Landsat영상과 GIS를 이용한 제주도 산사태 취약성분석
 Alternative Title
 Landslide Susceptibility Analysis Using Landsat Image & GIS Technics in Jeju
 Abstract
 The purpose of the following study is to compare three different methods AHP, LRA and ANN in generation of a landslide susceptibility map in Jeju Island. First, we made the DEM based on the counter line obtained from a 1:25,000 digital map. To evaluate the landslide susceptibility, we also generated a slope map and aspect map from the DEM. Then, we built a geographical information system as well as a soil map, forest map, rainfall intensity map, geological map, land cover map and so on. According to the importance of factors, which had an influence on landslide susceptibilities, we classified the factors and calculated the weight values by using three different methods.
To calculate the weight of each factor by using the AHP method, which has an influence on landslide, we suggested 8 proposals. These 8 proposals were classified according to the important relation between each factor in a landslide occurrence as 8 different cases and the important value of each factor such as slope and aspect that has an influence on a landslide was defined and we calculates the weight values by using the AHP method.
After calculating the weight values, we progressed a geometric correction through a treatment process to a thematic map of the factors, which has an influence on a landslide and the attribute values() of the layer used in overlay, and set up a slope map, aspect map, rainfall intensity map, geology map, soil map, forest map, and land use map. In the process of a calculation, a weighted overlay operation was used for a transformation function() applied to the overlay and layers were piling up one another after the generated model was considered all weights of factors which had an influence on a landslide susceptibility. The attribute value() generated by the overlay was made out a landslide susceptibility map of the study areas and the landslide susceptibility areas were expressed by red and yellow color pixels.
In the logistic method, the landslide occurrence belonged to an independent variable and it was assigned for "0" if no landslide was present or "1" if a landslide was present. Dependent variables contained slope, aspect, soil, geology, forest, rainfall intensity and land cover. The relation between dependent and independent variables could be written as the following equation:
Finally, the landslide susceptibility map was generated by using the above equation and the GIS method.
In the ANN method, the neuron number of the hidden layer was 14 when calculated by the equation suggested by Kanellopoulas and Wilkinson(1997), the number of the input layer was 7, the number of the output layer was 1, the number of the weights between the input and hidden layer was 98, and the number of the weights between the hidden and output layer was 14. Generally 50000 times of circulation were proceeded to achieve a designed error value. Finally, the relative importance of the landslide factors was calculated by the neural network method and the landslide susceptibility map was generated by using the GIS method.
After the susceptibility maps were generated by using the AHP method, the logistic regression method and the artificial neural network method respectively, each area belonging to each level in the susceptibility maps was compared with one another and with NDVI images to find a more dangerous area from susceptibility areas.
To analyze an accuracy of the susceptibility maps generated by three different methods, we used six areas where a landslide had occurred and six areas considered more dangerous after a field survey.
Through the comparison between three susceptibility maps and field data, we finally knew that the susceptibility areas generated by the ANN method was more accurate among three methods.
To inspect an accuracy of the susceptibility areas, we also performed a landslide probability analysis of some susceptibility areas and safety areas in the susceptibility map generated by the ANN method. Finally, we knew that probability values belonging to the susceptibility areas were very high and probability values belonging to safety areas were very low. Thus, it indicated that the susceptibility map generated by the ANN method was accurate.
 Author(s)
 權赫春
 Issued Date
 2009
 Awarded Date
 2009. 2
 Type
 Dissertation
 URI
 https://oak.jejunu.ac.kr/handle/2020.oak/14921
 Alternative Author(s)
 Quan,HeChun
 Affiliation
 제주대학교
 Department
 대학원 토목해양공학과
 Table Of Contents
 CHAPTER 1. INTRODUCTION 1
1.1 Background and objectives 1
1.2 The study trend 3
1.2.1 The study trend in oversea 4
1.2.2 The study trend in Korea 5
1.3 The materials and methods 7
CHAPTER 2. THE BASIC THEORY 9
2.1 The definition and causes of the landslide occurrences 9
2.2 The landslide analysis methods and stages 14
CHAPTER 3. THE GIS DATA IN JEJU 19
3.1 The study area 19
3.2 The GIS data characteristics 20
3.2.1 The topographical characteristic in Jeju Island 20
3.2.2 The slope characteristic in Jeju Island 22
3.2.3 The aspect characteristic in Jeju Island 23
3.2.4 The geological characteristic in Jeju Island 25
3.2.5 The soil characteristic in Jeju Island 29
3.2.6 The forest characteristic in Jeju Island 31
3.2.7 The precipitation and rainfall intensity
characteristics in Jeju Island 35
3.2.8 The land cover characteristic in Jeju Island 39
CHAPTER 4. LANDSLIDE SUSCEPTIBILITY ANALYSIS 42
4.1 The landslide susceptibility analysis by using
the AHP method 42
4.1.1 Theoretical consideration of the AHP method 42
4.1.2 The application of AHP method 46
4.1.3 The landslide susceptibility analysis using
the GIS overlay method and weights 60
4.2 The landslide susceptibility analysis using
the logistic regression analysis method 65
4.2.1 The conception of the logistic regression analysis 65
4.2.2 The application of the LRA method to the landslide
susceptibility analysis 67
4.3 The landslide susceptibility analysis using
the artificial neural network 69
4.3.1 The conception of the artificial neural network 69
4.3.2 Mathematics of the artificial neural network 71
4.3.3 The application of ANN to the landslide
susceptibility analysis 75
CHAPTER 5. IMAGE CHANGE DETECTION 80
5.1 Landsat TM data 80
5.2 Landsat TM image analysis 81
5.3 NDVI image change detection 89
5.3.1 The conception of NDVI 89
5.3.2 The detection of the NDVI changes 91
CHAPTER 6. DISCUSSION 95
6.1 The analysis of the susceptibility areas 95
6.2 The accuracy analysis of the susceptibility areas
by field data 99
6.3 The landslide probability analysis 102
CHAPTER 7. CONCLUSIONS 106
REFERENCES 108
요약 113
 Degree
 Doctor
 Publisher
 제주대학교
 Citation
 [1]權赫春, “Landsat영상과 GIS를 이용한 제주도 산사태 취약성분석,” 제주대학교, 2009.

Appears in Collections:
 General Graduate School > Civil and Ocean Engineering
 Authorize & License

 Files in This Item:

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.