Climate Risk Management When the disaster strikes: Gendered (im)mobility in Bangladesh

Gender influences people’s behaviour in various ways. This study investigates gendered (im) mobility during cyclone strikes in Bangladesh. During such strikes people have described being unable to move away from environmentally high-risk locations and situations. The Q-based Discourse Analysis used by this study shows how and why gender-roles (im)mobilised people in three coastal locations during the cyclones. People (and especially women) explained that failing to evacuate to the cyclone shelters when a disaster strikes was not uncommon. Gender, or feminine and masculine social roles, played a significant role in these evacuation decisions while facilitating or constraining their mobility. The gendered subjectivities presented different accepted social behaviours and spaces for women and men. In this way, immobility (social, psy- chological, and geographical) was strongly gendered. Masculine roles were expected to be brave and protective, while female ‘mobility’ could be risky. Women’s mobility therefore ended up being constrained to the home. In other words, when the disaster strikes, everyone did not have the same ability to move. These empirical insights are important to inform climate policy in a way that it better supports vulnerable populations worldwide as they confront global environmental changes today and in the future.


Introduction
The uneven impacts on women from global environmental changes put gender at the frontline of all three 2015 climate agendas (Cutter, 1995(Cutter, , 2017Wahlström, 2015;UNFCCC, 2015). Empirical evidence illustrating gendered vulnerabilities to environmental stress will be crucial to the development of the robust policy frameworks. These scenarios include immobility as vulnerable populations may be unable to escape environmental risks. The literature on climate-induced immobility largely relates to Trapped Populations 2 (Ayeb-Karlsson et al., 2018) -a concept put forward as an effective policy tool to safeguard vulnerable people in a climate changed world (Foresight, 2011;Black et al., 2011Black et al., , 2013Black and Collyer, 2014). The emergence of an immobility perspective expanded the environmental migration debate. People are known to migrate away from environmentally risky locations and situations but may also be unable to move (Nawrotzki and DeWaard, 2018;Farbotko and McMichael, 2019;Thornton et al., 2019). Initially, people were proposed to be involuntary 'trapped' but are now recognised to alternatively be voluntary immobile (or choose to stay). However, as noted in the conceptualisation of the term, distinguishing between those who choose and those who are forced to stay behind is extremely complex (Black et al., 2013;Black and Collyer, 2014;Ayeb-Karlsson et al., 2018).
The conceptual idea of Trapped Populations can be useful to better understand gender vulnerability. To date, the term was primarily used to describe rural non-migration behaviour due to financial constraints, but the body of literature is rapidly growing Nawrotzki and DeWaard, 2018;Farbotko and McMichael, 2019;Thornton et al., 2019;Zickgraf, 2019;Blondin, 2020). In line with this, more people-centred studies (Ayeb-Karlsson et al., 2016;Tebboth et al., 2019;Conway et al., 2019) better acknowledging the human aspects of climate change, including internal psychological, subjective and emotional processes, are encouraged Tschakert et al., 2019;Ayeb-Karlsson et al., 2020). The UNFCCC climate policy space has similarly observed the conceptual creation and development of 'Non-Economic Loss and Damage' (UNFCCC, 2013;Boyd et al., 2017;Tschakert et al., 2019).
This article investigates gendered (im)mobility, or how gender influences people's evacuation behaviour, during cyclone strikes in Bangladesh. Gender roles are important psychosocial factors that can enable and constrain (im)mobility decisions. The analysis supports an explanation of why people (often women due to power relations) fail to evacuate when a cyclone strikes. The Q-based Discourse Analysis (DA) examines people's subjective values around hazards and risks and their connections to reasonings on seeking shelter or staying behind.

(Im)mobility and gendered disasters
Disaster immobility is mainly framed around 'non-evacuation behaviour' and 'involuntary immobility' during specific events such as US Hurricanes Katrina and Rita (Elliott and Pais, 2006;Stein et al., 2010;Thiede and Brown, 2013), while Trapped Populations is slowly emerging in the disaster literature (Penning-Rowsell et al., 2013;Logan et al., 2016;McCaughey et al., 2018).
Women often suffer disproportionally from environmental hazards due to social inequality. Effects observed in everything from food insecurity, mental (and physical) ill health, and gender-based violence during and after climatic stress. This conceptually aligns with the critical understanding of hazards turning into disasters when they negatively impact vulnerable populations. Hazard severity is therefore a reflection of location, intensity and number of vulnerable people hit by the disaster (Blaikie et al., 1994;Kelman, 2019;Ayeb-Karlsson et al., 2019a). Food hierarchies may pose greater risks of malnutrition and water scarcity for children and women (especially when pregnant or breastfeeding) during population movements or displacement (Rivers, 1982;Cutter, 2017;Watts et al., 2018Watts et al., , 2019. Similarly, income inequality and financial dependence make it more difficult for women, children, disabled and elderly to bounce back after hazards, or independently decide whether to migrate, stay put, or evacuate (Enarson, 1998;Jordan, 2018;Ayeb-Karlsson et al., 2019a. In Bangladesh, women, children and the elderly are often left behind while men migrate towards the cities. Women also tend to delay more in responding to cyclone warnings or are left behind while others evacuate (Bhatta et al., 2015;Ahsan et al., 2016). Studies show that most women cannot swim, that clothing and responsibilities for children restrict their movements, and that fear of sexual and physical abuse increases outside the home (Sultana, 2007;Fordham, 2012;Jordan, 2018).
More women than men died in Bangladeshi cyclone strikes such as Sidr in 2007 and Gorky in 1991. 90% of the fatalities in 1991 cyclone Gorky, for example, were women (Chowdhury et al., 1993;Alam and Rahman, 2014). Gender-based disaster mortality differences extend to a global level, while acknowledging potential underestimates and uncertain approximates (Brunkard et al., 2008;Jonkman et al., 2009;Bradshaw and Fordham, 2013). Some disaster mortality investigations show that women face higher risks of dying during cyclones in the global south (Doocy et al., 2013;Cutter, 2017). Critical studies analysing the root causes of gender vulnerability will support the understanding of diverse vulnerabilities (Blaikie et al., 1994;Cannon, 2000;Juran and Trivedi, 2015). Structural social and political processes coupled with socio-cultural patriarchal power systems generally enforce gender imbalances (Anbarci et al., 2005;Neumayer and Plümper, 2007;Rhodes et al., 2010). Gender vulnerability (including socially and geographically reduced mobility) must be understood as part of larger psychosocial factors including gender identity, patriarchal attitudes, religious or traditional practices, shame and honour, gendered division of labour, and internal psychological processes such as feelings and emotions (Rashid and Michaud, 2000;Stephens et al., 2013). When access to disaster preparedness training, early warning messages, shelter space or post-disaster relief is gendered, social positions determine life and death (Fordham, 2012;Nadiruzzaman and Wrathall, 2015).
This study addresses some important gaps in the current literature body. Most investigations exploring gendered disasters and vulnerabilities in Bangladesh tend to be overly simplistic, presenting linear comparisons between women and men. Women are often perceived as one coherent social group with collectively shared vulnerability. This idea neglects vulnerability differences due to socio-economic status, age, religion, social background, marginalisation and stigma. Outlining these nuances around gendered disaster (im)mobility and vulnerability contradicts such generalisation. This Q-based Discourse Analysis explores discursive values around cyclone evacuation in Bangladesh and whether people felt constrained, socially immobile or 'trapped' by their gendered subjectivity.

Method and study sites
Q-methodology (Stephenson, 1935(Stephenson, , 1986Brown, 1980Brown, , 1996Stenner, 2008;Watts and Stenner, 2012) encapsulates people's subjectivities around a topic or within a social or geographical group. Q is often called the 'science of subjectivity' (Stephenson, 1935(Stephenson, , 1986, where subjectivity is understood as 'how an individual understands oneself within a social context, and what it momentarily feels like to exist in this set of social relations, space and place' (Morales and Harris, 2014:706). To date, studies using Q to understand viewpoints on climate change policy, environmental risks, migration or immobility are still scarce (Niemeyer et al., 2005;Ockwell, 2008;Wolf et al., 2009;Morinière and Hamza, 2012;Oakes, 2019;Ayeb-Karlsson et al., 2020).
After Q-methodology was introduced within Human Geography in the early 2000s (by Robbins and Krueger, 2000) a group critical discourse scholars proposing more reflexive and innovative ways to execute Q emerged (López-i-Gelats et al., 2009;Duenckmann, 2010;Jepson et al., 2012;Sneegas, 2020;Ayeb-Karlsson et al., 2020). The group challenged propositions of Qmethodology as an 'objective' identification of subjectivities and argued that to be of real value it must be combined with (Critical) Discourse Analysis (see, for example, Wodak, 2011;Fairclough, 2013). Yet, Q-studies attempting to combine Q and Discourse Analysis are still few (but include Dryzek, 1994;Stainton-Rogers, 1997;Barry and Proops, 1999;Ockwell, 2008;Wolf et al., 2009;Morinière and Hamza, 2012;Ayeb-Karlsson et al., 2020). Some Q scholars suggest that Discourse Analysis has been excluded from Qcircles 'discursively' -according to similar forces separating binary opposites such as black-white, woman-man, south-north, messiness-tidiness, or DA-Q (see, for example, Sneegas, 2020 andAyeb-Karlsson et al., 2020). This study applies an innovative methodological approach that combines Q and Critical Discourse Analysis to empirically investigate climate-induced (im)mobility (first carried out in Ayeb-Karlsson et al., 2020).
The Q-element in this study followed a standardised Q-procedure of a Q-set creation, Q-sorting exercise, and factor analysis in PQMethod 3 (Watts and Stenner, 2012;Sneegas, 2020). The Q-analysis was then combined with an extensive textual Critical Discourse Analysis of the unstructured post-sorting interview and survey responses (Ayeb-Karlsson et al., 2020). 40 statements generated from 100 qualitative field quotes from qualitative research sessions conducted in the three study sites between 2014 and 2015 on gendered disaster (im)mobility were used in the Q-sorting activity (Fig. 1). The 100 original field statements were grouped into re-occurring topics, themes and storylines, to achieve an effective, balanced and well-covered final Q-set. The Q-sorting activity was combined with an (im)mobility survey 4 that offered a discursive understanding of the informant's (im)mobility history and demographic background.
The recorded and factor analysed Q-sorts identified 15 different discourse (or factor) groups 5 (see Table 1). The three study sites' 62 Q-sorts (in total 186 snowball sampled informants of 93 women and 93 men) were first factor analysed together as a whole, and then grouped and analysed according to study sites. In addition to ensuring a gender-balance, the selected informants reflected the socio-economic, religious, age, and livelihood makeup of the study sites. The discursive narratives appearing in the 15 discourse groups identified, (five in Dalbanga South, four in Mazer Char and six in Gabtola) were critically analysed textually as well as with the help of the informant details registered in the survey questionnaire.
The selection of study sites used was justified by the socio-environmental diversity and different experiences of cyclone strikes in each. Consultations with local stakeholders and people during initial site visits guided the selection. Dalbanga South had not suffered extensive losses of human lives during recent cyclones, but reported lacking disaster preparedness measures such as a functional early warning system and their only cyclone shelter could not accommodate the whole village. Mazer Char islanders expressed being cut off or isolated from the mainland during cyclone strikes as people could not reach the mainland by boat. Meanwhile, people in Gabtola described having sufficient shelters, but feeling traumatised by having lost approximately one third of the village population in cyclone Sidr in 2007 (Nadiruzzaman and Wrathall, 2015;Ayeb-Karlsson et al., 2016). Q-grid used in study. As recommended by Watts and Stenner (2012:80) and previously applied in an empirical Q-based Discourse Analysis to investigate climate-induced (im)mobility, an 11-point (-5 to +5) distribution Q-grid was used for the Q-sorting activities (Ayeb-Karlsson et al., 2020: 6). Table 1 Q-Statement overview. List of the 40 Q-statements used in this study and how each discourse group ranked them during the Q-sort activity. The selected statements were based on the qualitative fieldwork sessions carried out in the three study sites between the years of 2014 and 2015.
No Q-statements  One can never know whether you will reach the shelter safely. The outcome is not in my hands.

Disentangling gendered (im)mobility
The discourse groups were identified through centroid factor analysis to detect factor patterns or inter-correlation between the Qsorts, and rotated using Varimax in PQMethod (Watts and Stenner, 2012:96-100). Varimax rotation ensured that each Q-sort (or participant sorting of Q-statements) only loaded on or reflected the viewpoints of one factor group. The significant factor loading was calculated through the equation (2.58 × (1/√ No. of Q-sorts in Q-set e.g. (2.58 × (1/√62) = 0.33). Q-sorts loading on more than one factor group (cofounded) or non-significant Q-sorts (below 0.33) were not selected for further analysis. Eigenvalue above 1.00 served as selection criteria for factor extraction (Watts and Stenner, 2012:127-128, 197-199).
Out of the 15 discourses, a remarkable 14 were predominantly associated with either female or male participants (Fig. 2). The analysis will refer to male and female discourses (groups predominated by male or female participants) while acknowledging that gendered characteristics are social constructs reproduced by men and women. The groups named after the Q-based Discourse Analysis represent different discourses of cyclone evacuation (im)mobility (Table 2). For transparency, the interview extracts are referenced in brackets with informant number, statement number and sort value e.g. (informant: statement sort value) or (014:2 +5) for informants and (2 +5) for group ranking.

Disaster preparedness can save lives but women are still not safe
Five discourses were identified in Dalbanga South where the selected un-rotated factors explain 48% of the study variance. 34 of the 62 participants were highly associated with the discourses. Study variance is the proportion of shared meaning (or factors) that the factor groups explain. Factor analysis (or the data reduction technique) inspects the correlation that identifies patterns of regularity in the Q-sorts (or viewpoints) expressed by the Q-participants (Watts and Stenner, 2012:98).
The planner (female) discourse group expressed the feeling that warning messages and evacuation save lives (Fig. 3). References to early warning messages included natural and technological systems. Natural early warning systems include practices, observations and indigenous knowledge used to estimate or project an oncoming cyclone. The natural warning systems used combine observations of changes in weather, movements of animals, and directions of wind and rivers: Q1: Our elders are very experienced. They understand from the nature what is happening (014:40 +4). Q2: We start preparing when we get the warning signal (013:1 +4).  Table 2 Discourse groups' survey informant background. An overview of the discourse groups' informant background, as well as their migration and evacuation histories that were captured in the survey questionnaire and post-sorting interview.

Discourse group
The planner The troubled The protector The follower The stayer Most attended school up to 5 years (7).
All attended school for 5-9 years.

Migration history
Most have never migrated anywhere. Only one informant migrated seasonally as a household for 1 month of work.
Most have never migrated anywhere. Three households have had family members migrated to Dhaka or Barisal several times for work (fishing and garment).
Most have never migrated anywhere. Two of the men have migrated to Dhaka in the past for day labour work.
Have not tended to seasonal, temporary or permanent migration.
Most have never migrated anywhere. One woman spent 10 years total in Dhaka working in the harbour.  The realist

Migration history
Four men migrated seasonally for about 7-10 years in total to cut wood or construction work. Two of them also migrated to Dhaka permanently but later returned.
Most have never migrated anywhere (4). One man migrated seasonally to Dhaka for two months to work in a garment factory, and another man permanently to One woman estimate that her husband have been in the Sundarbans for 20 years in total cutting wood, and one man in Dhaka seasonally for 8 years in One man and one woman migrated seasonally for garment factory work for a total of 4 years each.
The man migrated permanently to Dhaka and opened up a shop there for a year. The woman work in a garment factory in Chittagong.
(continued on next page) The planner clearly separated the home and shelter as unsafe and safe spaces. The shelter was safer since a house could not save lives, or withstand flooding and storm surges: Q3: The shelter is safer than my house so I must go there (002:20 +4). Q4: We will not be able to save our lives if we stay at home (053:38 -5). Q5: If the water level raises too high, a house will not be able to save us (055:38 -5). 7 The troubled (male) expressed diverse fears related to shelter evacuation and the cyclone strikes. Their unwillingness to evacuate was often explained by fear of leaving valuables behind. Future scenarios of what may happen to the valuables and how this could affect the family's wellbeing created distress. The Bengali term 'mulloban jinispotro' was used to define 'valuable things' -mainly portable valuables such as documents of value (often for property, assets and agreements) and jewellery. Valuables were often behind delays or failure to evacuate to the shelter: Q9: It has taken me hard work to get my valuables together so I cannot just leave them behind (025:7 +5). Q10: It is very hard to leave without my valuables (027:7 +4).
The narratives also captured fear around what may happen on the way to the shelter. The dangers described included injuries and deaths from falling trees. In other storylines the severity of the dangers were clarified with adjectives such as 'terrible' or 'horrendous': Q11: Sometimes when we go, people get injured or even die from falling trees (034:33 +4). Q12: We are very afraid that something terrible may happen to us (035:33 +4).
Despite their fears around evacuation, The troubled strongly emphasised that everyone, women and men, had the right to The protector (male) shared some values with The troubled by outlining the right for everyone to seek shelter. The protector, however, added the importance of disaster training and that a man must protect his whole family (not just himself). The move to the shelter was narrated as a responsibility to keep family members safe so that a man ought never to leave their family behind. The narratives described collective mobilities where evacuation always had to come first. The time to discuss why people may not have wanted to evacuate was when everyone was in safety in the shelter: Q17: I must go to be able to keep my family safe (015:15 +4). Q18: Everyone must go there, then after arriving we can discuss (029:15 +5). Q19: I have a responsibility towards my family (028:12 +4).
Concerning The protector's individual evacuation decision-making, no priorities came before safety. It would simply not be wise to refrain from seeking shelter: Q20: Being safe is my first priority (015:19 -4). Q21: That would be a bad decision [to not go to the shelter] (018:20 -5). 9 The follower (female) also brought up the importance of natural early warning messages. It is interesting that several female discourses pointed this out. According to The follower, it would be wrong of a man to evacuate and leave his wife behind. A husband's first priority must be to protect his wife. This 'natural' order was explained by the superior knowledge and ability of the male. A man was simply smarter than a woman and knew better how to handle difficulties: Q24: I will know first of all due to the nature (012:40 +5). Q25: A wife should be a husband's first priority (011:35 +4). Q26: A husband knows how to handle things smarter (012:35 +4).
The shelter was not particularly unsafe for women but The follower did not feel safe there. When it came to the word of a disaster volunteer it did not measure up to the word of a husband. The follower would not necessarily evacuate because it was the order of a volunteer since the husband was the ultimate decision-maker: Q27: My house is safer than the shelter (010:14 -4). Q28: They can very well tell me to evacuate but I will follow my husband's instructions (012:34 -4). 10 Female The stayer was certain that women were unsafe in the cyclone shelters -a woman belongs at home. The lack of safety in the shelter is related to the crowds that increased risks to females. The storylines described other women's lack of safety as well as their own. The stayer expressed strongly disliking the shelter space and not feeling comfortable in there: Q35: Crowds create an increased risk for women so they are not safe in the shelter (057:4 +5). Q36: I agree with this, women are not safe in the shelter (017:4 +4). Q37: I do not like the shelter. I must agree with this (017:35 +5).
The stayer also clarified that the lack of safety had nothing to do with fear of the shelter collapsing or that delayed warning messages would pose increased evacuation risks. The stayer did not lack time to evacuate, nor did they get the warning messages too late. The discourse group felt strongly that they should not seek shelter outside the home. This tied into their female responsibility over the house and its family members. Their ultimate role was to stay behind and look after the cattle, furniture and household: Q38: I cannot just leave my house that easy. I should try to stay until the end. I have a responsibility to my household, to my cattle, and my furniture (004:26 +5). Q39: I usually get the warning quickly (004:23 -4). Q40: I do not think there is any chance of the shelter breaking (057:36 -4). 11 Out of the five, only The stayer disagreed with the statement 'Evacuating is the right thing to do (15-2)'. The narrative strongly illustrated how gender and social roles regulated normative behaviour.

'Natural' female intuition and masculine protection
Female and male discourses in Dalbanga South paid importance to early warning systems. However, interestingly, only female discourses mentioned natural systems. Women described trusting and depending on nature, feelings and intuition to anticipate the cyclones. Meanwhile, male discourses described a responsibility and a duty to protect and guard over finances and personal safety. It was problematic to leave valuables or family members behind. The family's survival was a husband's core liability. The follower (female) confirmed this obligation as resulting from male psychological and emotional pressure. The distress was articulated by male discourses through the fear, worry and anxiety surrounding the risks that evacuation posed upon family members.
The analysis revealed different viewpoints around women's safety, their right to seek shelter outside the home, and the nature of inappropriate spaces -particularly for unmarried women. It should be acknowledged that 'unmarried women' often referred to 'unmarried girls' as in rural Bangladesh many women marry at a relatively young age. Male discourses outlined everybody's right to safety and shelter place (including unmarried women), but female The follower, although agreeing with their right to be safe, clarified that no place is safe. As nobody truly knew their destiny, the house felt safer. Meanwhile, The stayer (female) pushed women's lack of safety in the shelter further as crowds implied increased risks to women. The concrete building may be physically strong, but is viewed as unsafe due to social risks. Women were narrated as better off waiting out the cyclones at home and felt obliged to stay behind to care for the household. Long evacuation distances also posed dangers so the correct choice was to stay at home. The discourses revealed several gendered immobility aspects that delicately illustrate the value of analysing 'trapped' populations through a gender-lens. The gendered disaster immobility, and its aligned evacuation reasoning, ran though female as well as male discourses.

Safe uncertainty -Fearing death but trusting God's decision-making power
Four discourses, whereof three female, were identified on Mazer Char island. The selected un-rotated factors explain 48% of the study variance and 38 of the 62 participants were highly associated with the discourses. Overall, people focused more on the lack of space in the shelter(s) and less on women's safety. The cautious (female) explained how important early warning systems are for people as they anticipate the approaching dangers (Fig. 4). This included natural systems such as the observation of rivers and animals: Q44: That is how I will know when the danger is coming (065:1 +5). Q45: We read various natural signals. We, for example, observe the sound and movement of the river and how the crabs move (068:40 +5 The cautious felt strongly that men never ought to leave women or children behind during the cyclone strikes. Men were described as stronger and more intelligent and therefore must protect and keep their families safe: Q46: A husband is responsible and must protect his wife (074:35 +4). Q47: A husband should keep his wife and family safe as he is stronger and more intelligent (069:35 +4).
The cautious opposed any likelihood of staying behind and waiting out the cyclones at home. To survive they had to evacuate to the shelters emphasising that the shelter was made for everyone: Q48: If we want to live we will have to go to the shelter (093:38 -5). Q49: The shelter is made for everyone (076:31 -4). 12 The realist (male), a young group in their 20-30 s, thought the main issue was the lack of space in the shelter. The village simply needed more cyclone shelters. It was not that people did not want to evacuate, but that the limited space prevented them from trying: Q57: There is not enough space in the shelter but we still have to go there (087:24 +5). Q58: We need more cyclone shelters (078:24 +5). Q59: If there is not enough space there, then people will stop going (089:24 +4).
Besides the limited space, most people in this group also lived too far away from the shelters. The long distances meant that people often could not come to the shelter without putting their lives at risk: Q60: It is because of the long distance that most people cannot come to the shelter (100:22 +5). Q61: The shelter is quite far from us. People may die on the way there (085:22 +4).
The realist did not think that the warning messages were irregular, confusing, or unclear, but that they were heard loud and clear. Women (including unmarried) were also safe in the shelter, but The realist felt that they should be accompanied by a father or brother who could protect them: Q62: An unmarried woman should be accompanied by her father or brother (085:29 -5). Q63: They are safe in the shelter (087:4 -4). Q64: I hear the warning loud and clear (089:18 -4). 13 The believer (female) had to evacuate to the shelters for their own safety. Evacuating was the right thing to do so one should listen to villagers or accommodate around things making sure to seek shelter when encouraged. In the end however, The believer explained, who was to live and who was to die would all be up to God's will: Q80: We need to go there for our safety (072:15 +5). Q81: Allah can do what he wants as he is the Almighty (116:28 +4). Q82: Allah knows everything (120:3 +5).
The believer clarified that the shelter did not lack space but that the space was badly organised. The warning messages came through various channels including TV, radio and volunteers so the system worked fine. Finally, The believer clarified that unmarried women had the same human rights as others. They should therefore evacuate or seek shelter just like the rest of the village: Q83: I get the warning from the volunteers, TV and radio. There are no obstacles here (072:18 -4). Q84: There is enough space but it is not organised (120:24 -4). Q85: Unmarried women are human beings too (118:29 -4) .14 The anxious (female), in their 40 s, differed substantially from other female discourses. The discourse did not think people relied on natural warnings as people required more time to read them. The anxious explained that they sometimes did not even evacuate when told to by the disaster preparedness volunteers: Q95: Most people do not wait or listen to natural early warning messages (102:40 -5). Q96: Sometimes we do not go to the shelter although the volunteers tell us to evacuate (110:34 0).
The reasoning around the non-evacuation behaviour was emotionally loaded. The anxious expressed pure terror around what could happen to them. They could very well end up dying on the way to the shelter, a fear that may be incomprehensible for those who never experienced a cyclone strike. All in all it was simply better to die at home than in the shelter: Q97: If we try to go to the shelter after the cyclone has started, we could all end up dying on the way there (110:33 +5). Q98: He who has not experienced a cyclone may never understand its danger (109:16 +4). Q99: I very much agree with this statement. I want to die in my house (102:37 +4).
Despite the fear and hesitance of evacuating The anxious thought it was wrong of a man to leave a woman behind. Everyone should stay together as a family -survive or die together. The shelter space was not described as an inappropriate space for women, but anything could happen there so unmarried women better stay home: Q100: Your family may die. We have to stay together. If we die, then we all die together (109:12 +5). Q101: Anything could happen to unmarried girls (097:29 +4). 15 It appeared that some discourses had accepted living in a state of safe uncertainty 'when the lack of certainty around future climate threats do not necessarily limit people to feel secure' (Mason, 1993;Ayeb-Karlsson et al., 2019b). It was highly uncertain when the next cyclone would strike and whether people would survive, but diverse strategies to achieve safety in the uncertainty streamed through the discourses. For example, leaning in and accepting God as the ultimate decision-maker appeared to be a way to ease female fear over what could happen, and male anxiety over failing to protect the lives of family members.

Male lack of physical place -Female lack of social space
Several female discourses emphasised the importance of natural warning signals, although The anxious disagreed as people did not have time for intuition. It should, however, be noted that as the data was analysed separately for each study site there will be some overlaps and similarities in the results. Male The realist felt that people did not evacuate to the shelters due to their locations and lack of space. People were afraid that they would not reach the shelter in time, or that it would be full when they got there. Female The believer disagreed and said that the available space was disorganised.
Male discourses proposed that women (including unmarried) were safe in the shelters, while female The believer clarified that women's safety and survival depended on God. The Almighty Allah controlled all death and survival leading them to put their faith in God's hands. The anxious (female) disagreed with unmarried women evacuating. Terrible things could happen to them on the way to the shelter or once there -separating married and unmarried women. Only The anxious described their own immobility, while others referred to collective (im)mobility. Not only did they fear dying en route, but they also preferred to die at home than in the shelter. 13  The narratives circled around several levels of death anxiety, such as the fear of dying alone, on the way to, or in the shelter. The risk of the shelter collapsing due to cracks in the building posed another danger. Mistrust around the warnings of disaster volunteers and villagers of approaching cyclones were other reasons to stay behind.

Allah the almighty decision-maker encourages home-based safety
Six discourses were identified in Gabtola where the selected un-rotated factors explain 54% of the study variance. 30 of the 62 participants were highly associated with the discourses. The discursive values here were strongly linked to religion and described selfexperienced immobility rather than collective immobility.
The religious (female) clarified that life and death was essentially up to God. It was therefore irrelevant if a person remained at home or had made it to a shelter since their survival remained in God's hands (Fig. 5). If it were God's will for a person to be saved from the cyclone, then they would be saved within the four walls of the home. If God's will was instead for that person's time to have come, they would have to face death, whether in the shelter or at home. As only God knew the outcome it was better to remain at home than to seek shelter. In that way one could die with respect: Q110: If God wants to save us, he will save us at home (147:37 +5). Q111: [It is better to die at home] because I want to die with respect (172:37 +5).
The narrative of The religious strongly circulated around the Almighty Allah as the ultimate decision-maker and world creator. Allah knows everything, decides everything and nobody has the power to change the decisions or determine their own fate. Interestingly, this devotion and certainty around God's will eased fear and anxious feelings around the cyclone strikes -The religious were not afraid of dying at all: Q112: Allah knows everything (173:6 +5). Q113: Allah decides everything (173:28 +4). Q114: No one can change God's decision. Allah knows everything (147:28 +4). Q115: If Allah decides to save us, then nobody can lead us down the path of death (173:16 -4).
The religious made an important distinction between availability and safety in regards to the shelter space. The shelter is available for all, but lack of safety for women. This unsafety was narrated around religious devotions. Muslim women in particular were not safe due to the lack of religious privacy and obligations. Unmarried women faced problems in the shelter as they could not secure their privacy and young men would tease and harass them. It should be acknowledged that 'teasing' in a South Asian context is a euphemism referring to 'Eve-teasing', a term used to describe sexual harassment and molestation of women by men, often in public places and by strangers. As Eve refers to the biblical creation story, and refers to the temptress 'nature' of Eve it places responsibility on women while minimising the seriousness of the offense: Q116: That women are not safe in the shelter is important in relation to religious privacy -especially for Muslims (182:4 +4). Q117: Women with family members are safe in the shelter (147:4 -4). Q118: Unmarried women face problems in the shelter. Young men sexually harass or 'tease' them and they cannot maintain their privacy (181:29 0). Q119: I am a poor man, so if I have the right to be safe then why would not others have the same right (172:31-5).
Similarly to other female discourses, The religious believed that natural warning signals and traditional knowledge should be acknowledged. In terms of the existing technological early warning systems they saw no issues: Q120: I believe and trust in what the elders tell me (135:40 +4). Q121:I receive the warning each time and very quickly (135:18 -4). 16 The saviour (male) felt strongly that their ultimate responsibility was to protect and save the lives of their family membersparticularly women in the household. Female family members are their sole responsibility and should therefore never be left behind. Women should therefore be taken to the shelter with the children by The saviour before they evacuate themselves. Interestingly, it seemed that the option for women to wait out the cyclone at home had not even crossed their minds 'Where is she supposed to go if not to the shelter?', they asked. The saviour saw their saviour role as fundamental to their existence -any questions beyond the survival of their family were met with ambivalence as losing them meant losing their reason to live: The saviour also felt that the death of every single person depended on Allah, but there was an interesting distinction from the female discourse; God only saved those who attempted saving themselves. People must at least try to evacuate in order for God to save their lives: Q133: The death of every single man depends on Allah (143:28 +4). Q134: We believe in Allah, but Allah only helps those who helps themselves. So we have to at least try to survive (151:6 0).
The saviour saw a clear reason to evacuate. It would be foolish not to as it could end up costing them their lives. People were truthful when it came to cyclone warnings including from strangers -nobody would lie about such a thing. The saviour declined the existence of potential mistrust or suspicion expressed by others around 'outsiders' as being thieves who would try to steal belongings after people have evacuated: Q135: This would be foolish. If we do not go, we will end up losing everything (151:20 -5). Q136: This is foolishness, you will end up dying thinking like that (140:20 -4). Q137: We are all truthful about these things, nobody would give false information about a cyclone strike (140:27 -5) .17 The rational (male) felt that nobody wanted to leave their belongings behind. The Bengali word 'jinispotro' was used for belongings referring to portable and non-portable things including possessions, items, and goods. The emotional effort made to force oneself to leave belongings behind was strongly captured in the narratives. Leaving family members behind however, was seen as unacceptable as women were every 'man's responsibility': Q150: Nobody wants to leave their belongings behind. They try their best to leave behind what they can in accordance to their very best efforts (166:25 +4). Q151: There is no other way. I have to leave my house and go to the cyclone shelter (170:20 -5). Q152: Nobody can leave their family behind, family is a man's liability (167:35 +4).
The rational had strong reason to evacuate and always got the early warning messages in time. The discourse added an interesting aspect to the natural warning signals, by clarifying that they are important and should work, but due to climate change they cannot be relied upon anymore: Q153: I always get the warning in time (166:23 -5). Q154: Nowadays, the climate has changed so the natural signs are not really working anymore (170:40 -4).
The rational noted that lack of space in the shelters was a serious concern. One could never be certain that there would be space available once reaching the shelter. Space restrictions could potentially force people to continue their evacuation route to another shelter, or make them try to survive the cyclone on the stairs leading up to the shelter, resulting in plenty of death in the village. Finally, one could die on the way to the shelter, painting a picture of the immensely difficult life and death decisions people face during the cyclone strikes: Q155: It is true that there is not enough space in the shelter. Plenty of death has occurred due to this reason (165:24 +4). Q156: I never know whether I will reach the shelter safely. My neighbour died on his way there (156:3 0). 18 The housewife (female) expressed strong feelings about wanting to wait out the cyclones at home. The home was described as their 'everything'. This was where they wished to die, in the comfort of their own homes. Adding to this, they strongly disliked the crowds that gather in the public cyclone shelters. Their place was at home and nowhere else, The housewife emphasised: The religious discourse expressed similar fearlessness around death as other religious female groups. The housewife felt no fear, just as God had created them, He would one day bring them back. This belief aligned with being a good Muslim. One needed to believe, give in to, and trust, that they were not in control. God had a higher plan for them all and would ultimately decide who lived and died. This gave fundamental peace to The housewife that eliminated fear, stress and anxiety around the cyclone strikes and their outcomes: Q172: Only Allah can decide who lives and dies (139:28 +4). Q173: I have no fear of death. Allah created us and one day he will take us away (179:16 -5).
The housewife strongly felt that there was no point in evacuating. One woman had lost her son and therefore did not want to go to the shelter. Another expressed disliking the chaos in the shelters. She did not want to go anywhere and would stay in her house until the day she died. A third woman described the extreme emotional distress she felt in the shelter by comparing it with suffocation. She felt as if she was gasping for air in the restricted space. The housewife also agreed that unmarried women should not evacuate to the shelters. They could not be stopped from going but should not be doing so: Q174: I lost my son so I do not want to go to the shelter (139:20 +5). Q175: I do not want to go anywhere. I do not like the chaos. I want to stay in my house up until the day I die (179:20 +4). Q176: There is no space. I feel suffocated in there (180:19 +4). Q177: Sometimes I think this is true, that unmarried women should not go to the shelter, but I suppose you cannot stop anyone from going (179:29 0). The housewife made sure to clarify that their non-evacuation behaviour had nothing to do with early warning system failure or lack of warning messages. They got the warnings fine and everyone was allowed into the shelters: Q178: We get the signal on time (180:18 -4). Q179: Everyone is allowed in (169:31 -4). 19 The guardian (male) gave detailed elaborations around the value and importance of disaster preparedness training and early warning systems. They expressed proudness around their preparedness knowledge, such as 'I know just how to prepare', and explained in detail why disaster training and early warning systems are important. 'The training helps us understand what to do and what not to do', they said, and 'warning messages are the most important thing to us as they tell us when to start preparing': Q187: If we get training we will be better prepared (158:10 +5). Q188: The training is very important as it helps us understand what to do, and what not to do. It gives us a complete overview of how to prepare (134:10 +4). Q189: Early warning messages are the most important thing for us to be able to know when to start preparing (134:1 +5). Q190: I know just how to prepare. The volunteers taught us about the flags (134:13 +4).
The guardian felt strongly that a man should never leave his family behind. One man even clarified that nobody can do that unless they did not understand the true value of a family. He knew this as his family was not present at that moment. Having a family and being able to protect them were described as core-elements of masculine social value: Q191: No one can do this. I know the true value of family. My family is not here (157:12 +4). Q192: Everyone must reach the shelter safely (157:15 +5).
Similar to other male discourses who felt a responsibility to safeguard the lives of their family, The guardian expressed stress and anxiety around the evacuation and cyclone strike. Anything could happen on the way to the shelter, and once there the shelter could collapse in the storm: Q193: This is true. The shelter can break down anytime (163:36 +5). Q194: Anything can happen to us on the way to the shelter. We do not have a clue (158:3 0).
The guardian also expressed strong feelings around statements relating to non-evacuation behaviour. One man simply shouted out 'This is wrong'. People were not so malicious that they would find time to steal valuables that had been left behind during the cyclones. Everyone would be occupied with surviving the strike, they stated: Q195: If one wants to live, he will have to go to the shelter (163:26 -5). Q196: This is wrong -we will die if we do not go (158:20 -5). Q197: When people are facing life and death situations, who would take their valuables (134:7 -4). 20 Finally, mixed The optimist(mixed), a young group under the age of 34, believed that evacuation is a great lifesaver and that people must go to the shelters to stay safe and alive: Q206: Evacuating is a great lifesaver (159:15 +5). Q207: To keep ourselves safe we have to go to the shelters (138:15 +4).
These values ran through the storylines of The optimist. For example, the guidelines of the disaster preparedness volunteers must be followed as they have the most accurate weather information. The optimist also explained that disaster preparedness training was useful and helpful as people learned how to prepare for the cyclones: Q208: We can learn many things on how important it is to prepare during a cyclone (138:10 +4). Q209: The volunteers have the most updated information so I would follow his advice (138:34 +5).
Statements relating to non-evacuation behaviour were met with strong reactions. In response to the statement 'the shelter is not a place where I should be' one man, for example, screamed out 'Who said such a thing?' The man clarified that it was his place just as much as it was anyone else's. The shelter was where he would be. Interestingly, one woman identified with the statement referring to unmarried women staying behind and said that although she indeed is unmarried (outlining the strong social value of matrimony) she had a right to be safe. Additionally, The optimist felt that one could not just give up. One had to try evacuating to the shelter and hope for a positive outcome: Q210: Even though I am an unmarried woman, I still have the right to be safe (138:29 -5). Q211: Who said such a thing? It is my place and the shelter is where I will be (159:19 -4). Q212: I have to at least try. I cannot give up that easy (138:37 -4). 21 The discourses revealed valuable reasoning around gendered evacuation behaviours and the explanations people gave for moving or staying behind. It was clear that the social roles and socially expected behaviours were different for men and women. Interestingly, however, the same factors were often used to argue for both male mobility and female immobility. For example, religion and the efforts men and women made to adhere God's will were in clear contrast. 4.6. God's will is for men to go but for women to stay All six discourses in Gabtola felt strongly about God's desire and decision-making power. These religious narratives were mostly linked to life and death outcomes. An important distinction between female and male discourses was how God helped men who helped themselves, while God would save a woman no matter where she waited out the cyclone -an attitude that encouraged female non-evacuation behaviour. For a woman it is better to wait out the cyclone at home since it was safer, more comfortable and this was where she could die in dignity. When God decided that her time had come she would die regardless of whether she had stayed behind or evacuated to the shelter.
Another interesting element was how statements such as 'There is no point going to the shelter', 'My place is at home' and 'The shelter is not a place where I should be' angered the male discourses. The 'foolishness' of these statements upset men, while the some women strongly agreed with them as they increased their security and safety. The housewife avoided going to the shelters as they felt unsafe due to crowds and chaos. One woman even felt as if she could not breathe or was suffocating in the space. This powerful feeling assured her that she should not be there. Another woman stopped going after she lost her son, confirming that she needed male family member's protection to be safe. The shelter was mainly described as an unsafe place for women by the female discourses. The religious linked the lack of safety back to religion as it was a serious problem for Muslims. Meanwhile, the female discourses made a distinction between the vulnerability of married and unmarried women. All discourses confirmed that everybody was allowed into the shelter, but the female discourses clarified that this did not mean that women were safe or should be there. Unmarried women in particular, did not belong in the shelters and should therefore refrain from evacuating. Nobody would try to stop them from going, but they would lack personal and religious privacy and risk facing sexual harassment and serious abuse.

Discussion
This Q-based Discourse Analysis demonstrated important empirical insights into the ways in which diverse gender roles have guided male and female behaviours during the cyclone strikes in Bangladesh. The findings revealed gendered subjective reasonings around why some should evacuate while others ought to stay behind (Fig. 6). The gendered power system affected women and men differently (not simply negative for women or positive for men). The system also aligned with other social roles in the way that age and marital status, for example, could influence someone's (im)mobility. Unmarried women did not have the same mobility options as married women.
The discourses described diverse layers of being 'trapped' that can be useful in the further conceptualisation of the term. To give an example, people commonly described feeling mentally and emotionally constrained in their decisions to move or evacuate to the shelter(s) despite such immobilising factors receiving limited attention within the 'trapped' populations literature. Psychological trauma (such as having witnessed someone die in a previous cyclones) could be immobilising, for example. Some women described preferring to wait out a cyclone at home after having lost family members. Men sometimes reported finding the very thought of evacuating unbearable after having 'failed' in their male responsibility to get their family safely to the shelter. These findings illustrate how gender roles affect societies (women and men) as social norms encourage female and male behaviours (O'Brien et al., 2007;Bradshaw, 2010;Arora-Jonsson, 2011).
The socially immobilising elements of women and men, unmarried and married or poor and rich, shed light on the conceptual wording of Trapped Populations. The analysis perfectly showcase the fact that references to a homogenous group of people or a population can be misleading. Semantically grouping people into overly coarse classifications (or framing 'populations' rather than an individuals 'trapped') can therefore create unnecessary generalisations and lead to confusion around how to most effectively apply the concept. Encouraging scholars to widen the use of the concept to include Trapped Individuals and Persons could address this conceptual weakness as one group or population is unlikely to be immobilised by the same factor.
Discourses, social norms, and subjective feelings around what was right and wrong were different for women and men. The reproduction of discourses and their meanings, values and norms in this study ended up constraining women's mobility more than men's. Male social behaviours and spaces were not always 'acceptable' for women. To give an example, no discourse (out of fifteen) narrated the possibility of a man being 'harmed' by another person in the shelter, while this was described (by men and women) as a risk for women in several discourses. Similarly, in contrast to the case for women, no discourse claimed that men could generate 'bad luck' when leaving the house or that they did not 'belong' in the shelter. The unequal effects upon women during the cyclone strikes differentiated in shape and form but the root cause of the gender vulnerability was social. Immobility (social and geographical) must therefore be analysed through a psychosocial lens (Blaikie et al., 1994;Cutter, 2017).

Conclusion
This gendered disaster immobility study makes a few important contributions to the existing literature body. Methodologically it illustrates an innovative way of applying Q-based Discourse Analysis within climate-induced (im)mobility resulting in rich empirical insights that revealed complex psychosocial layers. The study is also an important contribution to the scarce critical gender and disaster literature, and unique in the way that it interlinks gender, disasters and immobility. Finally, it offers a much needed contribution to the Trapped Populations literature; one that is urgently lacking both empirical insights and publications that conceptualise gender as a potential socially immobilising factor.
The analysis presented here has delicately illustrated involuntary and voluntary immobility as well as the grey areas in between these positions. Complex and shifting aspirations have been revealed, sometimes expressed by one and the same discourse or even one and the same person. These are useful insights for the conceptualisation of Trapped Populations and as a means through which better incorporate climate-induced immobility within climate policy. The contribution is part of a growing body of critical literature that applies the concept empirically and which is key to building knowledge-based policy recommendations that protect vulnerable populations while preserving subjective agency, desire and dignity.
Gender has been shown to be immobilising or 'trapping' people in different ways during the cyclone strikes in Bangladesh. Women and men faced psychologically and emotionally paralysing effects from their expected gender roles. When the disaster strikes, not all people had the same ability to move. More empirical evidence on who, how and why people in different geographical and sociocultural settings become immobile during disaster strikes is needed to create more robust policy frameworks. Trapped Populations clearly has the capacity to become a powerful climate policy tool that better protects vulnerable people worldwide to the impacts of global environmental changes in the future. However, to fully realise this potential the abstract notion of the populations in question must be more thoroughly explored and the nuanced composition and behaviour of diverse and homogenous groups of people adequately accommodated for.

Declaration of Competing Interest
The author declares no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.