碩士 / 國立中興大學 / 水土保持學系所 / 103 / The Jing-Shan River is a tributary of Da-An River watershed, which is located at Li-Yu-Tan reservoir in the downstream. The facility functions as not only an electricity generator, flood preventer, farmlands irrigator as well as a tourist attraction but also the water supplier of Miao-Li and Tai-Chung district. Recently, the torrential rain that come with typhoons and extreme weathers has caused many landslides in the watershed and severely shrunk the capacity and usability of reservoir.
This study used the inventories of landslide established by Central Geological Survey as the landslide data. Logistic regression, Instability index method and Support vector machine (SVM) were selected to establish the landslide susceptibility models and obtain the landslide susceptibility maps in the upstream areas of Jing-Shan River. Ten landslide causative factors were first chosen as the landslide causative factors, according to the previous studies. A selection procedure was then performed to efficiently reduce the number of factors. Further, the receiver operating characteristic curve was used to evaluate the accuracy of model results.
As a result, Logistic regression and Instability index method both show that the terrain roughness is a critical factor on the susceptibility value. The instability index method is not only lead to possible underestimation around the river side but also the number of factor classification can impact the result. SVM establish the model by classifying the landslide data. The landslide susceptibility values is not prone to particular factors. Therefore, the results of model prediction is not influenced by the weights of factor. Furthermore, the landslide susceptibilities is classified into four ranks, including: low, medium, medium-high, and high. SVM and logistic regression is suggested to be superior to Instability index method due to recognize the landslides located at the medium-high susceptibility areas. The analysis of area under the curve (AUC) showed AUC of 0.825 for SVM, 0.721 for the logistic regression, and 0.718 for the instability method. This further suggests SVM is a preferred method among the others in assessment of landslide in the research areas.
| Identifer | oai:union.ndltd.org:TW/103NCHU5080033 |
| Date | January 2015 |
| Creators | Yu-Ting Wen, 温祐霆 |
| Contributors | 詹勳全 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 88 |
Page generated in 0.0166 seconds