Barberopoulou519

Full text

519
Bulletin of the Seismological Society of America, Vol. 96, No. 2, pp. 519–535, April 2006, doi: 10.1785/0120050090
Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget
Lowland: Observations and Modeling
by A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Abstract Analysis of strong-motion instrument recordings in Seattle, Washing-
ton, resulting from the 2002 M
w
7.9 Denali, Alaska, earthquake reveals that ampli-
fication in the 0.2- to 1.0-Hz frequency band is largely governed by the shallow
sediments both inside and outside the sedimentary basins beneath the Puget Lowland.
Sites above the deep sedimentary strata show additional seismic-wave amplification
in the 0.04- to 0.2-Hz frequency range. Surface waves generated by the M
w
7.9 Den-
ali, Alaska, earthquake of 3 November 2002 produced pronounced water waves
across Washington state. The largest water waves coincided with the area of largest
seismic-wave amplification underlain by the Seattle basin. In the current work, we
present reports that show Lakes Union and Washington, both located on the Seattle
basin, are susceptible to large water waves generated by large local earthquakes and
teleseisms. A simple model of a water body is adopted to explain the generation of
waves in water basins. This model provides reasonable estimates for the water-wave
amplitudes in swimming pools during the Denali earthquake but appears to under-
estimate the waves observed in Lake Union.
Introduction
Seismic waves produced by the M
w
7.9 Denali, Alaska,
earthquake of 3 November 2002, initiated a series of water
waves that damaged at least 20 houseboats along the shores
of Lake Union and Portage Bay in Seattle, Washington, at
an epicentral distance of 2400 km (Table 1; Barberopoulou
et al., 2004). Observers reported a series of waves, or runup
of water on the shore on a calm day, at the approximate
arrival time of the surface waves from the Denali earthquake.
The houseboats are routinely subjected to waves from large
ships and windstorms, so the damage during the earthquake
suggests water waves with unusual amplitudes or periods.
In Table 2 we summarize water-wave observations in Wash-
ington state during several large distant and local earth-
quakes to show that seiching is a recurrent phenomenon in
Washington.
Seiche is a common term used to describe the standing
oscillations setup in a lake or harbor due to a perturbation
of the water level (Russell and Macmillan, 1952). Seismi-
cally induced seiches fall into two categories: (1) free seiches
initiated by displacement of water due to fault motion or
landslides (Russell and Macmillan, 1952; Ichinose et al.,
2000); and (2) forced seiches caused by the passing of seis-
mic waves. The latter category, which we focus on in this
article, can initiate oscillations either through the tilting of
the water body or from the horizontal motion of the sides
(Kvale, 1953; Russell and Macmillan, 1952; Donn, 1964;
McGarr, 1965; McGarr and Vorhis, 1968; Barberopoulou et
al., 2004; Cassidy and Rogers, 2005). McGarr (1965) argues
that for small bodies of water the angle of tilt is so slight
that horizontal motions of the sides of the water body are a
much more effective wave generator.
The unusual water waves observed in the Puget Low-
land of Washington state following the Denali earthquake
were apparently initiated by large seismic surface waves di-
rected preferentially along the west coast of North America
by the earthquake source mechanism (Eberhart-Phillips et
al., 2003; Barberopoulou et al., 2004; Cassidy and Rogers,
2005). The concentration of water wave reports above the
Seattle sedimentary basin (Barberopoulou et al., 2004), sug-
gests that local amplification of seismic waves by the basin
may have played a role.
Unusual water waves in Washington state have coin-
cided temporally with seismic-wave arrivals from past re-
gional and distant earthquakes. Unlike previous earthquakes
known for generating seiches (e.g., the 1964 Alaska earth-
quake), the Denali earthquake was well recorded on a variety
of seismometers worldwide and has provided an opportunity
to examine ground motions that initiate seiches in water bod-
ies at large distances. Although documented damage from
seiches during past seismic events has been minor in Wash-
ington state the question remains whether amplified long-
period waves by sedimentary basins could cause high-
amplitude water waves during large earthquakes on local
crustal faults or on the Cascadia subduction zone. Examples

520 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Table 1
Reports of Unusual Water Activity in Washington* on 3 November 2002
Location
Duration
of Water
Activity in
Seconds
(Unless
Otherwise
Noted)
Period of
Oscillation
in Seconds
(Unless
Otherwise
Noted) Damage Reported
Horizontal
Motion
Runup
or
Amplitude
of Vertical
Oscillation Other Comments/Observations
1 Portland, OR Several
inches
Floating walkway was moving up and down by
several inches.
2 Seattle, WA 18–23 2–3ft Water at the Volunteer park reservoir oscillated
20–30 ft crest to crest about 8–10 times in
3 min. Observer appears to have noticed
2 different types of water motion.
3 Lake Wenatchee
State Park, WA
5 ft Lake shoreline water surged 5 ft up the beach.
4 Seattle, WA
†
8–10 ft Barge in a shipyard along with other vessels
ranging from 1000 to 5000 tons experienced
a violent sway.
5 Bellingham,
WA
1 min 5 Reporter specified rocking of boat was in the
north–south direction.
6 Seattle, WA
†
Houseboat sewer,
phone, and
power lines
were broken.
Five houseboats disconnected from dock. Dock
moved few inches.
7 Seattle, WA
†
1 min Houseboat slammed
against pllings,
lost phone, sewer,
and power lines.
Mooring attachments
damaged or thrown
out of alignment.
Similar damage was noticed by one resident during
the 1964 Alaska earthquake but not during the
Nisqually event.
8 Blaine, WA Several
minutes
1 ft Swimming pool sloshed back and forth.
9 Bellevue, WA Swimming pool sloshed over its sides.
10 Bothell, WA 20 Water moved up and down and canoe fell off
the dock into the water.
11 Silver Creek,
WA
11–12
inches
Extreme water motion not accompanied by felt
ground motion. Violent shaking of the lake was
compared with stirring like shaking a jug of water.
Lake was calm before and after the event.
12 Bayview, ID 15 min 20 inches Docks moving laterally.
13 Seattle, WA Few min Swimming pool closed for the winter splashed.
14 Seattle, WA
†
30 Rocking of boat.
15 Woodinville,
WA
Water splashed out of the pool in the north–south
direction.
16 Kirkland, WA Sloshing observed in swimming pool.
17 Bellevue, WA Water sloshed over the swimming pool.
18 Seattle, WA 2–5 min 1 ft Seattle University Center: two pools showed unusually
large water waves. Swimmers experienced dizziness
and symptoms like sea sickness. People in other
parts of the building experienced nothing.
19 Everett, WA Sloshing observed in swimming pool.
20 Seattle, WA 10
inches
Sloshing observed along the length of the pool in
city of Seattle pool.
21 Seattle, WA
†
Houseboats
slammed against
dock pilings with
water and sewer
lines damaged.
Bystanders commented on this different wave
action compared with that from a passing ship
or caused by winds.
22 Everett, WA 30 Extreme shaking of boats back and forth.
23 Seattle, WA
†
45 Houseboats unfastened from their docks (heavy
appliances fell out).
24 Stanwood, WA Water motion in Lake Martha was compared with tidal.
Some swirling also observed.
(continued)

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 521
Table 1
Continued
25 Bainbridge
Island, WA
Water moved back and forth and flowed out of the
swimming pool.
26 Forest Grove, OR 3–4 ft Henry Haag Lake had water waves of 3–4 ft moving
across lake.
27 Clyde Hill, WA Pond flooded, with 2 ft waves present.
28 Bellevue, WA
†
Boats moored at
marina were
violently moved
against the
dock structures
(minor damage
at the docks).
Boat moving violently. Could not stand still on it
without support.
29 Bellevue, WA Water at swimming pool sloshing back and forth on a
very sunny calm day without any wind.
30 Snoqualmie,
WA
1 min Thin layer of ice covering Melakwa Lake broke drawing
the attention of hikers from the cracking sounds.
Water rose and fell.
31 Seattle, WA 3 min 6
inches
Pool on the 35th floor of a steel/concrete building
started sloshing back and forth in the north–south
direction.
32 Bainbridge
Island, WA
Sloshing observed on the swimming pool which
overflowed.
33 Seattle, WA 180 15 4–5 ft Docks moved violently in a north–south direction
with a lot of noise.
34 Seattle, WA 60 Boat in marina moved.
35 Snoqualmie
Pass, WA
Observer noticed water moving back and forth
carrying ice that had broken from the icy surface.
36 Seattle, WA 30–60 3 ft Boat started moving sideways. No vertical displacement
noticed.
37 Seattle, WA
†
Observer on a boat in Lake Union, moved while
observing many other boats moving back and forth.
38 Seattle, WA 5 min Water sloshed back and forth resulting in swimming
pool overflowing.
39 Seattle, WA
†
4–5
inches
Observer noticed water runup on the shore as the
level of the water rose and fell (kayak with observer
moved east–west). 4–5 cycles of complete oscillation
were noticed.
40 Des Moines,
WA
2–8 min Water in fish bowl started swaying along with other
things in the room, like hanging plants.
41 Des Moines,
WA
⬍30 Boat slammed
against dock.
No wind, water quiet when suddenly the boat slammed
against the dock.
42 Seattle, WA 15ⳭBoats bumping
against docks.
Boats moving back and forth. No wind was present.
Shaking was quite violent.
43 Everett, WA 15 Boats in the shed moved sideways.
44 Everett, WA 60 Sloshing observed on swimming pool.
45 Mt. Rainier
National Park,
WA
Several
minutes
Small logs
washed on
the shore.
3 ft On Blue Lake near Mt. Baker, motion of the water was
compared to water in a bathtub (on southwest–
northeast direction).
46 Clinton, WA 30 4 ft Observer on top of a boat noticed a sudden movement
of boat and what appears to be 4 ft of runup on shore.
47 Lopez Island,
WA
3
inches
Concentric wave pattern on Mud Bay, Lopez Island.
48 Marysvllle, WA 5 min Dock swayed sideways (east–west); a boat stretched
its mooring lines.
49 University
Place, WA
5 Movement in pond observed.
50 Centralla, WA Reservoirs of capacity of 3.5 and 4 million
gallons oscillated.
51 Ione, WA 1 ft Waves on Long Lake near Spokane.
52 Kenmore, WA 5 min
at least
Door swang closed, river water sloshing in the
north–south direction, big boat banging.
*Few other locations outside of Washington are also listed.
†
Locations in Lake Union or Portage Bay.
‡
Locations in Lake Washington.

522 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Table 2
Reports of Unusual Water Waves in Seattle during other Earthquakes
Event Observations
29 November 1891
Port Townsend earthquake
Lake Washington, on the east side of town, was lashed into foam and the water rolled onto the beach two feet
above the mark of the highest water and eight feet above the present stage (The Oregonian, 29 November 1891;
Holden, E. S (1898). A Catalog of Earthquakes on the Pacific Coast 1769–1897, Smithsonian Miscellaneous
Collections, 1087).
Yakutat Bay 1899
Alaska earthquake
A 6–10 foot upheaval was observed in the center of Lake Chelan. Waves rolled toward the shore but no wind was
blowing at the time (Dow, 1964).
18 April 1906
(San Francisco,
California, earthquake)
“Lake Washington Feels The Shock” Surface of water on west side violently agitated.
• Floats tossed about.
• Three distinct tidal waves reported to have come from northeast.
Seattle got a seismic disturbance, too, though it was quite short and luckily did hardly a dollar’s worth of damage.
About 6 o’clock yesterday morning Lake Washington on the west shore was agitated so violently that house boats,
floats and bathhouses were jammed and tossed about like leaves on the water. Beyond breaking a few moorings
and causing a small-sized fright, no damage (Seattle Post-Intelligencer, 19 April 1906). A man, who was on the
bank of the Wishkah River early in the morning, said that waves five feet high rolled up the river from the sea,
and that he was sure that there was some disturbance (Seattle Post-Intelligencer, 25 April 1906).
M7.1 13 April 1949
Olympia earthquake
A crew towing two gravel barges reported the disturbance from their boat (The Seattle Times, 14 April 1949).
A diver in the Seattle Port of Embarkation, working on the propeller shaft of an Army transport reported the
ship “jumping.” Water tanks in Snohomish overflowed.
M9.2 27 March 1964
Good Friday
Alaska earthquake
Effects felt in many states, including Washington (McGarr and Vorhis, 1968) where many houseboats
were damaged.
M6.5 28 April 1965
Seattle-Tacoma earthquake
Green Lake in North Seattle was “sloshing back and forth like soup in a shallow bowl” (The Seattle Times,
30 April 1965).
of such damage to docks and ships, coastal inundation, ero-
sion of coastal areas, damage to floating bridges, or trigger-
ing of landslides on eroding bluffs has occurred in past earth-
quakes in other areas (e.g., Korgen, 1995; Ruscher, 1999).
The 1959 M7.1 Hebgen Lake, Montana, earthquake was
responsible for damage to the Hebgen dam when a seiche
caused water to overtop the dam several times. The 2003, M
8.3 Tokachi-oki earthquake in Japan caused sloshing in oil
and naphtha tanks, resulting in sparks that started a fire that
burned for several days (Koketsu et al., 2005).
Seismic waves from the Denali earthquake recorded on
the Pacific Northwest Seismic Network (PNSN; Fig. 1) show
substantially larger amplitudes and longer durations within
the Seattle basin (Fig. 2) (Barberopoulou et al., 2004) than
at surrounding sites. We use spectral ratios of these record-
ings (Fig. 2) to examine the amplification of the shear and
surface waves from the Denali earthquake caused by the
Seattle basin and the shallow glacial deposits inside andout-
side the basin. To estimate the influence of these amplified
seismic waves on the overlying water bodies we model the
generation of water waves by the amplified ground motions
from the Denali earthquake using a simple 1D model (Lamb,
1932; Russell and Macmillan, 1952; Proudman, 1953;
McGarr, 1965; Wilson, 1972). Although simple, the model
successfully reproduces the character of waves observed in
swimming pools during the Denali event but probably un-
derestimates the observed water motions in Seattle-area nat-
ural waterways. We also discuss the apparent absence of
seiche reports during the Nisqually earthquake. We use the
2001 M
W
6.8 Nisqually earthquake ground motions to drive
the model because the Nisqually-type events are the region’s
most common type of damaging earthquakes.
Puget Lowland Geology
The Puget Lowland of Washington state is part of the
Puget-Willamette forearc basin above the subducting Juan
de Fuca oceanic plate (Crosson and Owens, 1987; Galster
and Laprade; 1991). The Seattle basin is one of three large
sedimentary basins beneath the Puget Lowland, the others
being the Everett and Tacoma basins (Fig. 1) (Johnson et
al., 1994, 1996; Pratt et al., 1997; Brocher et al., 2001). The
Seattle basin is a 30 km by 60 km depression in the Eocene
volcanic basement rocks, filled with up to 9 km of low-
density, low-velocity sedimentary rocks and unconsolidated
sediments (Johnson et al., 1994; Pratt et al., 1997; ten Brink
et al., 2000; Brocher et. al., 2001). The south end of the
Seattle basin is formed by the Seattle Fault zone, an east-
trending reverse or thrust fault separating thick sediments to
the north from shallow bedrock and thin sediments to the
south (Johnson et al., 1994; Pratt et al., 1997; ten Brink et
al., 2000, Brocher et al., 2004). The Seattle Basin is bounded
on the west by the Olympic Mountains, but its eastern
boundary is not well constrained. To the north the basin
sediments thin onto the antiformal Kingston Arch (Johnson
et al., 1994; Pratt et al., 1997; ten Brink et al., 2000). The

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 523
Figure 1. Map showing the strong-motion stations
of the Pacific Northwest Seismic Network (PNSN) that
were operating during the Denali earthquake. Stations
GNW and ERW (top and left of figure) are the bedrock
sites we used as reference stations for the spectral
ratios. Contours show the outlines of the deep sedi-
mentary basins and the dash-dotted contour the ap-
proximate edges of the Quaternary glacial deposits
that underlie the Puget Lowland (Booth, 1994).
Everett and Tacoma basins to the north and south are less
well known but also appear to be fault-bounded basins of
comparable size and age to the Seattle basin (Van Wagoner
et al., 2002; Johnson et al., 1996, 2004; Snelson et al., 2000;
C. M. Snelson et al., unpublished manuscript, 2006).
The upper portions of the sedimentary basins are Qua-
ternary glacial deposits that also extend across the Puget
Lowland outside of the basins (Fig. 1) (Booth, 1994; Jones,
1996). These glacial deposits were laid down during a series
of Quaternary ice-sheet advances and retreats, the most re-
cent retreat being 14,000 to 12,000 years ago (Thorson,
1980). The glacial strata consist of advance deposits that
were compacted when overridden by the ice and uncom-
pacted recessional deposits that have never been covered by
glaciers. The deep valleys (Puget Sound, Lake Washington)
were carved by subglacial meltwater (Booth, 1994).
Sedimentary basins in the Puget Lowland are docu-
mented to amplify seismic waves at periods of 0.14 to 10
sec (Frankel and Stephenson, 2000; Frankel et al., 2002;
Pratt et al., 2003; Pratt and Brocher, 2006). Impedance con-
trasts and resonance within the basin sediments explain, in
part, the observed amplification, but locally generated basin
surface waves likely play an important role in the long du-
rations of ground motion. Pratt et al. (2003) measured the
amplification across the Seattle basin relative to bedrock
sites in the Olympic mountains, concluding that the maxi-
mum spectral amplification is about 16 at a frequency of
about 0.33 Hz (3-sec period). They could not separate the
effects of the deep basin from those of the shallow deposits
because of their limited distribution of seismographs.
It has been suggested that basin subsidence results in
the concentration of water bodies over the sedimentary ba-
sins of the Puget Lowland (Fig. 1) (Pratt et al., 1997). The
Tacoma basin underlies the many broad arms of southern
Puget Sound. The Seattle basin underlies sections of Hood
Canal, Puget Sound, Lake Washington, and Lake Samma-
mish. The Everett basin underlies Puget Sound where it
widens near the Strait of Juan de Fuca. This concentration
of water bodies over basins that amplify long-period seismic
waves could explain the many observations of water waves
during large earthquakes as noted by McGarr and Vorhis
(1968).
Amplification of Seismic Waves from the Denali,
Alaska, Earthquake by the Deep Seattle Basin
Strata and Shallow Glacial Deposits
Data
To measure the effects of local geology on the ampli-
tudes of the seismic waves, we calculated spectral ratios (SR)
of the Denali shear and surface waves recorded at nonbed-
rock sites, relative to shear and surface waves recorded at
bedrock sites ERW and GNW. Data consisted of three-
component recordings from 46 PNSN strong-motion stations
distributed around the Puget Lowland, both inside and out-
side the Seattle sedimentary basin (Fig. 1). The data ana-
lyzed in this article were recorded by force-balance accel-
erometers with flat response to acceleration for the frequency
range 0–50 Hz (K2 and Guralp instruments). PNSN broad-
band stations also recorded arrivals from the Denali earth-
quake, but nearly all these instruments (many colocated with
a strong-motion instrument) clipped during the surface-wave
arrivals and were therefore not useful. The strong-motion
instrument recordings showed an emergent but clear S-wave
arrival, but dispersed surface waves were by far the largest
waves (Fig. 2). Basin sites consistently had the greatest
amplitudes and longest duration (300 sec; Fig. 2). Large-
amplitude surface waves radiating to the southeast from the
epicenter resulted in long-period effects across a wideregion
of Canada and the United States, similar to those from the
1964 Good Friday, Alaska, earthquake (Cassidy and Rogers,
2005; Eberhart-Phillips et al., 2003; Barberopoulou et al.,

524 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Figure 2. Seismograms from bedrock and basin sites, with traces scaled to equalize
their maximum values (acceleration units are m/sec
2
). The traces show the east–west
component of motion recorded at stations SEA (Seattle basin), SMNR (nonbasin) and
SP2 (Seattle basin). Note the extremely large surface-wave amplitudes. The maximum
acceleration was about 5 cm/sec
2
, and the maximum displacement was about 20 cm
(40 cm peak-to-peak). On the right are acceleration spectra of these traces showing the
dominant frequency in the signal to be at about 0.08 to 0.1 Hz (10- to 12.5-sec periods).
Figure 3. Plots of signal-to-noise ratio for shear
and surface waves shown in Figure 2 and reference
sites ERW and GNW. The signal-to-noise ratio is
above 2 for the frequency range of 0.01–0.8 Hz at
most of the sites except at rock sites, where the signal-
to-noise ratio approaches 2 at about 0.3 Hz. Thus,
spectral ratios taken with respect to the average of the
two rock sites are reliable to frequencies up to 0.3 Hz.
Between 0.3 and 0.8 Hz the spectral ratios are con-
taminated by noise in the rock signals, indicating that
the computed amplifications will be underestimated
because noise will make the rock reference signals
artificially large.
2004). Maximum acceleration on strong-motion records was
5 cm/sec
2
(0.004g) at basin sites and 1.5 cm/sec
2
at bedrock
sites.
The recorded waves provide a frequency spectrum for
analysis of 0.01–0.5 Hz (2- to 100-sec period), with the larg-
est amplitudes being at frequencies of 0.08–0.1 Hz (10 to
12.5-sec periods; Fig. 2). To avoid results that are unduly
influenced by a single reference station, our spectral ratios
are computed relative to the average of the spectra from two
bedrock sites. Reference site GNW is located on Green
Mountain near Bremerton on intrusive bedrock (Eocene
mafic igneous rock) and ERW is on Fidalgo Island near An-
acortes on volcanic basement rock (Mesozoic intrusive rocks
and associated metamorphosed strata). Relative to a time
window before the P-wave arrivals, the signal-to-noise ratio
for both shear and surface waves at the two bedrock refer-
ence sites (ERW and GNW), which show some of the lowest
signal strengths, is greater than 2 at frequencies between
0.01 and 0.3 Hz (3.33- to 100-sec periods; Fig. 3). The
signal-to-noise ratio at other sites, which generally have
larger amplitudes than the reference sites, exceeds 2 up to
frequencies of about 0.8 Hz (1.25 sec periods; Fig. 3). Thus,
the spectral ratios we derive are reliable between 0.01 and
0.3 Hz (3.33- to 100-sec periods). Between 0.3 and 0.8 Hz,

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 525
noise at the reference sites likely will cause the spectralratios
to underestimate the amplification factors.
Processing
The spectral ratios we computed include the effects of
the source parameters, wave path, station location, and in-
strument response (e.g., Hartzell, 1992). For teleseisms, the
source and path effects can be considered the same for
nearby sites because the differences in radiation angle and
path are minor (azimuthal angles for the Denali earthquake
ranged between east-southeast (127⬚) and east–southeast
(130⬚) degrees). The SR technique assumes that the hard-
rock site does not materially affect the amplitude spectrum
of the waves, and the frequency spectra at the rock sites are
therefore characteristic of the input source spectrum. Under
these assumptions, the SR technique gives the amplification
factor at the site of interest.
We chose a 70-sec time window to compute the spectra
of the shear-wave arrivals, with the window length being
limited by the arrival of the surface waves. A 300-sec win-
dow starting at the Love wave arrivals was used for surface-
wave analysis; this time window covered the largest-ampli-
tude arrivals (Fig. 2). Data were tapered with a 5% Hanning
taper before computing the spectra. Both horizontal com-
ponents of the seismograms (vector sum of the spectra) were
used in the calculation of the spectral ratios. The spectral
ratios were smoothed with a three-point mean smoothing
algorithm. Because we have recordings from stations on
both deep-basin (20 sites) and shallow glacial deposits (15
sites), we have the opportunity, unlike previous studies, to
characterize the seismic-wave amplification in these two set-
tings. We divided the sites into three categories to differ-
entiate the amplification effects caused by the shallow and
deep sediments: (1) sites over the basins (more than 2 km
of strata), (2) sites on glacial deposits and older sediments
outside of the deep basin (less than 2 km of sedimentary
strata), and (3) sites where bedrock is very near the surface
(estimated to be ⬍20 m based on proximity to bedrock out-
crops). We classified our sites as “deep-basin” if they lie
over the deep portions of the basins defined by the contours
in Figure 1. All our deep-basin seismograph sites lie on the
Seattle basin. Seismographs at non-basin sedimentary sites
lie on glacial deposits and Tertiary sedimentary rocks out-
side of our deep-basin contours, meaning the total thickness
of sediment with a P-wave velocity less than 3.5 km/sec is
about 2 km or less based on tomography and gravity mea-
surements (Van Wagoner et al. 2002). We used the 3.5 km/
sec contour as representing strong, consolidated sedimentary
rocks lying near crystalline basement. Bedrock sites are
known to lie on only a thin (⬍20 m) cover of till or sediment
above metamorphic or volcanic rock.
Results
The most obvious result evident from our spectral ratios
is the large amplification of seismic waves above about
0.04 Hz at all nonbedrock sites (Figs. 4a, b and 5). In con-
trast, sites on shallow bedrock other than bedrock reference
sites ERW and GNW show little or no amplification except
above 0.4–0.5 Hz, where noise levels are starting to be a
factor (Fig. 4c). All sites on glacial deposits, both withinand
outside the deep basins, show strong amplification (5 or
more) of both shear- and surface-wave arrivals above 0.1 Hz
relative to our bedrock reference sites. The peak amplifica-
tion occurs at 0.2–0.7 Hz, where amplification values often
reach 10 or more. Lack of station coverage does not permit
a thorough study of Everett or Tacoma basin effects, but
surface-wave amplification reached 10 at the few sites near
the Everett basin (Figs. 1 and 4a, b).
Although rock site spectral ratios are clearly distin-
guished from those at sedimentary sites, the distinction be-
tween the deep-basin sites and the nonbasin sedimentary
sites is more subtle. All the sites on glacial deposits show
little amplification below 0.04 Hz (25-sec period and
greater), but between 0.04 Hz and 1 Hz the sedimentary
strata substantially amplify seismic waves compared with
bedrock. The influence of the deep-basin sediments is seen
on the basin spectral ratios (Fig. 4a) as a distinct amplifi-
cation at and below 0.1 Hz that is missing from the nonbasin
sedimentary sites (Fig. 4b). Graphs of the average amplifi-
cations for the three classes of sites (Fig. 6) show that be-
tween 0.04 and 0.2 Hz the deep-basin sites have about twice
the amplification as the other sedimentary sites. These results
indicate that the relatively shallow (⬍2 km) glacial deposits
cause most of the observed amplification above 0.2 Hz, and
the deep-basin strata increase the amplification in the 0.04-
to 0.2-Hz range.
The amplifications we document are in agreement with
previous studies of amplification by the Seattle basin (Pratt
et al., 2003), which also show peak shear-wave amplifica-
tions in excess of 10 at frequencies of 0.3–0.5 Hz (2- to 3.33-
sec period). Pratt et al. (2003) noted that the largest ampli-
fications were over the east side of the basin. Our results are
consistent with this observation (Figs. 2, 5), but we have few
permanent stations on the west side of the basin to rigorously
test this hypothesis (Fig. 1). Potential explanations for the
seismic-wave amplification can be found in Pratt et al.
(2003) and include resonance, focusing, surface waves, and
the low impedance of the basin sediments. The spectral ra-
tios we present here also are consistent with inferencesabout
basin amplification noted elsewhere (e.g., Joyner, 2000). For
high frequencies (⬎1 Hz), in general, it is assumed that the
effects of the shallow strata dominate the site response, thus
causing sites on glacial deposits outside of the deep basins
to show amplification. Both the shallow- and deep-basin
strata influence the site response at low frequencies through
long-period resonance, focusing, and surface waves (e.g.,
Joyner, 2000; Pratt et al., 2003).
The large concentration of water-wave observations in
the Puget Lowland on top of the basins suggests that the
water bodies are especially susceptible to ground motions in
the 0.04- to 0.2-Hz frequency range. The maximum ampli-

526 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Figure 4. (a) Deep-basin sediment sites. Plots of spectral ratios for all stations (shear
and surface waves for basin sites) (Seattle and Everett) with respect to the average
spectrum of bedrock sites GNW and ERW. The yaxis corresponds to amplitude of
spectral ratios. (continued)
tudes of the Denali arrivals occurred in this frequency range
(Fig. 2b). The Puget Lowland may be especially susceptible
to seiches during large earthquakes because of the low-
frequency seismic-wave amplification by sedimentary basins
and other sedimentary structures, plus a concentration of wa-
ter bodies above the basins.
Seismically Induced Water Waves
Observations from the Denali Earthquake
Long-period seismic waves from the Denali earthquake
initiated large waves in water bodies in western Canada
(Cassidy and Rogers, 2005) and the United States (Table 1;
reports to the PNSN and National Earthquake Information
Center for the Pacific Northwest). Water was reported to
have surged 1.5 m (5 feet) on the shoreline at Lake We-
natchee in the Cascade Mountains. Broken ice and logswere
carried and washed over the shores of several lakes in Wash-
ington state. Standing oscillations (seiches) or a series of
large waves were set up in other water bodies where floating
walkways started moving and boats were slammed against
docks.
Some of the most prominent water waves triggered by
the Denali earthquake were in Lake Union, in Seattle, Wash-

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 527
Figure 4. (continued). (b) Nonbasin glacial sediment sites. Plots of spectral ratios for
all stations (shear and surface waves for nonbasin sites) at sites outside of the Seattle or
Everett basins that are located at sedimentary sites. Spectral ratios are computed with
respect to the average spectrum of bedrock sites GNW and ERW. The yaxis corresponds
to the amplitude of spectral ratios.
ington, and in an arm of Lake Union named Portage Bay
(Table 1) (Barberopoulou et al., 2004). Lake Union is a shal-
low, Y-shaped lake with the main body having dimensions
of about 2 km by 1 km and maximum depths in the different
arms varying between 6 and 14 m. More than 20 houseboats
on these water bodies sustained minor damage consisting of
buckled moorings and broken sewer and water lines. The
damage reports were concentrated on the east and west
shores of Lake Union and Portage Bay (Barberopoulou et
al., 2004). Although this damage pattern might imply water-
wave motion in a specific direction, the pattern may also
reflect the greater concentration of houseboats along the east
and west shores of the lake. Water waves were reported on
Lake Washington, which also overlies the Seattle basin, but
because of the lack of houseboats on the shores of Lake
Washington there were fewer observations of damage there
than around Lake Union.
Sloshing action was also reported in swimming pools
and ponds around Seattle (Table 1). Residential swimming
pools overflowed or showed large water oscillations in the
north–south direction, which is approximately the direction
from which the seismic waves arrived and may indicate the
effect of Rayleigh waves (Table 1). Many of the pools with
reported oscillations in Seattle are public facilities or private
clubs with pool lengths of 17–25 m. A typical residential
swimming pool where sloshing was observed was 11 m (36
feet) long and 5.5 m (18 feet) wide with an average depth
of2m.
The distribution of these water-wave reports is obvi-
ously biased by population density and demographic factors,
but the density of reports shows a close correspondence with
the largest ground motions recorded on PNSN seismographs.
In particular, from the 49 reports collected for Washington
state 41 fall within the latitude and longitude limits of Figure
1. All the observations in the Seattle basin (thirty in total)
were clustered over the deepest, central part of the basin
despite large areas of dense population outside of the bound-
ary of the Seattle basin near a variety of water bodies for
which reports are lacking. Based on this correlation we be-
lieve that basin amplification was an important factor con-
tributing to damaging water waves over the Seattle basin
(Fig. 6). McGarr and Vorhis (1968) suggested that seiches
from distant earthquakes most commonly occur in areas
where the seismic waves with periods of 5–15 sec are am-
plified by local geologic structures such as sedimentary ba-
sins. As noted earlier, our comparison of spectral ratios in-

528 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Figure 4. (continued). (c) Rock and nearly rock sites. Plots of spectral ratios corre-
sponding to stations situated at bedrock or nearly bedrock sites. Missing shear-wave
spectral ratios correspond to stations with very high signal-to-noise ratio or malfunc-
tioning instrument with no signal recorded. ERW and GNW are the reference sites used.
side and outside the basins (Fig. 6) suggest that seismic
waves in the frequency range of 0.04 to 0.2 Hz caused the
most prominent water waves to be concentrated over the
deep sedimantary basins.
Large water waves from past regional or distant earth-
quakes have been reported in local newspapers in Seattle
(Table 2), and Lake Union, in particular, has experienced
many seismically induced water waves in the past. News-
paper accounts show that large-amplitude water waves oc-
curred in various water bodies across Washington state dur-
ing the 1891 Port Townsend earthquake (The Oregonian,29
November 1891), the 1899 Yakutat Bay Alaska earthquake
(Dow, 1964), the 1906 San Francisco earthquake (Seattle
Post-Intelligencer, 19 and 25 April 1906), the M7.1 1949
Olympia earthquake (The Seattle Times, 14 April 1949), the
M
w
9.2 1964 Alaska earthquake (The Seattle Times,29
March 1964), and the M6.5 1965 Seattle-Tacoma earth-
quake (The Seattle Times, 19 and 30 April 1965). Damage
during the 1964 Alaska earthquake was, as during the Denali
earthquake, concentrated around Lake Union and Portage
Bay in Seattle.
With one exception, we have no water-level recordings
of seismically induced water waves in Washington state
from the Denali earthquake, or from any previous earth-
quakes, that are suitable to analyze wave motions in detail.
Water levels are generally sampled only once every few
minutes, or less frequently, to measure tide or reservoir lev-
els. The gauges are often damped to filter out the short-
period waves caused by wind and boats. Although long-
period seiches in bodies of water with fundamental periods
of hours are adequately sampled, short-period seiches in-
duced by seismic waves are not. Therefore, our analysis is
almost entirely limited to testing whether modeling results
match the approximate wave heights reported by eyewitness
accounts.
National Oceanographic and Atmospheric Administra-
tion (NOAA) tide gauge stations in Puget Sound provide
some scant evidence of water waves caused by the Denali
earthquake. The highest data rate recorded by these stations
is 6-min sample intervals and each water-level datum is av-
eraged over 3 min, centered on the reporting time. As ex-
pected, the averaged water-level data do not show a clear
record of seismically induced seiches. However, NOAA also
computes root-mean-square (rms) wave heights that are sam-
pled every 1 sec but averaged in the same way as the water
level. Several Alaska tide gauges show rms peaks resulting
from seismic waves from the Denali earthquake, and in
Washington, a tide gauge in Tacoma shows a particularly
prominent rms peak at the expected time (Fig. 7a, c).
A theoretical rms was computed to test whether the Ta-
coma rms wave-height recording was indeed a record of a
seiche due to ground shaking from long-period seismic
waves. We computed the rms we would expect at the Ta-
coma tide gauge from the Denali earthquake assuming that
the seiche amplitude time-history is a scaled version of the
accelerogram recorded at a station near Tacoma (seismic

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 529
Figure 5. Dot maps showing amplification of horizontal ground motion at strong
motion sites relative to the motion on bedrock for wave periods of 2, 4, 8, and 16 sec
(0.5, 0.25, 0.125, and 0.0625 Hz). The four maps in the top row show amplification of
Swaves. The four maps at the bottom show surface-wave amplification. The compu-
tations are done for horizontal “transverse” motion (motion perpendicular to a line
between the epicenter and the Puget lowland). Dot size is proportional to the spectral
ratio (SiteSpectrum/RockSpectrum) at each period. The rock spectrum is the average
of two rock sites (ERW,GNW). Stations enclosed by the heavy contours are defined as
basin sites. Area is the same as Figure 1.
trace at the top of Figure 7). We computed the resulting rms
wave height versus time using the same algorithm as is used
at the NOAA tide gauges. If we constrain the maximum val-
ues of the observed and computed rms amplitudes to be
equal, we obtain an excellent match for the duration of the
observed and computed rms time histories (Fig. 7a). From
the computed rms amplitude, we conclude that the maximum
amplitude of the waves at the Tacoma tide gauge was about
26 cm (0-peak) which is of the same order as amplitudes
anecdotally reported from other bodies of water in the Puget
Lowland (Table 1).
The rms wave-height peaks at other tide gauges in the
Puget Sound region were not nearly as prominent as at Ta-
coma. We believe this is because the Tacoma gauge is lo-
cated in a ship channel about 140 m wide that is bounded
by vertical walls. It is therefore likely that any seismically
generated waves observed at the gauge would be locally gen-
erated within the channel, rather than waves that had prop-
agated across Tacoma port/Commencement Bay. The am-
plitudes of such waves within the channel must vary greatly
with location, especially relative to distance from the side-
walls (e.g., cross-channel slosh modes). Hence it would be
difficult to extrapolate the interpretation of seiche measure-
ments at the gauge to other locations in the Port of Tacoma
(Commencement Bay). The tide gauge station in Seattle
would be the obvious choice to analyze in a similar fashion,
but the instrument is located at the Ferry terminal on a busy
waterfront and the data are therefore contaminated by high-
frequency noise due to waves from ship traffic (Fig. 7c).
With the exception of the Tacoma tide-gauge record,
we can only compare our theoretical results with observers’
reports. Most observers estimated that water-wave action in
the Seattle area lasted 1 to 5 minutes, which corresponds
well with the approximately 300-sec duration of the largest

530 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Figure 6. Average of spectral ratios (see Fig. 4) of basin, nonbasin, and rock sites
with respect to two reference stations (GNW,ERW). (a) Surface waves. (b) Shear waves.
Nonbasin sites show significant amplification because they also overlie sediments, al-
though not as thick as the basin sediments. The thin black line represents the ratio of
basin to nonbasin spectral ratios. The difference between the basin and nonbasin curves
is most pronounced for surface waves at frequencies of 0.04–0.2 Hz.
surface waves recorded by the PNSN seismographs (Table
1). Some observers reported wave periods of 5–20 sec (Table
1), in general agreement with the 10–12-sec period of the
largest surface waves. A few reports contained a rough es-
timate of the number of cycles (8–10), but these estimates
must reflect just the largest wave motions because the ob-
served duration of motion was reported to be considerably
longer than 80–100 sec. Many observers reported water
moving back and forth horizontally with little vertical mo-
tion. Reported vertical amplitudes, observed primarily at
swimming pools where the waves overtopped the sides of
the pools, were typically 30 cm. On large bodies of water
people reported horizontal runup of 0.6–3 m on the shore,
with most observers reporting about 1 m.
Water Wave Modeling
Lamb (1932) developed the shallow-water approxima-
tion for standing water waves in basins of rectangular cross
section (uniform depth) and variable depth. The earliest
studies of earthquake-induced seiches may be attributed to
Kvale (1953) and Proudman (1953). A basic mechanism for
the generation of seismically induced water waves by the
horizontal motion of the sides of the water body has been
presented by McGarr (1965). McGarr and Vorhis (1968),
and Ruscher (1999) have also studied seismically induced
seiches. However, the fluid mechanics literature contains
analyses of the sloshing of fluids in rectangular tanks, in
particular, as applied to tanks in ships (e.g., Waterhouse,
1994; Faltinsen et al., 2000; Hill, 2003). The application of
these latter methods to lakes of arbitrary shapes withvariable
sides has not been tested. Wind- and tsunami-induced
seiches in lakes and open bodies of water have been better
studied, with recent work carried out by Ichinose et al.
(2000), Zacharias (2000), and Rueda and Schladow (2002).
In the 36 years since the McGarr and Vorhis study,
seiches induced by seismic events have received little ex-
amination. Unlike previous earthquakes known for gener-
ating seiches (e.g., the 1964 Alaska earthquake), ground mo-
tions during the Denali earthquake were well recorded on a
variety of seismometers worldwide. The Denali earthquake
therefore provides us with an opportunity to examine ground
motions that initiate seiches in water bodies at large dis-
tances.
Barberopoulou et al. (2004) did some analysis for the
surface oscillations setup in water bodies during the Denali
earthquake. It was found that a residential swimming pool
of 10 m width and 2 m depth has a fundamental period of
oscillation of 4.5 sec (0.2 Hz), a period that was strongly
excited by seismic waves from the Denali earthquake
(Fig. 2). This estimate is in agreement with the observations
of sloshing in swimming pools during the earthquake
(Table 1). Therefore resonance in the fundamental mode is
a reasonable explanation for the surface oscillations induced
in swimming pools during the Denali earthquake. Figure 8
shows that the Denali earthquake, with dominant wave pe-
riods of 10 to 30 sec, could set shallow water bodies with
widths of 20 to 300 m into fundamental mode resonance.
The 10- to 30-sec periods (0.1–0.033 Hz) are also the periods
that the deep Seattle basin amplifies relative to the glacial
deposits, making these periods the most likely to produce
fundamental mode seiches.
The predicted fundamental periods of sections of Lake
Union (approximately 200 sec; Barberopoulou et al., 2004)
are longer than the dominant periods of the seismic surface
waves, and the water-wave periods described in the eyewit-
ness reports. Seismic amplitudes from the Denali earthquake
at 100 sec (0.01 Hz) are relatively low, far smaller than the
amplitudes at 10-sec periods (0.1 Hz) that dominate the sur-
face-wave signal (Fig. 2). Furthermore, long-period seismic

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 531
Figure 7. (a) Heavy line shows the recorded rms
deviation of wave height recorded at the NOAA Ta-
coma, Washington, tide gauge on 3 November 2002.
Root mean square (rms) is sampled every 6 min, and
each point is computed for a 3-min window centered
at the sample time. The rms is computed from wave-
height deviations that are sampled every second. The
light line is the computed rms deviation assuming the
water waves are a scaled version of an actual seis-
mogram of the Denali earthquake recorded at a station
near Tacoma (illustrated at the top of the figure using
the same time scale). To scale the wave heights so
that the peaks of both rms curves match at 6.6 cm
requires that the maximum 0-to-peak water wave am-
plitude at the gauge was about 26 cm. (b) Computed
response of a channel 8 meters deep to surface waves
(east component recorded at station WISC). The
graph shows the maximum vertical amplitude of wave
motion as a function of channel width using a reflec-
tion coefficient of 0.9 at the walls of the channel.
(c) rms heights recorded in Alaska and Washington
states. The names of the tide gauge locations appear
on the left. The top four traces are from locations in
Alaska whereas the bottom five correspond to loca-
tions in Washington state. The origin time of the
earthquake in local time in Alaska is shown by the
solid vertical line.
Figure 8. Lines showing the change in fundamen-
tal period as a function of width for a swimming pool
with the range of depths shown at the inset. Merian’s
formula has been used for the calculation of the period
(Lamb, 1932; Proudman 1953).
waves (⬎10 sec) undergo relatively modest amplification
within the Seattle basin compared with waves of shorter pe-
riod (Figs. 4 and 6). Such long water-wave periods also con-
flict with observers’ reports, which describe a series of
waves 5–20 sec apart.
The mismatch between the fundamental period of sec-
tions of Lake Union and the periods of observed water waves
can be explained by higher-order modes of resonance ex-
cited by the seismic forcing. Surface waves from the Denali
earthquake provided 10–30 cycles of motion over a 100- to
300-sec period (Fig. 2). This is a sufficiently long duration
of forcing to excite higher modes of resonance.
We model the response of swimming pools and portions
of Lake Union using a simple model of a water body
(“canal”) of rectangular cross section and infinite length
(Lamb, 1932; Russell and Macmillan, 1952; Proudman,
1953; McGarr, 1965; Wilson, 1972). Following Lamb (1932)
and Proudman (1953) we assume a water body with vertical
boundaries at distances x⳱0 and L. We constrain the mo-
tion in the x-zplane (one-dimensional flow) and simplify the
geometry by assuming a flat, horizontal bottom. In this
model, horizontal motion of the sides of the water body gen-
erate water waves that constructively interfere with waves
generated at the opposite wall.
In the rectangular model we assume that energy is fed
into the water body through the horizontal movement of
its boundaries (McGarr, 1965). Because we consider long-
period motion (seismic surface waves have wavelength of
several kilometers) we assume the forced walls of the water
body to be moving in the same direction at the same time
(i.e., the water body is much smaller than a half-wave-
length). Then it can be shown that the water elevation gat
any time tand point xis given by equation (1) (McGarr,
1965):
t
⬁ⳮ1
4hcos
[
(2nⳭ1)pxL
]
g(x,t)⳱兺
冮
pc2nⳭ1
n⳱00
(1)
(2nⳭ1)pc(tⳮs)
ⳮk(tⳮs)/2
F(s)esin ds,
冤冥
L
where Fis the forcing function (horizontal ground acceler-
ation), ka damping constant, hthe depth of the water body
(constant in our case), the wave velocity, and L
c⳱gh
冪
the distance between the boundaries of the water body.

532 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
Figure 9. Expected wave heights for three rectan-
gular water bodies subjected to the horizontal ground
motion recorded at Seattle basin station WISC during
the Denali earthquake. (a) Velocity record section of
east component at WISC station is used as input to
our model. (b) Computed wave height at the edge of
a large pool whose dimensions (width, 22.1 m; depth,
2 m) are chosen to provide a fundamental period of
10 sec. (c) Computed wave height in a small pool of
width 10 m and depth 2 m. Note scale factor of 10
ⳮ3
.
(d) Computed wave height in Lake Union (width,
800 m; depth, 8 m). In all cases a reflection coefficient
of 0.9 is assumed for waves that bounce back and
forth between the walls of the basin.
Figure 10. Expected wave heights for three rec-
tangular water bodies subjected to the horizontal
ground motion recorded at reference station GNW
during the Denali earthquake. (a) Velocity record sec-
tion of east component at GNW station is used as input
to our model. (b) Computed wave height at the edge
of a large pool whose dimensions (width, 22.1 m;
depth, 2 m) are chosen to provide a fundamental pe-
riod of 10 sec. (c) Computed wave height in a small
pool of width 10 m and depth 2 m. (d) Computed
wave height in Lake Union (width, 800 m; depth,
8 m). In all cases a reflection coefficient of 0.9 is
assumed for waves that bounce back and forth be-
tween the walls of the basin.
In this model the forcing is applied at the boundaries
(x⳱0, L) and the response of the water surface is “mea-
sured” at a selected position xbetween the walls. As a forc-
ing function, we used a ground motion recording from sta-
tion WISC near Lake Union (Fig. 9a). For computations we
used a modified version of equation 1 (H. Mofjeld, personal
comm., 2004) with the forcing input being ground velocity
rather than acceleration. The elevation of the water at any
point xand time tis given by the superposition of reflected
and incident waves off the boundary walls.
The rectangular model (Fig. 9) gives reasonable esti-
mates for the large swimming pools, with computed wave
heights in excess of 0.3 m being consistent with observations
of water sloshing over the sides of the pools. We set the
width of the large pool (22.1 m) to match the period of the
pool’s fundamental mode with the period of the largest seis-
mic waves, in an attempt to find the maximum expected
water-wave amplitudes. Such dimensions for a swimming
pool are within the range of 17–25 m that are standard for
pools in schools, country, and sport clubs where many of
the observations of sloshing originated. For small pools the
model predicts only 4-cm wave heights, which are large
enough to be noticeable but too small to cause sloshing over
the sides of the pools (Fig. 9). The simple model indicates
that a cross section of Lake Union (width, 800 m) would
produce water waves with amplitudes of up to 20 cm. We
know that amplitudes were large enough to be noticeable,
with the collection of reports suggesting water-wave heights
of at least 30 cm along the shores of the lakes. Considering
the uncertainty and incompleteness in the observations, the
model agrees fairly well with the vertical motions of the
water observed in the swimming pools but may slightly un-
derestimate those in Lake Union.
To demonstrate the effect the Seattle basin has on water
bodies we have computed water levels using the accelero-
gram recorded by station WISC (Seattle basin) and station
GNW (bedrock; Figs. 9 and 10). The seismic waves recorded
at GNW would not produce large surface oscillations in any
of the water bodies we considered. The predicted water-
surface oscillations for location GNW do not exceed 2 cm.
Water bodies located on the basins would produce water
waves about 10 times as large (Figs. 9 and 10).
Most estimates of the vertical amplitude of the water
level during the Denali earthquake are derived from obser-
vations of swimming pools. We do not have many estimates
around the lakes and only a few estimates refer to vertical
motion at Lake Union. Most observations from lakes in the
Puget Lowland refer to horizontal displacements or runup
distance on the shore. These runup distances are often large
(1–5 m) but depending on the sloping beach the vertical

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 533
Figure 11. Expected wave heights for three rec-
tangular water bodies under Nisqually earthquake
forcing. (a) Velocity record section at WISC station
is used as input to our model. (b) Computed wave
height at the edge of a large pool whose dimensions
(width, 22.1 m; depth, 2 m) are chosen to provide a
fundamental period of 10 sec. (c) Computed wave
height in a small pool with width, 10 m, depth, 2 m.
(d) Computed wave height in Lake Union (width,
800 m; depth, 8 m). In all cases a reflection coefficient
of 0.9 is assumed for waves that bounce back and
forth between the walls of the basin.
heights they correspond to can be much smaller (i.e., 20–30
cm). The lack of water-level measurements makes itdifficult
to conclude how well our simple model approximates the
true wave motions.
Despite these uncertainties we can still use the rectan-
gular model to estimate wave heights in Lake Union during
other types of earthquakes. Since the accelerometers of the
PNSN were installed, the only major local earthquake re-
corded was the 2001 M
L
6.8 Nisqually earthquake. We have
used the recording of the Nisqually earthquake from seismic
station WISC in the Seattle basin as input to the model
(Fig. 11). Under Nisqually earthquake forcing we see water-
wave amplitudes that are comparable to those calculated for
the Denali event (Fig. 11) although their frequency is much
higher than the water waves produced by the low-frequency
Denali surface waves. We note that no unusual water waves
were reported during the Nisqually earthquake.
The absence of observed seiches and the results of the
model appear to agree that water waves were not considered
unusual during the Nisqually earthquake. The frequency
content of the Nisqually earthquake signal was much higher
than that of the Denali. This would not be surprising given
that the Nisqually earthquake is a local deep (52 km) event
generating relatively small surface waves compared with
near-surface events. The signal from the distant Denali earth-
quake in Washington included long periods and duration and
generated unusual water motion (easily noticeable) unlike
that produced by ship traffic or winds. In contrast the higher-
frequency wave motion during the Nisqually earthquake
probably was either confused with other generating sources
(i.e., ship traffic) by observers who were also distracted by
the strong shaking and damage caused by the earthquake, or
the water activity was not surprising enough for observers
to report. A few observations of water waves were made
during the 1949 and 1965 earthquakes (Table 2). These
earthquakes were similar to the Nisqually earthquake in lo-
cation, magnitude, and depth (54 and 63 km).
Discussion and Conclusions
We used the spectral-ratio technique to document the
amplification of long-period seismic waves by the Seattle
basin, shallow glacial sediments, and bedrock sites. We find
large amplification at frequencies of 0.2–1 Hz, which is in
agreement with previous studies (Pratt et al., 2003), but we
also see amplification at longer periods. Most of the ampli-
fication we observe is caused by shallow deposits, whereas
the deep basin strata further amplify waves at frequencies of
0.04–0.2 Hz (5 to 25-sec period). Spectral ratios confirmthat
there is amplification at other locations outside of the Seattle
basin where sedimentary deposits overlay the bedrock.
In an effort to explain the generation of water waves in
Lake Union, we have used a simple model of a rectangular
basin to compute water-wave heights due to prolonged seis-
mic forcing of water bodies. The forcing functions we used
were seismic recordings of the distant, Denali earthquake
and recordings made in Seattle during the deep Nisqually
earthquake of 2001. This model predicts amplitudes in Lake
Union of ⬃20 cm during the Denali earthquake and com-
parable water-wave amplitudes during the Nisqually earth-
quake. The absence of reported water activity during the
Nisqually may be due to the higher frequency content and
shorter duration of the Nisqually signal.
Despite demographics, there is a remarkable correlation
between the distribution of water-wave reports during the
Denali earthquake and the largest seismic wave amplifica-
tion in the Puget Lowland. The documented long-period am-
plification by the sedimentary basins underlying the Puget
Lowland, and the results of our simple model suggest that
the water waves observed in Lake Union were caused by
prolonged seismic forcing caused by the amplified waves in
the Seattle basin.
Although the results do provide suggestions for what
caused the unusual water motion they also highlight the need
for further investigation using more realistic models and ex-
amination of other types of earthquakes. In particular, acom-
plete investigation would have to include the full geometry
of the lake. The latter is important since variable bathymetry
and shoreline (among other things) may be responsible for
enhanced wave heights in selected locations through focus-
ing. Since the 1964 Alaska earthquake, a comprehensive
water-level-recording network, including tide gauges oper-
ating at the time of that earthquake, has largely been re-

534 A. Barberopoulou, A. Qamar, T. L. Pratt, and W. P. Steele
moved, resulting in much less water-level data to work with.
Estimates could be significantly improved using instrumen-
tation to record water levels at high-frequency rates in key
locations because this would provide data for model testing.
Acknowledgments
We thank the Advanced National Seismic System (ANSS) for partial
financial support of the strong-motion network and Ruth Ludwin for the
information about the 1899 Alaska earthquake. We especially thank Harold
Mofjeld for his help with the tide gauge instruments and the computer
modeling. We thank Tom Brocher, Art McGarr, Paul Bodin, and the two
anonymous reviewers. Their suggestions significantly improved the final
version of this manuscript.
References
Barberopoulou, A., A. Qamar, T. L. Pratt, K. C. Creager, and W. P. Steele
(2004). Local amplification of seismic waves from the Denali Earth-
quake and damaging seiches in Lake Union, Seattle, Washington,
Geophys. Res. Lett. 31, L03607, doi 10.1029/2003GL018569.
Booth, D. B. (1994). Glaciofluvial infilling and scour of the Puget Lowland,
Washington, during ice-sheet glaciation, Geology 22, 695–698.
Brocher, T. M., R. J. Blakely, and R. E. Wells (2004). Interpretation of the
Seattle Uplift, Washington as a passive roof duplex, Bull. Seism. Soc.
Am. 94, 1379–1401.
Brocher, T. M., T. Parsons, R. J. Blakely, N. I. Christensen, M. A. Fisher,
and R. E. Wells, and SHIPS Working Group (2001). Upper crustal
structure in Puget Lowland, Washington: results from the 1998 Seis-
mic Hazards Investigation in Puget Sound, J. Geophys. Res. 106,
13,541–13,564.
Cassidy, J. F., and G. C. Rogers (2005). The M
w
7.9 Alaska earthquake of
3 November 2002: felt reports and unusual effects across western
Canada, Bull. Seism. Soc. Am. 94, no. 6B, S53–S58.
Crosson, R. L., and T. L. Owens (1987). Slab geometry of the Cascadia
subduction zone beneath Washington from earthquake hypocenters
and teleseismic converted waves, Geophys. Res. Lett. 14, 824–827.
Donn, W. L. (1964). Alaskan earthquake of 27 March 1964: remote seiche
simulation, Science 145, 261–262.
Dow, E. (1964). Passes to the North: History of Wenatchee Mountains,
Wenatchee Bindery and Printing Company, Wenatchee, Washington.
Eberhart-Phillips, D., P. J. Haeussler, J. T. Freymueller, A. D. Frankel,
C. M. Rubin, P. Craw, N. A. Ratchkovski, G. Anderson, G. A. Carver,
A. J. Crone, T. E. Dawson, H. Fletcher, R. Hansen, E. L. Harp, R. A.
Harris, D. P. Hill, S. Hreinsdo´ttir, R. W. Jibson, L. M. Jones, R.
Kayen, D. K. Keefer, C. F. Larsen, S. C. Moran, S. F. Personius,
G. Plafker, B. Sherrod, K. Sieh, N. Sitar, and W. K. Wallace (2003).
The 2002 Denali fault earthquake, Alaska: a large magnitude, slip-
partitioned event, Science 300, 1113–1118.
Faltinsen, O. M., O. F. Rognebakke, I. A. Lukovsky, and A. N. Timokha
(2000). Multidimensional modal analysis of nonlinear sloshing in a
rectangular tank with finite water depth, J. Fluid Mech. 407, 201–
234.
Frankel, A., and W. Stephenson (2000). Three-dimensional simulations of
ground motions in the Seattle region for earthquakes on the Seattle
fault zone, Bull. Seism. Soc. Am. 90, 1251–1267.
Frankel, A. D., D. L. Carver, and R. A. Williams (2002). Nonlinear and
linear site response and basin effects in Seattle for the M6.8 Nisqually,
Washington, earthquake, Bull. Seism. Soc. Am. 92, 2090–2109.
Galster, R. W., and W. T. Laprade (1991). Geology of Seattle, Washington,
United States of America, Bull. Assoc. Eng. Geol. 28, 235–302.
Hartzell, S. (1992). Site response estimation from earthquake data, Bull.
Seism. Soc. Am. 82, 2308–2327.
Hill, D. F. (2003). Transient and steady-state amplitudes of forced waves
in rectangular basins, Phys. Fluids 15, 1576–1587.
Ichinose, G. A, J. G. Anderson, K. Satake, R. A. Schweickert, and M. M.
Lahren (2000). The potential hazard from tsunami and seiche waves
generated by large earthquakes within Lake Tahoe, California–
Nevada, Geophys. Res. Lett. 27, 1203–1206.
Johnson, S. Y., R. J. Blakely, W. J. Stephenson, S. V. Dadisman, and M. A.
Fisher (2004). Active shortening of the Cascadia forearc and impli-
cations for seismic hazards of the Puget Lowland, Tectonics 23,
TC1011, doi 10.1029/2003TC001507.
Johnson, S. Y., C. J. Potter, and J. M. Armentrout (1994). Origin and evo-
lution of the Seattle fault and Seattle basin, Washington, Geology 22,
71–74.
Johnson, S. Y., C. J. Potter, J. M. Armentrout, J. J. Miller, C. Finn, and
C. S. Weaver (1996). The southern Whidbey Island fault: an active
structure in the Puget Lowland, Washington, Geol. Soc. Am. Bull.
108, 334–354.
Jones, M. A. (1996). Thickness of unconsolidated deposits in the Puget
Sound Lowland, Washington and British Columbia, U.S. Geol. Surv.
Water Resour. Invest. Rept. 94-4133.
Joyner, W. B. (2000). Strong motion from surface waves in deep sedimen-
tary basins, Bull. Seism. Soc. Am. 90, S95–S112.
Koketsu, K., K. Hatayama, T. Furumura, Y. Ikegami, and S. Akiyama
(2005). Damaging long-period ground motions from the 2003 M
w
8.3
Tokachi-oki, Japan, Earthquake, Seism. Res. Lett. 76, 67–73.
Korgen, B. J. (1995). Seiches, Am. Scientist 83, 330–341.
Kvale, A. (1953). Seismic seiches in Norway and England during the
Assam earthquake of August 15, 1950, Bull. Seism. Soc. Am. 45,
93–113.
Lamb, H. (1932). Hydrodynamics, Dover, New York.
McGarr, A. (1965). Excitation of seiches in channels by seismic waves, J.
Geophys. Res. 70, 847–854.
McGarr, A., and R. C. Vorhis (1968). Seismic Seiches from the March
1964 Alaska Earthquake, U.S. Geol. Surv. Profess. Pap. 544E, E1–
E43.
Pratt, T. L., and T. M. Brocher (2006). Site response and attenuation in the
Puget Lowland, Washington State, Bull. Seism. Soc. Am. 96, no. 2,
536–552
Pratt, T. L., T. M. Brocher, C. S. Weaver, K. C. Miller, A. M. Trehu, K. C.
Creager, and R. S. Crosson (2003). Amplification of seismic waves
by the Seattle basin, Washington State, Bull. Seism. Soc. Am. 93, 533–
545.
Pratt, T. L., S. Johnson, C. Potter, W. Stephenson, and C. Finn (1997).
Seismic reflection images beneath Puget Sound, western Washington
state: the Puget Lowland thrust sheet hypothesis, J. Geophys. Res.
102, 27,469–27,489.
Proudman, J. (1953). Dynamical Oceanography, Methuen, London.
Rueda, F. J., and S. G. Schladow (2002). Surface seiches in lakes of com-
plex geometry, Limnol. Oceanogr. 47, 906–910.
Ruscher, C. R. (1999). The sloshing of trapezoidal reservoirs, Ph.D Thesis,
University of Southern California, Los Angeles, California.
Russell, R. C. H, and D. H. Macmillan (1952). Waves and Tides, Hutch-
inson’s Scientific and Technical Publications, New York.
Snelson, C. M., K. C. Miller, T. M. Brocher, T. L. Pratt, A. M. Tre´hu, and
C. S. Weaver (2000). Results of the 1999 Dry SHIPS (Seismic Hazard
Investigations of Puget Sound) Seismic Experiment in western Wash-
ington, E
OS
Trans. AGU 81, F871.
ten Brink, U. S., P. C. Molzer, M. A. Fisher, R. J. Blakely, R. C. Bucknam,
T. Parsons, R. S. Crosson, and K. C. Creager (2000). Subsurface
geometry and evolution of the Seattle fault zone and the Seattle basin,
Washington, Bull. Seism. Soc. Am. 92, 1737–1753.
Thorson, R.M. (1980). Ice-sheet glaciation of the Puget Lowland, Wash-
ington, during the Vashon Stade (Late Pleistocene), Quat. Res. 13,
303–321.
Van Wagoner, T. M., R. S. Crosson, K. C. Creager, G. Medema, L. Preston,
N. P. Symons, and T. M. Brocher (2002). Crustal structure and re-
located earthquakes in the Puget Lowland, Washington from high
resolution seismic tomography, J. Geophys. Res. 107, no. B12, 22-1–
22-23.

Long-Period Effects of the Denali Earthquake on Water Bodies in the Puget Lowland: Observations and Modeling 535
Waterhouse, D. D. (1994). Resonant sloshing near a critical depth, J. Fluid
Mech. 281, 313–318.
Wilson, B. W. (1972). Seiches,inAdvances in Hydrosciences, V.-T. Chow
(Editor), Academic Press, New York.
Zacharias, I. (2000). Verification of seiching processes in a large and deep
lake (Trichonis, Greece), Mediterr. Mar. Sci. 1, 79–89.
Department of Earth and Space Sciences
University of Washington
Seattle, Washington 98195
(A.B., A.Q., W.P.S.)
Anthony Qamar died in an automobile accident on 4 October 2005.
U.S. Geological Survey, School of Oceanography
University of Washington
Seattle, Washington 98195
(T.L.P.)
Manuscript received 2 May 2005.