Bangladesh's vulnerability to cyclonic coastal flooding — Source link

. In the Ganges-Brahmaputra-Meghna delta, covering most of Bangladesh, more than 165 million people live in low-lying coasts facing major extreme climatic events, such as cyclones. This article reviews the current scientific literature publications (2007-2020) in order to define vulnerability in the context of coastal Bangladesh facing cyclonic flooding. Based on this review, a new metric, called ‘socio - spatial vulnerability index’ (SSVI), is defined as function of both the probability of the cyclonic flood hazard and the sensitivity of delta 15 inhabitants. The main result shows that the districts of Shariatpur, Chandpur and Barisal situated in the tidal floodplain of the Ganges-Brahmaputra-Meghna delta, are in the fourth quartile, i.e. highest category. These districts are very densely populated (from 870 up to 1400 inhabitants/km2) and exposed to inundation hazards, and host a large number of vulnerability factors. Finally, t he delta’s mouth was identified as a very vulnerable area to cyclonic flooding. to frequent cyclone flooding. This study explores the potential for cyclone flood SSVI to highlight socio-spatial inequalities between the coastal districts in the case of Bangladesh, this country being considered as one of the most vulnerable in the world regarding extreme climatic events. In this region, 165 million people live on the low-lying land of the Ganges-Brahmaputra-Meghna (GBM) delta and face inter alia to catastrophic storm surge floods caused by cyclone (Ali, 1999; Becker et al., 2020; Nicholls et al., 2018; Uddin et al., 2019). Like mentioned above, several studies have 70 already been carried out to assess the cyclone flood vulnerability of Bangladesh coastal areas. However, as mentioned by Hoque et al. (2019), only two studies have proposed the production of a multi-parameter index as a tool to characterize the physical vulnerability in the central and western part of the GBM delta (Islam et al., 2016; Islam et al., 2015). Regarding the work of Mahmood et al., (2020), they proposed a cyclone flood vulnerability assessment located on the mouth area of the GBM delta and based on the coastal vulnerability index developed by 75 Hoque et al. (2019). Only physical parameters such as topography, shoreline change, coastal slope (among others) were taken into account. Therefore, here also the social dimension of cyclone flood vulnerability was not assessed. The paper is organized as follows: section 2 gives an overview of current paradigms used in vulnerability conceptualization; section 3 presents the region study; section 4 provides insights from a literature review into 80 as in the Chandpur Sadar upazila at the mouth of the GBM delta. In Khulna, the height is lower but still significant when crossed with the high population density. Shatkhira and Khulna, west coastal districts, population density is lower but as the people of this area is highly nature resources-dependent, the flooding of Sundarbans that groups or communities are equiprobably exposed to the hazard in a given spatiality. C onsequently, we define a ‘socio - spatial vulnerability index’ (SSVI), as function of both the probability of the cyclonic flood hazard and the sensitivity of delta inhabitants. Through this index, we quantify and map levels of this vulnerability. SSVI is also a new methodology to highlight socio-spatial 540 inequalities between the coastal districts of Bangladesh, by considering social, physical and infrastructural aspects of vulnerability. The concept is thought in a systemic and interdisciplinary framework, considering this vulnerability aspects from social and natural sciences and their interactions between both. With this index and through its different components, we can better reveal why some groups or communities, at the districts level in case of our study, are more vulnerable to cyclonic flooding, and could be also slower to recover than others. At final, we identify places the most vulnerable that are understudied, opening to consider differently the coastal area for State, NGOs and OIs to prioritize their Preparedness and Recovery programs. Climate change could increase the intensity and frequency of tropical cyclones in the Bay of Bengal, therefore access to appropriate cyclonic flood hazard, from hydrological modelling efficient tools, is a key parameter to predict the future socio-spatial vulnerability patterns.


Introduction
Tropical cyclones are among the most devastating extreme natural events. When a cyclone approaches coast, it can induce an increase of sea level, called storm surge. This storm surge can cause severe flooding of vast coastal low-lying areas and leads to profound societal and economic impacts. This is especially the case for the regions surrounded by the semi-enclosed sea of the Bay of Bengal. Although the frequency of the cyclone is weaker than 25 in Atlantic and Pacific Oceans, most of them make landfall (Balaguru et al., 2014). That, combined with low-lying and high-populated areas, causes catastrophic impacts (Hoque et al., 2019). In April 1991, for example, in Bangladesh 15 million people were affected by the flooding that followed the cyclone Gorky landfall, causing about 140'000 fatalities and 1.8 billion of dollars' worth of damage (source International Emergency Events how vulnerability to cyclone flooding in Bangladesh is questioning and according to what indicators; and section 5 proposes a socio-spatial index for standardizing vulnerability to cyclone flooding in Bangladesh. Finally, two sections are devoted to the discussion and conclusion.

Overview about the vulnerability concept and its indicators
In the following, we provide a necessary overview of different vulnerability paradigms in order to understand how 85 they were defined in these past decades, and how its conceptualization can be achieved. Following the authors, it is seen either as static or dynamic, including exposure or not. Chambers (1989) initiated a formal definition as "exposure to contingencies and stresses and the difficulty which some communities experience while coping with such contingencies and stresses". Hence, he identified two binaries of vulnerability: an external (the exposure to external shocks) and an internal vulnerability (the incapacity to cope). Lastly, he suggested that it is determined 90 by the management and accessibility of assets. Here, exposure is included but vulnerability is mainly considered as a state (Roberts et al., 2009).
The pressure and release model proposed by (Blaikie, 1994) considers that vulnerability is socially produced (Menoni et al., 2012). It is defined as: "the characteristics of a person or group in terms of their capacity to 95 anticipate, cope with, resist, and recover from the impact of a natural hazard''. It focuses on both the social and temporal dimensions of a disaster (Kuhlicke et al., 2011) and proposes that vulnerability can be reduced through people's access to resources (Menoni et al., 2012).
Further developed by (Wisner et al., 2004), the model includes a sense of progression into the vulnerability, which 100 gradually evolves according to: root causes (the unequal distribution of power and assets), dynamic pressures (the society characteristics, such as population density and growth) and unsafe conditions (situations evolving as its expression when the hazard occurs). This model conceptualizes it as a process, but it underestimates the exposure in the risk equation (Roberts et al., 2009). Vulnerability as a dynamic process was additionally developed by Birkmann, (2006) in the BBC model which incorporates coping capacity into the vulnerability of the risk equation. 105 It also distinguishes between its social, economic and environmental components (Romieu et al., 2010). However, his model does not integrate physical vulnerability as impacting the social aspects. In most of these models, it is emphasized how it is a dynamic concept that does not only include socio-demographic characteristics.
Vulnerability is also the results of human interactions and networks and it may change across time. De Marchi and 110 Scolobig (2012) argue that the vulnerability of social systems is derived from characteristics of individuals, groups, but also from institutions. For those reasons, later authors offered to characterize its various components according to its associated people and threatened assets. In addition, these authors adopt a systemic perspective in which they distinguish between the individual and the institutional components of social vulnerability in order to explore their interactions. Individual vulnerability stresses the perceptions and attitudes impeding people from preparing and 115 coping sufficiently in case of disaster. However, institutional vulnerability implies that agencies and services may paradoxically weaken the overall capacity of a specific system to deal effectively with disasters (De Marchi and Scolobig, 2012). For example, the authors demonstrate how the presence of structural devices (barriers, dams and embankments) can lessen the perception of risk and mislead to a "symbol of complete safety". This institutional https://doi.org /10.5194/nhess-2021-8 Preprint. Discussion started: 9 February 2021 c Author(s) 2021. CC BY 4.0 License. vulnerability leads to a situation where the bigger the defense system is, the greater the impression of 120 invulnerability increases. This effect is well known in the literature as "safety paradox" (Burby, 2006;Ferdous et al., 2020): increasing levels of flood protection can generate a sense of complacency among the protected people, which can reduce preparedness, thereby increasing vulnerability. Therefore, to assess vulnerability level, classical indicators are often social inequality indicators, such as : age, 125 gender, household composition, employment, income, educational level, ethnicity, race (Blaikie, 1994;IPCC AR5, 2014). Other socio-economic indicators are added by the United Nations Organisation and the World Bank, such as: gross domestic product (GDP) per inhabitant, human poverty index (HPI), inflation rate, population characteristics (density, growth, life expectancy at birth, illiteracy rate). These classical indicators of vulnerability may lack explanatory power and so this type of parameters using quantitative data should be discussed and put 130 into perspective with qualitative indicators (Kuhlicke et al., 2011). Indicators relying only on census data lack its full understanding because they do not take into account people's perceptions and culture of risk, norms, traditions, beliefs and behaviors. Its other social dimensions, like "the ability of individuals and communities to cope seems to depend also on some intangible aspects, which can be described by terms such as energy, vigor, vitality, strength, courage, nerve, fortitude, and their antonyms" (De Marchi and Scolobig, 2012). Those aspects shape people's 135 vulnerability, together with personal, contextual and material conditions. As mentioned above, according to disciplinary approaches, the physical and social dimensions must be studied together and fit within a territory. This spatial dimension then appears unavoidable to assess vulnerability (Mazumdar and Paul, 2018). Its spatial dimension enables, from a local to a national level, to define specific 140 adaptation or mitigation strategies for each region, through the identification of hot spots, and to use the map as means of communication to meet the decision-makers' needs. As mentioned by Mazumdar and Paul (2018), vulnerability is specific according to population, factors and locations. This is why we have chosen to assess the vulnerability to cyclone flooding through a socio-spatial vulnerability index (SSVI), a tool that enables both quantifying and mapping its levels. Socio-spatial vulnerability takes into account social conditions and measures 145 the resistance or resilience of populations to a hazard; at the same time, it integrates the potential exposures of these same populations in particular territories England and Knox, 2016;Forrest et al., 2020;O'Hare and White, 2018). Thus, we consider the vulnerability in a structural perspective (Hufschmidt, 2011) and where "exposure", i.e. the spatial intersection of flood hazard and population density, is included as one of its components. 150

Region study
Bangladesh is situated between India, Myanmar and the Bay of Bengal. This country is one of the most vulnerable countries in the world regarding extreme climatic events. The entirety of coastal Bangladesh is under 10 meters' elevation (Becker et al., 2019) and is highly populated with a mean of 1'240 people per km 2 in 2018 (World Bank, 2018), which increases its exposure to coastal floods. Though people of the deltas have learnt how to live with 155 periodic coastal submersion, they face numerous constraints like ground instability because of fluvial alluvium and subsidence, insalubrity of swamps, and ground salinity, which compelled the population to adapt its cultural activities like planting new rice varieties (Ali, 1999) or shifting from rice cultivation to shrimp farming. https://doi.org /10.5194/nhess-2021-8 Preprint. Discussion started: 9 February 2021 c Author(s) 2021. CC BY 4.0 License.
Since 1960s, Bangladesh has a widespread system of embankments and polders (land surrounded by 160 embankments), which initially improve agriculture and water level management (Brown et al., 2018). However, as these earthen embankments systems were built for tidal flooding -without considering storm surgesthey were not able to protect from extreme case of cyclonic flooding. Besides, due to a lack of maintenance, most of coastal embankments nowadays are partially or completely eroded . In 1972, a Cyclone Preparedness Program (CPP) was created by the Bangladesh's government to prepare weather forecasts and 165 disaster warnings , whereby cyclone shelters were erected as multi-storey buildingsfor example, in the form of "School cum Cyclone Shelter", to promote education as well as serve as shelter during storm-surges. Recent progress in hydrodynamic modeling for the GBM delta showed that to accurately reproduce patterns of cyclonic coastal flooding, several parameters are essential as accurate bathymetry and topography, inclusions of dikes, well-tuned bottom roughness, and forcing the model with accurate representation of the cyclonic wind and 180 pressure fields (Khan et al., 2020;Krien et al., 2017). Using this robust hydrodynamical modelling schema, Khan et al. (2019) provided probabilistic cyclonic flood hazard maps over coastal GBM from a dataset of statistically and physically consistent synthetic cyclone events (Emanuel et al., 2006). We presented in Figure 1.a the cyclonic flooding level with a 50-year return-period estimated from Khan et al. (2019). This means that at a particular location in the GBM, the reported flooding height, due to cyclonic flooding, is expected to be observed on average 185 once in 50 years. This type of metric aims to anticipate the cyclone flood occurrence and its consequences in order to alert the inhabitants and put in place effective strategies. In dark blue on Figure 1.a, one observes the places which are exposed to flooding with the most extreme heights, up to eight meters from the mean sea level at the junction between the Ganges and the Meghna rivers. All the coastal islands areas are exposed to a 1-in-50-year flood, though the biggest ones, such as Hatiya, Sandwip, and Kutubdia islands are not totally exposed to the 190 cyclonic inundation on their total area. For these large coastal islands, the presence of dykes makes a very large gradient in inundation hazard. Along almost all the coastline, blue colors are darker -excepting the South of Laxmipur district -meaning that the embankments and their polders behind are strongly exposed to the inundation.
This directly threatens the houses settled by the dykes as authors underlined .

195
In addition to lots of places exposed to flooding with the most extreme heights, the population is aggregated around the main cities of the South: Khulna, Barisal, Chittagong and Cox's Bazar, as well as on big islands such as Sandwip and Kutubdia island (Fig. 1b). While the area in the southwestern coast, behind the Sundarbans forest is not very populated, compared with other parts of the delta, the districts of the central and east coasts are much https://doi.org/10.5194/nhess-2021-8 Preprint. Discussion started: 9 February 2021 c Author(s) 2021. CC BY 4.0 License. denser, such as in the Chandpur Sadar upazila (sub-district) at the mouth of the GBM delta. In Khulna, the water 200 height is lower but still significant when crossed with the high population density. In Shatkhira and Khulna, the west coastal districts, population density is lower but as the people of this area is highly nature resourcesdependent, the flooding of the Sundarbans increases the vulnerability of the people in the area.  (IPCC) set the significance of vulnerability social factors over the international research community. This review excludes non-peer-reviewed articles, in the aim to obtain a state of the art's review of the current knowledge extracted from scientific literature only, to make sure of high scientific quality standards in the selected articles 220 (Räsänen et al., 2016). Finally, only English-written literature was included in order to keep references that can be widely read, to enable reproduction, and to deepen this review. The search engines used for the review are Google Scholar, ISTEX, Science Direct, Scopus and Web of Science. The following search sequence used in the electronic databases includes the hazard, the area of study and key words related to vulnerability: ("coastal flood*" OR "sealevel rise" OR "Storm surge" OR "Cyclonic storm" OR "Disaster risk reduction") AND ("Bangladesh" OR 225 "Brahmaputra delta") AND ("*vulnerability" OR "human exposure" OR "coping*"). Those terms were searched for in the titles, abstracts and keywords of the articles (see Appendix A1 for more details on the strategy used).
From this research, 49 publications were kept for this review (see Appendix A2-A4). b.

Summary of the literature analysis
The first result shows the universities behind the analyzed articles: 85% of the studies were conducted by Western 230 or Asian scholars. Only seven articles were directed by Bengali nationals, and the rest is made from a cooperation with Bengali. Amongst the authors' disciplines, the applied sciences are overrepresented with 26 articles (especially in the field of disaster risk management), followed by the humanities and social sciences and, finally, the natural sciences.

235
The methodologies used in half of the studies are mixed methods, using both quantitative (e.g., surveys) and qualitative data (focus group discussion and key informant interviews, for example). Inductive approach (bottomup) was used in three quarters of the articles, explaining that most of the studies were done at a small scalevillages and unionsto study some aspects of population vulnerability. While still more than half of the studies used surveys (structured questionnaires), five articles used the Participatory Rural Appraisal (PRA) method to 240 allow the participant to fully participate even if they were illiterate (Ahmed et al., 2010;.
Most of the authors touch upon vulnerability in their articles, but surprisingly, not all give a definition of it, often taking the concept for granted. Among the studies defining the theory upon which they grounded their research, 245 three articles define what they consider is vulnerability. First, for , "vulnerability to disasters refers to the inability of a society and its people to withstand adverse impacts from multiple stresses to which they are exposed. It has two sides: an external side of risks, shocks, and stress to which an individual or household is subject; and an internal side which is defenselessness, meaning a lack of means to cope". Second, Quader et al.
(2017) mention "three main characteristic groups of social vulnerabilities (dimensions), namely the socio-250 demographic and economic characteristics of the population, their access to basic facilities and the proportion of physically or ethnically marginalized people". Then, Ishtiaque et al., (2019) define vulnerability "as the degree or extent to which a system is likely to be exposed and sensitive to a hazard, and the capacity of that system to adapt to the effects of climate impacts". However, those definitions do not have a lot in common but are complementary.
Therefore, those results show how little consensus exists on this concept in this region. 255 Finally, this literature review enables us to outline below the vulnerability factors' main characteristics discussed in the articles (physical and systemic infrastructures, social and economic aspects). Some authors agree that in the coastal zone globally, but especially in Bangladesh, it is imperative to consider both physical and socio-economic factors in defining vulnerability . According to the authors, the lack of proper road communication networks and transport infrastructure is an important factor of people's vulnerability to cyclone disasters. In the coastal area, roads are not uniformly 265 distributed , most of them are narrow, unpaved and made of dirt tracks, since they often mark boundaries for crop fields , and embankments are also used as roads. During the rainy https://doi.org/10.5194/nhess-2021-8 Preprint. Discussion started: 9 February 2021 c Author(s) 2021. CC BY 4.0 License. season, they are destroyed and the mud makes any physical mobility difficult . The lack of road infrastructure is particularly important when authorities should disseminate the early warnings before a cyclone ), but they are mainly inaccessible to motorized vehicles. When the hazards are imminent, 270 roads in good shape are needed for the population to travel to reach the protective shelters .
Finally, after the cyclone, developed road communication networks enable effective evacuation and rescue during post-disaster relief operations (Hossain, 2015). Hence, it is concluded by authors that remote areas rarely get emergency assistance .

275
Situation of the shelters are studied by many authors, showing how the long distances to reach them, the lack of infrastructure facilities, and their deterioration is an issue to the nearby households when they have to decide to find shelter or not. Proximity, poor hygienic condition , lack in communication means and delivery of sanitation and drinking water  are some example for additional reluctant aspects for many of the families. People are also reluctant to take refuge in shelters because they are suspicious of cyclone alerts, 280 but also because they are afraid of having all their belongings stolen. There is also a memory loss of the past events impacts (Ishtiaque et al., 2019). If each shelter can accommodate approximately 1000 people , they are often insufficient to host the dense population living in the area, not uniformly distributed on the territory or are also too far from the houses in need Hossain, 2015;). The distance to a shelter may considerably increase the vulnerability of the people if they decide to walk to one . 285 The embankments are often in very bad maintenance, including the sluice gates .
Embankment failure can also induce problem with drainage congestion and water logging .
Those coastal walls are also breached on purpose to be penetrated by pipes that fill the basins of shrimp farms (Hossain, 2015). 290 It is worth noting that in the small island polders there is no electricity and, in remote areas, it is harder to find hospitals or clinics as well as available doctors, because they are located mainly in big towns and cities . Those aspects definitely decrease people's coping capacities.

Houses: Material and emplacement 295
Houses' materials and emplacement are the main criteria of the household's vulnerability. Four main kinds of dwelling houses may be described in coastal Bangladesh. The pucca houses are the most solid and permanent houses, made of brick and concrete, they are very rare and held by wealthy households. The semi-pucca buildings are made of semi-permanent material in the sense that they have a floor of mud, brick or cement, but their walls are bamboo mats or timber and the roof of corrugated iron (CI) sheets. The kutcha houses have floors made of 300 mud; the walls are bamboo, sticks or straw; and the roof made of paddy (rice) or wheat straw . Finally, the jhupri houses are mainly huts or made of CI sheets. Semi-pucca houses are not cyclone resilient, but kutcha and jhupri houses are organic-made structures totally vulnerable to this hazard . In the majority of the rural studies analyzed in the literature review, most of the houses are kutcha houses . 305 Houses' material fragility is accentuated by their settlement place: the more the villages are isolated and scattered, the more they are vulnerable. Moreover, there is a dreadful situation when houses are settled along the embankments: they face huge casualties and damages with strong winds and waves, or in case of any breach of dykes (Hossain, 2015). This also happens 310 for people living out of the dykes: when their living places remain flooded in the days and weeks following a cyclonic inundation, people take shelter on embankments, building tent-like huts . The main reason explaining people's choice to install on embankments is because they are the major public spaces owned by the government along the coast which have become the last area above sea level not privately owned where displaced people can construct settlements . Many authors highlight that after cyclonic inundations, sanitation facilities are particularly challenged. The first main problem is the drinking water supply, which is disrupted and becomes scarce . Polluted by waste or animal carcasses (Hossain, 2015), as well as human feces due to the lack of 325 sanitation , the water is often contaminated by water-borne diseases like diarrhea, or even cholera and typhoid .
Drinking water is also contaminated by saltwater intrusion, which is a serious problem for most the families that use pond springs Ishtiaque et al., 2019;. Besides failures in access to 330 sanitation and clean water, households are significantly disrupted, even times no electricity after a cyclone , which can be an issue for communication purposes and economic recovery. It is worth noting that in the small island polders there is no electricity at the first place.

A significant indicator influencing people's vulnerability is education. It was shown that less educated people 335
were less able to understand the disaster forecast and to foresee the appropriate materials and food to store in case of evacuation . The analysis of the literature review establishes that five main social groups are particularly vulnerable during cyclonic inundations: women, children, elderly, disabled people and religious minorities Rabby et al., 2019). Firstly, in customary social order in Bangladeshi rural areas, men are usually the heads of household, being the ones who decide how to prepare the 340 family for a cyclone, where to take shelter and if to eventually evacuate from the house (Ahmed et al., 2010).
Traditionally, women take care of family resources and have to wait for their husbands the consent to go to a protective shelter . During the cyclone and post-disaster, large families (more than 10 members) are considered as more vulnerable, in the sense that elderly, children and people with disabilities are considered to be dependent people, for the evacuation as well as in the recovering phase 345 (Hossain, 2015;. Generally, women have less access to incomes and assets because social https://doi.org/10.5194/nhess-2021-8 Preprint. Discussion started: 9 February 2021 c Author(s) 2021. CC BY 4.0 License. norms limit their advocacy in the public sphere, although they can get access to microcredits offered by NGOs to support their recovery. Nonetheless, authors pointed out that financial relief is often distributed to particular religious, political and social groups, neglecting women, poor and religious minorities, a situation that mainly gives advantage to the supreme social class ; this social class, in rural zone, corresponds to 350 the wealthy persons who regularly become elected representatives in the local government .
However, some NGOs emphasize that policies aimed at improving infrastructure to reduce vulnerability must be accompanied by a policy of sensitizing populations to the disaster in order to reduce illiteracy, poverty and increase livelihoods (Ishtiaque et al., 2019).

355
Beyond these social indicators, other cultural and/or political variables are mentioned in the literature. For example, despite the Cyclone Preparedness Program, many critics were found in the literature about the early warning system's effectiveness, either because of its dissemination or its understanding. The first problem in the preparedness phase is the population's access to the alert, since a large part does not have access to a radio, the communication dissemination remains poor Paul and Routray, 360 2011;. Moreover, several researchers noticed the surprise of the population at the arrival of the cyclone disasters because they failed to understand the warning systems .
Other authors also report people are indifferent to the weather forecast because they use their own traditional forecast methods through indigenous signs (Hossain, 2015;. Therefore, signs in the environment (wind directions, clouds, sea birds and insects' behaviors) are used instead to perceive impending hazards (Garai, 365 2017). When the information of the early warning has been effectively transmitted to the people and they are aware of the imminent hazard, many scientists underline the population's resignation that a cyclone is God's will. They regard disasters as "common events"  since they happen frequently. In case of total flooding and destruction, people may end up having to move to another place. Nonetheless, authors observed that if people can temporarily move out of their village, they do not consider permanent relocation as alternatives (Ahmed et al., 370 2016). Some authors argued that the polders "create a false sense of security" (Sultana, 2010) and therefore increases people's vulnerability (Ishtiaque et al., 2017). This feeling of security can also diminish the efficiency of a cyclone warning system. In addition, authors note that memory loss from previous extreme events can be observed (Isthiaque et al., 2019).

Economic factors 375
Some labor categories, tied to the coastal economic development, are more vulnerable to cyclones, namely rice farmers, fishermen, and shrimp farmers. Cyclones particularly affect cultivated lands and fishing facilities, often leading to extreme poverty for the people who cannot practice their job anymore . Shrimp farmers are especially concerned with this problem, where a quarter of them suffered more damages than agricultural farmers after cyclone Aila in 2019 (Ishtiaque et al., 2017). The expansion of shrimp farms in 380 Bangladesh, especially in the West and East coasts have raised a certain number of issues both for the people and for the environment. This culture has been developed mainly in the three last decades because of its great potential of exportation, bringing back value from foreign currency, until it became one of the most important sectors in the country especially concentrated in the coastal zones of Khulna, Satkhira, Bagherhat and Cox's Bazar . 385 https://doi.org /10.5194/nhess-2021-8 Preprint. Discussion started: 9 February 2021 c Author(s) 2021. CC BY 4.0 License.
Finally, to conclude this diagnosis, let us stress that according to , the variables of population density, poverty and geomorphology (coastal and flat/low-lying country face to sea level rise) will be the most aggravating factors of future vulnerability. This literature review allows identifying main characteristics to assess the vulnerability. While from a theoretical point of view, it can be defined on the basis of all these physical, socio-390 cultural and economic indicators, its spatial component is still not addressed and the interrelation between a population and its territory of life is not analyzed or taken into account here. Yet, the spatial distribution of its different components (physical, socio-cultural and economic) enables defining its level for a territory. This is why the index approach was chosen because it allows synthesizing all this information into a single variable and mapping it. 395

A socio-spatial vulnerability index
To identify places that are highly vulnerable to cyclonic flooding hazard, we focus on 16 coastal districts, the district being the spatial unit of reference (Fig. 1a). There are two popular approaches to construct vulnerability index: the variable reduction and variable addition. The first one is an inductive approach, a large set of variables is used, assuming that they have potentially more or less influence (i.e. weight) on the index calculation. Principal 400 Components Analysis, although fairly complex, is often used to estimate these weights to assign to each indicator . On the contrary, the variable addition is a deductive approach, where only variables have been determined as influential in previous studies are added (with equal weights). We chose to apply this simple method, from the vulnerability factors highlighted by our literature review, to determine the SSVI. 405

Methodology
From the literature review (section 4) three components based on local experts' knowledge are highlighted: physical and infrastructures, socio-cultural and economic components. Like explained above, exposure is included in the vulnerability, a fourth component is added, cyclone protection and exposure.

410
The calculation of SSVI is based on the methodology developed by Flanagan et al. (2011). To construct it, each of the variables selected was ranked from highest to lowest vulnerability degree. A percentile rank was then calculated, by using the formula Percentile Rank = (Rank-1)/(N-1) where N is the total number of data points, for each district over each of these factors. In addition, a percentile rank is calculated for each of the four components or domains (adapted from Flanagan et al. (2011)), defined as (1)  The interpretation of the SSVI is as follows: for example, a district being in the 85th percentile (0.85 ranking), it means that 85% of these districts are either below, or equal, to that particular district regards to SSVI. 420

Data and sources
From the literature review, 17 variables were identified to assess the level of vulnerability of each coastal district (Fig. 2). The representativeness of each variables, defining the four components, is characterize according to the To describe the Socio-economic status, two variables are used: poverty rate which corresponds to percentage of total of poor comprising both lower and upper the poverty line; and education rate which corresponds to percentage of children who have more than 10 years of school Islam et al., 2014;Isthaique et al., 2019). 435 The Household composition and disability domain is characterized by 5 variables Ishtiaque et al., 2019;Rabby et al., 2019): Age 14 or younger (%) and age 60 or older (% ) (ages 14 and 60 are chosen as limits according to ) correspond to addition of categories between ages 0-14 years and categories above 60 years, respectively and expressed in percentage, female rate (%), disability represents the 440 percentage of disable people per district and religion, described by percentage of Muslim and percentage of religious minorities included Buddhist, Hindu, Christian and other Ishtiaque et al., 2019;Rabby et al., 2019;Roy and Blaschke, 2015).
The third domain, called Housing and infrastructures, includes percentage of houses made of organic materials 445 (kutcha and jhupri that are structures totally vulnerable to coastal flooding ), unpaved road corresponds to percentage of length of road that are not paved, access to electricity, mobile and sanitary (toilet with water and sealed) represent the percentage of population which have access to these infrastructures, unsafe drinking water represents percentage of population which have not access to safe drinking water (unsafe drinking water corresponds to all spring water that are not coming from the tap or from tube wells) and then 450 number of hospitals per district Ishtiaque et al., 2019;.
The last domain is Cyclone protection and exposure and is described by three variables: dykes and embankments Hossain, 2015;Rabby et al., 2019) which corresponds to the length of dykes and embankments 455 into each district, shelter capacity computed by the difference between the shelter capacity (number of people) and the number of people present within a one-kilometer radius of the shelter; and then, exposure corresponding to percentage of people per district concerned by the cyclonic inundation with a 50-year return-period (Fig. 1). As we have seen above (section 3), this cyclonic inundation probability, resulting from hydrodynamical modeling, integrates components as topography, bathymetry, slope, dikes, embankments, soil roughness and vegetation, 460 which allows to indirectly integrate the physical vulnerability in the SSVI. Although authors proposed population density as a variable to characterize the vulnerability Ishtiaque et al., 2019;, none used the population density actually affected by cyclonic floods.   Figure 3 presents the SSVI. We deduced that the inhabitants of the districts at the mouth of the GBM rivers, i.e.

Socio-spatial vulnerability index to cyclonic flooding mapping
Chandpur, Shariatpur and Barisal, and Bagerhat in the west coast, where the SSVI is in the highest category, are 470 more vulnerable to cyclonic flooding relative to other districts. It appears clearly that these districts are very densely populated, poor, with a low percentage of high school completion and extremely exposed to the inundation hazard with insufficient cyclone protections. However, this analyze does not mean that the population of other districts are not vulnerable, but only less vulnerable.  For the districts that appear the least vulnerable according to the index (Fig. 3), the Cyclone protection and exposure domain is the one that appears the most critical for these districts, as example, Feni, Khulna, Pirojpur 480 and Patuakhali (Fig. 4)

Main findings and contributions
Although, considerable effort has been made in Bangladesh, as shown in the literature review (section 4), to 490 understand patterns of social, economic and environmental variables that render inhabitants more or less vulnerable to cyclonic flood, much less effort has been spent to consider the socio-spatial vulnerability. This study provides a methodology and an integrated index, called 'socio-spatial vulnerability index' (SSVI), for assessing the ways in which socio-spatial vulnerability to cyclonic flooding is distributed across the coastal territory. This new metric is a function of the probability of the cyclone flood hazard and the sensitivity of inhabitants. 495 The SSVI shows that an interdisciplinary approach is needed to take into account the full complexity of the definition of vulnerability. The exposure of the population to the risk of cyclonic flooding must be integrated into vulnerability assessment, which distinguishes our approach from research dealing only with coastal vulnerability or social vulnerability (Hoque et al., 2019;Rabby et al., 2019). We showed that there are uneven geographical distributions in cyclonic flood vulnerability in Bangladesh; and that these different geographical patterns seem 500 linked to distinct types of physical and systemic vulnerability. The SSVI map can help to show where there are higher concentrations of sensitive people to cyclonic flood and therefore where actions and which actors are best able to deliver efficient support to mitigate the disaster level that will inevitably accompany cyclone landfall.
However, we saw also that vulnerability for people nature-dependent to the Sundarbans Forest was underestimated. Moreover, in this prawn farming region, it is amplified by the illegal or forced occupation of land 505 often lead to violence, and by intensive, potentially hazardous, and poorly paid working conditions, especially for women (Ito, 2002;. Nonetheless, specific populations as the one living in slums or Rohingya refugees are under-represented in our study Swapan et al., 2020). In the same way, it should be noted that there are a lot of new proxies of social, environmental and economic dataset (social network, mobile phone, satellites) that would allow estimation of the SSVI over finer geographical resolution to better represent 510 the cyclone flooding vulnerability. https://doi.org/10.5194/nhess-2021-8 Preprint. Discussion started: 9 February 2021 c Author(s) 2021. CC BY 4.0 License.

Limits of the research
A significant limitation to this work is the approach intrinsically used in this literature review: the exclusion criteria left out the analysis of grey literature to focus solely on peer-reviewed articles. By doing so, we excluded reports from local NGOs in Bangladesh that could bring to light new elements only known at the local scale after a long 515 period of time. Nonetheless, we obtained a clear view of what is done throughout scientific articles in terms of coastal vulnerability. Additionally, we obtained 42 out of 49 analyzed articles which involved either Western or Asian scholars likely to assign concepts to Bangladesh that would normally work only into their own foreign country. Thus, the outcome of socio-spatial vulnerability is defined mainly from a Western perspective, because the majority of the articles included Western researchers. Moreover, bias in results the top-down approach used in 520 this paper results from an aggregation of data from studies that used different methodologies. Consequently, this bias could happen since by aggregating the data, we lost some accurate information and perhaps considered a concept defined with the same word but conceptualized differently between different studies. Another issue raised with the top-down approach is the absence of local community participation in this assessment. This feature is necessary to give a complete validity to such an evaluation, by taking into account people's opinions and 525 perceptions about their own vulnerability Sattar and Cheung, 2019). It is also through their participation that possible solutions and strategies of adaptation to disasters may appear and be formulated .

Conclusion
In this article, we defined vulnerability to cyclonic flooding in coastal Bangladesh based on the scientific literature 530 publications (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020). This enables both to test scientifically the research procedure to assess the socio-spatial vulnerability and to state the coastal places in Bangladesh which are under studies or which lack of attention, in order to steer forward researches where need is identified for coastal population. The mouth of the Ganges thus appears to be a relatively little studied area, although it is very exposed to cyclonic flooding and vulnerable due to the density of the population (Chandpur and, to a lesser extent, Shariatpur). 535 Thus, we argue than a better knowledge on the socio-spatial vulnerability can help sensitive classical cyclone flooding vulnerability analysis that tends to consider that groups or communities are equiprobably exposed to the hazard in a given spatiality. Consequently, we define a 'socio-spatial vulnerability index' (SSVI), as function of both the probability of the cyclonic flood hazard and the sensitivity of delta inhabitants. Through this index, we quantify and map levels of this vulnerability. SSVI is also a new methodology to highlight socio-spatial 540 inequalities between the coastal districts of Bangladesh, by considering social, physical and infrastructural aspects of vulnerability. The concept is thought in a systemic and interdisciplinary framework, considering this vulnerability aspects from social and natural sciences and their interactions between both. With this index and through its different components, we can better reveal why some groups or communities, at the districts level in case of our study, are more vulnerable to cyclonic flooding, and could be also slower to recover than others. 545 At final, we identify places the most vulnerable that are understudied, opening to consider differently the coastal area for State, NGOs and OIs to prioritize their Preparedness and Recovery programs. Climate change could increase the intensity and frequency of tropical cyclones in the Bay of Bengal, therefore access to appropriate cyclonic flood hazard, from hydrological modelling efficient tools, is a key parameter to predict the future sociospatial vulnerability patterns.