A survey-based assessment of perceived flood risk in urban areas of the United States

A survey-based assessment of perceived flood risk in urban areas of the United States

Journal Pre-proof A Survey-Based Assessment of Perceived Flood Risk in Urban Areas of the United States Sharon L. Harlan, Mariana J. Sarango, Elizabet...

4MB Sizes 1 Downloads 82 Views

Journal Pre-proof A Survey-Based Assessment of Perceived Flood Risk in Urban Areas of the United States Sharon L. Harlan, Mariana J. Sarango, Elizabeth A. Mack, Timothy A. Stephens

PII:

S2213-3054(19)30028-1

DOI:

https://doi.org/10.1016/j.ancene.2019.100217

Article Number:

100217

Reference:

ANCENE 100217

To appear in: Received Date:

1 February 2019

Revised Date:

4 August 2019

Accepted Date:

5 August 2019

Please cite this article as: Harlan SL, Sarango MJ, Mack EA, Stephens TA, A Survey-Based Assessment of Perceived Flood Risk in Urban Areas of the United States, Anthropocene (2019), doi: https://doi.org/10.1016/j.ancene.2019.100217

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier.

A Survey-Based Assessment of Perceived Flood Risk in Urban Areas of the United States Sharon L. Harlan 1, Mariana J. Sarango 2, Elizabeth A. Mack 3, Timothy A. Stephens 4 1

Department of Health Sciences and Department of Sociology & Anthropology, Northeastern University, 360 Huntington Avenue, Boston, MA 02115; [email protected] 2

Department of Health Sciences, Northeastern University, Boston, MA 02115; [email protected] 3

of

Department of Geography, Environment, and Spatial Sciences, Michigan State University, East Lansing, MI 48824; [email protected] 4

ro

Institute for Resilient Infrastructure Systems, University of Georgia, Athens, GA 30602; [email protected]

-p

Highlights (3-5 bullet points, 85 characters each) ● In a sample of US urban residents, 27.5% had experienced a major flood in their lifetimes. Socially vulnerable people were more likely to perceive higher risks of future floods.



Potential exposure to flood hazard increased the odds of perceiving higher flood risk.



Own flood experiences and exposure to flood news increased perceived flood risk.



Experiencing > 1 major flood doubled the odds of perceiving higher flood risk.

lP

re



Jo

ur na

Abstract How people perceive the risks of climatic hazards is currently a major research thrust in the field of risk perception. In the wake of recent flood disasters in all regions of the United States and globally, more researchers are investigating social vulnerabilities as well as the role of cognition in explaining risk perceptions. This study analyzed how people in the United States perceive the risk (i.e., likelihood and seriousness) of flooding via a layered analysis that considered several plausible and intertwined lines of inquiry from the risk perception literature. We surveyed 9,250 individuals within nine major urban areas, including the largest city and one smaller city in each region. The National Flood Hazard Layer product provided data for deriving their potential exposure to flood hazards. The analyses tested and confirmed several hypotheses drawn from Social Vulnerability Theory and from Protective Motivation Theory: characteristics associated with social vulnerability (older, female, race/ethnic minorities, low income), previous experiences with and awareness of flood news, and potential exposure to flood hazard (local fraction of flood prone area) significantly increased risk perceptions of floods. Selfconfidence in ability to cope with a future flood disaster lowered risk perceptions. This study is the first snapshot of US flood risk perceptions nationwide. It points to needs for more theoretically-driven research about flood risk perceptions and behaviors, flood risk communication within local communities, and more social and economic support for vulnerable populations.

Keywords 1

flood hazard; flood risk management; flood risk perception; protective motivation theory; social vulnerability theory; survey research Introduction In the United States (US), approximately 3% of the population lives in 100-year coastal floodplains (1% chance of flooding annually) (Crowell et al., 2010). Fifty-two percent of the population lives in counties that intersect coastal watersheds and 39% lives in counties along the shorelines of oceans, major

of

estuaries, or the Great Lakes, where sea-level rise and intensifying storms are most directly increasing flood risks (NOAA, 2013). Almost one million people in the US are exposed annually to inland flooding

ro

from rivers and streams (Zbigniew et al., 2013). Increasing urbanization and its impacts on the landscape

-p

and atmosphere further amplify flood hazards by increasing peak runoff and flood volumes, reducing the time to flood peaks, and altering the magnitude and frequency of storm events (Leopold, 1968;

re

National Research Council, 2012; Poff et al., 2006; Shepherd, 2013; Welty, 2009). Floods are not distributed equally but are disproportionately visited on older, non-white, and poorer residents and

lP

communities, as well as women and people in poor health (Cutter et al., 2014; Debbage, 2018).

ur na

Flood management is high on the agendas of many US municipal governments because of the disruption, hardship, and expense involved in the aftermath of floods. Indeed, flood policy is transitioning from flood defense, or keeping water out, to flood risk management, or learning to live with floods and adapt to their consequences (Butler and Pidgeon, 2011; Scott, 2013). Some coastal

Jo

cities, such as New York (OneNYC, 2018), Miami (Miami-Dade Green, 2018), and Boston (City of Boston, 2018) are proactively developing technical and engineering initiatives led by academics and planning professionals to deal with sea level rise. Yet, with each new flood disaster, it is increasingly clear that engineering alone cannot completely protect people from floods (O’Hare and White, 2018). Managers believe that effective risk management balances scientific and technical knowledge with an understanding of the public’s attitudes toward and beliefs about flood hazards in order to develop 2

workable strategies to avoid and recover from future floods, as well as ways to minimize the short- to medium- term impacts of flood disasters (McDaniels et al., 1997).

Prior research on the human dimensions of floods has created risk profiles for spatially and socially vulnerable populations (Maantay and Maroko, 2009; Maldonado et al. 2016), examined the health impacts (Alderman et al., 2012; Fernandez et al., 2015; Pietrzak et al., 2012; Renner, 2017) and

of

assessed the economic burdens (Changnon, 2008; Smith and Katz, 2013) of floods. Studies of how

ro

people perceive the risks of climate hazards are currently a major thrust (O’Connor et al., 1999; Renn, 2011; Wachinger et al., 2013; Valkengoed and Steg, 2019) in the decades-old field of risk perception

-p

studies in the social sciences. Although debate is ongoing about whether and how risk perceptions predict individual adaptive behaviors aimed at reducing flood risks, a preponderance of social scientific

re

evidence supports a positive relationship between perception and people’s intentions to adapt, if not

lP

actual behaviors (Valkengoed and Steg, 2019). Understanding how environmental risk perceptions are formed may lead to new hypotheses about the interface of risk perceptions with adaptions. Understanding whether socially vulnerable groups, who are at the highest risk of harms, perceive their

ur na

risks differently than more privileged groups is especially important.

Social scientists conceptualize risk perception as “the process of collecting, selecting, and interpreting

Jo

signals about uncertain impacts of events, activities, or technologies” (Wachinger et al. 2013, 1049). Broad consensus exists that people incorporate multiple sources of information, experiences, and subjective intuition into formulating mental models of perceived risk. In the wake of recent major flood disasters in all regions of the US and globally (e.g., Chan, 2015; Dewan, 2015) and projections of increasing global flood risks (Hirabayashi et al. 2013; Ward et al., 2014), more researchers are investigating social vulnerabilities (Collins et al., 2018; Rufat et al., 2015; Zahran et al., 2008) as well as

3

the role of cognition in explaining risk perceptions and motivation for adaptation to extreme events, such as floods (Grothmann and Patt, 2005).

This study sought to understand how a broad sample of the US urban population perceives their own individual risk to major floods (i.e., the likelihood and seriousness of floods). The study utilized a layered analysis that considered several plausible and intertwined lines of inquiry from the literature on

of

risk perception. This first study to use national US survey data answers the following research

ro

questions:

-p

1) Are socially vulnerable individuals more likely than others to perceive higher risks from floods? 2) What is the role of cognition, as manifested in individuals’ experience of past floods, flood news

re

awareness, or potential exposure to regulatory flood hazard areas, in predicting perceived risk?

lP

3) How does an individual’s appraisal of their private adaptive resources and confidence in the problem-solving abilities of their communities and social institutions affect perceived risk from

ur na

floods?

1.0 Material and Methods

This study utilized the Survey of Water Indicators and Socioeconomic Status of Households (SWISSH),

Jo

developed previously by the authors, as part of a federally funded research network. The network, with more than 20 university partners in the US, aims to find technological and behavioral solutions that promote sustainable urban water systems. The purpose of the SWISSH was to generate data for addressing questions about people’s perceptions and behaviors relating to a broad range of water pressures in a sample of US urban households. As such, the SWISSH is a cross-sectional, online survey of the adult population in 9 US urban areas, which asked respondents questions about a variety of waterrelated issues. Questions concerned the cost of water bills, sustainability of water systems, 4

contamination of drinking water, polluted waterways, water recreation, and perceived risks of water hazards, such as floods and drought, in different local socioeconomic and cultural contexts. The survey also included questions related to sociodemographic characteristics, social capital, and health. The Data Statement (Harlan et al., 2019) provides greater detail about the survey methodology.

1.1.1 Sample Design

of

Working closely with consultants at Qualtrics, Provo, Utah (2019), a nationally recognized survey

ro

research firm that conducts online surveys for businesses and universities, we designed a non-

proportional quota sample of residents in nine major urban areas of the US where scientists in our

-p

network are conducting research (hereafter referred to as “study regions”). We selected the largest city and one smaller city in each study region. These study regions (and cities) are: Eastern

re

Massachusetts (Boston - Worcester); Front Range-Colorado (Denver – Fort Collins); Mid-Atlantic

lP

(Washington, DC – Baltimore, Maryland); Pacific Northwest-Oregon (Portland – Eugene); Piedmont Atlantic (Atlanta, Georgia – Charlotte, North Carolina); Southeastern Florida (Miami – Palm BayMelbourne); Southeastern Michigan (Detroit – Flint); Southern California (Los Angeles – San

ur na

Bernardino); Sun Corridor-Arizona (Phoenix – Tucson).

We used ZIP Code Tabulation Areas (ZCTAs), which are geographical areal units created by the US Census Bureau for the purpose of tabulating population statistics (US Census, 2018), to define a

Jo

geographical boundaries for each study region in the SWISSH sampling frame. ZCTAs are proxies for ZIP code areas, which the US Postal Service uses as a tool for mail delivery. The Data Statement (Harlan et al., 2019) explains the four criteria that a ZCTA needed to meet in order to be included in our sampling frame. Survey respondents for the SWISSH reside in 1,752 ZCTAs.

Qualtrics sent email messages to potential respondents in each study region through their Online Sample platform. The invitation asked recipients to participate in the survey and provided a link to 5

eligibility/screening questions. Qualtrics managed administration of the survey, ensuring exclusion of duplicate responses. We utilized non-proportional quotas (i.e. a minimum of 100 of each race/ethnicity and income group) within each study region to ensure adequate numerical representation of typically underrepresented racial/ethnic and income groups and the statistical power needed for multivariate analyses. We administered the survey from December 2017 through March 2018 in three waves. We inserted pauses during the survey period (i.e., between waves) to ensure that data were recording

ro

of

properly and that responses were received from as many ZIP code areas as possible.

Participants completed the survey online using a computer or a mobile device. Of 24,527 contacts who

-p

answered a screener question, 49% (n=12,024) were eligible to participate because their ZIP code area was included in the SWISSH sampling frame. After checking eligible surveys for completeness, 77%

re

(n=9,250, across 1,752 ZCTAs) of the respondents were retained for this analysis For details, see Data

lP

Statement (Harlan et al., 2019). We applied sample weights to adjust for the non-proportional quota sampling strategy (Kalton and Flores-Cervantes, 2003). First, a sample weight was calculated for each respondent’s 1) race/ethnicity and 2) income group. These weights reflect the probability of selection

ur na

into the sample based on our selected quota limits (i.e. minimum of 100) and the population size of the respondent’s racial/ethnic or income group in that study region, respectively. Population size estimates (by race/ethnicity and income group) were obtained from the US Census 2011-2015 American

Jo

Community Survey. Second, the iterative process of rim weighting, or raking (Rose, 2000), was used to combine race/ethnicity and income weights into one final probability weight per respondent.

1.1.2 Measurement of the Dependent Variable: Risk Perception Perceived risk of an environmental hazard is often expressed as a combined measure of perceived likelihood (probability) that an event will occur and perceived severity (seriousness) of the consequences of such an event (Bubeck et al., 2012; Grothmann and Patt, 2005). We measured 6

perceived risk of flood as an ordinal variable that assigned each respondent an integer score between 0 and 7. Respondents were first asked, “How likely or unlikely do you think it is that a flood will affect your residential area within the next 10 years?” (0=very unlikely, 1=somewhat unlikely, 2=neither likely nor unlikely, 3=somewhat likely, 4=very likely). Respondents were then asked, “If a flood occurred, do you think the consequences would be serious or not serious for you and your household?” (0=not at all serious, 1=a little serious, 2=somewhat serious, 3=very serious). Scores for the risk perception variable

of

were the sums of answers to the question about perceived likelihood of flooding in a respondent’s

ro

residential area and the question about a respondent’s appraisal of how serious the consequences of

-p

such a flood would be. The lowest perceived risk is 0 and the highest is 7.

1.1.3 Social Vulnerability Predictor Variables

re

The study determined age (in years) based on year of birth. Gender was designated by participants as

lP

male, female, or non-binary/gender non-conforming; 0.23% (n=22) of the sample selected the latter response and were excluded from this analysis because cases were too few to include in multiple regression models. We constructed race/ethnicity as a 5-level categorical variable (Hispanic, Non-

ur na

Hispanic White, Non-Hispanic Black, Non-Hispanic Asian, Non-Hispanic Other Race or Mixed), from responses to two questions, 1) “Do you consider yourself to be Hispanic, Latino/a, Mexican, MexicanAmerican or of Spanish background?” (yes, no) and 2) “With which racial or ethnic group(s) do you

Jo

identify yourself?” (Check all that apply: African-American or Black, Asian or Asian-American, MiddleEastern, Native American or American Indian, Native Hawaiian or Pacific Islander, White, Other).

We measured educational attainment using a five-level ordinal variable: did not complete high school, completed high school, completed community college or vocational/technical school, completed fouryear college (bachelors degree), completed graduate or professional school. We measured household income using the question, “What was the total combined income before taxes of everyone in your 7

household in 2017?” Participants selected less than $50,000, $50,000 - $100,000, or more than $100,000 a year. We measured home ownership status using a dichotomous variable indicating whether the respondent reported owning their current home versus renting or other living arrangements. We created a dichotomous variable to represent access to basic needs. This variable indicated whether the participant has access to five basic needs: 1) participant had enough money in past year to pay for housing (yes, no); 2) participant had enough money in past year to pay for

of

gas/electricity bill (yes, no); 3) participant had enough money in past year to pay for health care and

ro

medications (yes, no); 4) participant’s household had enough of the kinds of food they wanted (yes, no); and 5) participant had health insurance (yes, no). Respondents who answered affirmatively to all

re

1.1.4 Variables Representing Flood Experience

-p

questions were counted as having access to basic needs.

lP

We measured direct experience with past flood events using a three-level ordinal variable. Respondents were asked, “In your lifetime, have you ever experienced a flood where water entered the property or living space of your home?” (no, one time, more than one time). Respondents who had

ur na

experienced a flood at least once in their lifetimes were asked the following four follow-up questions: 1) “Did the flood cause property damage to your home?” (lost everything or major property damage, minor property damage, no property damage); 2) “Did the flood cause anyone in your household to

Jo

become injured or physically ill?” (yes, no); 3) “Did the flood cause anyone in your household to become emotionally or mentally ill?” (yes, no); and 4) “Did a major flood occur within the past year?” (yes, no). We measured indirect experience of flood events using a four-level ordinal variable based on responses to the item, “Which answer comes closest to how often you have heard or seen news reports about floods in the past four months?”(almost daily or several times a week, several times a month, once a month, never).

8

1.1.5 Appraisal of Adaptation Resources The study measured individuals’ appraisals of their available resources to adapt to or cope with the consequences of serious floods at the institutional, community, and household levels. The institutional confidence scale used responses to five questions that asked how confident the participant is in the following institutions and their respective leaders: local water utility, local police, city/town government, local Drainage and Flood Control District, and the Federal Emergency Management

of

Agency. For each item, respondents answered 1=not at all confident, 2=hardly any confidence,

ro

3=neutral, 4=somewhat confident, or 5=very confident. For each respondent, the average of these five response values yielded a continuous institutional confidence score, ranging from 1 (low institutional

-p

confidence) to 5 (high institutional confidence). Perceived likelihood of community problem-solving was measured as perceived likelihood that residents living near respondent would cooperate to try and

re

solve the problems associated with floods, using a five-level ordinal variable (very likely, somewhat

lP

likely, neither likely nor unlikely, somewhat unlikely, very unlikely). Confidence in personal ability to cope with the consequences of a flood event, was a 4-level ordinal variable (very confident, somewhat

ur na

confident, neutral, hardly any/no confidence).

1.2 Potential Exposure to Flood Hazard

The study quantified the potential exposure to flood hazards within a ZCTA to evaluate its association

Jo

with a respondent’s perception of flood risk. Ideally, a respondent’s potential exposure to flood hazards would measure the location of their residential street address relative to flood-prone areas. Qualtrics Panel rules, however, do not permit collecting the exact location of respondents’ residences.

The study utilized the Federal Emergency Management Agency’s (FEMA) regulatory 100-year floodplains that define flood prone areas. We obtained this information through FEMA’s National Flood Hazard Layer product, a geospatial database containing current regulatory flood hazard boundaries 9

(FEMA, 2019). In this product, hydrologic and hydraulic models estimate flood hazards, as represented in the area inundated by the 100-year flood event, including coastal and riverine flood events. For details, see Data Statement (Harlan et al., 2019). We used dasymetric mapping techniques to estimate exposed populations (Debbage, 2018; Ferguson and Ashley, 2017; Maantay and Maroko, 2009). Dasymetric mapping techniques disaggregate datasets, such as ZCTA level demographics, to a finer resolution based on ancillary categorical data (Mennis, 2003). We applied an urban filtering dasymetric

of

mapping approach using the 2011 National Land Cover Dataset classes 21 -24 (Homer et al, 2015) and

ro

the 100-year regulatory floodplains to estimate areas with potentially exposed people as the

intersection of flood hazard areas and developed land. The fraction of developed land in the ZIP code

-p

area delineated as a special flood hazard area (SFHA) served as a proxy for respondents’ potential

re

exposure to flood hazard (continuous variable ranging from 0-1).

lP

1.3 Regional Variability in Selected Flood-Related Variables

The nine study regions in this study exhibit variation in important flood-related characteristics, such as perceived flood risk, potential exposure to flood hazard, and the local population’s experience with

ur na

recent flooding. The maps in Figure 1 illustrate the location of the study regions (panel a) and differences in these flood-related characteristics (panels b, c, and d). Regional distributions in risk perception (panel b) varied in their moments. A one-way ANOVA comparing mean risk perception

Jo

scores by region was statistically significant (F (8, 9179) = 43.69, p < 0.001). Southeastern Florida had the highest proportion of respondents perceiving higher flood risk, followed by Southeastern Michigan.

Regional variability in potential flood hazard exposure (panel c) was generally minimal. Southeastern Florida was a noticeable outlier, however, where on average, 28.5% of a ZCTA contained potentially exposed people. Low relief topography and the nature of coastal flood hazards in Southeastern Florida resulted in a larger aerial coverage of regulatory flood hazard areas relative to the inland riverine flood 10

hazard areas of the other sampled regions. Southeastern Florida contained the highest proportion of respondents experiencing a flood within the last year (9%), followed by Southeastern Michigan (7%) and Eastern Massachusetts (6%) (panel d). Observations of increasing flood frequency in Southeastern Michigan (Mallakpour and Villarini, 2015) support the relatively higher proportion of flood experience in

Jo

ur na

lP

re

-p

ro

of

this region.

Figure 1. Regional summaries of selected flood-related variables for SWISSH respondents: (a) sampled study regions in the US, (b) distribution of perceived flood risk scores, (c) average percent of developed land classified as special flood hazard area (SFHA) averaged over all ZCTAs represented in the sample for each region; and (d) proportion of respondents reporting a local flood within the last year.

1.4 Statistical Analysis

11

We used a discrete choice modeling framework to estimate the likelihood that study respondents perceive themselves as at risk for flooding. Specifically, a proportional odds logit model (Harrell, 2015; McCullagh, 1980) was estimated to account for the extra information about the order of the data in the dependent variable (perceived risk of flood values range 0-7). We selected this type of model instead of a multinomial logit model to account for the ordered nature of our dependent variable. Our models were multilevel in that they included variables collected at the household level and one variable

of

collected at the ZCTA level (potential exposure to flood hazard). We chose to specify ZCTA as a fixed

ro

effect given that we had nearly 90% of the ZCTAs that comprise our sampling framework represented in our sample and, therefore, accounted for most of the variation between ZCTAs. We estimated the

-p

models using the software Stata versions 13 and 15. Prior to running the models, we used the svyset command to specify the appropriate probability weights for the models and the strata for the models.

re

Models were estimated with the combined race/ethnicity weights derived from rim weighting and the

lP

study regions specified as strata. Model results were estimated with Taylor linearized standard errors which are appropriate for survey data (Demnati and Rao, 2004). The specification of the proportional

ur na

odd logit model is as follows:

𝑃𝑟[𝑦 ≥ 𝑗|𝑋] =

1 1 + 𝑒𝑥𝑝[𝑎(𝛼 + 𝑋𝛽)]

Jo

where j=0,1,2, 3,4, 5, 6, or 7

X is a matrix of covariates or explanatory variables 𝛽is a matrix of coefficient estimates for each of the explanatory variables

The estimated parameters of the models are presented in Tables 2 and 3 as odds ratios. Odds ratios measure the strength of association between each predictor variable and an outcome as the estimated 12

increase (or decrease) in the log odds of the outcome per unit increase in the predictor. Interpreting an odds ratio requires subtracting “1” from the estimate of the odds ratio. For odds ratios less than one, this procedure yields a negative number and for odds ratios greater than one, the number is greater than zero. These numbers then provide interpretations as percentages. For example, a value of 1.50 means that a given variable increases the risk perceptions of survey respondents by 50% as compared to the reference group. A value of 0.83 means that a variable decreases the risk perception of survey

of

respondents by 17% as compared to the reference group. A confidence interval around the odds ratio is

ro

calculated using the standard error of the estimate (e.g., 0.86 (OR) ± 0.07 (SE) = CI, 0.79, 0.93). Model results were estimated with standard errors robust to heteroskedasticity and weighted by sample

-p

probability weights.

re

2.0 Theory

lP

We tested hypotheses drawn from Social Vulnerability Theory (SVT), (Emrich and Cutter, 2011; Flanagan et al., 2011) and from Protective Motivation Theory (PMT) (Lindell and Perry, 2004; Rogers, 1975). SVT centers individual and community social characteristics as highly salient in determining people’s actual

ur na

risks of experiencing environmental hazards and recovering from the impacts of extreme events (Wisner et al., 2004). Studies also show that social and cultural frames of reference are defined by the social groups to which people belong. These frames influence individual attitudes, values, and interpretations

Jo

of information about their surroundings, including perceived risks of environmental threats (Maibach et al., 2009; Ogbu, 1993).

PMT examines the role of cognition in shaping risk perceptions and decision-making about whether people will act to reduce their risks from environmental threats (Bubeck et al., 2012; Grothmann and Patt 2005; Grothmann and Reusswig 2006; Horney et al., 2010). According to Grothmann and Patt (2005, 205), “the objective ability or capacity of a human actor only partly determines whether that 13

actor will take an adaptive response. . . . Risk perception and perceived adaptive capacity are factors we believe to be important, but which most adaptation models in the climate change literature have so far omitted.” Some studies have uncovered a “risk perception paradox:” high risk perception can either result in action or, under other circumstances, it can motivate people to avoid the problem or to willfully ignore it (Grothmann and Reusswig 2006; Wachinger et al. 2013). Thus, understanding the cognitive

of

factors that influence perceptions of risk is important.

ro

We used standard measures of social vulnerability to establish a baseline of differences in risk

perceptions among social groups. Then, we adopted the PMT framework, which centers on human

-p

cognition to test whether subjective appraisals of previous and perceived flood exposure add explanatory power to predictive models of risk perception that use conventional sociodemographic

re

variables. We extend the modeling to include measures of resource appraisal for adaptation to floods,

available resources.

ur na

2.1 Social Vulnerability

lP

which allow us to assess individual-level heterogeneity in the relationship between perceived risk and

Hypothesis 1: Socially vulnerable individuals are more likely than others to perceive higher risks of future

Jo

floods.

Socioeconomic status predicts risk perceptions because people take cues from their membership in social groups that denotes their places in social hierarchies. Sociodemographic variables are markers of social, economic, and institutional resource privileges and constraints with respect to risk from environmental hazards. Research shows that more vulnerable social groups – oppressed racial or cultural minorities and white women – are more fearful of environmental risks than white men (Bord et

14

al. 1998; Kellstedt et al. 2008; O’Connor et al. 1999), and that vulnerable groups are more likely to believe that the local environment is changing for the worse (Ruddell et al. 2012).

2.2 Past Experiences with Floods Hypothesis 2: Individuals who have been exposed to floods through direct or indirect experiences will

of

perceive higher risks of future floods.

ro

Individuals’ subjective judgments about personal risk are based on past experiences that invoke

memories of strong emotions (Slovic 1987). Consistently, studies find that experiences are related to risk

-p

perceptions and that memory plays a key role in experiential heuristics. People conceptualize environmental threats in terms of local contexts centered on place and time - that is, perceptions of risk

re

are highest when the likelihood of an event is both close by and poses immediate dangers (Aitken et al.

lP

1989; Dunlap and Catton 1979; Uzzell, 2000). Experiences help people envision more accurately what can happen in a future that holds many uncertainties (Bubeck et al. 2012; Grothmann and Patt 2005). Experience with past floods reduces uncertainty about the possibility of a flood occurring because the

ur na

person has already experienced it, making the threat less deniable.

2.3 Potential Exposure to Flood Hazard

Jo

Hypothesis 3: An individual’s potential exposure to flood hazards, defined by the percent of developed land delineated as a National Flood Insurance Program (NFIP) regulatory 100-year floodplain in a respondent’s ZIP code, will not affect the respondent’s perceived risks of future floods.

The NFIP defines the regulatory floodplain (i.e. flood hazard area) as the area inundated by the 100-year flood (the flood with a 1% chance of occurring in any given year). Investments in technical expertise for modeling and mapping the spatial distributions of regulatory floodplains have given risk managers 15

greater knowledge about which places and populations are exposed to flood hazards, as well as a tool to communicate risk to the public (Bell and Tobin, 2007; Patterson and Doyle, 2009). Case studies have consistently found, however, that potential flood hazard exposure, such as living in a regulatory floodplain or in areas exposed to sea level rise, do not highly correlate with people’s perceptions of flood risks (Horney et al., 2010; Ludy and Kondolf, 2012; McPherson and Saarinen, 1977; Siegrist and

of

Gutscher, 2006).

ro

2.4 Appraisal of Adaptation Resources

Hypothesis 4: Individuals who perceive that they have greater capacity to adapt to floods (i.e., access to

-p

adaptive resources) will have lower perceived risk of future floods.

re

Beyond considerations of social group memberships (SVT) and past experiences (PMT), people hold

lP

other beliefs and embodied experiences that are more complex, intersectional, and intertwined with individual and community circumstances in society. Many kinds of resources – those held privately within households and those held collectively within communities and societies – can assist people in

ur na

coping with disasters (Wisner et al., 2004). An individual’s privately-held resources, such as money, knowledge, and capacity for independent action empower people to protect themselves (i.e. selfefficacy). Social cohesion, also known as social capital, is a communal resource for members of

Jo

communities or societies that helps to solve problems collectively (i.e. collective efficacy) and promotes social benefits for everyone (Coleman, 1988; Putnam, 1995), as well as confidence (and trust) in institutions (Paxton, 1999; Twenge et al., 2014). Socially cohesive communities can alleviate an absence of material deficits that may help to protect people during and in the aftermath of disasters (Walker and Burningham, 2011). Beliefs about access to those resources may also influence individuals’ risk perception because if people believe they have resources to cope with an extreme event, the consequences would appear less serious. 16

3.0 Results Descriptive statistics for the SWISSH sample (n, percent, and where appropriate, mean +/- SD and range) for all variables and all respondents included in the analysis are presented in Table 1. By quota design, numbers of race/ethnic minorities and low-income respondents were overrepresented in our sample. With appropriate weights applied, the counts and percentages of racial/ethnic and income groups listed

of

in Table 1 are an accurate representation of those groups in the sample population. Thirty-four percent

ro

(34.1%) of respondents were not meeting their household’s basic needs. The sample was more female, and more educated than the general US population, which is common in survey research (Green, 1996;

-p

Holbrook et al., 2008). Twenty-eight percent (27.5%) had experienced a major flood in their lifetimes, 4.4% reported experiencing a flood within the last year, and 39.5% heard floods news at least several

re

times per month. Half (49.1%) were very or somewhat confident that they can cope with the

was 3.9 on a 7-point scale.

lP

consequences of a flood. Respondents’ mean perceived risk of a major flood within the next 10 years

ur na

Table 1. Survey of Water Indicators and Socioeconomic Status of Households: Sample Characteristics Characteristic % (n)

Jo

Response wave Wave 1 (Dec. 6 – Dec. 12, 2017) Wave 2 (Dec. 28, 2017 – Jan. 2, 2018) Wave 3 (Jan. 18 – Mar. 28, 2018) Study region Eastern Massachusetts (Boston – Worcester) Front Range - Colorado (Denver – Fort Collins) Mid-Atlantic (Washington, DC – Baltimore, Maryland) Pacific Northwest - Oregon (Portland – Eugene) Piedmont Atlantic (Atlanta, Georgia – Charlotte, North Carolina) Southeastern Florida (Miami – Palm Bay – Melbourne) Southeastern Michigan (Detroit – Flint) Southern California (Los Angeles – San Bernardino) Sun Corridor - Arizona (Phoenix – Tucson) Social Vulnerability Age, in years

Mean + SD (min, max)a

9.3 (855) 33.5 (3098) 57.3 (5296) 11.1 (1024) 11.2 (1034) 10.9 (1006) 11.3 (1044) 11.2 (1036) 11.0 (1013) 11.1 (1029) 10.9 (1009) 11.4 (1055) 49.5 + 15.3 (25.0,96.0)

17

64.1 (5920) 57.4 (5311) 20.7 (1918) 13.2 (1216) 6.2 (570) 2.5 (235)

ro

40.2 (3716) 30.2 (2789) 29.7 (2745) 34.1 (3152) 65.4 (6005)

re

lP

ur na

Jo

Confidence in personal coping ability Very confident Somewhat confident Neutral No confidence at all/hardly any confidence Perceived likelihood of community problem-solving Very likely Somewhat likely Neither likely nor unlikely Somewhat unlikely Very unlikely

of

1.9 (177) 17.9 (1646) 25.4 (2342) 32.2 (2964) 22.6 (2079)

-p

Female Race/ethnicity Non-Hispanic White Hispanic Non-Hispanic Black Non-Hispanic Asian Non-Hispanic Other Race or Mixed Educational attainment Less than high school High school Community college or vocational/technical school 4-year college Graduate/professional school Income group <$50k $50-$100k >$100k Does not have access to basic needs Homeowner Exposure to Floods Potential exposure to flood hazardb Experienced major flood, lifetime More than once Once Never Frequency of flood news Almost daily/Several times a week Several times a month Once a month Never Experienced major flood, past yearc Experienced flood-related property damage, lifetimec Lost everything/Major property damage Minor property damage No property damage Experienced flood-related injury or physical illness in household, lifetimec Experienced flood-related emotional or mental illness in household, lifetimec Appraisal of Adaptation Resources

6.7 + 12.2 (0.0, 98.5)

10.1 (914) 17.4 (1572) 72.6 (6574) 14.0 (1280) 25.5 (2334) 37.1 (3388) 23.4 (2142) 4.4 (407) 4.1 (378) 15.5 (1438) 6.6 (611) 1.9 (178) 2.6 (244)

16.2 (1488) 32.9 (3029) 27.9 (2568) 23.0 (2112) 43.7 (3840) 35.3 (3103) 10.2 (897) 7.5 (661) 3.3 (293)

18

3.7 + 0.8 (1.0, 5.0) 3.9 + 1.8 (0.0,7.0)

of

Confidence in institutions, score Risk Perception Perceived risk of major flood event, score Note. All sample estimates are weighted based on the non-proportional quota sample design, i.e., by race/ethnicity and income. Individuals with missing data (i.e. preferred not to answer, item not asked due to skip pattern) on a given variable were excluded from the calculated percentage for that variable. a SD = standard deviation, min = minimum value, max = maximum value. b Percent of a respondent’s developed ZCTA land designated as SFHA. c Item only asked of respondents who reported experiencing major flood in lifetime; percentage presented is of total sample (N=9250).

The odds ratio coefficients, standard errors, and probability estimates for four proportional odds logit

ro

models (Models 0 – 3) predicting perceived risk of flood in the SWISSH sample are shown in Table 2. The

-p

baseline model (Model 0) controls for wave of survey administration because during the four months the survey was in the field, summer hurricane/flood season receded further into the past. Later wave 3

re

respondents were 21% less likely to perceive higher flood risk than early wave 1 respondents. As more variables were added to subsequent Models 1 - 3, however, wave effect on perceived risk of flood was

lP

not statistically significant. The baseline model (Model 0) included study regions. The largest difference in regional odds ratios, similar to differences in Figure 1b, was that respondents in Southeastern Florida

ur na

were three times more likely to perceive higher flood risk than the reference category (Piedmont Atlantic). Eastern Massachusetts and Sun Corridor – Arizona also had significantly higher perceived risks than the reference category (Model 0).

Jo

In Model 1, which added social vulnerability predictor variables, the differences of perceived flood risk between Southeastern Florida and Eastern Massachusetts with Piedmont Atlantic remained statistically significant. Perceived flood risk in Pacific Northwest-Oregon was also significantly higher in Model 1, controlling for social vulnerability variables. In Model 1, older people were slightly more likely to perceive higher flood risk. Age squared was incorporated to test for nonlinearities in the relationship between age and flood risk. The statistically significant odds ratio for age squared indicates that, as age 19

increases, risk perception increases at a declining rate. Women were 35% more likely to perceive higher flood risk than men. Hispanics were 22% more likely and non-Hispanic Blacks were 26% more likely to perceive higher risk than non-Hispanic Whites. Respondents in the low and middle annual household income groups were significantly more likely to perceive higher flood risk than respondents with incomes greater than $100,000 (< $50,000, 14% more likely; between $50,000 and $100,000, 18% more likely). Respondents without basic access to health insurance, housing, healthcare, electricity/gas, and

ro

of

food were 40% more likely to perceive higher risk than those with basic access to those goods.

In Model 2, which added flood experience variables, respondents who had experienced a major flood

-p

once in their lifetimes were 65% more likely to perceive higher risk than people who had never experienced a flood. People who had experienced a major flood more than once were about two times

re

more likely to perceive higher risk. The amount of indirect exposure to floods, experienced as flood

lP

news, was also statistically significant. Compared to the reference category, never heard flood news, respondents who heard occasional news about floods were 41%-64% more likely to perceive higher flood risk, while respondents who heard about floods almost daily were twice as likely to perceive flood

ur na

risk. Potential exposure to flood hazard (i.e., the local fraction of residential land in flood prone area) significantly increased the odds of perceiving higher flood risk. The social vulnerability variables, which were statistically significant in Model 1, remained significant in Model 2. Additionally, having controlled

Jo

for experience variables, Asians were 15% more likely to perceive higher flood risk than non-Hispanic Whites. Homeowners were 9% less likely than renters/other living arrangements to perceive high flood risk.

20

Model 1b (n=9039)

Model 2c (n=8786)

Model 3d (n=8147)

OR

SE

P>z

OR

SE

P>z

OR

SE

P>z

OR

SE

P>z

0.83 0.79 1.33 0.88 1.09 1.12 2.86 1.35 0.93 0.82

0.06 0.05 0.10 0.07 0.08 0.09 0.23 0.10 0.07 0.06

0.008 0.000 0.000 0.081 0.268 0.162 0.000 0.000 0.321 0.012

0.88 0.91 1.42 0.94 1.14 1.22 3.13 1.34 0.91 0.86 1.02 1.00 1.35 1.22 1.26 1.06 0.93 1.05 0.87 0.99 0.93 1.14 1.18 1.40 0.94

0.06 0.06 0.11 0.07 0.09 0.10 0.25 0.10 0.07 0.07 0.01 0.00 0.05 0.07 0.07 0.07 0.08 0.17 0.06 0.06 0.05 0.07 0.06 0.06 0.04

0.069 0.156 0.000 0.394 0.103 0.012 0.000 0.000 0.215 0.059 0.020 0.000 0.000 0.001 0.000 0.329 0.420 0.780 0.034 0.824 0.178 0.032 0.001 0.000 0.202

0.89 0.88 1.30 0.95 1.07 1.25 2.69 1.21 0.87 0.87 1.03 1.00 1.38 1.19 1.29 1.15 0.97 1.15 0.93 1.02 0.96 1.14 1.15 1.29 0.91 1.98 1.65 2.19

0.07 0.06 0.10 0.08 0.09 0.10 0.27 0.10 0.07 0.07 0.01 0.00 0.06 0.07 0.08 0.07 0.08 0.20 0.06 0.06 0.05 0.07 0.06 0.06 0.04 0.43 0.08 0.15

0.120 0.077 0.001 0.511 0.395 0.007 0.000 0.019 0.093 0.093 0.007 0.000 0.000 0.005 0.000 0.027 0.697 0.419 0.241 0.767 0.396 0.030 0.006 0.000 0.036 0.002 0.000 0.000

0.90 0.90 1.29 0.94 1.12 1.25 2.63 1.22 0.87 0.89 1.02 1.00 1.32 1.15 1.27 1.09 0.96 1.19 0.92 1.01 0.95 1.17 1.16 1.29 0.91 2.14 1.66 2.20

0.07 0.07 0.11 0.08 0.09 0.11 0.27 0.10 0.07 0.07 0.01 0.00 0.06 0.07 0.08 0.07 0.09 0.21 0.06 0.06 0.05 0.07 0.06 0.06 0.04 0.48 0.09 0.15

0.171 0.132 0.002 0.482 0.166 0.008 0.000 0.016 0.095 0.177 0.022 0.000 0.000 0.032 0.000 0.173 0.695 0.338 0.231 0.818 0.347 0.011 0.005 0.000 0.056 0.001 0.000 0.000

ur na

Jo

re

-p

ro

Model 0a (n=9188)

lP

Wave 2 (Dec. 28, 2017 – Jan. 2, 2018) Wave 3 (Jan. 18 – Mar. 28, 2018) Eastern Massachusetts Front Range - Colorado Mid-Atlantic Pacific Northwest - Oregon Southeastern Florida Southeastern Michigan Southern California Sun Corridor - Arizona Age, y Age squared Female Hispanic Non-Hispanic Black Non-Hispanic Asian Non-Hispanic Other Race or Mixed < High school High school Community college or vocational/technical school 4-year college <$50k $50-$100k Does not have access to basic needs Homeowner Potential exposure to flood hazarde Experienced major flood once in lifetime Experienced major flood more than once in lifetime

of

Table 2. Results From Proportional Odds Logit Models Assessing Predictors of Perceived Risk of Flood Score in 9 Urban Regions of the United States

21

Jo

ur na

lP

re

-p

ro

of

Exposed to flood news almost daily/several times a week 2.08 0.14 0.000 2.06 0.14 0.000 Exposed to flood news several times a month 1.64 0.09 0.000 1.59 0.09 0.000 Exposed to flood news once a month 1.41 0.07 0.000 1.40 0.07 0.000 No/hardly any confidence in personal coping ability 1.42 0.08 0.000 Somewhat confident in personal coping ability 1.06 0.05 0.212 Very confident in personal coping ability 0.66 0.04 0.000 Perceives community problem-solving as very unlikely 0.83 0.11 0.166 Perceives community problem-solving as somewhat unlikely 0.94 0.09 0.532 Perceives community problem-solving as somewhat likely 1.10 0.08 0.170 Perceives community problem-solving as very likely 1.10 0.08 0.172 Confidence in institutions, score 0.95 0.03 0.045 Note. All models were weighted based on survey quota sample design, i.e., by race/ethnicity and income. OR = Odds Ratio; SE = Standard Error; P>z=Probability of OR. Bolded values indicate association is statistically significant at the 0.95 confidence level. Reference groups for categorical variables are (in listed order): Wave 1 (Dec. 6 – Dec. 12, 2017), Piedmont Atlantic, Male, Non-Hispanic White, Graduate/professional school, >$100k, Has access to basic needs, Non-homeowner, No experience of major flood in lifetime, Never exposed to flood news, Neutral level of confidence in personal coping ability, Perceives community problem-solving as neither likely nor unlikely. a Model 0: Wald X2 (d.f.=10) = 286.84 (p=0.00), Pseudo Log Likelihood = -17827.416 b Model 1: Wald X2 (d.f.=25) = 688.15 (p=0.00), Pseudo Log Likelihood = -17290.078 c Model 2: Wald X2 (d.f.=31) = 857.80 (p=0.00), Pseudo Log Likelihood = -16623.51 d Model 3: Wald X2 (d.f.=39) = 917.52 (p=0.00), Pseudo Log Likelihood = -15313.167 e Percent of a respondent’s developed ZCTA land designated as SFHA.

22

Respondents’ appraisals of their personal coping abilities to protect themselves and recover from floods was statistically significant in Model 3. Compared to the reference category (neutral about coping ability), people who felt very confident were 34% less likely to perceive higher risk from floods. People who did not feel confident were 42% more likely to perceive high risk. Respondents’ confidence in community problem-solving did not affect the odds of perceiving higher flood risk. Higher levels of confidence in institutions reduced perceptions of flood risk by 5% compared to lower confidence. A one-

of

way ANOVA test showed that non-Hispanic Blacks and non-Hispanic other races each had significantly

ro

lower institutional confidence scores than the other three groups (F (4, 8,981) = 66.58, p < 0.001). Therefore, we ran another version of Model 3 (“amended Model 3” - not shown) with an added

-p

interaction term between the race/ethnicity variables and confidence in institutions score. Interaction terms in the amended Model 3 showed that, for non-Hispanic Blacks, increased levels of confidence in

re

institutions significantly reduced the odds of perceiving higher flood risks more than for non-Hispanic

lP

Whites. In Model 3, after controlling for perceived coping variables, the odds coefficient for Asian ethnicity was lower and not statistically significant.

ur na

In sum, socially vulnerable respondents were more likely than others to perceive higher risks of future floods. Respondents with direct or indirect flood experiences perceived higher risks of future floods. Across all models in Table 2, the statistical significance of variables remained stable, indicating that SVT

Jo

and PMT variables have independent effects on risk perceptions. Contrary to previous studies, respondents with greater potential exposure to flood hazard in our study did perceive higher flood risk. Respondents with more positive appraisals of their personal adaptation resources were less likely to perceive higher flood risk. Confidence in government institutions lowered the perceived risk of floods, especially among African Americans.

23

We continued to explore the same four hypotheses in Models 2a and 3a using only the subsample that had experienced a lifetime flood (i.e., flood survivors). These equations were structured identically to Models 2 and 3, but five new variables were added to assess the effects of respondents’ experience intensity: flood experienced within the past year; minor or major loss of property; physical injury/illness; emotional illness. In this group of flood survivors, only some of the SVT variables (female, Hispanic, low income, and basic needs not met) significantly raised the odds of higher risk perception. High exposure

of

to flood news raised the odds as well. Potential exposure to flood hazards was not statistically

ro

significant. Having no self-confidence in personal coping ability increased risk perception by 52%, but higher self-confidence did not have a significant effect on risk perception (Model 3a). Of the new

-p

variables, experiencing a major flood in the past year increased the odds of higher risk perception by 3536% (Models 2a and 3a). Flood-related physical injury or illness increased risk perception by 51-53%

lP

re

(Models 2a and 3a).

ur na

Table 3. Results From Proportional Odds Logit Models Assessing Predictors of Perceived Risk of Flood Score, Among Population that has Experienced Lifetime Major Flood Event, in 9 Urban Regions of the United States

Jo

Wave 2 (Dec. 28, 2017 – Jan. 2, 2018) Wave 3 (Jan. 18 – Mar. 28, 2018) Eastern Massachusetts Front Range - Colorado Mid-Atlantic Pacific Northwest - Oregon Southeastern Florida Southeastern Michigan Southern California Sun Corridor - Arizona Age, y Age squared Female Hispanic Non-Hispanic Black Non-Hispanic Asian Non-Hispanic Other Race or Mixed

Model 2aa (n=2383)

Model 3ab (n=2250)

OR

SE

P>z

OR

SE

P>z

0.88 0.95 1.55 0.98 1.38 1.16 2.59 1.47 1.13 0.81 1.03 1.00 1.18 1.28 1.03 1.00 0.75

0.12 0.13 0.25 0.16 0.22 0.20 0.51 0.22 0.20 0.14 0.02 0.00 0.09 0.15 0.11 0.13 0.12

0.375 0.676 0.006 0.896 0.041 0.403 0.000 0.010 0.498 0.202 0.071 0.013 0.038 0.030 0.760 0.993 0.079

0.85 0.95 1.48 0.96 1.40 1.12 2.41 1.45 1.07 0.78 1.03 1.00 1.16 1.30 1.01 0.95 0.73

0.13 0.13 0.24 0.16 0.23 0.20 0.48 0.22 0.19 0.14 0.02 0.00 0.09 0.16 0.12 0.12 0.12

0.283 0.686 0.016 0.792 0.043 0.543 0.000 0.014 0.692 0.152 0.106 0.023 0.060 0.026 0.898 0.668 0.068

24

Jo

ur na

lP

re

-p

ro

of

< High school 1.43 0.40 0.204 1.35 0.41 0.321 High school 0.91 0.12 0.449 0.90 0.12 0.399 Community college or vocational/technical school 1.14 0.13 0.244 1.16 0.13 0.187 4-year college 1.00 0.10 0.986 1.01 0.10 0.952 <$50k 1.44 0.18 0.003 1.51 0.19 0.001 $50-$100k 1.25 0.13 0.031 1.27 0.13 0.026 Does not have access to basic needs 1.43 0.12 0.000 1.43 0.13 0.000 Homeowner 1.01 0.09 0.914 1.06 0.10 0.533 Potential exposure to flood hazardc 1.76 0.68 0.147 1.95 0.78 0.093 Exposed to flood news almost daily/several times a week 1.54 0.20 0.001 1.56 0.21 0.001 Exposed to flood news several times a month 1.25 0.14 0.049 1.28 0.15 0.033 Exposed to flood news once a month 1.19 0.13 0.116 1.24 0.14 0.058 Experienced major flood, past year 1.35 0.15 0.006 1.36 0.15 0.007 Lost everything/major property damage 1.23 0.16 0.103 1.20 0.16 0.170 Minor property damage 0.88 0.08 0.165 0.83 0.08 0.050 Experienced flood-related injury or physical illness in household, lifetime 1.51 0.26 0.016 1.53 0.27 0.016 Experienced flood-related emotional or mental illness in household, lifetime 1.01 0.15 0.970 0.98 0.15 0.916 No/hardly any confidence in personal coping ability 1.52 0.17 0.000 Somewhat confident in personal coping ability 1.18 0.12 0.089 Very confident in personal coping ability 1.11 0.14 0.402 Perceives community problem-solving as very unlikely 1.21 0.31 0.468 Perceives community problem-solving as somewhat unlikely 1.14 0.21 0.459 Perceives community problem-solving as somewhat likely 1.22 0.17 0.147 Perceives community problem-solving as very likely 1.23 0.17 0.129 Confidence in institutions, score 1.02 0.05 0.684 Note. All models were weighted based on survey quota sample design, i.e., by race/ethnicity and income. OR=Odds Ratio; SE= Standard Error; P>z=Probability of OR. Bolded values indicate association is statistically significant at the 0.95 confidence level. Reference groups for categorical variables are (in listed order): Wave 1 (Dec. 6 – Dec. 12, 2017), Piedmont Atlantic, Male, NonHispanic White, Graduate/professional school, >$100k, Has access to basic needs, Non-homeowner, Never exposed to flood news, Did not experience major flood in past year, No property damage, No lifetime experience of flood-related injury or physical illness in household, No lifetime experience of flood-related emotional or mental illness in household, Neutral level of confidence in personal coping ability, Perceives community problem-solving as neither likely nor unlikely. a Model 2a: Wald X2 (d.f.=34) = 226.41 (p=0.00), Pseudo Log Likelihood = -4332.1413 b Model 3a: Wald X2 (d.f.=42) = 233.59 (p=0.00), Pseudo Log Likelihood = -4049.2336 c Percent of a respondent’s developed ZCTA land designated as SFHA.

4.0 Discussion

This study contributes to the body of scholarship on how people form beliefs about their personal risks from environmental harms. We use floods as an example of a hazard that is intensifying primarily because of global and regional changes in climate and human settlement patterns on increasingly

25

impervious urban landscapes. Major floods affect people in cities worldwide, with causes ranging from coastal storm surges to overflowing rivers and streams and urban drainage systems (Rufat et al., 2015).

Most prior research on perceptions of flood risk has relied on case studies of local floods and motivations for self-protective behaviors (Bubek et al., 2012; Wachinger et al., 2013; Rufat et al., 2015). To our knowledge, however, this study is the first national snapshot of perceived flood risks in the US.

of

We distributed a social survey to a large cross-section of the population in nine major US urban areas,

ro

which allowed us to examine risk perceptions across a wide range of sociodemographic characteristics,

-p

flood experiences, and exposures to different kinds of flood hazards. .

We tested four theory-driven hypotheses that evaluate the importance of people’s social circumstances,

re

cognition, and coping resources in predicting their risk perceptions of floods. Results from the

lP

proportional odds logit models (Tables 2 and 3) confirmed the first hypothesis, drawn from Social Vulnerability Theory, that particular social groups are more susceptible to environmental harms (elderly, women, ethnic minorities, and low-income people) and are more fearful of uncertainties that portend

ur na

threats to their well-being. One possible reason is that they are more likely to live in unprotected, floodprone areas (Burton and Cutter 2008; Montgomery and Chakraborty 2015). A second reason may be that these social groups are accustomed to living with constraints on their abilities to avoid risks or cope

Jo

with consequences of disaster (Brody et al. 2004). For example, lack of access to basic human needs (food, housing, medical care) and lack of mobility (Chakraborty et al., 2019a) represent hardships that constrain adaptive capacity. Similarly, renting a home rather than owning property restricts choices about where to live, housing quality, and constraints on protective actions (Walker and Birmingham, 2011). In the aftermaths of Hurricanes Katrina and Harvey in 2005 and 2017, respectively, studies have documented that floods affect female, poor, and African American individuals and communities more

26

severely than others (Bullard and Wright, 2009; Walker and Birmingham, 2011; Chakraborty et al., 2019b).

Results also confirmed the second hypothesis, drawn from Protective Motivation Theory: independently of vulnerability-related characteristics, flood risk perceptions were highly influenced by memories of past experiences, how recently events had occurred, and frequency of exposure to news that evokes

of

flood imagery. Prior studies have found that direct personal experience, that is, experiencing an

ro

environmental disaster such as a flood, can heighten risk perceptions (Grothmann and Reusswig, 2006; Wachinger et al., 2013). The more recent the past event, the more risk perception is heightened

-p

(Grothmann and Patt 2005). In this study, respondents with firsthand experience with a flood) were much more likely to have high risk perceptions, especially if the flood had occurred recently. Indirect

re

experience with flooding, acquired through media or personal connections, was also associated with

lP

increased risk perceptions (Leiserowitz, 2006; Uzzell, 2000). Frequent news reports of national and global floods, for example, may trigger individuals’ memories of past events and raise risk perceptions for everyone, independently of individual experiences (Wachinger et al. 2013). Thus, our findings in a

ur na

sample survey of US urban residents support case studies of smaller areas. These studies also show how social vulnerability, subjective assessments, salience of event memories, and complex cognitive

Jo

processes influence the perception of flood risk.

Only some of the results in Table 3 agreed with case studies of flood survivors. These studies have shown the aftermath of previous flood experience increases perceived risk of future floods. For example, psychological distress and illness such as post-traumatic stress disorder are common among survivors of flood disasters (e.g., Pietrzak et al., 2012) and have been found to heighten risk perception. We did not find a positive effect on risk perceptions for emotional illness, but physical illness/injury showed a positive association. 27

Interestingly, findings from this study did not support the third hypothesis that potential exposure to flood hazard (measured by percent of ZCTA developed land classified as a flood hazard area) would not affect risk perceptions (Table 2). Prior case studies highlight widespread misunderstanding of federal floodplains (Horney et al. 2010; Siegrist and Gutscher, 2006), and that expert knowledge of local flood hazards is poorly communicated to the public (Bell and Tobin 2007; Brody, 2018; Galloway et al., 2006).

of

For example, people living in flood-vulnerable lands below sea level in the state of California, including

ro

well-educated people with higher incomes, did not understand that their homes were at risk of flooding (Ludy and Kondolf, 2012). Large proportions of residents in low-income and predominantly minority

-p

communities in the states of Delaware and Massachusetts expressed high concern about the seriousness of flood risk after being informed by researchers. They had no knowledge, however, of what

re

causes sea level rise or what they could do to protect themselves (Douglas et al., 2011; Perez and Egan,

lP

2016). A study in the state of North Carolina found a mismatch between actual and perceived risk: slightly over half the participants’ risk perceptions matched their actual level of flood risk and only 20%

ur na

of residents living in a 100-year floodplain correctly identified high flood risk (Horney et al., 2010).

Nevertheless, and despite the fact that the respondents’ street addresses were unavailable to us (therefore, their exact locations relative to floodplains were unknown), results for our full national

Jo

sample in Table 2 (Models 2 and 3) showed that respondents’ flood risk perceptions increased as the percent of developed ZCTA land classified as a flood hazard area increased. We speculate that more frequent, widespread, and costly storms over the last few years (Changnon, 2008; Rahmstorf, 2017; NOAA, 2019) have brought floods “closer to home” for many. This situation may have raised local awareness of floodplains, even for people who have not experienced a flood themselves.

28

In the subsample of lifetime flood survivors (Table 3, Models 2a and 3a), the flood hazard variable did not produce a significant outcome. For the subsample, we established, with a one-way ANOVA, that experiencing a past flood did not correlate with the flood hazard variable (F (2, 9,067) = 1.03, p = .358). This is potentially because survivors have moved away from places where they experienced floods. We also speculate that memory salience of past personal experiences among flood survivors may overwhelm facts about current potential flood hazard. These findings are potentially important for flood

of

managers concerned about communicating flood risks to the public. Given the measurement error in

ro

our exposure variable, however, this hypothesis requires further investigation with more precise

-p

residential data.

Data from this study partially supported the fourth hypothesis that individuals with more confidence in

re

their access to adaptive resources perceive lower risks from floods than individuals with less confidence.

lP

Model 3 results in Table 2 showed that being very confident of personal coping abilities significantly reduced the odds of higher risk perception, whereas having little or no confidence increased the odds of higher risk perception by nearly half. On the other hand, belief in residential community problem-solving

ur na

ability had a negligible effect. This finding was surprising because a great deal of sociological research has shown that, even among disadvantaged neighborhoods, communities are safer and healthier when neighbors engage in collective problem-solving and take action for the common good (Sampson et al.,

Jo

1999; Sampson et al., 2002). Trust in authorities for flood protection has been shown to reduce risk perceptions (Wachinger et al., 2013) and to increase the likelihood of flood-adaptive behaviors (Valkengoed and Steg, 2019). In this study, higher confidence in institutions reduced perceived risk of flood, especially for non-Hispanic Blacks. Non-Hispanic Blacks (African Americans) in our sample, had less confidence in flood management institutions compared to non-Hispanic Whites. Thus, the larger social significance of our finding is the need for government institutions and flood managers to gain the

29

confidence of the public, particularly African Americans, in the wake of well-publicized social and environmental injustices in flood exposure, emergency management, and recovery.

Results of this study can be applied to improving communication about local flood hazards so that the public has better information about their risks and appropriate mitigation behaviors. This study has shown that indirect experience in the form of news strongly influences risk perception, but whether that

of

kind of knowledge helps people accurately perceive their own risk is questionable. Reading about flood

ro

devastation in Miami may unnecessarily raise risk perceptions for people in Denver with no basis in fact. Artificially elevated risk perception may be counterproductive by inducing fear and hopelessness about

-p

the efficacy of making flood preparations, and this may be especially true for socially vulnerable populations. Better communication is needed between flood managers at all levels of government and

lP

re

community residents about their community’s local flood hazards.

Some flood researchers have made progress using participatory methods to work with vulnerable communities on increasing flood-awareness and encouraging local “collective flood management”

ur na

(López-Marrero and Tschakert, 2011; Perez and Egan, 2016). As we know from research in the social and natural sciences, however, effective flood management is a multi-scalar problem in need of federal and local government cooperation and support. It clearly intertwines with broader societal issues of

Jo

socioeconomic inequalities. It must overcome the lack of confidence and institutional obstacles that hinder some groups from being active collaborators in flood protection and adaptation (López-Marrero, 2010; O’Hare and White, 2018). It is also important to temper the mindset that individual protective action is enough to a difference in actual flood outcomes, especially without more social and economic policy supports for vulnerable populations.

5.0 Conclusions 30

This study used social survey data from nine large urban areas in the US to answer three research questions about people’s perceptions of their individual risks from major flood hazards. Based on statistical analysis of the data, results are as follows: 1) Social vulnerability contributes significantly to higher perceived risks from floods. Individuals’ perceptions of the likelihood and severity of flood risk are higher for the elderly, women, ethnic minorities, and low-income people without access to basic economic needs; 2) Cognition, as manifested in individuals’ experience with past floods, flood news

of

awareness, and potential exposure to regulatory flood hazard areas, is important in predicting higher

ro

perceived flood risk; 3) Individuals’ appraisal of private adaptive resources is more important than

beliefs about the efficacy of their communities and social institutions in reducing their perceived threats

-p

from floods.

re

This paper assessed Social Vulnerability and Protective Motivation theories in a single analysis to reflect the idea that risk perceptions have multiple independent and intersecting predictors. This is a departure

lP

from prior studies, which to this point, have evaluated theories about risk perceptions independently of one another. This successful evaluation highlights that it is both possible and productive to assess

ur na

theories of risk perceptions jointly and helps to understand how individuals in different social circumstances perceive their environmental risks, appraise resources for coping with disasters, and decide whether to take protective actions. Ample evidence in the broader literature supports SVT, which shows that socially vulnerable groups have greater exposure to flood hazard, as well as resource

Jo

constraints that make the costs of undertaking protective actions untenable. PMT researchers could incorporate social vulnerability variables into their models as predictors of risk perceptions and moderators of risk perception and protective action. Chronic underprivileged circumstance is a plausible reason why women, ethnic minorities, low-income, elderly and disabled people have enduring perceptions of high risk, and yet at the same time, have long-term constraints on their choices that prevent them from doing little or nothing to protect themselves (Eisenmann et al., 2007). Study designs 31

that include all the relevant variables for a heterogeneous population sample in a longitudinal (time series panel) survey could shed more light on the on-going debate about the relationship between risk perception and protective action. It is also important to study populations over time because risk perceptions are dynamic and subject to change with life circumstances and the occurrence of new flood impacts or receding memories of floods.

Jo

ur na

lP

re

-p

ro

of

Acknowledgement This work was funded by the National Science Foundation Sustainability Research Network Cooperative Agreement 1444758 and Supplement 1444758, Urban Water Innovation Network. The authors thank Rachel Domond for programming the survey.

32

References Adger, N., 2006. Vulnerability. Global Environmental Change. 16(3), 268-281. Aitken S.C., Cutter S.L., Foote K.E., Sell J.L., 1989. Environmental perception and behavioral geography, in: Gaile G.L, Wilmott C.J. (Eds.), Geography in America. Merrill. Columbus, pp. 218–238. Alderman, K., Turner, L.R., and Tong, S., 2012. Floods and Human Health: A systematic review. Environment International. 47, 37-47.

of

Bell, H.M. and Tobin, G.A., 2007. Efficient and Effective? The 100-Year Flood in the Communication and Perception of Flood Risk. Environmental Hazards. 7(4), 302-311.

ro

Bord R.J., Fisher A., O’Connor R.E., 1998. Public perceptions of global warming: United States and international perspectives. Climate Research 11, 75–84.

-p

Brody S.D., Peck B.M., Highfield W.E., 2004. Examining localized patterns of air quality perception in Texas: a spatial and statistical analysis. Risk Analysis 24:1561–1574.

re

Brody, S.D., Sebastian, A., Blessing, R., & Bedient, P. B., 2018. Case study results from southeast Houston, Texas: identifying the impacts of residential location on flood risk and loss. Journal of Flood Risk Management, 11, S110-S120.

lP

Bubeck, P., Botzen, W.J.W., and Aerts, J.C.J.H., 2012. A Review of Risk Perceptions and Other Factors That Influence Flood Mitigation Behavior. Risk Analysis 32(9), 1481-1495. Bullard, R.D. and Wright, B., 2009. Race, Place, and Environmental Justice after Hurricane Katrina: Struggles to Reclaim, Rebuild, and Revitalize New Orleans and the Gulf Coast. Boulder, CO: Westview.

ur na

Burton, C. and Cutter, S.L., 2008. Levee failures and social vulnerability in the Sacramento-San Joaquin Delta Area, California. Natural Hazards Review. 9(3), 136-149. Butler, C., Pidgeon, N., 2011. From ‘flood defence’ to ‘flood risk management’: exploring governance, responsibility and blame. Environ Plan C 29(3), 533-547.Chakraborty, J., S. Grineski, and T. Collins, 2019a. Hurricane Harvey and people with disabilities: Disproportionate exposure to flooding in Houston, Texas. Social Science & Medicine 226, 176-181.

Jo

Chakraborty, J., T. Collins, and S. Grineski, 2019b. Exploring the environmental justice implications of Hurricane Harvey flooding in Greater Houston, Texas. American Journal of Public Health 109(2), 244-250 Chan N.W. 2015. Impacts of Disasters and Disaster Risk Management in Malaysia: The Case of Floods. In: Aldrich D., Oum S., Sawada Y. (eds) Resilience and Recovery in Asian Disasters. Risk, Governance and Society, vol 18. Springer, Tokyo. Changnon, S.A., 2008. Assessment of Flood Losses in the United States. Journal of Contemporary Water Research & Education. 138 (April), 38-44.

33

City of Boston, 2018. Climate Ready Boston. https://www.boston.gov/departments/environment/climate-ready-boston (accessed January 26, 2019). Collins, T.W., Grineski, S.E., and Chakraborty, J., 2018. Environmental injustice and flood risk: a conceptual model and case comparison of metropolitan Miami and Houston, USA. Regional Environmental Change. 18(2), 311-323. Crowell, M., Coulton, K., Johnson, C., Westcott, J., Bellomo, D., Edelman, S., and Hirsch, E., 2010. Journal of Coastal Research. 26(2), 201-211.

ro

of

Cutter, S. L., Solecki, W., Bragado, N., Carmin, J, Fragkias, M., Ruth M., and Wilbanks, T.J., 2014. Urban Systems, Infrastructure, and Vulnerability. Climate Change Impacts in the United States: The Third National Climate Assessment, in Melillo, J., Terese, M., Richmond T.C., and Yohe, G. W., Eds., U.S. Global Change Research Program, 282-296. DOI:10.7930/J0F769GR.

-p

Debbage, N.A., 2018. Urban Flooding Vulnerability: A Multifaceted Comparative Assessment of the Charlanta Megaregion (Doctoral dissertation, University of Georgia). Demnati, A. and Rao, J. N. K. (2004). Linearization variance estimators for survey data. Survey Methodology, 30(1), 17-26.

lP

re

Dewan, T.H. 2015. Societal impacts and vulnerability to floods in Bangladesh and Nepal. Weather and Climate Extremes 7, 36-42.

ur na

Douglas, E.M., Kirshen, P.H., Paolisso, M., Watson, C., Wiggin, J., Enrici, A., and Ruth, M., 2011. Coastal Flooding, Climate Change and Environmental Justice: Identifying Obstacles and Incentives for Adaptation in Two Metropolitan Boston Massachusetts Communities. Mitig. Adapt. Strateg. Glob. Change. 17, 537-562. Dunlap R.E., Catton W.R. Jr., 1979. Environmental sociology. Annu Rev Sociol. 5, 243–273. Eisenman, D.P., Cordasco, K.M., Asch, S., Golden, J.F. and Gilk, D., 2007. Disaster Planning and Risk Communication with Vulnerable Communities: Lessons from Hurricane Katrina.

Jo

Emrich, C.T., and Cutter, S.L., 2011. Social vulnerability to climate-sensitive hazards in the United States. Weather, Climate, and Society. 3, 193-208.

Federal Emergency Management Agency. 2019. National Flood Hazard Layer. https://www.fema.gov/national-flood-hazard-layer-nfhl (accessed July 31, 2019) Ferguson, A. P., & Ashley, W. S., (2017). Spatiotemporal analysis of residential flood exposure in the Atlanta, Georgia metropolitan area. Natural Hazards, 87(2), 989-1016.

Fernandez, A., Black, J., Jones, M., Wilson, L., Salvador-Carulla, L., Astell-Burt, T., and Black, D., 2015. Flooding and Mental Health: A Systematic Mapping Review. PLoS ONE. 10(4), e0119929.

34

Flanagan, B.E., Gregory, E.W., Hallisey, E.J., Heitgerd, J.L., and Lewis, B., 2011. A social vulnerability index for disaster management. Journal of Homeland Security and Emergency Management. 8(1), https://doi.org/10.2202/1547-7355.1792 (accessed January 26, 2019). Galloway, G. E., Baecher, G. B., Plasencia, D., Coulton, K. G., Louthain, J., Bagha, M., & Levy, A. R., 2006. Assessing the adequacy of the national flood insurance program’s 1 percent flood standard. Water Policy Collaborative, University of Maryland. College Park, Maryland. Grothmann, T. and Patt, A., 2005. Adaptive Capacity and Human Cognition: The Process of Individual Adaptation to Climate Change. Global Environmental Change. 15, 199-213.

of

Green, K.E., 1996, Sociodemographic factors and mail survey response. Psychology and Marketing 13(2): 171-184.

ro

Grothmann, T. and Reusswig, F., 2006. People at Risk of Flooding: Why Some Residents Take Precautionary Action While Others Do Not. Natural Hazards. 38, 101-120.

-p

Halpern-Felsher, B.L., Millstein, S.G., Ellen, J.M., Adler, N.E., Tschann, J.M., and Biehl, M., 2001. The role of behavioural experience in judging risks. Health Psychology. 20(2), 120-126.

re

Harlan, S.L., Sarango, M.J., Mack, E.A., and Stephens, T.A., 2019. Data Statement for “An Assessment of Perceived Flood Risk in Urban Areas of the United States.”

lP

Harrell, F. E., 2015. Ordinal logistic regression. In: Regression modeling strategies. Springer, Cham, pp. 311-325. Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., Kim, H., and Kanae, S. 2013. Global flood risk under climate change. Nature Climate Change 3:816–821.

ur na

Holbrook, A.L., J.A. Krosnick, and A. Pfent, 2008. The causes and consequences of response rates in surveys by the news media and government contractor survey research firms. Pp. 499-678 in Advances in Telephone Survey Methodology, Edited by J.M. Lepkowski, C. Tucker, J.M. Brick, E. de Leeuw, L. Japec, P.J. Lavrakas, M.W. Link, and R. L. Sangster. John Wiley & Sons, Inc.

Jo

Homer, C.G., Dewitz, J., Yang, L., Jin, S., Danielson, P., Xian, Coulston, J., Herold, N., Wickham, J. and K. Megown. 2015. Completion of the 2011 National Land Cover Database for the conterminous United States – representing a decade of land cover change information, Photogrammetric Engineering and Remote Sensing, Vol. 81, 345-353. Horney, J., MacDonald, P.D.M., Van Willigen, M., Berke, P.R., and Kaufman, J.S., 2010. Individual Actual or Perceived Property Flood Risk: Did it Predict Evacuation from Hurricane Isabel in North Carolina, 2003? Risk Anal. 30, 3. Kalton, G. and Flores-Cervantes, I., 2003. Weighting methods. Journal of official statistics, 19(2), p.81. Kellens, W., Terpstra, T., and Maeyer, P.D., 2013. Perception and Communication of Flood Risks: A Systematic Review of Empirical Research. Risk Anal. 33(1).

35

Kellstedt, P.M., Zahran S., Vedlitz A. 2008. Personal efficacy, the information environment, and attitudes toward global warming and climate change in the United States. Risk Anal. 28(1), 113–126. Leiserowitz, A.A., 2006. Climate change risk perception and policy preferences: the role of affect, imagery, and values. Climatic Change 77, 45–72. Lindell, M.K. and Perry, R.W., 2004. Communicating environmental risk in multiethnic communities. Sage, Thousand Oaks, CA.

of

Leopold, L., 1968. Hydrology for urban land planning: A guidebook on the hydrologic effects of urban land use. United States Geological Survey Circular 554. https://pubs.usgs.gov/circ/1968/0554/report.pdf (accessed Jan. 24, 2019).

ro

Lopez-Marrero, T., 2010. An Integrative Approach to Study and Promote Natural Hazards Adaptive Capacity: A Case Study of Two Flood-Prone Communities in Puerto Rico. The Geographical Journal. 26(2), 150-163.

-p

Lopez-Marrero, T. and Tschakert, P., 2011. From Theory to Practice: Building More Resilient Communities in Flood-Prone Areas. Environment and Urbanization. 23(1), 220-249.

re

Ludy, J. and Kondolf, G.M., 2012. Flood Risk Perception in Lands “Protected” by 100-Year Levees. Nat. Hazards. 61, 829-842.

lP

Maantay, J., & Maroko, A. (2009). Mapping urban risk: Flood hazards, race, & environmental justice in New York. Applied Geography, 29(1), 111-124.

ur na

Maibach, E., Rosner-Renouf, C., and Leiserowitz, A. 2009. Global Warming’s Six Americas: An Audience Segmentation Analysis. Yale Project on Climate Change and the George Mason University Center for Climate Change Communication https://cdn.americanprogress.org/wpcontent/uploads/issues/2009/05/pdf/6americas.pdf (accessed May 19, 2019). Maldonado, A., Collins, T., and Grineski, S., 2016. Hispanic immigrants' vulnerabilities to flood and hurricane hazards in two US metro areas. Geographical Review 106(1), 109-135. Mallakpour, I., & Villarini, G. (2015). The changing nature of flooding across the central United States. Nature Climate Change, 5(3), 250.

Jo

McCullagh, P., 1980. Regression models for ordinal data. J. Roy. Stat. Soc. B. 42, 109–142. McDaniels, T.L., Axelrod, L.J., Cavanagh, N.S., and Slovic, P., 1997. Perception of Ecological Risk to Water Environments. Risk Anal. 17(3), 341-352. McPherson, H.J. and Sarrinen, T.F., 1977. Flood Plan Dwellers’ Perception of the Flood Hazard in Tucson, Arizona. The Annals of Regional Science 11(2), 25-40.

Mennis, J., 2003. Generating Surface Models of Population Using Dasymetric Mapping, The Professional Geographer, 55:1, 31-42.

36

Miami-Dade Green, 2018. Climate Change. https://www.miamidade.gov/green/climate-change.asp (accessed January 26, 2019). Montgomery, M.C. and Chakraborty, J. 2015. Assessing the environmental justice consequence of flood risk: a case study in Miami, Florida. Environmental Research Letter 10(2015) 095010. http://dx.doi.org/10.1088/1748-9326/10/9/095010 National Research Council, 2012. Urban Meteorology: Forecasting, Monitoring, and Meeting Users' Needs. National Academies Press, Washington DC. https://doi.org/10.17226/13328.

of

(NOAA) National Oceanic and Atmospheric Administration, 2019. Weather Fatalities. National Oceanic and Atmospheric Administration. www.nws.noaa.gov/om/hazstats.shtml (accessed Jan 24, 2019).

ro

(NOAA) National Oceanic and Atmospheric Administration and Department of Commerce, 2013. National Coastal Population Report: Population Trends from 1970 to 2020.

-p

O’Connor R.E., Bord R.J., Fisher A., 1999. Risk perceptions, general environmental beliefs, and willingness to address climate change. Risk Anal. 19, 461–471. Ogbu J.U., 1993. Differences in cultural frames of reference. Int. J. Behav. Dev. 16, 483–506.

re

O’Hare, P. and White, I., 2018. Beyond ‘just’ flood risk management: the potential for-and limits toalleviating flood disadvantage. Reg. Environ. Change. 18, 385-396. DOI 10.1007/s10113-017-1216-3.

lP

OneNYC, 2018. Progress Report 2018. https://onenyc.cityofnewyork.us/ (accessed January 26, 2019). Patterson, L. A., & Doyle, M. W. (2009). Assessing Effectiveness of National Flood Policy through Spatiotemporal Monitoring of Socioeconomic Exposure 1. JAWRA Journal of the American Water

ur na

Resources Association, 45(1), 237-252.

Perez, V.W. and Egan, J., 2016. Knowledge and Concern for Sea-Level Rise in an Urban Environmental Justice Community. Sociological Forum. 31(S1), 885-904. Pietrzak, R.H., Tracy, M., Galea, S., Kilpatrick, D.G., Ruggiero, K.J., Hamblen, J.L., Southwick, S.M., and Norris, F.H., 2012. Resilience in the Face of Disaster: Prevalence and Longitudinal Course of Mental Disorders Following Hurricane Ike. 7(6), e38964.

Jo

Poff, N. L., Bledsoe, B. P., and Cuhaciyan, C. O., 2006. Hydrologic variation with land use across the contiguous United States: geomorphic and ecological consequences for stream ecosystems. Geomorphology. 79(3-4), 264-285. Qualtrics, 2019. Qualtrics Online Sample. https://www.qualtrics.com/online-sample/ (accessed May 21, 2019). Rahmstorf, S. 2017 ahead of print. Rising hazard of storm-surge flooding. Proceedings of the National Academy of Sciences Oct 2017, 201715895; DOI: 10.1073/pnas.1715895114 Renn, O., 2011. The social amplification/attenuation of risk framework: application to climate change. Wiley Interdisciplinary Reviews: Climate Change, 2(2): 154-169. 37

Renner R., 2017. The Deadliest Period of a Hurricane? After its Over. Washington Post. https://www.washingtonpost.com/news/posteverything/wp/2017/09/12/the-deadliest-time-during-ahurricane-after-its-over/?noredirect=on&utm_term=.da0d7b47e9f4 (accessed Jan. 24, 2019). Ruddell, D., Harlan, S., Grossman-Clarke, L., Gerardo, S. and C., 2012.. Scales of perception: Public awareness of regional and neighborhood climates. Climatic Change. 111, 581-607. 10.1007/s10584011-0165-y. Rogers, R. W., 1975. A protection motivation theory of fear appeals and attitude change. J. of Psychology. 91, 93–114.

of

Rose, D. (Ed.), 2000. Researching social and economic change: the uses of household panel studies. New York, NY and London, UK: Routledge.

ro

Sampson, R.J., Morenoff, J.D., and Gannon-Rowley, T., 2002. Assessing “neighborhood effects”: social processes and new directions in research. Annual Review of Sociology 2002 28(1), 443-478.

-p

Sampson, R.J., Raudenbush, S.W., and Earls, F., 1997. Neighborhoods and Violent Crime: A Multilevel Study of Collective Efficacy. Science. 277(5328) 918-924.DOI: 10.1126/science.277.5328.918

re

Siegrists, M. and Gutscher, H., 2006. Flood Risks: A Comparison of Lay People’s Perceptions and Expert’s Assessments in Switzerland. Risk Anal. 26(4), 971-979.

lP

Scott, M., 2013. Living with flood risk. Plan. Theory Pract. 14(1), 103-106. Shepherd, J. M, 2013. Impacts of urbanization on precipitation and storms: Physical insights and vulnerabilities. Climate Vulnerability. 5, 109-125.

ur na

Slovic, P., 1987. Risk Perception. Science. 236(4799), 280-285. Smith, A. B., & Katz, R. W., 2013. US billion-dollar weather and climate disasters: data sources, trends, accuracy and biases. Natural Hazards, 67(2), 387–410.

Jo

US Census Bureau, 2018. Geography Program. Zip Code Tabulation Areas (ZCTAs). https://www.census.gov/programs-surveys/geography/guidance/geo-areas/zctas.html (accessed July 27, 2019). Uzzell, D.L., 2000. The psycho-spatial dimension of global environmental problems. J Environ Psychol. 20, 307–318 Wachinger, G., Renn, O., Begg, C., and Kuhlicke, C., 2013. The Risk Perception Paradox-Implications for Governance and Communication of Natural Hazards. Risk Anal. 33(6), 1049-1065. Walker, G. and Burningham, K., 2011. Flood Risk, Vulnerability and Environmental Justice: Evidence and Evaluation of Inequality in a UK Context. Critical Social Policy. 31(2), 216-240. Ward, P.J., Jongman, P., Kummu, M., Dettinger, M.D., Weiland, F.C.S.,and Winsemius, H.C. 2014. Strong influence of El Niño Southern Oscillation on flood risk around the world. PNAS 111 (44), 15659-15664. 38

Welty, C., 2009. The urban water budget. The water environment of cities. Springer, Boston, MA, pp. 17-28. Wisner, B., Blaikie, P., Cannon, T., Davis, I., 2014. At Risk: Natural Hazards, People’s Vulnerability and Disasters. 2nd ed. New York: Routledge. van Valkengoed, A. M., & Steg, L., 2019. Meta-analyses of factors motivating climate change adaptation behaviour. Nature Climate Change, 9(2), 158.

of

Zahran, S., Brody, S.D., Peacock, W.G., Vedlitz, A., and Grover, H., 2008. Social vulnerability and the natural and built environment: a model of flood casualties in Texas. Disasters 32(4), 537-560.

Jo

ur na

lP

re

-p

ro

Zbigniew W. Kundzewicz et al., Flood risk and climate change: global and regional perspectives. Hydrological Sci. J. 59(1), 1–28. https://doi.org/10.1080/02626667.2013.857411

39