Storm surge and evacuation in urban areas during the peak of a storm

Full text

Final manuscript of H. Takagi
et al
. Coastal Engineering 108 (2016) pp. 1–9
Storm surge and evacuation in urban areas during the peak of a storm
Hiroshi Takagi a,⁎, Siyang Li a, Mario de Leon b, Miguel Esteban c, Takahito Mikami d, Ryo Matsumaru e,
Tomoya Shibayama d, Ryota Nakamura d
a
Tokyo Institut e of Technology, Graduate S chool of Science and Engineeri ng, 2-12-1 Ookayama, Meguro-ku, To kyo 152-8550, Japan
b
De La Salle University, C ivil Engineering Departm ent, 2401 Taft Avenue, Manila 1004, Philippin es
c
The University of Tokyo, Graduate School of Frontier Sciences, 5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8563, Japan
d
Waseda University, Department of Civil and Environmental Engine ering, 3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan
e
Toyo University , Faculty of Regional Development Studies, 5-28-20 Hakusan, Bunky o-ku, Tokyo 112-8606, Japan
a r t i c l e i n f o
Article history:
Received 4 April 2015
Received in revised form 22 October 2015
Accepted 2 November 2015
Available online xxxx
Keywords:
Typhoon Haiyan (Yolanda)
Tacloban
Storm surge
Pedestrian evacuation
Numerical si mulation
Depth
–
velocity product
a b s t r a c t
The present paper examines the impact of floodwater caused by the storm surge brought about by Typhoon
Haiyan in 2013, focusing on downtown Tacloban in Leyte Island, the Philippines. A reliable numerical model
for predicting such flooding was developed by calibrating the results of field investigations, including footage
from a video clip taken during the storm surge. The simulation reveals that flow velocities along the streets in
downtown Tacloban reached up to 7 m/s due to flow contraction along the high-density blocks of houses, and
how water levels reached their peak in just 10 min. According to the depth–velocity product criteria, often
used for evaluating the vulnerability of people and buildings to floodwaters, only 8% of the length of streets in
downtown Tacloban were within t he safe limits that allow pedestrian evacuation. Based on these findings, the
present research concludes that pedestrian evacuation in the middle of a storm surge generated by a strong ty-
phoon is a high-risk behavior. Thus, clearly and objectively, evacuation during this time should not be encouraged,
even when seawater intrudes the houses of local residents. In this respect, it would appear imperative that prior
to the arrival of the typhoon all residents should evacuate areas at risk of being flooded. Though the flood height
was significant in the downtown area, the damage to these houses was limited. If it was not possible for some
reason to evacuate prior to the arrival of the typhoon, those in solid houses should first consider vertical evacu-
ation and the possibility that they could survive in their place, rather than courageously evacuating in an unpre-
dictable water flow.
1.
Introduction
Typhoon Haiyan (Yolanda, according to its local name) struck the
Philippines on November 8, 2013, causing enormous damage to Leyte,
Samar and many other islands. 6245 individuals were reported dead,
28,626 were injured and 1039 are still missing (NDRRMC, as of
6 March 2014). Such a large death toll was caused not only by the
large size of the storm surge, but also due to issues related to the level
of knowledge and awareness by local residents on what is a “storm
surge” (Esteban et al., 2014, 2015). Indeed, despite the fact that two his-
torical storm surges had previously devastated Tacloban, one in 1897
killing up to 1500 and another in 1912 killing 15,000 (Lagmay et al.,
*
Corresponding author at: 2-12-1-S6-212 Ookayama, Meguro-ku, Tokyo 152-8550,
Japan.
E-mail addresses:
takagi@ide.titech.ac.jp (H. Takagi)
2015), local awareness about such events was non-existent (Esteban
et al., 2014). The total economic loss associated with infrastructure
and agriculture was estimated to be around 34,366 million pesos
(776 million USD), possibly the most expensive natural disaster in the
history of the country (TIME, 2013).
Haiyan was one of the strongest typhoons known to have ever made
landfall, not only in the Philippines but the entire world (Lin et al., 2014;
Schiermeier, 2013). The forward speed of the typhoon, reaching around
41 km/h at landfall, was also unusual among other events of comparable
intensity during the past 6 decades in the Western North Pacific (Takagi
et al., 2015a). Haiyan can be characterized as both the fastest moving
and strongest typhoon measured in the Philippines. The return period
for a Haiyan-class typhoon to make landfall was estimated to be
200 years (Takagi and Esteban, 2015). As a result of its strong intensity,
the typhoon caused a massive storm surge in many islands in the
Philippines. The storm surge inundated most of the coastline of Leyte
Gulf, causing particularly large damage to Tacloban City, the biggest
city in Leyte Island. A maximum inundation height of up to 6–7m was
observed in this city, where the largest number of casualties took
place (Mas et al., 2014; Shibayama et al., 2014; Takagi et al., 2015a;

2
H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
Nakamura et al., 2015). It is important to note the local amplification of
the water surface elevation due to seiche effects inside Leyte Gulf (Mori
et al., 2014).
High inundation heights were observed even outside Leyte Gulf
along the east coast of Eastern Samar, which faces the Pacific Ocean
and the deep Philippine Trench (Tajima et al., 2014). Maximum hindcast
significant wave heights during the storm reached up to 19 m off
Eastern Samar, though they decreased to less than 5 m in Leyte Gulf
near Tacloban (Bricker et al., 2014).
Understanding the nature and degree of exposure of coastal areas to
flooding is important for reducing its effect on people and property. Such
assessments may be developed through filed observations, modeling
studies, or some combination of the two (Brown et al., 2007). Flood mit-
igation agencies need to designate floodplains and high hazard zones to
identify the risks of flood damage or loss of life. For adequate mapping of
flood hazards it is important to examine where dangerous zones are lo-
cated within a particular area (Jonkman and Penning-Rowsell, 2008). For
the case of Tacloban city, an early warning system was in place prior to
the arrival of Haiyan, which made use of hazard maps and flood drills.
These hazard maps were created as part of the READY project, a joint ef-
fort in risk mapping among government and donor agencies (Lagmay
et al., 2015). However, these maps greatly underestimated the extent
of the inundation (Lagmay et al., 2015).
An early evacuation, based on a reliable and accurate warning sys-
tem, is crucial to survive typhoons and their subsequent storm surges.
Local authorities might either advice people to leave or sometimes
even conduct forced evacuations, and it appears that a combination of
the two was used by the various barangays (smallest administrative
divisions in the Philippines) in Leyte and Samar (Esteban et al., 2015).
However, despite such evacuation orders and typhoon and storm surge
warnings by local authorities and media before the arrival of Typhoon
Haiyan, many people did not evacuate. A number of reasons were
given by local residents to explain why they stayed in their residences,
such as uncertainties in the expected typhoon level, attempting to secure
their houses and properties, or problems with poorly maintained evacu-
ation centers (Esteban et al., 2015; Leelawat et al., 2014).
If residents fail to evacuate prior to the arrival of the typhoon, they
may find themselves in a very difficult situation. At that stage it would
be hard for them to rationally determine whether they should evacuate
or stay home, particularly when extremely strong gusts and heavy rain
affects the area during the pass of the typhoon. The primary objective of
this research is thus to rationally review if it was actually advisable and
possible to evacuate during the passage of super typhoon Haiyan. To do
so, the authors first conducted a series of field surveys to reveal the
extent of inundation in downtown Tacloban. The observed inundations
were compared with the water levels computed by a storm surge simu-
lation to validate the model's accuracy. Once the reliability of the nu-
merical model was established, it was possible to examine whether
pedestrian evacuation during the passage of strong typhoon would be
safe, focusing on Tacloban and typhoon Haiyan as a case study.
2.
Field survey
2.1.
Methodology
The authors conducted field surveys at three different times after the
disaster: (1) December 4–13, 2013, (2) May 1–6, 2014, and (3) October
16–21, 2014. The primary purpose of these surveys was to identify the
extent of the inundation due to the storm surge along the coastline of
Leyte Island. The last two surveys involved a topographical survey of
Tacloban, conducted at the same time as the flood investigations (see
Fig. 1). The storm surge inundation height was established by checking
physical evidence and interviewing local residents at each site. Laser
range finders and GPS receivers were used to establish the inundation
height at each location (with reference to the local sea level) to within
a few centimeters accuracy, which was adjusted to take into account
the tidal levels at the time of the survey and during the passage of the
typhoon. Laser range finders were also used for surveying the ground el-
evation of different parts of Tacloban's downtown, essentially densely
populated residential areas built on top of both flat and hilly terrains.
Another important factor investigated in the present study was the
flow velocity at a particular location in the downtown area of Tacloban.
(c) Inundation height (d) Topographical survey
Fig. 1.
Site investigation on May 2014 at Tacloban in Leyte Island, the Philippines.
(a) Inundation height
(b) Inundation height

H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
3
Interestingly, a video clip taken from a hotel in this area (Hotel Alejandro)
during the peak of the storm surge shows that the flow was relatively
slow, and nearby local residents could walk across the flooded street to
evacuate from their house to the hotel shown in Fig. 2(a) (iCyclone,
2013). The flow velocity and direction was roughly estimated by visual
investigation through the elapsed time shown imbedded in the video
clip. The distance moved by pieces of floating rubbish captured in the
video was estimated by reference to a photo showing a measuring
staff, taken by the authors during their surveys (Fig. 2(c)). Finally, the
velocity could be calculated by dividing the time it took for the debris
to cross the points shown in the measuring staff. Obviously, a certain
degree of estimation error is present in the visual measurement with
a floating object. However, such errors can be minimized by considering
and calculating the speed of the multiple objects that simultaneously
flow within the video.
2.2.
Results
Fig. 3 shows the topography around downtown Tacloban, which was
reproduced by the authors using the measured elevations, together with
a satellite image showing the city before the typhoon. In this area many
houses were built on relatively low-lying land, only 1–3 m above sea
level. The storm surge heights, which were measured from the astronom-
ical tide at the time of typhoon landfall, are plotted in Fig. 4, showing that
inundation took place even in the middle part of the downtown area.
Though the flood height was significant (in the rage of 0.6 to 4.2 m),
the damage to these houses was not as severe as in other areas of the
city that had informal settlements, as shown in Figs. 1 and 2. Essentially,
houses in the downtown area where typically rather strong cement or
concrete two-story or higher buildings, and thus floodwaters intruded
through the streets (rather than passing over the roofs of houses).
The flow velocity at the street in front of Alejandro Hotel was
estimated to be about 0.6 m/s. This agrees well with field survey obser-
vations, as water flow at such low speed will not cause catastrophic
damage to houses, though minor damage could be observed in all
areas that were flooded. It also explains why it was possible for local
residents to walk through the water flow shown in Fig. 2(a).
3.
Storm surge simulation
3.1.
Methodology
The authors performed numerical simulations to estimate the spatial
distributions of the storm surges using three different scales (Fig. 5).
First, the simulation was carried out for a wide area which encompasses
most of the Philippines, in order to assess which areas of the country
were affected. Then, a more detailed simulation was carried out for
San Pedro Bay in Leyte Gulf, an area where the storm surge engulfed
and claimed thousands of lives. The reliability of the numerical simula-
tion for these two domains was already confirmed by Takagi et al.
(2015a), showing a storm surge larger than 3 m along the entire bay,
and up to 6 m high in its inner part. Such simulation results agree well
with the observations of field surveys (Lagmay et al., 2015; Takagi
et al., 2015a).
In the present research the authors carried out detailed inundation
simulations that focused on Tacloban's downtown (1.7 km × 1.7 km
square grids). The simulation used a 10 m fine computational grid and
incorporates details of the city down to the block level to examine
how the urban geometry can mitigate or intensify the impact of flooding.
In previous studies, many types of numerical codes have been ap-
plied to reproduce coastal floods due to recent major storm surges
or tsunamis (e.g. the COMCOT model for the 2004 Indian Ocean tsunami
Fig. 2.
Field investigations nearby Alejandro Hotel: (a) iCyclone's video clip taken from the hotel during the surge peak, showing family members exiting their house and evacuating to the
hotel, (b) interview with the family fea tured in iCyclone's video to understand more de tails about the situation during the storm surge, and (c) a photo with a mea suring staff taken from
the same position where the video clip was
fi
lmed, and which was used for measuring the
fl
ow velocity during the
fl
ood.

4
H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
Fig. 3.
(a) Satellite image of Tacloban taken before the typhoon (©Google Earth ), (a
′
) de nsely populated residential area around Ho tel Alejandro, and (b) topography of Tacloban which was
used in the simulation.
(Wijetunge et al., 2008), the ADCIRC model for Hurricanes Katrina and
Rita (Dietrich et al., 2011), or the FVCOM model for the 2008 Cyclone
Nargis (Tasnim et al., 2015) and for the 2011 Tohoku Earthquake Tsuna-
mi (Sasaki et al., 2012)).
In this study, the authors applied the parametric typhoon model
developed by Takagi et al. (2012, 2014, 2015c), coupled with the
fluid dynamics model Delft3D-FLOW to estimate the extent of the
storm surge. The typhoon model calculates both the pressure and
Fig. 4.
Storm surge heights (black) adjusted to the tidal level at the time of passage of the typhoon, and inundation depths above ground level (blue) measured around Tacloban downtown
area (unit: meters). (Fo r interpretation of the referen ces to color in this
fi
gure legend, the reader is referred to the web version of this article.)

H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
5
Fig. 5.
Three-scale computational domains: (A) country scale, (B) bay scale (both after Takagi et al., 2015a), and (C) city scale, focusing on Tacloban (present research). Color contours
present the maximum st orm surge heights during the passage of the typhoon. ( For interpretation of the reference s to color in th is
fi
gure legend, th e reader is referred to the web version
of this article.)
wind fields using parameters obtained from the typhoon track dataset,
namely, the central positions and pressures at every recording period.
Delft3D-FLOW was then used to simulate how a storm surge travels
from the deep sea to shallow water, and eventually flows over down-
town Tacloban. Though this model can be applied to a 3D domain, the
present study used a 2D horizontal grid, and thus the code is equivalent
to a non-linear long wave model, commonly used for storm surge
simulation.
The standard drying and flooding algorithm in Delft3D-FLOW is effi-
cient and accurate for coastal regions, tidal inlets, estuaries, and rivers.
In combination with the flooding scheme for advection in the momen-
tum equation, the algorithm known as “drying and flooding” in this nu-
merical model is effective and accurate for rapidly varying flows with
large water level gradients due to the presence of hydraulic jumps
or the occurrence of bores as a result of dam breaks (Stelling and
Duinmeijer, 2003; Stelling et al., 1986).
The numerical settings used in the model are summarized in
Table 1.
Data regarding the typhoon and bathymetry in the target area were
obtained from the Japan Meteorological Agency and The General
Bathymetric Chart of the Oceans (GEBCO) datasets, respectively. The
simulation was run for 18 h, which encompasses the final approach of
the typhoon towards Tacloban, and also its passage through the rest
of Leyte, Cebu and Panay islands. The bathymetry for the detailed
simulation was produced by digitizing a nautical chart of NAMRIA
(National Mapping and Resource Information Authority). The
astronomical tide at Tacloban at the time of landfall was calculated
by using the software WXTide32, showing that the tidal level was
very similar to MLLW (Mean Lower Low Water, which corresponds
to the chart datum), with the difference to this level being only a
few centimeters. Therefore, the bathymetry of the chart was used
directly in the simulation (i.e. without it having to be modified for
tide levels).
Manning's n coefficient is one of the most important parameters for
the accurate simulation of inundation over land. For the case of tsunami
inundation simulations MLIT (2012) recommends using a Manning's
n value in the range of 0.04 to 0.08 for the case of residential areas.
Arcement and Schneider (1984) propose that the estimation of n values
for a flood plain should depend on soil surface, surface irregularities, size
of the cross section, obstructions, vegetation and sinuosity of the flood
plain. Taking the maximum numbers from the range of possible values
of each of these factors, the maximum n value calculated by their
method is 0.25. It is thus clear that Manning's n value can take a wide va-
riety of values depending on the conditions considered. Furthermore, it
appears that the n value for coastal floods in an urban area has not been
researched in detail because such extreme flood events rarely occur. In
light of the constraints regarding the information available, the authors
chose to conduct a series of simulations with different settings of
Manning's n, in order to find an appropriate value that best describes
the flow velocity in front of Alejandro Hotel.
Table 1
Numerical mode l and settings.
Typhoon path JMA typhoon b est track
http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-
pub-eg/trackarchives.html
Typhoon model Pressure: e mpirical estimation by Myers formula, wind:
gradient winds considering super-gradient wind effect
Fluid dynamics model Delf t3D
fl
ow ver. 3.42
Domain Cartesian (UTM51 N), grid @3000 m (Philippines),
@100 m (San Pedro Bay), and @ 10 m (Tacloban)
Bathymetry GEBCO_08 grid (Philippines)
NAMRIA chart (San Pedro Bay and Tacloban)
Terrain data ASTER GDEM 30 m resolution (San Pedro Bay) Measured
by the authors (Tacloban)

6
H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
3.2. Results
Fig. 6.
Flow velocity varies with Manning's
n
value.
Inundation simulations were conducted for a variety of values
of Manning's n coefficient in the range between 0.02 and 0.32. An
n
=
0.04 was found to provide the results that best fit the flow velocities
in front of Alejandro Hotel (Fig. 6). This value falls within the 0.04 to
0.08 range recommended by MLIT (2012). However, Takagi and
Bricker (2014) and Bricker et al. (2015) point out that the recommend-
ed values by MLIT would underestimate the mitigating effect of build-
ings when not taking into account urban geometries (such as streets
and residential blocks). The present study confirms that the simulation,
based on the recommended values by MLIT (2012), could yield a good
estimation when urban geometries are appropriately incorporated into
the model.
Fig. 7 presents a time series of the simulated flow velocity with
n
=
0.04, comparing it to the estimated velocity calculated through
the observation of the iCyclone video. Though the simulated peak ve-
locities agree well with the observed velocity, there is a slight differ-
ence in the arrival time of the floodwater. According to the video the
strongest flow occurred before 8:00 AM, as indicated in Fig. 2, whereas
the flood in the simulation starts after 8:00 AM. This 30-minute differ-
ence could have been caused by a discrepancy between the typhoon's
location recorded by the JMA Typhoon Best Track Data and the actual
typhoon course. Basically, the JMA Typhoon Best Track Data is given
at 3 hour intervals and therefore the position of the simulated typhoon
between these data points might be slightly inaccurate. Nevertheless,
the flow would otherwise appear to be well simulated, and for
Fig. 7.
Comparison between simulated and observed velocities, together with simulated
fl
ow direction (direction can be ident i
fi
ed by the azimuth in Fig. 4).
Fig. 8.
Comparison of storm surge between measured and simulated heigh ts at 15 loca-
tions in Tacloban downtown.
example, during the period of peak flow water moves along the street
in the southwest direction, which corresponds exactly with the actual
direction shown in the video.
Fig. 8 shows how the simulated and observed storm surge heights at
the various locations shown in Fig. 4 are in good agreement at many
points, such as Alejandro Hotel. In fact, the coefficient of determination
R2, which is a number that indicates how well data fits a statistical
model, was calculated as 0.574. This implies that the calibration of
Manning's n value, by using the observed flow velocity at even just
one point, can be expected to improve the accuracy of the simulation
of storm surge heights.
However, the authors do acknowledge that there are still many un-
certainties that present in the numerical results. The best estimated
Manning's n could also vary according to the local topography, which
can create various levels of resistance to the flow due to differences in
bottom friction, friction on the walls and parked cars. Despite this, the
n value in the present study was set to be uniform throughout the entire
land domain. Mignot et al. (2006) attempted to improve the reliability
of an urban flood modeling with two different n values of 0.025 and
0.050, according to different road widths. Changes in Manning's n in
this manner could be one possible option to further improve the estima-
tion of velocities and inundation depths. It should also be noted that the
flow condition at some particular locations in the downtown area can
be considered to be supercritical. The energy loss during such a regime
cannot be correctly evaluated because the formulation in terms of fric-
tion used in the present model merely assumes subcritical flow (though
Manning's n can be still applied to a rapid flow, such as that induced by a
tsunami's overflow, by using an appropriate value, see Bricker et al.,
2015). Therefore, it is important to recognize that the numerical settings
used in this study do not necessarily provide the best options for other
cases or areas.
4.
Discussion
This section will discuss the characteristics of the floodwater in gen-
eral and the flow velocity along the streets in particular, which has a
great influence on the chance that evacuees caught in the flood can sur-
vive. According to an experimental study using a large sized water
flume (Suga et al., 1995), a 60 year old male did not experience any
fear when he walked against a water flow as high as his crotch with
speeds up to 0.8 m/s. In contrast, he felt intense fear and could not

H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
7
Fig. 9.
Maximum
fl
ow veloc ity simulated along the streets of Tacloban downtown.
walk without supporting ropes when the flow reached a height of 1 m
or a velocity of 1 m/s. Nishihata et al. (2005) estimated that the maxi-
mum inundation depth at which people can safely evacuate is around
0.7 m. Likewise, in areas with high flow velocity people may not
be able to evacuate because of friction instability (sliding), no matter
how shallow the inundation depth. Therefore, the authors feel safe to
presume that the family featured in iCyclone's video could cross the
flooded street because the flow in front of Alejandro Hotel was relative-
ly slow, with a velocity 0.6 m/s or lower, as Fig. 7 indicates. Fig. 9 shows
the maximum flow velocities calculated during the passage of the ty-
phoon. According to this it appears that the flow velocity at the hotel
was relatively calm compared to other parts of the city, which enabled
them to evacuate. In contrast, flow appears to have been particularly
intense along certain streets, possibly due to flow contraction, with
flow velocities of up to 7 m/s in places.
It is worth noting that friction instability appears to occur earlier
than moment instability (toppling) for the combination of shallow
depth and high velocities (Jonkman and Penning-Rowsell, 2008). Also,
individuals who are males, middle-aged, heavy, tall, in good physical
condition, and lightly clad are able to, on average, remain standing in
higher depth–velocity products (Abt et al., 1989). Some authors have
Fig. 10.
Histogram of the maximum depth
–
velocity product in T acloban downtown.
Fig. 11.
Variations in water depth and velocity at St. 1 (the location shown in Figs. 4 and 9).
proposed a critical depth–velocity product to indicate a person's insta-
bility in flood waters, which would be a combination of the depth
and the velocity of the flood (Jonkman and Penning-Rowsell, 2008).
Wright et al. (2010) summarizes that the depth–velocity product of
1.0 m2/s as the safe limit for pedestrians. Fig. 10 presents the histogram
of the maximum depth–velocity product derived by summing up the
values at all the computational grids along the streets surrounded by
building blocks, i.e. excluding the roads that directly face the sea. This
figure shows that only 8% of the total street areas in the downtown sec-
tion of Tacloban could have been classified as being within the safe limit
for pedestrian evacuation during the peak of the storm surge.
The depth–flow velocity product is also considered as a critical pa-
rameter leading to the damage or the collapse of buildings. Pistrika
and Jonkman (2010) proposes residential building damage categories
that can be determined by the depth–velocity product (dv), as follows:
1)
dv b 3 m2/s: “inundation damage”
2)
3 m2/s ≦ dv b 7 m2/s: “partial damage”
3)
dv ≧ 7 m2/s: “total destruction”.
According to these criteria residential houses can withstand much
stronger floodwaters than people. Although the simulated depth–
velocity product in Fig. 10 reached up to 10 m2/s, it can be noted
that water does not flow normal to the houses but typically along
the roads, which would further alleviate the impact of the flood. In
fact, in the course of the surveys conducted by the authors there was lit-
tle evidence that any residential houses nearby the locations surveyed
in Fig. 4 suffered any significant physical damage, as shown also by
Figs. 1 and 2.
The rise time of the storm surge was estimated to be less than half an
hour (Takagi et al., 2015b), which correlates well with interviews with
survivors (Esteban et al., 2015). Fig. 11 also shows how water depth
and velocity would have risen quickly at the street indicated by St. 1
in Fig. 9, with the depth–velocity product exceeding the safety limit of
1.0 m2/s in just 10 min. Therefore, the case of the family featured in
the iCyclone video, who successfully evacuated across their street to
Hotel Alejandro, can be perceived as a case of good luck, brought
about by the fact that the flood around the area was much less violent
than in other areas in downtown Tacloban (Fig. 5).
Although this would appear rather intuitive, in fact comparatively
little research has focused on ascertaining such dangers and explaining
why anecdotal evidence of evacuation by pedestrians during the
Haiyan's storm surge was possible. The present research establishes
that success in such cases was often a matter of luck and timing, as in
most cases flood inundation depths and velocities would have been
too high to allow any evacuation. In this sense, the results also explain
why witnesses often reported how it was impossible to swim in the
flood waters due to the high turbulence, and how many caught in
them died (Esteban et al., 2015).

8
H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
5.
Conclusion
In the present work the authors reported on field investigations car-
ried out in downtown Tacloban to reveal the impact of floodwater
caused by the storm surge due to Typhoon Haiyan in 2013. Inundation
levels were confirmed by visual inspection and interviews with local
residents. The results of a numerical model, calibrated by the flow pat-
terns captured in a video clip by iCyclone, agree well with the observed
inundation heights.
Local residents of coastal areas have to consider evacuating from
their houses when facing a strong typhoon that can cause extensive
flooding. Particularly, it is inevitable that they will be wanting to evacu-
ate once floodwaters directly enter their houses, though at such point it
is often hard for them to rationally determine whether they should
evacuate or stay in order to survive. By using the model it was then
possible to understand the hazard posed by storm surge flooding, and
the risks involved in pedestrian evacuation during the passage of a
strong typhoon. Based on such analysis it can be concluded that pedes-
trian evacuation during a storm surge can be considered as high-risk
behavior, at least for the case of downtown Tacloban during the peak
of storm.
Also, it was clear that many residential houses in downtown
Tacloban withstood the flood without suffering any significant physical
damage, despite high flow velocity along the streets, which reached up
to 7 m/s in places. From these findings, it would appear that pedestrian
evacuation during the peak of typhoons could be more hazardous than
staying in houses, provided that the house was of sufficient height to
allow vertical evacuation to its upper floors.
However, the present research only deals with the case of Tacloban
City and Typhoon Haiyan in 2013, and is clearly not sufficient to gen-
eralize such conclusions to all cases. Therefore, more detailed investi-
gations for other stretches of the coastline and typhoons would be
needed before a definitive conclusion can be reached. As a preliminary
conclusion, and although this would appear somehow obvious in the
context of storm surge risk management, evacuation should be encour-
aged prior to the arrival of a typhoon, particularly for those living in
low-level and low-strength housing. Even if it was not possible for
some reason to evacuate prior to the arrival of the typhoon, evacuation
should not be encouraged when people find that seawater intrudes into
their properties. Though the flood height was significant in the down-
town area, the damage to these houses was limited. Particularly, those
in such solid houses should first consider vertical evacuation and the
possibility that they could survive in their place, rather than coura-
geously evacuating in an unpredictable water flow.
Acknowledgments
The funds for the present research were provided by JSPS KAKENHI
Grant Number 26702009 (Tokyo Institute of Technology), J-RAPID
Program of Japan Science and Technology Agency (JST) (grant to Tokyo
Institute of Technology), and Strategic Research Foundation Grant-
aided Project for Private Universities from Ministry of Education, Culture,
Sports, Science and Technology (MEXT) (Waseda University).
References
Abt, S.R., Wittier, R.J., Taylor, A., Love, D.J., 1989. Human stability in a high flood hazard
zone. J. Am. Water Resour. Assoc. 25, 881–890. http://dx.doi.org/10.1111/j.1752-
1688.1989.tb05404.x.
Arcement Jr., G.J., Schneider, V.R ., 1984. Guide for s electing manning's rou ghness coef
fi
-
cients for natural channels and
fl
ood plains. Report No. FHWA-TS-84-2 04. Federal
Highway Admin istration.
Bricker, J.D., Takagi, H., Mas, E., Kure, S., Adr iano, B., Yi, C., Roeber, V., 2014. Spatial varia-
tion of damage due to storm surge and waves during typhoon Haiyan in the
Philippines. J. Jpn. Soc. Civil Eng. Ser. B2 Coast. Eng. 70 (2), 231–235.
Bricker, J.D., Gibso n, S., Takagi, H., Ima mura, F., 2015. On the need for larger Manning's
roughness coef
fi
cients in depth-integrated tsunami inundation models. Coa st. Eng.
J.
57 (02). http://dx.doi.o rg/10.1142/S0578563415500059 (13 pp.).
Brown, J.D., Spencer, T., Moeller, I., 2007. Modeling storm surge
fl
ooding of an urban
area with particular reference to modeling uncertainties: a ca se study of Canvey
Island, United Kingdom . Water Resour. Res. 43, W06402. http://dx.doi.o rg/10.1029/
2005WR004597.
Dietrich, J.C., Zijlema, M., Westerink, J.J., Holthuijsen, L.H., Dawson, C., Luettich, R.A.,
Jensen, R.E., Smith, J.M., Stelling, G.S., Stone, G.W., 2011. Modeling hurricane waves
and storm surge using integrally-coupled, scalable computations. Coast. Eng. 58 (No.
1), 45–65.
Esteban, M., Matsumaru, R., Takagi, H., Mikami, T., Shibayama, T., Leon, M., V alenzuel a,
V.P., Thao, N.D., 2014. Study on disaster information dissemination and people's
response for evacuation — the case of the 2013 Typhoon Yolanda (Haiyan). J. Jpn. Soc.
Civ. Eng. Ser. B3 Ocean. Eng. 70 (No. 2), 1218–1223. http://dx.doi.org/10.2208/
jscejoe.70.I_1218.
Esteban, M., Valenzuela, V.P., Yun, N.Y., Mikami, T., Shibayama, T., Matsumaru, R., Takagi,
H., Thao, N.D., De Leon, M., Oyama, T., Nakamura, R., 2015. Typhoon Haiyan 2013
evacuation preparations and awareness. J. Sustain. J. 3 (1) (9 pp.).
iCyclone, 2013. Super Typho on HAIYAN (YOLANDA): Eyewall Video. http://www.icyclone.
com/ (Accessed on December 20, 2013).
Jonkman, S.N., Penning-Rowsell, E., 2008. Hu man instability in
fl
ood
fl
ows.
J.
Am. Wa ter
Resour. Assoc. 44 (4), 1
–
11.
Lagmay, A.M.F., Agaton, R.P., Bahala, M.A.C., Briones, J.B.L.T., Cabacaba, K.M.C., Caro, C.V.C.,
Dasallas, L.L., Gonzalo, L.A.L., Ladiero, C.N., Lapidez, J.P., Mungcal, M.T.F., Puno, J.V.R.,
Ramos, M.M.A.C., Santiago, J., Suarez, J.K., Tablazon, J., 2015. Devastatin g storm surges
of Typhoon Haiyan. Int. J. Disaster Risk Reduct. 11, 1–12.
Leelawat, N., Mateo, C.M.R., Gaspay, S.M., Suppasri, A., Imamura, F., 2014. Filipin os' views
on the disaster information for the 2013 Super Typhoon Haiyan in the Philippines.
Int. J. Sustain. Future Hum. Sec. J. Sustain. 2 (2), 16–28.
Lin, I.I., Pun, I.F., Lien, C.C., 2014. “Category-6” supertyphoon Haiyan in global warming
hiatus: contribution from subsurface ocean warming. Geophys. Res. Lett. http://dx.
doi.org/10.1002/2014GL061281.
Mas, E., Bricker, J., Kure, S., Adriano, B., Yi, C., Suppasri, A., Koshimura, S., 2014. Field survey
report and satellite image interpretation of the 2013 Super Typhoon Haiyan in the
Philippines. Nat. Hazards Earth Syst. Sci. Discuss. 2, 3741–3767.
Mignot, E., Paquier, A., Haider, S., 2006. Modeling floods in a dense urban area using 2D
shallow water equations. J. Hydrol. 327, 186–199.
MLIT, 2012. Guide to determining the potential tsunami inundation (Version 2.00). Sea-
coast Office, Water and Disaster Management Bureau, Ministry of Lan d, Infrastructure,
Ministry of Land, Infrastructure, Tran sport and Tourism.
Mori, N., Kato, M., Kim, S., Mase, H., Shibutani, Y., Takemi, T., Tsubok i, K., Yasuda, T., 2014.
Local a mplification of storm surge by Super Typhoon Haiyan in Leyte Gulf. Geophys.
Res. Lett. 41, 5106–5113. http://dx.doi.org/10.1002/2014GL060689.
Nakamura, R., Oyama, T., Shibayama, T., Esteban, M., Takagi, H., 2015. Evaluation of storm
surge caused by Typhoon Yolanda (2013) and using weather-storm surge-wave-tide
model. 8th International Conference on Asian and Pacific Coasts (APAC2015). Ma-
dras, India http://dx.doi.org/10.1016/j.proeng.2015.08.306.
NDRRMC, 2014. Effe cts of Typhoon “YOLANDA” (HAIYAN). SitRep No.106 (6 March 2014).
Nishihata, T., Moriya, Y., Tamura, T., Takimoto, K., Miura, H., 2005. Experimental study on
the condition of evaluation under
fl
ood situation caused by tsunami. Ann.
J.
Coast.
Eng. Coast. Eng. JSCE 52, 1256–1260.
Pistrika, A.K., Jonkman, S.N., 2010. Damage to re sidential buildings due to
fl
ooding of New
Orleans after Hurricane Katrina. Nat. Ha zards 54, 413
–
434. h ttp://dx.doi.org/10.1007/
s11069-009-9476-y.
Sasaki, J., Ito, K., Suzuki, T., Wiyono, R.U.A., Oda, Y., Takayama, Y., Yokota, K., Furuta, A.,
Takagi, H., 2012. Behavior of the 2011 Tohoku earthquake tsunami and resultant dam-
age in Tokyo Bay. Coast. Eng. J. 54 (1). http://dx.doi.org/10.1142/S057856341250012X
(26 pp.).
Schiermeier, Q., 2013. Did climate change cause Typhoon Haiyan. Nature. http://dx.doi.
org/10.1038/nature.2013.14139.
Shibayama, T., Matsumaru, R., Takagi, H., Leon, M., Esteban, M., Mikami, T., Oyama, T.,
Nakamura, R., 2014. Field survey and analysis of storm surge caused by the 2013
Typhoon Yolanda (Haiyan). J. Jpn. Soc. Civ. Eng. B3 Ocean. Eng. 70 (2), 1212–1217.
http://dx.doi.org/10.2208/jscejoe.70.I_1212 (2014).
Stelling, G.S., Duinmeijer, S.P.A., 2003. A staggered conservative scheme for every Froude
number in rapidly varied shallow water flows. Int. J. Numer. Methods Fluids 43,
1329–1354.
Stelling, G.S., Wjersma, A.K., Willemse, J.B.T.M., 1986. Practical aspects of accurate tidal
computations. J. Hydraul. Eng. 112 (9), 802–817.
Suga, K., Uesaka, T., Yoshida, T., Hamaguchi, K., Chen, Z., 1995. Preliminary stu dy on fea si-
ble safe evacuation in flood disaster. Proc. Hydraul. Eng. JSCE 39, 879–882.
Tajima, Y., Yasuda, T., Pacheco, B., Cruz, E., Kawasaki, K., Nobuoka, H., Miyamato, M.,
Asano, Y., Arikawa, T., Origas, N.M., Aquino, R., Mata, W., Valdez, J., B riones, F., 2014.
Initial report of JSCE-PICE joint survey on the storm surge disaster caused by Typhoon
Haiyan. Coast. Eng. J. 56 (1). http://dx.doi.org/10.1142/S0578563414500065.
Takagi, H., Bricker, J.D., 2014. Assessment of the effectiveness of general breakwaters in
reducing tsunami inundation in Ishinomaki. Coast. Eng. J. 56 (04). http://dx.doi.org/
10.1142/S0578563414500181.
Takagi, H., Esteban, M., 2015. Statistics of tropical cyclone landfalls in the Philippines
—
unusual characteristics of 2013 Typhoon Haiyan. Nat. Hazards. http://dx.doi.org/10.
1007/s11069-015-1965-6 (12 pp.).
Takagi, H., Thao, N.D., Esteban, M., Tam, T.T., Knaepen, H.L., Mikami, T., 2012. Assessment
of the coastal disaster risks in Southern Vietnam. J. Jpn. Soc. Civ. Eng. B3 Ocean. Eng.
68 (2), 888–893. http://dx.doi.org/10.2208/jscejoe.68.I_888 .
Takagi, H., Thao, N.D., Esteban, M., 2014. Tropical cyclones and storm surg es in South ern
Vietnam. In: Thao, N.D., Takagi, H., Esteban, M. (Eds.), Coastal Disasters and Climate
Change in Vietnam, pp. 3
–
16. http://dx.doi.org/10.1016/B978-0-12-800007-6.
00001-0.

H. Takagi et al. / Coastal Engineering 108 (2016) 1
–
9
9
Takagi, H., Esteban, M., Shibayama, T., Mikami, T., Matsumaru, R., Nguyen, D.T., Oyama, T.,
Nakamura, R., 2015a. Track analysis, simulation and field survey of the 2013 Typhoon
Haiyan storm surge. J. Flood Risk Manage. http://dx.doi.org/10.1111/jfr3.1 2136.
Takagi, H., Mario, L.D., Esteban, M., Mikami, T., Nakamura, R., 2015b. Storm surge due to
2013 Typhoon Yolanda in Leyte Gulf, the Philippines. In: Esteban, M., Takagi, H.,
Shibayama, T. (Eds.), Handbook of Coastal Disaster Mitigation for Engineers and
Planners, 1st ed. Elsevier. http://dx.doi.org/10.1016/B978-0-12-801060-0.00008-3.
Takagi, H., Thao, N.D., Esteban, M., Mikami, T., Cong, L.V., Ca, V.T., 2015c. Coastal disasters
in Vietnam. In: Esteban, M., Takagi, H., Shibayama, T. (Eds.), Handbook of Coastal
Disaster Mitigation for Engineers and Planners, pp. 235–255. http://dx.doi.org/10.
1016/B978-0-12-801060-0.00012-5.
Tasnim, K.M., Shibayama, T., Esteban, M., Takagi, H., Ohira, K., Nakamura, R., 2015. Field
observation and numeri cal simulation of past and future storm surges in the Bay of
Bengal: case study of cyclone Nargis. Nat. Hazards 75 (2), 1619
–
1647. http://dx.d oi.
org/10.1007/s1 1069-014-1387-x.
TIME, 2013. The typhoon's toll. TIME Mag. 182 (22) (http://content.time.com/time/
magazine/0,9263,7601131125,00.html#ixzz2lY0n8 A4C, Accessed on December 20,
2013).
Wijetunge, J.J., Wang, X., Liu, P.L.-F., 2008. Indian Ocean Tsunami on 26 December
2004: Numerical modeling of inundation in three cities on the south coast of
Sri Lanka. Journal of Earthquake and Tsunami Vol.2 (No.2). http://dx.doi.org/
10.1142/S1793431108000293.
Wright, K., Doody, B.J., Becker, J., McClure, J., 2010. Pedestrian and motorist flood saf ety
study: a review of behaviours in and around floodwater and strategies to enhance
appropriate behaviour. GNS Science Report 2010/51 (91 pp.).