2. 'Worst-case scenario' study for Seattle
• Combination of methods from seismology, eng.
geology, geotechnical eq. engineering
• New: generated broadband synthetic seismograms
including site amplification, to simulate landslide triggering
• Mined rich geologic/geotechnical datasets that already exist
for Seattle
• Just one of many possible scenarios for one city...
3. Bigger picture
• Seismically induced landslide hazard warrants more
attention in Seattle and many other prone areas
• Details of ground motion are important
• Future?
– library of plausible scenarios
– probabilistic hazard maps
– real-time triggered landslide hazard assessment
10. Past seismically induced landslides
in the area
• Worst quakes haven’t happened yet since western
settlement
• Landslides triggered by at least 13 quakes since mid-1800s
• 2001 Nisqually quake - $34 million in landslide damage
Salmon Bay/ Olympia/Cedar River after Nisqually quake (from Highland, 2003)
13. What if Seattle Fault
Earthquake Happened
Today?
Karlin et al., 2004
14. Approach
• Simulate shallow landsliding triggered by a Mw7 Seattle fault
earthquake
• 5m grid, City limits of Seattle
• Newmark method
15. Newmark Method
Ground Motions
q
ac = critical acceleration
a = ground acceleration
ac = (FS -1)gsinq
Factor of safety slope
(from Harp et al., 2006) from
Lidar
Newmark (1965)
16. Fault rupture
• 45km segment
• Blakely et al., 2002 surface projection of frontal fault
• Dips 45o toward south
• Seismogenic from 3 to 15km depth (17km rupture width)
18. Generate Broadband Synthetics
Methods from Frankel (2009)
Long Periods (<1Hz): Short Periods (>1Hz), Stochastic
Finite Difference method using Andy method for each subevent
Delorey’s 3D basin velocity model popping off at rupture velocity
Combine with crossover at 1Hz
19. Realistic broadband seismograms
(Available if you want to use them for other purposes)
N
Distance from fault (surface projection)
M7 Cape Mendocino, Rrup 7km
2
2
S
M7.6 Chi-chi, Rrup 10km
Accelerograms
PEER Ground Motion Database
20. Validation of base synthetics against
2008 NGA attenuation relations
3 km (± 1km)
2 Synthetic seismograms
±1std
1.8 Campbell and Borzognia
1.6 Abrahamson and Silva
1.4 Chiou and Youngs
Mean of NGA
1.2
SA (g)
1
0.8
0.6
0.4
0.2 Directivity
0 pulse added to
NGA relations
0.01 0.1 1 10
(Shahi and
Vs30 = 600 m/s Period (T) Baker, 2011)
21. Basin effects become more apparent
further into basin
15km Synthetic seismograms
±1std
0.8
3D velocity Campbell and Borzognia
0.7 model Abrahamson and Silva
Chiou and Youngs
0.6 Mean of NGA
0.5
Basin effects
SA (g)
0.4
0.3
0.2
0.1
0
0.01 0.1 1 10
Period (T)
Vs30 = 600 m/s
23. Built representative Vs profiles for entire city
Vs (m/s)
0 500 1000
0
Vashon Till
10
20
Advance
30
Outwash
Depth (m)
40
50
60
70 Lawton
80
Clay
90
Pre-Vashon
Example Transfer Function
from ProShake
Equivalent linear methods to
approximate transfer functions
24. Validation of transfer functions
Compared with spectral ratios of Nisqually earthquake
ground motions
25. Max PGA after applying transfer functions
Up to 2g’s
in some
localized
areas
(thin low
velocity
layers over
high
velocity
layers)
Non-linear
effects
moderate
gm in fill
areas
Before After
26. Relating Newmark Displacement
to probability of failure
• One curve exists:
Relative Hazard Zones Jibson et al., 2000
• Based on Northridge
Very landslide inventory
High High
Moderate
• This project: Used
Low
Jibson curve to define
hazard zones and
designate slope
This topic needs work! failures for scenario
Incorporate new post-quake
landslide inventories
27. Validation: Run simulation using Nisqually Ground
26 landslides
Motions, Dry soils
triggered Dearth of landsliding triggered – only in a few VERY steep areas
28. If soils had been saturated = different story
7,500 All soils
completely
landslide saturated
sources to surface
triggered
(Thankfully,
not a likely
scenario)
29. How about the
Seattle Fault
scenario quake?
• Dry soils
• 5,000 landslides
triggered
• Hanging wall and
coastal bluffs most
affected
30. How about the
Seattle Fault
scenario quake?
• Dry soils
• 5,000 landslides
triggered
• Hanging wall and
coastal bluffs most
affected
% of land that fails
31. And if soils are
completely
saturated?
• 30,000 landslides
triggered
• Hanging wall and
coastal bluffs worst
• but inland and North
Seattle affected too
% of land that fails
37. How do these results compare to
static landslide hazard?
– 36% of failed cells are outside City of Seattle's
'Potential sliding areas'
– 28% of failed cells from this study are in low and
medium static landslide hazard zones from Harp
et al. 2006
41. More research needed:
• Relation between Newmark displacement
and probability of failure
• Frequency dependence of landslide triggering
• How to include deep seated slides?
42. •
Summary fault quake
Landsliding triggered by a Seattle
will be extensive and potentially devastating
• Prone areas don't necessarily overlap with static
landslide hazard zones
• Details of ground motion make a huge difference
• Methods can be expanded to more
scenarios, probabilistic, and real-time
applications
Editor's Notes
-Talk about study I worked on with Art and John-look at seismically induced landslide haz in this city-This was a focused study of one scenario – a M7 quake on the Seattle fault
-Considered a worst-case scenario-What makes this study different from others is that we generated broadband seismograms for the scenario earthquake and used those to trigger LS-To do this, we integrated methods-Rich datasets already available-BUT just one scenario earthquake for one city and you may be wondering why do I care? I don't live in Seattle (unless you do)
-And I hope to demonstrate that this proejct we worked on serves to raise awareness, first of all, that seismically induced landslide hazard has the potential to be devastating to Seattle, but has received very little attention relative to other earthquake hazards, and I believe this is the case in other areas as well-I hope to show you that the details of the ground motion, spatial variability and site amplification, make a huge difference in the outcome, so using a single simplified ground motion parameter can severely underpredcit-and finally, that the methods we developed for this project would be fairly easy to expand to more scenarios for various earthquake types and various levels of ground saturation-If we do enoough scenarios and quantify the uncertainties, they can even be expanded to developing a probabilistic seisimcally induced landslide hazard map-and could also be set up to run automatically when an earthquake occurs using ground motions from our ever densifying strong motion networks to identify potential problem areas during rescue and recovery efforts
LS triggered by EQs have been a big problem in many historical earthquakes-direct casualties-block access-disrupt infrastructure and slow down recovery after quake-but not every place – need landslide prone steep slopes, earthquake nearby (shallow crustal), and dense populations
...like SEattle Earthquake sources, subduction, in subducted plate, and in shallow crust above
Southern edge of that basin is the Seattle fault zoneS dipping thrust fault Fault that most directly threatens the cityRecurrence interval not well known, every couple thousand years, last one 900 A.D., past recurrence intervals have been 200 to 12,000 years
Southern edge of that basin is the Seattle fault zoneS dipping thrust fault Fault that most directly threatens the cityRecurrence interval not well known, every couple thousand years, last one 900 A.D., past recurrence intervals have been 200 to 12,000 years
Make things even better, the city is made up almost entirely of unconsolidated soilsSteep in many places, cut by Puget sound, lake washington, rivers – steep landslide prone slopes shown in redParticular troublesome layers = permeable sand over impermeable clay, get water seeping out of hillsides at this contact
And then of course we have our dense population of >600,000 people, important commericial and industrial areas on top of all this
-There are signs of seismically induced landsliding here in the distant and not so distant past–Seds in Lake WA contain slides from at least 7 coherent events in last 3500 yrs, some sunken forests, lakewide turbidity currents-Nisqually eq caused $34 million in damages, dammed rivers, mostly shallow landslides (<3m), not much in Seattle proper-Other previous earthquakes also caused landslidinglandslide deposits from previous large earthquakes are preserved in Lake Washington->M6.8 Nisqually Earthquake 2001 - $34.4 million in damages from landsliding (Highland, 2003)Other Puget sound area eq’s in 1949 (M7.1) and 1965 (M6.5) and others also caused landslides
The most ominous evidence is in the geologic evidence from the last major quake on this faultin the form of the remains of giant block landslides in the bottom of lake washington7 quakes over past 3500 years
many from last quake in 900 AD, black ones on this map are actually forests that slid right into the water
-shallow only, few meters thick-deep seated requires site specific info/detail we don't have/difficult on the scale
since salish didn't have seismometers
Mean magnitude of 4.6, moment and slip distribution modeled as a spatial random field with a correlation length corr to magnitude (mai and beroza, 2002)Rupture velocity is 70% of Vs, slows toward surface. Rupture vel randomized by 25%, never exceed Vs, rake varies by +-20degrees. Rise time depends on shear wave velocity at point of rupture.
-First, put rupture model into a 3D velocity model that includes the Seattle basin developed by Andy Delorey-finite diff code propagates waves through the model and we record it on a fine grid throughout the city-computational limitations restrict us to running it only up to 1Hz-So for the short period portion of the signal, we use a stochastic method developed by Boore where each subevent is assigned a theoretical spectrum which is then multiplied by gaussian noise in the frequency domain to make it more realistic when transformed to the time domain-and then these subevent time series are sent off as each subevent occurs and they are added up at each station. -Then we combine the long and short periods with a matched filter.
Available on 210 meter grid throughout the city if anyone wants to use them
Validation of 1D amplificationCould be improved, but do a surprisingly good job
Soil types were different
-Nisqually didn't trigger much landsliding at all in the city because soils were dry-so we ran the landslide triggering simulation using the Nisqually ground motions as input-and found that it triggered hardly any landsliding-The few landslides triggered were in -shows that our model gives reasonable results
Google earth flyover of results for dry scenarioInstead of individual failures, I show hazard zones for more complete pictureLandslide areas interspersed with private residences upslope and downslopeCritical roads potentially at risk
Herren home in 1997 landslide set1916 Alki landslide
1965 rail damage
Train derailed Dec 18, 2012
We cannot assume that seismically induced landslides will only be triggered in areas already designated as hazardous by studies focusing on static slope stability