GIS Application in Landslide Hazard Analysis An Example from the Shihmen Reservoir Catchment Area in Northern Taiwan Chyi-Tyi Lee Institute of Applied Geology, National Central University, No.300, Jungda Rd., Jungli City, Taoyuan County, 32001, Taiwan Email: ct@ncu.edu.tw PNC 2008 Annual Conference in Conjunction with ECAI and JVGC December 4-6, 2008 Hanoi, Vietnam 1
Introduction What is landslide hazard analysis? Stability analysis for a specific slope: Limit equilibrium analysis of a specific slope using strength parameters and hydrologic parameters to obtain safety factor of the slope. Landslide susceptibility analysis: Joining geomorphological, geological, hydrological, and environmental factors to find a linear function which could be used to divide a region into several successive landslide susceptible classes and could best interpret landslide distribution in that region. Landslide hazard analysis: To analyze the probability of landslide occurrence in a given region and in a given time period. Landslide risk analysis: To analyze the amount of lost in a given time period. 2
Introduction What is a landslide hazard map? Yunlin 475-year earth- quake 3
Introduction Method of landslide hazard analysis Deterministic method: Limit equilibrium analysis of a region using strength parameters and a hydrologic model to obtain safety factors for slopes under rainfall of certain return period. Probabilistic method: Using a long-period multi-temporal landslide inventories to analyze the landslide occurrence probability for each slope under certain return period. Discrete probabilistic method: To analyze the spatial landslide occurrence probability and the rainfall temporal probability separately. Landslide occurrence probability is obtained via event-based landslide susceptibility analysis. Temporal probability is obtained by hydrological frequency analysis of rainfall. 4
Introduction Using GIS in landslide hazard analysis Construction of landslide inventories: Using an orthographic satellite image or a airborne multi-spectral scanner image or a scanned and rectified air-photo, we can interpret and digitize a landslide in GIS and construct a landslide inventory. Processing of landslide causative factors: A vector factor, like geological map, may be processed in vector GIS like ArcGIS or MapInfo. A raster factor, like slope gradient, may be processed in raster GIS like Eardas Imagine. Mapping of landslide hazards: Result of landslide hazard analysis is hazard levels or landslide probabilities in a region. GIS should be used to produce a landslide hazard map of the region. 5
Introduction Problems in Shihmen Reservoir Catchment Area The Shihmen Reservoir is an important water resources reservoir in northern Taiwan (Fig. 1). It has a catchment area of 763.4 km2. From 23 to 25 August 2004, typhoon Aere crossed the northern part of Taiwan. The passage of typhoon Aere brought a maximum recorded rainfall of 1,578 mm and a maximum rainfall intensity of 88 mm/hr in the study area. During the typhoon, there occurred numerous landslides in the catchment area and caused the reservoir water become turbid, and the sediments even blocked the water intake and suspend water supply for 20 days. Millions of people and thousands of factories suffered this disaster. 6
Method and Working Procedure for Landslide Hazard Analysis 7
Landslide (area 1.8 ha.) SPOT5 image for extraction of landslides 8
Landslides prior to Typhoon Aere Landslides post Typhoon Aere SPOT5 image taken on 2004/02/10 SPOT5 image taken on 2004/11/02 Landslides extracted from SPOT5 images 9
Landslides triggered by Typhoon Aere 1,624 landslides ( total area 5.84km 2 ) were triggered. Of these, 633 were enlarged or reactivated old landslides. Most observed slope failures were shallow landslides on soil mantled slopes with depths less than 2 m. A GIS layer with attributes was constructed. This inventory will be used for training of a landslide susceptibility model in this study. 10
5m x 5m DEM Plunge Pool for extraction of topographic factors and wetness index Shihmen Dam Shihmen Reservoir Tahan Creek Source of data: Central Geological Survey, Taiwan 11
Geologic Map for extraction of lithology and fault lines SCALE:1/50,000 Source of data: Central Geological Survey, Taiwan 12
SPOT image taken prior to Typhoon Aere for extraction of vegetation covers 13
Rainfall Records for processing maximum rainfall intensity for triggering factor of landslides LEGEND Catchment area Gauge station, WRA Gauge station, CWB 14
Data Processing Both multi-spectral (XS) and panchromatic (PAN) SPOT5 images were used. They were fused into a higher resolution (2.5m x 2.5m) false-color image for landslide interpretation. Multi-spectral image taken prior to typhoon Aere was used for calculation of normalized differential vegetation index (NDVI). Digital terrain model (DEM) of 5x5m resolution was corrected for erroneous data and filtered for noises, and then the DEM was reduced into a 10x10m resolution for further use. Geological map, rainfall data, NDVI, etc. are trnsfered into 10x10m grid data for further use. Slope gradient, slope aspect, slope height, etc. which are derived from DEM are also in 10x10m grid data format. Tools used are MapInfo vector GIS and EARDS Imagine raster GIS. 15
Preliminary Selection of Factors Lithology Slope gradient Slope aspect NDVI Topographic roughness Slope roughness Topographic curvatures Relative slope height Total slope height Distance to road Distance to fault Distance to river bend Distance to river head Wetness index Maximum rainfall intensity Total rainfall 16
Selection of Effective Factors Frequency 4% 3% 2% 1% Non-landslide Landslide D=0.790 0% 0 40 80 120 160 200 Slope, % Exp. Cum. Prob. P-P Plot Obs. Cum. Prob. Prob. of Failure 10% 8% 6% 4% 2% Probability of Failure Curve 0% 0 40 80 120 160 200 Slpoe, % Portion of Landslide 1 0.8 0.6 0.4 0.2 Success Rate Curve AUC=0.767 0 0 0.2 0.4 0.6 0.8 1 Portion of Area Visual inspection of frequency distribution of the two groups, and calculation of discriminator D. Test of normal distribution of the factor. Examination of probability of failure curve to see if landslide probability increases with the factor value. Examination of success rate curve to check the ability of interpreting landslides of the factor. Discriminator D j :, where, A j is average of landslide group, B j is average of non-landslide group, S Pj is pooled standard deviation of two groups, j indicates j th factor. 17
Effective Factors Selected The following factors are tested to be effective in interpreting landslides: Lithology Slope gradient Slope roughness Profile curvature Relative slope height Total slope height NDVI Wetness index Distance to fault Maximum rainfall intensity (triggering factor) 18
Effective Factors SLOPE SLOPE ROUGHNESS 19
Effective Factors RELATIVE SLOPE HEIGHT TOTAL SLOPE HEIGHT 20
Schematic map showing the definition of slope terms A: Elevation of crest B: Horizontal distance to river C: Height relative to riverbed D: Elevation of toe E: Total slope height F: Height relative to crest G: Height relative to toe H: Horizontal distance to crest I: Horizontal distance to toe J: Horizontal distance between crest and toe K: Slope length Relative slope height = Slope height (height relative to toe) Total slope height 21
Effective Factors PROFILE CURVATURE NDVI 22
Effective Factors WETNESS INDEX DISTANCE TO FAULT 23
Effective Factors LITHOLOGY MAXIMUM HOURLY RAINFALL 24
Logistic Regression Data in the Shihmen Reservoir catchment area for shallow landslides triggered by typhoon Aere were used for the analysis. The data set from landslide group contains 38,403 cells, 20,000 cells were randomly selected for building the model and the other 18,403 cells were used for validation. A randomly selected non-landslide data set of similar size was used in the logistic regression and validation. p -1.148 F + 0.805 F + 0.897 F + 0.631 F + 0.127 F - 0.306 F + 0.062F ln = 1 2 3 4 5 6 7 1 p +0.277 F + 0.126 F - 0.685 F + 0.074 F - 0.747 F +1.304 F - 2.617. 8 9 10 11 12 13 Occurrence probability p is taken as susceptibility index in this study. F 1 : sandstone and shale unit, F 2 : indurate sandstone and shale unit, F 3 : argillite unit, F 4 : quartzite and argillite unit, F 5 : slope gradient, F 6 : NDVI, F 7 : slope roughness, F 8 : profile curvature, F 9 : total slope height, F 10 : relative slope height, F 11 : wetness index, F 12 : distance to a fault, F 13 : maximum rainfall intensity. 25
Landslide Spatial Probability Maps Typhoon Aere Rainfall 100-year Return Period Rainfall 0.1 Probability of Failure, P ls 0.08 0.06 0.04 0.02 0 P ls λ = 0.007587( ) 1 λ 0.6403 0 0.2 0.4 0.6 0.8 1 Susceptibility Index, λ 1 Success Rate Curve Portion of Landslide 0.8 0.6 0.4 0.2 0 Logistic AUC=0.8561 0 0.2 0.4 0.6 0.8 1 Portion of Area 26
Model Applications The landslide hazard model could be used for the prediction of future landslides providing a scenario rainfall distribution is given. The 100-year return period landslide hazard map could be used for general purposes, like for regional planning, engineering site selection and hazard mitigation policy making. The landslide hazard model could also be used for sediment prediction of a catchment area. 27
Sediment Prediction (1/3) From probability of landslide occurrence map, we may estimate shallow landslide area A sl as follow, A = ap sl ls i i where, P ls i is probability of failure at cell i, a is area of a cell; it is 100 m2 in this study. From spatial landslide probability map of typhoon Aere event, we can estimate landslide area to be 28,928,800m2, Actual landslide from the inventory is 2,872 landslides with a total area of 21,053,000m2. The difference indicates that mapping of the landslide inventory may have missed some landslides in shadows of the SPOT images. Furthermore, missing of small-scale landslides and deleting of repeated landslides in the event-based landslide inventory may also reduce the total area of landslide as compared to the prediction ones. 28
Sediment Prediction (2/3) We further adopted a soil-thickness prediction model from Chung (2008) as follow, h = 4.9429 1.0939ln S where, h is soil thickness (m),s is slope ( ). Inserting slope data into above equation, we can get soil thickness at each grid-cell in the study area and further estimate landslide volume V sl as follow, V = (494.29P 109.39P ln S ) sl lsi lsi i i Using the above equation, we could estimate landslide volume induced by typhoon Aere to be 25,510,700 m 3. Actual landslide volume using landslide area from inventory is 17,996,290 m3. Again, the difference is because of we have missed small-scale landslides, landslides in shadows and repeating landslides. 29
Sediment Prediction (3/3) To this step, we can successfully predict landslide location, area and volume in a drainage basin or catchment area using GIS. However, the amount of sediment yield in a catchment still requires estimation of soil erosion on the slope and sediment transportation in the stream. A distributed hydrological and sediment transport modeling should be further carried out so that sediment problem like that in the Shihmen Reservoir catchment area could be interpreted. GIS is a useful tool for managing and processing the model factors and also good for construction of a hazard map and application in regional planning, hazard mitigation, and sediments yield estimation. 30
Thanks for your attention! 31