台灣因山區地形陡峻,加上近年來平原地區之開發又趨飽和,坡地之開發利用已是必然趨勢。近年又受極端氣候的影響,每逢颱風或暴雨來襲,在山區便容易因集中性降雨而引發崩塌現象,造成山崩或是土石流等災害,進而伴隨著嚴重的生命財產損失。因此,適切地發展山崩潛勢評估模式,以及探討降雨對崩塌地分佈特性的影響,對山坡地災害防治實為重要之課題。 本研究運用基因演算自動演化類神經網路技術於四個不同年度颱風降雨前後之研究區衛星影像的判釋分類,藉以擷取地表與崩塌資訊,並利用多變量不安定指數建置山崩潛勢評估模式。在山崩潛勢模式中所考量的影響因子包括「坡度」、「坡向」、「高程」、「地質」、「距水系距離」、「距斷層距離」、「坡地擾動程度」、「坡度粗糙度」與「有效累積雨量」等。研究中,亦透過地理資訊系統之空間分析工具,及數值高程模型,萃取研究範圍之山脊與水系,藉以探討崩塌空間分佈特性。 影像判釋結果顯示,八幅不同時期衛星影像之一致性係數Kappa指標平均達0.77,具有中高程度的精確度。另由多變量不安定指數所建立之評估模式可知,有效累積雨量與地質因子之影響最大,距斷層距離、坡向、土地擾動程度、高程及坡度之影響次之。 研究結果亦發現,不論那一次颱風降雨後,研究範圍之崩塌數量皆明顯增長;整體來看,2009年莫拉克颱風降雨後之崩塌地分佈相當平均且有多處較大面積的崩塌,但相較於莫拉克颱風,其它三年度降雨事件(凡那比颱風、南瑪都颱風及蘇拉颱風)後之崩塌分佈較偏向山脊,且規模較小,而在鄰近溪流之崩塌形況則是較大型之崩塌情況,但數量較少,但其規模較大。此外,可能因植被復原完整與否之故,莫拉克颱風後,接續之2010年凡那比颱風亦造成研究區許多中型崩塌,然2011年南瑪都颱風及2012年蘇拉颱風後,研究範圍則大多是新增小型的崩塌。而且,四段不同時期降雨前後崩塌分佈結果皆顯示,有效累積雨量與崩塌數量大致上呈正比之趨勢。
About two-thirds of Taiwan’s total area is covered by mountains and hills. Furthermore, due to technological and economic development, the original plain area in Taiwan has been nearly fully developed and human development has extended to the hillside area. Human development coupled with the global impact of extreme weather, typhoons and heavy rains caused the hillside disaster. The scope and impact extent of the damage are more serious than ever before. For this reason, evaluation of rainfall-induced landslide potential is essential to disaster prevention. This study employed genetic adaptive neural networks in the classification of high-resolution satellite images before-and-after four typhoons from which data related to surface conditions and hazard log in slope areas were derived. Meanwhile, using the weighting ratios of various disaster causing factors inferred from the multivariate dangerous value method, this study explored the degree of slope land disturbance. Then, this study incorporated the relationship between rainfall and landslides to draw a landslide potential map using the discriminant analysis approach combined with the geographic information system platform. The slope, aspect, elevation, geology, distance from the river, distance from the fault, slope disturbance, slope roughness, and effective accumulative rainfall are used as fundamental factors in this research. The spatial characteristics of landslide distribution were explored. Results of image classification show that the values of coefficient of agreement for eight different time periods are at intermediate-high level. The predicted potential of landslide is in reasonable confidence level. This research employs multivariate dangerous value method to estimate the weight of factors. The result shows that weights of effective accumulative rainfall and geology are the highest, followed by weights of distance from the fault, aspect, slope disturbance, elevation, and slope. The findings of this research also show that the amount of landslide in research area increases despite the rainfall of typhoons. In general, the distribution of landslides after Typhoon Morakot in 2009 is very uniform, accompanied by some large scale landslides. The distribution of landslides after the other typhoons (Fanapi, Nanmadol, and Saola) tends to near the ridges with small scale. Larger scale landslides but with small amount occurs near rivers. Medium-scale landslides occur after Typhoon Fanapi in 2010, whiles small-scale landslides occur after Typhoon Nanmadol in 2011 and Typhoon Saola in 2012. This may be due to if the vegetation is restored. Moreover, the landslide distribution after aforementioned typhoons shows that the amount of landslide is proportional to the effective accumulative rainfall.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。