St. Kitts and the 2017 hurricanes: sheltering, evacuation, disaster preparation and communication insights from a transient island population perspective Open Access

Disasters triggered by weather-related hazards have escalated around the globe in recent years (IPCC, 2023). In the Atlantic Basin, there has been a marked increase in the number of named storms, hurricanes and major hurricanes (category 3+) (Kleptsova et al., 2021). These hurricanes and storms and their associated flooding, high winds and storm surges pose a serious annual threat to the Caribbean region. While projections of hurricane activity further into the 21st century are varying and uncertain, some experts are predicting stronger, slower and wetter hurricanes in the future (Van Beusekom et al., 2018).

Hurricanes present opportunities to learn about improved means to protect against loss of life and infrastructural damage (Lewis, 2022). Research and implementation of lessons learned from major events such as Hurricane Katrina (2005) in the United States (USA) (Boyd et al., 2009; Henderson et al., 2009) have had significant impacts on improved response and decreased loss of life. However, much of the research to understand how to improve responses to hurricanes and other disasters triggered by weather-related hazards is focused on events that occur on large land masses and mainly in well-developed nations. In comparison, there is much less data on mitigation within island environments, which are unique. Caribbean nation situations are tremendously different from continental nations in their topography and natural ecosystems, sociocultural beliefs and practices, financial and political landscape, technological capability and infrastructure. Importantly, with current economic growth and development, island populations are unique considering subpopulations of people, including nationals, long-term residents (i.e. expatriates), transients (those who temporarily move for school, work or business) and visitors (i.e. tourists) all cohabitating but each with potentially their own needs in the face of disasters (Sutton, 2006). As a result of the comprehensive distinctness that comprises an island's identity, Caribbean islands' response capabilities to adverse events are also starkly different.

One significant difference between Caribbean island nations and large land mass nations is the use of evacuation to decrease exposure of people to a hazard or subsequent disaster. On larger and developed land masses, people can evacuate to areas that are not expected to be impacted. Evacuation might not be an option for island environments for many reasons, and these may differ based on the population. Reasons for nationals include but are not limited to: a lack of economic means, sociocultural fears (e.g. fear of leaving family and losing personal belongings to theft and damage), spatial limitations to relocate due to island size and topography; physical barriers resulting directly and indirectly from causative weather events such as flooding and land/mudslides (Medina et al., 2023). Transients and tourists may not evacuate as they lack awareness of the danger and are unfamiliar with what action to take in the face of harm (Fathianpour et al., 2023). Logistically speaking, insular locations may have limitations in the number of flights and other transportation options feasible to remove people to safer areas. Island environments also differ in the frequency of exposure to disasters triggered by weather-related hazards such as hurricanes, which could result in a more prepared population. Studies by Gargano et al. (2015) and Malmin (2021) demonstrated that previous exposure to a disaster can increase resident preparedness and willingness to prepare. In this regard, with ten hurricanes in 2017, eight hurricanes and seven additional tropical storms in 2018 and six hurricanes plus an additional 12 tropical storms in 2019, Caribbean island populations could be well prepared for disasters, resulting in better mitigation and less loss of life. However, preparedness might not increase with exposure due to many socioeconomic factors and warning fatigue (Cong et al., 2023; Chan et al., 2014).

In 2017, St. Kitts, a Caribbean island in the Lesser Antilles, was impacted by two hurricanes, Irma and Maria. St. Kitts was considered to be in the direct path of Irma and residents were requested to prepare with this expectation. Given the size of the island (23 miles long and 5 miles wide at the widest section, with an area of 69 square miles) (St. Kitts and Nevis Bureau of Standards, 2023) and the average size of hurricanes being approximately 300 miles wide with centers of 20–40 miles wide (National Oceanic and Atmospheric Administration, n.d.), evacuation within the island was unlikely to be beneficial. Evacuation to off-island locations was challenging due to flight limitations and significant flight costs in light of the minimum wage in St. Kitts being $9.00 Eastern Caribbean dollars/hour in 2017, making plane tickets unaffordable for many residents (Law Commission, 2017).

While St. Kitts has not been struck by a major hurricane since 1995, the island is exposed to tropical storms and category 1 and 2 hurricanes every hurricane season. Hence, St. Kitts provides an opportunity to learn from residents and identify means of improving mitigation in an island setting. The first aim of this research, in which we used a mixed population of transients and long-term residents, was to explore factors in the decision-making process with regard to shelter at home or shelter in a designated facility. The second aim was to corroborate that previous exposure to disasters results in better preparation and obtain a better understanding of what “prepared” means. The third aim was to determine the sources of information used by those in island environments prior to and during a weather-related disaster.

The Caribbean region represents one of the world's most unequal regions. It is an area that is heavily affected by poverty, but also holds tremendous wealth disparity. There is widespread illiteracy and poor educational performance, inadequate public health and negative health outcomes and lagging economic development, among other negative social determinants (Busso and Messina, 2020). Colonialism and the legacy of the institution of slavery still haunt the region with social and racial inequality (Weiss et al., 2018). Concurrently, large numbers of transients and tourists embark upon this region for education, business and recreational purposes (Erikson and Lawrence, 2008). All of this exists in a region in the global south that is well recognized as one of the most prone to natural hazards (López-Marrero et al., 2013), and the Caribbean's vulnerability to the anticipated tertiary effects of climate change (i.e. harm to agri-aquaculture, tourism, water availability, energy supply, etc.) is only expected to worsen (Taylor et al., 2012). The researchers' approach to this study was to examine disasters in the Caribbean through the lens of inequality. It is important to give nationals and residents in this region, who face limitations in resources, access to every advantage in facing hazards and disasters. To do this, we must understand the decisions and impacts of transient and visiting populations and their respective organizations on local resources. We intend for this research to explore decisions of transient populations as a first step in exploring differences in these populations and their potential impacts on resources. This research may call attention to the obligation of organizations to bear some responsibility for their personnel in these localities, especially in the context of disasters, so as not to overwhelm islands' emergency response. We intend for research and emergency response to consider the distinctive local context to promote efforts and actions that will address residents' mental models and available resources, and facilitate cooperative work to build capacity towards safer and stronger communities in the Caribbean.

The study and survey were approved by the Ross University School of Veterinary Medicine (RUSVM) Institutional Review Board (19-06-EX) and all participants were informed that participation was voluntary and anonymous. The survey was developed using Qualtrics (Provo, Utah, USA; https://www.qualtrics.com) with a link sent via email to faculty, staff and students known to have been at RUSVM in 2017. The survey was available from 26 November 2019 to 6 February 2020. The email and first page of the survey contained consent information, the voluntary aspect of the survey, the approximate time to complete the survey and contact information if the respondent had questions. Respondents were prevented from completing the survey multiple times using the default method within Qualtrics.

The survey consisted of demographic questions followed by questions specific to the aims (see Supplemental File 1). Additional questions requiring validation for future surveys on pet ownership and actions during disasters were included but not presented here. Questions were multiple choice, drop-down, yes/no, Likert scale and open-ended. Respondents who completed <70% of the survey were excluded. In addition, respondents who provided conflicting information regarding location (on St. Kitts or not) during the hurricanes were omitted.

Data were transferred from Qualtrics to Microsoft Excel and then imported into Minitab^® (version 21.1.1; LLC, State College, PA, USA). Analyses were performed using Minitab^® and VassarStats (http://vassarstats.net/). Data were tabulated and responses to questions were coded for analysis (e.g. answer “Yes” converted to 0, “No” to 1). Additionally, for Likert scale questions, responses were combined with those that were negative or neutral (e.g. not a consideration, considered a little and not a major consideration) coded together and those that were positive (e.g. important and very important) coded together. Age categories used reflected generational groups and residency time on St. Kitts categories were meant to reflect the division between transient populations (primarily students and temporary employees) and those with potentially more place attachment (permanent employment or residency).

The first hurricane, Irma, occurred shortly before the beginning of a new semester at the university, with some students, local staff and faculty not yet returned to St. Kitts. These individuals largely returned to St. Kitts between the two hurricanes. Therefore, most respondents were either present for both or only the second hurricane. For questions not specific to the hurricanes (e.g. previous disaster exposure, communication sources), data from all respondents were assessed. For questions specific to sheltering, data from only those present during Irma or who specifically evacuated due to the pending arrival of Irma were analyzed. Data from those present only for Maria were excluded due to the potential influence of Irma on sheltering decisions for Maria.

For questions regarding sheltering during the storm, those residing in a university dormitory were considered to be in a designated shelter already and excluded from the analysis. Respondents were categorized as having moved to an official shelter if they moved from off-campus to on-campus or to a designated government facility. For questions regarding preparedness, those who obtained items or developed disaster plans or kits between the hurricanes were classified as unprepared, given that hurricane season started several weeks earlier.

A chi-square test was used to assess the association of any exposure and the number of exposures to a previous disaster on perceived preparedness. Chi-square test also compared previous exposure to a disaster and perceived preparedness to true preparedness, defined as having critical disaster items (plan and kit), using a chi-square test. Availability of items recommended as part of disaster preparedness, sheltering decision data and data on sources of information used were tabulated. Additionally, a Chi-square test was used to assess differences between respondents from the Caribbean region and not from the region, by length of residency on St. Kitts, and by age.

The survey was accessed by 376 potential respondents, 55 of whom were excluded for completing <70% of the survey and an additional 28 of whom were excluded for not being on St. Kitts before or during either storm due to unrelated reasons (e.g. planned time off island for a conference). Responses from people who evacuated specifically because of Irma were retained for portions of the survey. A total of 293 respondents are included in the analyses, although not all respondents answered all questions, resulting in <293 responses in some cases.

The majority of the respondents lived in North America (198 in the USA and 16 in Canada) prior to moving to St. Kitts. Fifty-eight respondents were from St. Kitts or another Caribbean Island, inclusive of Puerto Rico. Four respondents did not indicate where they lived prior to St. Kitts and the remaining respondents (21) were from Europe, Australia or Asia. While not all respondents indicated their US state of origin, those who did respond represented 37 states, with California and New York being the most represented (22 and 23, respectively).

Over 69% (203 of 293) of respondents were students at RUSVM at the time of the hurricanes. The majority of the respondents were female (80%, 233 of 293) and ≥20 to ≤30 years of age (68%, 199 of 293), which is reflective of veterinary school student populations (Morello et al., 2021). Over one-third (33%, 67 of 203) of the student respondents had resided on St Kitts for <1 month, while 66.7% (58 of 87) of the non-student respondents had resided on St Kitts for over 3 years (Table 1).

Prior to hurricane Irma, 277 of the respondents were on St. Kitts. Of these, ten students and two faculty evacuated from the island of St. Kitts (destinations unknown). The evacuees included seven people not from a Caribbean island, three from Puerto Rico and two who did not state where they resided prior to St. Kitts. One of the evacuees had been on St. Kitts for <1 month, seven for 2–12 months, inclusive, three for 13–36 months, inclusive and one for over 36 months. While not statistically analyzed, there did not appear to be any trends for evacuees with regard to origin or length of residence on St. Kitts.

For those present for the hurricanes, 72 considered evacuating for Irma and 47 considered evacuating for Maria. Five of those who considered evacuating for Maria had not considered evacuating for Irma. Primary reasons for not evacuating despite considering it were cost (50 respondents), flight availability (39 respondents) and concern of missing classes/work (47 respondents; Supplemental File 2, Table I). More students than faculty (35.9% (66 of 184) vs 14% (11 of 78)) and more people previously residing outside of the Caribbean than those who were from the Caribbean (29.6% (59 of 199) vs 11.3% (6 of 53)) considered evacuating (Supplemental File 2, Table II). Residence time on St. Kitts was relatively evenly distributed for those who did or did not consider evacuation, taking into account that students typically spend 28 months on St. Kitts.

Of the 277 respondents on St. Kitts during Hurricane Irma, 51 resided in a RUSVM dormitory and were excluded from analysis regarding the decision of where to shelter during the storm. Of the remaining respondents, 179 were renters, 38 owned their homes and the remainder did not indicate their type of residence. For those who were renters, 3 (1.7%) went to a designated shelter, while 1 (2.6%) of those who owned their homes went to a designated shelter. For those who did not go to a shelter, 201 responded to some or all of the questions regarding the importance of factors in influencing their decision. Feeling safe in one's chosen place to reside during the storm and not wanting to be around many other people ranked highest in selection of reasons not to go to a designated shelter (Table 2).

People exposed to a disaster triggered by weather-related hazards in the past felt more prepared than those with no prior experience (Table 3, p = 0.0236). However, the number of previous exposures did not influence a feeling of preparedness (p = 0.0586). In addition, exposure to a disaster was not related to true preparedness (i.e. securing necessary household disaster items) (Table 4, p = 0.1821 for disaster plan and p = 0.7518 for disaster kit). Feeling prepared did relate to being prepared (Table 4, p < 0.0001 for disaster plan and disaster kit). A large portion of the respondents had key items for disaster preparation (e.g. flashlight, medicine, important documents) as part of their household items and almost all respondents had a minimum of a 3-day supply of food and water prior to Irma (Table 5). There was no apparent correlation between feeling prepared and having key items (Supplemental File 2, Table III).

Respondents relied the most on National Oceanic and Atmospheric Administration (NOAA) National Hurricane Center for information on the impending hurricanes, with local sources being used the least (Table 6). However, given the different populations (citizens/residents vs temporary students), data were analyzed to see if there was a difference in information sources (Supplemental File 2, Table IV). With the exception of NOAA and other Internet sources, those from the Caribbean and those not from the Caribbean relied on resources differently. This difference might have been due to age differences, with the non-Caribbean population being largely students and younger. As with origin, age did not impact reliance on The Weather Channel, NOAA, or other Internet sources (Supplemental File 2, Table IV).

Island nations are particularly vulnerable to disasters triggered by weather-related hazards. These vulnerabilities can be due to (mostly) common existing factors such as poverty, inadequate delivery of basic services, small undiversified economies, rapid urbanization and population density (especially coastal development) and an overall lack of capacity to deal with these precarious predisposing conditions (López-Marrero and Wisner, 2012). Given many of the differences between island nation environments versus larger landmasses, understanding decision making, preparedness and communication within island environments is needed.

Studies have shown that stress, loss and recovery efforts could be improved if people evacuated the area or were in designated shelters prior to the disasters (Rivera, 2020). On large land masses, the decision to evacuate, shelter in a designated place or shelter at home is influenced by several factors, with the type of disaster, income, education level and pet ownership being especially important (Chadwin, 2017). Our study explored reasons for not evacuating or going to designated shelters. In the study presented herein, evacuation was not a readily available option to most people due to limited flights and associated cost. This held true even for temporary residents who had potentially greater financial resources than citizens and had off-island locations in which to go. Interestingly, missing work and/or classes was also a driving factor in the decision not to evacuate. While this concern may have been population specific (students), it also might reflect island situations where internal recovery of the function of public and private entities could occur sooner than resumption of flights. Generally speaking, this could be an important consideration for individuals engaged in business, who might risk financial loss and job security by leaving the island with an ambiguous return. As an example, after Irma and Maria, electricity, water and work schedules resumed within a few days. However, flights to St. Kitts were backlogged, with some of those who evacuated waiting for return flights for over 2 weeks. Even if evacuation flights were affordable and available, island residents might not select to evacuate without assurances regarding the ability to return. This should be considered in developing island-wide disaster plans. As industries such as tourism and medical education grow in the Caribbean, it begs respective organizations to pay attention to and support transient individuals in the face of crises as an industry practice. Ideally, this should be planned for and communicated in advance of any crisis. Execution of this practice would be twofold; it would help ensure personal well-being and safety for those involved, and it should increase capacity for residents who critically rely on island resources in emergencies. Such was the case with RUSVM when, after St. Kitts' COVID-19 island shutdown, the university provided return charter flights to students and faculty to ensure the continuity of the veterinary medical education program, and not dominate the available airline flights to the island (SKNVibes, 2020).

For those who did not evacuate and chose to remain on the island, the survey investigated the number of respondents choosing to shelter at home versus shelter at a facility. What could not be determined in our study is whether warning fatigue and concerns about protecting property (Haynes et al., 2018) or previous experience influence these decisions. The role of place attachment was also not investigated (Swapan and Sadeque, 2021). However, over 80% of our respondents were not from St. Kitts and slightly over 50% had been resident on St. Kitts for less than 1 year, so the influence of place attachment might be low.

Our research did not investigate the reasons respondents felt safe in their residences, a primary reason for not going to a shelter. Guidance for residents on what entails a safe residence/building could potentially be beneficial so that people can make informed decisions on where and when to shelter. For example, flood zone data could be made readily available as well as home inspection services to evaluate sound construction well ahead of any impending storm. The importance of knowing if a residence is actually safe is supported by incidents that occurred during Hurricane Dorian in the Bahamas in 2019. In this storm, many residents stayed in their homes and realized their vulnerability as they attempted to ride out the storm. During the eye of the storm, there was extensive movement of people from residences to shelters (Marazita, 2020). Better preparation of residents and communication have been identified as vital contributing factors to disaster mitigation for any disaster (Levac et al., 2012). Based on work by Levac et al. (2012), there is a positive correlation between previous disaster exposure leading to increased belief in preparedness for future disasters. The protection motivation theory suggests that past disaster experiences could prompt individuals' expectation of disaster exposure, estimation of consequences and the belief that preparedness actions are effective in disasters and that they are able to do something to prepare for a future event, all of which may contribute to better preparedness (Gargano et al., 2015; Becker et al., 2017). While many studies, including ours, have shown a positive correlation between previous disaster exposure and feelings of preparedness, some have also shown that previous disaster experiences may not improve actual preparedness for future disasters (Rincon et al., 2001; Chan et al., 2014). In the study presented herein, with 76% (174 of 228) of those with previous exposure not being from the region, a potentially novel finding is that the type and geographical location of previous disaster exposure are not critically impactful in the degree of preparedness. Importantly, in this study, previous exposure to a disaster did not increase actual preparedness in regard to having a disaster plan or disaster kit. However, those who felt more prepared were more likely to have a disaster plan or disaster kit. An explanation for the difference in participants' feeling prepared vs truly being prepared with disaster items may be explained by the phenomenon of psychological preparedness. Individuals who have previously experienced this type of disaster may psychologically feel prepared, while in actuality, may not have taken anticipatory actions to physically prepare (Carroll et al., 2022). Many of the respondents who were not prepared for Irma but who made preparation efforts in anticipation of Maria were classified as “not prepared” for purposes of this study, due to lack of preparation in advance of the storms or when entering hurricane season. This lack of preparation might be attributed to respondents who were new to the island and did not have time to prepare.

In regard to specific items, individuals who believed they were prepared typically had key disaster kit items. Individuals tended to have these items at a ratio that appeared to be no different than that of those who felt less prepared. The significant number of respondents who indicated having a flashlight, emergency contact at a distant location, designated means of emergency communication plan, first aid kit, medication supply and important documents might reflect the island environment and necessity of having specific items on hand, as utilities may be frequently interrupted. Specifically, having a flashlight could signify the frequency with which electricity is lost (Times Caribbean Online, 2024), while having a supply of medications could indicate the challenges of obtaining certain medications on an island (A. Pierre, personal communication, June 2024).

However, the emergency contact and designated means of emergency communication plan might reflect the study population, with most respondents not being from St. Kitts. Many hurricane preparation pamphlets have extensive lists of items and recommendations for preparation prior to a storm, many of which tend to remain consistent across geographical locations (island, mainland). In an island environment such as St. Kitts, these items are often routine household items. As a storm nears, broadcast communications such as radio, television and Internet could therefore focus on recommending frequently omitted items which might improve overall preparedness.

Risk communication in disasters aims to prevent and mitigate harm from disasters, prepare the population before a disaster, disseminate information during disasters and aid subsequent recovery (Bradley et al., 2016). Communication research is diverse and frequently varies by disaster type, purpose of communication and communication source. This study focused only on how respondents gathered information about the hurricane and how they ranked the reliability of the source. A study by Senkbeil et al. (2020) evaluated aspects of communication with Hurricane Michael in Florida in 2018. It found that 26% of people reported using local media, 15% utilized national media, 20% used Wireless Emergency Alerts and community-generated cell phone alerts and 10% reported they used NOAA. In a survey on crisis communication research by Park and Avery (2018) it was found that regardless of crisis type, people who seek information about a crisis using traditional media (e.g. TV and radio) reported higher intentions to follow directives than social media and website users did. In the study presented herein, NOAA was the most utilized tool by the respondents and local media was the least used, which contrasts with the study by Senkbeil et al. (2020).

While all communication methods should provide reliable updates on a hurricane, given respondents' reliance on NOAA for storm-related information, there are opportunities for local media to provide more tailored information on community conditions. This could improve public safety by providing recommendations, including lists and locations of shelters for each parish, and emphasizing which communities are at greatest risk for life-threatening events (e.g. storm surge), which may increase community members' willingness to follow evacuation recommendations. These more tailored recommendations could also include updates on road, bridge and ghaut conditions and space availability in shelters. This information may be especially useful during the eye of storms when residents may attempt evacuation due to damage in their current sheltering location.

A large portion of respondents were short-term residents, specifically students at RUSVM. This transient population, representative of several geographical locations, enabled an assessment of preparedness based on diverse previous disaster experiences. While many islands are tourist destinations and can have large transient populations, these subpopulations, including our student respondents, might be less representative of island citizens and understanding their actions during disasters is important in emergency planning. Response actions that target transient populations have the potential to divert attention and resources from local relief efforts. For example, post-Hurricanes Irma (St. Maarten) and Maria (Dominica), evacuation of transient residents became a priority (Dyer, 2017). Emergency management decisions such as this could inadvertently put local residents at risk and contribute to the inequality felt in the region, with visitors getting seemingly preferential treatment over residents. In this study, some of the analyses were separated by residency length. While not analyzed for all factors, these populations responded similarly for a number of the aspects assessed. Further studies, including other islands where populations of residents are less transient, could confirm our findings.

Given recent trends of industry growth in the region, especially tourism, Caribbean Island populations represent a heterogeneous mix of resident and transient populations. This is an important distinction as these populations are fundamentally different in their respective sociocultural backgrounds, perceptions and lived experiences. This study demonstrated that transient and resident island populations can differ in their decisions regarding sheltering, ability to evacuate from at-risk areas and general preparedness in the face of disasters. Transient populations may have unfamiliarity with insular systems and infrastructure, limiting sheltering and evacuation options. This study elucidated that the idea of feeling safe in one's choice of place to shelter was a major driving factor in both transient and resident populations. This provides an opportunity for preparedness messaging by emergency managers to help residents make objective safety assessments of property and homes for the ability to withstand a hurricane and associated storm surge. Lastly, this study showed that transient and resident populations had many recommended disaster items already in their households, owing to life in the Caribbean, where services can be unreliable. Dissemination of disaster preparation information to island populations should therefore consider standard items found in these households, and targeted messaging for missing and needed items and preparations could be beneficial.

Operationally speaking, public and private organizations in the Caribbean need to recognize the dissimilarity between transient and resident populations in order to be stewards of effective crisis management plans. How tourists are managed in crises has been attracting specific attention as an industry practice and represents both a challenge and an opportunity for tourism and hospitality industries. This research may serve as a contribution to highlight this increasing concern. As private industry investment grows in this region, organizations could take a more active and responsible role in supporting transplanted or visiting individuals under their supervision. This also calls for coordinated planning and response by public and private entities for effective operational management of emergencies and disasters. Given the differences between island and resident populations and the increasing number and magnitude of disasters occurring in the Caribbean, more research focused on island population needs could contribute to better mitigation, preparedness, response and recovery efforts in this resource-limited but heavily traveled area.

The authors wish to acknowledge past members of the Ross University School of Veterinary Medicine Disaster Research Working Group present during the inception of this survey, including Elpida Artemiou, Heather Fenton, Pedro DePedro, Doris Castillanos, Robin Alexander and the Ross University School of Veterinary Medicine Research Center for Veterinary Education, Diversity and Data Analytics, who provided support.

The supplementary material for this article can be found online.