Seeking Shelter: The factors that in ﬂ uence refuge since Cyclone Gorky in the Coastal Area of Bangladesh

Considering damageandloss oflives, cyclones arethemost severenatural disasterinBangladesh.One ofthemost important cyclone mitigation measures is to temporarily evacuate the vulnerable population on receipt of warning and lookafterthem byarrangingshelters withadequate facilities. To ensureeffectiveandef ﬁ cient evacuation, itisimpor-tanttobetterunderstandpeople'sprotectiveactionsandhazarddecisionmaking.Thispaperprovidesanassessmentof the factors associated with evacuation to cyclone shelters in coastal areas in Bangladesh over the past 30 years. It is based on the comparative study of Cyclone Amphan (2020) and three major historical cyclones: Gorky (1991), Sidr (2007) and Aila (2009). Primary data sources included surveys, focus group discussions, key informant interviews and formal and informal conversations; and secondary data sources included reports, policies and design documents. 210 participants from seven coastal districts severely affected by Cyclone Amphan are selected for the study. While substantial improvements in the disaster response infrastructure were made in the last three decades, shelter evacua-tionrates arestill low.Itwas found that themajorityofpeopleare still takinga ‘ wait-and-see ’ approachmainlydueto the fearoflosingproperty. Itwas foundthat partialevacuation, observing environmentalcuesprior to evacuationand seeking refugein neighbours' houses, is indirectlyassociatedwith the fearof losing property. Thereis also inadequate spatial distribution of cyclone shelters with minimal facilities. This study provides recommendations to intensify risk-based planning for cyclone shelters and gender-responsive efforts to ensure a safe environment for the community to seek refuge at shelters with proper planning at the community level


Introduction
Bangladesh is one of the most climate vulnerable countries in the world, heavily exposed to natural hazards.Low-lying coastal districts along the Bay of Bengal are particularly vulnerable to monsoon flooding, cyclones and related storm surges [3].On average, five cyclones originate in the Bay of Bengal every year during the monsoon period [62].Cyclones are most severe natural disasters in Bangladesh, with almost half of all deaths caused by cyclones globally being reported in Bangladesh [42].The death toll from the cyclones in Bangladesh from 1484 to 2020 totalled to more than a million [63] the most catastrophic cyclones being Bhola (1970), Gorky (1991) and Sidr (2007) [41,52] (Fig. 1).
One of the most important cyclone mitigation measures is temporary evacuation of the vulnerable population on receipt of warning.This evacuation may be horizontal (moving out of the danger zone) and vertical (moving into a designated shelter in a danger zone [64].While horizontal evacuation is preferred in developed countries [20], it is rarely used in developing countries due to larger population, inadequate infrastructure and insufficient resources.Therefore, developing countries including Bangladesh rely on the evacuation to cyclone shelters in disaster risk management efforts. Following Cyclone Bhola (1970), the Government of the People's Republic of Bangladesh took measures to minimise the impacts of cyclones.Central to these efforts was the Cyclone Preparedness Programme [36].Evacuation of the population at risk to shelters was expedited through a highly organized volunteer network at the village level (consisting of 49,365 volunteers, of which one-third were female) [59].Following Cyclone Gorky (1991), which exposed inadequacy of the shelter system, the construction of multi-purpose cyclone shelters was accelerated.
The number of fatalities from the cyclones during the 1970-2007 period decreased at the rate of 2.5% per year [25].This may be attributed to the increase in both the shelter capacity and the evacuation rate.In 2007, Bangladesh coastal area had 2000 shelters with the capacity to accommodate only 1.5% of the coastal population [62].The average evacuation rate from Cyclone Gorky (1991) to Cyclone Sidr (2007) has changed from 27% to 33%, representing a 6% increase over 16 years [7].During http://dx.doi.org/10.1016/j.pdisas.2021.1001792590-0617/© 2021 The Authors.Published by Elsevier Ltd.This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Contents lists available at ScienceDirect
Progress in Disaster Science j o u r n a l h o m e p a g e : w w w .e l s e v i e r .c o m / l o c a t e / p d i s a s the Cyclone Aila (2009), majority of people at risk evacuated but only a small proportion sought refuge in cyclone shelters [44].
To ensure effective and efficient evacuation, it is important to better understand people's protective actions and hazard decision making.A number of complex factors influence people's decisions to evacuate, including socio-demographic characteristics such as age, gender, household income, education level and ethnicity [66].While various scientists [e.g., 7,29,43,48,50,58] investigated the behaviour of people in Bangladesh that do not seek shelter during a cyclone event, there is limited research that assesses the factors that drive decision to evacuate in cyclone shelters.
This paper provides an assessment of the factors associated with evacuation to cyclone shelters in coastal areas in Bangladesh over the past 30 years.It is based on the comparative study of Cyclone Amphan (2020) and three major historical cyclones: Gorky (1991), Sidr (2007) and Aila (2009).Detailed analysis of the factors associated with the noncompliance in the study areas is provided in a narrative format to deepen understanding of the actual scenario during emergency.The findings from this study will help researchers, practitioners and government agencies to revisit and redesign the conventional approach to disaster risk reduction system.

Impact of major cyclones in Bangladesh
Table 1 shows key information about the five major cyclones that hit the Bangladesh coast (Cyclone Bhola, Gorky, Sidr, Aila and Amphan).This study excluded Cyclone Bhola as there was limited information and the construction of cyclone shelter had not been undertaken at a mass scale.
Cyclone Bhol (Category 3, 12 November 1970) severely affected the coastal region of Bangladesh.It claimed 500,000 lives and caused a damage worth USD 450 million, making it the deadliest natural disaster in Bangladesh and human history [24].
Cyclone Gorky (Category 4, 29 April 1991) killed 138,882 people and caused a damage estimated at USD 1.8 billion, equivalent to one-third of the nation's gross domestic product (GDP) [41].Around 4.5 million people were directly or indirectly affected by Gorky [29].Department, 2007;43].This could be attributed to timely forecast, effective dissemination of warnings, and increased number of cyclone shelters.However, more than 27 million people were affected in 30 districts of Bangladesh.More than 500,000 houses and an additional 900,000 houses were destroyed and highly damaged [48,54].
Cyclone Amphan (Category 1, 20 May 2020) made a landfall in the southwestern Bangladesh, affecting 2.6 million people, destroying 55,767 houses, damaging 76 km of embankment and 440 km of road across more than 19 districts.The death toll was 22.

Cyclone shelters in Bangladesh
Cyclone shelters play an important role during disasters in ensuring successful adaptation steps and in preventing the loss of human lives, in particular for those living in the coastal areas and offshore islands of Bangladesh. 2  These shelters are reinforced concrete multi-storeyed buildings elevated several metres from the ground to resist storm surges.
The government and nongovernmental organisations undertook a series of programmes to construct cyclone shelters in vulnerable areas.However, such initiatives have not been fully effective, mainly due to inadequate maintenance and financial resources.
In the late 1960s, the government initiated the construction of 2000 cyclone shelters, but completed only 132 shelters due to limited financial resources [28,40].During Cyclone Bhola (1970), the number of cyclone shelters was 44 (reduced from 132, mostly due to poor usage of building materials, which contributed to early dilapidation).Following Cyclone Bhola, the government planned to construct 12,500 shelters, but completed 236 shelters by 1975 [7,41].By the mid-1980s, the project resumed and the Bangladesh Red Crescent Society constructed additional 60 shelters, with the capacity to accommodate more than 500 people per shelter.When Cyclone Gorky (1991) hit the country, there were 512 shelters which could accommodate 10% of the 14 million population in the cyclone risk zone [33].The number of shelters increased to only 542 by 1992 [7].
In 1993, the Multi-Purpose Cyclone Shelter Project was initiated in the cyclone risk areas.During the non-emergency period, the multipurpose cyclone shelters serve as schools, community centres and temporary offices.By the mid-2000s, the number of shelters increased to 3976.However, 1576 shelters were severely affected due to poor management and maintenance.An overwhelming number of multi-purpose shelters has limited maintenance funds.Hence, the shelters which were in a relatively proper condition could accommodate only 15% of the coastal population [43].According to the [26], the number of shelters in operation was 3751 in 15 coastal districts; out of these, 7% have become dilapidated and unusable due to the river erosion and inadequate maintenance [7].Hence, the cyclone shelters could accommodate only 7.3% of the coastal population [28].
During Cyclone Amphan (2020), 6788 additional informal shelters (upazila offices, hospitals and neighbours house) were made accessible due to inadequate formal cyclone shelters.It is not clear how many cyclone shelters are usable, moderately useable, or possess basic amenities.Source: Authors.
Blue line shows that in 1970 around 10% of Bangladesh population lived in the coastal area and that in 2030, it is projected to have around 30% of the population along the coastline.There is a constant gap between the shelter capacity and the population needs, which has significantly increased since 2007.For 2030, it is projected that only 5% of the coastal population could be accommodated in shelters during a cyclone event.
Fig. 3 shows two kinds of shelters: a single-purpose shelter used during cyclone events only and a multi-purpose cyclone shelter which can be used for other purposes when there is no cyclone event happening.
In 2010, the World Bank estimated that more than 5500 multi-purpose cyclone shelters were required to be constructed as part of a disaster management strategy.Although literature states [38] that each cyclone shelter can accommodate 500-2500 people, in practice a typical cyclone shelter, particularly in villages, could accommodate only 250 to 500 people.The shelters are not properly distributed along the coast based on population density.Table 2 reveals that 5.7% of the cyclone shelters are not operational.Most of the shelters which are usable or moderately usable lack Fig. 2. Shelter capacity to accommodate coastal population in Bangladesh. 1 A landfall is the storm moving over the land after its intensification in the heat source.Therefore, a tropical cyclone is said to make landfall when the centre of the storm moves across the coast. 2 People residing in coastal areas with low elevation levels are most exposed to cyclones.
basic facilities like separate toilet facilities for women.Killa is another important facility added to some of the cyclone shelters to accommodate the livestock in a raised land above storm height level close to the cyclone shelter [1].

Approach and methodology
A mix method using quantitative and qualitative assessment was used to analyse data obtained through multiple information sources (literature review, household survey, and phone interviews).Primary data were collected from surveys and informal conservations with the local people and government officials.Data were collected during nationwide COVID-19 lockdown, thus face-to-face conversations and surveys could not have been undertaken in all study areas.Information gathered over phone interviews were cross-checked with the villagers of the adjoining villages to verify the data source and authenticity.Secondary data were collected from peer reviewed articles and government reports.
According to the Joint Need Assessment report by CARE Bangladesh [16], eight districts were affected: Satkhira, Khulna, Jashore, Bagerhat, Pirojpur, Barguna, Bhola and Patuakhali.Satkhira and Khulna had a higher number of populations affected by Cyclone Amphan.Seven villages from the seven most affected coastal districts (Satkhira, Khulna, Jashore, Bhola, Barguna, Pirojpur and Patuakhali) were selected for this study (Fig. 4).The villages were selected using purposive sampling of the villages with no or inadequate cyclone shelters.Time constraints and insufficient resources restricted the access to these areas physically.
210 participants (75 female and 135 male) were selected, which is on average 30 participants per village.Discrepancies in the number of female and male respondents are attributed to the socio-cultural aspects of rural areas.Some females felt discomfort to talk to strangers, even to female volunteers, in front of their husbands, and some were engaged in the household chores.
In the coastal areas, there are volunteers from the Cyclone Preparedness Programme.They were selected on the basis of their interest to participate in the study to collect data from the participants.The volunteers were mostly school teachers, madrasa teachers, housewives, businessmen and high school students.They all can read and write Bengali.Five volunteers (two female and three male) from each village agreed to conduct surveys.Female volunteers were selected to gain access to information from the female participants.The number of volunteers varied from one village to another, however, the purpose of consulting 30 participants from each village was served.
Households were randomly selected to eliminate any kind of biases.Union Parishad members 3 of their respective villages provided mobile numbers of 8-10 villagers from each village.Each volunteer conducted a phone interview of four to six households.Key informants include local government officials, school teachers and Union Parishad members.10-15 villagers from each village were randomly telephoned to ensure authenticity of the information collected.
Information were collected through a pre-tested questionnaire administered by 35 interviewers, five in each village.The questionnaire comprised of questions pertaining to the status of the cyclone shelters and number of factors affecting decision to seek refuge at the cyclone shelters.Photography of cyclone shelters were sent via mobile phones.
Key informant interviews were conducted to understand how people responded to the Cyclone Amphan.Extensive literature review on factors influencing people to seek refuge at shelters during Cyclone Gorky, Sidr and Aila was conducted.
The factors identified by authors may not represent the actual scenario of the event of a disaster; however, they provide an understanding of common factors of not seeking refuge at the cyclone shelters.All variables associated with people not seeking refuge at the cyclone shelters have been categorised in accordance with the three time periods (1991, 2007 and 2009).Source: Bangladesh Department of Disaster Management, 2020.Source: Authors.

Evacuation debacle: people's perception
Ikeda [33], Chowdhury et al. [14] and Paul et al. [50] stated that distance is the major impediment in seeking refuge at the shelters.During cyclones Gorky, Sidr and Aila the mean distance from the shelters was 5.00 km, 3.00 km and 3.14 km respectively [1,52].Most people in Bangladesh do not go to shelters until the surge water enters their courtyards, hence walking more than 2.0 km becomes difficult [14].Most locals can travel between 0.8 km and 1.0 km to a shelter during a cyclone [47].Unelevated and undulating road to the shelter coupled with surge water obstructed people to go to shelters during cyclones Gorky, Sidr and Aila [14,44,46,48,51,52].
The capacity to accommodate people is often insufficient due to inadequate number of shelters.During Cyclone Gorky, death rates were highest due to the absence of shelters or insufficient space [29].A study conducted by Paul [46] found that 17% of the respondents came back home due to over-crowdedness during Cyclone Sidr.During Cyclone Aila, over 77% of the non-evacuees came back home due to insufficient space in the cyclone shelter [1].In most cases, cyclone shelters were occupied by local elites, hence the poor people seeked refuge on embankments, trees and other vulnerable and low-lying areas [52].A study by Parvin et al. [44] shows that 69% of the evacuees sought refuge on elevated land such as road and embankment.
False sense of security is common among the inhabitants of the coastal area [30].A study by Paul [43] shows that 12% of the non-evacuee respondents believed that their houses would withstand cyclone surges during Sidr.A study by Paul [46] shows that around 12% of the non-evacuee respondents refuse to go to the cyclone shelter since they thought it would not be able to withstand the intensity of the cyclone.This could be attributed to the fact that villagers adopt indigenous practices to reinforce their houses and avoid inundation from storm surges.A study by Paul et al. [50] identified that some crude housing reinforcing measures were used to withstand the impacts of cyclones.Almost all houses of the respondents were submerged during Cyclone Gorky.To avoid inundation induced by storm surges, around 73% of the respondents raised their yards at varied intervals.One-third of the respondents installed pillars to strengthen the foundations and another one-third tied their houses with strong ropes to trees.Such practices may have provided them false sense of security during cyclones.
society of Bangladesh.In a conservative society, Muslim women are not allowed to leave home to take shelter because of purdah norm4 [67,8].
The practice of purdah is strictly observed in the coastal areas of Bangladesh, where men still exhibit reluctance in taking their wives to the cyclone shelter [68].Many women do not go to the cyclone shelter as they do not have separate toilets.A study by Ahsan et al. [1] shows that more than 35% of the non-evacuee respondents did not seek refuge to the cyclone shelters due to the lack of separate toilets and spaces for men and women.Poor condition and lack of adequate facilities of the cyclone shelters such as limited number of latrines, provision for drinking water and storage for food have also been considered as important factors.
One of the major socio-economic factors is the fear of losing belongings if left abandoned during disasters.The majority of the coastal population is living at or below the poverty line hence losing their belongings would contribute to extreme poverty, hunger and perhaps even death.A study by Haque and Blair [29] shows that 57.6% and 31.6% of the non-evacuee respondents residing in Halishahar and Jahanabaj areas feared that their household items will be looted, and did not go to the cyclone shelter.Households whose principal source of income is livestock rearing exhibited reluctance to go to the cyclone shelter since there were no killas nearby.A study by Ahsan et al. [1] shows that 72% of the respondents did not go to the cyclone shelter since there were no killas nearby.

Use of shelters during an emergency: Cyclone Amphan
Key factors that contributed to refusal of seeking refuge at the shelter during Cyclone Amphan are: fear of losing property, inadequate shelter, poor conditions and false sense of security or people living in a relatively strongly built house.
The majority of the respondents did not go to the cyclone shelter due to fear of losing their properties.35 (17%) of the non-evacuee respondents did not go to the cyclone shelter because they thought staying at home would help prevent their houses being destroyed or looted.Out of 35 respondents, four respondents from Bodropara anchored their boats at khal5 and took refuge inside their boats with family members and valuable belongings.14 male respondents from Char Anda sent their wives and children to cyclone shelters and stayed at home to protect their houses.One respondent from Char Anda reported that staying at home may not help to protect his house from being swept away; however, it would give him peace knowing that he tried till the end.He kept a rope at hand to tie oneself against a tree provided the wind intensifies.17 male respondents from Bodropara sent their wives and children to another village with the trawler and they stayed at home and observed the weather patterns.When the situation worsened, they took refuge at the cyclone shelter.One respondent reported coming back from the shelter to protect house being damaged and swept away.Informal conversations with the villagers of Bodropara revealed that, on request of the Upazila Nirbahi Officer, fishing trawlers were sent with the coast guards to evacuate the people of Bodropara and the adjoining villages.These attempts of saving property reflect how the fear of losing property influences people to evacuate to the shelter.
Partial evacuation was found to be one of the indicators of fear of losing property [9].It also reflects the male dominance in the rural areas indicating that women are not capable of providing assistance in times of disaster.Informal conversation with the female respondents revealed that they managed to take children and livestock, along with some dry food, to the shelter during cyclones in the past.This reflects that women play a huge part in disaster risk reduction.
The survey results revealed that, even if a shelter is 1 km away from the residence, some people exhibited reluctance in seeking refuge at the shelter.This is particularly true for the people with small poultry farm businesses, since it is difficult to carry them to the shelter.It reflects that fear of losing asset dominates over the distance to shelters.30 respondents (14%) reluctantly took refuge at their neighbours' houses at the last moment.Among them, 13 respondents have small poultry farms and 12 respondents sought refuge to the neighbour's place closer to their houses to observe their property.Three respondents from Purbo Kalinagar and two from Singkhali sought refuge at the neighbour's house for the fear of being infected with COVID-19 at the shelter.The number of confirmed COVID-19 cases was 35,585 in Bangladesh until 25 May 2020 [65].According to the respondents, some shelters provided masks to the evacuees, however, very few were wearing them.
About 17% (36) respondents did not go to the cyclone shelter.Among 36 respondents, nine respondents have pucca house and 27 respondents reinforced houses with woods with elevated plinths.Fig. 5 shows an example of houses in the study area.
The majority of the respondents have improved their housing condition particularly following the cyclone of 1991.Jhupri houses, where the roofs of the houses were previously made from the leaves of Nipa Palm locally known as Golpata, are now roofed with corrugated tin.None of the participants have houses made from the leaves of Golpata or straw.Hence the very definition of jhupri has been changed.Such improvements have been reflected in the housing condition of the study areas.Table 4 shows the number of katcha, pucca, semi-pucca and jhupri houses6 across the study areas (See Table 5).
Insightful data were received when the villagers were asked on choosing the best disaster management measure.15 respondents believe that mending or constructing embankments would solely save them during cyclones.65 respondents believe that construction of cyclone shelters and embankments would save their lives and livestock.25 respondents believe that constructions of embankments and cyclone shelters would save them from disaster and would provide an opportunity for their children to go to schools since the government constructs mostly school cum cyclone shelters.Others (105 respondents) preferred receiving financial assistance to reinforce or build their houses with bricks.Irrespective of the presence of multi-purpose cyclone shelters 7 for more than two decades, people perceive them solely as cyclone shelters.
67 respondents (32%) observed various environmental cues such as intensity of the wind, colour of the sky and height of storm surge, irrespective of the evacuation orders and meteorological warnings prior to evacuation.This group includes people who stayed at home by sending their wives and children to the cyclone shelters and people who sought refuge at their neighbour's house and cyclone shelters at the last moment.Observing the weather pattern until the situation worsens reflects that the cyclone shelters or other places are considered as the last resort during an emergency.One respondent from Singkhali reported about leaving the house with family members to the neighbours' place when the wind intensified.
56 respondents (27%) stated that inadequate cyclone shelters caused them trouble during an emergency.Among the 56 respondents, 20 respondents returned home since the cyclone shelters were full or crowded and 13 respondents went to neighbour's house.23 respondents went to multiple places to seek refuge prior to disasters.One respondent from Gazipara took his wife to a cyclone shelter in another village since the shelter in his village was under construction.He returned home and took his livestock to a nearby killa and tied himself to a tree near his house.There is only one cyclone shelter available for Purbo Kalinagar, Gazipara and Char Anda villages (Table 4).In Gazipara, there is one cyclone shelter under construction which was not completed due to the COVID-19 outbreak.Many people sought refuge at the under-constructed cyclone shelter.In Char Anda, there are two cyclone shelters; one in a very poor condition and the other one is in a better condition, however too small to accommodate all villagers.About 23% (49) of the respondents fully evacuated with their family members to the cyclone shelters.Out of them, 13 respondents kept their livestock at neighbour's place and went to the cyclone shelter.
Informal conversations with female respondents revealed that purdah is no longer considered as a major hindrance while seeking refuge at the cyclone shelter.All female respondents (75 respondents) prioritised saving their lives over traditional belief.This could be attributed to the fact that women and children were mostly sent to the cyclone shelters in the study areas.However, they exhibited discomfort of staying inside overcrowded cyclone shelters.One female respondent from Purbo Kalinagar reported feeling suffocated inside the crowded shelter.They expressed concern over having less or no separate toilets and space for men and women.The majority of the female respondents were escorted by husbands to the cyclone shelter during an emergency.This reflects that women are not allowed to go outside of their houses alone.

Conclusions and recommendations
The study explored factors which influence people in seeking refuge at cyclone shelters during the major cyclones Gorky, Sidr, Aila and compared Fig. 5. Example of brick-built home, jhupri and semi-pucca houses.Source: Authors.

Table 4
Types of housing across the study areas.
them with the most recent Cyclone Amphan to identify major factors that influences them in making decision to evacuate.Despite significant improvement in disaster management in Bangladesh, variables influencing non-compliance with evacuation orders have not changed significantly over the last 30 years.
Inadequate cyclone shelters in proximity of the villages are the major deterrent to shelter evacuation.There are three major factors for not seeking refuge at the cyclone shelter found in this study: fear of losing property, inadequate shelter, and false sense of security or people living in a relatively strongly built house.Fear of losing property is a dominant factor.The majority of people considered evacuation as a last resort.Partial evacuation, observing environmental cues prior to evacuation and seeking refuge at cyclone shelter and neighbours' houses at the last moment, is indirectly associated with the fear of losing property.The study suggested several recommendations, as follows:

Housing condition and compliance with national building codes
Housing conditions of the village could be integrated into planning and selecting sites for cyclone shelters.As cyclones are most likely to intensify as a result of climate change, design of cyclone shelters and buildings in the cyclone hazard areas need to compliance with national building codes.Other than shelters, other adaptation options such as embankments, nature based solutions (i.e.mangrove) should incorporate aspects of changing climatic patterns such as intensification of surge water and wind speed.Existing building structures should be reinforced and additional building structures, if required, should be constructed in proximity to larger communities so that they can keep assets and seek refuge at the cyclone shelter during cyclones.

Assessment of cyclone exposure and shelter needs
The study found that the houses in Mainapur village are newly build and well-constructed and thus people may not need to go to the formal evacuation place.A proper needs assessment would ensure number of shelters necessary in the risk areas.The hazard assessment also inaccurate due to lack of high-resolution topography and advance modelling.With the advancement in the modelling and current data, risk-based coastal inundation information would enable them to take appropriate action.

Gender responsive shelter management plan
Women's and men's vulnerability to the impact of severe cyclone events is determined not only by biology but also by differences in their cultural, social roles and responsibilities.Men and women are affected with different vulnerabilities, particularly, when it comes to safeguarding their food security, livelihoods and other social issues.Acquiring information on the male to female ratio of a village and their vulnerabilities would help to effectively integrate gender-sensitive aspects (separate toilets and space for male and female) into the construction and develop facilities in the cyclone shelter.

Cyclone shelter management information system and budget
Government and nongovernmental organisations should intensify efforts to obtain international funds to reinforce dilapidated houses, particularly located in risk and high-risk zones, and establishing a centre or a structure to keep their assets during an emergency.

Awareness campaign and exercise
A community-based programme to raise awareness of appropriate actions during an emergency (e.g., remaining calm, evacuate immediately following receiving warning messages, taking dry food prior to evacuating to the cyclone shelter) should be conducted.Such messages can be disseminated through popular theatre (theatre which delivers messages to people in their language), street theatre and FM radio.

Ensuring livestock safety for communities
The construction of killas in proximity to cyclone shelters and larger communities should be intensified and adequate measures should be undertaken to minimise the threat of theft during cyclones.People often feel discomfort for leaving livestock in outdoors and at risk of being stolen.A pilot initiative by the government may be undertaken integrating fencing and sheds into the design of killas.During a non-emergency period, these killas can be used as local markets or local people can take their livestock there during rainy seasons.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Cyclone risk map in the coastal area of Bangladesh.

Fig 2
illustrates the trend of the coastal population in Bangladesh (blue line) against the capacity to accommodate the coastal population (orange line).

Fig. 4 .
Fig. 4. Study areas and cyclone shelters in the coastal area of Bangladesh.Source: Authors.

Table 3
lists common factors categorised in five themes: infrastructure deficits, invasion of privacy, insufficient sanitary facilities, religious and cultural barriers, and other barriers.

Table 2
Distribution, types and use of cyclone shelters.

Table 3
Reasons for not seeking refuge at the cyclone shelters.