Applied Geography 98 (2018) 22–33

Contents lists available at ScienceDirect

Applied Geography
journal homepage: www.elsevier.com/locate/apgeog

Assessing tropical cyclone risks using geospatial techniques T

a,b,∗ a a c
Muhammad Al-Amin Hoque , Stuart Phinn , Chris Roelfsema , Iraphne Childs
a
Remote Sensing Research Centre, School of Earth and Environmental Sciences, The University of Queensland, Brisbane, QLD 4072, Australia
b
Department of Geography and Environment, Jagannath University, Dhaka 1100, Bangladesh
c
Queensland Centre for Population Research, School of Earth and Environmental Sciences, The University of Queensland, Brisbane, QLD 4072, Australia

A R T I C LE I N FO A B S T R A C T

Keywords: Tropical cyclones are one of the most catastrophic natural disasters, regularly affecting coastal areas across the
Tropical cyclone world. The intensity and extent of impacts by these disasters have remained very high throughout history. An
Natural disaster appropriate mapping approach is essential for producing detailed risk assessments to plan for, and reduce the
Hazard future impacts of cyclones on people, property and the environment. This study developed and evaluated a
Vulnerability
spatial multi-criteria approach for mapping the risk levels of areas to tropical cyclone impacts using remote
Risk
sensing, field data and spatial analysis at a local scale covering < 1000 km2. We used the Sarankhola Upazila
Analytical hierarchy process
(151 km2), a local government area, from coastal Bangladesh to test the applicability of this approach. Three risk
components: vulnerability and exposure; hazard; and mitigation capacity, and their 14 relevant criteria were
considered. Thematic raster map layers were prepared for every criterion and weighted using an Analytical
Hierarchy Process (AHP). A weighted overlay technique was used for generating individual maps for every risk
component and then risk map was produced from them. The resulting maps successfully identified the spatial
extents and levels (very high, high, medium, low, and very low) of risk. Our results indicate that eastern, south
and south-western parts of the study area are likely to be subject to higher risk from tropical cyclone's surge
height, strong wind and intensive rainfall than other parts. The approach provided was validated by confidence
level assessment and a map of past cyclone impacts. Our preliminary findings indicate this approach has the
potential to assess risks from tropical cyclone impacts in the affected countries by producing risk maps able to be
used by policymakers and administrators.

1. Introduction year (Burroughs, 2007). Moreover, it is predicted and expected by re-

cent studies that the intensity and frequency of tropical cyclones will
Tropical cyclones are one of the most catastrophic hydro-meteor- increase enormously under future climate change scenarios (Deo &
ological natural disasters in coastal environments (Li & Li, 2013). These Ganer, 2014; Krishnamohan, Mohanakumar, & Joseph, 2014;
deadly disasters are associated with strong winds, heavy rainfall and Mendelsohn, Emanuel, Chonabayashi, & Bakkensen, 2012; Yin, Yin, &
large storm surges resulting in widespread human, economic, social and Xu, 2013). Consequently, coastal lives, resources and environments will
environmental losses (Hoque, Phinn, & Roelfsema, 2016a; Puotinen, be under constant threats in the future (Hoque, Phinn, & Roelfsema,
2007). The occurrences and impacts of tropical cyclones are very 2017a).
common in many coastal areas across the world (Peduzzi et al., 2012). Prevention and reduction of tropical cyclone disaster impacts are
According to the statistics, at the global level, 637 tropical cyclones core disaster management activities (Yin et al., 2013). The key infra-
were reported during the period of 1970–2010 (Weinkle, Maue, & structure and areas at risk in relation to spatial location, level of risk
Pielke, 2012). Over the last two centuries, about 1.9 million people and factors liable for risk are required information in the context of
have lost their lives due to tropical cyclones worldwide (Shultz, Russell, finding suitable cyclone mitigation options (Gao et al., 2014; Li, Ahuja,
& Espinel, 2005). Tropical cyclones cause the largest economic and & Padgett, 2011). This type of information is usually derived from risk
environmental losses compared to the other hydro-meteorological assessments. Risk assessments are the essential groundwork to prevent
natural disasters (Pielke et al. 2008; Hoque, Phinn, Roelfsema, & Childs, and reduce the impacts of cyclones (Birkmann, 2007; Hoque, Phinn,
2016b). Tropical cyclone disasters in the Gulf and east coast of the Roelfsema, & Childs, 2017b). An effective risk assessment procedure
United States cause, on average, nearly US 5 billion dollars damage per includes the mapping of potential hazards, along with their resultant

∗
Corresponding author. Remote Sensing Research Centre, School of Earth and Environmental Sciences, The University of Queensland, Brisbane, QLD 4072, Australia.
E-mail addresses: [email protected] (M.A.-A. Hoque), [email protected] (S. Phinn), [email protected] (C. Roelfsema), [email protected] (I. Childs).

https://doi.org/10.1016/j.apgeog.2018.07.004
Received 9 March 2017; Received in revised form 22 June 2018; Accepted 2 July 2018
0143-6228/ © 2018 Elsevier Ltd. All rights reserved.
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

vulnerability and exposure levels, and the capacity of existing mitiga- (AHP) is an effective tool for combining a larger number of input
tion (Cutter, Boruff, & Shirley, 2003; Dewan, 2013a). Hazards are variables, for example on hazard and vulnerability, in a multi-criteria
events (Blaikie, Cannon, Davis, & Wisner, 2014) which affect life, decision making process to derive risk assessments (Malczewski, 1999).
property, and the environment (Dewan, Islam, Kumamoto, & Nishigaki, AHP allows for the analysis of the spatial multi-criteria layers through
2007; Rashid, 2013). Vulnerability is linked to the exposure of element the generation of a hierarchical structure providing weighting and
at risk and the extent to which a community and environment are likely ranking to support the spatial decision making process with the views of
to be affected by a hazard (Eckert, Jelinek, Zeug, & Krausmann, 2012; experts and user (Dewan, 2013b). A weighted overlay technique is used
Rashid, 2013; Turner et al., 2003). Exposure is proportional to the to integrate weighted and ranked spatial criteria together. AHP has
population and property value located in hazard-prone areas. Mitiga- been successfully used for mapping hazards, vulnerability and risk of
tion capacity involves structural and non-structural initiatives (e.g. other natural disasters such as floods, landslides, and earthquakes,
cyclone shelter, warning system) used to reduce the impacts of tropical hence it is considered suitable for tropical cyclone risk assessment
cyclone hazards (Coppola, 2006; Khan, 2008). While vulnerability, (Abella & Van Westen, 2007; Chen, Ito, Sawamukai, & Tokunaga, 2015;
exposure, and hazard act simultaneously to contribute the risk of nat- Panahi, Rezaie, & Meshkani, 2013; Roy & Blaschke, 2013).
ural hazard, but mitigation capacity reduces the risk. The maps pro- This study developed and examined a risk mapping approach for
duced by risk assessment have been used by policymakers and admin- tropical cyclone impacts using AHP to combine information derived
istrators to devise effective management plans with targeted prevention from remote sensing, field data and spatial analysis at a local scale. The
and reduction measures (Mahapatra, Ramakrishnan, & Rajawat, 2015; study was divided into three specific objectives: (1) develop a spatial
Masuya, Dewan, & Corner, 2015; Yin et al., 2013). Thus, the risk as- AHP based multi-criteria integrated risk mapping approach for tropical
sessment procedure can minimise disaster losses and even provide cyclone impacts; (2) examine the developed approach for mapping
protection from disaster occurrence. tropical cyclone risk at the local scale covering < 1000 km2 in
Several studies were conducted using remote sensing and spatial Sarankhola upazila, Bangladesh; and (3) assess the validation of the risk
analysis for mapping the hazard, vulnerability and risk of tropical cy- mapping approach.
clone disasters using different approaches by Li and Li (2013); Yin et al.
(2013); Rafiq, Blaschke, and Zeil (2010); Yin et al. (2010); Kumar and
Kunte (2012); Saxena, Purvaja, Suganya, and Ramesh (2013). Accurate 2. Study area
and detailed tropical cyclone risk information is determined by the
selection of appropriate and sufficient criteria, scale as well as com- The study was carried out in Sarankhola Upazila, a local govern-
ponents of risk (Dewan, 2013a; Rashid, 2013). Adequate criteria se- ment area, (about 151.24 sq. km) of Bagerhat District which is located
lection under each of the risk components (e.g., hazard, vulnerability in coastal Bangladesh (Fig. 1). The study area was chosen as tropical
and exposure, and mitigation) and their processing serve as a founda- cyclones are the most common disaster in coastal Bangladesh, striking
tion to assess the risk in a more reliable way (Birkmann, 2007). Simi- the country almost every year with a varied scale (Choudhury, Paul, &
larly, the size of the study area (local or regional scale) is a vital issue in Paul, 2004; Islam & Peterson, 2009). About 508 cyclones originated in
deriving detailed risk information (Wang, Li, Tang, & Zeng, 2011). the Bay of Bengal in the last 100 years, and out of them 17% struck
Detailed local level risk information is required to identify the type of coastal Bangladesh (Paul, 2009). The selected study area is one of the
risk accurately as well as to select the best mitigation options in the most tropical cyclone prone coastal Upazilas located on the mouth of
process of appropriate management plans (Blaikie et al., 2014). How- the Bay of Bengal in the southern part of the country (Islam, Bala,
ever, most available studies have been conducted based on limited Hussain, Hossain, & Rahman, 2010; Nadiruzzaman & Paul, 2013). The
criteria and were at the regional scale. area is open to the sea surface as well as subject to high storm surges by
Another fundamental issue with risk assessment is the selection of the coastal Baleshwar and the Bhola River which surround it on both
risk components. Generally, risk is considered as the product of vul- sides. As a lowland and densely populated area, human as well as all
nerability, exposure and hazard (Birkmann, 2007). The evaluation of kinds of natural and physical resources are highly vulnerable to tropical
actual risk to a particular community, resources and the environment cyclone impacts by strong winds, flooding and storm surges (Islam
not only depends on the level of vulnerability, exposure and hazard, but et al., 2010). This coastal area is characterised by a humid climate and
also on the existing mitigation capacity to prevent and reduce the dis- annual rainfall is 2200 mm. Tropical cyclones typically originate and
aster impacts (Cutter, 1996; Frazier, Thompson, & Dezzani, 2013a). To strike the area in the period of March–July and September–December
assess the actual risk information, it is essential to incorporate mitiga- due to the dominance of humid air in the nearby ocean (Islam &
tion capacity in the risk assessment procedure (Cutter et al., 2003). Peterson, 2009).
Currently, very few studies are available in the existing literature fo-
cusing on multi-criteria integrated tropical cyclone risk assessment at
3. Materials and methods
the local scale covering < 1000 km2 and incorporating mitigation ca-
pacity using remote sensing and spatial analysis.
3.1. Overview
Geospatial techniques using remote sensing and spatial analysis are
efficient and accurate tools for deriving the information sources re-
The geospatial multi-criteria assessment technique in the form of
quired for tropical cyclone risk assessment (Rafiq et al., 2010;
AHP is adopted in this study to integrate many natural and anthro-
Taramelli, Melelli, Pasqui, & Sorichetta, 2008; Yin et al., 2013). Remote
pogenic criteria in the risk assessment process. This approach can in-
sensing provides a capability to deliver maps of environmental features
tegrate and aggregate many criteria easily and can present output in
at repeated time frames from spatial scales of a few meters to entire
very simple manure (Dewan, 2013b; Yin et al., 2013). Several risk
continents. This essential environmental information enables the spatial
equations are available for risk mapping for any kind of natural or man-
risk assessment of natural hazards (Mahendra, Mohanty, Bisoyi, Kumar,
made hazard. A well-developed and complete risk equation with its
& Nayak, 2011; Martino, Ulivieri, Jahjah, & Loret, 2009). Spatial ana-
components gives a better risk evaluation. On the basis of reviewing
lysis supports data collection and analysis as well as the integration of
existing literature, the following risk equation has been considered for
spatial and aspatial data for spatial decision making (Cutter et al., 2003;
evaluation of tropical cyclone risk in this study:
Dewan, 2013a; Kunte et al., 2014).
The spatial decision making processes weighting and ranking are Risk = hazard × vulnerability × exposure/mitigation capacity (1)
required to integrate multi-criteria of risk components in the risk as-
sessment process (Roy & Blaschke, 2013). Analytical Hierarchy Process The processes used in this study are summarised in Fig. 2.

23
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

Fig. 1. (a) Study area, Sarankhola Upazila under Bagerhat district in Bangladesh drawn on Landsat 8 OLI (Operational Land Imager) image (Bands: 6:5:1 combi-
nations) of 15 April 2014, (b) Coastal region of Bangladesh and location of study area.

international organisations as well as from field work. The field work

was carried out in the study area from March to May 2015 to collect
mitigation capacity and validation data. Mitigation capacity data in-
cluded cyclone shelters, health infrastructure and cyclone warning
systems.
WorldView-2 imagery was used to produce land use and land cover
map. The point-based reference data for accuracy assessment of land
use and land cover map were collected from the study area in 2015.
Table 1 outlines the data type, source, period and particular mapping
application used in the present study.

3.3. Risk evaluation criteria, alternatives and mapping

The criteria and alternatives in this study were selected based on a

literature and data review considering their relevance as well as their
influence in terms of tropical cyclone risk. The selected alternatives of
each criterion were mapped to convert into spatial thematic layers. For
this study, 14 different spatial thematic criteria layers were produced
under three risk components. The Universal Transverse Mercator
(UTM) zone 45 north and World Geodetic System (WGS)-1984 datum
were used as the referencing system for all the raster layers. The spatial
resolution of each raster layer was set to 15 m × 15 m cell size. In ad-
dition, the natural break statistical method was used to classify maps
where it was required. We used the ArcGIS software (version 10.3) for
processing and preparing most of spatial criteria layers. The importance
of selected criteria and their mapping procedures are explained in the
following paragraphs.
Fig. 2. Flowchart of the risk assessment approach used in this study.
3.3.1. Criteria for vulnerability and exposure mapping
The vulnerability and exposure to tropical cyclones are controlled
3.2. Dataset and sources
by some drivers and these have been selected as criteria for this ana-
lysis. In total, six criteria were selected categorizing into vulnerability
Most of the risk component criteria used for this study was dynamic
and exposure. The vulnerability was comprised of elevation, slope,
in nature as its datasets are updated periodically. A wide variety of data
proximity to coast and proximity to cyclone track and these factors are
was required and used from different sources. The criteria layers were
mainly associated with physical vulnerability. The land use and cover
derived from these distinctive data using spatial analysis and remote
along with population density were the features of the exposure ana-
sensing techniques. The data were collected from national and
lysis.

24
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

Table 1
Dataset and sources used for this study.
Data type Source Period Mapping output

Worldview 2 (1.84 m resolution) DigitalGlobe foundation 13-11-2012 Land use and land cover
IKONOS-2 (4 m resolution) DigitalGlobe foundation 09-02-2014 Coastal vegetation
SRTM-DEM (30 m resolution) United States Geological Survey (USGS) Earth 11-02-2000 Elevation, slope, storm surge height
explorer
SPOT-5 (10 m resolution) Airbus Defence and Space 22-10-2007 Risk map validation
23-11-2007
Population Bangladesh Bureau of Statistics (BBS) Population Census Population density
2011
Cyclone track International Best Track Archive for Climate 1942–2015 Proximity to cyclone track, cyclone frequency
Stewardship (IBTrACS)
Historical cyclone Bangladesh Meteorological Department (BMD) 1968–2015 Cyclone frequency, storm surge height
and published research articles
Precipitation Bangladesh Meteorological Department (BMD) 1960–2015 Precipitation intensity
Cyclone wind speed NOAA Regional and Mesoscale Meteorology 2007 Cyclone Sidr Cyclone wind speed
Branch (RAMMB)
Cyclone shelter, health infrastructure, cyclone Field work March–May 2015 Distance to cyclone shelter and health
warning system infrastructure, cyclone warning system

Coastal slope and elevation have a high importance in terms of probability based on the calculation of the historical database and
vulnerability assessment. Areas with gentle slopes are highly vulnerable present observations. Location, time, intensity and frequency of the
to tropical cyclone induced storm surges than areas of higher elevation event are all considered to determine hazard probability (Dewan,
with steep slope (Mahapatra et al., 2015). In the present study, we used 2013a). The selected criteria used to detect potential hazard for this
the Shuttle Radar Topography Mission (SRTM) modified Digital Ele- study are storm surge height, cyclone wind speed, cyclone frequency,
vation Model (DEM) at 30 m resolution for generating an elevation and and precipitation intensity.
slope map (Fig. 3a and b). Tree offsets are present in the SRTM data as Intensive inundation and flooding in the coastal areas are controlled
radar used for the SRTM mission cannot penetrate the tree canopies by storm surge heights. In the present study, a storm surge height map
fully (Lewis, Bates, Horsburgh, Neal, & Schumann, 2013). To produce a was produced using a model integrated with historical storm surge data
bare earth DEM, the tree offsets were removed using a tree height and a bare earth DEM modified from SRTM data at 30 m resolution
offsets estimation method provided by Gallant, Read, and Dowling (Fig. 4a). The historical storm surge records from 1960 to 2015 were
(2012). Then bare earth DEM was further processed to remove other considered to determine the maximum surge height in the course of a
errors including void filled areas and spikes converting into points and 20-year return period. Based on frequency analysis, the study found
re-interpolated. that 6.10 m was the maximum surge height (Dasgupta et al., 2011). A
In general, the areas close to the cyclone track and coastline are surge decay coefficient was calculated for this surge height. Using a
considered highly susceptible compared to the areas which are located surge decay coefficient and a bare earth DEM, various raster calculator
further away. It is very important to include proximity to the cyclone formulas within the ArcGIS environment were used to produce the
track and the coastline as both of these play a major role in the vul- storm surge height map. The detailed procedures for calculating surge
nerability assessment. International Best Track Archive for Climate decay coefficient as well as producing storm surge height map through
Stewardship (IBTrACS) data from 1968 to 2015 was used for mapping different raster calculator formulas are discussed in Hoque, Phinn,
the proximity to the cyclone track (Fig. 3c) (Knapp, Kruk, Levinson, Roelfsema, and Childs (2018).
Diamond, & Neumann, 2010). By contrast, the proximity to the coast- Intensity of cyclone wind speed determines how much devastation
line data was generated by measuring the distance of the study area can occur in the man-made and natural features of a landscape. In this
from the coastline using the Google earth pro ruler and this data was study, a cyclone wind speed spatial layer was generated using cyclone
used for preparing the proximity to the coastline spatial layer (Fig. 3d). Sidr (2007) wind speed distribution data acquired from the NOAA re-
Certain types of land covers are more vulnerable to damage and gional and mesoscale meteorology branch (RAMMB) (Fig. 4b). Owing
destruction by tropical cyclones. For this study, WorldView-2 imagery to the unavailability of previous cyclone wind speed spatial distribution
was used for land use and land cover mapping (Fig. 3e). An object- data, only the cyclone Sidr (2007) wind speed was considered in the
based classification approach was adopted for classifying seven land use present study. Cyclone Sidr was a category 4 storm which struck with
and land cover classes (Table 2): vegetation; settlement; bare land; an average wind speed of 223 km/h on 15 November 2007, and the
closed water bodies; open water bodies; crops land, and infrastructure. study area was completely devastated by this cyclone.
We used the Trimble eCognition Developer 9.0 software for all of the Cyclone frequency helps to make a prediction of the probable time
object-based classification processes. The accuracy assessment of the interval of tropical cyclone formation and landfall. To prepare cyclone
land use and land cover map was carried out by collecting 400 random frequency map, previous historical cyclone data with their tracks over
points from the field (Field work, March to May 2015) with a minimum the period of 1950–2015 and their impacts on the study area were
of 50 points per cover class by using the stratified random sampling taken into consideration in this study. Historical data and field-based
method. The overall accuracy of the classified image was 91%. investigation indicated that the study area was affected by 22 tropical
Coastal populations are highly exposed to tropical cyclone disasters cyclones over the last 65 years. Cyclone track and local people's views
and population density also has a critical influence on the level of helped to estimate the frequency of cyclones in the particular areas, and
vulnerability (Frazier, Thompson, Dezzani, & Butsick, 2013b). The these data were then converted into spatial layers using spatial analysis
population density layer was prepared using the population census data (Fig. 4c).
of 2011 which was acquired from the Bangladesh Bureau of Statistics Heavy rainfall is associated with tropical cyclones. The spatially
(BBS) (Fig. 3f). distributed average rainfall data associated with landfall tropical cy-
clones are not available for the study area. Therefore, to map pre-
3.3.2. Criteria for hazard mapping cipitation intensity, the daily precipitation data (1950–2015) collected
The likelihood of a tropical cyclone hazard strike is determined by from the Bangladesh Meteorological Department was used in this study.

25
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

Fig. 4. Hazard assessment criteria layers: (a) storm surge height, (b) cyclone
wind speed, (c) cyclone frequency, and (d) precipitation intensity.

3.3.3. Criteria for mitigation capacity mapping

The criteria chosen as mitigation initiatives to lessen hazard impacts
are cyclone shelters, health infrastructure, coastal vegetation and cy-
clone warning systems.
Cyclone shelter construction is an effective mitigation measure
which helps to give temporary accommodation for affected people prior
to cyclone landfall (Mallick & Rahman, 2013). Local health infra-
structure can also provide emergency health care for affected people
and make them less vulnerable. For the present study, spatial cyclone
shelter and health infrastructure (hospital, union health centre and
functional community clinic) data were directly collected from the field
Fig. 3. Vulnerability and exposure criteria layers: (a) elevation, (b) slope, (c) using a Global Positioning System (GPS) device. The ‘Euclidian dis-
proximity to cyclone track, (d) proximity to coastline, (e) land use and land tance’ technique within the ArcGIS environment was used for preparing
cover, and (f) population density. distance to cyclone shelter and health infrastructure spatial layers
(Fig. 5a and b).
Coastal vegetation operates as a strong form of protection for
Table 2
Land use and land cover classification scheme. people, property and the environment by reducing the impacts of cy-
clone winds and storm surges (Paul & Rahman, 2006). In the present
Land use/land cover type Description
approach, IKONOS-2 imagery was used to map the coastal vegetation
Vegetation Homestead forest, deciduous forest, mangrove forest, (Fig. 5c). An object-based image analysis was adopted as a classification
and others approach. Based on the coastal vegetation density found in the study
Settlement Rural and urban settlements area, the vegetation covers were classified into three classes: high,
Bare land Bare soil, fallow land, open space, and sand fillings medium and low vegetation cover. On the other hand, proper and
Closed water bodies Ponds, lakes, reservoirs, and low land
Open water bodies River, canal, and permanent open water
timely warnings can save many human lives and their valuable re-
Crops land Agricultural lands, crop fields, and vegetable lands. sources (Mallick & Rahman, 2013). The present warning system for the
Infrastructure Roads, bridges, and embankment study area was investigated through field visits including interviewing
local people, experts and administration. Based on the collected in-
formation, the study areas warning system dissemination was classified
To begin with, an average annual precipitation map of Bangladesh was into three categories; ineffective, moderate and effective warning
prepared interpolating all of 35 rainfall station data. The kriging in- system (Fig. 5d).
terpolation technique was used in ArcGIS environment. The part of the
study area was then extracted from this map (Fig. 4d).

26
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

These spatial criteria raster layers alternatives were then standardised

into a common scale 0 to 1 using the linear scale transformation
equation (2). This procedure helped to prepare all criteria layers in the
same scale unit to make the multi-criteria decision applicable.
x− min
p=
max−min (2)

where p is the standardised score, min and max represent the minimum
and maximum values associated with each dataset and x is the cell
value.

3.5. Weighting the criteria using AHP

The AHP technique was used in this study for weighting the criteria
of vulnerability and exposure, hazard, and mitigation capacity in the
risk assessment process. For weighting the selected criteria of vulner-
ability and exposure, hazard, and mitigation capacity, the individual
pairwise comparison matrix was established through the qualitative
judgments of five experts, and a user, each of whom considered the
influencing factors and alternatives which were integrated with every
criterion. For prioritisation of the criteria, experts were asked based on
the pairwise comparison 9 point scale (Table 4) developed by Saaty
(2008). Five experts were selected at the national level. The expert
selection was based on their related experience, research, in-depth
knowledge about study site and coastal Bangladesh. The selected ex-
perts were from academic, governmental, and research institutions. The
sum of all the scores of hazard, vulnerability and exposure, and miti-
gation capacity was 1, respectively.
The consistency ratio (CR) was calculated to justify the consistency
Fig. 5. Mitigation capacity criteria layers: (a) distance to cyclone shelter, (b)
of comparisons provided by experts and the user in the pairwise com-
distance to health infrastructure, (c) coastal vegetation, and (d) cyclone parison matrix. The comparisons in the pairwise comparison matrix are
warning system. considered consistent if the CR is equal or less than 0.1 (Malczewski,
1999). The CR was calculated using the following equation:
3.4. Alternative ranking and standardisation criteria layer CR = Consistency Index/Random Index (3)

where, Random Index (RI) represents the randomly generated average

The mapped alternatives under each spatial criteria layer were
consistency index, and Consistency Index (CI) which is defined as fol-
ranked into five levels of risk from 1 to 5 where rank 1 indicated a very
lows.
low risk and rank 5 represented a very high risk (Table 3). The ranking
of alternatives was performed based on the contribution of risk and the CI = (λ max − n)/(n − 1) (4)
AHP.
where λ max refers the largest eigenvalue of the matrix and n represents
All the criteria map layers were then converted into a 15 m pixel
the order of the matrix (Mahapatra et al., 2015).
raster layer to support the raster based weighted overlay technique.
The criteria weights derived from the pairwise comparison matrices

Table 3
Alternative ranking scheme based on the contribution to the risk of tropical cyclone disaster.
Component Criteria Ranking (Based on risk)

Very Low (1) Low (2) Moderate (3) High (4) Very High (5)

Vulnerability and Elevation (m) >4 3–4 2–3 1–2 <1

Exposure Slope (%) > 1.52 0.89–1.52 0.54–0.89 0.24–0.54 < 0.24
Proximity to coastline (km) 26–30 22–26 18–22 14–18 10–14
Proximity to cyclone track > 5.20 3.54–5.20 1.70–3.54 0.53–1.70 < 0.53
(km)
Land use and land cover Open water bodies, Closed water bodies Vegetation Crops ground Infrastructure, Settlement
bare land
Population density (sq. km) 424 424–593 593–729 729–832 832–961
Hazard Storm surge height (cm) < 10 10–150 150–250 250–340 > 340
Cyclone frequency (number) <6 6–8 8–10 10–18 > 18
Precipitation (mm) < 2120 2120–2143 2143–2167 2167–2190 > 2190
Cyclone wind speed (km/h) – < 170 170–176 176–180 > 180
Mitigation capacity Coastal vegetation – High cover Moderate cover Low cover No cover
Cyclone warning system – Ineffective warning Moderate warning Effective warning –
system system system
Distance to cyclone shelter < 500 500-1000 1000–1500 1500–2000 > 2000
(m)
Distance to health < 1000 1000–2000 2000–3000 3000–4000 > 4000
infrastructure (m)

27
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

Table 4
Scale of relative importance (adapted from Saaty, 2008).
Relative importance Definition Description

1 Equal importance Two factors influence equally to objective

3 Moderate importance Experience and judgement slightly favour one factor over another
5 Strong importance Experience and judgement strongly favour one factor over another
7 Very strong importance One decision factor is favoured strongly over another and its supremacy is established in
practice
9 Extreme importance The evidence favouring one decision factor over another is of the highest possible order of
validity
2, 4, 6, 8 Intermediate values between the two adjacent judgements When compromise is needed

Table 5 Table 6
Criteria weights produced from the pairwise comparison matrices and con- Confidence level scale rated according to the level of confidence experts had in
sistency ratio of comparisons through the qualitative judgments of five experts the developed mapping procedure that identified the specific spatial data
and a user based on the scale of relative importance (Table 4). (FAQs, 2012; Glanville et al., 2016).
Component Criteria Weight Rating Description

Vulnerability and exposure Elevation 0.19 Highest Known spatial data – Data expert knowledge, peer reviewed method
Slope 0.10 and supporting evidence (e.g. field validation); the mapped risk
Proximity to coastline 0.21 component criteria layer has been accurately identified.
Proximity to cyclone track 0.27 High Derived spatial data – Data expert knowledge, peer reviewed method
Land use and land cover 0.07 and there is a high confidence in the mapping procedure
Population density 0.16 Moderate Derived spatial data – Data expert knowledge, there is a moderate
Consistency ratio: 0.06 confidence in the mapping procedure
Hazard Storm surge height 0.45 Low Derived spatial data – Data expert knowledge, there is a low
Cyclone frequency 0.14 confidence in the mapping procedure
Precipitation 0.10 Unknown No history or knowledge about how data was collected or created.
Cyclone wind speed 0.30
Consistency ratio: 0.05
Mitigation capacity Coastal vegetation 0.17
(2007) damage extent and impact using SPOT-5 satellite images are
Cyclone warning system 0.30
Distance to cyclone shelter 0.45
discussed in Hoque et al. (2016b).
Distance to health infrastructure 0.08
Consistency ratio: 0.03
4. Results and discussion

and associated consistency ratio of comparisons are presented in 4.1. Vulnerability and exposure mapping
Table 5.
Vulnerability and exposure to tropical cyclones were mapped and
3.6. Risk assessment categorised into five levels (Fig. 6). The resulting map indicates that
moderate to very high vulnerability and exposure covered 74% of the
The weighted overlay technique was applied individually with study area, whereas very low to low vulnerability and exposure ac-
spatial criteria layers of vulnerability and exposure, hazard, and miti- counted for 24%. The south-eastern and south-western parts of the
gation capacity. This procedure generated the vulnerability and ex- study area were more vulnerable and exposed to the impact of tropical
posure, hazard, and mitigation capacity indices. The produced indices cyclones. The proximity to cyclone track and coastline, lower elevation,
were categorised into five classes namely very low, low, moderate, high gentle slopes, high population densities, vulnerable infrastructure and
and very high for producing vulnerability and exposure, hazard, and settlements etc. are major factors controlling high vulnerability and
mitigation capacity maps. According to the adopted equation (1) for exposure locations. In contrast, north and north-western areas exhibited
risk assessment, a risk index was produced multiplying hazard and lower vulnerability and exposure. This is because these areas are lo-
vulnerability with exposure indices, and then dividing by the mitigation cated farther from the coastline, water bodies, and cyclone tracks as
capacity index using the raster calculator in the spatial analysis en- well as being comparatively less populated.
vironment. The risk index values were then standardised using equation
(2) in the scale of 0–1 and classified into five levels of risk being very 4.2. Hazard mapping
low, low, moderate, high and very high.
A hazard map (Fig. 7) was produced by classifying the hazard index
3.7. Risk mapping validation values into five distinct zones. Fig. 7 shows that 15% of the study area
was located in the very high hazard zone, mostly close to the coastal
The risk map was validated using two validation approaches. river, with 25% of the area as a high hazard zone and around 18% of
Firstly, a qualitative confidence level scale was developed for the da- the area as a moderate hazard zone. The area classified as a low and
taset and mapping procedure. The confidence level scale was divided very low hazard zone accounted for 16% and 26% respectively, and this
into five categories (Table 6) and rated according to the level of con- is mostly towards the inland, away from the coast. The hazard map also
fidence in the developed mapping procedure and spatial dataset used revealed that compared to other parts, the eastern, south-eastern and
(Table 7). Secondly, the mapping results were compared with the extent south-western parts of the study area were located in the highly hazard
and impact of tropical cyclone Sidr (2007), as mapped by geospatial prone zones.
techniques (Hoque et al., 2016b). The extent and impact of cyclone Sidr The hazard intensity is considerably high in these areas because
were assessed by SPOT-5 satellite images at 10 m resolution using they are prone to high storm surge impacts due to the close proximity of
change detection technique acquired immediately before and after coastal rivers and waterbodies, high precipitation intensity with lower
landfall. The detailed procedures for mapping tropical cyclone Sidr elevation, as well as the high frequency of cyclones. Conversely, the

28
 Table 7
Confidence levels for dataset and methods used to produce risk map (see Table 6 for confidence rating definitions and Table 5 for weighting).
Components of risk Data Description Importance/Weight Data type and main method Confidence level Justification

Vulnerability and exposure Elevation 0.19 Processed imagery – spatial analysis High Modified by peer reviewed methods, good quality input data and well accepted
M.A.-A. Hoque et al.

approach
Slope 0.10 Processed imagery – spatial analysis High Modified by peer reviewed methods, good quality input data and well accepted
approach
Proximity to coastline 0.21 Google earth imagery – spatial analysis High High confidence input data and peer reviewed approach
Proximity to cyclone track 0.27 Cyclone track – spatial analysis High Peer reviewed method and high confidence input data collected
Land use and land cover 0.07 Processed imagery - object based Highest Peer reviewed methods, good quality input data and independently validated
Population density 0.16 Population - spatial analysis High Peer reviewed methods, high confidence input data and well accepted
Hazard Storm surge height 0.45 Information extraction - empirical storm surge High Peer reviewed methods and high confidence Input data
model
Cyclone wind speed 0.30 Information extraction - spatial analysis Moderate High confidence input data, valid approach but various assumptions
Cyclone frequency 0.10 Cyclone track-field survey - spatial analysis High Peer reviewed methods and good quality input data
Precipitation intensity 0.14 Precipitation - spatial analysis High Peer reviewed methods, high confidence input data and well accepted
Mitigation Distance to cyclone shelter 0.45 Cyclone shelter - field survey Highest Collected from field and analysed following peer reviewed method
Distance to health infrastructure 0.08 Health infrastructure - field survey Highest Collected from field and analysed following peer reviewed method
Coastal vegetation 0.17 Processed imagery - Object based Highest Peer reviewed methods, good quality input data and independently validated
Cyclone warning system 0.30 Field survey Highest Collected from field and analysed following peer reviewed method

29
zard.

high wind speeds.

4.3. Mitigation capacity mapping

of vulnerability and exposure to tropical cyclones.
Fig. 6. Vulnerability and exposure map showing spatial distribution and level

Mitigation capacity map (Fig. 8) was prepared by classifying the

prone due to the less likely influence of storm surges, heavy rainfall and
areas of the north and north-western parts were found to be less hazard
Fig. 7. Cyclone hazard map showing spatial distribution and the level of ha-
Applied Geography 98 (2018) 22–33
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

Fig. 8. Mitigation capacity map showing spatial distribution and level of mi- Fig. 9. Cyclone risk map showing spatial distribution and level of risk derived
tigation capacity against the consistent threats of tropical cyclone. from spatial vulnerability and exposure, hazard and mitigation capacity indices.

mitigation capacity index values into five levels of mitigation capacity. 4.5. Risk mapping validation
A higher mitigation capacity indicates an area with approaches in place
for reducing the level of risk. The derived mitigation capacity map The adopted approaches provided reliable outputs to validate our
shows that 56% of the study area is classified as having moderate to risk assessment approach, justifying the confidence for data and map-
very high mitigation capacity zones. These areas are close to cyclone ping procedures and verifying the risk results with past cyclone event.
shelters and health infrastructure as well as good cyclone warning The confidence levels were assigned for dataset and methods
systems. (Table 7) based on a confidence level scale (Table 6). Most of the data
It is worth noting that after the study area was highly devastated by and methods used for vulnerability and exposure mapping were rated
a category four tropical cyclone Sidr (2007) (Nadiruzzaman & Paul, as having a high level of confidence. This is because data were derived
2013), some noticeable mitigation strategies were implemented in the from good quality input and mapped using peer reviewed methods. In
area. As a result, the overall mitigation capacity of the area is good contrast, data and method used for land use and cover mapping were
compared to other more remote coastal areas in Bangladesh. On the rated with highest confidence level due to their independent field va-
contrary, very low and low mitigation capacity areas cover 10% and lidation. However, elevation, slope, land use and land cover, and po-
34%, respectively and these areas are deprived from appropriate miti- pulation density criteria were finally rated in the moderate level based
gation measures. on importance derived from AHP and confidence level, whereas
proximity to cyclone track and coastline were with high levels (Fig. 10
and Fig. 11). On the other hand, all of the data and methods used for
4.4. Risk mapping
hazard mapping were rated in the high confidence level for their high
quality input and peer reviewed methods except cyclone wind speed.
The approach adopted in this study produced a detailed risk map
The data and method used for cyclone wind speed mapping were rated
(Fig. 9) by categorizing the standardised risk index values into five
with moderate level as many approaches are associated with this
levels of risks. The resulting risk map shows that 6% of the study area is
located in the very high risk zone, while 16% area is covered by high
risk zone, both of the zones mostly close to the coastal river. These high
to very high risk zones covered comparatively few areas due to good
mitigation measures in the study area. The moderate risk zone con-
stituted 28% of study area. The area categorised as low and very low
risk zone accounted for 26% and 23% accordingly, mostly towards
inland areas away from the coast.
The present study reveals that the eastern, south and south-western
parts of the study region are under more risk of tropical cyclones
compared to other parts. The noticeable influencing factors behind this
high risk are considered to be the proximity to cyclone tracks and
coastline, high population density, high storm surge impacts, low ele-
vation and inadequate mitigation measures etc. In contrast, the north,
north-western and some central parts of the study region are classified
as comparatively low risk zones due to less vulnerability (Fig. 6), lower Fig. 10. Rating of importance and confidence criteria of tropical cyclone risk
hazard levels (Fig. 7) as well as well a range of mitigation measures. assessment.

30
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

Fig. 11. Component and criteria for tropical cyclone risk assessment with their individual rating of importance and confidence based on Fig. 10 and Table 7.

mapping. Nevertheless, in the final rating based on importance and

confidence level, storm surge height criteria was rated with very high
level, whereas cyclone frequency, wind speed and precipitation in-
tensity criteria were with moderate level (Figs. 10 and 11). Moreover,
all of the data and methods used for mitigation mapping were assigned
with the highest level confidence. Most of these data were collected
directly from field and validated. Mapping procedures were also well
accepted. However, in the final rating based on importance and con-
fidence level, distance to cyclone shelter and cyclone warning were
assigned with very high level, whereas distance to health infrastructure
and coastal vegetation were with moderate and high level (Figs. 10 and
11).
In addition, the risk assessment results were validated by compar-
ison with the extent and impact of tropical cyclone Sidr (2007). The
satellite image-based impacts assessment map (Fig. 12) indicated that
eastern, south and south western parts of the study area were severely
affected by tropical cyclone Sidr compared to other areas. There was
significant coincidence between cyclone Sidr impacts assessment map
(Fig. 12) and our assessment results.

5. Conclusions

In this paper, we developed a multi-criteria integrated risk mapping

approach for tropical cyclone impacts at the local scale using remote
sensing, field data and spatial analysis. We used Sarankhola Upazila, a
151 km2 local government area in coastal Bangladesh, to test the ap-
plicability of this approach. The risk mapping approach was validated
Fig. 12. The intensity of impact map of tropical cyclone Sidr (2007) produced
using qualitative confidence level assessment and a previous cyclone
from change detection analysis using satellite images before and after the cy-
impact map. The results demonstrated that detailed risk information
clone.
can be effectively mapped using geospatial techniques and an AHP at
the local scale. All the required criteria under each component of risk
were mapped using geospatial techniques. The AHP was found to be of the area in the risk assessment procedure. We found that in-
effective for generating the required multi-criteria decision making corporation of mitigation capacity significantly influenced the risk re-
process in the context of assigning the relative weights. sults in this study. In addition, validation of the results confirms that the
A local scale study with multi-criteria assessment can produce de- current approach can provide accurate and reliable risk information.
tailed and accurate risk information. However, acquiring data at the The findings of this study provide a framework for mapping vulner-
local level and processing it is not easy. This approach has been found ability and exposure, hazard, mitigation capacity and finally risk. The
to be efficient in the processing of multi-criteria at the local level to information generated from these maps can be used by administrators,
generate accurate and detailed risk information. Further, mapping the planners and decision makers to develop effective prevention and mi-
actual risk information requires incorporating the mitigation measures tigation measures to minimise disaster losses.

31
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

The results of this study were associated with a number of limita- hazard for land-use planning in Greater Dhaka of Bangladesh using remote sensing
tions. The study requires lots of criteria for effective risk assessment. and GIS techniques. Water Resources Management, 21, 1601.
Eckert, S., Jelinek, R., Zeug, G., & Krausmann, E. (2012). Remote sensing-based assess-
Acquiring quality data at the local level for each of the criteria is very ment of tsunami vulnerability and risk in Alexandria, Egypt. Applied Geography, 32,
challenging especially in the context of developing countries. Our study 714–723. https://doi.org/10.1016/j.apgeog.2011.08.003.
experienced these challenges. In particular, we used SRTM DEM for FAQs Gde (2012). WetlandInfo Queensland government. http://wetlandinfo.ehp.qld.gov.
au/wetlands/facts-maps/gde-background/gde-faq/, Accessed date: 20 July 2016.
mapping elevation and slope, and as a basic input for the storm surge Frazier, T., Thompson, C., & Dezzani, R. (2013a). In C. A. Brebbia (Ed.). Development of a
height map. High resolution DEM like LiDAR data can be used to pro- spatially explicit vulnerability–resilience model for community level hazard mitigation
vide better results. Tropical cyclone propagation speed was not in- enhancement. Disaster Management and Human Health Risk III' (pp. 13–24). .
Frazier, T. G., Thompson, C. M., Dezzani, R. J., & Butsick, D. (2013b). Spatial and tem-
cluded as a “criteria” in the hazard mapping due to a lack of data. poral quantification of resilience at the community scale. Applied Geography, 42,
Similarly, our study was limited to one form of cyclone wind speed 95–107.
spatial data for preparing the cyclone wind speed map due to data Gallant, J., Read, A., & Dowling, T. (2012). Removal of tree offsets from SRTM and other
digital surface models. International Archives of the Photogrammetry, Remote Sensing
unavailability. Moreover, all of the criteria used in this study were as-
and Spatial Information Sciences, 39, 275–280.
sociated with past and present situations. The criteria like sea level rises Gao, Y., Wang, H., Liu, G. M., Sun, X. Y., Fei, X. Y., et al. (2014). Risk assessment of
can be considered for future climate scenario prediction. In addition, in tropical storm surges for coastal regions of China. Journal of Geophysical Research:
the context of vulnerability assessment and exposure, our study was Atmospheres, 119, 5364–5374.
Glanville, K., Ryan, T., Tomlinson, M., Muriuki, G., Ronan, M., & Pollett, A. (2016). A
dominated by physical vulnerability indicators. Future research should method for catchment scale mapping of groundwater-dependent ecosystems to sup-
be more focused on the above limitations. Despite the above reported port natural resource management (Queensland, Australia). Environmental
drawbacks, the approach we employed is still considered to be effective Management, 57, 432–449.
Hoque, M., Phinn, S., Roelfsema, C., & Childs, I. (2016a). Modelling tropical cyclone
to aggregate the coastal environments in a flexible way and to support hazards under climate change scenario using geospatial techniques. IOP conference
efficient disaster risk reduction measures. Although the selection of Series: Earth and environmental science: Vol. 1, (pp. 1–7). Yogyakarta, Indonesia: IOP
criteria, data type and scale are site specific, this tested and validated Publishing. http://dx.doi.org/10.1088/1755-1315/47/1/012024.
Hoque, M. A.-A., Phinn, S., & Roelfsema, C. (2017a). A systematic review of tropical
approach can be applied to mapping tropical cyclone risk in other cyclone disaster management research using remote sensing and spatial analysis.
coastal environments. Ocean & Coastal Management, 146, 109–120.
Hoque, M. A.-A., Phinn, S., Roelfsema, C., & Childs, I. (2016b). Assessing tropical cyclone
impacts using object-based moderate spatial resolution image analysis: A case study
Acknowledgement in Bangladesh. International Journal of Remote Sensing, 37, 5320–5343. http://dx.doi.
org/10.1080/01431161.2016.1239286.
Funding for this study was supported by the International Hoque, M. A.-A., Phinn, S., Roelfsema, C., & Childs, I. (2017b). Tropical cyclone disaster
management using remote sensing and spatial analysis: A review. International
Postgraduate Research Scholarship (IPRS) and the University of
Journal of Disaster Risk Reduction, 22, 345–354. http://doi.org/10.1016/j.ijdrr.2017.
Queensland (UQ), Australia. Access to SPOT-5 images, field support and 02.008.
equipment were provided by the School of Earth and Environmental Hoque, M. A.-A., Phinn, S., Roelfsema, C., & Childs, I. (2018). Modelling tropical cyclone
Sciences (SEES), UQ. Field assistance was provided by Newton risks for present and future climate change scenarios using geospatial techniques.
International Journal of Digital Earth, 11, 246–263.
Howlader, Anisul Kabir, Abdur Rahim and Mehedi Hasan.WorldView-2 Islam, A. S., Bala, S. K., Hussain, M. A., Hossain, M. A., & Rahman, M. M. (2010).
and IKONOS-2 images were provided by DigitalGlobe Foundation. Performance of coastal structures during cyclone Sidr. Natural Hazards Review, 12,
Access to software and research facilities was provided by the Remote 111–116. http://dx.doi.org/10.1061/(ASCE)NH.1527-6996.0000031.
Islam, T., & Peterson, R. E. (2009). Climatology of landfalling tropical cyclones in
Sensing Research Centre (RSRC), SEES, UQ. The authors would also like Bangladesh 1877–2003. Natural Hazards, 48, 115–135.
to thank the two anonymous reviewers for their constructive comments Khan, M. S. A. (2008). Disaster preparedness for sustainable development in Bangladesh.
and suggestions in terms of improving the quality of this manuscript. Disaster Prevention and Management, 17, 662–671.
Knapp, K. R., Kruk, M. C., Levinson, D. H., Diamond, H. J., & Neumann, C. J. (2010). The
international best track archive for climate stewardship (IBTrACS) unifying tropical
References cyclone data. Bulletin of the American Meteorological Society, 91, 363–376.
Krishnamohan, K., Mohanakumar, K., & Joseph, P. (2014). Climate change in tropical
cyclones and monsoon depressions of North Indian Ocean. In U. Mohanty, M.
Abella, E. C., & Van Westen, C. (2007). Generation of a landslide risk index map for Cuba
Mohapatra, O. Singh, B. Bandyopadhyay, & L. Rathore (Eds.). Monitoring and pre-
using spatial multi-criteria evaluation. Landslides, 4, 311–325. http://dx.doi.org/10.
diction of tropical cyclones in the indian ocean and climate change (pp. 33–39). Springer.
1007/s10346-007-0087-y.
Kumar, A. A., & Kunte, P. D. (2012). Coastal vulnerability assessment for Chennai, east
Birkmann, J. (2007). Risk and vulnerability indicators at different scales: Applicability,
coast of India using geospatial techniques. Natural Hazards, 64, 853–872. http://dx.
usefulness and policy implications. Environmental Hazards, 7, 20–31.
doi.org/10.1007/s11069-012-0276-4.
Blaikie, P., Cannon, T., Davis, I., & Wisner, B. (2014). At risk: Natural hazards, people's
Kunte, P. D., Jauhari, N., Mehrotra, U., Kotha, M., Hursthouse, A. S., & Gagnon, A. S.
vulnerability and disasters. London: Routledge.
(2014). Multi-hazards coastal vulnerability assessment of Goa, India, using geospatial
Burroughs, W. J. (2007). Climate change: A multidisciplinary approach. New York:
techniques. Ocean & Coastal Management, 95, 264–281.
Cambridge University Press.
Lewis, M., Bates, P., Horsburgh, K., Neal, J., & Schumann, G. (2013). A storm surge in-
Chen, H., Ito, Y., Sawamukai, M., & Tokunaga, T. (2015). Flood hazard assessment in the
undation model of the northern Bay of Bengal using publicly available data. The
Kujukuri Plain of Chiba Prefecture, Japan, based on GIS and multicriteria decision
Quarterly Journal of the Royal Meteorological Society, 139, 358–369. http://dx.doi.org/
analysis. Natural Hazards, 78, 105–120. http://dx.doi.org/10.1007/s11069-015-
10.1002/qj.2040.
1699-5.
Li, Y., Ahuja, A., & Padgett, J. E. (2011). Review of methods to assess, design for, and
Choudhury, N. Y., Paul, A., & Paul, B. K. (2004). Impact of coastal embankment on the
mitigate multiple hazards. Journal of Performance of Constructed Facilities, 26,
flash flood in Bangladesh: A case study. Applied Geography, 24, 241–258.
104–117.
Coppola, D. P. (2006). Introduction to international disaster management. Burlington,USA:
Li, K., & Li, G. S. (2013). Risk assessment on storm surges in the coastal area of
Butterworth-Heinemann.
Guangdong Province. Natural Hazards, 68, 1129–1139.
Cutter, S. L. (1996). Vulnerability to environmental hazards. Progress in Human
Mahapatra, M., Ramakrishnan, R., & Rajawat, A. S. (2015). Coastal vulnerability as-
Geography, 20, 529–539. http://dx.doi.org/10.1177/030913259602000407.
sessment using analytical hierarchical process for South Gujarat coast, India. Natural
Cutter, S. L., Boruff, B. J., & Shirley, W. L. (2003). Social vulnerability to environmental
Hazards, 76, 139–159. http://dx.doi.org/10.1007/s11069-014-1491-y.
hazards. Social Science Quarterly, 84, 242–261. http://dx.doi.org/10.1111/1540-
Mahendra, R. S., Mohanty, P. C., Bisoyi, H., Kumar, T. S., & Nayak, S. (2011). Assessment
6237.8402002.
and management of coastal multi-hazard vulnerability along the Cuddalore-
Dasgupta, S., Huq, M., Khan, Z. H., Ahmed, M. M. Z., Mukherjee, N., Khan, M. F., et al.
Villupuram, east coast of India using geospatial techniques. Ocean & Coastal
(2011). Cyclones in a changing climate: The case of Bangladesh. London: Department for
Management, 54, 302–311. http://dx.doi.org/10.1016/j.ocecoaman.2010.12.008.
Environment, Food and Rural Affairs.
Malczewski, J. (1999). GIS and multicriteria decision analysis. John Wiley & Sons.
Deo, A., & Ganer, D. (2014). Tropical cyclone activity over the Indian ocean in the
Mallick, F., & Rahman, A. (2013). Cyclone and tornado risk and reduction approaches in
warmer climate. In U. C. Mohanty (Ed.). Monitoring and prediction of tropical cyclones
Bangladesh. In R. Shaw, F. Mallick, & A. Islam (Eds.). Disaster risk reduction ap-
in the indian ocean and climate change (pp. 72–80). New Delhi: Springer.
proaches in Bangladesh (pp. 91–102). Tokyo: Springer.
Dewan, A. M. (2013a). Hazards, risk, and vulnerability. Floods in a megacity (pp. 35–74).
Martino, L., Ulivieri, C., Jahjah, M., & Loret, E. (2009). Remote sensing and GIS techni-
Springer. http://dx.doi.org/10.1007/978-94-007-5875-9_2.
ques for natural disaster monitoring. In P. Olla (Ed.). Space technologies for the benefit
Dewan, A. M. (2013b). Vulnerability and risk assessment. Floods in a megacity (pp. 139–
of human society and earth (pp. 331–382). Springer.
177). Dordrecht: Springer.
Masuya, A., Dewan, A., & Corner, R. J. (2015). Population evacuation: Evaluating spatial
Dewan, A. M., Islam, M. M., Kumamoto, T., & Nishigaki, M. (2007). Evaluating flood
distribution of flood shelters and vulnerable residential units in dhaka with

32
M.A.-A. Hoque et al. Applied Geography 98 (2018) 22–33

geographic information systems. Natural Hazards, 78, 1859–1882. 0_2.

Mendelsohn, R., Emanuel, K., Chonabayashi, S., & Bakkensen, L. (2012). The impact of Roy, D. C., & Blaschke, T. (2013). Spatial vulnerability assessment of floods in the coastal
climate change on global tropical cyclone damage. Nature Climate Change, 2, regions of Bangladesh. Geomatics, Natural Hazards and Risk, 1–24.
205–209. Saaty, T. L. (2008). Decision making with the analytic hierarchy process. International
Nadiruzzaman, M., & Paul, B. K. (2013). Post-sidr public housing assistance in Journal of Services Sciences, 1, 83–98.
Bangladesh: A case study. Environmental Hazards, 12, 166–179. Saxena, S., Purvaja, R., Suganya, G. M. D., & Ramesh, R. (2013). Coastal hazard mapping
Panahi, M., Rezaie, F., & Meshkani, S. (2013). Seismic vulnerability assessment of school in the Cuddalore region, South India. Natural Hazards, 66, 1519–1536. http://dx.doi.
buildings in Tehran city based on AHP and GIS. Natural Hazards and Earth System org/10.1007/s11069-012-0362-7.
Sciences, 1, 4511–4538. Shultz, J. M., Russell, J., & Espinel, Z. (2005). Epidemiology of tropical cyclones: The
Paul, B. (2009). Why relatively fewer people died? The case of Bangladesh's Cyclone Sidr. dynamics of disaster, disease, and development. Epidemiologic Reviews, 27, 21–35.
Natural Hazards, 50, 289–304. http://dx.doi.org/10.1007/s11069-008-9340-5. Taramelli, A., Melelli, L., Pasqui, M., & Sorichetta, A. (2008). Estimating hurricane ha-
Paul, A., & Rahman, M. (2006). Cyclone mitigation perspectives in the islands of zards using a GIS system. Natural Hazards and Earth System Sciences, 8, 839–854.
Bangladesh: A case of Sandwip and Hatia islands. Coastal Management, 34, 199–215. Turner, B. L., Kasperson, Roger E., Matson, Pamela A., McCarthy, James J., Corell, Robert
Peduzzi, P., Chatenoux, B., Dao, H., De Bono, A., Herold, C., Kossin, J., et al. (2012). W., Christensen, Lindsey, et al. (2003). A framework for vulnerability analysis in
Global trends in tropical cyclone risk. Nature Climate Change, 2, 289–294. sustainability science. Proceedings of the National Academy of Sciences, 100,
Pielke, R. A., Jr., Gratz, J., Landsea, C. W., Collins, D., Saunders, M. A., & Musulin, R. 8074–8079.
(2008). Normalized hurricane damage in the United States: 1900–2005. Natural Wang, Y., Li, Z., Tang, Z., & Zeng, G. (2011). A GIS-based spatial multi-criteria approach
Hazards Review, 9, 29–42. for flood risk assessment in the Dongting Lake Region, Hunan, Central China. Water
Puotinen, M. (2007). Modelling the risk of cyclone wave damage to coral reefs using GIS: Resources Management, 25, 3465–3484.
A case study of the great Barrier Reef, 1969–2003. International Journal of Weinkle, J., Maue, R., & Pielke, R., Jr. (2012). Historical global tropical cyclone landfalls.
Geographical Information Science, 21, 97–120. Journal of Climate, 25, 4729–4735.
Rafiq, L., Blaschke, T., & Zeil, P. (2010). Application of satellite derived information for Yin, J., Xu, S., Wang, J., Zhong, H., Hu, Y., Yin, Z., et al. (2010). Vulnerability assessment
disaster risk reduction; vulnerability assessment for SW-coast of Pakistan. In U. of combined impacts of sea level rise and coastal flooding for China's coastal region
Michel, & D. L. Civco (Eds.). Earth resources and environmental remote sensing-gis ap- using remote sensing and GIS. Geoinformatics, 2010 18th international conference on
plications, vol 7831. Proceedings of SPIE-the international society for optical en- (pp. 1–4). IEEE.
gineeringhttp://dx.doi.org/10.1117/12.864858. Yin, J., Yin, Z., & Xu, S. (2013). Composite risk assessment of typhoon-induced disaster
Rashid, A. K. M. M. (2013). Understanding vulnerability and risks. In R. Shaw, F. Mallick, for China's coastal area. Natural Hazards, 69, 1423–1434. http://dx.doi.org/10.1007/
& A. Islam (Eds.). Disaster risk reduction approaches in Bangladesh. Disaster risk re- s11069-013-0755-2.
duction (pp. 23–43). Springer Japan. http://dx.doi.org/10.1007/978-4-431-54252-

33