Received December 25, 2013; revised January 25, 2014; accepted January 31, 2014
Copyright 2014 Abu Bakar Sambah, Fusanori Miura. This is an open access article distributed under the Creative Commons At-
tribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is prop-
erly cited. In accordance of the Creative Commons Attribution License all Copyrights 2014 are reserved for SCIRP and the owner
of the intellectual property Abu Bakar Sambah, Fusanori Miura. All Copyright 2014 are guarded by law and by SCIRP as a guar-
dian.
ABSTRACT
Disaster mitigation and reconstruction plan due to tsunami can be implemented with various actions. An inte-
gration of spatial analysis through Geographical Information System (GIS) application and multi-criteria analy-
sis through Analytical Hierarchy Process (AHP) is one of the methods for tsunami inundation and impact as-
sessment. In this study, vulnerability, inundation and impact assessment due to tsunami hazard in Ofunato city,
Iwate Prefecture, Japan was carried out. Appropriate input parameters were derived from Digital Elevation
Model data, and satellite remote sensing and field data were analyzed through GIS. We applied the parameter of
elevation and slope created from Aster GDEM version 2, coastline distance created from vector map of the study
area and vegetation density created from ALOS ANVIR-2 image. We applied AHP process for weighting the
parameter through pair-wise comparison using five iterations of normalized matrix. Five classes of vulnerability
were defined and analyzed for tsunami inundation mapping. We used weighted overlay through spatial analyst
in GIS to create the final map of tsunami vulnerability. The assessment results indicate that 7.39 square kilome-
ter of the study area was under the high vulnerability zone due to tsunami, and 8.13 square kilometer of building
area was under the inundation area. Our result showed good agreement with the observed data and historical
map. The result presented here can aid as preliminary information for the coastal zone management related to
disaster mitigation and for the evacuation process and management strategy during disaster.
KEYWORDS
Spatial Analysis; GIS; AHP; Tsunami
degree to which a person, community or a system is ing additional 250,000 scenes acquired after the previous
likely to experience harm due to an exposure to an ex- release. The data are posted on a 1 arc-second (approxi-
ternal stress. Vulnerability is described as a set of condi- mately 30 m at the equator) grid [12].
tions and processes resulting from physical, social, eco- We analyzed ALOS AVNIR-2 image for vegetation
nomic and environmental factors that increase the sus- density mapping. Moreover, we used run up and inun-
ceptibility of a community to the impact of hazards [5,6]. dation information of the 2011 Tohoku Earthquake
The application of remote sensing and GIS made the which is recorded in the report book of detailed map of
possibilities in order to map the distribution of damage the impacts of the 2011 Japan tsunami in Miyagi Prefec-
area due to disaster and to assess the vulnerability area. ture [13]. The step of analysis starts from data collection,
GIS through the spatial multi-criteria analysis helps in surface analysis of DEM data, vector data processing,
making a priority related to the decision making process vegetation density mapping, AHP process and ends with
using geo-reference data. In addition to the decision ma- raster overlay processing through GIS approach (Figure
kers preferences related to a parameter evaluation, spa- 1).
tial multi-criteria analysis needs both information on cri-
terion attributes and the geographical references [7,8]. 2.2. Study Area
Some previous works have applied GIS method for tsu-
The study area was Ofunato city in Iwate prefecture,
nami mapping and overlaid with official land-use map
Japan (Figure 2). This area is one of the areas affected
[9], and have analyzed tsunami risk using multi-scenario
by The 2011 Tohoku earthquake tsunami, which is the
approach [10], and have analyzed the vulnerability using
fourth huge earthquake that caused tsunami since 1900,
remote sensing data and integrated analysis using GIS to
besides the 1960 Chile Earthquake tsunami, the 1964
the physical built-up infrastructure, i.e. buildings, and
Alaska earthquake tsunami and the 2004 Sumatra earth-
identified the inundation area based on the contour and
quake tsunami [14].
the highest record of tsunami event related to the build-
The significant damage to city in previous event was
ing vulnerability and human vulnerability [1,11].
in 1896, 1933 and 1960. In 2013, the city has an esti-
The aims of this study are to map the vulnerability
mated population of 39,136. The total area is 323.28
area due to tsunami disaster and to assess its impact by
square kilometer. Population residing in inundated area
mapping the possible inundation area due to tsunami by
during the 2011 Tohoku Earthquake tsunami was 19,073.
using GIS. The vulnerability was calculated based on ele-
Ofunato faces the Pacific Ocean. Outside its bay, the
vation, slope, coastline distance and vegetation density.
warm and cold ocean currents meet. In this city, the pre-
In terms of vegetation density, we tried to apply Soil
fectural government provides hazard mapping to the mu-
Adjusted Vegetation Index (SAVI) instead of vegetation
nicipal government who then work with community
mapping using Normalized Difference Vegetation Index
groups to develop the evacuation maps and identify suit-
(NDVI). In addition, we also applied the Analytical Hie-
able places for refuge. In other hand there is limited illu-
rarchy Process for estimating weights of the parameters.
stration of evacuation routes on the maps [15].
2. Materials and Methods 2.3. Spatial Analysis
2.1. Data
2.3.1. Elevation
The elevation data used in this study was obtained from Elevation data is one of the principal datasets required
The ASTER Global Digital Elevation Model (ASTER for the model to generate vulnerability and inundation of
GDEM) version 2. The Advanced Space-borne Thermal the tsunami. In order to derive a set of parameters that
Emission and Reflection Radiometer (ASTER) GDEM is describe the physical vulnerability a digital elevation
a joint product developed and made available to the pub- model was generated by elevation map using ASTER
lic by the Ministry of Economy, Trade, and Industry GDEM version 2 data. This data was downloaded from
(METI) of Japan and the United States National Aero- ASTER GDEM website
nautics and Space Administration (NASA). It is generat- (http://gdem.ersdac.jspacesystems.or.jp/).
ed from data collected from a space-borne earth observ- We classified elevation into five classes of vulnerabil-
ing optical instrument, namely ASTER. ity based on the height of the surface. It describes that
The ASTER GDEM covers the entire land surface of low elevation (in meter) of the surface will have high
the Earth at high resolution. Version 2 of the ASTER vulnerability to tsunami wave, as shown in Table 1. Vul-
GDEM is employing an advanced algorithm to improve nerability mapping based on elevation data describe that
global digital elevation resolution and elevation accuracy most of coastal area is in the slightly high class of vulner-
and reprocessing a total of 1.5 million scene data includ- ability (Figure 3(a)).
Weighted process
Tsunami through AHP
vulnerability
criteria Raster overlay
(weighted overlay)
Raster
Vulnerability map
reclassification
Inundation map
Figure 3. Tsunami vulnerability map based on the parameter of (a) Elevation; (b) Slope; and (c) Coastline distance.
Table 1. Vulnerability class based on elevation [17], slope to reflectance values before generating the synthetic
[18], and coastline distance. NDVI and SAVI image [20,21]. The steps to create ve-
Elevation Slope Coastline distance Vulnerability getation density map as follows.
(meter) (percentage) (meter) class
<5 0-2 0 - 603.50 High 2.4.1. Reflectance Calculation
Radiance convertion from digital number values is need-
5 - 10 2-6 603.50 - 1451.76 Slightly high ed before generating reflectance. Calculation of radiance
10 - 15 6 - 13 1451.7 - 2453.60 Medium is the fundamental step in relating image data from mul-
tiple sensors and platforms into a common radiometric
15 - 20 13 - 20 2453.60 - 3572.06 Slightly low
scale [22]. The general equation for DN to radiance con-
>20 >20 >3572.06 Low version as in Equation (2) [23]
L = Grescale QCAL + Brescale (2)
2.3.2. Slope
L is spectral Radiance at the sensors aperture
Slope is the rate of maximum change in z-value from
(W/m2/sr/m), Grescale is rescaled gain, QCAL is Digital
each cell of the image. We created slope map using the
Number (DN), and Brescale is rescaled bias. Moreover,
third-order finite difference method proposed by [16].
Equation (3) was used to convert radiance to reflectance
Slope percentages range from 0 to near infinity. A flat
[24]:
surface is 0 percent, and as the surface becomes more
vertical, the slope percent becomes increasingly larger. = L d 2 ESUN cos s (3)
Slope map was created from ASTER GDEM version 2.
is unitless planetary reflectance, L is spectral ra-
We classified slope map into five classes according to the
diance at the sensors aperture, d2 is earth-sun distance in
value of tsunami vulnerability as described in Table 1.
astronomical units from nautical handbook, ESUN is
We applied slope classification based on tsunami vulner-
mean solar Exoatmospheric irradiances, and s is solar
ability class through reclassify process to generate the zenith angle in degrees.
vulnerability map (Figure 3(b)). High vulnerability areas
based on slope was identified in the urban area close to 2.4.2. NDVI Calculation
the river in the southeast part of Ofunato. The principle behind NDVI is that channel 1 is in the
red-light region of the electromagnetic spectrum where
2.3.3. Coastline Distance chlorophyll causes considerable absorption of incoming
The calculation of the distance from coastline to the land sunlight, whereas channel 2 is in the near-infrared region
using multi-buffering was done to divide area into five of the spectrum where a plants spongy mesophyll leaf
classes of vulnerability. The buffering distance was set structure creates considerable reflectance. Consequently,
based on the possibility range of the tsunami to reach the dense vegetation shows up very strongly in the imagery
land. The distance is depend on the historical report of where red spectrum reflectance is low and in contrast
the maximum run up in the area of study, and is ex- high near infrared reflectance [25-27].
pressed by Equation (1) [9,19]: NDVI is a measure of the difference in reflectance
4 Y between these wavelength ranges that takes values be-
=
log X max log1400 + log o (1) tween 1 and 1, with values 0.5 indicating dense vegeta-
3 10
tion and values < 0 indicating no vegetation including
Xmax represents the maximum reach of the tsunami water. NDVI was estimated using Equation (4) [28]
over land, and Yo is the tsunami height at the coast. Maxi-
( NIR VIS)
mum run-up of Ofunato area based on The 2011 Earth- NDVI = (4)
quake Tsunami Joint Survey Group was 30.1 meter and ( NIR + VIS)
minimum run up was 5.32. By using the algorithm above, NIR is Near Infra-Red band, and VIS is visible band of
we generated five range of coastline distance that vul- red band. ALOS AVNIR-2 band 3 is red and band 4 is
nerable to the tsunami as shown in Table 1. We applied NIR.
coastline distance in vector map of Ofunato through GIS
processing to generate vulnerability map based on coast- 2.4.3. SAVI Calculation
line distance (Figure 3(c)). SAVI is one of the algorithms developed to generate ve-
getation index by eliminating soil factor [29]. Although
2.4. ALOS AVNIR-2 Image Processing SAVI model is not significant in the intra-class analysis,
We tried to use SAVI for vegetation density mapping. in the similar spectral space, this model presents a rea-
The digital numbers of ALOS AVNIR-2 were converted sonable performance in the characterization of forested
and non-forested areas [21]. Atmospheric influences low- Table 2. Nine-point comparison scale by Saaty [37].
er the NDVI of vegetated areas whereas dark or wet soil Intensity of
Definition
backgrounds increase the NDVI for given vegetation importance
canopy [30]. SAVI was developed to minimize soil bri- 1 Equal importance
ghtness, including shadow, influences found in the NDVI
3 Weak importance of one over another
by accounting for first-order soil-vegetation spectral in-
teractions as in Equation (5) [31]. 5 Essential or strong importance
Figure 4. (a) SAVI map; (b) Vegetation density; and (c) Vulnerability map based on vegetation density.
vulnerability classes as shown in Table 3 was created Table 4. Vulverability class of building area.
using the subtraction of maximum and minimum value
5 4
and divided by number of class. Subsequently tsunami Vulnerability Class 3 2 1
vulnerability in five classes is mapped (Figure 6(b)). Inundation Area
Building area
3.35 4.78 6.25 1.71 6.01
3.3. Tsunami Inundation and Impact Assessment (square kilometer)
We compared our result to the survey data from Joint meter, while reflectance value was 0.13, and SAVI was
Survey Group of the 2011 Tohoku Earthquake, the ex- 0.05. Elevation map created from ASTER GDEM ver-
isting map of tsunami inundation area due to the 2011 sion 2 describes that the highest elevation in the inunda-
Tohoku Earthquake published by Geographical Survey tion area is 20.5 meter.
Institute (GSI) Japan (Figure 7), and detailed map of the
impacts of the 2011 Japan tsunami, vol. 1: Aomori, Iwate 4. Discussion
and Miyagi Prefecture [13].
The comparison describes that most of inundation area Vulnerability is related to the capacity to reduce the im-
was occurred in the area of high and slightly high vulner- pact of disaster and mitigation plan. It describes the po-
ability with an area of 1.10 square kilometer and 6.29 tential area to be damaged by natural disasters. Class of
square kilometer. Based on these areas we estimate that vulnerability could be based on a physical parameter,
total area of the possibility of inundation is 7.39 square such as elevation, slope, coastline distance, land cover
kilometer (Figure 8(a)), while the inundation area due to and coastal shape. Moreover, inundation can be defined
due to 2011 Tohoku earthquake based on GSI report is 8 as the result of a tsunami traveling a long distance inland
square kilometer. Maximum run-up of Ofunato area and is a horizontal measurement of the path of the tsu-
based on The 2011 Earthquake Tsunami Joint Survey nami. The analysis of satellite image, elevation data, and
Group was 30.1 meter and found in the area of Shiraha- survey data followed by multi-criteria analysis through
ma Sanrikucho Ryori. Minimum run up was 5.32 meter AHP and raster overlay in GIS processing can be used as
at the area of Sakaricho. This area is identified as the the basic information for vulnerability mapping, inunda-
high vulnerability area. We estimated that inundation tion mapping, and impact assessment due to tsunami dis-
area will cover 8.13 square kilometer of building area aster.
along coastal area of Ofunato (Table 4 and Figure 8(b)). The use of AHP method helps in the analysis of spatial
Inundation area as shown in Figure 8(a) describe that multi-criteria where all of the parameters used in this
the farthest area of inundation could reach until 3.5 ki- study were calculated based on its weight factor to create
lometer from coastline. In this area, the elevation was 12 vulnerability map. This study is a first attempt to assess
tsunami vulnerability by using the parameter of SAVI
Pair - wise comparison matrix
Elevation Slope Coastline distance Vegetation density
besides elevation, slope and coastline distance, and apply
Elevation
1 2 3 3
AHP methods combining with raster overlay through
Slope 12 1 2 21 2
Coastline distance 13 1 2 1 3 GIS processing in Ofunato area, Japan. According to the
Vegetation density 13 0.4 1 3 1
result of this study, a simple method for inundation pre-
Normalized matrix diction can be a valuable step for carrying out a prelim-
Elevation 0.4615 0.5129 0.4737 0.3158
Slope
0.2308 0.2564 0.3148
0.2632 nary tsunami vulnerability mapping and impact assess-
Coastline distance 0.1539 0.1282 0.1579 0.3158
Vegetation density 0.1539 0.1026 0.0526
0.1053
ment when the high resolution of DEM data and detailed
0.4595
0.4595 0.4595 0.4595
45.94%
topographic data is not available.
0.2554 0.2554 0.2554 0.2554 25.53%
5 th iteration
0.1671
0.1671 0.1671 0.1671
16.71%
Elevation is the parameter that has the highest weight
0.1181 0.1181 0.1181 0.1181 11.81%
and 2.31% of the area that identified as slightly high and
Figure 5. Pair-wise comparison and normalized matrix. high class of tsunami vulnerability found in the coastal
area which has a lower elevation. Most of the area is
Table 3. Vulnerability classes. highly developed areas with low vegetation density. Tsu-
Vulnerability Class Area (square kilometer) Percentage (%)
nami vulnerability map and inundation map generated in
this study can be used for determining a priority for risk
Low (1) 234.79 73.63
prevention, mitigation, and land-use planning to reduce
Slightly low (2) 71.10 22.30 the tsunami risk. Particularly, in the slightly high and
Medium (3) 5.59 1.75 high vulnerability area that can be identified as possible
inundation area describe that 11.85% of building area has
Slightly high (4) 6.29 1.97
potential consequences of a tsunami impact.
High (5) 1.10 0.34 In this research, five classes of vulnerability were used.
Figure 6. (a) Spatial analysis step; (b) Tsunami vulnerability map of Ofunato area.
Vulnerability map describes that 73.63% of the area is high vulnerability area. In addition, we assumed that ve-
low vulnerability, 22.30% is slightly low vulnerability, getation may play an important role as tsunami barriers
1.75% is medium, 1.97% is slightly high vulnerability, to reduce the impact of the tsunami destruction. Based on
and 0.34% is high vulnerability. The high vulnerability the vegetation density that created from SAVI value in
areas were mostly found in the coastal area with the the study area, we found that 52.09% of the inundation
sloping coast type. Inundation areas were predicted in area was in slightly low class of vegetation density and
areas that identified as high vulnerability and slightly 34.05% of the inundation area was in medium class of
Figure 7. Inundation map of Ofunato during the 2011 Japan Tsunami, published by GSI [38] (modified).
Figure 8. (a) Map of possible inundation area in ofunato; (b) Tsunami vulnerability class of building area in Ofunato.
vegetation density. Ofunato. The vulnerability map showed that most of the
coastal areas are vulnerable to tsunami. Most of buildings
5. Conclusions area and bare land along coastal area was in the high
vulnerability of tsunami. The inundation pattern as the
GIS application followed by satellite image processing result of this study has shown similar compared to inun-
and AHP approach is useful for tsunami vulnerability dation area of the 2011 Tohoku Earthquake in the area of
mapping and impact assessment. It can be used for the Ofunato. ASTER GDEM version 2 is useful for tsunami
evacuation and reconstruction plan due to tsunami disas- vulnerability mapping for the area where the high resolu-
ter. GIS indicates the vulnerability area due to tsunami tion of DEM data is not available. Building and residence
and describes the possibility area that could be affected area in potentially affected areas should be provided with
by tsunami wave. In this study we applied four parame- basic information on tsunami risk because awareness and
ters in order to create a tsunami vulnerability map and to preparedness are the most important factors to reduce
pre- dict the possibility of inundation area in the area of potential losses due to a tsunami impact.
tion Prediction, Proceedings of the 15th ASCE Confer- International Journal of Remote Sensing, Vol. 21, No.
ence on Coastal Engineering, Honolulu, 11-17 July 1976, 6-7, 2000, pp. 1331-1364.
pp. 1006-1024. http://dx.doi.org/10.1080/014311600210209
[20] A. R. Huete, A Soil-Adjusted Vegetation Index (SAVI), [29] P. Gong, R. Pu, G. S. Biging and M. R. Larrieu, Estima-
Remote Sensing of Environment, Vol. 25, No. 3, 1988, pp. tion of Forest Leaf Area Index Using Vegetation Indices
295-309. Derived from Hyperion Hyperspectral Data, IEEE Tran-
http://dx.doi.org/10.1016/0034-4257(88)90106-X sactions on Geoscience and Remote Sensing, Vol. 41, No.
[21] L. S. Araujo, J. R. dos Santos and Y. E. Shimabukuro, 6, 2003, pp. 1355-1362.
Relationship Between SAVI and Biomass Data of Forest http://dx.doi.org/10.1109/TGRS.2003.812910
and Savanna Contact Zone and the Brazilian Amazonia, [30] A. R. Huete, G. Hua, J. Qi, A. Chehbouni and W. J. D.
International Archives of Photogrammetry and Remote Van Leeuwen, Normalization of Multidirectional Red
Sensing, Vol. 33, 2000, pp. 77-81. and NIR Reflectances with the SAVI, Remote Sensing of
[22] G. Chander and B. Markham, Revised Landsat-5 TM Environment, Vol. 41, No. 2-3, 1992, pp. 143-154.
Radiometric Calibration Procedures and Post Calibration http://dx.doi.org/10.1016/0034-4257(92)90074-T
Dynamic Ranges, IEEE Transactions on Geoscience and [31] A. R. Huete, A Soil-Adjusted Vegetation Index (SAVI),
Remote Sensing, Vol. 41, No. 11, 2003, pp. 2674-2677. Remote Sensing of Environment, Vol. 25, No. 3, 1988, pp.
http://dx.doi.org/10.1109/TGRS.2003.818464 295-309.
[23] M. Bouvet, G. Chander, P. Goryl and R. Santer, Prelimi- http://dx.doi.org/10.1016/0034-4257(88)90106-X
nary Radiometric Calibration Assessment of ALOS AV- [32] J. R. Jensen, Remote Sensing of the Environmental: An
NIR-2, Geoscience and Remote Sensing Symposium, Earth Resources Perspective, Prentice Hall, Upper Sad-
IGARSS, Barcelona, 2007, pp. 2673-2676. dle River, 2000.
[24] National Aeronautics and Space Administration, Data [33] T. L. Saaty, A Scaling Method for Priorities in Hierar-
Users Handbook-Data Products, Chapter 11, 2011. chical Structures, Journal of Mathematical Psychology,
http://landsathandbook.gsfc.nasa.gov/pdfs/Landsat7_Han Vol. 15, No. 3, 1977, pp. 234-281.
dbook.pdf http://dx.doi.org/10.1016/0022-2496(77)90033-5
[25] C. J. Tucker, Red and Photographic Infrared Linear [34] T. L. Saaty, The Analytic Hierarchy Process, Planning,
Combinations for Monitoring Vegetation, Remote Sens- Priority Setting, Resource Allocation, McGraw-Hill,
ing of Environment, Vol. 8, No. 2, 1979, pp. 127-150. New York, 1980.
http://dx.doi.org/10.1016/0034-4257(79)90013-0 [35] T. L. Saaty, Decision Making with the AHP; Why Is the
[26] R. D. Jackson, P. N. Slater and P. J. Pinter, Discrimina- Principle Eigenvector Necessary, European Journal of
tion of Growth and Water Stress in Wheat by Various Operational Research, Vol. 145, No. 1, 2003, pp. 85-91.
Vegetation Indices Through Clear and Turbid Atmos- http://dx.doi.org/10.1016/S0377-2217(02)00227-8
pheres, Remote Sensing of Environment, Vol. 13, No. 3, [36] E. H. Forman and M. A. Selly, Decision by Objectives,
1983, pp. 187-208. World Scientific Publishing Company, 2001.
http://dx.doi.org/10.1016/0034-4257(83)90039-1
[37] T. L. Saaty, Decision Making for Leaders: The Analyti-
[27] C. J. Tucker, W. W. Newcomb, S. O. Los and S. D. Prin- cal Hierarchy Process for Decisions in a Complex World,
ce, Mean and Inter-Year Variation of Growing-Season The Analytical Hierarchy Process Series, RWS Publica-
Normalized Difference Vegetation Index for the Sahel tion, 1996.
1981-1989, International Journal of Remote Sensing,
Vol. 12, No. 6, 1991, pp. 1133-1135. [38] Geospatial Information Authority of Japan (GSI), Tsu-
http://dx.doi.org/10.1080/01431169108929717 nami Inundation and Land Use Area of the 2011 Japan
Earthquake, Geospatial Information Authority of Japan
[28] M. C. Hansen, R. S. Defries, J. R. G. Townshend and R. (GSI Japan), 2011.
Sohlberg, Global Land Cover Classification at 1 km
Spatial Resolution Using a Classification Tree Approach,