f. c. dai and c. f. lee

24
F. C. DAI AND C. F. LEE A Spatiotemporal Probabilistic Modelling of Storm-Induced Shallow Landslide Using Aerial Photographs and Logistic Regression 報報報 報報報 報報報報 報報報 報報報報2010/12/30

Upload: lajos

Post on 11-Jan-2016

60 views

Category:

Documents


1 download

DESCRIPTION

A Spatiotemporal Probabilistic Modelling of Storm-Induced Shallow Landslide Using Aerial Photographs and Logistic Regression. F. C. DAI and C. F. LEE. 報告者:蔡 雨 澄 指導教授:李錫堤 報告日期: 2010/12/30. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: F. C. DAI  and C. F. LEE

F. C. DAI AND C. F. LEE

A Spatiotemporal Probabilistic Modelling of Storm-Induced Shallow Landslide Using Aerial Photographs and Logistic

Regression

報告者:蔡雨澄指導教授:李錫堤報告日期: 2010/12/30

Page 2: F. C. DAI  and C. F. LEE

Varnes (1984) defined natural hazard as the probability of occurrence of a potentially damaging phenomenon

within a specified period of time and within a given area.

Mapping or delineating areas prone to landsliding is essential for land-use activities and management decision making in hilly or mountainous regions.

Page 3: F. C. DAI  and C. F. LEE

Study Area

Page 4: F. C. DAI  and C. F. LEE

Study Area

Mean annual rainfall for the study area over the period 1961–91 is in the range of 2000 to 2400 mm (Lam and Leung,1994).

Page 5: F. C. DAI  and C. F. LEE

Linear Model

i i iy x e +

Page 6: F. C. DAI  and C. F. LEE

Logistic regression

1 1 2 2 3 3

1( 1| )

11

( 1| ) 11 0

1( 1| ) 0

1

ii i i i i i

i i i

i i i

P y x x x xe

P y x

P y x

Page 7: F. C. DAI  and C. F. LEE

Logistic regression

1 1 2 2

1 1 2 2

1 1 2 2

1( 1| )

1 1

1( 0 | ) 1 ( 1| )

1

( 1| )Odds( 1)

( 0 | ) 1

ln Odds( 1) ln1

logit( )

i

i i

i

i

i i i i i

i i i i

i ii

i i

i i i i

i i

eP y x x x

e e

P y x P y xe

P y x Py e

P y x P

Py x x

P

Y x x

1 11 2 21

1 12 2 22

1 13 2 23

1 ( ....)

2 ( ....)

3 ( ....)

11

11

01

10

1

x x

x x

x x

Pe

Pe

Pe

Page 8: F. C. DAI  and C. F. LEE

Data

aerial photographs

The date taken on

spatial scales Landslide occurred

1st 1991/12/30 1:8000

2nd 1992/11/11 1:16000 103

3rd 1993/12/03 1:10000 132

cumulative maximum in any 24 h period

Page 9: F. C. DAI  and C. F. LEE

Data1992/7/18 1993/11/4~5

Page 10: F. C. DAI  and C. F. LEE
Page 11: F. C. DAI  and C. F. LEE

Data

DEM (2m×2m)Slope gradient ( <15,15-20,25-30… 50 ) (degree)

Slope aspect ( 8+1(flat) )

Elevation ( <50,50-100,100-150… 500 ) (m)

Slope shape ( LL, LX, LV, XL, XX, XV, VL, VX, VV )

Page 12: F. C. DAI  and C. F. LEE

Data

Lithology

1:5000 geological maps

Page 13: F. C. DAI  and C. F. LEE

Data

Land cover

(a) developed land

(b) grassed land

(c) shrub–grassed land

(d) forest–shrubbed land

(e) forested land

Page 14: F. C. DAI  and C. F. LEE

Data

Rainfall data 1992/7/18

1993/11/4~5

+

1990/9/11

1992/6/13~14

1990/9/11

1992/6/13~14

Page 15: F. C. DAI  and C. F. LEE

Data

number Value of ln(P/(1-P) )

Landslide grid cells 11955 1

Stable grid cells 12000 0

3000+3000+3000+3000

Page 16: F. C. DAI  and C. F. LEE

Modelling result

Statistical Package of Social Sciences (SPSS)

Page 17: F. C. DAI  and C. F. LEE
Page 18: F. C. DAI  and C. F. LEE

Modelling result

Error matrix

Observed data

occurred

Not occurre

d

Predicted result

occurred

10704 1813

Not occurre

d1251 10187

Accuracy 89.5% 84.9 %

Page 19: F. C. DAI  and C. F. LEE

Model Application

Page 20: F. C. DAI  and C. F. LEE

10-year 20-year

Page 21: F. C. DAI  and C. F. LEE

50-year 100-year

Page 22: F. C. DAI  and C. F. LEE

DISCUSSION AND CONCLUSIONS

For each landslide cell, the maximum rolling 24 h rainfall was designated as the dynamic variable.

The rainfall return periods conventionally used were assessed using data from only one site and should be applied only to that site.

The antecedent rainfall may have some influence on the

occurrence of landslides, but this effect is not accounted for in the predictive model as stated.

Page 23: F. C. DAI  and C. F. LEE

DISCUSSION AND CONCLUSIONS

Land-use planners may differ in the level of risk they can afford or accept. This model allows them to choose their own level of increased risk.

This model has been useful in identifying areas likely to have landsliding in a way that has not been possible previously.

Page 24: F. C. DAI  and C. F. LEE

End

Thank for your listening