花蓮地區地形東西短南北長,地勢陡急狹長,常受颱風豪雨 影響甚大,且近年來降雨量屢創新高,土砂災害問題頻傳,威脅 居民生命財產安全甚鉅。尤其花蓮縣內之秀姑巒溪、壽豐溪等各 重要治理集水區,更因其地形地勢陡峭,降雨集中坡地遭受崩 塌、沖蝕與土石流肆虐為害,危及花蓮地區重大公共工程建設與 居民生命財產安全更為顯鉅。 鑑於此,本研究乃針對花蓮地區秀姑巒溪及壽豐溪集水區內 131 處土砂災害發生區位之各項調查資料及其致災誘因進行GPS 定位與災害環境調查分析,並蒐集區域內相關自然環境、土砂災 害與衛星影像資料據以建構相關GIS 主題圖層,且加以萃取區域 內各項土砂災害之重要致災因子,進行Pearson 積差相關分析、 X2 獨立性檢定及因子融合性檢定等統計分析後,運用多變量不安 定指數分析、多元線性迴歸分析及倒傳遞類神經網路分析等三種 方法建立適用於花蓮地區之土砂潛勢災害風險評估模式,據以評 估區域環境之不同土砂災害發生風險程度,俾期能提供相關單位 作為花蓮地區土砂災害防治方案研擬之重要參考依據。
The topography of hillslopes located at Hualien county in the eastern Taiwan are quite steep combined with fragile geological structure. Hsiukuluan and Shoufeng watersheds at Hualien County were selected as two study areas in this research. Mechanism, behavior, and scale of these sediment disasters are analyzed to rule out their interaction of the factors mentioned above and the extent of the debris flow. In the study, applications of global positioning systems (GPS) would be necessary used to collect all fundamental informations, including site positioning, disaster scale investigation, and other disaster related factors determination. Geographic information systems (GIS) and remote sensing (RS) techniques are also integrated and used to establish the sediment related disaster database system based on mathematic methods. The data gained from 131 combat engineering structures in the hillslope areas and their surrounding disaster prevention measures were divided into 8 major affective factors by using GPS/GIS/RS integration. Multivariate analysis, multiple linear regression analysis, and back-propagation neuron network were introduced to establish the sediment disaster risk assessment model. All results established by this study can help Hualien County set up its own disaster prevention system to keep well development of city and rural in the near future.