Presentation on theme: "Modeling typhoon- and earthquake-induced landslides in a mountainous watershed using logistic regression Kang-Tsung Chang, Shou-Hao Chiang, Mei-Ling Hsu."— Presentation transcript:
Modeling typhoon- and earthquake-induced landslides in a mountainous watershed using logistic regression Kang-Tsung Chang, Shou-Hao Chiang, Mei-Ling Hsu Geomorphology 89 (2007) 335–347 報告者 : 蔡雨澄 指導教授 : 李錫堤 報告日期 :2011/06/09
Landslides occur when unstable rock and soil masses on slopes are disturbed by earthquakes, intense storms, human activities such as road construction, or a combination of these factors. Many researchers have used logistic regression to predict probabilities of landslide occurrence by analyzing the functional relationships between the instability factors and the past distribution of landslides.
We used landslides triggered by a typhoon in 1996 and a major earthquake in 1999 to build a typhoon-induced model and an earthquake-induced model, respectively.
Study area and Data
Area:92km 2 Elevation:770~2850m Slope:0 ° ~77 ° Hoshe basin Hoshe basin Topographic map (left) and oblique aerial photograph (bottom right) of the study area
Geological map of the study area Four Miocene lithologic formations Hoshe (Hs) Formation Nanchuang (Nc) Formation Kueichulin (Kcl) Formation T Terrace Two sets of strike–slip faults WNW NNE Climate: subtropical Mean annual temperature:16°C Mean annual precipitation:2300mm
Chi-Chi earthquake 1999 /9/21, magnitude of 7.6 epicenter: about 60 km to the northwest of the study area
Wetness index: ln(a/tan β) a: contributing area tan β:slope
Distribution of landslides triggered By typhoons and earthquake.
Methods- logistic regression
The logit model from a logistic regression has the following form: y : dependent variable x i : the i -th explanatory variable a : constant a bi : the i -th regression coefficient e : error term. The logit of y is the natural logarithm of the odds: p : the probability of the occurrence of y p/(1−p) : the odds To convert logit( y ) back to the probability p, it can be rewritten as:
The numbers of landslide cells for each event are as follows: Herb:783 Chi-Chi:4213 Toraji: 6294 Mindulle: 2161 Haitang:1146 Random sample: 500 landslide cells and 500 stable cells for developing each model Ran logistic regression analysis in the Statistical Package of Social Sciences (SPSS)
Results and Validation
TyphoonEarthquake ROC0.7170.758 Area concordance76.974.5 (Overlapped landslide area / Total landslide area on digital map)*100%
LL = (distance to channel)/(slope length) distance to channel slope length)
Typhoon-triggered landslides tended to be near stream channels and earthquake-triggered landslides were more likely to be near ridge lines. A major earthquake such as the Chi-Chi earthquake could still affect the spatial location of typhoon-triggered landslides 6 years after the event. By covering both types of landslides in a natural experiment, this study takes a step closer to understanding one of the most destructive natural hazards.