漸受到重視，諸多學者陸續透過崩塌地判釋或利用邊坡穩定分析軟體探討崩塌地，希冀降低災害發生風險。本研究透過土壤溼度指數，根據屏東縣霧臺鄉阿禮村境內的霧臺T002大規模崩塌潛勢區域內之地文與水文因子建置地下水面模型，並以Scoops3D進行穩定分析。本研究透過支援向量機（SVM）之輻狀基底函數（RBF Kernel）配合Hyperopt函式庫找出最佳正規化係數，以建置淨降雨率與地下水位增量、潛在崩塌深度、潛在崩塌總面積與潛在崩塌總體積量之間之回歸關係，之後利用聯合國政府間氣候變化專門委員會（IPCC）發布之第五次評估報告（AR5）當中的15種大氣環流模式（GCM），配合臺灣氣候變遷推估資訊與調適知識平台（TCCIP）所提供之統計降尺度日降雨資料，依據各情境與時期探討大規模崩塌之影響。根據研究結果顯示，當淨降雨率介於410 mm/day至420 mm/day，約莫日降雨量介於931.6 mm至1,007.5 mm時，境內南側聚落附近將生成面積約4.091公頃、且歷史尚未發生崩塌之潛在崩塌區。另在氣候變遷影響下，旱季（11月至4月）最大日降雨量中位數變化隨著時期變遷而有減少之趨勢，並且有79.2%為較基期減少；潛在崩塌平均深度、潛在崩塌總面積以及潛在崩塌總體積量之中位數變化幅度無論何種情境與時期皆不超過0.8%，故對於大規模崩塌之影響程度極低甚至無明顯影響。然而在雨季（5月至10月）時，其最大日降雨量中位數變化為隨著時期變遷而呈現增加趨勢，之中有75%為較基期增加，雨量增幅中位數超過10%者以RCP8.5情境世紀末期占最大宗；潛在崩塌平均深度、潛在崩塌總面積與潛在崩塌總體積量之中位數變化幅度則介於4%以內。由於雨季在氣候變遷下將可能有頻繁之強降雨發生，故未來須防範大規模崩塌災害發生之可能性。

關鍵字

氣候變遷；大規模崩塌；土壤溼度指數； Scoops3D ；支援向量機

並列摘要

In recent years, the climate change has been effected in the world, and could be caused to extreme weather events frequently such as heavy rainfall. As the issue about the large-scale landslide is gradually received attention, many scholars have successively interpreted the landslide features or used some stability analysis software (or model) to probe to reduce the risk of disasters of landslides. In this study, using soil wetness index (SWI) to build groundwater level models by topography and hydrology factors for Wutai T002 in Adiri, Wutai Township, Pingtung County, and then using stability analysis afterwards by Scoops3D. Using the radial basis function kernel (RBF Kernel) in support vector machine (SVM) to find the best regularization parameter with the Hyperopt library, and after that to build the relationship (regression) between net rainfall rate and average increase of groundwater level, mean depth (and total area & volume) of the potential landslide. Using 15 general circulation models (GCM) in the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC), and matching with the AR5 statistical downscaled daily rainfall data which is provided by Taiwan Climate Change Projection and Information Platform (TCCIP), to probe the impact for the potential large-scale landslide by scenarios and periods. The conclusions of the results for this study show that when the net rainfall rate is between 410 mm and 420 mm per day (equal to between 931.6 mm and 1,007.5 mm of the maximum daily rainfall (MDR)), the potential landslide which has never been occurred in history will be generated about 4.091 hectare near the settlement in the southern region of this area. Under the effect of climate change, the variation in the median of the MDR in the dry season (between November and April) will be decreased progressively as periods change, and 79.2 percent of they will be less than the baseline. The change in the median of the mean depth (and total area & volume) of the potential landslide exceed 0.8 percent neither scenarios nor periods, and therefore it extremely low or no obvious impact on large-scale landslide. The variation in the median of the MDR in the wet season (between May and October) will be increased progressively as periods change, and 75 percent of they will be more than the baseline; the median of the MDR in the end of 21st century of the RCP8.5 will be increased by above 10 percent, which is accounted for the largest proportion in each scenario and period. The change in the median of the mean depth (and total area & volume) of the potential landslide are mostly within 4 percent. The large-scale landslide disasters should be prevented due to the possibility of heavy rainfall frequently under the effect of climate change in the wet season in the future.