莫拉克災區土石流發生因子關連性探討

Abstract

莫拉克颱風在民國98年年8月7日至9日侵襲台灣，並在屏東、嘉義、高雄等南台灣山區降下了大量的降雨，因此本研究「莫拉克災區土石流發生因子關連性探討」分析莫拉克颱風期間，土石流可能致災原因、條件，探討莫拉克颱風時受災嚴重之六縣市（南投縣、嘉義縣、台南市、高雄市、屏東縣、台東縣）218條土石流潛勢溪流之土石流發生因子。 本研究參考眾多學者提出之因子，將地文因子分為地形、土砂來源等因子，地形因子部分包括幾何形狀因子、坡向因子、高程因子與坡度因子等，土砂來源主要為集水區內崩塌地因子，降雨因子則採用降雨強度及累積雨量，透過主成分分析、皮爾森相關性分析、發生因子成功率曲線及多變量交叉關連性分析等統計分析方法，分析98年莫拉克颱風災區土石流潛勢溪流發生土石流之致災因子關連性及其顯著因子特性。 本研究實際應用現地調查及莫拉克颱風災害資料，依地形、降雨、土砂來源來探討土石流發生的因子。成果包括 1. 利用主成分分析、皮爾森相關性分析等統計分析法，將因子由101個篩選至18個，並藉由AUC值瞭解單一因子預測土石流，結果發現AUC值最高為0.5670，這代表無法以單一因子有效的預測土石流發生。 2. 接著則採用泡泡圖及三維散布圖探討「集水區平均坡度」及「各潛勢溪流線50公尺環域範圍內之崩塌率」與雨量間相互關係，並歸納出莫拉克颱風時易發生土石流現象之因子區間。 3. 採隨機森林演算法(random forest, RF)針對多變量因子間關連性進行研究與探討，發現莫拉克颱風時，土石流顯著發生因子分別是「有效累積雨量」影響最大，「時雨量」次之，之後依序「各潛勢溪流線50公尺環域範圍內之崩塌率」、「10度以上有效集水區高度平均值」。 4. 由研究結果可知，莫拉克颱風引發土石流最顯著因子是降雨因子，其次才是土砂來源(崩塌地)、集水區高程因子、集水區形狀因子及其它地形指標影響。 5. 此外本研究並以南投縣神木地區作為示範，藉由雨量及歷史災害事件推求神木地區發生土石流事件之下邊界；針對凡那比颱風前後衛星影像進行崩塌地判釋，瞭解凡那比颱風前後崩塌地變化情形；編修4期歷史DEM，探討投縣DF199潛勢溪流由下游至上游的歷年地形變化趨勢。

Typhoon Morakot lashed out on Taiwan during Aug. 7 to 9, 2009. Morakot dumped heavy rainfall in southern Taiwan, especially around the Central Mountain Range in Pingtung, Chia-Yi, and Kaohsiung County. Comprehensive field investigation was carried out by government and private organizations after Typhoon Morakot, and valuable information of debris flow was gathered. In addition, after Typhoon Morakot, the occurrence factors of debris flow become more challenging in Taiwan, and many aspects need to be considered. The scope of this project covered mainly the discussion on the occurrence factors of debris flow in seriously damaged areas which include Nantou, Chia-Yi, Tainan, Kaohsiung, Pingtung, Taitung. A total 218 torrents were included. Field investigation data and disaster records of Typhoon Morakot were utilized to analyze the occurrence factors of debris flow in three aspects: terrain, rainfall, and sediment source. The project’s results were summarized as following.

1. After “principal component analysis” and “Pearson product-moment correlation analysis”, 101 factors were reduced to the most important 18. The highest AUC(Area Under Curve) value under SRC(Success Rate Curve) is 0.5670, which means that single factor could not predict the occurrence of debris flow.

2. Bubble chart and 3D scatter diagram were selected to describe the relationship of average slope of watershed, landslide rate along the stream within the 50m buffer zone, as well as the rainfall during Typhoon Morakot. The above charts were used to summarize the range of factor values which tends to occur in the debris flow in Typhoon Morakot. 3. Furthermore, RF(Random Forest) Theory was utilized to research the relationship with multi-variables. Significant factors which affect the occurrence of debris flow were effective accumulated rainfall, hourly rainfall, landslide rate along the stream within the 50m buffer zone, and average elevation value of effective watershed higher than 10 degrees sequentially. 4. According to the results, the most significant factor is clearly the rainfall factor during Typhoon Morakot. 5. Also, the Shenmu area of Nantou County was selected for demonstration, and rainfall data and historical disaster events were used to find the lower boundary line for debris flow events. Satellite images before and after Typhoon Fanapi were purchased for the interpretation of landslide area. Four historical DEMs were produced to explain the terrain change from the downstream to upstream in DF199 of Nantou County.