Apply Artificial Intelligence, Multivariate Hazards Evaluation Method and GIS in Construction of Vulnerability Assessment Model for Landslide Potential Areas
| Autor: | CHU, SHAO-CHIEN, 朱劭謙 |
|---|---|
| Rok vydání: | 2019 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 107 Global climate change, abnormal extreme weather phenomena have increased, and Taiwan’s geography conditions are poor, geology is fragile, plate movement is frequent, mountainous terrain is steep. Every typhoon and heavy rain is prone to short-term heavy rain, causing sediment disasters on slopes, causing the natural environment of the mountainous area changes. When the rain comes again, the probability of landslide will increase, and the safety of life and property of the residents in the remote mountainous areas will be threatened, which will increase the vulnerability of the settlement residents. Therefore, the establishment of a set of assessment models to analyze and explore the vulnerability of remote mountainous settlements under the impact of landslides is one of the important issues of slopeland disaster prevention. Based on the coverage of parts of Alishan Township in Chiayi County, this study selected a total of 8 typhoons and satellite images of the Formosat II before and after the rainfall incident from 2009 to 2013. The simple fuzzy adaptive resonance theory mapping (SFAM) and genetic adaptive neural network (GANN) of artificial intelligence were employed to classify satellite image interpretation and compare the advantages and disadvantages, to extract the current situation of slopes and information on landslide disasters. In the study, the multivariate hazards evaluation method was used to establish the evaluation model of landslide potential. The relationship between landslide and rainfall recurrence interval was also discussed. In addition, this study uses the five aspects of exposure, sensitivity, adaptability, slope disaster potential and settlement characteristics as the basic framework of vulnerability, and uses GIS spatial data analysis to estimate the vulnerability index and its classification to draw a vulnerability map in the study area. The results of image interpretation showed that the results of GANN and SFAM interpretation of image classification were moderate to medium-high precision, and the results of GANN were better. The evaluation results of the landslide potential show that the slope disturbance factor and geological factors have the greatest impact on the landslide, followed by rainfall intensity, slope and slope roughness. Furthermore, the greater the rainfall recurrence interval, the higher the number of landslide areas after rainfall. In addition, the results of the vulnerability assessment showed that the settlements with high vulnerability were Laiji Village, Dabang Village, Leye Village, Shanmei Village and Xinmei Village, and Laiji Village was the most vulnerable. According to the historical data comparison results, the large-scale sediment disaster data in the disaster-stricken areas have been roughly consistent with the results of the high vulnerability of residents assessed by this study. Therefore, the evaluation results are still credible. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|