seeking shelter: the factors that influence refuge since

9
Seeking Shelter: The factors that inuence refuge since Cyclone Gorky in the Coastal Area of Bangladesh Tahmina Hadi a , Md. Sirajul Islam b , Denise Richter c , Bapon (SHM) Fakhruddin c, a Climate Change Programme, BRAC, Bangladesh b North South University, Bangladesh c Tonkin and Taylor Limited, New Zealand ABSTRACT ARTICLE INFO Article history: Received 11 December 2020 Received in revised form 27 April 2021 Accepted 27 May 2021 Available online 07 June 2021 Considering damage and loss of lives, cyclones are the most severe natural disaster in Bangladesh. One of the most im- portant cyclone mitigation measures is to temporarily evacuate the vulnerable population on receipt of warning and look after them by arranging shelters with adequate facilities. To ensure effective and efcient evacuation, it is impor- tant to better understand people's protective actions and hazard decision making. 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). 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- tion rates are still low. It was found that the majority of people are still taking a wait-and-seeapproach mainly due to the fear of losing property. It was found that partial evacuation, observing environmental cues prior to evacuation and seeking refuge in neighbours' houses, is indirectly associated with the fear of losing property. There is 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. Keywords: Climate change Cyclones Refuge seeking Bangladesh 1. 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 ooding, cyclones and related storm surges [3]. On average, ve 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 evacu- ation may be horizontal (moving out of the danger zone) and vertical (mov- ing into a designated shelter in a danger zone [64]. While horizontal evacuation is preferred in developed countries [20], it is rarely used in de- veloping countries due to larger population, inadequate infrastructure and insufcient resources. Therefore, developing countries including Bangladesh rely on the evacuation to cyclone shelters in disaster risk man- agement efforts. Following Cyclone Bhola (1970), the Government of the People's Re- public 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 Cy- clone 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 19702007 pe- riod 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 ac- commodate only 1.5% of the coastal population [62]. The average evacua- tion rate from Cyclone Gorky (1991) to Cyclone Sidr (2007) has changed from 27% to 33%, representing a 6% increase over 16 years [7]. During Progress in Disaster Science 11 (2021) 100179 Corresponding author. E-mail address: [email protected] (B.(S.H.M.) Fakhruddin). http://dx.doi.org/10.1016/j.pdisas.2021.100179 2590-0617/© 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 journal homepage: www.elsevier.com/locate/pdisas

Upload: others

Post on 15-May-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Seeking Shelter: The factors that influence refuge since

Progress in Disaster Science 11 (2021) 100179

Contents lists available at ScienceDirect

Progress in Disaster Science

j ourna l homepage: www.e lsev ie r .com/ locate /pd isas

Seeking Shelter: The factors that influence refuge since Cyclone Gorkyin the Coastal Area of Bangladesh

Tahmina Hadi a, Md. Sirajul Islam b, Denise Richter c, Bapon (SHM) Fakhruddin c,⁎

a Climate Change Programme, BRAC, Bangladeshb North South University, Bangladeshc Tonkin and Taylor Limited, New Zealand

⁎ Corresponding author.E-mail address: [email protected] (B.(S.H

http://dx.doi.org/10.1016/j.pdisas.2021.1001792590-0617/© 2021 The Authors. Published by Elsevi

A B S T R A C T

A R T I C L E I N F O

Article history:Received 11 December 2020Received in revised form 27 April 2021Accepted 27 May 2021Available online 07 June 2021

Considering damage and loss of lives, cyclones are the most severe natural disaster in Bangladesh. One of themost im-portant cyclone mitigation measures is to temporarily evacuate the vulnerable population on receipt of warning andlook after them by arranging shelters with adequate facilities. To ensure effective and efficient evacuation, it is impor-tant to better understand people's protective actions and hazard decisionmaking. This paper provides an assessment ofthe factors associated with evacuation to cyclone shelters in coastal areas in Bangladesh over the past 30 years. It isbased 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 interviewsand 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. Whilesubstantial improvements in the disaster response infrastructure were made in the last three decades, shelter evacua-tion rates are still low. It was found that the majority of people are still taking a ‘wait-and-see’ approachmainly due tothe fear of losing property. It was found that partial evacuation, observing environmental cues prior to evacuation andseeking refuge in neighbours' houses, is indirectly associated with the fear of losing property. There is also inadequatespatial 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 toseek refuge at shelters with proper planning at the community level.

Keywords:Climate changeCyclonesRefuge seekingBangladesh

1. Introduction

Bangladesh is one of themost climate vulnerable countries in theworld,heavily exposed to natural hazards. Low-lying coastal districts along theBay of Bengal are particularly vulnerable to monsoon flooding, cyclonesand related storm surges [3]. On average, five cyclones originate in theBay of Bengal every year during the monsoon period [62]. Cyclones aremost severe natural disasters in Bangladesh, with almost half of all deathscaused by cyclones globally being reported in Bangladesh [42]. The deathtoll from the cyclones in Bangladesh from 1484 to 2020 totalled to morethan 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 temporaryevacuation of the vulnerable population on receipt of warning. This evacu-ationmay be horizontal (moving out of the danger zone) and vertical (mov-ing into a designated shelter in a danger zone [64]. While horizontalevacuation is preferred in developed countries [20], it is rarely used in de-veloping countries due to larger population, inadequate infrastructure and

.M.) Fakhruddin).

er Ltd. This is an open access artic

insufficient resources. Therefore, developing countries includingBangladesh rely on the evacuation to cyclone shelters in disaster risk man-agement efforts.

Following Cyclone Bhola (1970), the Government of the People's Re-public 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 ahighly organized volunteer network at the village level (consisting of49,365 volunteers, of which one-third were female) [59]. Following Cy-clone Gorky (1991), which exposed inadequacy of the shelter system, theconstruction of multi-purpose cyclone shelters was accelerated.

The number of fatalities from the cyclones during the 1970–2007 pe-riod decreased at the rate of 2.5% per year [25]. This may be attributedto the increase in both the shelter capacity and the evacuation rate. In2007, Bangladesh coastal area had 2000 shelters with the capacity to ac-commodate only 1.5% of the coastal population [62]. The average evacua-tion rate from Cyclone Gorky (1991) to Cyclone Sidr (2007) has changedfrom 27% to 33%, representing a 6% increase over 16 years [7]. During

le under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Page 2: Seeking Shelter: The factors that influence refuge since

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

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

the Cyclone Aila (2009), majority of people at risk evacuated but only asmall proportion sought refuge in cyclone shelters [44].

To ensure effective and efficient evacuation, it is important to betterunderstand people's protective actions and hazard decision making. Anumber of complex factors influence people's decisions to evacuate, in-cluding socio-demographic characteristics such as age, gender, householdincome, education level and ethnicity [66]. While various scientists[e.g., 7,29,43,48,50,58] investigated the behaviour of people inBangladesh that do not seek shelter during a cyclone event, there is lim-ited research that assesses the factors that drive decision to evacuate incyclone shelters.

This paper provides an assessment of the factors associated with evacu-ation to cyclone shelters in coastal areas in Bangladesh over the past 30years. 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 non-compliance in the study areas is provided in a narrative format to deepenunderstanding of the actual scenario during emergency. The findings

Table 1Characteristics of cyclones Bhola, Gorky, Aila, Sidr, Amphan in Bangladesh.

Year Cyclone Landfall area Maximum windspeed (km/h)

Category Tidalsurgeheigh(m)

1970 Bhola Chittagong 224 3 3.0–11991 Gorky Chittagong 225 4 3.7–62007 Sidr Khulna-Barisal Coast near Baleshwar

River223 4 4.6–6

2009 Aila West Bengal-Khulna Coast near SagarIsland

70–90 1 3.0

2020 Amphan West Bengal-Jammu Island 160 3 3.4–6

Source: [27, 5; 22; 61, 16].

2

from this study will help researchers, practitioners and government agen-cies to revisit and redesign the conventional approach to disaster risk reduc-tion system.

2. Impact of major cyclones in Bangladesh

Table 1 shows key information about thefivemajor cyclones that hit theBangladesh coast (Cyclone Bhola, Gorky, Sidr, Aila and Amphan). Thisstudy excluded Cyclone Bhola as there was limited information and theconstruction of cyclone shelter had not been undertaken at a mass scale.

Cyclone Bhol (Category 3, 12 November 1970) severely affected thecoastal region of Bangladesh. It claimed 500,000 lives and caused a damageworth USD 450 million, making it the deadliest natural disaster inBangladesh and human history [24].

Cyclone Gorky (Category 4, 29 April 1991) killed 138,882 people andcaused a damage estimated at USD 1.8 billion, equivalent to one-third ofthe nation's gross domestic product (GDP) [41]. Around 4.5 million peoplewere directly or indirectly affected by Gorky [29].

t

Deathtoll

Houses damaged (number) Damage (USD,million)

0.0 500,000 912,000 fully and partial damaged 864.7 138,882 1 million fully and partial damaged 2280.0 3363 565,000 totally destroyed over 950,000 significantly

destroyed1521

190 94,609 fully damaged48,097 partially damaged

1000

.0 22 55,667 fully damaged162,000 partially damaged

375

Page 3: Seeking Shelter: The factors that influence refuge since

Fig. 2. Shelter capacity to accommodate coastal population in Bangladesh.

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

Cyclone Sidr (Category 4, 15 November 2007) made a landfall1 acrossthe southwestern region of Bangladesh. It claimed 3363 lives, significantlyfewer than in 1970 and 1991 cyclones [Bangladesh Meteorological Depart-ment, 2007; 43]. This could be attributed to timely forecast, effective dis-semination of warnings, and increased number of cyclone shelters.However, more than 27 million people were affected in 30 districts ofBangladesh. More than 500,000 houses and an additional 900,000 houseswere destroyed and highly damaged [48,54].

Cyclone Aila (Category 1, 25 May 2009) affected 11 southwesterncoastal districts, claimed 190 lives, injured more than 7000 people and af-fected about 4.8 million people [56,58].

Cyclone Amphan (Category 1, 20 May 2020) made a landfall in thesouthwestern Bangladesh, affecting 2.6 million people, destroying 55,767houses, damaging 76 km of embankment and 440 km of road acrossmore than 19 districts. The death toll was 22.

3. Cyclone shelters in Bangladesh

Cyclone shelters play an important role during disasters in ensuring suc-cessful adaptation steps and in preventing the loss of human lives, in partic-ular for those living in the coastal areas and offshore islands of Bangladesh.2

These shelters are reinforced concrete multi-storeyed buildings elevatedseveral metres from the ground to resist storm surges.

The government and nongovernmental organisations undertook a seriesof programmes to construct cyclone shelters in vulnerable areas. However,such initiatives have not been fully effective, mainly due to inadequatemaintenance and financial resources.

In the late 1960s, the government initiated the construction of 2000 cy-clone shelters, but completed only 132 shelters due to limited financial re-sources [28,40]. During Cyclone Bhola (1970), the number of cycloneshelters was 44 (reduced from 132, mostly due to poor usage of buildingmaterials, which contributed to early dilapidation). Following CycloneBhola, the government planned to construct 12,500 shelters, but completed236 shelters by 1975 [7,41]. By themid-1980s, the project resumed and theBangladesh Red Crescent Society constructed additional 60 shelters, withthe capacity to accommodate more than 500 people per shelter. When

1 A landfall is the storm moving over the land after its intensification in the heat source.Therefore, a tropical cyclone is said tomake landfall when the centre of the stormmoves acrossthe coast.

2 People residing in coastal areas with low elevation levels are most exposed to cyclones.

3

Cyclone Gorky (1991) hit the country, there were 512 shelters whichcould accommodate 10% of the 14 million population in the cyclone riskzone [33]. The number of shelters increased to only 542 by 1992 [7].

In 1993, the Multi-Purpose Cyclone Shelter Project was initiated in thecyclone risk areas. During the non-emergency period, the multipurpose cy-clone 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 mainte-nance. An overwhelming number of multi-purpose shelters has limitedmaintenance funds. Hence, the shelters which were in a relatively propercondition could accommodate only 15% of the coastal population [43].According to the [26], the number of shelters in operation was 3751in 15 coastal districts; out of these, 7% have become dilapidated andunusable due to the river erosion and inadequate maintenance [7].Hence, the cyclone shelters could accommodate only 7.3% of the coastalpopulation [28].

During Cyclone Amphan (2020), 6788 additional informal shelters(upazila offices, hospitals and neighbours house) were made accessibledue to inadequate formal cyclone shelters. It is not clear howmany cycloneshelters are usable, moderately useable, or possess basic amenities.

Fig 2 illustrates the trend of the coastal population in Bangladesh (blueline) against the capacity to accommodate the coastal population (orangeline).

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 around30% of the population along the coastline. There is a constant gap betweenthe shelter capacity and the population needs, which has significantly in-creased since 2007. For 2030, it is projected that only 5%of the coastal pop-ulation could be accommodated in shelters during a cyclone event.

Fig. 3 shows two kinds of shelters: a single-purpose shelter used duringcyclone events only and a multi-purpose cyclone shelter which can be usedfor other purposes when there is no cyclone event happening.

In 2010, the World Bank estimated that more than 5500 multi-purposecyclone shelters were required to be constructed as part of a disaster man-agement strategy. Although literature states [38] that each cyclone sheltercan accommodate 500–2500 people, in practice a typical cyclone shelter,particularly in villages, could accommodate only 250 to 500 people. Theshelters are not properly distributed along the coast based on populationdensity. Table 2 reveals that 5.7% of the cyclone shelters are not opera-tional. Most of the shelters which are usable or moderately usable lack

Page 4: Seeking Shelter: The factors that influence refuge since

Fig. 3. Example of single-purpose and multi-purpose cyclone shelters. Source: Authors.

Table 2Distribution, types and use of cyclone shelters.

Districts Number of shelters Carryingcapacity

Number ofpeople thatsought refuge

Number oflivestocksheltered

Formal Informal Total

Bagerhat 420 611 1031 486,277 208,000 23,029Barguna 509 120 629 320,983 380,000 26,942Barishal 316 755 1071 249,150 229,870 16,020Bhola 704 400 1104 536,000 316,509 150,315Chadpur 324 1 325 103,457 16,700 2720Chittagong 511 1440 1951 200,000 64,213 13,021Cox's Bazaar 576 0 576 555,000 32,857 1965Feni 101 0 101 46,700 2200 0Gopalganj 4 1 5 1600 400 0Madaripur 92 0 92 27,600 1434 620Jhalokati 60 414 474 392,325 10,000 2175Khulna 814 0 814 416,650 111,600Lashimpur 101 151 252 71,000 14,885 2795Narail 3 0 3 2000 0 0Noakhali 450 450 900 328,200 21,122 13,909Patuakhali 701 206 907 655,100 381,989 89,623Pirojpur 235 477 712 405,270 270,340 35,690Satkhira 145 1762 1907 750,000 353,000 31,286Shariatpur 299 0 299 59,800 0 0Total 6365 6788 13,153 5,607,112 2,415,119 410,110

Source: Bangladesh Department of Disaster Management, 2020.Source: Authors.

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

basic facilities like separate toilet facilities for women. Killa is another im-portant facility added to some of the cyclone shelters to accommodate thelivestock in a raised land above storm height level close to the cycloneshelter [1].

3 Union is the lowest administrative unit of Bangladesh, comprised of few villages; UnionParishad is a directly elected body of local government representatives led by a Chairmanwitha few members.

4. Approach and methodology

Amixmethod using quantitative and qualitative assessment was used toanalyse data obtained through multiple information sources (literature re-view, household survey, and phone interviews). Primary data were col-lected from surveys and informal conservations with the local people andgovernment officials. Data were collected during nationwide COVID-19lockdown, thus face-to-face conversations and surveys could not havebeen undertaken in all study areas. Information gathered over phone inter-views were cross-checked with the villagers of the adjoining villages to ver-ify the data source and authenticity. Secondary data were collected frompeer 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 highernumber of populations affected by Cyclone Amphan. Seven villages fromthe 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

4

no or inadequate cyclone shelters. Time constraints and insufficient re-sources restricted the access to these areas physically.

210 participants (75 female and 135 male) were selected, which is onaverage 30 participants per village. Discrepancies in the number of femaleand male respondents are attributed to the socio-cultural aspects of ruralareas. Some females felt discomfort to talk to strangers, even to female vol-unteers, in front of their husbands, and some were engaged in the house-hold chores.

In the coastal areas, there are volunteers from the Cyclone PreparednessProgramme. They were selected on the basis of their interest to participatein the study to collect data from the participants. The volunteers weremostly school teachers, madrasa teachers, housewives, businessmen andhigh 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 fe-male participants. The number of volunteers varied from one village to an-other, however, the purpose of consulting 30 participants from each villagewas served.

Households were randomly selected to eliminate any kind of biases.Union Parishad members3 of their respective villages provided mobilenumbers of 8–10 villagers from each village. Each volunteer conducted aphone interview of four to six households. Key informants include localgovernment officials, school teachers and Union Parishad members.10–15 villagers from each village were randomly telephoned to ensure au-thenticity of the information collected.

Information were collected through a pre-tested questionnaire adminis-tered by 35 interviewers, five in each village. The questionnaire comprisedof questions pertaining to the status of the cyclone shelters and number offactors affecting decision to seek refuge at the cyclone shelters. Photogra-phy of cyclone shelters were sent via mobile phones.

Key informant interviews were conducted to understand how peopleresponded to the Cyclone Amphan. Extensive literature review on factorsinfluencing people to seek refuge at shelters during Cyclone Gorky, Sidrand Aila was conducted.

The factors identified by authors may not represent the actual scenarioof the event of a disaster; however, they provide an understanding of com-mon factors of not seeking refuge at the cyclone shelters. All variables asso-ciated with people not seeking refuge at the cyclone shelters have beencategorised in accordance with the three time periods (1991, 2007 and2009).

5. Evacuation debacle: people's perception

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

Page 5: Seeking Shelter: The factors that influence refuge since

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

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

Barriers for usingshelter

Reasons for not seeking shelter Cyclone

Gorky,1991

Sidr,2007

Aila,2009

Infrastructuredeficits

Inadequate shelters in the affectedareas

✓ ✓ ✓

Insufficient capacity ✓ ✓ ✓

Poor road condition ✓ ✓ ✓

Dilapidated condition of exisitingshelters

✓ ✓

No killa (animal shelter) ✓ ✓ ✓

Invasion of privacy Crowded shelters ✓ ✓

Purdah (personal privacy) ✓ ✓ ✓

Other gender-sensitive issues (i.e.toilet, hygiene, etc.)

✓ ✓

Insufficient sanitaryfacilities

No separate toilet ✓ ✓ ✓

Limited water and sanitation facilities ✓ ✓ ✓

Religious andcultural barriers

Allah's (God's) will ✓ ✓

Neighbour's house stronger and betterthan shelter

✓ ✓

False warning ✓

Occupied by local elite ✓

Other barriers Fear of burglary ✓ ✓

Distance from shelters ✓ ✓ ✓

Locational barrier ✓ ✓

False sense of security ✓ ✓ ✓

No previous experience of cyclone ✓ ✓

Source: [1, 14, 29, 30, 31, 33, 40, 44, 48, 47, 51, 50, 46, 45, 43, 52].

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

Ikeda [33], Chowdhury et al. [14] and Paul et al. [50] stated that dis-tance is the major impediment in seeking refuge at the shelters. During cy-clones Gorky, Sidr and Aila the mean distance from the shelters was 5.00km, 3.00 km and 3.14 km respectively [1,52]. Most people in Bangladeshdo not go to shelters until the surge water enters their courtyards, hencewalking more than 2.0 km becomes difficult [14]. Most locals can travel

5

between 0.8 km and 1.0 km to a shelter during a cyclone [47]. Unelevatedand undulating road to the shelter coupled with surge water obstructedpeople 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 inade-quate number of shelters. During Cyclone Gorky, death rates were highestdue to the absence of shelters or insufficient space [29]. A study conductedby Paul [46] found that 17% of the respondents came back home due toover-crowdedness during Cyclone Sidr. During Cyclone Aila, over 77% ofthe non-evacuees came back home due to insufficient space in the cycloneshelter [1]. In most cases, cyclone shelters were occupied by local elites,hence the poor people seeked refuge on embankments, trees and other vul-nerable and low-lying areas [52]. A study by Parvin et al. [44] shows that69% of the evacuees sought refuge on elevated land such as road andembankment.

False sense of security is common among the inhabitants of the coastalarea [30]. A study by Paul [43] shows that 12% of the non-evacuee respon-dents believed that their houses would withstand cyclone surges duringSidr. A study by Paul [46] shows that around 12% of the non-evacuee re-spondents refuse to go to the cyclone shelter since they thought it wouldnot be able to withstand the intensity of the cyclone. This could be attrib-uted to the fact that villagers adopt indigenous practices to reinforce theirhouses and avoid inundation from storm surges. A study by Paul et al.[50] identified that some crude housing reinforcing measures were usedto withstand the impacts of cyclones. Almost all houses of the respondentswere submerged during Cyclone Gorky. To avoid inundation induced bystorm surges, around 73% of the respondents raised their yards at varied in-tervals. One-third of the respondents installed pillars to strengthen thefoundations and another one-third tied their houses with strong ropes totrees. Such practices may have provided them false sense of security duringcyclones.

A section of the people believes that cyclones are a will of God or pun-ishment [2]. Religious belief is embedded into the culture of the coastal

Page 6: Seeking Shelter: The factors that influence refuge since

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

society of Bangladesh. In a conservative society, Muslim women are notallowed to leave home to take shelter because of purdah norm4 [67, 8].The practice of purdah is strictly observed in the coastal areas ofBangladesh, where men still exhibit reluctance in taking their wives tothe cyclone shelter [68]. Many women do not go to the cyclone shelter asthey do not have separate toilets. A study by Ahsan et al. [1] shows thatmore than 35% of the non-evacuee respondents did not seek refuge to thecyclone shelters due to the lack of separate toilets and spaces for men andwomen. Poor condition and lack of adequate facilities of the cyclone shel-ters such as limited number of latrines, provision for drinking water andstorage for food have also been considered as important factors.

One of the major socio-economic factors is the fear of losing belongingsif left abandoned during disasters. The majority of the coastal population isliving at or below the poverty line hence losing their belongings would con-tribute to extreme poverty, hunger and perhaps even death. A study byHaque and Blair [29] shows that 57.6% and 31.6% of the non-evacuee re-spondents residing in Halishahar and Jahanabaj areas feared that theirhousehold items will be looted, and did not go to the cyclone shelter.Households whose principal source of income is livestock rearing exhibitedreluctance to go to the cyclone shelter since there were no killas nearby. Astudy by Ahsan et al. [1] shows that 72% of the respondents did not go tothe cyclone shelter since there were no killas nearby.

6. Use of shelters during an emergency: Cyclone Amphan

Key factors that contributed to refusal of seeking refuge at the shelterduring Cyclone Amphan are: fear of losing property, inadequate shelter,poor conditions and false sense of security or people living in a relativelystrongly built house.

The majority of the respondents did not go to the cyclone shelter due tofear of losing their properties. 35 (17%) of the non-evacuee respondents didnot go to the cyclone shelter because they thought staying at home wouldhelp prevent their houses being destroyed or looted. Out of 35 respondents,four respondents from Bodropara anchored their boats at khal5 and tookrefuge inside their boats with family members and valuable belongings.14 male respondents from Char Anda sent their wives and children to cy-clone shelters and stayed at home to protect their houses. One respondentfrom Char Anda reported that staying at home may not help to protect hishouse from being swept away; however, it would give him peace knowingthat he tried till the end. He kept a rope at hand to tie oneself against a treeprovided the wind intensifies. 17 male respondents from Bodropara senttheir wives and children to another village with the trawler and they stayedat home and observed the weather patterns. When the situation worsened,they took refuge at the cyclone shelter. One respondent reported comingback from the shelter to protect house being damaged and swept away.Informal conversations with the villagers of Bodropara revealed that, onrequest of the Upazila Nirbahi Officer, fishing trawlers were sent with thecoast guards to evacuate the people of Bodropara and the adjoining vil-lages. These attempts of saving property reflect how the fear of losing prop-erty influences people to evacuate to the shelter.

Partial evacuation was found to be one of the indicators of fear of losingproperty [9]. It also reflects the male dominance in the rural areas indicat-ing that women are not capable of providing assistance in times of disaster.Informal conversationwith the female respondents revealed that theyman-aged to take children and livestock, alongwith some dry food, to the shelterduring cyclones in the past. This reflects that women play a huge part in di-saster risk reduction.

The survey results revealed that, even if a shelter is 1 km away from theresidence, some people exhibited reluctance in seeking refuge at the shel-ter. This is particularly true for the people with small poultry farm busi-nesses, since it is difficult to carry them to the shelter. It reflects that fear

4 Purdah literally means “a curtain” and is used figuratively to indicate seclusion of womenfrom men from any kinds of engagement with regard to conducting socioeconomic roles out-side of a household.

5 Canal

6

of losing asset dominates over the distance to shelters. 30 respondents(14%) reluctantly took refuge at their neighbours' houses at the last mo-ment. Among them, 13 respondents have small poultry farms and 12 re-spondents sought refuge to the neighbour's place closer to their houses toobserve their property. Three respondents from Purbo Kalinagar and twofrom Singkhali sought refuge at the neighbour's house for the fear ofbeing infected with COVID-19 at the shelter. The number of confirmedCOVID-19 caseswas 35,585 in Bangladesh until 25May 2020 [65]. Accord-ing to the respondents, some shelters providedmasks to the evacuees, how-ever, very few were wearing them.

About 17% (36) respondents did not go to the cyclone shelter. Among36 respondents, nine respondents have pucca house and 27 respondents re-inforced houses with woods with elevated plinths. Fig. 5 shows an exampleof houses in the study area.

The majority of the respondents have improved their housing conditionparticularly following the cyclone of 1991. Jhupri houses, where the roofs ofthe houses were previously made from the leaves of Nipa Palm locallyknown as Golpata, are now roofed with corrugated tin. None of the partic-ipants have houses made from the leaves of Golpata or straw. Hence thevery definition of jhupri has been changed. Such improvements have beenreflected in the housing condition of the study areas. Table 4 shows thenumber of katcha, pucca, semi-pucca and jhupri houses6 across the studyareas (See Table 5).

Insightful data were received when the villagers were asked on choos-ing the best disaster management measure. 15 respondents believe thatmending or constructing embankments would solely save them during cy-clones. 65 respondents believe that construction of cyclone shelters and em-bankmentswould save their lives and livestock. 25 respondents believe thatconstructions of embankments and cyclone shelters would save them fromdisaster and would provide an opportunity for their children to go toschools since the government constructs mostly school cum cyclone shel-ters. Others (105 respondents) preferred receivingfinancial assistance to re-inforce or build their houses with bricks. Irrespective of the presence ofmulti-purpose cyclone shelters7 formore than two decades, people perceivethem solely as cyclone shelters.

67 respondents (32%) observed various environmental cues such as in-tensity of the wind, colour of the sky and height of storm surge, irrespectiveof the evacuation orders and meteorological warnings prior to evacuation.This group includes people who stayed at home by sending their wives andchildren to the cyclone shelters and people who sought refuge at theirneighbour's house and cyclone shelters at the last moment. Observing theweather pattern until the situationworsens reflects that the cyclone sheltersor other places are considered as the last resort during an emergency. Onerespondent from Singkhali reported about leaving the house with familymembers to the neighbours' place when the wind intensified.

56 respondents (27%) stated that inadequate cyclone shelters causedthem trouble during an emergency. Among the 56 respondents, 20 respon-dents returned home since the cyclone shelters were full or crowded and 13respondents went to neighbour's house. 23 respondents went to multipleplaces to seek refuge prior to disasters. One respondent from Gaziparatook his wife to a cyclone shelter in another village since the shelter in hisvillage was under construction. He returned home and took his livestockto a nearby killa and tied himself to a tree near his house. There is onlyone cyclone shelter available for Purbo Kalinagar, Gazipara and CharAnda villages (Table 4). In Gazipara, there is one cyclone shelter under con-struction which was not completed due to the COVID-19 outbreak. Manypeople sought refuge at the under-constructed cyclone shelter. In CharAnda, there are two cyclone shelters; one in a very poor condition andthe other one is in a better condition, however too small to accommodateall villagers.

6 Katcha: houses built with corrugulated tin sheet; pucca: houses built with brick and ce-ment; semi-pucca: houses built with corrulgulated tin sheet with brick floor; jhupri: houses builtwith bamboo, straw and mud.

7 Multi-purpose shelters also serve as economic and social benefits serving as schools, hos-pitals and government offices during non-disaster periods.

Page 7: Seeking Shelter: The factors that influence refuge since

Fig. 5. Example of brick-built home, jhupri and semi-pucca houses. Source: Authors.

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

About 23% (49) of the respondents fully evacuated with their familymembers to the cyclone shelters. Out of them, 13 respondents kept theirlivestock at neighbour's place and went to the cyclone shelter.

Informal conversations with female respondents revealed that purdah isno longer considered as a major hindrance while seeking refuge at the cy-clone shelter. All female respondents (75 respondents) prioritised savingtheir lives over traditional belief. This could be attributed to the fact thatwomen and children were mostly sent to the cyclone shelters in the studyareas. However, they exhibited discomfort of staying inside overcrowdedcyclone shelters. One female respondent from Purbo Kalinagar reported

Table 4Types of housing across the study areas.

Village Pucca Katcha

Mainapur 6 18Purbo Kalinagar (No record) 401Bodro Para (No record) 350Angulkata 20 485Gazipara 15 189Char Anda 15 485Singkhali 26 69

Source: Upazila Parishad Members, Upazila Nirbahi Officer.

Table 5Cyclone shelters in the study areas.

District Upazila Union Village

Jessore Keshabpur Sufalakathi MainapurSatkhira Shyamnager Munshiganj Purbo KalinaBhola Char Fasson Dal Char Bodro paraBarguna Amtali Gulishakhali AngulkataKhulna Koyra Uttar Bedkashi GaziparaPatuakhali Rangabali Char Montaz Char AndaPirojpur Nazirpur Malikhali Singkhali

Source: Field survey, 5 June 2020.⁎ Note: This is a school cum cyclone shelter.

7

feeling suffocated inside the crowded shelter. They expressed concernover having less or no separate toilets and space for men and women. Themajority of the female respondents were escorted by husbands to the cy-clone shelter during an emergency. This reflects that women are notallowed to go outside of their houses alone.

7. Conclusions and recommendations

The study explored factors which influence people in seeking refuge atcyclone shelters during the major cyclones Gorky, Sidr, Aila and compared

Semi-pucca Semi-pucca and pucca Jhupri

30 (No record) (No record)(No record) 52 (No record)(No record) (No record) (No record)60 (No record) (No record)18 (No record) 4(No record) (No record) (No record)139 (No record) (No record)

Cyclone shelters (number) Eroded shelters (number)

1 No recordgar 0 No record

0 11 10 No record1 11⁎ No record

Page 8: Seeking Shelter: The factors that influence refuge since

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

them with the most recent Cyclone Amphan to identify major factors thatinfluences them in making decision to evacuate. Despite significant im-provement in disaster management in Bangladesh, variables influencingnon-compliance with evacuation orders have not changed significantlyover the last 30 years.

Inadequate cyclone shelters in proximity of the villages are the majordeterrent to shelter evacuation. There are three major factors for not seek-ing 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 relativelystrongly built house. Fear of losing property is a dominant factor. The ma-jority of people considered evacuation as a last resort. Partial evacuation,observing environmental cues prior to evacuation and seeking refuge at cy-clone shelter and neighbours' houses at the last moment, is indirectly asso-ciated with the fear of losing property. The study suggested severalrecommendations, as follows:

7.1. Housing condition and compliance with national building codes

Housing conditions of the village could be integrated into planning andselecting sites for cyclone shelters. As cyclones aremost likely to intensify asa result of climate change, design of cyclone shelters and buildings in thecyclone hazard areas need to compliance with national building codes.Other than shelters, other adaptation options such as embankments, naturebased solutions (i.e. mangrove) should incorporate aspects of changing cli-matic patterns such as intensification of surge water and wind speed.Existing building structures should be reinforced and additional buildingstructures, if required, should be constructed in proximity to larger commu-nities so that they can keep assets and seek refuge at the cyclone shelter dur-ing cyclones.

7.2. Assessment of cyclone exposure and shelter needs

The study found that the houses in Mainapur village are newly buildand well-constructed and thus people may not need to go to the formalevacuation place. A proper needs assessment would ensure number of shel-ters necessary in the risk areas. The hazard assessment also inaccurate dueto lack of high-resolution topography and advance modelling. With the ad-vancement in themodelling and current data, risk-based coastal inundationinformation would enable them to take appropriate action.

7.3. Gender responsive shelter management plan

Women's and men's vulnerability to the impact of severe cyclone eventsis determined not only by biology but also by differences in their cultural,social roles and responsibilities. Men andwomen are affectedwith differentvulnerabilities, particularly, when it comes to safeguarding their food secu-rity, livelihoods and other social issues. Acquiring information on the maleto female ratio of a village and their vulnerabilitieswould help to effectivelyintegrate gender-sensitive aspects (separate toilets and space for male andfemale) into the construction and develop facilities in the cyclone shelter.

7.4. Cyclone shelter management information system and budget

Government and nongovernmental organisations should intensify ef-forts to obtain international funds to reinforce dilapidated houses, particu-larly located in risk and high-risk zones, and establishing a centre or astructure to keep their assets during an emergency.

7.5. Awareness campaign and exercise

A community-based programme to raise awareness of appropriate ac-tions during an emergency (e.g., remaining calm, evacuate immediately fol-lowing receiving warning messages, taking dry food prior to evacuating tothe cyclone shelter) should be conducted. Such messages can be dissemi-nated through popular theatre (theatre which delivers messages to peoplein their language), street theatre and FM radio.

8

7.6. Ensuring livestock safety for communities

The construction of killas in proximity to cyclone shelters and largercommunities should be intensified and adequate measures should be un-dertaken to minimise the threat of theft during cyclones. People often feeldiscomfort for leaving livestock in outdoors and at risk of being stolen. Apilot initiative by the government may be undertaken integrating fencingand sheds into the design of killas. During a non-emergency period, thesekillas can be used as local markets or local people can take their livestockthere during rainy seasons.

Declaration of Competing Interest

The authors declare that they have no known competing financial inter-ests or personal relationships that could have appeared to influence thework reported in this paper.

Acknowledgments

Wewould like to thank to Dr. S. M. Munjurul Hannan Khan (AdditionalSecretary, Ministry of Environment and Forest), Mr. Netai Chandra DeySarker (Deputy Director, Department of Disaster Management), MahmoodHussain, (Associate Professor, San Francisco State University) andMs. SiratMahmuda (PhD Candidate, The University of Queensland) for their kindsupport.

References

[1] Ahsan MN, Takeuchi K, Vink K, Warner J. Factors affecting the evacuation decisions ofcoastal households during Cyclone Aila in Bangladesh. Environ Hazard. 2015;15(1):1–27.

[2] Nasif Ahsan Md, Khatun Amina, IslamMd Sariful, Vink Karina, Ohara Miho, FakhruddinBapon SHM. Preferences for improved early warning services among coastal communi-ties at risk in cyclone prone south-west region of Bangladesh. Progr Disast Sci. 2020;5.

[3] Ali A. Vulnerability of Bangladesh to climate change and sea level rise through tropicalcyclones and storm surges. Water Air Soil Pollut. 1996;92:171–9.

[5] Alam E, Dominey-Howes D. A new catalogue of tropical cyclones of the northern Bay ofBengal and the distribution and effects of selected landfalling events in Bangladesh. Int JClimatol. 2015;35:801–35. https://doi.org/10.1002/joc.4035.

[7] Ahsan MN, Takeuchi K, Vink K, Ohara M. A systematic review of the factors affectingthe cyclone evacuation decision process in Bangladesh. J Disaster Res. 2016;11(4):740–51.

[8] Bern C, Sniezek J, Mathbor GM, Siddiqi MS, Ronmans C, Chowdhury AMR, et al. Riskfactors for mortality in the Bangladesh cyclone of 1991. Bull World Health Organ.1993;71(1):73–8.

[9] Bangladesh Bureau of Statistics. District statistics 2011: Khulna. Retrieved on 05 Janu-ary 2021 from http://www.bbs.gov.bd/WebTestApplication/userfiles/Image/District%20Statistics/Khulna.pdf.; 2013.

[14] Chowdhury AMR, Bhuyia AU, Choudhury AY, Sen R. The Bangladesh Cyclone of 1991:why so many people died. Disasters. 1993;17(4):291–304.

[16] CARE. Joint Need Assessment Working Group. Retrieved on 01 June 2020 from https://reliefweb.int/sites/reliefweb.int/files/resources/cyclone_amphan_joint_needs_assessment_final_draft_31052020.pdf; 2020.

[20] Dash B. Evacuation for cyclone Phailin and Hudhud: revaluation of success. Econ PolWkly. 2016;51(53):130–7.

[22] Disaster Management Bureau. National plan for disaster management 2010–2015. Di-saster Management and Relief Division; 2010.

[24] EM-DAT. Top 10 natural disasters in Bangladesh during 1900-2014. Retrieved on 01June 2020 from http://www.emdat.be/result-country-profile; 2020.

[25] Haque U, Hashizume M, Kolivras KN, Overgaard HJ, Das B, Yamamoto T. Reduceddeath rates from cyclones in Bangladesh: What more needs to be done? Bulletin ofthe World Health Organisation. 2012;90(2):150–6.

[26] Government of Bangladesh. Emergency preparedness plan for cyclone in Bangladesh.April 2013. Retrieved on 01 June 2020 from http://kmp.dmic.org.bd/bitstream/handle/123456789/144/Emergency20Preparedness20Plan_Cyclone20Bangladesh202013.pdf?sequence=1; 2013.

[27] GoB (Government of the People'’s Republic of Bangladesh). Cyclone Sidr in Bangladesh:damage, loss, and needs assessment for disaster recovery and reconstruction. Dhaka:GoB; 2008.

[28] Haider MZ, Ahmed MF. Multipurpose uses of cyclone shelters: quest for shelter sustain-ability and community development. Int J Disast Risk Reduct. 2014;9(1) 1-.

[29] Haque CE, Blair D. Vulnerability to tropical cyclones: Evidence from the April 1991 cy-clone in coastal Bangladesh. Disasters. 1992;16(3):217–29.

[30] Haque CE. Climatic hazards warning process in experience of, and lessons from the AprilCyclone Bangladesh: 1991. Environ Manag. 1995;19(5):719–34. https://link.springer.com/article/10.1007/BF02471954.

Page 9: Seeking Shelter: The factors that influence refuge since

T. Hadi et al. Progress in Disaster Science 11 (2021) 100179

[31] Haque CE. Atmospheric hazards preparedness in Bangladesh: a study of warning, adjust-ments and recovery from the April 1991 Cyclone. Natural Hazrad. 1997;16(2–3):181–202. https://link.springer.com/article/10.1023/A:1007942712838.

[33] Ikeda K. Gender differences in human loss and vulnerability in natural disasters: a casestudy from Bangladesh. Indian J Gend Stud. 1995;2(2):171–93.

[36] Khan MR, Rhaman A. Relationship approach to disaster management in Bangladesh:critical policy assessment. Nat Hazards. 2007;41(1):359–78.

[38] Miyaji M, Okazaki K, Ochiai C. A study on the use of cyclone shelters in Bangladesh.Japan Architect Rev. 2020;3:590–600. https://doi.org/10.1002/2475-8876.12177.

[40] Mallick B, Vogt J. Social supremacy and its role in local level disaster mitigation plan-ning in Bangladesh. Disast Prevent Manag. 2011;20(5):543–56.

[41] Matsuda I. Loss of human lives induced by the Cyclone of 29–30 April, 1991 inBangladesh. GeoJournal. 1993;31(4):319–25.

[42] Nicholls RJ. Synthesis of vulnerability analysis studies. In: Beukenkamp P, et al, editors.Proceedings of the world coast conference 1993. The Hafue: National Institute forCoastal and Marine Management, Coastal Zone Management Center Publication;1995. p. 181–216.

[43] Paul BK. Factors affecting evacuation behavior: the case of 2007 Cyclone Sidr,Bangladesh. Professional Geographer. 2012;64(3):401–14.

[44] Parvin GA, Sakamoto M, Shaw R, Nakagawa H, Sadik MS. Evacuation scenarios of Cy-clone Aila in Bangladesh: investigating the factors influencing evacuation decision anddestination. Progr Disast Sci. 2019;2:1–13.

[45] Paul BK. Why relatively fewer people died? The case of Bangladesh’s Cyclone Sidr. NatHazards. 2009;50(2):289–304.

[46] Paul BK. Hazard warnings and compliance with evacuation orders: the case ofBangladesh's cyclone Sidr. Manhattan: Department of Geography, Kansas State Univer-sity; 2008.

[47] Paul A, Rahman MM. Cyclone mitigation perspectives in the Islands of Bangladesh: acase of Sandwip and Hatia islands. Coast Manag. 2006;34(2):199–215.

[48] Paul BK, Dutt S. Hazard warnings and responses to evacuation orders: the case ofBangladesh’s cyclone Sidr. Geogr Rev. 2011;100(3):336–55.

[50] Paul BK, Rashid H, Islam MS, Hunt LM. Cyclone evacuation in Bangladesh: tropical cy-clones Gorky (1991)vs. Sidr (2007). Environ Hazard. 2010;9(1):89–101.

9

[51] Paul SK, Routray JK. An analysis of the causes of non-responses to cyclone warnings andthe use of indigenous knowledge for cyclone forecasting in Bangladesh. In: Filho WL, ed-itor. Climate change and disaster risk management; 2011. p. 15–39.

[52] Paul SK. Determinants of evacuation response to cyclone warning in coastal areas ofBangladesh: a comparative study. Orinetal Geographer. 2014;55(1–2):55–84.

[54] Sorensen JH, Mileti DS. Public response to emergency warnings. Reston, VA: USGS;1987.

[56] Walton-Ellery Sandie. A review of the Cyclone Aila response 2009: IFRC-led EmergencyShelter Coordination Group; 2009 Dhaka.

[58] Saha CK. Dynamics of disaster-induced risk in southwestern coastal Bangladesh: an anal-ysis on tropical Cyclone Aila 2009. Nat Hazards. 2015;75(1):727–54.

[59] Saha SK, James H. Reasons for non-compliance with cyclone evacuation orders inBangladesh. Int J Disast Risk Reduct. 2017;21:196–204.

[61] Shawson A. Cyclone Amphan: Bangladesh may face losses worth Tk12,744cr. DhakaTribune. 2020(May 20) Retrieved on November 8, 2020 from https://www.dhakatribune.com/bangladesh/2020/05/20/cyclone-amphan-bangladesh-may-face-a-loss-of-tk12.

[62] Shamsuddoha M, Chowdhury RK. Climate change impact and disaster vulnerabilities inthe coastal areas of Bangladesh. Dhaka: Coast Trust; 2007.

[63] Tonkin, Taylor IL. Tropical cyclone disaster impact modelling – impact assessment forChittagong and Cox’s Bazar. Auckland: Bangladesh; 2020.

[64] Velotti L, Trainor JE, Engel K, Torres M, Myamoto T. Beyond vertical evacuation: re-search consideration for a comprehensive vertical protection strategy. Int J MassEmerg Disasters. 2013;31(1):60–77.

[65] World Health Organization. COVID-19: situation report No 13. Retrieved on 06 April2021 from https://www.who.int/docs/default-source/searo/bangladesh/covid-19-who-bangladesh-situation-reports/who-ban-covid-19-sitrep-13-20200525.pdf?sfvrsn=a15591c0_4; 2020.

[66] Petrolia D, Bhattacharjee S. Why don’t coastal residents choose to evacuate for hurri-canes? Coastal Management. 2010;38(2):97–112.

[67] Feldman S, McCarthy FE. Purdah and changing patterns of social control among ruralwomen in Bangladesh. Journal of Marriage and the Family. 1983;45(4):949–59.

[68] Kabir R. Bangladesh: surviving the cyclone is not enough. Stop Disasters. 1995;24.