|
|
|
Finn Krogstad |
|
University of Washington |
|
|
|
|
|
Soil, rain, vegetation, topography,... |
|
|
|
|
|
|
Rearranging gives: |
|
|
|
SHALSTAB - Critical Rainfall |
|
dSLAM -
Factor of Safety |
|
LISA - Probability of Landslide |
|
|
|
which are all correlated with landsliding |
|
|
|
|
Classifying hillslopes by |
|
local slope and curvature: |
|
SMORPH - Hazard |
|
|
|
|
Hazard based management says:
Slope 1
is less stable than slope 2. |
|
|
|
What we really want to know:
Is plan 1
better than plan 2? |
|
|
|
Is plan 1 okay? |
|
|
|
|
|
|
|
Soils |
|
Storms |
|
Traffic |
|
Vegetation |
|
Slope |
|
... |
|
|
|
|
|
|
Input road properties |
|
|
|
|
|
Output erosion |
|
|
|
|
|
|
|
Average over variability |
|
Explicitly |
|
WEPP |
|
Implicitly |
|
SEDMODL |
|
USLE |
|
|
|
We need to do the same thing with Landslides. |
|
|
|
|
|
Use observations of landsliding on similar
hillslopes to estimate landslide fraction or ‘probability’ |
|
|
|
p = #landslides / #hillslopes |
|
(if #hillslopes is large) |
|
|
|
|
|
|
|
|
|
What is a ‘Hillslope’? |
|
Slides/Acre |
|
Slides/Hectare |
|
Slides/Grid square |
|
What is ‘similar’? |
|
Topography |
|
management |
|
|
|
|
|
Each hazard and activity |
|
phn=Lhn/Hhn |
|
|
|
|
|
If |
|
pmr=Lmr/Hmr |
|
Then |
|
Lmr=pmr*Hmr |
|
|
|
|
|
If you add a spur road and harvest: |
|
What is the landslide probability before? |
|
What is the landslide probability after? |
|
The difference is due to management activity. |
|
The sum is the expected number of landslides. |
|
|
|
|
|
Combining Independent Events |
|
Pr(A or B) |
|
|
|
|
|
|
|
Combining any number of independent events |
|
|
|
|
Landslides Flow Downhill: |
|
Pr(blowout at X)=Pr(landslide upslope of X) |
|
|
|
|
Bigger table: more cells: more uncertainty |
|
More dimensions: More uncertainty |
|
|
|
|
|
Many hillslope have continuous properties |
|
Slope |
|
age |
|
retention |
|
contributing area |
|
soil cohesion |
|
root reinforcement |
|
... |
|
|
|
|
|
|
|
Regression is a powerful tool: |
|
y=a+bx+e |
|
|
|
Probability takes [0,1]
values |
|
|
|
|
|
Computers handle the math |
|
|
|
|
|
|
Fitting parameters from past landslides |
|
|
|
Allows prediction of future landslides |
|
|
|
|
|
Current landslide models poorly guide management
decisions. |
|
If we consider average landsliding: |
|
Estimate from past landsliding |
|
Predict future: |
|
landscape average landslide numbers |
|
probability of local event |
|
Logistic regression adds flexibility |
|
Notes