Rainfall-induced Shallow Landslide Prediction and Development of Warning Model in Korean Mountain
- Alternative Title
- 한국 산지에서 강우가 유발하는 얕은 산사태 예측 경보 모델 개발
- Abstract
- Shallow landslides and subsequent debris flow cause thousands of deaths and serious economic damage worldwide. The main changing events in the world is temperature and precipitation. South Korea is one of the countries that have been affected by extreme precipitation. Extreme rainfall events are the major driver of landslide occurrences in mountainous and steep terrain regions. The mountainous terrain and highly weathered and fractured rocks contribute excessive mass wast-ing in Korean mountain. Damages caused by landslides have been increasing be-cause of the greater frequency of occurrence of localized heavy rain. To prevent landslide disaster more efficiently, more studies related to predicting the shallow landslide initiation that cause damage and landslide early warning models are nec-essary.
This thesis represents a monograph comprised of the results of comprehensive research work on rainfall-induced shallow landslide prediction and development of warning model in Korean mountain with high-throughput computational and GIS methods and their implementation of regional and local warning systems.
Soil depth is one of the most important and well known factors controlling shallow landslide. A digital elevation model (DEM) based internal relief model was developed to estimate soil depth at high resolution over a mountainous area. This thesis establishes a multi-model procedure for the evaluation of landslide susceptibility on a medium scale. Bivariate and multi-variate models were used to make a consensus landslide susceptibility model with random forest technique. The investigation of combined models for landslide susceptibility assessment, as a means of exploring the optimal zonation, marks a new trend in landslide suscepti-bility and the role of soil depth in landslide susceptibility mapping is comprehen-sively defined.
A regional based rainfall-induced landslide early warning model was pre-pared using the rainfall condition of landslides were reconstructed using spatial distribution of threshold levels after that reconstructed rainfall threshold values were used to correlate with elevation, aspect and AWS proximity. Thus obtained scores were reclassified into four different warning zones. A matrix table was suggested to integrate between magnitude of landslide and possibility of occur-rence of rainfall. The landslide hazard susceptibility on each slope unit will be calculated automatically and updated immediately as the rainfall varies.
Similarly a local based rainfall-induced landslide early warning model was prepared using a hydrological coupled physically-based approach to assess the safety factors in hilly terrain for various rainfall circumstances. . The main differ-ence between the present study and previous approaches is that the rainfall threshold values were incorporated in physically-based model to assess the dif-ferent warning levels. For early warning, an ensemble approach was proposed. The proposed ensemble modelling is the process of running two or more related but different analytical models and then synthesizing the results into a single score. A benefit of ensemble model is that it can incorporate causative factors which are unsuitable in infinite slope model.
Finally, the runout propagation was modelled based on a multiple flow di-rection algorithm in Flow-R. The landslide susceptibility and spreading suscepti-bility were combined using a map algebra function and reclassified into very high, high, moderate and low zones.
This integrated results can be used by both land-development planners and landslide technical specialists, though they may be less helpful at the site-specific scale where geological and conditioning factor heterogeneities may prevail.
- Issued Date
- 2017
- Awarded Date
- 2017. 2
- Type
- Dissertation
- Publisher
- 부경대학교 대학원
- Affiliation
- 부경대학교 대학원
- Department
- 대학원 해양공학과
- Advisor
- 김윤태
- Table Of Contents
- CHAPTER I 1
1. INTRODUCTION 1
1.1. Significance of the problem 1
1.2. Ambition and research objectives of the work 5
1.3. Hypothesis and assumptions of the study 7
CHAPTER II 10
2. SOIL DEPTH MODELLING USING A DIGITAL ELEVATION MODEL AND LAND COVER ATTRIBUTES 10
2.1. Overview 10
2.2. Study area 11
2.3. Methodology 12
2.3.1. Soil depth model 13
2.3.2. Seismic method 14
2.4. Methodology 16
2.5. Model application 19
2.5.1. Mt Hwangyeong 19
2.5.2. Mt Umyeon 26
2.6. Discussion and findings 28
CHAPTER III 30
3. A CONSENSUS MODEL FOR ASSESSING LANDSLIDE SUSCEPTIBILITY MAPPING TO IMPROVE PREDICTION ACCURACY 30
3.1. Overview 30
3.2. Study area 34
3.3. Relevant data 38
3.3.1. Landslide inventory 38
3.3.2. Spatial data-set 39
3.3.2.1. Topographic factors 41
3.3.2.2. Hydrologic factors 42
3.3.2.3. Forest factors 44
3.3.2.4. Soil factors 45
3.3.2.5. Geological factors 49
3.4. Method 49
3.4.1. Revised information value (RIV) 50
3.4.2. Modified frequency ratio (MFR) 51
3.4.3. Logistic regression (LR) 53
3.4.4. Random Forest (RF) 54
3.5. Results 55
3.5.1. Landslide susceptibility using RIV 55
3.5.2. Landslide susceptibility using MFR 58
3.5.3. Landslide susceptibility using LR 59
3.5.4. Ensemble model using RF 64
3.6. Discussion 65
3.6.1. Measurement of performance and comparison of susceptibility models 66
3.6.2. Landslide susceptibility index (LSI) classification 70
3.6.3. Effect of soil depth on susceptibility model 73
CHAPTER IV 75
4. RAINFALL THRESHOLDS AND ITS CALIBRATION FOR POSSIBLE OCCURRENCE OF SHALLOW LANDSLIDE IN KOREA 75
4.1. Overview 75
4.2. Rainfall-induced landslide data catalogue 76
4.3. Methods for the definition of rainfall thresholds 77
4.4. Validation of rainfall thresholds 79
4.5. Results and discussion 81
4.5.1. Rainfall thresholds and warning levels 81
4.5.2. Comparison between the Frequentist and Quantile regression models 84
CHAPTER V 87
5. COMBINING RAINFALL THRESHOLDS AND LANDSLIDE SUSCEPTIBILITY TO FORECAST SHALLOW LANDSLIDE IN BUSAN, KOREA 87
5.1. Overview 87
5.2. Study area 88
5.3. Materials and methodology 89
5.3.1. Landslide inventory 91
5.3.2. Landslide susceptibility mapping 91
5.3.3. Rainfall data collection 92
5.4. Results and discussion 92
5.4.1. Landslide susceptibility model prediction 92
5.4.2. Assessment of rainfall threshold and data reconstruction 97
5.4.3. Integration of susceptibility and warning scores for warning system 99
CHAPTER VI 102
6. EVALUATION OF LANDSLIDE SUSCEPTIBILITY MODEL CONSIDERING EFFECTIVE CONTRIBUTING AREA FOR DRAINAGE TIME 102
6.1. Overview 102
6.2. Study area 104
6.3. Methodology 105
6.3.1. Determination of effective contribution area 105
6.3.2. Landslide susceptibility model 107
6.3.3. Data preparation 107
6.4. Assessment of model results 111
6.4.1. Effective contributing area simulation 111
6.4.2. Causative factor contribution and susceptibility analysis 111
6.5. Discussion 115
CHAPTER VII 120
7. AN ENSEMBLE LANDSLIDE HAZARD MODEL INCORPORATING RAINFALL THRESHOLDS IN KOREAN MOUNTAIN 120
7.1. Overview 120
7.2. Study area 121
7.3. Methodology 122
7.3.1. Rainfall threshold model for initiation of landslide 123
7.3.2. Physically-based model 124
7.3.3. Spatial database 125
7.4. Results and discussion 127
7.4.1. Rainfall threshold based warning levels 127
7.4.2. Ensemble warning model 131
CHAPTER VIII 135
8. SPATIAL MODEL INTEGRATION FOR SHALLOW LANDSLIDE SUSCEPTIBILITY AND ITS RUNOUT USING A GIS-BASED APPROACH IN MT UMYEON 135
8.1. Overview 135
8.2. Study area 137
8.3. Runout model 138
8.4. Results 140
8.4.1. Prediction of shallow landslide susceptibility zones 140
8.4.2. Prediction of propagation routes 141
8.4.3. Integrated landslide susceptibility modelling 143
8.5. Discussion 145
CHAPTER IX 148
9. CONCLUSIONS 148
9.1. General 148
9.2. Synthesis 148
9.2.1. Specific personal contributions 149
9.2.2. Findings 150
REFERENCES 156
ACKNOWLEDGMENTS 177
- Degree
- Doctor
- Files in This Item:
-
-
There are no files associated with this item.
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.