碩士 / 國立彰化師範大學 / 地理學系 / 106 / Sediment-related disasters triggered by intense rainfall occur frequently in Taiwan due to steep slopes and uneven spatial and temporal distribution of rainfall. However, Taiwan is still lack of timely nationwide early warning systems for sediment-related disasters to prevent the loss of life and damage to property. The Japan Meteorological Agency applies Soil Water Index (SWI) in sediment-related disaster warning system to issue timely sediment-related disasters warnings. Many studies indicated that using logistic regression model for landslide susceptibility analysis could be used to predict the location of possible landslide effectively. Nevertheless, it still unknown whether this method is suitable for estimating time of landslide occurrence. This study compared SWI with other triggering factors for their ability to predict the occurrence time and location of sediment-related disasters, and discussed the problems when predicting the time and place of occurrence of sediment-related disasters with logistic regression model. The study area is the whole area of Taiwan Island. We collected rainfall data, topographical data, and disaster records for the Typhoon Morakot（2009）. Firstly, we selected predisposing factors to predict the landslide location and used triggering factors to predict the time of landslide occurrence. To develop the temporal model, we adopted the Wald forward to calibrate the predisposing factors combination of optimal spatial modified success rate（MSR）, and selected the best triggering factors based on the prediction ability. Next, we combined these two factors to calibrate the model parameters of the best success rate of time prediction, the range of allowable error is three hours for the success criteria. The spatial model was developed directly from logistic regression model. Then, we compared the spatial MSR and the success rate of time prediction in two models and used the data from Typhoon Sinlaku（2008）to verify the predicted results. The results of this study showed that the predictive ability of SWI is better than other triggering factors. For Typhoon Morakot, the MSRs in temporal and spatial model are 71.09% and 83.33%, respectively; the success rate of time prediction are 28.44% and 22.02%, respectively. These two models shows similar success rate in the estimation of temporal and spatial distribution of sediment-related disasters and are both underestimated the landslide susceptibility. In addition, for Typhoon Sinlaku, the spatial MSR in temporal and spatial models are 53.88% and 69.23%, respectively, and the success rate of time prediction are 7.69% and 0% respectively. In summary, logistic regression model is capable of predicting spatial distribution of sediment-related disasters, and SWI can be used to predict temporal susceptibility. Nevertheless, the integration of this two models still can not predict the both of location and time of the disaster.
| Identifer | oai:union.ndltd.org:TW/106NCUE5136010 |
| Date | January 2018 |
| Creators | Lin, Ji-Wei, 林繼煒 |
| Contributors | Chen, Yi-Chin, 陳毅青 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 85 |
Page generated in 0.0125 seconds