Sensitivity Analysis of Source Parameters for Earthquake-

Generated Distant Tsunamis

Edison Gica1; Michelle H. Teng, M.ASCE2; Philip L.-F. Liu, F.ASCE3; Vasily Titov4; and Hongqiang Zhou5

Abstract: This paper presents a sensitivity analysis of the effects of different fault plane parameters on earthquake generated tsunamis
in the far field. The parameters studied include the location of epicenter, rake angle, dip angle, strike angle, fault plane dimensions, slip
displacement, and focal depth. The study was carried out by applying the verified Cornell COMCOT tsunami model to simulate distant
tsunamis generated by earthquakes in three different seismic regions, namely, Japan, Aleutian Islands, and Chile in the Pacific Basin. In
this study, each fault plane parameter was varied and its effect on the tsunami wave height at a distant location 共Hawaii offshore waters
as a case study兲 was examined. Our results showed that under the same earthquake magnitude, in general, variations within a reasonable
range of uncertainty in rake and dip angles, epicenter location, and focal depth do not affect the resulting tsunamis significantly while a
variation in the fault dimensions, strike angle, and slip displacement can cause a large change in the wave heights in the far field. The
study revealed that the sensitivity of the far field wave height to certain earthquake parameters including the fault plane dimensions, slip
displacement, and the strike angle does not always decrease as the distance between the earthquake and the far field increases. These
results indicate the importance of accurate earthquake information for accurate tsunami predictions and that even for wave heights in a far
field thousands of miles away, the earthquake may not be treated as a simple point source described by its epicenter location and
magnitude alone.
DOI: 10.1061/共ASCE兲0733-950X共2007兲133:6共429兲
CE Database subject headings: Tsunamis; Earthquakes; Parameters; Sensitivity analysis.

Introduction historical data so that a proper coastal management plan can be

established for tsunami risk areas. So far, theoretical and numeri-
Earthquake generated tsunamis can cause serious damage to cal models for predicting the initial tsunami wave generated by an
coastal communities both near the earthquake source and at dis- earthquake and for simulating tsunami propagation in the ocean
tant locations 共Liu et al. 2005兲. The 1946 Aleutian and the 1960 and inundation on coastal land have been developed by the tsu-
Chilean tsunamis are two examples of powerful distant tsunamis nami research community. For the generation of the initial tsu-
that caused severe property damage and loss of human lives both nami wave height due to an earthquake, the commonly applied
in the source region and in the Hawaiian Islands thousands of theory is the dislocation theory developed by Mansinha and
miles away from the earthquake. It is important for scientists and Smylie 共1971兲. This theory is based on the linear elastic theory
engineers to develop modeling tools that can accurately predict and predicts the seafloor displacement based on the fault plane
tsunami wave heights generated by an earthquake to assist civil characteristics described by several parameters including the
defense agencies to issue proper warnings during a tsunami event. location of the epicenter, focal depth, size of fault area, slip
These tools can also be helpful in performing reliable probabilis- displacement, strike, dip, and rake angles. Assuming that the
tic risk analysis of coastal flooding due to tsunamis based on rupture speed of the fault plane is much 共a couple of orders of
magnitude兲 larger than that of the phase speed of the generated
1
Research Scientist, UW-JISAO/NOAA Center for Tsunami Research, tsunami wave and that the water is incompressible, we expect that
NOAA/PMEL/OERD, Seattle, WA 98115. the initial sea surface displacement takes the same form as the
2
Associate Professor, Dept. of Civil and Environmental Engineering, seafloor displacement. Once a tsunami is generated by an earth-
Univ. of Hawaii at Mānoa, Honolulu, HI 96822 共corresponding author兲.
quake, its propagation and inundation can be simulated by hydro-
E-mail: teng@wiliki.eng.hawaii.edu
3
Professor, School of Civil and Environmental Engineering, Cornell dynamic tsunami models such as the Cornell COMCOT model
Univ., Ithaca, NY 14853. 共Liu et al. 1994, 1995; Wang and Liu 2005, 2006兲, NOAA Center
4
Research Scientist, UW-JISAO/NOAA Center for Tsunami Research, for Tsunami Research’s MOST model 共Titov and Synolakis
NOAA/PMEL/OERD, Seattle, WA 98115. 1998兲, and the Japanese TIME model developed by Imamura
5
Ph.D. Candidate, Dept. of Civil and Environ. Engr., Univ. of Hawaii 共Imamura et al. 1998; Imamura 1996兲.
at Mānoa, Honolulu, HI 96822. Currently, one of the most significant uncertainties in tsunami
Note. Discussion open until April 1, 2008. Separate discussions must wave height prediction is the specification of the initial sea sur-
be submitted for individual papers. To extend the closing date by one face elevation distribution associated with the earthquake. The
month, a written request must be filed with the ASCE Managing Editor.
dislocation theory itself has been shown to be able to predict the
The manuscript for this paper was submitted for review and possible
publication on July 24, 2006; approved on March 12, 2007. This paper is initial wave quite satisfactorily as long as all the required input
part of the Journal of Waterway, Port, Coastal, and Ocean Engineering, fault plane parameters are known and accurate. However, in real-
Vol. 133, No. 6, November 1, 2007. ©ASCE, ISSN 0733-950X/2007/6- ity, the main source of uncertainty comes from the fact that for
429–441/$25.00. most earthquakes, while the epicenter location and the magnitude

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007 / 429
can be measured relatively accurately and quickly after an earth-
quake occurs, most of the other fault plane parameters can be
difficult to determine and may remain unknown 共Synolakis et al.
1997兲. In these cases, we often apply empirical formulas to ap-
proximately estimate the values of the fault plane parameters.
Since the fault plane parameters are the required input parameters
for the tsunami prediction models, it is important to understand
how the uncertainty in these parameters would affect the final
prediction of the tsunami wave heights. Thus, a sensitivity analy-
sis of the fault plane parameters is necessary.
Titov et al. 共1999兲 conducted a numerical study of the 1996
Andreanov Island tsunami generated by an earthquake in the
Aleutian region and found that the initial leading tsunami wave is Fig. 1. Definition sketch of fault plane dimension and geometry
insensitive to certain earthquake fault plane parameters such as
the rake and dip angles. In addition, their results showed that the
detailed spatial variation of the initial wave may not affect the Johnson 1999兲. In this theory, the displacement of the seafloor
tsunami wave heights in a distant coastal area which may indicate due to a submarine earthquake is calculated based on several
that an earthquake may be considered as a point source for a equations 共which can be found in Mansinha and Smylie 1971;
distant location. These results were helpful and were the basis for Liu et al. 1998兲 that require input parameters describing the loca-
tsunami scenarios using a discrete set of unit sources for the linear tion, orientation, and rupture direction of the fault plane, specifi-
propagation model database in NOAA Center for Tsunami Re- cally, the strike, dip, and rake angles, fault width, fault length,
search’s method for tsunami forecasting 共Titov et al. 2005兲. How- focal depth, and slip displacement as illustrated in Fig. 1. Since an
ever, since tsunami generation and propagation is a very complex earthquake abruptly displaces the seafloor, we can assume that the
phenomenon, further sensitivity studies on more than one tsunami initial sea surface displacement, i.e., the initial tsunami wave,
event originated from different source regions are needed. The takes the same profile as that of the seafloor displacement.
present study extends the work by Titov et al. 共1999兲 to further After the initial sea surface displacement is calculated, the
investigate the effects of fault plane parameters on tsunami gen- propagation of the tsunami wave in the open ocean and its runup
eration by examining more tsunami scenarios originated from in a coastal area are predicted by solving the shallow water equa-
Japan, Aleutian Islands, and Chile. In addition, the parameters tions 共SWEs兲 numerically. In Cartesian coordinates, the equations
that were not studied by Titov et al. 共1999兲, for example, the focal are given as
depth, are investigated in the present study in order to achieve a
more complete understanding of the sensitivity of tsunamis to the ⳵␨ ⳵ P ⳵Q
+ + =0 共1兲
earthquake fault plane parameters by using the Pacific Basin as a ⳵t ⳵x ⳵y
case study. We would like to mention that in Titov et al.’s study
共1999兲, the range of variation in certain parameters such as the
dip angle was relatively small. In the present study, a larger range
⳵ P ⳵ P2
+
⳵t ⳵x H
+冉 冊 冉 冊
⳵ PQ
⳵y H
⳵␨
+ gH + ␶xH − f
⳵x
Q
H
=0 冉冊 共2兲
of variation is examined. Geist 共1999兲 also carried out a very
valuable study on the effect of different fault plane parameters on
local tsunamis, i.e., the initial tsunami waves close to the earth-
quake fault plane location. He provided very useful information
⳵ Q ⳵ PQ
+
⳵t ⳵x H
+冉 冊 冉 冊
⳵ Q2
⳵y H
⳵␨
+ gH + ␶yH + f
⳵y
P
H
=0 冉冊 共3兲

on earthquake fault plane mechanism and common values of dif- where ␨⫽wave elevation above the unperturbed sea level; P and
ferent fault plane parameters. However, he did not examine the Q⫽depth-averaged horizontal fluxes; H⫽total depth; ␶x and
wave heights at a distant location in detail, although he mentioned ␶y⫽bottom friction terms that are modeled by either the Chezy or
that a distant tsunami would most likely depend mainly on the the Manning’s formula; and f⫽Coriolis coefficient. In general
earthquake magnitude and will not be sensitive to other fault simulations of tsunamis by the Cornell COMCOT model, the
plane parameters. The focus of the present study is the quantita- equations are solved in spherical coordinates in the open ocean
tive analysis of wave heights in the far field, i.e., in the offshore and in Cartesian coordinates in coastal regions by applying the
area of a distant location, and to study how the uncertainty or finite difference scheme described in Liu et al. 共1994, 1995,
variations in different fault plane parameters may affect the far 1998兲. For the particular simulation of tsunami propagation in the
field wave heights. open ocean 共or in deep ocean兲 and at offshore locations in the
present study, we used the linearized nondispersive shallow water
equations by neglecting the second and third terms in the momen-
Theory and Methodology tum Eqs. 共2兲 and 共3兲. Based on the long wave theory, the disper-
sive effect is measured by the dimensionless parameter 共h / ␭兲2
To study earthquake generated tsunamis, we need theoretical and the nonlinear effect by a / h, where h⫽water depth;
models that can predict all three processes involved, namely, the ␭⫽wavelength; and a⫽wave amplitude. For tsunamis propagat-
generation of the initial tsunami wave by an earthquake, the tsu- ing in the open Pacific Ocean with a mean depth of 5,000 m,
nami propagation in the open ocean, and the final inundation and the wave amplitude is usually less than 1 m while the wavelength
runup in a coastal area. In the present study, the initial wave can be more than 500 km. These data lead to an estimate of
generation is predicted based on the elastic deformation theory of 共h / ␭兲2 = 共5 / 500兲2 = 0.0001 and a / h = 1 / 5,000= 0.0002, which in-
the earth’s crust where the earth is assumed to follow the laws of dicate that both the nonlinear and dispersive effects are extremely
the classical linear elastic theory which treats it as a homoge- small and can be neglected. Similar linear nondispersive shallow
neous, isotropic, and elastic material 共Mansinha and Smylie 1971; water equations are also applied in the MOST model at the

430 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007
NOAA Center for Tsunami Research for simulating tsunami Table 1. Earthquake Fault Plane Parameters for 1960 Chilean Tsunami
propagation in the deep ocean. Although the basic governing Earthquake parameter Parameter values
equations for the COMCOT and MOST models are the same, the
two models apply different numerical schemes to solve the equa- Epicenter 74.5° W, 39.5° S
tions. Details of the numerical techniques applied in the MOST Strike angle 10°
model can be found in Titov and Synolakis 共1998兲. As tsunamis Dip angle 10°
enter coastal regions with shallow depth, both the nonlinear and Rake angle 90°
dispersive effects may become important and may not be ne- Slip 共m兲 24
glected. In those cases, the nonlinear and dispersive Boussinesq- Focal depth 共km兲 53
type equations can be applied. Fault length 共km兲 800
As mentioned earlier, many fault plane parameters cannot Fault width 共km兲 200
be determined directly based on the initial earthquake data
共Synolakis et al. 1997兲. Therefore scientists often estimate the
parameters by applying the following empirical equations based ated by a combined earthquake and an earthquake triggered land-
on the known parameter of earthquake moment magnitude M w slide 共Fryer et al. 2004兲. Unfortunately, a recent seafloor survey
共Synolakis et al. 1997; Johnson 1999兲 by the Scripps Institute failed to discover the evidence of a past
M w = 共2/3兲 log M 0 − 6.03 共4兲 landslide at the site 共this does not imply that the landslide did not
happen, however兲. Thus, the fault plane mechanism for the 1946
Aleutian tsunami remains an open question. On the other hand,
M 0 = ␮DLW 共5兲
the fault plane parameters for the 1960 Chilean earthquake have
where M 0⫽seismic moment; ␮⫽shear modulus 共also called been better understood and determined, and therefore this event is
rigidity, N / m2兲 of the earth; D⫽average slip displacement 共m兲; used in the present study to verify the COMCOT model. We
L⫽fault length 共m兲; and W⫽fault width 共m兲. The value of should mention that the first verification study of the COMCOT
rigidity varies with location and ranges usually from 1.0 to model by simulating the 1960 Chilean tsunami runup at Hilo Bay
6.0⫻ 1010 N / m2. Rigidity depends mainly on the geological prop- in Hawaii was carried out by Liu et al. in 1994 共Liu et al. 1994兲.
erties of a location, for example, whether the fault sits on soft Since then, more accurate bathymetry data for the Pacific Ocean
sediment or hard rock. For each given location, geologists can and especially for the Hawaiian coastal areas have become avail-
usually give a good estimate for the rigidity. For our study, we able thanks to the new bathymetric surveys conducted by the U.S.
used a typical value of ␮ = 4.0⫻ 1010 N / m2 for the Pacific Rim Army Corps of Engineers and other agencies in recent years. In
regions 共see Johnson 1999兲. Applications of Eqs. 共4兲 and 共5兲 will the present study, we apply the COMCOT model to resimulate the
be discussed further in the section on “Sensitivity Analysis.” 1960 Chilean tsunami by using the improved bathymetry data and
compare the simulated wave height with the field data at a differ-
ent location in Hawaii, i.e., at Honolulu Harbor on Oahu, Hawaii,
Model Verification for further verification. Table 1 lists the values of the earthquake
fault plane parameters for the 1960 Chilean tsunami as given by
In this study, the Cornell COMCOT tsunami model is applied to Kanamori and Cipar 共1974兲. It should be noted that even though
simulate earthquake generated distant tsunamis in the Pacific the 1960 event is better understood, there may still be a small
Basin. This model combines all three modules, namely, the simu- uncertainty in the values of the fault plane parameters, for
lation of initial wave generation from an earthquake; tsunami example, the focal depth may range from 16 to 53 km as evalu-
propagation; and final coastal runup and inundation, in one For- ated in Kanamori and Cipar’s study 共Kanamori and Cipar 1974兲.
tran code. To enhance computational efficiency, the model uses a However, when combined with the estimate for the fault width
multilayered grid system with coarser grids for the open ocean as 200 km, a focal depth of about 50 km was suggested by
and finer grids in coastal regions where a higher resolution is Kanamori and Cipar 共1974兲. The focal depth given by the NOAA
necessary to predict wave height, inundation, and overland flows. National Geophysical Data Center 共NGDC兲 for this event is
Since the COMCOT model has been verified for long wave 60 km. In the present study, we used the upper value of 53 km
propagation and runup on beach and island models in laboratory from Kanamori and Cipar’s paper 共Kanamori and Cipar 1974兲.
experiments 共Liu et al. 1995; Zhou et al. 2005兲, the focus of The fact that there are existing uncertainties in fault plane param-
verification in this study is to examine the accuracy and efficiency eters is the motivation of the present study, and the effect of focal
of the COMCOT simulation model when applied to predict actual depth on distant tsunamis will be discussed in detail in a later
tsunamis in the field. In the Pacific Ocean, there are two past section of this paper.
significant tsunamis with field data, namely, the 1946 Aleutian In the simulation of the 1960 Chilean tsunami for predicting
and 1960 Chilean tsunamis, which can potentially be used for the wave heights in Honolulu Harbor in Hawaii 共location shown
model verification purposes. However, after examining both in Fig. 2兲, two layers of nested grids were used 共see Fig. 3, the
cases, we decided that the 1946 Aleutian tsunami may not be a lower right section兲. The first layer of the computational domain
proper event for verifying a tsunami model. The reason is that the covers a large domain of open Pacific Ocean 共from 165° W to
fault plane mechanism for the 1946 event is still in debate. The 65° W and from 45° S to 30° N兲 that includes the earthquake lo-
1946 Aleutian earthquake was initially measured to have a mag- cation and the Hawaiian islands. The bathymetry data used for
nitude of about 7.3, which is not powerful enough to generate a this layer is based on NOAA’s ETOPO5 data set, which has a grid
giant trans-Pacific tsunami that killed 159 people in Hilo Bay, size of 5 min 共about 9 km兲. The second layer is a smaller rectan-
Hawaii. Consequently, many researchers have assigned a much gular domain that encloses all the major Hawaiian islands 共from
larger earthquake magnitude, for example, 8.6 共Johnson 1999兲, in 162° W to 152° W and from 15° N to 25° N兲 and is based on the
their simulations of this event. However, in recent years, a new best available bathymetry data obtained from the Office of Marine
hypothesis emerges and argues that the 1946 tsunami was gener- Data Archive at the University of Hawaii at Manoa 共compiled

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007 / 431
 is considered as more than sufficient to simulate wave propaga-
tion. In fact, we have conducted a convergence test which showed
that using different grid sizes of 4.5, 9.0, and 18 km leads to
almost the same simulation results. The time step used is 1.0 and
0.5 s for Layers 1 and 2, respectively. These time steps satisfy the
Courant–Friedrichs–Levy stability criterion.
The total time that the 1960 Chilean tsunami took to travel
from the Chilean earthquake location to Hawaii’s shores was
about 14 h. A comparison between the simulated and recorded
wave height by Honolulu Harbor is presented in Fig. 4. The simu-
lated wave height was evaluated at the grid point closest to the
tidal gauge location 共W = 157.8667°, N = 21.275°兲. The results
show that the overall amplitude and envelope of the wave train
match the recorded data relatively well although there is a dis-
crepancy in wavelength for the first leading wave. Considering
the complexity of the physical phenomenon being studied, the
Fig. 2. Sketch of locations of Honolulu Harbor and Hilo Bay as well agreement between the simulated and recorded wave height is
as 12 other offshore locations in Hawaii where tsunami wave heights reasonably good, thus verifying the COMCOT model. More veri-
are simulated 共represented by black dots兲 fication cases, including simulation of tsunami inundation and
comparing the simulated runup and inundation with the field data
at Hilo Bay for the 1960 Chilean tsunami, can be found in Gica
from the following sources: USGS DEM, USGS Digital Data 共2005兲. In the simulation of tsunami inundation at Hilo Bay, a
Services for Hawaii DDS-55 Hawaii, NGDC NOS hydrographic four layer grid system was used, and the grid size of Layer 4 at
bathymetry, Smoot/King NAVO bathymetry, George Pararas- Hilo Bay was as fine as 10 m. The simulated wave runup and
Carrayannis, and Army Corps of Engineers’ SHOALS data兲 com- inundation showed good agreement with the field measurement
bined with NOAA’s ETOPO2 共finer than ETOPO5兲. The second 共Gica 2005兲.
layer is discretized to a grid size of 30 s 共approximately 1 km
depending on latitude兲. These spatial grid sizes may seem large,
however, since tsunamis are extremely long waves, the grids are Sensitivity Analysis and Results
proper and efficient for simulating tsunami propagation in the
open ocean and for predicting offshore wave heights. For ex- The objective of this study is to examine and quantify the effect
ample, if the wavelength of a tsunami is 500 km, a grid size of of different fault plane parameters on the generated tsunamis in
9 km will ensure more than 50 grid points per wavelength which the far field, i.e., in the offshore area at a distant location. We are
especially interested in understanding how sensitive the predicted
tsunami wave height is to the variations or uncertainties in the
fault plane parameters that include epicenter location, rake, dip
and strike angles, fault length and width, slip displacement, and
the focal depth. According to Dr. Gerard Fryer and Dr. Barry
Hirshorn at the Pacific Tsunami Warning Center 共personal com-
munications, 2005兲, the scientists at the warning center are
usually able to determine the following fault plane parameters
relatively fast 共30 min after the earthquake occurs兲: epicenter lo-
cation accurate to ±50 km, magnitude to ±0.2 and hypocenter
focal depth to ±15 km. They explained that these are estimated
numbers, and the uncertainty in these parameters for any particu-
lar event may be much larger. Here the hypocenter is the location
of the initial rupture. The centroid focal depth of the entire rupture
area is more difficult to determine as the size and dimensions
共length and width兲 of the entire rupture area are usually not
known immediately after the earthquake. The final centroid loca-
tion can be hundreds of kilometers away from the initial hypo-
center location. Similarly, other parameters such as the fault
dislocation, dip, and rake angles are also difficult to determine as
they may not be known from the earthquake recordings directly.
Some may be estimated based on aftershock information and oth-
ers need to be approximated through empirical equations. Syno-
lakis et al. 共1997兲 pointed out that the short-term errors for the
fault plane parameters immediately after the earthquake can range
Fig. 3. Sketch of two-layer grid systems for tsunamis originated from from 25 to 50% except for the length and width, distribution of
Japan, Aleutian Islands, and Chile. Dashed lines show first layer and the slip, and its strikewise variation, where the errors can be as
solid lines show second layer surrounding Hawaiian Islands. Small high as 75–90%. Errors in longer term estimates can be lowered
rectangles show three fault plane areas in Japan, Aleutian Islands, and by a factor of two. Sato 共1989兲 developed empirical relationships
Chile with star marking epicenter location. between the earthquake magnitude and fault plane dimensions

432 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007
 Fig. 4. Comparison of predicted wave elevation 共solid lines兲 with tidal record 共dashed line兲 at Honolulu Harbor for 1960 Chile tsunami

共width and length兲 and other parameters based on statistical fact the actual 1960 Chilean earthquake, and the epicenter loca-
analysis of past earthquakes in the Pacific Ocean. His results have tion for the Aleutian earthquake scenario is selected to be the
been helpful to Japanese researchers in estimating the fault plane same as the location for the 1946 Aleutian earthquake. In all the
parameters in tsunami prediction. Currently, for dip and rake simulations, a similar two-layer grid system as well as similar
angles, the Japanese tsunami scientists usually apply the common time steps as described in the previous section were used.
values of 20– 25° and 90°, respectively 共Professor Tomoyuki Specifically, Layer 1 共dashed line in Fig. 3兲 covers a large region
Takahashi of Akita University, personal communications, 2005兲. of the Pacific Ocean that includes the source location and the
In the present study, we would like to examine the effect of Hawaiian Islands while Layer 2 共solid line in Fig. 3兲 covers the
each fault plane parameter by varying the parameter while main- major Hawaiian Islands only. To investigate how the variations in
taining the same earthquake magnitude and the other parameters. the fault plane parameters affect the predicted wave height in the
One of the related questions that we would like to study is far field, we focus on studying one parameter at a time, namely,
whether an earthquake can be considered as a point source de- when we study the effect of a particular parameter, we vary it
scribed by its epicenter location and magnitude alone. In this while holding the other parameters, especially the earthquake
study, three tsunamigenic earthquake scenarios in three different magnitude, the same. This can help us to examine whether the
regions are chosen for the case studies. The detailed fault plane far field wave height depends only on earthquake magnitude or
parameter values for the three cases are given in Table 2. Based it is sensitive to other fault plane parameters as well. The location
on Eqs. 共4兲 and 共5兲, the moment magnitude for these earthquakes of the numerical station where the wave height is plotted in
is calculated to be 9.3. This magnitude is the largest earthquake Figs. 5–11 is offshore Hilo Bay with coordinates 155.0417° W
magnitude in the Pacific Basin based on historical earthquake and 19.7917° N. The local water depth is 126.4 m.
records for the past 200 years 共Whiteside et al. 2000兲 and is
selected to be studied here as the worst-case earthquake scenario.
Effect of Epicenter
In all the simulations in our study, wave height at 14 offshore
locations in Hawaii is predicted. Due to page limitation, only the In the present study, the epicenter location is varied by a maxi-
results near the mouth of Hilo Bay in Hawaii 共location shown in mum of 200 km with respect to the epicenter location of each
Fig. 2兲 are plotted in detail in Figs. 5–11. The results for the other reference scenario 共i.e., the reference scenario in Table 2兲. The
locations are summarized and presented in Tables 3–9. In these wave height generated by the original reference earthquake is
tables, the data presented are calculated for the leading wave compared with that by earthquakes with varied epicenter location.
height. The sample simulated results for the waves at Hilo Bay are shown
The locations of the three earthquakes being studied are illus- in Fig. 5. In this figure, a legend such as “100 km N” indicates
trated in Fig. 3. The earthquake scenario in Chile in Table 2 is in “100 km north of the original earthquake location.” The results

Table 2. Fault Plane Parameters for Three Reference Earthquake Scenarios

Earthquake
parameter Japan Aleutian Chile
Epicenter 143.84° E, 37.71° N 163.19° W, 53.32° N 74.5° W, 39.5° S
Strike angle 共deg兲 200° 250° 10°
Dip angle 共deg兲 10° 10° 10°
Rake angle 共deg兲 90° 90° 90°
Slip 共m兲 24 24 24
Focal depth 共km兲 53 53 53
Fault length 共km兲 800 800 800
Fault width 共km兲 200 200 200

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007 / 433
Fig. 5. Effect of epicenter shift on resulting wave height in Hilo Bay
Fig. 6. Effect of rake angle variation on resulting wave height at Hilo
due to earthquake in: 共a兲 Aleutian Islands; 共b兲 Japan; and 共c兲 Chile
Bay due to earthquake in: 共a兲 Aleutian Islands; 共b兲 Japan; and 共c兲
Chile
show that for earthquakes in all three regions, i.e., Japan, Aleu-
tian, and Chile, a 200 km variation in epicenter location will not
change the resulting tsunami wave height in Hilo 共and in Hawaii age relative wave height change in Hawaii due to the 200 km
in general兲 significantly. In our study, wave heights at 14 offshore variation in epicenter location as 24, 25, and 15%, for earthquakes
locations 共see Fig. 2兲 in Hawaii are simulated and examined. For originated in Aleutian, Japan, and Chile, respectively. 共In this
each location, the change in wave height of the leading wave due paper, all results from the sensitivity analysis are based on the
to the epicenter variation is calculated. Then the relative change wave height of the leading wave.兲
in wave height with respect to the wave height based on the Although in the previous section we mentioned that the epi-
original reference earthquake is averaged over the 14 offshore center location can usually be determined accurate to ±50 km, in
locations. Based on our numerical results, we obtained the aver- the numerical simulation we examined wave changes due to epi-

434 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007
Fig. 7. Effect of dip angle variation on resulting wave height in Hilo
Bay due to earthquake in: 共a兲 Aleutian Islands; 共b兲 Japan; and 共c兲 Fig. 8. Effect of strike angle variation on resulting wave height in
Chile Hilo Bay due to earthquake in: 共a兲 Aleutian Islands; 共b兲 Japan; and 共c兲
Chile

center location variation of ±50, ±100, ±150, and ±200 km. The the maximum allowable block size within which a deviation of
reason we studied a larger variation range than ±50 km is that the earthquake epicenter location will not introduce a large
for any particular earthquake event, there may exist a possibility change in the wave height in the far field.
that the uncertainty in epicenter location is larger than the usual
estimate. Another reason is that we are currently conducting a
Effect of Rake Angle
separate but related study on developing an indirect probabilistic
method for predicting tsunami inundation risk through earthquake In the study by Titov et al. 共1999兲 for tsunamis generated by an
statistics. In this method, we need to distribute scenario earth- earthquake event in the Aleutian area, the sensitivity of wave
quake blocks along the Pacific Rim. We would like to determine heights to rake angle was studied by varying the rake angle from

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007 / 435
Fig. 9. Effect of variations in fault dimensions on resulting wave Fig. 10. Effect of slip variation on resulting wave height in Hilo Bay
height in Hilo Bay due to earthquake in: 共a兲 Aleutian Islands; 共b兲 due to earthquake in: 共a兲 Aleutian Islands; 共b兲 Japan; and 共c兲 Chile
Japan; and 共c兲 Chile

共see Fig. 1 for definition sketch of rake angle兲, the earthquake

90 to 135°. It was found that a 50% 共or 45°兲 change in rake angle generates the highest waves under the condition that the other
would introduce about 20% change in the wave heights in the far parameters are kept the same. This is consistent with the physics
field. In the present study, we revisited this issue and examined of a fault plane and the mechanism of an earthquake in the field,
the sensitivity of wave heights to rake angle by varying it for a namely, a 90° rake angle will induce the largest vertical displace-
larger range, i.e., from 30 to 135°, and by studying earthquakes ment of the fault plane, and therefore generate larger waves. In
originated from more regions in the Pacific Ocean including our study, the following rake angles were studied: 30, 70, 80, 90,
Japan, Aleutian, and Chile. The simulated results are presented in 100, 110, 120, 130, and 135° with the case of 90° as the standard
Fig. 6. These results show that when the rake angle equals 90° and reference scenario. Relative change in the leading wave

436 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007
 Table 3. Average Relative Wave Height Change due to Variations in
Rake Angle
Relative wave
Rake angle height changea
Earthquake location 共deg兲 共%兲
Aleutian 30° 54
70° 8
110° 4
135° 26
Japan 30° 33
70° 1
110° 12
135° 41
Chile 30° 49
70° 9
110° 4
135° 20
a
With respect to the wave height in the reference scenarios with 90° rake
angle.

less than 45° from the common value of 90°, the variation in the
far field wave heights will not be significant. However, when the
rake angle is varied by 60° 共e.g., from 90 to 30°兲, the change in
wave heights in the far field can be larger than 50%, as shown in
the case where the earthquake is originated from the Aleutian
Islands. We should mention that, even though the rake angle is
difficult to determine accurately immediately after an earthquake
occurs, the uncertainty is usually less than 45°. Typical rake angle
values are within the range of 70– 110°. A variation of the rake
angle in this range causes very little change in the wave heights in
the far field.

Effect of Dip Angle

As pointed out by Geist 共1999兲, the dip angle is a permanent
feature of the fault and can be determined relatively accurately
based on aftershock information. From existing earthquake data
共Johnson 1999; Geist 1999; Sato 1989兲, the common values of dip
angle are usually between 10 and 35°. For a particular earthquake,
however, any value may be possible. In this study, we varied the
dip angle from 10 to 90°, a range much larger than the 10– 20°
range examined in the study by Titov et al. 共1999兲. The simulated
results are shown in Fig. 7 and Table 4.
These results show that, within the common range of dip angle
values, a change from 10 to 20° in dip angle causes only a mod-
erate variation in the wave height in the far field for all the three

Table 4. Average Relative Wave Height Change due to Variations in Dip

Angle
Fig. 11. Effect of variations in focal depth on resulting wave height Relative wave
in Hilo Bay due to earthquake in: 共a兲 Aleutian Islands; 共b兲 Japan; and Earthquake Dip angle height changea
共c兲 Chile location 共deg兲 共%兲
Aleutian 20° 19
height due to variations in rake angle was calculated and averaged 90° 64
over the 14 offshore locations in Hawaii. A summary of the quan- Japan 20° 21
titative results for several rake angles is presented in Table 3. 90° 56
We note that our results for the 135° rake angle for the Chile 20° 23
Aleutian source region are consistent with the results from Titov 90° 26
et al. 共1999兲 for the 1996 Andreanov Island tsunami. From the a
With respect to the wave height in the reference scenarios with 10° dip
results in Table 3, we can see that if the rake angle is varied by angle.

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007 / 437
Table 5. Average Relative Wave Height Change due to Variations in Table 7. Average Relative Wave Height Change due to Variation in Slip
Strike Angle Displacement
Strike angle Relative wave Relative wave
Earthquake variation height changea Earthquake Variation in slip height changea
location 共deg兲 共%兲 location 共m兲 共%兲
Aleutian 245– 255° 12 Aleutian from 24 to 48 22
Japan 195– 205° 19 Japan from 24 to 48 39
Chile 5 – 15° 84 Chile from 24 to 48 70
a a
With respect to the wave height produced by the lowest strike angle With respect to the reference scenario cases with slip of 24 m.
value.
of increasing distance along the path of maximum energy radia-
earthquake source regions being studied. However, when the tion on wave sensitivity to fault plane parameters will be further
dip angle is changed to 90°, the variation in the wave height discussed in a later section.
becomes significant, especially if the earthquake is originated
from Aleutian and Japan which are relatively close to Hawaii.
Effect of Fault Dimensions
Based on the present results 关and the results from Titov et al.
共1999兲兴, it is safe for us to conclude that reasonable variations and The fault dimensions 共length and width兲 are among the most un-
uncertainties in dip angle will not introduce significant changes in certain fault plane parameters. Synolakis et al. 共1997兲 commented
the wave height in the far field. that “the deformation area estimated from aftershocks is often
twice that estimated from geodetic data” and that the errors for
length and width can be as high as 75%. Even for the fairly well
Effect of Strike Angle
understood 1960 Chilean earthquake, its width still seems to have
Strike angle indicates the orientation of the fault plane and can be an uncertainty. For example, the width was given by Kanamori
estimated based on the aftershock area. In this study, we varied and Cipar 共1974兲 as 200 km while in Geist 共1999兲, the width was
the strike angle by 10° with respect to the reference cases, and listed as 300 km 共p. 148兲. In this study, we examine how a large
compared the resulting wave height with the original wave height variation in length and width can affect the far field wave height.
based on the reference cases. The results for waves at the mouth As shown in Eq. 共5兲, the seismic moment magnitude is a func-
of Hilo Bay are plotted in Fig. 8 and the results for sensitivity tion of both fault dimension and average slip displacement. In
analysis averaged over 14 offshore locations are presented in order to maintain the same moment magnitude when the fault
Table 5. We can see that a 10° variation in strike angle will not length is varied, the fault width is adjusted as well so as to retain
introduce a significant change to the resulting wave heights at a the same fault area and the same slip displacement in this study.
distant location in Hawaii for earthquakes in the Aleutian and This is slightly different from Titov et al.’s study 共Titov et al.
Japan regions. However, for an earthquake in Chile, a 10° change 1999兲 where the fault dimensions and the slip displacement were
in the strike angle is seen to cause a large relative change in wave adjusted together when determining the effect of fault dimensions
heights in Hilo Bay. This result shows that the sensitivity of the on the leading tsunami wave height.
far field wave heights to a particular earthquake fault plane In the present study, for earthquake scenarios in each of the
parameter does not always decrease as the distance between the three regions, we increased the fault plane length 共width兲 by 50
earthquake location and the far field location increases. The ori- and 100% and examined how these variations would affect the
entation of the earthquake with respect to the far field location, wave height in the far field. In these cases, if the length 共width兲 is
namely the directivity of the initial tsunami wave energy propa- increased by 50%, it means that the width 共length兲 is reduced by
gation, may also affect the wave height in the far field. The effect 33.3% in order to maintain the same magnitude and the same slip.

Table 6. Average Relative Wave Height Change due to Variations in Fault Length and Width
Relative wave
Earthquake height changea
location Change in fault plane length and width 共%兲
Aleutian 50% increase in length and 33.3% reduction in width 12
100% increase in length and 50% reduction in width 46
50% increase in width and 33.3% reduction in length 22
100% increase in width and 50% reduction in length 46
Japan 50% increase in length and 33.3% reduction in width 18
100% increase in length and 50% reduction in width 35
50% increase in width and 33.3% reduction in length 8
100% increase in width and 50% reduction in length 18
Chile 50% increase in length and 33.3% reduction in width 19
100% increase in length and 50% reduction in width 40
50% increase in width and 33.3% reduction in length 11
100% increase in width and 50% reduction in length 11
a
With respect to the wave height in the reference scenarios with 800 km length and 200 km width; the results are averaged over the 14 offshore stations
in Hawaii.

438 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007
Table 8. Wave Height in Hilo due to Different Focal Depth of Table 9. Effects of Distance on Sensitivity
Earthquake in Aleutian Islands
Maximum relative
Focal depth 共km兲 10 20 33 53 73 93 change in leading
Source Variation in fault Distance from wave heighta
Wave height 共m兲 3.67 4.30 4.46 4.08 3.51 2.94
region parameter source 共%兲
Japan 200 km in epicenter Near the source 19
The simulated results are presented in Fig. 9 and Table 6. In Midway 22
the figure, only the results for 100% increase in length or width Offshore Hawaii 22
are shown for waves at Hilo. These results show that even when 100% change in length Near the source 21/40 共L / W兲
we maintain the same magnitude, slip displacement, and total 共L兲 or width 共W兲 Midway 22/33 共L / W兲
fault plane area, a variation in the length/width ratio may affect Offshore Hawaii 26/30 共L / W兲
the wave height in the far field very strongly. In general, an in-
24– 48 m change in Near the source 62
crease in the length 共with a simultaneous reduction in width兲
slip dislocation Midway 43
causes more change in the wave heights in the far field than an
Offshore Hawaii 39
increase in width with a reduction in length. Specifically, a 100% a
increase in fault length coupled with a 50% reduction in width Compared with the wave height in the reference case.
can cause a 35–46% change in the wave heights in the far field. In
addition, for the Chilean scenario, there is a maximum shift of rameters that he investigated include seismic moment, radiation
about 40 min in arrival time for the leading wave when the fault pattern/directivity, focal depth/focal mechanism, and influence of
length is increased by 100%. These results indicate that even sedimentary layering. He solved for analytical solutions for the
when the earthquake magnitude, epicenter location, and fault far field wave heights by applying the linear wave equations as-
plane area remain the same, an uncertainty in the fault length and suming a constant water depth. He found that the focal depth
width ratio can cause a large variation in the wave heights in the plays only a minor role in tsunami generation.
far field. To further investigate and quantify the effect of focal depth on
Geist 共1999兲 pointed out that for large earthquakes, the width tsunami wave generation, we varied the focal depth from 10 to
tends to saturate while the length can keep increasing. Therefore, 93 km in the present numerical study, and present the compara-
for large earthquakes, the fault area will most likely have an elon- tive results in Fig. 11 and Table 8. The results in Fig. 11 show that
gated shape, and the width will not often exceed 300 km. a variation in the focal depth does not affect the wave height in
Hilo Bay significantly for earthquakes originated from a suffi-
ciently far region such as Japan and Chile; however, a variation in
Effect of Slip Displacement
focal depth in an earthquake source in the Aleutian Islands can
The sensitivity analysis results for the slip displacement are cause a moderate change 共e.g., 34%兲 in the generated wave height
shown in Fig. 10 and Table 7. Here the slip displacement is at Hilo Bay. The results in Table 8 show that as the focal depth
doubled from the reference case for all three earthquake source decreases from 93 to 33 km, the wave height increases. This is
regions. Since the earthquake magnitude is maintained the same, expected because if an earthquake happens closer to the earth
based on Eq. 共5兲, the source area needs to be reduced by 50%. In surface, it would cause a stronger disturbance which will generate
other words, slip displacement is a fault parameter that we cannot a higher tsunami wave. However, it is also interesting to note that
completely isolate in our sensitivity analysis as we need to change as the focal depth becomes extremely shallow, for example to less
it with fault area together in order to maintain the same earth- than 20 km, the simulated wave amplitude is seen to become
quake magnitude. In this study, we reduced the fault length and smaller, the reason for which is not directly obvious. In Okal’s
width equally by 29.3% in order to reduce the fault plane area by study 共Okal 1988兲, a similar phenomenon was also observed at
50%. The results show that a 100% increase in the slip displace- focal depth equal to 10 km but for a dip-slip fault model only
ment can cause quite a significant change 共increase兲 in the wave 共i.e., dip angle= 90°兲 and not for other cases with smaller dip
heights in the far field even when the earthquake magnitude is angles as shown in the present study. One possible explanation for
kept the same and the fault area is reduced by half. We note again the “irregular” results with 10 km focal depth can be that the
that the relative change in wave height, or the sensitivity of wave dislocation theory is a linear theory based on elastic deformation.
height to slip displacement, does not decrease as the distance When the focal depth is too shallow, the earth surface displace-
between the earthquake and the far field increases. As shown by ment may become much larger and nonlinear, and therefore a
the case of Chile, which is more than 14 h of traveling time away linear theory will no longer be valid. Another reason could be that
from Hawaii, a 100% increase in slip displacement in an earth- the theory assumes a homogeneous material for the earth, which
quake in Chile can almost double the wave height in Hawaii’s may not be true in reality, especially in the shallow region where
offshore waters. sediment can be layered on top of hard rock 共Okal 1988兲. Further
We would like to comment that since the slip displacement and studies are needed to resolve this question. Before that, we may
the fault area are changed together in this part of the study, we like to be cautious when applying the existing linear elastic theory
cannot attribute the change in wave height to the variation in slip to extremely shallow earthquake sources such as those with focal
displacement alone. For example, when the area is changed, the depth around 10 km.
directivity of the tsunami propagation may change which can also
affect the far field wave height.
Effect of Increasing Distance on Wave Sensitivity
to Fault Plane Parameters
Effect of Focal Depth
The numerical results we have obtained so far indicate that the
Okal 共1988兲 carried out a very helpful study to examine seismic wave height in the far field can be sensitive to several detailed
parameters controlling the far field tsunami amplitude. The pa- fault plane parameters in addition to earthquake magnitude and

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007 / 439
Table 10. Summary of Wave Height Change due to Fault Plane Parameter Variations under Same Earthquake Magnitude
Maximum wave
Earthquake heighta
Parameter Variation location 共%兲
Rake angle 70– 110° Aleutian 8b
Japan 16b
Chile 12b
Dip angle 10– 20° Aleutian 19
Japan 21
Chile 26
Epicenter location 200 km Aleutian 24
Japan 25
Chile 15
Strike angle 10° Aleutian 12
Japan 19
Chile 84
Focal depth 60 km 共33– 93 km兲 Aleutian 34
Japan 11
Chile 10
Fault dimensions 100% increase in Aleutian 46 共22兲
length 共and 50% Japan 35 共8兲
increase in width兲 Chile 40 共11兲
Slip dislocation 24– 48 m Aleutian 22
Japan 39
Chile 70
a
Maximum relative change in the leading wave averaged over the 14 selected coastal locations in Hawaii.
b
Change in leading wave height between the cases with 70 and 110° rake angles with the 70° case as the reference base.

epicenter location. We have found that even for waves at a loca- Summary and Conclusion
tion thousands of miles away from the earthquake source, their
amplitude can still be affected by a deviation in a particular fault In this study, the Cornell COMCOT model is applied to simulate
plane parameter such as the fault area aspect ratio and slip dis- earthquake generated distant tsunamis in the Pacific Ocean. The
placement. As pointed out by Okal 共1988兲 and many other tsu- objective was to examine the effect of different earthquake fault
nami scientists, the directivity of maximum energy radiation is a plane parameters on tsunami wave heights in the far field. Spe-
very important factor in predicting the far field tsunami wave cifically, we are interested in studying how sensitive the tsunami
height. To further investigate the effect of increasing distance on wave heights are to the variations in each fault plane parameter,
far field wave sensitivity to the earthquake parameters, we per- and whether an earthquake can always be considered as a point
formed numerical simulations for tsunamis generated from Japan source for the waves in the far field. After the COMCOT tsunami
and propagating towards Hawaii along the maximum energy ra- model was verified with the field data from the 1960 Chilean
diation direction. The earthquake scenario is the same as the ref- tsunami, it was used in the sensitivity analysis based on three
earthquake scenarios in three seismic regions, namely, Japan,
erence case in Japan. Along the radiation path, three locations are
Aleutian Islands, and Chile. All the fault plane parameters re-
selected for recording the simulated wave heights. They are lo-
quired by tsunami prediction models were studied. They are:
cated near the earthquake, midway between Japan and Hawaii,
epicenter location, rake, dip, and strike angles, fault dimensions,
and offshore of Hawaii, as shown by the three dark dots in Fig. 3.
dip displacement, and focal depth. Wave heights at 14 offshore
The sensitivity of wave height at these three locations to the four
locations around the Hawaiian Islands were computed. Table 10
selected fault plane parameters, namely, epicenter, length and presents a summary of wave height changes due to variations in
width, and slip displacement, is calculated and compared. The fault plane parameters. In this table, the variations in earthquake
detailed results are presented in Table 9. parameters are considered as reasonable or commonly encoun-
From these results we can see that along the path of maximum tered variations and uncertainties. For example, although we in-
energy radiation, the sensitivity of waves to certain fault plane vestigated how wave height can change if the dip angle varies
parameters does not decrease with distance significantly. This in- from 10 to 90°, in reality, the uncertainty in dip angle is rarely
dicates again that the far field wave height can still depend on the this large. In fact, researchers would consider earthquakes with
detailed fault plane parameters. Magnitude and epicenter alone dip angle of 10 and 90° as two different events. Therefore in
may not be sufficient to predict the far field wave height accu- Table 10, we present the wave change due to a reasonable uncer-
rately. Note that the results for “offshore Hawaii” are for one tainty of 10° 共i.e., from 10 to 20°兲 in dip angle, the same as in
single point shown in Fig. 3 and are not averaged over the previ- Titov et al.’s study 共Titov et al. 1999兲 for the 1996 Andreanov
ous 14 offshore locations. Island tsunami.

440 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007
 Our results showed that the far field tsunami wave height in Mar. Geol., 203, 201–218.
Hawaii is not sensitive to the variations in rake and dip angles and Geist, E. L. 共1999兲. “Local tsunamis and earthquake source parameters.”
the epicenter location of the earthquake in all three seismic re- Adv. Geophys., 39, 117–198.
gions studied. This part of the results confirms the previous pre- Gica, E. 共2005兲. “Risk analysis of coastal flooding due to distant tsuna-
dictions by Geist 共1999兲 and Titov et al. 共1999兲 that wave heights mis.” Ph.D. thesis, Univ. of Hawaii at Mānoa, Honolulu.
in the far field may not be affected by certain detailed earthquake Imamura, F. 共1996兲. “Review of tsunami simulation with a finite differ-
fault plane parameters. However, by examining the effects of all ence method.” Long-wave runup models, H. Yeh, P. L.-F. Liu, and C.
fault plane parameters, our study also revealed some new find- Synolakis, eds., World Scientific, River Edge, N.J., 25–42.
ings. Specifically, in this study, the fault dimensions, the slip dis- Imamura, F., Shuto, N., and Goto, C. 共1988兲. “Numerical simulations of
placement, and the strike angle are found to affect the wave the transoceanic propagation of tsunamis.” Proc., 6th Congress APD-
heights the most significantly even for far fields located thousands IAHR, 265–272.
of miles away from the earthquake location. In addition, the Johnson, J. M. 共1999兲. “Heterogeneous coupling along Alaska-Aleutians
sensitivity of the far field wave heights to certain earthquake pa- as inferred from tsunami, seismic, and geodetic inversions.” Adv. Geo-
rameters including the strike angle, fault dimensions, and slip phys., 39, 1–110.
displacement does not seem to decrease as the distance between Kanamori, H., and Cipar, J. J. 共1974兲. “Focal process of the great Chilean
the earthquake and the far field increases. This indicates that even earthquake May 22, 1960.” Phys. Earth Planet. Inter., 9, 128–136.
for a coastal location thousands of miles away, the earthquake Liu, P. L.-F., Cho, Y.-S., Briggs, M. J., Kanoglu, U., and Synolakis, C. E.
may not be treated as a point source described by earthquake 共1995兲. “Runup of solitary waves on a circular island.” J. Fluid
magnitude and epicenter location alone. Finally, we observed that Mech., 302, 259–285.
the linear dislocation theory commonly used by the tsunami com- Liu, P. L.-F., Cho, Y.-S., Yoon, S.-B., and Seo, S.-N. 共1994兲. “Numerical
munity may not be valid when the focal depth is extremely shal- simulations of the 1960 Chilean tsunami propagation and inundation
low such as less than 20 km. at Hilo, Hawaii.” Recent development in tsunami research, M. I.
We would like to mention that tsunami is a very complex El-Sabh, ed., Kluwer Academic, Dordrecht, The Netherlands, 99–115.
phenomenon to predict accurately when certain earthquake pa- Liu, P. L.-F., Lynett, P., Fernando, H., Jaffe, B. E., Fritz, H., Higman, B.,
rameters are not known. Tsunami researchers have contributed Morton, R., Goff, J., and Synolakis, C. 共2005兲. “Observations by the
their best efforts in developing forecasting and predicting models international tsunami survey team in Sri Lanka.” Science, 308, 1595.
for the past several decades. Their efforts and results are very Liu, P. L.-F., Woo, S.-B., and Cho, Y.-S. 共1998兲. “Computer programs for
valuable and useful for hazard mitigation and coastal manage- tsunami propagation and inundation.” Technical Rep., Cornell Univ.,
ment even if some of the earthquake parameters are estimated Ithaca, N.Y.
through empirical formulas due to the limitation of current tech- Mansinha, L., and Smylie, D. E. 共1971兲. “The displacement fields of
nology for measuring all earthquake parameters accurately. The inclined faults.” Bull. Seismol. Soc. Am., 61, 1433–1444.
results from this study are helpful to quantify the effects of dif- Okal, E. A. 共1988兲. “Seismic parameters controlling far-field tsunami
ferent fault plane parameters on tsunami generation, and to esti- amplitude: A review.” Natural Hazards, 1, 67–96.
mate the accuracy of tsunami predictions when some of the fault Sato, R. 共1989兲. Handbook of earthquake fault parameters in Japan,
plane parameters are not known and must be approximated. Kajima Institute Publishing, Kajima, Japan 共in Japanese兲.
Synolakis, C., Liu, P. L.-F., Carrier, G., and Yeh, H. 共1997兲. “Tsunami-
genic seafloor deformations.” Science, 278, 598–600.
Acknowledgments
Titov, V. V., Gonzales, F. I., Bernard, E. N., Eble, M. C., Mofjeld, H. O.,
Newman, J. C., and Venturato, A. J. 共2005兲. “Real-time forecasting:
This study was partially funded by the NOAA Sea Grant College
challenges and solutions.” Natural Hazards, 35, 41–58.
Program and the Joint Institute for Marine and Atmospheric Re-
Titov, V. V., Gonzales, F. I., Mofjeld, H. O., and Newman, J. C. 共1999兲.
search 共JIMAR兲 at the University of Hawaii. It is also funded by
“Offshore forecasting of Alaska-Aleutian subduction zone tsunamis in
the Joint Institute for the Study of the Atmosphere and Ocean Hawaii.” NOAA Technical Memorandum No. ERL PMEL-114,
共JISAO兲 at the University of Washington under NOAA Coopera- NOAA, Seattle.
tive Agreement No. NA17RJ1232, Contribution No. 1197. This Titov, V. V., and Synolakis, C. E. 共1998兲. “Numerical modeling of
publication is considered as Contribution No. 2877 from NOAA/ tidal wave runup.” J. Waterway, Port, Coastal, Ocean Eng., 124共4兲,
Pacific Marine Environmental Laboratory. The third writer P. 157–171.
L.-F. Liu would like to acknowledge the support from the Na- Wang, X., and Liu, P. L.-F. 共2005兲. “A numerical investigation of
tional Science Foundation, NOAA Sea Grant Program in New Boumerdes-Zemmouri 共Algeria兲 earthquake and tsunami.” Comput.
York, and the University of Alaska for the development of the Model. Eng. Sci., 10共2兲, 171–184.
Cornell COMCOT tsunami simulation model. Helpful discussions Wang, X., and Liu, P. L.-F. 共2006兲. “An analysis of 2004 Sumatra earth-
with Dr. Gerald Fryer and Dr. Barry Hirshorn of the Pacific Tsu- quake fault plane mechanisms and Indian Ocean tsunami.” J. Hy-
nami Warning Center, Dr. Barbara Keating of the University of draul. Res., 44共2兲, 147–154.
Hawaii, Dr. Yong Wei of the NOAA/PMEL Center for Tsunami Whiteside, L. S., Dater, D. T., Dunbar, P. K., Racey, S. D., Buhmann,
Research, and Professor Tomoyuki Takahashi of Akita University R. W., and Hittelman, A. M. 共2000兲. Earthquake seismicity catalog,
in Japan are greatly appreciated. Vols. 1 and 2, 共CD-ROM兲, National Geographical Data Center,
NOAA, Boulder, Colo.
References Zhou, H., Teng, M. H., and Lin, P. 共2005兲. “The Cornell 2D numerical
model for long wave runup.” Proc., Joint ASME/ASCE/SES Conf.
Fryer, G. J., Watts, P., and Pratson, L. F. 共2004兲. “Source of the great on Mechanics and Material (McMat-2005) 共CD-ROM兲, Baton Rouge,
tsunami of 1 April 1946: A landslide in the upper Aleutian forearc.” La.

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / NOVEMBER/DECEMBER 2007 / 441