Nat Hazards (2014) 71:2043–2061

DOI 10.1007/s11069-013-0995-1

ORIGINAL PAPER

Application of fragility curves to estimate building

damage and economic loss at a community scale: a case
study of Seaside, Oregon

Dane M. Wiebe • Daniel T. Cox

Received: 22 April 2013 / Accepted: 6 December 2013 / Published online: 18 December 2013
Ó Springer Science+Business Media Dordrecht 2013

Abstract Community-scale estimates of building damage and economic loss are mod-
eled for Seaside, Oregon, for Cascadia subduction zone events ranging from 8.7 to 9.3 MW
with corresponding slip distances of 3–25 m considering only the effects of the tsunami.
Numerical simulations are obtained from the National Oceanic and Atmospheric Admin-
istration’s method of splitting tsunami model which includes a source model, subsidence,
and calculations of the propagation and inundation flow characteristics. The damage
estimates are based on fragility curves from the literature which relate flow depth with
probability of damage for two different structural materials of buildings. Calculations are
performed at the parcel level for the inundation hazard without including damage caused
by the earthquake itself. Calculations show that the severity of building damage in Seaside
is sensitive to the magnitude of the event or degree of slip because the majority of the city
is located on low-lying coastal land within the estimated inundation zone. For the events
modeled, the percentage of building within the inundation zone ranges from 9 to 88 %,
with average direct economic losses ranging from $2 million to $1.2 billion.

Keywords Tsunami  Inundation  Fragility  Damage  Forces  MOST 

Economic loss  Cascadia subduction zone

1 Introduction

The 2011 Tohoku earthquake and tsunami event illustrated the destructive power of tsu-
namis. This disaster had casualties in the tens of thousands of people and destroyed
hundreds of thousands of buildings and other infrastructure (Mori et al. 2011). Post-event

D. M. Wiebe (&)  D. T. Cox

School of Civil and Construction Engineering, Oregon State University, Corvallis, OR 97331, USA
e-mail: [email protected]; [email protected]
D. T. Cox
e-mail: [email protected]

123
2044 Nat Hazards (2014) 71:2043–2061

field surveys from the 2011 Tohoku event have been used to develop fragility curves to
estimate building performance to tsunami loads (Suppasri et al. 2012). These curves can
aid both design engineers with building new infrastructure and policy makers with eval-
uating likely damage scenarios from future events. The latter is the focus of this study
where Seaside, Oregon, is used as a case study.
The Cascadia subduction zone (CSZ) measures 1,000 km in length and extends from the
Mendocino Ridge off the coast of northern California to northern Vancouver Island, British
Columbia (Fig. 1). A near-field event generated from the CSZ is expected to cause
widespread damage to the northwest Pacific coast of North America with the first waves
arriving in the tens of minutes (CREW 2013).
The last great CSZ event occurred more than three centuries ago on January 26, 1700 and
was a full-length rupture. The event is estimated to have had a moment magnitude (MW)
between 8.7 and 9.2, and a slip of 19 m (Satake et al. 2003). Over the past 10,000 years, the
CSZ has shown three typical ruptures scenarios: a rupture of 200–450 km of the southern
margin with 18–20 events on the order of 8.2 MW, a rupture of 650 km starting at the southern
margin with 3–4 events on the order of 8.5 MW, and a full-length rupture with 19–20 events
on the order of 8.9 MW (Goldfinger et al. 2012). The average recurrence interval between full-
length CSZ events is 530 years, and the next event is estimated to have a 7–12 % probability
of occurrence by 2060 (Goldfinger et al. 2012).
To asses the hazard from the CSZ event, Gonzalez et al. (2009) developed probabilistic
tsunami wave heights at Seaside, Oregon, for the 100- and 500-year events. The 100-year
hazard is most often caused by the far-field Alaska-Aleutian subduction zone with wave
heights less than 4 m. The 500-year hazard is most often caused by the near-field CSZ with
wave heights in excess of 10 m near the shoreline. As acknowledged by the authors, this
paper does not include values of velocity or momentum flux which are better estimator of
damage and only includes the maximum wave height which can lead to underestimates of
the tsunami hazard in some areas.

Fig. 1 Regional map of study area showing the Cascadia subduction zone

123
Nat Hazards (2014) 71:2043–2061 2045

Using the tsunami regulatory maps (Oregon Senate Bill 379) developed by the State of
Oregon Department of Geology and Mineral Industries, Wood (2007) assessed the expo-
sure and sensitivity of coastal Oregon communities to the tsunami hazard. Using geo-
graphic information system (GIS) software and land use, population, and tax lot datasets,
the work focused on documenting developed lands, populations, economic assets, and
critical facilities relative to the tsunami inundation zone. For economic assets, aggregate
quantities within the inundation zone were totaled to determine the community’s exposure
level and compared to the total assets to determine the sensitivity. The work identified
Seaside, Oregon, as the most vulnerable community along the Oregon coast to the tsunami
hazard in terms of exposure and sensitivity. While the purpose of this study was not to
estimate the probable economic loss from the tsunami event, the aggregate value of assets
within the inundation zone calculated by Wood provides the upper bound of possible loss.
Dominey-Howes et al. (2010) provided a refined probable economic loss estimate of
Seaside, Oregon, by accounting for the spatial variation in the hazard. For this study, the
authors used the 500-year tsunami event from the Seaside pilot study (Tsunami Pilot Study
Working Group 2006) and Clatsop County tax lot data as input into the Papathoma tsunami
vulnerability assessment (PTVA) model to estimate probably maximum loss. The model is
quite extensive in that it accounts for various aspects of the hazard such as flow depth,
building row from the sea, building material, number of stories, orientation, building
condition, surroundings, and land cover; however, the vulnerability score is calculated by a
summation of the standardized scores multiplied by a weighting coefficient. The problem
with this approach arises from assigning an appropriate weighting coefficient to each
criterion. For this study, the authors state that the values are based on expert judgment, but
the PTVA weighting coefficients appear to be somewhat arbitrary, linearly decreasing from
8 to 1. The authors acknowledge a lack of credible fragility curves available at the time of
this study and mention that these curves could be incorporated into the PTVA model as
they become available.
In this paper, we provide a framework for estimating economic loss at the community
scale for the tsunami hazard, using Seaside, Oregon, as a case study. The numerical model,
method of splitting tsunami (MOST), is used to generate the tsunami hazard, and probable
damage calculations are performed at the parcel level using tax lot data and fragility
curves. Based on the probability of damage, three accounting methods are used to
aggregate the total probable economic loss from an event. For this study, only the value of
the buildings lost were estimated, which is a subset of the true economic loss. To evaluate
the sensitivity of economic loss to event magnitude, five hypothetical events were gen-
erated and compared. The intention of this paper is to show the overall methodology for
combining hydrodynamic modeling, fragility modeling, and estimates of economic loss to
buildings including the sensitivity to the inundation levels using Seaside, Oregon as an
example, rather than an estimate of the total economic losses for this area in particular.

2 Numerical model simulations

The numerical simulation model, method of splitting tsunami (MOST), maintained by the
Pacific Marine Environmental Laboratory (PMEL) of the National Oceanic and Atmo-
spheric Administration (NOAA), is used in the paper to model a series of tsunami events at
Seaside, Oregon. MOST is a finite difference model, based on the depth integrated non-
linear shallow water wave equations, accounts for wave dispersion, includes a Manning’s
term for friction, and uses a set of three nested grids (PMEL 2006; National Tsunami

123
2046 Nat Hazards (2014) 71:2043–2061

Hazard Mitigation Program 2012). The model proceeds in three distinct phases: defor-
mation, propagation, and inundation. The deformation phase provides the initial hydro-
dynamic tsunami parameters for a specified event and is based on the Okada deformation
model (Okada 1985). All of the major subduction zones around the world have been
delineated by PMEL into 50 9 100 km tiles, which act as unit sources, and include the
strike, dip, and rake angles, and the depth of the epicenter, which are used to calculate the
deformation for a given magnitude event. The propagation phase uses the non-linear
shallow water wave equations to propagate the waves generated by the deformation across
the oceans to the shoreline. The inundation phase uses a 1D algorithm, derived from the
Vasily Titov Costas Synolakis (VTCS) model (Titov and Synolakis 1995), to move the
shoreline position based on a horizontal projection of the water level. Further information
regarding the governing equations, numerical implementations, and validation can be
found in Titov and Synolakis (1998), PMEL (2006), and Tang et al. (2009).
Community Model Interface for Tsunami (ComMIT) is a java program which provides
a graphical user interface to run MOST. It also provides access to a database of pre-
computed tsunami events for each 50 9 100 km unit source (http://sift.pmel.noaa.gov/
data). Using this propagation database, tiles representing the fault of interest are selected,
and the magnitude of the event is specified. ComMIT accesses the shared database and
generates the initial and boundary conditions for the grids. Further details of ComMIT and
the propagation database are available in Titov et al. (2011) and Gica et al. (2008).

2.1 Digital elevation models (DEM)

DEM data for this study was obtained from high-resolution DEMs provided by NOAA’s
National Geophysical Data Center (NGDC). The DEMs included Astoria (gov.-
noaa.ngdc.mgg.dem:4090), Garibaldi (gov.noaa.ngdc.mgg.dem:249), and Seaside,
(Gonzalez et al. 2009) Oregon, and the NW Pacific coast (NOAA NGDC, U.S. Coastal
Relief Model) with coverage from 116°W to 128°W and 44°N to 49°N. The DEMs are
referenced to mean high water (MHW) which is common convention with model tsunami
inundation as it provides a conservative estimate. Table 1 lists the details, extent, and
resolution of each grid, and Fig. 2 shows the details and extents.

2.2 Modeled events

Earthquake intensity is commonly reported in terms of moment magnitude, MW, which is a

function of the shear modulus of the fault material, the area displaced, and slip distance.

Table 1 Digital elevation model

Grid A Grid B Grid C
parameters
Lat north (°) 48.00 46.54 46.05
Lat south (°) 44.00 45.65 45.90
Lon north (°) -123.00 -123.51 -123.89
Lon south (°) -127.00 -124.40 -124.04
Cell size (arc sec) 36.00 4.00 1.00
Ncols 400 800 538
Nrows 400 800 645
Nodes 160,000 640,000 347,010

123
Nat Hazards (2014) 71:2043–2061 2047

Fig. 2 Digital elevation models for grids a (left), b (middle), and c (right). The boxed regions within grid
a and b represent the regions of grids b and c, respectively

For the full-length CSZ ruptures, the displaced area remains relatively constant (Goldfinger
et al. 2012), so the increased MW acts to increase the slip distance. For subduction zone
earthquakes, slip is a main mechanism in determining the magnitude of a tsunami. Other
key parameters include the displaced area, the fault type, and the angles at which the
displacement occurs.
Figure 3 shows the sensitivity of the slip as a function of MW for rupture lengths
ranging from 400 to 1,200 km. In all cases, the rupture width is 100 km, which corre-
sponds to the width of the CSZ, for equivalent rupture areas of 40,000–1,200,000 km2. For
constant 9.0 MW events with rupture lengths of 400 and 1,200 km, the corresponding slip
distances are 22.2 and 7.4 m, respectively, a threefold decrease. The calculated slip dis-
tances of a few recent and historically significant events are also plotted on this figure.
Note there is uncertainty in the estimated moment magnitudes and rupture lengths due to
fault asperities. The historical values were taken from Yoshida et al. (2011), Delouis et al.
(2010), Hirata et al. (2006), Ichinose et al. (2007), and Goldfinger et al. (2012).
For this paper, a full-length CSZ rupture (100 km wide by 1,000 km long) is modeled
for five simulated tsunami events ranging from 8.7 to 9.3 MW. These are hypothetical
events used to illustrate the sensitivity of slip distance to tsunami severity and should
provide an envelope of likely damage states. Table 2 lists the details of each simulated

Fig. 3 Fault slip as a function of earthquake magnitude for various rupture length, and notable historical
event

123
2048 Nat Hazards (2014) 71:2043–2061

event. For these events, the slip distance ranges from 3.15 to 25.00 m. Figure 4 shows the
deformation from a hypothetical 9.0 MW earthquake, with a rupture length and width of
1,000 and 100 km, respectively. Over this area, the average strike and dip angles are 345°
and 10°, respectively. At the fault, the maximum uplift is 3.4 m, and along the coast, the
maximum subsidence is 1.5 m. The extents of grids A, B, and C are also shown in the
figure. Note that while Suppasri et al. (2010) have shown that fault rupture velocity can
effect the wave height and arrival time of tsunamis, MOST treats the rupture process as an
instantaneous event and is assumed to have minimal impact on the results.

2.3 Numerical results

Figure 5 shows the maximum extent of inundation for the five modeled events. The extent
of inundation is sensitive to slip, which is directly related to MW since the displaced area is
held constant for this work. For 3 m of slip, the extent of inundation closely resembles the
existing shoreline (blue line) and it does not overtop the first peninsula formed by the
Necanicum River. The 6-m slip event (green line) inundates the first peninsula but does not
inundate the second half of the city between the Necanicum River and Neawanna Creek.

Table 2 Synthetic tsunami events modeled

Case MW Segment Length (km) Width (km) Slip (m)

1 8.7 acsz–56a to acsz–65b 1,000 100 3.15

2 8.9 acsz–56a to acsz–65b 1,000 100 6.28
3 9.0 acsz–56a to acsz–65b 1,000 100 8.87
4 9.1 acsz–56a to acsz–65b 1,000 100 12.53
5 9.3 acsz–56a to acsz–65b 1,000 100 25.00

Fig. 4 Deformation along the CSZ from a 9.0 MW event for a 1,000 km by 100 km rupture

123
Nat Hazards (2014) 71:2043–2061 2049

The 9-m slip event (yellow line) inundates most of the low-lying portion of Seaside. The
13- and 25-m slip events increase the extent of inundation to the base of the small coastal
mountain range. The white box in Fig. 5b highlights a section of detail that will be used in
later figures.
Figure 6 shows a time series of the offshore wave amplitude at the 10 m bathymetric
contour for the first 4 h for the 5 events. The wave amplitude is defined as the vertical
displacement of the free surface referenced to mean high water (MHW). The location of
the virtual wave gage is shown in Fig. 5a. For all 5 events, the maximum wave amplitude
is associated with the first wave and reaches the shoreline in approximately 35 min. The
maximum wave amplitude at the location of the wave gage for the 3-, 6-, 9-, 13-, and 25-m
slip events are 1.9, 3.7, 5.1, 7.0, and 12.2 m, respectively.
Figure 7 shows the maximum values of the wave amplitude or flow depth, velocity, and
momentum flux for the 13-m slip event for the C grid. In this paper, all figures relating to
the free surface report the wave amplitude for offshore values and flow depth for onshore
values. The flow depth is defined as the vertical distance from the land elevation to the free
surface elevation. At the shoreline, the boundary between these two regions, the wave
amplitude and flow depth are equal by definition. The momentum flux per unit mass per
unit width is defined as a flow depth multiplied by a squared velocity and is sometimes
termed the specific momentum flux. In this paper, we retain ‘‘momentum flux’’ for brevity.

Fig. 5 Regional plan view (a) and community plan view (b) of the maximum extent of inundation for the
five modeled events. The white box bounds the downtown region which is examined in detail. The Secchi
disk in a is the location of a virtual wave gage for Fig. 6. Slip = 3 (blue), 6 (green), 9 (yellow), 13 (orange),
and 25 m (red)

123
2050 Nat Hazards (2014) 71:2043–2061

Fig. 6 Time series of the offshore wave amplitude at the 10 m contour. Slip = 3 (blue), 6 (green), 9
(yellow), 13 (orange), and 25 m (red). The location of the wave gage is shown in Fig. 5a

The maximum momentum flux is the peak of the momentum flux time series, and it is
noted that the flow depth and velocity associated with the momentum flux are less than
either the maximum values of flow depth or velocity for a given event. For the a 13-m slip
event, which is most similar to the 1,700 event, the maximum offshore wave amplitude,
velocity, and momentum flux near the shoreline reach values as high as of 8 m, 10 m/s, and
200 m3/s2, respectively.
Figure 8 shows the detailed maximum values of flow depth, velocity, and momentum
flux for the 13-m slip event for the region represented by the white box in Fig. 5a. The
dimensions are 450 m east–west by 600 m north–south and indicate one of the main
commercial centers of Seaside. The black dotted lines represent the road network, and a
circle is used to indicate the roundabout at the end of Broadway Street marking a main
thoroughfare running perpendicular to the shoreline. Along the western most road near the
circular roundabout, the flow depth decreases from 8 to 3 m, while the velocity and
momentum flux increases from 4 to 8 m/s and from 75 to 150 m3/s2, respectively. The
spatial and temporal variation in maximum flow depth, velocity, and momentum flux has
been shown to be significant (Park et al. 2013). This spatial variation in the hazard
highlights the need and importance of understanding how building failure can be estimated
from flow depth, velocity, or momentum flux (Yeh et al. 2005; FEMA 2008; Muhari et al.
2011) and that damage predictions based solely on flow depth, velocity, or momentum flux
would give markedly different results.
It is noted that the calculations assume a bare earth DEM model, and the influence of the
macro-roughness or large buildings is not included. The addition of buildings to the DEM would
significantly alter the hydrodynamics (Park et al. 2013; Rueben et al. 2011; Cox et al. 2008),
such as accelerating flows between buildings. As well, some buildings may be expected to fail
during the tsunami, and the DEM would require a temporal variation during the model run to
correctly model the hydrodynamics. Therefore, a bare earth DEM model with an appropriate
friction factor will provide a conservative estimate. Kaiser et al. (2011) showed the importance
of the spatial variation in roughness, such as vegetated and urban areas, where the increased
roughness can decrease both the flow velocity and extent of inundation.
Figure 9 shows the maximum flows depths associated with the 6-, 9-, and 13-m slip
events. The overall trend of high and low values remain constant as the magnitude of the
event increases, but the average flow depths over the domain increases from 1.5 to 2.9 to

123
Nat Hazards (2014) 71:2043–2061 2051

Fig. 7 Maximum values of a wave amplitude and flow depth b velocity and c momentum flux in the C grid
of MOST for a 13-m slip event

Fig. 8 Maximum values of a flow depth b velocity and c momentum flux in downtown Seaside for a 13-m
slip event. Road network (dotted lines)

4.7 m, respectively. For the cases shown, a 0.1 increase in the seismic moment increases
the slip and flow depth approximately 3 and 1.5 m, respectively.

3 Method of damage estimation

Damage estimates were derived from fragility curves published in the literature. A fragility
curve is a statistical function which describes the performance (or damage state) for a given
demand (or loading condition). The curves are typically S-shaped, which describes the
uncertainty in the system’s capacity to withstand a loading condition (Schultz et al. 2010).
For example, a gradual curve implies a high uncertainty in the performance for a given
demand, whereas a steep curve implies a high certainty in the performance. Fragility
curves with high uncertainty may lead to an under prediction of performance at low

123
2052 Nat Hazards (2014) 71:2043–2061

Fig. 9 Maximum flow depths in downtown Seaside for slip = 6 (a), 9 (b), and 13 m (c). Road network
(dotted lines)

demands, and over prediction of performance at high demands (Schultz et al. 2010). There
are typically four methods used to develop fragility curves: judgmental, empirical, ana-
lytical, and hybrid (Schultz et al. 2010).
For tsunami performance, fragility curves have typically been developed using an
empirical approach of field observations, laboratory experiments, and numerical simula-
tions (e.g., Koshimura et al. 2009). An advantage of using fragility curves is that they
incorporate all of the hazards and uncertainty into a single function. However, extreme care
must be taken to ensure that the fragility functions are appropriate for the intended
application. For example, fragility curves developed by Koshimura et al. (2009), Suppasri
et al. (2011), and Murao and Nakazato (2010) from the 2004 Indian Ocean tsunami for
Indonesia, Thailand, and Sri Lanka would be inappropriate for application along the
Oregon coast for a CSZ event or other areas with different building standards as suggested
by Koshimura et al. (2009). First, the construction quality and building standards in the US
are generally different than those in southeast Asian countries. Therefore, the performance
under the same loading condition is expected to be different. Second, the curves developed
for regions far from the epicenter, such as Sri Lanka, do not include the near-field
earthquake hazard. Park et al. (2012) have shown a numerical approach of combining the
probabilistic seismic and tsunami hazards; however, this was not performed for this study.
While building standards and conditions in the US and Japan are different, of the available
tsunami fragility curves developed to date, they are the most similar. Therefore, the recent
fragility curves developed from the 2011 Tohoku event are the most appropriate for our
application (e.g., Suppasri et al. 2012).
Figure 10 shows the fragility curves developed by Suppasri et al. (2012) from the 2011
Tohoku event which are used in this study. These curves have the advantage of providing
four performance levels ranging from minor to complete damage for both wooden and
concrete/steel buildings. Table 3 lists descriptions of the severity of minor, moderate,
major, and complete damage. A limitation of these fragility curves are that they are all
based on the flow depth, and fragility curves for velocity and momentum flux from the
2011 event are still in development. The curves developed by Suppasri et al. (2012) show

123
Nat Hazards (2014) 71:2043–2061 2053

Fig. 10 Tsunami fragility curves for damage levels to wooden house and concrete/steel buildings as a
function of flow depth. Minor damage (solid), moderate damage (dashed dotted), major damage (dashed),
and complete damage (dotted), wooden houses (black) and concrete/steel buildings (gray). Data replotted
from Suppasri et al. (2012)

Table 3 Description of various

Damage level Description
damage levels taken from Sup-
pasri et al. (2012)
Minor Window is damaged but no damage on wall
Moderate Window and one part of wall are damaged
Major Window and large part of wall are damaged
Complete Window, wall, and column are damaged

an initiation of probabilities of damage for flow depths around 1 m and complete damage
beyond 7 m (Fig. 10).
Damage estimates are performed on the parcel level in our study area. Tax lot data from
2011 collected by the Clatsop County assessor of the Assessment and Taxation Department
were used to obtain information on each parcel. Each parcel is categorized with a three-
digit property classification which is used to assign a building type (wooden or concrete/
steel). As a simplifying assumption, all residential structures are taken as wooden and all
commercial structures are taken as concrete/steel. Based on several site visits, this
assumption is reasonable, particularly for residential buildings which are nearly all wood
and for large, newer hotels which typically have modern construction. Some of the older,
smaller hotels and small businesses are a mix of wood and concrete construction. The
assignment of building type could be refined by performing a more detailed field survey
and assigning types individually or assigned probabilistically based on a percentage of
buildings types. However, this second approach would introduce a random spatial distri-
bution which would require statistical analysis (e.g., Monte Carlo simulation). The real
market value (RMV) of each parcel is divided between land value and building value. For
this analysis, only the building values were used and we did not model loss to assets within
the buildings or changes to the land value.
Figure 11 shows aerial imagery, building type, and building value for downtown Sea-
side. The shoreline is fronted by large hotels, the north landward area is a commercial

123
2054 Nat Hazards (2014) 71:2043–2061

Fig. 11 Downtown Seaside: aerial imagery (a), construction type (b), and building value (c)

district, and the south area is manly residential dwellings (gray shaded building in
Fig. 11b). We show only this portion of Seaside for illustration, and we calculate damages
based on the inundation zones shown in Fig. 5.

3.1 Estimate of physical damage

Figure 12 shows the probability of moderate damage for the 9-, 13-, and 25-m slip events
by applying the fragility analysis (Fig. 10) to the building types (Fig. 11b) assumed from
the tax lot data and the flow depths estimated by MOST (Fig. 9). In general, the figure
clearly shows the increasing damage as the degree of slip increases. The damage patterns
in Fig. 12a directly correlate with the flow depth patterns in Fig. 9c. The only differences
are lower probabilities of damage in the northern commercial area in comparison with the
southern residential area due to increased performance of concrete/steel buildings over
wooden buildings (compare Fig. 12a with Fig. 11b). The probability of moderate damage
increases significantly between these events. For the 9.0-m slip event, most of the buildings
have less than 50 % probability of moderate damage, whereas for the 25-m slip event, the
probability increases to greater than 75 %.

Fig. 12 Probability of moderate damage for slip = 9 (a), 13 (b), and 25 m (c)

123
Nat Hazards (2014) 71:2043–2061 2055

Fig. 13 Probability of moderate (a), major (b), and complete (c) damage for slip = 13 m

Figure 13 shows the probabilities of moderate, major, and complete damage for the
13-m slip event. The majority of the buildings have a probability of moderate damage
greater than 75 %, but a probability less than 50 % for complete damage. The spatial
scatter of probability of damage shown in this figure is similar to the level of scatter shown
in the vulnerability score of Dominey-Howes et al. (2010).
As mentioned previously, the DEMs used to compute the hydrodynamics are bare earth
models and do not account for the spatial sheltering afforded by other buildings which may
lower damage estimates, or accelerate flows between building which may increase damage
estimates. Numerical modeling of tsunami inundation for constructed environments has
been successfully performed and verified by physical laboratory studies (e.g., Park et al.
2013). However, using the hydrodynamics for community scale, damage estimates intro-
duce new challenges due to the temporal variation in building failure which would need to
be incorporated into the DEM to properly model the hydrodynamics and the debris field
which may change the severity of damage in some areas (Cercone et al. 2013).
Table 4 lists the number of buildings within the inundation zone for the five modeled
events and the number of buildings for various probable damage states. The center portion
of this table lists the number of buildings with probabilities of moderate damage greater
than 25, 50, and 75 %. The left portion of this table lists the number of buildings with
minor, moderate, major, and complete damage with probabilities of occurrence greater
than 50 %. Within the study area (C grid), there are a total of 10,043 buildings. Over the
five modeled events, the number of buildings within the inundation zone increased from
908 to 8,846, which account for 9–88 % of all buildings.
Comparing the total number of buildings with probabilities of occurrence greater than
25, 50, and 75 % for a given event provides the sensitivity of damage for a single damage
state. For example, for the probability of moderate damage for the 9-m slip event, the
number of buildings impacted decreases from 1,754 to 963 to 448 which represents a 45
and 53 % decrease in the number of buildings impacted between 25 and 50 %, and 50 and
75 %, respectively.
Comparing the total number of buildings with probabilities of occurrence greater than
50 % for minimal, moderate, major, and complete damage states provides the sensitivity
for a given event. For example, for the 9-m slip event, the number of buildings impacted
decreases from 1819 to 963 to 200 to 132 which represents a 47, 79, and 34 % decrease in
the number of buildings impacted between minor and moderate, moderate and major, and
major and complete damage states, respectively.

123
2056 Nat Hazards (2014) 71:2043–2061

Table 4 Total number of buildings within the inundation zone, total number of buildings with a probability
of occurrence greater than 25, 50, and 75 % for an estimate of moderate damage, and total number of
buildings estimated to have greater the 50 % probability of occurrence for minor, moderate, major, and
complete damage
Slip Total P Mod. Damage Number of buildings P [ 50 %

(m) # Build. [ 25 % [ 50 % [ 75 % Min. Mod. Maj. Com.

3 908 48 39 0 48 39 0 0
6 4,508 233 149 67 234 149 52 48
9 5,883 1,754 963 448 1,819 963 200 132
13 6,661 5,326 4,585 3,417 5,356 4,585 2,328 1,469
25 8,846 8,021 7,775 7,572 8,032 7,775 7,414 7,282

And comparing the total number of buildings with probabilities of occurrence greater
than 50 % and a given damage state for multiple events provides the sensitivity to events.
For example, for moderate damage with probabilities of occurrence greater than 50 %, the
number of buildings impacted decreases from 7,775 to 4,585 to 963 to 149 to 39 which
represents a 41, 79, 85, and 74 % decrease in the number of buildings impacted between
25- and 13-, 13- and 9-, 9- and 6-, and 6 and 3-m slip events, respectively.
Figure 14 shows the number of buildings with probabilities of occurrence greater than
50 % for minor, moderate, major, and complete damage for 3-, 6-, 9-, 13-, and 25-m slip
events which are listed in Table 4. This figure illustrates the sensitivity of damage for
Seaside to events near 9 m of slip. For 6 m of slip relatively few buildings are expected to
be damage, only 2.3 % of buildings have probabilities of minor damage above 50 %. For
13 m of slip, 53 % of buildings have probabilities of minor damage above 50 %.

3.2 Estimate of monetary damage

For this paper, only the direct tangible economic loss was tabulated, which is damage to
buildings. Direct intangible loss, such as loss of life, and indirect tangible loss, such as the
disruptions to the regional economy, were not considered. It is noted that the tax lot data
included information on the land value and the improved value. Only the improved value
(value of the structure) was used and did not include assets to business (e.g., dry goods or
furnishings) inside the buildings.
To calculate the economic loss using the probabilities of damage and the real market
value (RMV) of the building from the tax lot dataset, the authors developed three methods
which are listed below:
Method (1) Aggregates a percentage of the RMV for all buildings with probabilities
greater than zero.
X
ELoss ¼ RMV  P
Method (2) Aggregates the full RMV for all buildings with probability of damage
greater than a threshold of 50 %.
X
ELoss ¼ RMV ifP [ 50 %
Method (3) Aggregates a percentage of the RMV based on the probability of damage for
three damage states.

123
Nat Hazards (2014) 71:2043–2061 2057

Fig. 14 Estimated number of damaged buildings for five events and four damage levels for probabilities of
occurrence greater than 50 %: minor (circle), moderate (square), major (triangle), complete (diamond)

Table 5 Probable cost in millions of dollars calculated using methods 1, 2, and 3 for 3, 6, 9, 13, and 25 m
of slip
Event Method 1 ($M) Method 2 ($M) Method 3

Mw Min. Mod. Maj. Com. Min. Mod. Maj. Com. ($M)

3 3.5 1.9 0.5 0.4 2.4 1.8 0.0 0.0 2.0

6 45.8 23.6 7.9 6.1 16.8 10.4 2.4 2.4 25.8
9 276.7 161.9 57.1 39.7 262.1 97.4 15.2 10.9 165.2
13 697.6 565.5 333.0 249.6 742.4 600.3 308.4 166.9 532.1
25 1,176.8 1,129.9 1,059.2 1,033.9 1,188.6 1,121.7 1,046.2 1,024.8 1,122.0

X RMV  
ELoss ¼ PMinor þ PModerate þ PMajor
3
Table 5 lists the probable cost for 3-, 6-, 9-, 13-, and 25-m slip events and four damage
levels using the three methods presented above. Within the study area (C grid), the total
RMV of all assets totals $1.5 billion. For method 3, the estimated loss ranges from $2 to
$1122 million.
Figure 15 shows the probable cost estimated using methods 1, 2, and 3 for minor,
moderate, major, and complete damage for 3-, 6-, 9-, 13-, and 25-m slip events. Similar to
Fig. 14, this figure illustrates the sensitivity of damage to Seaside for events near 9 m of
slip. For 6 m of slip, the probable loss is estimated to a total between $2.4 and $45.8
million, and for 13 m of slip, the probable loss is estimated to a total between $166.9 and
$742.4 million. As a percentage of the total RMV of all assets within the study area, the
range of probable cost for 6- and 13-m slip events are 0.2 to 3.1 % and 11.4 to 50.7 %,
respectively.
For 17 of the 20 cases, method 1 provided a higher estimate of loss in comparison with
method 2, and in those 3 cases, the difference was less than 6 %. The highest discrepancy
in terms of percentage between methods 1 and 2 is for the 3- and 6-m slip events where
relatively little damage is expected. The highest discrepancy for the larger events was
33 % which was for the estimate of complete damage for the 13-m slip event. This is most
likely attributed to a large number of buildings with probabilities of complete damage just

123
2058 Nat Hazards (2014) 71:2043–2061

Fig. 15 Probable cost (in billions of dollars) for five events and four damage levels for probabilities of
occurrence greater than 50 %: minor (circle), moderate (square), major (triangle), complete (diamond)
damage. Method 1 (solid gray line), method 2 (dashed black line), method 3 (dotted black line)

Table 6 Summary comparison of damage estimates with previous studies

Wood Dominey-Howes This study
(2007) et al. (2010)

Hazard 379 Line 500-year event 3, 6, 9, 13, and 25 m of slip

Assessment In/out & tax PTVA model & Fragility curves & tax lot data
lot data tax lot data
Assets inundated ($millions) 843 500 119, 507, 671, 768, and 1,324
Estimate loss ($millions) – 116 2, 26, 165, 532, and 1,122

below 50 %. Method 3 assigns an equal weighting to the probabilities of minor, moderate,

and major damage and provides an average loss estimate in comparison with the other
curves.

3.3 Comparisons to previous damage estimates

Table 6 lists a summary comparing this study with the methods and results of Wood
(2007) and Dominey-Howes et al. (2010). The ‘‘379 Line’’ which refers to the line of
maximum inundation used by the state of Oregon for planning purposes was used as the
extent of inundation in Wood (2007). The ‘‘379 Line’’ extent of inundation is most
similar to the 13-m slip event modeled by MOST for this paper. The probabilistic
500-year event used by Dominey-Howes et al. (2010) based on the NOAA study is most
similar to the 6-m slip event in comparison with the inundation extent and the 13-m slip
event is comparison to flow depths modeled by MOST in this paper. As expected,
comparing the total value of assets within the inundation zones provide similar results as
similar tax lot data were used for all three studies. The estimated loss however, is
markedly different. Dominey-Howes et al. (2010) aggregated the full RMV of all
buildings with ‘‘vulnerability class’’ medium–high and high and did not include any
values for buildings of lower ‘‘vulnerability class.’’ In the PTVA model used by Dom-
iney-Howes et al. (2010), the flow depth had the highest weighting on the vulnerability
score, so comparison with the 13-m slip event is more appropriate than comparison with

123
Nat Hazards (2014) 71:2043–2061 2059

the 6-m slip event. The difference in the value of assets within the extent of inundation
between the probabilistic 500-year event and the 13-m slip event is $268 million. The
difference in the estimated loss between Dominey-Howes et al. (2010) and this study
using fragility curves is $416 million. Accounting for the additional $268 million in
assets, the estimated loss from this study increases that of Dominey-Howes et al. (2010)
by $148 million. Although the estimated loss by Dominey-Howes et al. (2010) is rea-
sonable and falls within the bounds of the 6–13 m events, the PTVA model lacks the
physics based approached of fragility curves. Similar to Dominey-Howes et al. (2010)
improving on Wood (2007) work by incorporating the spatial variability of the hazard,
this work improves on Dominey-Howes et al. (2010) work by incorporating fragility
curves. Our estimates could be improved by refining the building classification using
recently published fragility curves by Suppasri et al. (2013). These new curves are
developed for wood, steel, reinforce concrete, masonry, and mixed building materials for
1-, 2-, and 3-storey buildings. The damage estimates could also account for key infra-
structures such as roads, bridges (Shoji and Moriyama 2007), and utilities (water/
wastewater, power, gas, and telecommunication networks), which are critical to com-
munity recovery. Other aspects which are important to include but hard to measure are
damage to agricultural lands and ecosystems (Yanagisawa et al. 2009). Estimates could
also be improved by incorporating the effects of macro-roughness or by relating prob-
ability of damage to momentum flux.

4 Summary and conclusions

This paper provides a methodology in the application of fragility curves for estimating
building damage on a community scale. Using Seaside, Oregon, as a case study, a com-
munity-scale estimate of damage and economic loss was calculated at the parcel level
using fragility curves from the literature for events ranging from 3 to 25 m of slip. The
numerical model MOST was used to obtain hydrodynamic values, and tax lot data was
used to determine building type and value. Fragility curves were applied to Seaside,
Oregon, for the first time to estimate probable damage states for a range of CSZ events.
Damage to buildings in Seaside was found to be sensitive to the magnitude of the events,
with minor damage expected for events less than 9 m of slip and severe damage expected
for events greater than 9 m of slip. This sensitivity is due to the community being situated
on low-lying coastal lands. For the 3-, 6-, 9-, 13-, and 25-m slip events, the percentage of
buildings within the inundation zone are 9, 45, 59, 66, and 88 %, respectively, with an
average economic loss ranging from $2 million to $1.2 billion.
The estimates of water level, velocity, and momentum flux show considerable spatial
variation (Fig. 8), and it is likely that a fragility analysis using velocity or momentum flux
would lead to different damage estimates locally. Future work should consider the use of
fragility curves based on velocity and momentum flux, as well as inundation model sen-
sitivity to provide a better range of possible damage levels. Moreover, the fragilities
analysis could be extended to include seismic damage due to the initial earthquake and
damage to other infrastructure (e.g., bridges) and assets (e.g., vessels) as new fragility
curves are developed. For economic loss, future work should include economic input/
output models to calculate the indirect tangible losses to the regional economy. The
probabilities of each modeled event could also be related to the damage estimates, similar
to the probabilistic hazard from the Seaside pilot study (Tsunami Pilot Study Working
Group 2006) to provide a probabilistic damage estimate.

123
2060 Nat Hazards (2014) 71:2043–2061

Acknowledgments This project was supported by Oregon Sea Grant (R/CNH-22). The authors would like
to thank the two anonymous reviewers for their comments and suggestions to improve the manuscript.

References

Cox D, Tomita T, Lynett P, Holman RA (2008) Kinematics of breaking tsunami wavefronts: a data set from
large scale laboratory experiments. In: Proceeding 31st international conference on coastal engineer-
ing. ASCE, pp 1421–1432
Cercone C, Naito C, Riggs R, Cox D (2013) Tsunami generated debris and structural performance guide-
lines. J Waterway Port Coastal Ocean Eng
CREW (Cascadia Region Earthquake Workgroup) (2013) Cascadia subduction zone earthquakes: a mag-
nitude 9.0 earthquake scenario. Oregon department of Geology and Mineral Industries Open File
Report O-05-05
Delouis B, Nocquet JM, Vallée M (2010) Slip distribution of the February 27, 2010 Mw = 8.8 Maule
Earthquake, central Chile, from static and high-rate GPS, InSAR, and broadband teleseismic data.
Geophys Res Lett
Dominey-Howes D, Dunbar P, Varner J, Papathoma-Kohle M (2010) Estimating probable maximum loss
from a Cascadia tsunami. Nat Hazards 53:43–61
FEMA P646 (2008) Guidelines for design of structures for vertical evacuation from tsunamis. Federal
Emergency Management Agency, Washington
Gica E, Spillane MC, Titov VV, Chamberlin CD, Newman JC (2008) Development of the forecast prop-
agation database for NOAA’s short-term inundation forecast for tsunamis (SIFT): NOAA Technical
Memorandum OAR PMEL-139. Pacific Marine Environmental Laboratory, Seattle
Goldfinger C, Nelson CH, Morey AE, Johnson JE, Patton JR, Karabanov E et al (2012) Turbidite event
history—methods and implications for holocene paleoseismicity of the Cascadia subduction zone. U.S.
Geological Survey Professional Paper 1661–F, 170 p
Gonzalez FI, Geist EL, Jaffe B, Kanoglu U, Mofjeld H, Synolakis CE et al (2009) Probabilistic tsunami
hazard assessment at Seaside, Oregon, for near- and far-field seismic sources. J Geophys Res 114:1–19
Hirata K, Satake K, Tanioka Y, Kuragano T, Hasegawa Y, Hayashi Y et al. (2006) The 2004 Indian Ocean
tsunami: tsunami source model from satellite altimetry. Earth Planets Space 195–2001
Ichinose G, Somerville P, Thio HK, Graves R, O’Connell D (2007) Rupture process of the 1964 Prince
William Sound, Alaska, earthquake from the combined inversion of seismic, tsunami, and geodetic
data. J Geophys Res
Kaiser G, Scheele L, Kortenhaus A, Lovholt F, Romer H, Leschka S (2011) The influence of land cover
roughness on the results of high resolution tsunami inundation modeling. Nat Hazards Earth Syst Sci
11:2521–2540
Koshimura S, Oie T, Yanagisawa H, Imanura F (2009) Developing fragility functions for tsunami damage
estimation using numerical model and post-tsunami data from Banda Aceh, Indonesia. Coast Eng J
51:243–273
Mori N, Takahashi T, Yasuda T, Yanagisawa H (2011) Survey of 2011 Tohoku earthquake tsunami
inundation and run-up. Geophys Res Lett 38
Muhari A, Imanura F, Koshimura S, Post J (2011) Examination of three practical run-up models for
assessing tsunami impact on highly populated areas. Nat Hazards Earth Syst Sci 11:3107–3123
Murao O, Nakazato H (2010) Vulnerability functions for buildings based on damage survey data in Sri
Lanka after the 2004 Indian Ocean tsunami. International conference on sustainable built environment,
pp 371–378
National Tsunami Hazard Mitigation Program (2012) Proceedings and Results of the 2011 NTHMP Model
Benchmarking Workshop. Boulder, CO: Department of Commerce/NOAA/NTHMP; (NOAA Special
Report). 436 p
Okada Y (1985) Surface deformation due to shear and tensile faults in a half-space. Bull Seismolog Soc Am
75(4):1135–1154
Park S, van de Lindt JW, Cox DT, Gupta R, Aguiniga A (2012) Successive earthquake-tsunami analysis to
develop collapse fragilities. J Earthq Eng 16:851–863
Park H, Lynett PJ, Cox DT, Wiebe DM, Shin S (2013) Tsunami inundation modeling in constructed
environments: a physical and numerical comparison of free-surface elevation, velocity, and momentum
flux. Coast Eng
PMEL (2006) Method of splitting tsunami (MOST) software manual. NOAA
PMEL. Community Model Interface for Tsunami (ComMIT). NOAA

123
Nat Hazards (2014) 71:2043–2061 2061

Rueben M, Holman R, Cox D, Shin S, Killian J, Stanley J (2011) Optical measurements of tsunami
inundation through an urban waterfront modeled in a large-scale laboratory basin. Coast Eng
58:229–238
Satake K, Wang K, Atwater BF (2003) Fault slip and seismic moment of the 1700 Cascadia earthquake
inferred from Japanese tsunami descriptions. J Geophys Res
Schultz MT, Gouldby BP, Simm JD, Wibowo JL (2010) Beyond the Factor of Safety: Developing Fragility
Curves to Characterize System Reliability SR-10-1. US army corps of engineers: engineering research
and development center
Shoji G, Moriyama T (2007) Evaluation of the structural fragility of a bridge structure subjected to a
tsunami wave load. J Nat Disaster Sci 29:73–81
Suppasri A, Imamura F, Koshimura S (2010) Effect of the rupture velocity of fault motion, ocean current
and initial sea level on the transoceanic propagation of tsunami. Coast Engineering Journal
52(2):107–132
Suppasri A, Koshimura S, Imamura F (2011) Developing tsunami fragility curves based on the satellite
remote sensing and the numerical modeling of the 2004 Indian Ocean tsunami in Thailand. Nat
Hazards Earth Syst Sci 11:173–189
Suppasri, A., Mas, E., Koshimura, S., Imai, K., Harada, K., & Imamura, F. (2012). Developing tsunami
fragility curves from the surveyed data of the 2011 Great East Japan tsunami in Sendai and Ishinomaki
plains. Coast Eng J
Suppasri A, Mas E, Charvet I, Gunasekera R, Imai K, Fukutani Y et al. (2013) Building damage charac-
teristics based on surveyed data and fragility curves of the 2011 Great East Japan tsunami. Nat Hazards
319–341
Tang L, Titov VV, Chamberlin CD (2009) Development, testing, and applications of site-specific tsunami
inundation models for real-time forecasting. J Geophys Res 114:C12
Titov VV, Synolakis CE (1995) Modeling of breaking and nonbreaking long wave evolution. J Waterway
Port Coast Ocean Eng 121(6):308–316
Titov VV, Synolakis CE (1998) Numerical modeling of tidal wave runup. J Waterway Port Coast Ocean Eng
124(4):157–171
Titov VV, Moore CW, Greenslade DJ, Pattiaratchi C, Badal R, Synolakis CE et al (2011) New tool for
inundation modeling: community modeling interface for tsunamis (ComMIT). Pure Appl Geophys
168:2121–2131
Tsunami Pilot Study Working Group (2006) Seaside, Oregon, Tsunami Pilot Study—Modernization of
FEMA Flood Hazard Maps. Joint NOAA/USGS/FEMA Special Report 94
Wood N (2007) Variations in city exposure and sensitivity to tsunami hazards in Oregon. U.S. Geological
Survey Scientific Investigations Report 2007-5283
Yanagisawa H, Koshimura S, Goto K, Miyagi T, Imamura F, Ruangrassamee A et al (2009) The reduction
effects of mangrove forest on a tsunami based on field surveys at Pakarang Cape, Thailand and
numerical analysis. Estuar Coast Shelf Sci 81:27–37
Yeh H, Robertson I, Preuss J (2005) Development of design guidelines for structures that serve as tsunami
vertical evacuation sites. Washington Division of Geology and Earth Resources, Olympia: Open File
Report 2005-4.
Yoshida Y, Ueno H, Muto D, Aoki S (2011) Source process of the 2011 off the Pacific coast of Tohoku
Earthquake with the combination of teleseismic and strong motion data. Earth Planets Space
63:565–569

123