Eyeing the storm: How residents of coastal Florida see hurricane forecasts and warnings

Abstract

This paper examines the societal dimensions of warning decisions during extreme weather events in one of the most hurricane-prone areas in the U.S., Miami-Dade County, Florida. With the aim of informing improvements in the hurricane forecast and warning system, and better understanding warning decisions in extreme weather events, we explore how members of the public obtain and use hurricane forecasts and warnings in decision making. Results from in depth mental models interviews with members of the public (N=28) and survey data from three counties in Florida (N=460) show that a large majority of respondents have some hurricane experience, which influences their thinking about storm impacts, individual actions to mitigate the hazard, and vulnerability to the hazard. Comparison with results from previous research with warning system professionals (National Weather Service forecasters, media broadcasters, and public officials) indicates several gaps between professionals and laypeople including different perceptions of hurricane risks overall and related to flooding from storm surge. The findings suggest several areas for improvements in the hurricane forecast and warning system.

1. Introduction

In 2005, over a thousand people perished in Hurricane Katrina even though forecasters’ predictions of the hurricane’s landfall and potential impacts were fairly accurate (Knabb et al., 2005, Rappaport 2014). Hurricanes Harvey, Irma and Maria in 2017 and Sandy in 2012 also resulted in daunting human and economic losses despite forecasts and warnings. Hurricanes present unique challenges for forecast and warning systems and decision making in the face of risk and uncertainty. Long and increasing lead times accompanied by decreasing yet still significant track and impact uncertainties contribute to the complexity and intensity of multi-actor hurricane forecast and warning processes.

Recent calls for research to address the challenges of extreme event warnings call for interdisciplinary research and collaboration on hazards and their management (Aitsi-Selmi et al., 2015a,b; Jones and Golding, 2014; Morss et al., 2017; NASEM, 2017b; NRC, 2006b; NSF, 2002). Assessments underscore the importance of better understanding the entire warning process, and the promotion of “risk-wise” behavior, particularly with respect to forecasts and warnings (NASEM, 2017b; NSB SDR, 2005; also Wei et al., 2014). One of the most significant challenges for risk communicators (including hurricane forecasters, public officials, and media personnel) is to issue forecasts and warnings and communicate risk information and information about protective action in ways that inform and promote appropriate governmental and other organizational decisions and self-protective actions by at-risk populations. These are critical for reducing economic impacts, loss of life, and injuries during extreme weather-related events. To tackle such challenges requires better characterization of “the reactions of both the general public and government officials to hurricane-related information and the manner in which such information is most effectively processed and shared” (NSB, 2007: p. 20; see also NRC 2006a, 2010; NASEM, 2017b). This paper responds to these calls by exploring how laypeople in a hurricane-prone area of the U.S. perceive hurricane risks, hurricane forecast and warning information, and decisions about hurricane risk management. The study reported here aims to inform improvements in hurricane forecast and warning systems, and warning decisions in extreme weather events, as part of a larger project (Bostrom et al., 2016; Lazo et al., 2015; Lazrus et al., 2016; Morss et al., 2015, 2016b).

Developing a broader understanding of the roles different parties play in warning decision making and communication processes can inform improvements in hazard warning systems. Social scientists have a longstanding interest in extreme event warning processes (Anderson, 1969; Mileti, 1975, 1999; Mileti et al., 2004; Mayhorn et al., 2006; NRC 2011; Mileti and Sorensen, 1990; Sorensen, 2000; Sorensen and Mileti, 2017 a,b,c). Many studies in this area focus on factors that influence decision making with respect to warning compliance, for example evacuation from hurricanes or tropical cyclones. Empirically based causal models of warning responses show that individual and household decisions are shaped by a variety of factors, including not only attributes of forecasts and warnings themselves, but also experiential, perceptual, psychosocial, socio-economic, household, and community-level factors (e.g., Ahsan et al., 2016; Bowser and Cutter, 2015; Cova et al., 2017; Huang et al., 2016; Lazo et al., 2015; Lindell and Perry 1992; Sadri et al., 2017 in press; Sharma and Patt, 2012; Sharma et al., 2009; Sorensen, 2000; Stewart, 2011; Thompson et al., 2017; Tierney et al., 2001). Much is already known about warning decision processes from this growing body of research, for example that official warnings and mandatory evacuation orders increase evacuation intent and behavior (e.g., Thompson et al., 2017), and that the quality of shelters can be influential (Kulatunga et al., 2014). Nonetheless, much remains poorly understood. For example, most research on self-protective behavior in extreme weather events does not focus on what specific elements of warnings were most influential in motivating actions or mitigating the impact of forecast and warning uncertainty on those actions (for recent exceptions see Cuite et al., 2017; Morss et al., 2016a). Moreover, few studies utilize research designs that explicitly connect such findings to the forecast and warning system.

Filling the knowledge gap on warnings and warning systems requires integrated research across disciplinary boundaries to develop an end-to-end understanding of the processes involved in forecasts and warnings. Questions regarding warning assessment, dissemination, and compliance clearly require multi-disciplinary and interdisciplinary exploration, with collaboration among the sciences focusing on organizational, group, and individual behavior with respect to extreme weather warnings.

However, weather researchers and forecasters have developed much of the forecast and warning information for hurricanes and other extreme weather events, with a focus on analysis of hydrometeorological data, forecast production, and forecast dissemination as the primary function of their professional responsibilities. Geophysical and atmospheric scientists seek to make their data, models, and forecasts as accurate and timely as possible, but often without knowing whether additional accuracy actually encourages more effective warning decisions by those affected.

Although physical scientists and forecasters have an interest in creating and communicating forecasts in ways that aid effective decision making, the expertise and data necessary to support this effort has, until recently, been lacking (e.g., NRC 2006a; Broad et al., 2007; Demuth et al., 2012; Pielke 1999; Stern and Easterling, 1999; Morss et al., 2008). The meteorological literature has little empirical research on how to most effectively enhance forecast value to diverse decision makers, particularly for extreme weather events. Often, physical scientists assume a direct relationship between forecast quality and the value of forecasts to users (including positive warning outcomes). In other words, they believe that improving a forecast will enhance forecast value, and that improving forecast quality is one of the best ways to increase forecast value. Yet, ample evidence suggests these premises do not always hold (e.g., Murphy, 1993; Roebber and Bosart, 1996). For example, the aforementioned damage caused by Hurricanes Katrina, Sandy, and Harvey occurred despite accurate forecasting.

This paper explores how to enhance the value of forecast and warning information by examining public perceptions in the context of the entire warning process. As noted above, the warning process for extreme weather events, particularly hurricanes due to their long lead time, is an interactive and iterative process that encompasses many important actors and factors. Building on research with hurricane warning system professionals (e.g., Demuth et al. 2012, Bostrom et al. 2016), this paper examines how public perceptions and decisions relate to the roles, decisions and products from forecasters, public officials, and broadcast media. Specifically, this paper addresses the following questions:

1. How do members of the public perceive hurricane risks, impacts, forecast and warning information, and risk management decisions?

2. How do members of the public obtain and use forecast and warning information for hurricanes? What decision processes and mental models underlie their behaviors with respect to hurricane warnings?

3. How can this information be used to make hurricane warnings more effective?

To address these questions, we report here on public interviews and surveys collected as part of a larger project designed to study the warning process holistically, with an emphasis on how information—including information on uncertainty—is interpreted, shared and used by citizens (Bostrom et al., 2016; Lazo et al., 2015; Lazrus et al., 2016; Morss et al., 2015, 2016b). The project aims to identify commonalities and variations in both information preferences and the use of forecast and warning information across different users; explore more extensively the decision processes and mental models that underlie behavior with respect to warnings for extreme weather events; and develop guidance for improving warning system effectiveness that is informed by the results.

Building on data from semi-structured interviews and survey research, this study combines descriptive induction with tests of three general hypotheses, corresponding to the first two research questions above: (1a) stakeholders differ in their perceptions of the warning process; (1b) there are gaps and inconsistencies in perceptions, beliefs, and goals between forecasters and forecast intermediaries (public officials and media), and information users (here, members of the public); and (2) current communication products (common at the time of the interviews) that represent uncertainty in hurricane and flash flood forecasts and warnings communicate ineffectively or are confusing to decision makers.

The next section of the paper describes our data collection, coding and analytical methods. The subsequent results section is structured as followed: (a) demographics, (b) hurricane culture, (b) storm impacts, (c) individual mitigation, (d) vulnerability and risk, (e) other parts of the forecast and warning system, (f) awareness and omissions, and (g) direct recommendations from interviewees regarding how to improve the hurricane forecast and warning system. The concluding discussion section summarizes implications for the forecast and warning system and future research.

2. Methodology

This research aims to understand how members of the public perceive hurricane impacts, exposure, and mitigation (including the forecast and warning system) by eliciting their mental models. Mental models are internal representations of external realities. The concept of small-scale models that the mind produces and employs to anticipate events and guide experiences can be traced back to Bartlett (1932), Craik’s 1943 book The Nature of Explanation, and more recently to Gentner and Stevens’ book Mental Models (1983). In this project people’s internal representations are elicited through the use of mental model interviews and a survey. A broader understanding of their mental models will help to identify their specific thinking about risk, and the decision process that they engage in and will provide insights into how the public uses forecasters’ warnings (de Bruin and Bostrom, 2013; Bostrom, in press; Morgan et al., 2002).

2.1 Miami/Dade County Interviews

2.1.1 Interview Sample Selection

Miami Market Research was contracted to recruit interviewees in August 2010 in one of the most hurricane prone areas in the United States- Miami/Dade County, Florida. From a total of 753 random digit telephone calls, 28 members of the public were recruited and interviewed individually and completed an accompanying questionnaire. The interviews were conducted in English by three interviewers involved in the research project using standardized protocols, in person at Miami Market Research facilities. The professionals interviewed were also from Miami/Dade County, as described in Bostrom et al. (2016).

2.1.2 Individual Mental Model Interviews

The goal of the interviews was to assess how people understand and make decisions about hurricane risk while imposing as little as possible of other people’s ideas, perspectives, and terminology (Morgan et al., 2002). A snapshot of the first page of the protocol for interviewing the public participants is provided in Figure 1. This protocol was modified slightly for the professional interviews, to align with their professional roles in the hurricane forecast and warning system (see Bostrom et al., 2016).

Figure 1.

Introductory section of the interview protocol.

As shown in Figure 1, the interviews were conducted using open-ended questions that began with very general questions about hurricanes. These were followed by questions about exposure to and effects of hurricanes both generally and in Miami-Dade (e.g., “What risks are there from hurricanes?” and “What might happen to a person who was in a hurricane?”), forecasts and warnings, the role of forecasts and warnings in decision making, and mitigation decisions (e.g., “What can or should be done, if anything, to reduce risks from hurricanes?”), as well as questions exploring how these in turn influence the consequences of hurricanes.

In the interviews, these open-ended questions were followed by structured, closed-ended questions about (1) the precise meanings and interpretations that interviewees have of sample forecasts and warnings, and (2) based on these interpretations, the behaviors that interviewees would recommend to their friends and family. This included the exploration of additional factors that interviewees think would influence their own and their friends’ and family’s decision making and actions. Finally, interviewees were asked to think aloud as they made a hypothetical decision based on one or two experimental products, such as the Experimental Tropical Cyclone Hazards impact graphic from the NWS.

The public interviews were transcribed verbatim by professional transcriptionists. They were then quality controlled by interviewers who compared several of the transcriptions to audio recordings to ensure that discussions were captured verbatim.

2.1.3 Interview questionnaires

After each individual interview, the interviewee filled out a written questionnaire. Designed primarily to pre-test the structured questions for the subsequent public survey, the questionnaire presented interviewees with increasingly structured questions about their general risk perceptions, probability estimates for hurricane events, beliefs about the impacts from hurricanes and understanding of forecasts and warnings. Basic demographic information was also collected. The results from this questionnaire are reported as part of the interviews, and not as part of the survey. This pretest questionnaire was subsequently revised for the larger survey.

2.1.4 Interview coding and analysis

The final coding scheme applied to the interviews includes 189 different codes, and covers forecast and warning processes from hurricane vulnerability and preparedness through storm development and monitoring, decision making—including forecast and warning product development, dissemination, and response decisions—and consequences (Figure 2). The scheme is hierarchical; lower level codes are more specific variations of their parent codes. Details of the interview coding scheme development and data analysis are provided in Appendix 1.

Figure 2.

General structure of coding scheme (from Bostrom et al., 2016). Storm impacts are discussed in section 3.4, Individual and Community Mitigation in section 3.5, Vulnerability in section 3.6, and other elements of the system in 3.7, including forecast communication products in section 3.7.2. For the complete coding scheme see Appendix 2.

Concepts that are mentioned by a high proportion of the sample in response to an open-ended question are interpreted as salient to the public. Overall coverage of the model expresses breadth of public understanding, whereas omitted (not mentioned) concepts represent potential knowledge gaps. We present the results using both quantitative analysis of the percentage of interviewees mentioning different concepts, along with quotes illustrating key points.

To enrich the context for interpreting key results of our public interviews, we compare them to responses from select samples of professionals in Miami-Dade County, including U.S. National Weather Service (NWS) forecasters (weather forecasting office WFO), emergency managers (public officials PO), and media broadcast meteorologists (BR) from Bostrom et al (2016). Mental models were elicited from these professionals using a similar interview protocol to that used here, and coded and analyzed using the same coding scheme (Bostrom et al., 2016).

2.2 Florida survey

2.2.1 Survey development and instrument

The survey enables an assessment of the reliability and comparability of open-ended responses, to permit generalization beyond the small interview sample. Survey development drew on related prior and concurrent research on the sources, communication, perceptions, uses, and value of hurricane information (Lazo and Waldman, 2011; Lazo et al., 2010); on mental models of hurricanes and flash floods (Bostrom et al. 2016; Morss et al. 2016; Lazrus et al. 2016); and on communicating hurricane information (Demuth et al. 2012; Lazrus et al. 2012; Morss et al 2016). Questions on hurricane experience, perceived risk, perceived use of information, and demographics are reported here. As noted above, we piloted many of the questions in the survey in the questionnaire administered at the end of the interviews. In addition, the survey was pretested in three cognitive interviews (Willis, 2015) using a hard copy of the draft survey with individuals in Boulder, Colorado who had previously lived in Miami. Following revisions based on the cognitive interviews, the survey was programmed online by Knowledge Networks (KN) and pretested with a random subset of 33 individuals from the full survey sample. Because a significant portion of the population in Miami is primary Spanish speaking, the survey was translated into Spanish by KN and offered in English and Spanish to all respondents.

2.2.2 Survey sampling and implementation

The target population consists of members of the public who are at least 18 year old and reside in one of the three coastal Florida counties in the Miami area (Broward, Miami-Dade, Palm Beach). The sampled households are from the KN (now GfK) KnowledgePanel, a probability-based web panel designed to be representative of the United States (see Rodkin and Lawrence, 2012 for details on KnowledgePanel® recruitment methodology), with sampling weights to support generalizations to the study area population.

The survey was implemented from May 4, 2012 through May 24, 2012. Email reminders to non-responders were sent on day three of the field period. The completion rate for the survey was 61.6% overall. The data analyzed here are from the 460 KnowledgePanel adult (18 and older) panelists in Florida who completed the survey, 70% in English and 30% in Spanish. The median time for survey completion was 26 minutes. KN provided demographic profile data for all respondents, including socio-demographic information of specific interest for our work (e.g., length of residence in a hurricane-vulnerable area), elicited at the end of the survey. Respondents were geo-located as described in Lazo et al. (2015).

2.2.3 Survey data analysis

Following compilation of the data set, we quality controlled the data and assessed summary statistics and missing values. For the current analysis, missing values have been replaced with the median, mean value, or more conservative response category as appropriate. The majority of items had less than 1% missing data, and none had greater than 5% missing data. These data are from the same survey used in Lazo et al. (2015), but here we report only the Florida respondents, and only those survey items relevant to our interview results.

3. Results

3.1 Demographics of the interview sample

Of the 28 public interviewees in Miami-Dade County Florida, 17 were female, 21 owned their own homes, condominiums, or townhomes. Nine of these were white males (including “other-Hispanic”), seven were white females, two were black males, two were black females, and one was a mixed-race male. Of these, five were Hispanic or primarily Spanish-speaking. Interviewees were between 18 and 73 years of age, with an average of 47 years. Eight of the homeowner interviewees worked full-time, four worked part-time, six were retirees (of which one was a “part-time retiree” and was counted as both part-time and retiree), two were homemakers, and two were unemployed. On average, each household had three residents, with a mean household income of $60,000-$74,999 (range: less than $15,000 to over $200,000). The overall pool of 28 interviewees was highly educated: 89% of interviewees had some kind of college or technical schooling, including 11 Bachelor degrees, three Master degrees and one Professional or Doctoral degree. Of the three interviewees who had not been to college, two had earned a high school diploma or equivalent. For comparison, in Miami-Dade County, Florida at the time the study was conducted the population was 52.1% female and 64% Hispanic, with a median age of 38.6, 17.5% were aged 60 or over, and 27.3% held a Bachelors degree (Miami-Dade County, 2015).

The majority of homeowner interviewees (19) reported that they had lived in their current location for between 5–25 years; 4 had lived there longer, and 5 for a shorter period of time. The owned homes were located an average of 9.6 miles from the beach, with a range of 0 to 30 miles. Only 11 of the 21 owners responded to the question about elevation above sea level, and of those who did respond many were unsure. All but two reported having properly functioning hurricane shutters.

Of the seven renters, four were white males; the others included a white female, a black female, and a black male. Of these, three were Hispanic or primarily Spanish-speaking. Five of the renters worked full-time and the other two were retired. On average, each rental household had 1.3 residents with a mean household income of $30,000-$44,999 (range: less than $15,000 to $100,000). The rentals were located an average of 15 miles from the beach, with a range of 1 to 50 miles. All but one reported having properly functioning hurricane shutters.

All but four of the interviewees reported having been personally affected by at least one previous hurricane, such as Hurricanes Andrew (1992) or Wilma (2005). Respondents are identified in the results by their interview number; in addition, the first time each interview is noted, information is provided about their race/ethnicity, gender, age, homeownership status, and where reported, the name of the most recent hurricane they experienced.

3.2 Demographics of the survey sample

Of the Florida survey sample (N=460), 45% were male, 55% female. The mean age was 50; 28% percent were age 60 or older, and 15% were 18–29 years old. Forty-two percent of the sample self-identified as Hispanic. Education levels varied, with 36% of the sample reporting less than high school or a high school education, and 35% as having a Bachelors degree or higher (average 14 years of education). By comparison with the Miami-Dade County statistics in section 3.1, the sample was a little less Hispanic, older, and more educated than the population of Miami-Dade County at the time.

Survey respondents had lived an average of 25.6 years in a hurricane vulnerable area, and 70% owned their residence, while 51% lived in a single detached house. Ninety-five percent reported having Internet access. Note that KN provided access to the online survey for those who did not otherwise have Internet access.

As in the interviews, most of the survey respondents reported some form of personal experience with hurricanes. More than three quarters of those surveyed said they had personally been affected by at least one previous hurricane; and 69% of survey respondents said that they had experienced moderately to extremely severe hurricane impacts (Figure 3a). Figure 3b provides some insight into their personal experiences. Distress is fairly common among those who report that they have been personally affected by a hurricane (68.6%) but also reported—those less commonly—by those who report they have not been personally affected (40.5%) (Figure 3b). Predominantly reported by those who identify themselves as having been affected are personal or household experiences of property damage, evacuation and financial losses (Figure 3b; see Lazo et al., 2015 for additional discussion).

Figure 3.

a. Hurricane experiences in the Florida survey sample, showing severity of experience subdivided by whether or not the respondent says they have personally been affected by a hurricane.

b. Specific hurricane experiences in the Florida survey sample, subdivided by whether or not the respondent says they have personally been affected by a hurricane (N=430, Florida).

3.3 Hurricane culture and its elements

The hurricane culture in Florida emerged as a key theme across many of the interviews. As noted by one interviewee in response to the opening prompt, “Hurricanes in Miami. Well um it’s a big part of Miami—living in Miami.” (#17, white Hispanic male age 21, homeowner, Wilma)

In their responses to the initial, open-ended general questions about hurricanes, public interviewees discussed primarily storm development, storm behavior, vulnerability to hurricanes, and mitigation efforts. With regard to storm development, they talked about a storm’s location, wind speed and category designation, season (timing), water temperature, pre-existing disturbances, and long-term trends and patterns. With regard to storm behavior, interviewees mentioned the wind speed designation at which point storms either get a name or reach hurricane designation. In these responses, interviewees discussed their vulnerability to hurricanes in terms of their personal experiences and perceptions of risks. The mitigation efforts they mentioned included hurricane education and evacuation procedures (e.g., where to seek shelter).

In the remainder of this section, we analyze responses across the interviews, with results organized primarily by the categories in the hurricane forecast and warning system model in Figure 2 and corresponding coding scheme. The model comprises nine high-level categories (Figure 2), explained in more detail in Bostrom et al. (2016). Of these nine categories, three are mentioned by over 75% of public interviewees: storm impacts, individual mitigation, and vulnerability. Results from the interviews within each of these salient categories are followed by presentation of results from the other six categories, then interviewees’ recommendations for improving hurricane forecast and warning. Results from relevant survey questions are presented after the interview results within each section as appropriate. In this way, the survey data are used to complement key themes that emerge in the mental models interview analysis, and to investigate the extent to which these themes generalize to a broader population in coastal Florida.

3.4 Storm Impacts

The public interviewees discussed multiple aspects of the destruction that can come from hurricanes (Figure 4), likely due to their own experiences with hurricanes (discussed above). As shown in Figure 4, all of the public interviewees mentioned death and property damage as impacts of a hurricane, and over 75% of them mentioned loss of power (89.3%), injury (82.1%), inland flooding (78.6%), or social/economic impacts (78.6%) as other major concerns. Most interviewees acknowledged that hurricanes of all sizes can cause damage. Some noted that faster-moving hurricanes may do less damage or at least “[do] damage real quickly,”(#7, white female, age 63, homeowner, Wilma) while slower-moving, wetter hurricanes stick around longer in one place and continue to cause damage.

Figure 4.

Percentage of interviewees who mentioned each of the storm impact concepts in the coding scheme. Percentages are not rolled up (i.e., the coding scheme is hierarchical, but the numbers in the figures do not include counts of mentions of more detailed codes). Red indicates fewer mentions, green more. NWS Weather Forecasting Office (WFO), emergency manager (public official, PO), and broadcaster (BR) data (representing mentions of concepts by different warning system professionals) were developed as reported in Bostrom et al (2016).

Physical Impacts

For both the professional and public interviewees, the two most commonly mentioned forms of physical impacts from hurricanes are damage to property (100%) and power outages (89%). The public interviewees discussed a range of property damage, from as minor as an uprooted tree or doghouse, to as serious as lost roofs, structural cracking, and fire. Interviewees noted how damage from wind or flooding often results in structures’ loss of integrity. In some cases, there is a risk of completely losing one’s home or place of business, which for some interviewees makes a strong case for properly preparing for hurricanes in advance. Regarding loss of power, a common and potentially life altering impact discussed by interviewees is the length of time without power. Interviewees’ experience ranged from as little as three days without power to months in the case of powerful storms like Andrew or Katrina. “Wilma we were without electricity like for two weeks. […] No electricity for two weeks. Wilma did extensive damage. Of course, Andrew was worse but Wilma did extensive damage.” (#14, black female, age 46, homeowner, Wilma) This potential for extended loss of power is one of the main reasons people are cautioned to gather supplies to ride out the storm and its aftermath. As one respondent (#23, white male, age 65, homeowner) noted: “When the power is out, then that has like a rolling effect on everything.” In a severe hurricane, interviewees noted that life may not immediately return to normal and people will have to fend for themselves until emergency officials can arrive.

Along with the loss of power come downed power lines, which become especially dangerous when combined with standing flood waters. Several public interviewees mentioned electrocution in the presence of downed power lines, when describing death as a hurricane-associated risk: “Well the one with the worst consequences is the…is the power lines. I think. I mean if you get electrocuted, you’re pretty much done for.” (#17)

Slightly over half of public interviewees mentioned damage to land or land reconfiguration (60.7%) and damage to other infrastructure systems such as roads and public transport (57.1%). Slightly less than half mentioned damage to drinking water systems and the following lack of [clean] water that can accompany a hurricane (46.4%).

Human Health Impacts

All of the public interviewees acknowledge that death is a serious hurricane risk, followed closely by injury (Figure 4). Interviewees discussed a variety of types of risks during and after hurricanes, including getting swept up by wind, being hit by flying debris, drowning and electrocution. As this illustrates, hurricane risks are not taken lightly. Interviewees generally discussed the need to seek shelter – either in a secure home or at a shelter to stay safe from the dangers of hurricanes. As one interviewee noted, however, sheltering decisions do not always correspond directly to health impacts: “I’ve seen a lot of homeless people go under bridges and all that and survive it… And yet I’ve seen people in houses not survive it.”(#21, white Hispanic male, age 51, renter, Andrew)

Public interviewees also noted the risk of death or injury during hurricane preparation and post-hurricane cleanup as well as during throughout the actual event. As one interviewee noted: “A lot of people die from the generators.”(#11, white male, age 56, renter) The injuries mentioned by interviewees had a variety of causes, including falling off roofs, dehydration, water-related infections and diseases, and driving in unsafe road conditions.

While most public interviewees mention the need for people to stay alert and be cautious to avoid harm during a hurricane, many interviewees mention the death and injury of people who do not heed hurricane warnings. Some recalled those who failed to evacuate or take shelter in prior hurricanes, for example those choosing to surf or throw hurricane parties. One interviewee explained: “You’d see ‘em on TV. They’d interview ‘em, they would say well we’re having a hurricane party. We’re just going to get drunk and – and ride it out. And to tragic ends sometimes…I mean people died because they didn’t take it seriously enough.” (#23) As one interviewee said about hurricanes, “you have to take ‘em seriously.” (#7)

Further information about Floridians’ perceptions of hurricane health risks is available from the survey, which asked a question about the likelihood of five possible causes of death in a hurricane. Survey respondents judged electrocution from downed power lines as most likely (mean=3.28 on a scale from 1=not at all likely to 5=extremely likely, s.e.=0.057), and drowning as least likely (mean=2.41; paired t-test compared to downed power lines, t=13.33, df=421, p<0.001).

Another health impact, mentioned by over half of public interviewees, is psychological trauma (53.6%). This includes fear and anxiety. For example, when asked if any other effects of hurricanes came to mind, one renter said:

“Yes, stress. Stress and anxiety on – on a lot of people. Especially homeowners. And um even people who – who don’t own homes, but like elderly. People that are scared a lot. Children and – and people who are afraid of things in general. I’d say they really get panicky during the hurricanes. It causes anxiety and stress levels really go up. It causes different effects in different people. Differently. Affects ‘em psychologically differently. Also animals.” (#11)

Social and Economic Impacts

Over three quarters of interviewees mention social and economic impacts (Figure 4); examples include temporary and permanent job loss, criminal activity, loss of general services, and homelessness. Half of interviewees mentioned cash shortages (50%). Many discussed spending large sums of money to purchase supplies and hurricane proof their homes. Multiple public interviewees mentioned crime, especially looting, as likely both before and after hurricanes. Thus, one homeowner remarked that if so inclined, people “should have a weapon.” (#23) Other negative social impacts mentioned included insurance fraud and—correspondingly—difficulty getting insurance. Also mentioned were increased prices at grocery stores and home improvement stores, which—one interviewee perhaps jokingly mentioned— might hope for a hurricane to drive up profits (#11). Similarly, when asked about hurricanes in Miami, another respondent opened with “Well a lot of people are complacent about it. And then they get serious, you know, because we get so many near misses. And we all think it’s, you know, drummed up by Home Depot and Publix to sell water and plywood.” (#21)

Public interviewees also tied psychological, social and economic impacts together in their discussions. As one interviewee explained:

“Besides just the damages and the cost. The expenses and the fear that it creates. And the havoc. I mean people freaking out before the hurricane comes, and then I guess after the hurricane. Just all the you know living without no power and all that. You know you become a third-world country all of a sudden. You know overnight. There’s no food. There’s no electricity. There’s huge lines for gas and water, the little bit there is. And you know trying to get your generators. I mean it’s funny how a whole—you know the whole economy gets affected. Or the whole city. All a sudden you’re you know a third-world country overnight.” (#19, white male, age 43, renter).

Impacts from Wind and Flooding Hazards

Hurricanes bring with them multiple hazards, including strong winds, flooding due to rain and storm surge, and tornadoes. All of the public interviewees mentioned wind speed as a storm characteristic, which suggests a strong understanding that wind is a strong contributor to this destruction. Approximately two thirds also explicitly discussed the damage that hurricanes can cause by wind (Figure 4).

Flooding from hurricanes is both complicated and central to the risks of hurricanes. Two major types of flooding arise from hurricanes: flooding from rain (often inland) and coastal flooding from storm surge. Both types of flooding are prominent in professionals’ discussions of hurricane impacts (Figure 4). Among public interviewees, however, over three-quarters of interviewees mentioned inland flooding, but only 7% (2 of 28) mentioned storm surge flooding, although some may have been thinking of storm surge when they mentioned flooding in coastal areas. While this may be because storm surge is not as major an issue in Miami as it is in many other coastal areas, due to the local bathymetry, it is still a surprisingly low percentage.

Commonly, hurricanes bring large amounts of precipitation, as often mentioned by interviewees. Many described torrential downpours causing flooding to their property. Others discussed their own hurricane experiences and recalled staying inside listening to the wind and rain outside. When asked about hurricane decisions, one respondent recalled the following about Hurricane Wilma: “even if there’s school, once the flooding in our area—the water was high. We wouldn’t send our son to school. I wouldn’t go to work. That was some of the decisions we made.” (#14) A few also discussed the heightened threat to those on the coast, living in flood zones, and those in homes that are not hurricane proof. When asked if any specific locations in the Miami/Dade area where hurricanes were more likely to have an impact, one respondent said “I think it’s all pretty you know, it’s on sea level so it’s all kind of pretty risky. But definitely the coastal—coast line. And the Keys.” (#2, white male, age 28, homeowner, Andrew)

Survey responses were consistent with the interviews (Figure 5), in that inland flooding was judged more likely than flooding from storm surge. High winds, intense rain, and blowing objects or debris were judged the most likely hazards associated with hurricanes with each of these rated by over half of respondents rating as extremely likely (Figure 5).

Figure 5.

Judgments of the likelihood of different hazard conditions from a hurricane, from the Florida survey sample. The number of responses ranges from 420 to 453 per weather condition, due to item nonresponse. Survey question: “How likely would each of the following conditions be in the general area where you live if a major hurricane (Category 3 or higher) hit your area?”

3.5 Individual and Community Mitigation

The public interviewees discussed a variety of ways that individuals can mitigate the impact of a hurricane (Figure 6). More than three quarters mentioned education (96.4%), home protection (96.4%), emergency supplies (96.4%), evacuation (89.3%), or securing loose property (75%). More than half mentioned insurance (67.9%) or building codes and land use planning (57.1%) as additional ways to diminish hurricane impacts. Fewer mentioned trimming trees (39.3%) or moving away from the danger area (28.6%).

Figure 6.

Percentage of interviewees who mentioned individual and community mitigation and decision concepts in the coding scheme. Red indicates fewer mentions, green more. WFO, PO, and BR reported in Bostrom et al (2016).

Hurricane Education

According to many of the interviewees, education is a key to helping people better prepare for hurricanes. As described by the public interviewees, much of the information they see comes from the Internet, TV segments, and pamphlets available at grocery stores. Interviewees remarked that the education they receive in the form of taped information segments on TV, brochures, signs, and other methods of information dissemination provides them with useful information.

Some interviewees mentioned that the city of Miami-Dade prepares them well for hurricanes, while others would like to see more information made available. These differing opinions ranged from “They’re very good in Miami about… telling us where to get pamphlets and information” (#14) to another who would like to see this dissemination made mandatory and given out in schools and private businesses. One interviewee, appreciative of the work Miami-Dade does to prepare its residence said “they don’t want anybody getting hurt and anybody…. You know and the news media and the papers and everybody’s very conscious of this.” (#10, white male, age 73, homeowner)

Home Protection

Nearly all of the public interviewees mentioned the importance of people preparing their homes for an impending hurricane, no matter the category. As one interviewee recalled: people should “[make] sure that [they] have [their] hurricane shutters… at least have you some plywood around and duct tape. Plan for prevention of damage to your house.” (#18, Hispanic female, age 43, renter, Andrew) While plywood can prove useful and new hurricane resistant windows were purchased by a few, most interviewees turn to hurricane shutters. Example comments included “my house is well protected. Shutters, everything” and the simple “we always put up the shutters.” (#12, white female, age 61, homeowner, Andrew)

Public interviewees recognized that properly preparing homes saves lives. As one homeowner stated: “I think that if your house has been built, or is up to code…hurricane code…and you have shutters, or hurricane-proof glass and you’re in a sound structure, I would say maybe you don’t need to worry so much about you know death or damaged property, or the loss of your home.” (#25, white female, age 32, homeowner, Wilma). Another homeowner discussing home protection explained: “they can make sure their home is ready…and that’s shutters, protection from wind damage over the windows and doors” (#23). Most interviewees focused primarily on home protection from wind rather than flooding.

Emergency Supplies

Consistent with the interviewees’ perceptions of hurricane impacts (discussed above), nearly all public interviewees noted that to prepare for hurricanes, people must take action not just with regard to structures, but also to their own personal well-being, which means ensuring adequate supplies are purchased in advance. As one interviewee shared, “We started preparing from the time we heard about the watch. I would go to the store and start buying… By the time the warning came around, we had everything.”(#14) Necessary supplies discussed by the interviewees included water, non-perishable food, power sources like generators and fuel, flashlights, additional batteries, medication if necessary, and, according to a couple of interviewees, weapons to fend off looters.

As the quote above illustrates, some public interviewees discussed going to the store to purchase their supplies—either well in advance or upon hearing hurricane warnings. Interviewees mentioned news outlets as a source of tips for better preparation including where to purchase items, where to find gasoline for cars and generators, and where the shortest lines are at different stores like Publix (a grocery store chain), gas stations, and Home Depot. One homeowner explained:

“You can find out generally where one’s [a hurricane] going by news media. Um you know all of the media stay on top of it. As a matter of fact, they go—again—they run their truck down to Home Depot at the first dust storm that leaves Africa. You know, ‘we’re here in front of Home Depot’ and then there’s some buzzwords that they use. We joke about it. But in the end they do a good job, as it gets closer and…. And also they have that list. They publish that preparedness thing too. They make that available, and uh all of the TV stations do.” (#23)

Securing Loose Property, and Pets

Also consistent with their discussions of hurricane impacts, many interviewees noted the importance of bringing in or tying down loose objects, due to concerns about flying debris. If left unsecured, items like patio furniture can be lost or damaged or can become deadly if picked up by the wind. Some bring their furniture inside if not tied down, or, according to one interviewee, in order to best secure patio furniture, “you keep ‘em in the pool.” (#16, white male, age 21, homeowner)

Securing loose objects also involves looking out for pets; some recalled lost dog houses and missing pets that “were found far away” (#2) or whose owners could not be identified. Not surprisingly, and as is evident from the earlier discussion of stress to animals as well, pet safety is seen by some interviewees as an important part of hurricane preparation.

Evacuation

Although 89% of the public interviewees mentioned evacuation as a form of protection from hurricanes (Figure 6), for most of the interviewees, evacuation to other counties or states prior to a hurricane is rare and reserved only for the strongest hurricanes and those people in poorly constructed homes. Most of the interviewees chose to ride out hurricanes in their own homes or homes of friends and family. Many discussed sheltering in a safe room in a house like “an inside room… a bathroom, or a closet.” (#7) This perspective on evacuation is consistent with interviewees’ lack of emphasis on storm surge risks, discussed above, since potential for loss of life due to storm surge is typically the primary motivation for ordering hurricane evacuations.

3.6 Vulnerability and risk (other than mitigation)

For the concepts in this section of the coding scheme (Figure 7), all of the public interviewees mentioned some sort of previous experience with hurricanes. Nearly all mentioned that people “should” have a hurricane plan (96.4%), and the effects of perceived hurricane risk (92.9%). Physical aspects of vulnerability and risk such as building preparedness (75%) and policy factors such as land use development patterns (60.7%) were also mentioned by many interviewees. Fewer mentioned other human factors such as population size (0%), density (3.6%), age (25%), or language diversity (21.4%) as contributors to vulnerabilities and risk.

Figure 7.

Percentage of interviewees who mentioned vulnerability concepts (other than mitigation). Red indicates fewer mentions, green more,).

Hurricane Experience

The hurricane experiences mentioned by public interviewees ranged from the extreme devastation of hurricanes like Andrew to hurricanes that failed to make landfall. Many interviewees take risks before, during, and after hurricanes seriously, and these worries are affected by prior hurricane experience and related feelings of preparedness. Interviewees discussed the worries they face when it comes to preparing, evacuating, or taking shelter. The more memorable storms for interviewees were those for which they felt less prepared, especially Andrew, when they did not realize the extreme risk ahead of time. As one interviewee recalled, “nobody really knew squat about preparing. And nobody ever did anything … That was definitely memorable ‘cause then you feel like you’re behind the eight ball from the beginning.”(#9, white female, age 45, homeowner, Wilma)

Risk Perception

As the previous section illustrates, public interviewees’ perceptions of hurricane risk were connected to their hurricane experience. Interviewees’ thoughts on the strength and devastation of hurricanes vary, ranging from “very scary” (#12) to “a pain in the butt” (#19).

To further examine Florida residents’ hurricane risk perceptions, we use data from a survey question that asked about the likelihood of different types of hurricane impacts (Figure 8). Results suggest that, across the different types of risks, professionals (forecasters, public officials, broadcasters) tend have somewhat higher risk perceptions than do public respondents. Further, public interviewees see injuries as less likely than property damage (paired t-test=4.995, df=26, p<0.001), while professionals see injuries as about as likely as damage to buildings (Figure 8, see bottom two sets of bars).

Figure 8.

Mean perceived risk of different types of hurricane impacts, for Florida survey respondents (N=460) compared to the WFO, PO and BR professionals (N=12) studied in Bostrom et al. (2016). Survey question: “A “major” hurricane is any hurricane of Category 3, 4, or 5 – that is a hurricane over 110 miles an hour. How likely do you think each of the following Impacts would be in the general area where you live if a major hurricane hit your area?”

Hurricane Plans

Public interviewees noted that having a hurricane plan helps reduce both the risks and the worries that people face. They discussed a hurricane plan as knowing where to go in a hurricane, having shutters up, and buying supplies: “Make sure you have shutters. Make sure you – you’re not anywhere by the ocean. Make sure you don’t walk in floods. You know the type of things – there’ll be little headlines on the TV or something that’ll tell you what to do. …They tell you tricks and stuff like what to do.” (#16)

Some interviewees mentioned the acceptability of being over-prepared in the context of planning, with statements such as, “if it doesn’t come and we have to take the shutters down, that’s fine. We—we don’t see it as a problem. We see it as a good thing.” (#12)

Building Preparedness

As evident from the discussion above of hurricane impacts and home protection, much of the vulnerability in hurricanes comes from the integrity of the buildings wherein people shelter. Public interviewees discussed preparing homes with hurricane shutters or plywood as important, but in that context noted that actual construction makes a bigger difference, and that older homes or buildings where contractors took shortcuts are especially vulnerable. As one interviewee recalled, “that’s why a lot of people lost their homes—because they were poorly constructed. And we had a couple people killed because of their homes.” (#23) Many remark that making buildings hurricane proof (for example by developers) should be mandatory in places like Miami-Dade that face a higher threat of hurricanes making landfall.

3.7 Other parts of the forecast and warning system

In this section we describe results from those parts of the coding scheme (Figure 2) that were less frequently mentioned than impacts, mitigation, and vulnerability.

3.7.1 Storm Behavior, Tracking and Forecasting

Aspects of storm behavior mentioned by most public interviewees were high winds (100%), heavy rainfall (75%), and tornadoes (64.3%). In particular, every interviewee discussed the severity of wind speed as a strong indicator of the strength and speed at which the hurricane is moving.

Almost everyone (92.9%) mentioned storm location and landfall. Though some interviewees noted trusting forecasters and other professionals for their knowledge and experience, they do not expect the professionals to accurately track where a hurricane will make landfall. When thinking about past hurricanes, many interviewees remarked that the most memorable ones are those that either hit unexpectedly or veered off course at the last minute leaving people with supplies and storm shutters up but no damage. As noted by one respondent: “So now they you know they kind of cover themselves that way, you know. Because you know there was uh some hurricane in Broward County or something. I forget which one it was. Wilma or Katrina, one of those. And there was a lot of damage and they weren’t expecting it ‘cause they thought it was going to go somewhere. And people…and people didn’t prepare themselves. And then they got a lot of grief and there was a lot of damage.” (#19)

3.7.2. Media Coverage and Forecast and Warning Communication products

Formal watches and warnings were mentioned by 75% of public interviewees, who noted that all major news stations go to constant coverage under a hurricane watch or warning. As one explained, hurricane coverage is “Pretty much so on every channel. Yeah. If it’s—if it’s not something um horrible…. I mean if it’s not something that it’s really coming, they just um tell you like in the news. You know like 6 o’clock news, or 11 o’clock news. But if it’s something that is supposed to come for sure, and it’s something—it’s a hurricane that’s strong—then they would tell you all the time. You know, constantly.” (#12)

As illustrated by an earlier quote, many public interviewees discussed how network broadcasters, National Hurricane Center experts, and websites across the media outlets provide up to date coverage on what to expect including the size, speed, projected landfall, and approximate time. Some also noted that heeding this information as imperative: “if you ignore these warnings, you are in serious trouble.” (#22, black female, age 71, renter, Andrew)

Every person interviewed (100%) mentioned hurricane forecast dissemination in the form of TV news messages, radio, Internet, or occasionally even cell phone communications. Opinions on this widespread dissemination as hurricanes approach was mixed; some said they “get tired of hearing” (#9) the wall-to-wall coverage and would prefer to look online, while others said they are glad of the news onslaught, stating “they’re giving us the warning that we should have.” (#18)

Public interviewees also discussed their opinions about information from different sources. For example, some discussed the usefulness of Internet data and their preference for the new media: “I heard about it on the 8 o’clock news in the morning and now I’m hearing it noon. And now I’m hearing it at 6. Now I’m hearing it at 11. And three days later I’m still hearing about it. It’s like I have the Internet. If I want even more detail, I can find it.” (#9). Another explained: “Maybe NOAA is actually the most accurate because that’s what they do. Pure weather. But um I think they’re all in tune with what’s going on and they’re giving us the warning that we should have. There’s not one channel or radio station that’s better than another in terms of disseminating information to us.” (#18)

As discussed in Bostrom et al. (2016), very few of the many U.S. National Hurricane Center (NHC) and National Weather Service (NWS) forecast products available for hurricanes were mentioned by interviewees not employed by NHC or NWS. The public interviewee results show this as well (Figure 9). Of the specific National Hurricane Center products in the coding scheme, only the cone graphic was named by public interviewees, by approximately one third of them.

Figure 9.

Percentage of interviewees mentioning different forecast communication products. Red indicates fewer mentions, green more. WFO, PO, and BR reported in Bostrom et al (2016). Note that some product categories were included in the coding scheme a priori, based on the NHC and NWS data collection and analysis reported in Bostrom et al. (2016), and not all of these were mentioned by the interviewees reported here (whereof the dashed lines in red rows).

Prior research has found that the cone graphic, designed to illustrate the hurricane’s anticipated path and associated uncertainty, is widely misunderstood, misinterpreted sometimes as the extent of the storm or the area of impact (Ruginski et al., 2016, Broad et al., 2007; c.f. Savelli and Joslyn, 2013). Our data corroborate this, as illustrated by this exchange: (interviewer prompt) “Can you tell me any more specifics about the hurricane warning or forecast? “Anything specific?” “Yeah. No, I mean just that you really don’t know what it’s going to do until it actually gets there. ‘Cause they’re kind of…they’re –yeah they have that track, that cone that they follow, but I mean it could really do anything up until like it’s pretty much on top of you already.” (#17) Another interviewee offered this as the warning that first came to mind: “Well I heard it on the news. And uh they—they show like the…the map of Florida. And they show hurricane warning for this part of Florida to this part of Florida. And it’s all in red. And if the hurricane’s going to for sure come then this cone of death – they call it the cone of uncertainty, we call it the cone of death.” (#26, mixed race male, 38, homeowner, Katrina).

Most of the more specific concepts in three categories were mentioned by 25% or fewer public interviewees. One of these categories is forecast communication products: Eighty-five percent of the products were mentioned by fewer than 25% of interviewees. The exceptions include: media products like brochures or pamphlets (mentioned by 61%); watches or warnings in effects for tropical storms, winds, hurricanes, floods or tornados (mentioned by 75%); precautionary or preparedness actions (mentioned by 86%); and evacuation guidance such as signs on the road pointing out the evacuation route (mentioned by 64%) (Figure 9).

Forecast/watch/warning dissemination, and forecast decision making were also sparsely mentioned in the public interviews. Eighty-two percent of the forecast decision making codes were mentioned by fewer then a fourth of interviewees. The few notable exceptions include, for example, forecast errors, the extent of the watch or warning, and the National Hurricane Center watch or warning decision, each of which were mentioned by about half of respondents.

Forecast decision making and watch/warning dissemination are fairly technical categories that include concepts of direct concern mostly to the warning system professionals interviewed in Bostrom et al. (2016). Thus, it is possible that public interviewees do not mention them either through lack of interest, concern, or understanding. Further, these categories may simply be not as relevant to the public interviewees’ situations are they are to hurricane professionals.

Despite differing opinions on information volume and sources, public interviewees’ overall assessment of the information available is that it is useful. Survey results corroborate this (Figure 10), with all of the types of information included in the question judged as useful to extremely useful. Interestingly, those who have experienced hurricanes tend to view information provided with hurricane forecasts as more useful than do those who have not experienced hurricanes, with one of the exceptions being how to prepare and respond (Figure 10),. Notably, information about storm surge depths was deemed less useful than all other types of information suggested to survey respondents, consistent with the absence of attention to storm surge in the public interviews.

Figure 10.

Judgments of the usefulness of different forms of information from the Florida survey sample, subdivided by whether or not the respondent says they have personally been affected by a hurricane. *p<0.05 for difference by hurricane experience.

3.8 Awareness and omissions

There is widespread awareness among the public interviewees of five key concepts related to hurricanes. In addition to all interviewees noting obtaining their information from local and national media, especially TV and radio but including social media, all mentioned wind speed and intensity, death, property damage, and previous hurricane experience.

A total of 22 different concepts (as coded) were mentioned by at least 75% of public interviewees. Most of the concepts mentioned fall under the categories of storm impacts (32%), vulnerability (18%), and mitigation (23%). Two general categories, storm tracking and forecasting and forecast dissemination, include only one specific code each that is mentioned by over 75% of interviewees. However, those specifics rank highly: storm location and landfall is mentioned by 92.9% of interviewees and all (100%) of interviewees reference local and national media coverage.

Of these 22 codes, four concepts are important to the public interviewees, but not to the professionals discussed in Bostrom et al. (2016) (as defined by at least 75% mentioning), in the categories of storm impacts, forecast communication products, vulnerability, and mitigation. These findings indicate that Miami-Dade residents’ hurricane considerations cut across the hurricane forecast and warning process, emphasizing having a hurricane plan in place, securing loose property, and watch/warning products aired through the media, as well as social and economic impacts (often looting).

Almost as many codes are salient to professional interviewees (mentioned by more than 75%) and not to the general public. These range across most broad categories: storm development, storm behavior, storm impacts, storm tracking and forecasting, watch/warning decision factors, forecast dissemination, vulnerability, and mitigation. Three codes in the categories of storm behavior and storm impacts codes are rarely mentioned at all by public interviewees (25% or less mention): storm surge, flooding from storm surge, and increasing wind speed. All professionals (100%) mention storm surge, but only 14.3% of the public (4 out of 28 people) mention it. Even more telling, 89% of professionals (all but one) discuss flooding from storm surge while only 7.1% of the public (2 out of 28 people) mention storm surge flooding. Professionals agree that storm surge and the resulting flooding are very real hazards facing the public, but in these interviews and survey results there is a clear disconnect between professional knowledge and public interest or attention on this point. The third term—increasing wind speed—is mentioned by 78% of experts and only 25% of the public. However, 100% of the public and professional interviewees discuss wind speed and intensity.

3.9 Recommendations from interviewees

Many of the public interviewees volunteered direct recommendations to improve the forecast and warning system. Recommendations ranged from comments on building standards and codes, to governmental preparedness, what utilities should do, and recommendations for education and media coverage. One interviewee thought developers should bear the responsibility and cost of preparing for hurricanes: “You waste a lot of money – waste a lot of money having to do all around preparing for storms […] Some of those costs should be incurred by the people who build the buildings, from the start. Not after the fact.” (#11) Another suggested support in advance of hurricanes for those who cannot afford preparations: “They can set up … a plan for to provide for those people that don’t have the cash to go out in an emergency crisis and just go ahead and start buying supplies… not after the hurricane—before the hurricane.” (#18)

Multiple public interviewees recommend less media hype, especially for weaker storms. For example, one notes that there should be “Knowledgeable, insightful media coverage, but less 24-hour coverage for a tropical depression would be great so that people take it seriously when it’s an actual risk.”(#8, white female, age 40, homeowner, Katrina) This sentiment is captured by another interviewee who characterizes the hurricane broadcast cycle this way: “The over-concentration by the news media sort of just makes you tune out a lot. At least it does me. … You just get tired of hearing it. I mean it’s the same thing over and over and over. You know, you hear it weeks at a time.” “Even when they come out with the initial predictions, way before hurricane season, you’ll hear nothing but that for a solid week… it’s just a prediction.” (#9)

At the very end of the questionnaire public interviewees filled out, they were asked to provide any thoughts they might have about the National Hurricane Center issuing a separate storm surge warning in addition to hurricane wind speed warnings. Public interviewees were of two minds about this, with some favoring this idea, others not, as these two comments illustrate: “Hurricane center has always done a good job. A separate storm warning is a good info to know.” (#1, white Hispanic male, age 59, homeowner, Andrew, Katrina, Wilma). “ I think it gets confusing when they give multiple warnings. They need to stick with warning and people will take them more serious.” (#2) Others commented that storm surge warnings would be a “Good thing for people near the coast.” (#13, white male, age 72, homeowner, Wilma)

4. Conclusions and Discussion

This study aimed to address three questions, beginning with how members of the public perceive hurricane risks, impacts, forecast and warning information, and risk management decisions. The interview and survey findings reported here paint a rich picture of a Miami hurricane culture, in which there is widespread experience, knowledge, and interest in hurricanes, from storm development to impacts and mitigation.

Public interviewees emphasize hurricane impacts, vulnerability and mitigation, much as one would hope and expect them to do, in line with risk behavior theories. A few somewhat surprising disconnects emerge. First, the importance placed by our public interviewees on social and economic impacts, watch and warning products, having a hurricane plan, and securing loose objects was not reflected to the same degree in the interviews with hurricane professionals reported in Bostrom et al. (2016). Our interviewees are concerned about vulnerabilities, aware of inequities, and attentive to a wide range of hurricane impacts, from anxious animals to post-hurricane recovery risks.

Second, our interviewees express little knowledge of storm surge and the resulting flooding, nor of increasing wind speed, though these are hazards of great importance to professionals, especially the dangers of storm surge. As noted earlier, this may be because in Miami, due to the coastal configuration, storm surge is not as major an issue as it is in other areas. However, this does not mean that it is not an issue; Hurricane Irma caused dangerous storm surge flooding in downtown Miami on September 10, 2017. One can speculate that some of the professionals may have been thinking of a broader geographic area than just Miami-Dade, or of the increasing risks of storm surge with sea-level rise and climate change (Gardner et al., 2017; Rahmstorf, 2017), and thus would have discussed it more. A reasonable conclusion might be that these findings are because the interviews were conducted a few years ago, and that storm surge in Sandy and other more recent events should have raised awareness of such risks. However, recent research conducted after Sandy concludes also that coastal community residents seem to be underestimating the likely future risks from storm surge exacerbated by climate change and sea-level rise (Morss et al., this issue; Wong-Parodi et al., 2017).

Further, lack of water and damage to transportation are potentially important, because this is another aspect of the aftermath that people need to be prepared for (like loss of power). However, these hurricane impacts were mentioned by fewer public interviewees than public officials or broadcasters.

Our second question concerns how people perceive and obtain forecast and warning information for hurricanes, and what decision processes and mental models underlie public behaviors with respect to hurricane warnings.

Consistent with a “spiral of reinforcement” (Slater, 2015; Zhao, 2009), those with hurricane experience rate most types of information about hurricanes as more useful than those who have not experienced a hurricane. Twenty-four-seven (all day every day) coverage of hurricanes by the broadcast media was a source of annoyance to many public interviewees, who in some cases recommended reduced coverage. This annoyance nevertheless illustrates the reach of those communications. Interviewees commented favorably on being able to choose their level of information on the Internet, and on the local availability of information on how to prepare, for example brochures in grocery stores. Despite some respondents recommending less media hype, a large appetite for forecast and warning information is apparent, with discrimination between the National Hurricane Center as a more knowledgeable and reliable source compared to other sources.

Also evident is the persistence of misunderstandings about existing hurricane communication products, such as the cone graphic, and absence of evidence of attention to many other hurricane communication products. The comments that public interviewees made regarding the representation and interpretation of uncertainty in such products suggest a continuing need for further exploration of this topic, to inform improvements in communications products and processes.

Our third question regards the implications of these findings for improving hurricane forecasts and warnings. While appreciation of hurricane impacts, vulnerabilities, and what to do about hurricanes is evident in our results, further research might examine the effects of emphasizing one of these elements over the others, of addressing the disconnects between how different groups understand each of them, and of evaluating and better coordinating how they are communicated across the forecast and warning system. This may require thinking about forecast and warnings communications from a systems perspective, to take into account the interrelations of information content and format, the many organizations and individuals involved, and then expanding diversity of communications channels and social contexts (NASEM, 2017a).

Many hurricane information products lack salience in the forecast and warning system. While these are arguably critical products according to some professionals, the lack of mention by public interviewees (and in many cases, non-NWS professionals; Bostrom et al., 2016) suggests that they do not figure largely in thinking about hurricane forecasts and warning. This interpretation is consistent with other findings, such as the relative neglect of storm surge in the interviews. Two recent changes in National Weather Service policies and practices may be closing gaps such as those identified in this research: new storm surge communications products (Morrow et al., 2015), and a shift toward impact-based decision support (Uccelini, 2016). Findings herein on how people think about storm surge risks and about impacts should inform these efforts.

Designing information that better addresses how people currently understand storm behavior and uncertainties should make that information more comprehensible and useful. This would likely require weather forecasting, broadcaster and public official professionals to better understand public concerns about vulnerabilities and beliefs about mitigation, in order to improve their forecast, warning, and information design and dissemination decisions. In this regard, the results presented here should be useful to all parties in the hurricane forecast and warning system, and to researchers who strive to model the forecast and warning system, for example using agent based modeling (Morss et al., 2017).

Appendix 1. Details of interview coding scheme development, analysis, and interpretation

Coding scheme development and coding process

Broadly, the coding process went through the following steps: A coding scheme was initially developed based on the forecaster group interview and applied to subsequent individual interviews. Two to three coders independently coded the first interview of each class of professionals (NHC, WFO, Public Officials, and Broadcasters; for additional details see Bostrom et al., 2016). Based on discussion with other coders, new concepts could be accommodated by adding or amending codes in the scheme. These initial coders then compared their coded interviews to generate coding agreement statistics and discuss any coding reliability issues that may have arisen. After this discussion, each coder independently re-coded the first interview in each class of professionals. After satisfactory inter-coder reliability was achieved for the broad question types and full interviews for the professional interviews, the first public interview was coded independently by the two coders. Reliability was satisfactory, as detailed in the following, so the rest of the public interviews were coded by a single coder.

Standardized guidance regarding how to apply codes from the scheme to corresponding interviewee responses was applied for coding individual interviews. The general structure of the coding scheme is shown in Figure 2. While coding, coders recorded notable comments that did not fit into the coding scheme.

Interview data analysis and interpretation

After completing all coding, we calculated inter-coder reliability. A binary coding method was used to format the data in preparation for inter-coder reliability calculations. Terms from the coding scheme were coded “1” if they were mentioned and “0” if they were not. Because each response could in theory be coded as reflecting anywhere from 0 to 189 different codes, we calculated reliability within each broad question type, across the entire coding scheme. For example, if both coders coded a response as code 11100 Intensity (wind speed) (from the section of codes for Storm Behavior), that would be coded as an agreement (a match). Absence of code 11130 Winds greater than 64 KTS/74 MPH -hurricane designation (also from Storm Behavior) in both coded transcripts would also be coded as agreement. Reliability was calculated using Freelon’s (2010) inter-coder reliability web service, ReCal. The output from these calculations was reported as average pairwise percent agreement between two coders and Cohen’s Kappa, which is an agreement statistic that corrects for the likelihood of agreeing by chance. Percent agreement between two coders was calculated as the number of concepts coded the same way (present or absent) by both coders, divided by the total number of terms in the coding scheme. Cohen’s Kappa was calculated by cross-tabulating frequency counts for all coding category agreement (e.g., both coding “1”), determining marginal proportions, then calculating an observed and expected agreement to achieve the ending output.¹ For the public interviews, overall agreement on the sample interview that was coded by two coders independently was 86%, with a Cohen’s Kappa of 0.66. This is a modestly acceptable reliability, in light of the complexity of the coding scheme, but not strong, which is another motivation to increase transparency by reporting direct quotes from the interviews.

Descriptive and comparative statistics were calculated for all codes for each respondent group. We compared the relative frequencies of concepts mentioned, both by respondent group, and across concept (i.e., code) categories, to examine patterns of response and learn how people thought about different elements of the forecast and warning system, as well as commonalities and differences across groups.

Appendix 2

Full heatmap for public and professional interviews (WFO=Weather Forecasting Office, PO=Public Official/Emergency Manager, BR=Broadcaster). See Bostrom et al. (2016) for details on the interviews of professionals. Numbers in cells represent the percentage of respondents in that category who mentioned that concept in their interview.