xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001935120
007 cr mnu|||uuuuu
008 080421s2007 fluab sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002294
Bell, Heather M.
Situating the perception and communication of flood risk :
b components and strategies
h [electronic resource] /
by Heather M. Bell.
[Tampa, Fla.] :
University of South Florida,
ABSTRACT: Loss prevention and distribution must begin well before a flood event at multiple levels. However, the benchmarks and terminology we use to manage and communicate flood risk may be working against this goal. U.S. flood policy is based upon a flood with a one percent chance of occurring in any year. Commonly called the "hundred year flood," it has been upheld as a policy criterion, but many have questioned the effectiveness of hundred year flood terminology in public communication. This research examined public perceptions of the hundred year flood and evaluated the comparative effectiveness of this term and two other methods used to frame the benchmark flood: a flood with a one percent chance of occurring in any year and a flood with a 26 percent chance of occurring in thirty years.This research also explored how flooding and flood risk messages fit into the larger context of people's lives by modeling the relationships between flood related understanding, attitude and behavior and the situational and cognitive contexts in which these factors are embedded. The final goal was to come up with locally based suggestions for improving flood risk communication. Data were collected in the Towns of Union and Vestal, New York. Participants were adult residents of single family homes living in one of two FEMA designated floodplains. Face to face surveys and focus groups were used to gather information on respondents' flood experience and loss mitigation activities; general perception of flood risk and cause; flood information infrastructure; perceptions associated with specific flood risk descriptions; and basic demographic data. Focus groups were also asked to suggest improvements to flood risk communication.Results indicated that experience was the most influential factor in perception and behavior. Additionally, there was little evidence that understanding led to "appropriate" behavior. The 26 percent chance description was the most effective when both understanding and persuasion were included, but interpretations of probabilistic flood risk messages were highly individualized. Finally, regulatory practice likely influences attitude and behavior and may emphasize the likelihood of a particular flood at the expense of the possibility of flooding in general.
Dissertation (Ph.D.)--University of South Florida, 2007.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 266 pages.
Advisor: Graham A. Tobin, Ph.D.
National flood insurance program.
t USF Electronic Theses and Dissertations.
Situating the Perception and Communication of Flood Risk: Components and Strategies by Heather M. Bell A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Geography College of Arts and Sciences University of South Florida Major Professor: Graham A. Tobin, Ph.D. S. Elizabeth Bird, Ph.D. Jayajit Chakraborty, Ph.D. Burrell E. Montz, Ph.D. Eric Oches, Ph.D. Elizabeth Strom, Ph.D. Date of Approval: November 2, 2007 Keywords: flood hazards, hazard percepti on, risk communication, flood policy, national flood insurance program Copyright 2007, Heather M. Bell
Acknowledgments Thank you to all who have given me encouragement and/or assistance over the last few years. My advisor Graham Tobin has provided inspiration, guidance, and the occasional kick in the pants. I am grateful. I would also like to thank the members of my committee: I appreciate your feedback as well as the scholarly example each of you has set. Burrell Montz and Paul Covey, Mary Br idge, and Laura Covey were kind enough to let me share their homes during field work. Thank you for the excellent company and the soft beds. I am also indebted to Steve Cars on, who braved the New York winter to help me in the field, and to Lucius Willis for providing me with spatial data. Both are from Binghamton University. Special thanks to the generous residents of Union and Vestal who opened their homes to me and shared th eir experiences and opi nions. You are in my thoughts. Lastly, IÂ’d like to thank my husband Mike for putting up with the highs and lows of research and my family for their continued love and support.
i Table of Contents List of Tables iv List of Figures ix Abstract xi Chapter 1: Background And Problem 1 Introduction 1 Early Flood Management 3 The National Flood Insurance Act 5 The Hundred Year Flood as Benchmark 8 Uncertainty and The Hundred Year Flood 11 Communication and The Hundred Year Flood 15 Problem Statement 22 Chapter 2: Conceptual Framework and Research Questions 24 Perception and Behavior 24 Project Framework 25 Components of Perception and Behavior 26 Conceptual Framework 35 Research Objectives 36 Research Questions 37 Chapter 3: Data Collection 38 Site Selection 38 Structured Questionnaire 41 Focus Groups 44 Chapter 4: The Study Area 48 Physical Context: The Sus quehanna River And Basin 48 Basin Physiography and Geology 49 Climate 53 Vegetation 54 Physical Context: The Upper Susquehanna Basin 55 Physiography and Geology of the U pper Basin and the Binghamton Area 56
ii Climate 58 Vegetation 59 Social Context 61 A Brief History of the Study Area 62 Neighborhoods 64 Current Demographics 69 Historical Flooding in Union and Vestal 71 June, 2006 Floods 74 Hydrology 74 Impacts 77 Chapter 5: Descriptive Results 87 Situational and Cognitive Factors 87 Location 87 Socio-Economic Factors 89 Experience 93 Flood Risk Information Infrastructure 96 Cognitive Factors 110 General Flood Related Perception and Behavior 117 Understanding Processes and Uncertainty 117 Threat Perception: General 118 Behavior: General 120 Behavior: Event Specific 122 Perception Associated with Specific Descriptions 123 Relative Likelihood 123 Relative Size 125 Understanding of Uncertainty over Time and Space 126 Relative Concern 128 Chapter 6: Analysis And Discussion 132 Exploring Figure 2.1: The Gene ral Model of Perceptual a nd Behavioral Influences 133 Data and Methods 133 Perception I: Understanding of Flood Related Uncertainty 139 Perception II: Perception of Threat 141 Behavior: General 146 Behavior: Event Specific 154 Relationships of Outcome Variables 160 Summary and Conclusions 163 Exploring Figure 2.2: The Model of Specifi c Flood Risk Messages, Settings, and Perception 167 Data and Methods 167 Perceived Relative Size 168 Perceived Relative Likelihood 171 Uncertainty over Space and Time 175 Relative Concern 181
iii Summary and Conclusions 188 Judging Relative Effectiveness 190 Data and Methods 190 Understanding of Flood Related Uncertainty 192 Relative Concern 197 Combined Measure of Effectiveness 202 Summary and Conclusion 204 Improving Flood Risk Communication 205 Participant Descriptions of Flooding 206 Participant Concerns about Flooding 209 Improving Flood Risk Communication 213 Summary and Conclusions 218 Chapter 7: Summary and Conclusions 220 Summary of Results 221 Exploring Figure 2.1: The General Model of Perceptual and Behavioral Influences 221 Exploring Figure 2.2: The Model of Specifi c Flood Risk Messages, Settings, and Perception 224 Judging Relative Effectiveness 225 Improving Flood Risk Communication 226 Conclusions 228 Contributions and Generalizability 231 Future Research 233 References Cited 235 Appendices 250 Appendix A: Questionnaire 251 Appendix B: Focus Group Materials 262 About the Author End Page
iv List of Tables Table 3.1 Demographic Profile of Focus Groups 47 Table 4.1 Census 2000 Demographic Data for Unincorporated Union, Vestal and Broome County 70 Table 4.2 Five Largest Floods Recorded by USGS Gage at Vestal 72 Table 5.1 Respondents in Floodplain by Town 88 Table 5.2 Distance in Feet from Major Waterway 88 Table 5.3 Gender, Race/Ethnicity and Ownership 89 Table 5.4 Property Residence Time in Years 91 Table 5.5 Age 91 Table 5.6 Completed Education 92 Table 5.7 Household Income Levels 92 Table 5.8 Severity of Impact 94 Table 5.9 Total Number of Time s Home or Property Flooded 96 Table 5.10 Information Sources Searched and Received During Flood 98 Table 5.11 Information Types Searched and Received During Flood 98 Table 5.12 Information Sources Searched and Received after Impact 100 Table 5.13 Information Types Searched and Received after Impact 100 Table 5.14 Sources Searched and Received since Response 103
v Table 5.15 Information Types Search ed and Received since Response 104 Table 5.16 Original Source of NFIP Information 105 Table 5.17 Total Number of S ources Searched and Received 105 Table 5.18 Total Number of Information Types Searched and Received 105 Table 5.19 Source Credibility for Total Sample 107 Table 5.20 TV Credibility 108 Table 5.21 Family Me mber Credibility 108 Table 5.22 Town Credibility 109 Table 5.23 County Credibility 109 Table 5.24 Average Percentage of information Sources Sought 111 Table 5.25 Overall Satisfaction with Flood Information 111 Table 5.26 Factors Believed to Contribute to Flooding 113 Table 5.27 Self Rated Flood Knowledge 114 Table 5.28 Self Rated Familiarity with the NFIP 115 Table 5.29 I have Control over What Happens to Me 115 Table 5.30 I Am Constantly Worrying about Something 116 Table 5.31 Primary Responsibility fo r Protecting Against Flood Damages 117 Table 5.32 Understanding of Flood Relate d Uncertainty over Space and Time 118 Table 5.33 Perceived Risk Level 119 Table 5.34 Flooding Is One of My Top Concerns 119 Table 5.35 General Consideration of Flooding 120 Table 5.36 Insurance 121 Table 5.37 Event Specific Measures 122
vi Table 5.38 Flood Thought Most Likely to Occur within Year 124 Table 5.39 Flood Thought Least Likely to Occur within Year 124 Table 5.40 Flood Thought Biggest in Size 125 Table 5.41 Flood Thought Smallest in Size 126 Table 5.42 Flood Could Happen More than Once in a Year 127 Table 5.43 Size of Flood Could Change over Time 127 Table 5.44 Flood Thought Most Concerning 128 Table 5.45 Flood Thought Least Concerning 130 Table 6.1 Measurement of Situational and Cognitive Factors 136 Table 6.2 Measurement of Outcomes 137 Table 6.3 Logistic Regression for Unde rstanding Uncertainty over Time and Space 139 Table 6.4 Logistic Regression for Medium-High Flood Risk 142 Table 6.5 Logistic Regression for Flooding as a Top Concern 145 Table 6.6 Logistic Regression for Insurance 148 Table 6.7 Logistic Regression for M odification of House or Property 150 Table 6.8 Logistic Regression for No Action or Consideration of Flooding 153 Table 6.9 Logistic Regression for Evacuation 156 Table 6.10 Logistic Regression for Protec tion of Home or Personal Property 158 Table 6.11 Improved Models by Ou tcome Variable and Grouping 162 Table 6.12 Logistic Regression fo r Biggest Flood: 100 year Flood 169 Table 6.13 Logistic Regression for Sm allest Flood: 1% Chance Flood 170 Table 6.14 Logistic Regression for Sm allest Flood: 26% Chance Flood 171
vii Table 6.15 Logistic Regression for L east Likely Flood: Total Sample 173 Table 6.16 Logistic Regression for Mo st Likely Flood: 1% Chance Flood 174 Table 6.17 Logistic Regression for Mo st Likely Flood: 26% Chance Flood 175 Table 6.18 Logistic Regression for More than Once per Year 176 Table 6.19 Logistic Regression for Change in Size over Time: All 178 Table 6.20 Logistic Regression for Cha nge in Size over Time: DonÂ’t Know 179 Table 6.21 Logistic Regression for More than Once per Year: DonÂ’t Know 180 Table 6.22 Logistic Regression for More than Once per Year: 15 Chance Flood 181 Table 6.23 Logistic Regressi on for Equally Concerning 182 Table 6.24 Logistic Regression for Equally Concerning: High and Low Concern 184 Table 6.25 Logistic Regression for Most Concerning: 100 Year Flood 185 Table 6.26 Logistic Regression for Mo st Concerning: 26% Chance Flood 185 Table 6.27 Logistic Regression for Least Concerning: 100 Year Flood 186 Table 6.28 Logistic Regression for L east Concerning Flood: 1% Chance Flood 187 Table 6.29 Variation in Possibility of Occurring More than Once per Year: CochranÂ’s Q 193 Table 6.30 Participant Comments on Flooding More than Once per Year 194 Table 6.31 Participant Comments on Change in Size over Time 196 Table 6.32 Variation in Possibility of Change in Size over Time: CochranÂ’s Q 197 Table 6.33 Participant Comments on Concern 198 Table 6.34 Variation in Most Concerning Flood: CochranÂ’s Q 199 Table 6.35 Variation in Least Concerning Flood: CochranÂ’s Q 199 Table 6.36 Variation in Relativ e Concern: Friedman Test 202
viii Table 6.37 Variation in Combined Effectiveness: Friedman Test 203 Table 6.38 Descriptions of th e Largest Experienced Flood 207 Table 6.39 Most Concerni ng Thing about Flooding 210 Table 6.40 Regression for 1% Chance Flood as Least Concerning: Size or Likelihood? 213 Table 7.1 Perceptions and Con cern: 26% Chance Description 229
ix List of Figures Figure 2.1 General Model of Perceptu al and Behavioral Influences 35 Figure 2.2 Model of Specific Flood Risk Messages, Settings and Perception 36 Figure 3.1 Location of Union and Vest al in Broome County, New York 40 Figure 4.1 Susquehanna Basin and Sub-basins 51 Figure 4.2 Physiographic Provinces of the Susquehanna Basin 52 Figure 4.3 The Upper Susquehanna Sub-basin 56 Figure 4.4 Vestal Gage Site and Topography 57 Figure 4.5 Average Monthly Precipita tion at Binghamton Regional Airport 59 Figure 4.6 Population of Broo me County, New York: 1900-2000 61 Figure 4.7 West Corners and West Endicott 65 Figure 4.8 Endwell 66 Figure 4.9 Fairmont Park and Westover 67 Figure 4.10 Twin Orchards/Ideal Terrace 68 Figure 4.11 Castle Gardens 69 Figure 4.12 Peak Annual Discharge for the Susquehanna at Vestal: 1935-2006 72 Figure 4.13 The Susquehanna at Vestal, NY: June 26th to July 3rd, 2006 75 Figure 4.14 24 Hour Precipitation in Nort hern PA and Southern NY: June 27th-28th, 2006 76
x Figure 4.15 Areal Extent of Flooding in June, 2006 78 Figure 4.16 FEMA 100 and 500 Year Floodplains 79 Figure 4.17 West Corners and We st Endicott Flood Control 81 Figure 4.18 Endwell 100 Year Floodplain 82 Figure 4.19 Fairmont Park Flood Control 83 Figure 4.20 Westover Flood Control 84 Figure 5.1 Concern Levels for Equal Concern: All 129 Figure 5.2 Distributional Differences in Least Concerning Level: 100 and 500 Year Floodplains 131 Figure 6.1 Relationships between Si tuational and Cognitive Factors and Understanding, Attitude and General Behavior 165 Figure 6.2 Relationships between Si tuational and Cognitive Factors and Understanding, Attitude and Event Behavior 166 Figure 6.3 Relationships of Situational and Cognitive Factors to Perceptions of Specific Descriptions 189
xi Situating the Perception and Communication of Flood Risk: Components and Strategies Heather M. Bell ABSTRACT Loss prevention and distribution must begi n well before a flood event at multiple levels. However, the benchmarks and term inology we use to manage and communicate flood risk may be working against this goal. U.S. flood policy is based upon a flood with a one percent chance of occu rring in any year. Commonl y called the Â“hundred year flood,Â” it has been upheld as a policy criterion, but many have questioned the effectiveness of hundred year flood terminology in public communication. This research examined public perceptions of the hundred year flood and evaluated the comparative effectiveness of this term and two other methods used to frame the benchmark flood: a flood with a one per cent chance of occurring in any year and a flood with a 26 percent chance of occurring in thirty years. Th is research also explored how flooding and flood risk messages fit into the larger context of peopleÂ’s lives by modeling the relationships between flood rela ted understanding, att itude and behavior and the situational and cognitive contexts in which these factors are embedded. The final goal was to come up with locally base d suggestions for improving flood risk communication.
xii Data were collected in the Towns of Un ion and Vestal, New York. Participants were adult residents of single family ho mes living in one of two FEMA designated floodplains. Face to face surveys and focus gr oups were used to gather information on respondentsÂ’ flood experience and loss mitiga tion activities; general perception of flood risk and cause; flood information infrastructu re; perceptions associated with specific flood risk descriptions; and ba sic demographic data. Focus groups were also asked to suggest improvements to flood risk communication. Results indicated that experience was the most influential factor in perception and behavior. Additionally, there was little evid ence that understanding led to Â“appropriateÂ” behavior. The 26 percent chance descripti on was the most effective when both understanding and persuasion were included, but interpretations of probabilistic flood risk messages were highly individualized. Fi nally, regulatory practi ce likely influences attitude and behavior and may emphasize the likelihood of a particular flood at the expense of the possibility of flooding in general.
1 CHAPTER 1: BACKGROUND AND PROBLEM INTRODUCTION From 1980 through 2005, 67 weather-related disasters in the United States exceeded one billion dollars each in direct damage (Ross and Lott, 2006). In over half of them, flooding was either the primary cause or a significant com ponent of a compound disaster like a hurricane. More property dama ge was caused by and more lives lost to flooding than any other disaster type in the twentieth centur y (Perry, 2000). According to an updated version of Pielke et alÂ’s 2002 reanalysis of National Weather Service data (flooddamagedata.org), direct physical fl ood losses over 72 years prior to 2003 exceeded 171 billion dollars. This statistic does not in clude the catastrophic losses related to Hurricane Katrina; FEMA (2006a) re ported that almost $16 billion in claims were paid out by the National Flood Insura nce Program (NFIP) in 2005. Home ownerÂ’s insurance does not c over flooding and the National Flood Insurance Program (NFIP) covers only those who pay premiums. The only people required to have flood insurance ar e those living in financed homes located in areas subject to whatÂ’s known as the Â“hundred year flood.Â” However, this Â“high riskÂ” floodplain delineation is neither completely accu rate nor static. Over half of U.S. flood losses occur outside the hundr ed year floodplain, in the fi ve hundred year floodplain (described as having moderate risk), or outside both mapped fl oodplains (Smith, 2000;
2 Frech, 2005). Land behind approved levees is also exempt, though not immune to flooding, as the aftermath of Katrina illustrated. Effective loss prevention and distributi on must begin well before a flood event on individual, local, state and fe deral levels, but the benchmar ks and terminology we use to manage and communicate flood risk may be work ing against this goal. The usefulness of the hundred year flood as a policy criteri on has been upheld (NRC, 2000), though it is still debated (GFWNFPF, 2004; AFSPM, 2007). Pr actitioners and researchers alike have questioned the effectiveness of hundred ye ar flood terminology in public communication (NRC, 1995; NRC, 2000; Smith, 2000; Gruntfest et al, 2002; GFWN FPF, 2004). Its use may emphasize risk dichotomies and mask th e irregularity and uncertainty associated with both the timing and consequences of flooding. In a 2006 publication, the National Research Council (NRC) linked the misundersta nding of uncertainty to poor decisions with potentially disastrous results in the face of hazardous events. There is worry that misunderstandings associated with hundred year flood terminology might attenuate concern and reduce the likelihood of mitigative behavior. Alternative descriptions of the benchmark flood ha ve been introduced; the hundred year flood, a flood with a one percen t chance of occurring in any year, and a flood with a 26 percent chance of occurring in thirty years al l represent an event of the same size and likelihood. None of these term s have been adequately tested, however. How do people respond to these other descri ptions? Are they associated with better perception of flood related un certainty? Are they associated with higher threat perception?
3 This research examined public perceptions of the hundred year flood and evaluated the comparative effectiveness of this term and two other methods commonly used to describe and frame policyÂ’s benchm ark flood (a flood with a one percent chance of occurring in any year and a flood with a 26 percent chance of occurring in thirty years). Before individual risk messages were tested, though, this research explored how flooding and flood risk messages fit into the la rger context of peopl eÂ’s lives. In addition to comparing flood risk descriptions, this project modeled the relationships between understanding, attitude and beha vior associated with floodi ng and the situational and cognitive contexts in which these factors are embedded. This chapter covers the establishment of the hundred year flood as a U.S. policy benchmark and form of flood risk communicatio n and lays out the problem addressed in this research. Chapter 2 establishes the con ceptual framework and re search questions for the project. Methods of data collection are ex plained in Chapter 3, while the physical and social characteristics of the study area are outlined in Chapter 4. Chapters 5 and 6 make up the bulk of this work and contain result s and discussion of th e analyses conducted. A summary is provided in Chapter 7, along with general conclusions and suggestions for future research. EARLY FLOOD MANAGEMENT In 1875, a congressional report lamented the lack of a unified flood control program, stating that Â“the experience of one hundred and fifty years has utterly failed to create judicious laws or effective organiza tion in the several states themselves, and no systematic cooperation has ever been atte mpted between them. The latter is no less
4 important than the former, for the river has no respect for State boundariesÂ” (House Doc. No.127, in PWRPC, 1950). This was an early call for streamlining flood management. Prior federal involvement, like the Sw amp Land Acts of 1849 and 1850, the Flood Control Act of 1917, and the 1927 Rivers and Ha rbors Act, focused on specific projects or prioritized navigation. Most flood control, however, was le ft in the hands of individual communities and states. In spite of the 1875 plea, federal i nvolvement remained spotty, though funding and research increased. It was not until the 1936 Flood Control Act that flood control became the official responsibility of the fe deral government (USWRC, 1971). The Act, like others preceding and following it, was passed in order to reduce losses caused by flooding. Delegating responsibility for flood cont rol to a single body (the Army Corps of Engineers) created both organization a nd cooperation between states, though the cooperation was, perhaps, invol untary. Multi-state projects emphasizing flood control as well as navigation became more feasible. In 1936, flood control meant structural m itigation. Organized federal floodplain management ended with dams and levees. Aside from aid, non-structural mitigation continued to be left to the states. In 1958, only seven states had encroachment provisions of any kind (Murphy, 1958). None used the hund red year flood as a guideline and most did not enforce permit requirements. In his ev aluation of state encr oachment provisions, Murphy (1958) states, Â“It appeared that, lack ing firm criteria of channel encroachment, the states tend to establish requirements th at are not in major conflict with existing developments nor unduly restrictive to new developmentsÂ” (Murphy, 1958, 20). These
5 considerations took the form of Â“reasonablenessÂ” when th e National Flood Insurance Act was passed in 1968. THE NATIONAL FLOOD INSURANCE ACT In the 1940Â’s and 1950Â’s, Gilbert White (1945, 1964), Francis Murphy (1958), the Bureau of the Budget (1952), and others recommended a floodplain management program that went beyond structural mitigat ion. This was in no small part due to the approximately seven billion dollars spent on river maintenance and Â“improvementsÂ” from 1936 to 1966 (USWRC, 1971; USWRC, 1979). Resear ch indicated that, in spite of the outlay for flood control, flood losses were not decreasing (Reuss, 1993; Kusler, 1982; Holmes, 1961; Renshaw, 1961). In order to stem the outward flow of cash, a combination of zoning and encroachment regulations, bui lding codes, insurance, and financial incentives and disincentives was suggested. Murphy (1958) identified several requ irements for successful non-structural floodplain management. If regulations were to be enforceable, they must be clear and concise, and set with consistent criteria Though Murphy believed that the state level would be the most appropriate for administer ing flood regulations, he saw that the states had very different management philosophies an d feared that the resulting differences in criteria would cause inequit y. The federal governmentÂ’s financial resources, existing staff, and infrastructure also made it a more suitable choice for esta blishing and recording criteria. Murphy suggested that the criteria should lie between the 65 and hundred year floods in order to be eff ective in reducing losses.
6 Many of the above suggestions were incorporated into the National Flood Insurance Act of 1968. As the 1936 Flood Control Act streamlined structural mitigation, the 1968 Act was designed to create a framew ork for non-structural mitigation. The Act sought to both reduce losses and distribut e those incurred. The National Flood Insurance Act followed the Southeast Disaster Ac t of 1965, a 1966 report on flood insurance commissioned by HUD, and House Document 465, a report entitled A Unified National Program for Managing Flood Losses. The report was produced by a task force headed by Gilbert White and presented to the Bureau of the Budge t in 1966. The National Flood Insurance Act was based in large part on House Document 465, though White was not entirely satisfied with the ma nagement results (Reuss, 1993). The National Flood Insurance Program (N FIP) established in the 1968 Act had three goals: to better indemn ify individuals for flood losses through insurance; to reduce flood damages through management and regulat ion; and to reduce federal expenditures for disaster assistance a nd flood control (FEMA, 2002). These goals emphasized modifying vulnerability and the loss burde n while discouraging non-action and placing event modification (structural mitigation) within a broader management framework. Three inter-related components (mapping, mana gement, and insurance) were designed to function as a whole in achieving the goa ls of the Act by en couraging preferred community responses. In the current framework, mapping is intende d to increase awareness and assist in both floodplain management and the creation of rate maps. Management includes the use of zoning, codes, and permitting in order to d ecrease vulnerability. Insurance shares the loss burden and is intended to reduce reliance on federal ai d. Insurance availability and
7 rates were based on mapped risk zones a nd documented management practices. Basic compliance is enforced through the withholdi ng of insurance, loans, and (in theory) disaster relief. Beyond-compliance is encour aged through the Community Rating System, which offers rate reductions to communities th at initiate additional structural and nonstructural programs (Emergency Management Institute, 2007). The focal point of each of these compone nts is the hundred year floodplain. This is the concise, consistent cr iteria that Murphy believed wa s necessary for the successful administration of a flood management pr ogram. While flood maps may include other information, the information most important fo r administering policy is the designation of the Base Flood Elevation and the Special Fl ood Hazard Area. Both of these are based on the predicted parameters of a flood with a return period of one hundred years. Permits are required for all developmen t within the Special Flood Hazard Area (SFHA). All new construction must be raised at least one foot a bove designated Base Flood Elevation. The National Flood Insuranc e Program prohibits any construction within the hundred year floodway that raises the Base Flood Elevation (BFE) one foot or more. Insurance is required for all new buildi ng within the SFHA, but is unavailable to communities not identified as having SFHAÂ’s or those that do not meet or enforce the management requirements above. Levees rate d to a hundred year level of protection exempt the land behind them from the SFHA. Both the management and insurance components of the National Flood Insurance Program depend on mapping the hundred year flood parameters; these lines are represented as absolute.
8 THE HUNDRED YEAR FLOOD AS BENCHMARK How did the hundred year return period become the benchmark of U.S. flood policy? The Tennessee Valley Authority (TVA) used the Â“intermediate regional floodÂ” as a structural guideline (Reuss, 1993; Godda rd, 1961; Kates and White, 1961), but the one percent chance flood was not widely used prio r to the National Flood Insurance Act. The intermediate regional flood is now equated with the hundred year flood. Reasonableness, efficiency, and the individuals involved all pl ayed a role in its adoption as the regulatory standard. Murphy emphasized reasonableness in his 1958 report. He was speaking of reasonableness regarding specific communities; in 1968 administrators were looking for a reasonable benchmark for all communities. In order to get the program off the ground, administrators Â“initially had to have some figure to useÂ” (Reuss, 1993). MurphyÂ’s Â“reasonablenessÂ” took into account the predic ted and historical parameters of flooding, along with community use and need (Murphy, 1958). In 1968, reasonableness was based on a general cost benefit model. In addition to its association with the TVA, the one percent chance flood was chosen because it Â“constitutes a reasonable compromise between the need for building restrictions to minimize potential loss of life and property and the economic benefits to be derived from floodplain developmentÂ” (Krimm, 1998). Reasonableness was not assessed on a contextu al basis, but a theoretical one. What passed as reasonable in policy formation wa s assumed to be reasonable in practice throughout the country. Though s uggestions were made by re searchers like White, the public was not involved in a policy dialogue. Both Murphy (1958) and White (Reuss, 1993) felt that different criteria for different situations would be more effective than
9 applying a single criterion to co mmunities with different needs, but other forces were at work. The National Flood Insurance Program was broad in scope and its functionality depended on producing easily readable maps. There was pressure to get the program off the ground and in a 1985 intervie w, Gilbert White recalls the first FIA administrator as Â“committed to blanketing the countryÂ” (Reuss, 1993, 55). There was no pilot program; George Bernstein began by Â“making large commitments for surveys, for mapping programs, and for doing this not using the regular federal agenci es, but bringing in consulting engineersÂ” (Reuss, 1993, 55). In order to map the nation as quickly as possible, using multiple organizations, e fficiently coded policy requirements were imperative. The resulting maps were also to be clear, concise, and consistent, able to travel through time and space without obvious alteration. The maps, too, then, needed to be coded efficiently. Increased efficiency and higher levels of codification become more important as distance and the number of receivers and transm itters increases. Contexts differ, however, and codification may change with them. The more highly coded and efficient a message is, the more data it loses and lumps, and the more easily a receiver a ttaches his or her own associations (Boisot, 1995). What is actua lly packed into the efficient, economizing package of phrase or gesture depends on the contextual history of the individuals and groups using it, as well as the situation at hand. Meaning is, in part, socially constructed (Peters, 1995). Apparent similarity in soci al codification (i.e. language) may lead to misunderstanding if differences in contexts are not recognized. Existing methods of categorization tend to be reinforced, and new information may be overlooked. People
10 favor information that confirms rather th an challenges existing beliefs and attitudes (Lingwood, 1974); it is more efficient to ignore or transform other pe rspectives than to create new categories for conflicting information. This tendency makes the effective co mmunication of complicated concepts, especially those that are efficiently coded w ith a few familiar words (as in the case of the hundred year flood), extremely difficult, whet her efficacy is judged through persuasion (Shannon and Weaver, 1949; Boisot, 1995) or understanding. Many hazards and risk researchers have contested the equation of a communication actÂ’s effectiveness with its success in influencing behavior in a ma nner desired by the sender (F. Johnson, 1991; Belsten, 1996; Parker, 2000; Trumbo, 2000; Ka semir, 2003), but persuasion continues to be the practical, if not theo retical, model of much risk communication. More inclusive models of policy making and flood risk comm unication have begun to be explored and implemented, however (e.g. Larson and Plas encia, 2001; Environmental Agency, 2005; Frech, 2005). The initial goal in adopting the hundred year flood criterion was not effective communication of risk or risk policy, but efficient administration and implementation. The effectiveness of this criterion in enc ouraging desired behavior, preventing loss, and engendering understanding of ri sk and policy was not tested. It has yet to be tested thoroughly. Because efficiency was prioritized, a single cr iterion was used, chosen because of its perceived Â“reasonablenessÂ”. Th e efficiency with which the FIA set about Â“blanketingÂ” the country, combin ed with the efficient verbal and visual coding of the hundred year floodplain, may have led to the institutionalization of the Â“hundred year
11 floodÂ” in flood related policy and risk communi cation before its usefulness was proven in either arena. UNCERTAINTY AND THE HUNDRED YEAR FLOOD Flood risk is defined as Â“the probability that one or more events will exceed a given flood magnitude within a specified period of yearsÂ” (USWRC, 1977). This definition reflects the most comm on definition of risk currently used in hazards research, that of probability of occu rrence (Cutter et al, 2003). For the hundred year flood, this means that an event of a specific size or larger can be expected, on average every hundred years. This does not mean it can not happen multiple times in the same year. The return period is based on analysis of the historical record, or flood frequency curves. Historical data can be augmented by comparisons with similar watersheds and precipitation analysis. The basic formul a for obtaining the return period is:  Tr = n + 1/ m where n equals the number of data entrie s and m equals the rank of a specific flood magnitude (Dunne and Leopold, 1978). Probability is simply the reciprocal, so a flood with a one hundred year return period has a one percent chance of o ccurring in any given year (a 0.01 probability). As with any analysis based on a sample, there is an associated error and a set of confidence limits. These Â“uncertainties can be decreased only by obtaining more or better data and by using better statistical methodsÂ” (USWRC, 1977). Policy standards for discharge are calculated based on the guidelines outlined in Water Resources Council Bulletin 17B and use the Log-Pearson Type III distribution
12 represented by Equation 2 (USWRC, 1982; Ha mmett and DelCharco, 2005); error is reduced and better accounted for, but it is not eliminated.  log QT = M + kS QT is the flood discharge (cfs) for a se lected return period T (ie. 100 year) M is the mean of the logarithms of the annual peak discharges K is the Pearson Type III frequency factor, a function of the skew coefficient of the logarithms of the annual peak discharges and the recurrence interval S is the standard deviation of the logarithms of the annual peak discharges Statistical error is only one potential s ource of uncertainty in the concept of the 100 year flood. Using an example from th e IAWCD, the National Research Council (2000) has described probability distribution as a reflection of Â“nat ural variabilityÂ” and the error bounds as Â“knowledge uncertainty.Â” Climate change can be included in natural variability, but the resulti ng probabilities may be skewed. One must also include mechanical error and operator error in crea ting the frequency curv es. In addition, these curves, and the probabilities based on them, may change as land use patterns change in the floodplain and watershed. Many researcher s have linked development to larger, flashier flood responses (Kates and White, 1961; Kusl er, 1982; Burby, 2000; Changnon, 2000; Tang et al, 2005). Maps based on these probabilities may quickly become obsolete (USWRC, 1982). The National Flood Insura nce Reform Act of 1994 mandated that hundred year floodplains be update d every five years, but la ck of available money and personnel has been an obstacle to execution (FEMA, 2003). In addition to the uncertainties asso ciated with determining probability, flood heights, velocities, and consequences of sim ilarly rated events vary from place to place, introducing another type of uncer tainty. Smith has used this type of uncertainty to argue that Â“it is impossible to set definitions for a designated flood that are universally applicableÂ” (Smith, 2000, 255). His argument echoes those set forth by Murphy (1958)
13 and White (Reuss, 1993) regarding the suitab ility of a single criterion. If a risk benchmark is not associated with both tem poral and consequential probability, what does it really mean? Is it useful for either policy or communicatio n? These concerns have been raised by the National Research Council ( 2000) and by Slovic (1986) but the underlying concepts need to be tested. As the hundred year flood designation move d out of the arena of implementation and into the political and public spheres, it lost its associated uncertainty. Scientists assessing hundred year flood parameters proba bly shared a code a nd underlying context. Embedded in the flood parameters is uncertain ty resulting from bot h natural variability and human error. Parameters change with changing conditions and may not include all relevant information. At least some of th ese uncertainties were likely internalized by those conducting the studies. While recent research has indicated that the public favors including uncertainty bounds in weather related comm unication (NRC, 2006), other research has shown that lay pe opleÂ’s contexts and coding syst ems may not be as tolerant of uncertainty and the probabilities used to communicate it (Burton and Kates, 1964; Slovic, 1986; Mileti and P eek, 2002). Some cope with uncertainty by mentally eliminating irregularity or denying a thr eat exists (Tobin and Montz, 1997). An efficiently coded message will perhap s be filtered differently by politicians and by the general public than by those who orig inally produced its content. Uncertainty is likely to be eliminated as the incomi ng message is recoded into existing mental categories. The message will be particularly malleable if the code is recognizable and potentially relevant, as is the case with the hundred year flood. Differences in contextual meaning may not be immediately obvious. It ha s been noted that messages for a general
14 audience also hit contextual subgroups (C allaghan, 1987). It is also the case that messages bound for specific subgroups may be intercepted, in whole or in part, by unintended receivers. It is conceivable, even probable in the case of the hundred year flood, that what is believed to be a message coded for a general audience was originally, and practically speaking, continues to be, targeted toward a subgroup. The wider the variety of s ituations to which uncertain knowledge is applied, the greater the uncertainty and error it produces. But the bigger a policyÂ’s scope, the more important it is to produce absolute benchmar ks. The more visible and politically powerful a policy becomes, the more important that it appear free of uncertainty. Wynne (1992) has argued that the language of risk polic y plays on our distaste for ambiguity and Â“falsely reduces uncertainties to the more comforting illusion of controllable probabilistic processesÂ” (Wynne, 1992, 150). This patterning of risk uncertainty into probability allows risk policy to be made, but uncertainti es may be further patt erned when policy is communicated to the public. The bigger a policyÂ’s scope, the further its message must travel. The more certain a message appears, the more efficient it beco mes. Thus, as a policyÂ’s reach extends, two factors might work towards the elimination of uncertainty in risk policy communication: a general aversion to uncertain ty and the need for policyÂ’s efficient transmission. In its Â“certainÂ” efficiency, it is repl icated and ingrained through mental short cuts and coding mechanisms. As the hundred year flood was quickly institutionaliz ed, its uncertainty morphed into certainty, Â“rapidly transforme d in the community mind to a definition of flood free and flood prone, with areas above th e designated flood perceived to be flood free Â– a misconception often reinforced by flood maps that shade only those portions that
15 are subject to the designat ed floodÂ” (Smith, 2000, 255). More research is needed to confirm this, but the assertion is plausible. General risk co mmunication research supports this conclusion (Wynne, 1992; Hance et al, 1988). Kates and White (1961) have also indicated that map lines, like levees, may produc e a false sense of security (see Tobin, 1995 for description of the levee effect), potentially increasin g losses. Efficient communication, perhaps inadvertently, has succeeded in making effective communication very difficult. If people think in binaries, as Mileti a nd Peek (2002) have argued, the current efforts in many areas to remove neighborhoods ( politically rather th an physically) from the designated hundred year floodplain may be potentially harmful, emphasizing the flood-free rather than the flood-prone. When th is removal is described as Â“insurance relief,Â” one must wonder what behavior is desired, as this type of communication may undermine the goals of loss prevention, lo ss distribution, and reduction of federal expenditure. Each of these goals requires individual and community action discouraged by these campaigns. COMMUNICATION AND THE H UNDRED YEAR FLOOD Powell and Leiss (1997) identified thr ee phases in the development of risk communication. The first phase, dating fr om approximately 1975-1984, focused on comparable risk assessment and emphasized t echnical expertise and the categorization of physical risk. Scientists and communicators were assumed to be altruistic and objective (Kasperson and Stallen, 1991) and communi cation was authoritarian. Audience characteristics were generally igno red. During the second phase (1985-1994),
16 communication remained instrumental and one sided, but made an effort to recognize public wants and needs. Communication was effective when it persuaded the target population to think or act in a manner deemed Â“appropriateÂ”. Altruism continued to be assumed, though total objectivity was quest ioned. Powell and Leiss (1997) did not mention Gilbert White or others in the Ch icago School, but their early work in risk perception (White, 1945; 1964; Murphy, 1958; Kates and White, 1961) made the second stage of risk communication possible. The current phase encourages the build ing of public trust in governmental organizations and experts. Several resear chers have claimed that trust in the communicator and message source is imperativ e for risk communication to be persuasive (Covello et al., 1989; B. Johnson, 1989; Slovic, 1993). Others believe that the emphasis on trust is misguided. Trust implies a level of acceptance that Trettin and Musham (2000) and Leiss (1995) have argued is neither realistic nor necessary and may not be desirable. They believe that credibility, rather than trus t, is the key factor. In both cases, however, communication emphasizes social context and atte mpts to initiate a stakeholder dialogue rather than an official monologue. In this third stage, there has been a move towards understanding and consensus in risk asse ssment and communication. Effectiveness has begun to part ways with persuasiveness, though the extent to which this is practical or possible is debatable. Even though risk communication has outwa rdly emphasized exchange and tried to move away from the expert driven linear m odel, research has show n that expert disdain for lay audiences has persisted (Heath and Ga y, 1997; Cook et al, 2004), as has the idea that opposition can be Â“remediedÂ” through education. This position assumes that
17 difference arises only through lack of info rmation or understanding and that once the information experts possess is acquired, the Â“problemÂ” will be solved (NRC, 1989; Cook et al, 2004). Many risk communi cation campaigns overtly cite behavioral change as the purpose of risk communication (NRC, 1989; 2006) Others have argued that the major goal of risk communication is to improve knowledge (Read and Morgan, 1998), though the implicit assumption in most cases is that once individuals have knowledge of risk and risk processes, they wi ll behave Â“appropriatelyÂ”. Since the implementation of the Nati onal Flood Insurance Program, many (e.g. Fordham, 2000; NRC, 2000; Smith, 2000) have recommended the adoption of public participation in flood risk analysis and polic y making, an approach that reflects PowellÂ’s and LeissÂ’s (1997) third phase. However, the National Flood Insu rance Act was passed prior to this stage and most communication regarding floodi ng continues to work from the expert driven, persuasive, li near model of the first two ph ases. Public participation in official flood risk communicati on continues to lag behind that of some technological risks and associated policies. This may in pa rt be due to the more recent emphasis on technological risk, as well as the lack of outrage associat ed with hazards like nuclear power and the consequent differences in perception (Sandman, 1989). All of the above phases repr esent risk communication as an intentional exchange. In 1986 and 1987, Covello et al. identified f our components of the risk communication process (message source, message design, deli very channel, and target audience) and argued that Â“effective risk communication mu st be understood as a two-way interactive process that is based on mutual respect a nd trustÂ” (Covello et al., 1987, 9). In theory, either side could be the message source, but, reflective of the second phase of risk
18 communication, the focus was on the source as scientist or government. Risk communication was primarily seen as transm itting technical or scientific information regarding risk from the experts to the public. Covello et alÂ’s (1986) linear model is described by Ka sperson and Stallen (1991) as Â“the engineering approach.Â” Communicati on assumes intentionality and a goal, targets an audience, has an expert point source, and flows along designated channels. Much hazards research in communication has b een done assuming a narrow model. Two noteworthy examples however, helped shif t the focus from expert to audience: development of the Â“hear-perceive-respondÂ” model by Mileti (see Mileti et al, 1991; Mileti and OÂ’Brien, 1992; Mileti and Peek, 2000) and Lave and LaveÂ’s (1991) work regarding flooding. Both focused on the relation ship of intended messages, perception, and response. However, the exchange of non-sc ientific types of risk information and the role of social networks were not addre ssed. Beacco et al. (2002) have argued that communication of any scientific knowledge canno t be adequately conceptualized as a one way exchange between expert and public, be cause the media and everyday exchanges are also relevant. Krimsky and Plough (1988) identified five s lightly different components of risk communication (intention of the communicator content of the message, nature of the audience, source of the message, and direc tion of the message) and added a latitude factor. Each component may be interpreted broadly or narrowly. A broad interpretation does not assume a goal and m odels communication as coming from any source through any channel to any audience. It assumes le ss control over the re sult of communication, acknowledges multiple types of risk informa tion, and better anticipates the unintended
19 consequences of risk information taking multiple paths through multiple sources to recipients outside a target audience. It is perhaps a more appropriate representation of reality in the case of hundred year fl ood terminology. Social and communicative networks are also better account ed for and the model begins to break from the assumption that messages (intended or not) are cons umed by disconnected individuals Â– an assumption that rarely reflects real ity (Valente and Schuster, 2002). Others have also moved away from a narrow conceptualization to varying degrees. Trettin and Musham (2000) have de fined risk communication as a purposeful exchange of information and opinion regard ing hazards. While this definition includes non-expert discussion, the emphasis on purpose ignores the influence of non-directed exchanges of non-scientific information a nd social norms. Rohrmann (2000) tried to account for these factors by describing risk communication as a social process that informs, influences and allows for particip ation in decision making. These broad models have not consistently been applied to flood risk communica tion. The truly broad interpretation takes into account the poten tial for communicative free-for-alls, a possibility that narrow models cannot anticipate well. In a narrow model of risk communication, flood risk information is assumed to be transferred from experts to a general a udience via maps, pamphlets, and policy. The message travels through intended channels. B ecause of the efficiency of hundred year flood terminology and the shared language of experts and public, a shared code was initially assumed. However, in using the hundr ed year flood as the basis of verbal and visual communication, risk communicators substituted a contextual sub-group (the experts) for the public as th e target audience. No rese arch on message appropriateness
20 was done prior to the implementation of the NFIP. Sub-contexts and symbol sets were merged and potential differences in contex t were not addressed. By the time risk communicators were concerned with context, new meanings were already entrenched. A narrow model does not take into acc ount informal commu nication. It cannot explain the compounding of error as a message makes its way through different contexts on unintended routes. It assume s that flood risk information is received only through the channels in which it was initially sent a nd that it is not consumed as a group. The assumption of the narrow model of risk communication perhaps contributed to the rapid institutionalization of what has been called Â“the most sp ectacular failure of public communication for any scientific concept of our timeÂ” (in Frech, 2005, 63). Other paths, other sources, and other contexts were not considered. Future communication of flood risk would be wise to adopt a broad communi cation model in order to better anticipate perception and behavior, to a ppropriately tailor a message, and to encourage a dialogue of stake holders (Fordham, 2000). Aspects of th e broad model have also been encouraged by the National Research Council (1995; 2000). In the beginning, communication of flood policy to the public was a secondary concern. Communication is becoming a prim ary concern, but its study and practice continues to assume a narrow, persuasive m odel. Effectiveness is judged on compliance with the NFIP, not on the understanding of its principles. Many flood risk communicators now believe that persuasion is contingent upon unde rstanding of the un certainty concepts originally associated with the hundred y ear flood (NRC, 1995; 2000), but it is not clear that a better perception of uncertainty and probability will bring the risk perception of expert and general public closer together or induc e a desired response.
21 Networks and the social construction of risk information have begun to be addressed in general risk literature, as have social infl uences on risk perception (Heath and Gay, 1997; Steg and Siever, 2000; Mile ti and Peek, 2002; Valente and Schuster, 2002; Grasmuck and Scholz, 2005), but have not b een consistently applied to a general or specific natural hazards context. Further research is needed to explore the relationships of specific flood risk messages to meaning cons truction, infrastructures, understanding of flood irregularity, uncertainty a nd causation, attitude, and beha vior in specific contexts. Then, perhaps, more realistic and usef ul methods of flood policy development, enforcement and communication may be employed and losses reduced. In 2000, the National Research Council suggested several messages thought to better convey uncertainty, though the usefulne ss of the hundred year floodplain as a policy benchmark was upheld. Suggested term inology included the one percent chance flood, percent chance of flooding during a 30 year mortgage (essentially a one in four, or 26 percent chance), and an analogy linking th e chance of flooding in 50 years to the chance of tossing a coin and coming up with heads (NRC, 2000). Including damage potential with this description of probability was also suggested. Other organizations, like the Lincoln Institute of Land Policy, have s uggested a similar shift in terminology (Faber, 1996). The TVA uses the 26 percent chance in a 30 year period in its communication (Newton, 1987). Research has shown, however, th at this description of flood risk may have serious drawbacks (Bell and Tobin, 2007). The Army Corps of Engineers has altered its Principles and Guidelines in an attempt to recapture uncertainty; the objective is to use the probability term Â“one percent chance floodÂ” instead of using the return period. Most emergency management agencies have done the same. Publications,
22 however, show that the Guidelines are not always adhered to (USACE, 1994) and anecdotal evidence from areas using the one pe rcent chance description indicates that this message may also be ineffective. After al l, thereÂ’s Â“a 99 percent chance of it not happening!Â” (coxsmeadow.homestead.com). Research is needed to determine if any or all these messages perform differently under different conditions. Arkin (1987) has emphasized the importanc e of carefully tes ting messages prior to official communication. Practitioners and othe r researchers have also indicated a need to research the effectiveness of differe nt messages (Connelly and Knuth, 1998; NRC, 2006). However, the suggested terms, like the hundred year flood before them, have not yet been tested as to thei r efficacy in influencing eith er public understanding of flood related uncertainty concepts, atti tudes, or behavior. It may be that they reflect the same communicative problems as the hundred year flood. Nor is there much evidence that they will produce similar qualitative perceptions of risk or levels of concern across contexts, even when uncertainty is accepted. In a ddition to looking at specific messages and contexts, research must also address th e relationship between perception of flood irregularity, perception of threat, and response. PROBLEM STATEMENT Over half of flood losses occur outside of the Special Flood Hazard Area (Faber, 1996; Smith, 2000). Government agencies recogn ize the inadequacy of Â“the hundred year floodÂ” in communicating flood risk and the dang er of reinforcing risk dichotomies (NRC, 1995, 2000; GFWNFPF, 2004). Gruntfest et al (2002) have also argued that hundred year terminology needs to be replaced. There ha s been a move to adopt more effective
23 terminology without forsaking the efficiency necessary to carry out a nationwide program, but the introduced messages, like th e hundred year flood, have had only limited testing (Bell and Tobin, 2007). This research fills that gap. In order to evaluate flood risk messages, however, a better contex tualization of the factors related to the perceptions and behaviors associated with flooding is necessar y. Such an examination will help situate the comparative results. Additionally, communicators may need to move beyond risk messages based on return periods and probabilitie s. This research also explores locally generated means of improving flood risk communication.
24 CHAPTER 2: CONCEPTUAL FRAMEW ORK AND RESEARCH QUESTIONS PERCEPTION AND BEHAVIOR The ultimate goal of most formal hazards communication programs is reduced and/or distributed losses through a more accurate perception of uncertainty, a higher threat perception, and the adoption of some form of mitigative behavior. For the past half century, hazards researchers have directly and indirectly tried to make communication and response more effective through the st udy of perception (e.g. White, 1945; Burton and Kates, 1964; Smith and Tobin, 1979; La ve and Lave, 1991; Mileti et al, 1991). However, most of these studies are modeled narrowly and the assumption, of course, is that the three are directly linked. Much hazar ds research operates under the belief that awareness and understanding will lead to desired at titudes and behavior. Other research shows that the links are more complic ated. Some have indeed found that awareness and knowledge of risk are connected to both increased perception of threat and, in turn, increased purchase of insurance or other pr o-active behavior (Palm and Hodgson, 1992). Mileti and Peek (2002) an d Gruntfest (2001), though, cautioned that risk awareness does not mean ri sk internalization or action. Ev en perception of large risk may not be accompanied by fear or con cern (Sjoberg, 2000; Beehler et al, 2001). Additionally, research (Slovi c, 2000; Bell and Tobin, 2007) i ndicates that understanding
25 of uncertainty and perception of risk does not always tran slate into action; behavior change may precede attitude adjustment or understanding (Valente and Schuster, 2002). PROJECT FRAMEWORK Like previous work, this research m odeled general and specific communication, perception, and behavior. It attempted, however to account for some of the influences identified in broader models and did not assu me that behavior is dependent on perception. Major components addressed in relation to communication, perception and response within the context of flood risk included: 1. Location a. Distance from river b. Floodplain status c. Community 2. Socio-economic factors a. Education b. Age c. Income d. Race/Ethnicity e. Gender f. Home Ownership g. Length of Residence 3. Experience a. Frequency of Impact b. Severity of Impact 4. Risk Infrastructure a. Information Sources b. Information Type c. Information Channels d. Credibility e. Frequency 5. Cognitive Factors a. Knowledge b. Information Sufficiency c. Information Seeking d. General Outlook
26 6. Information Processing/Cognitive Setting a. Systematic/Heuristic Components of Perception and Behavior Each componentÂ’s relationship to risk perception and behavior has been discussed, in some form, in hazards literature, though their relative co mbined effects have not been addressed. Gilbert White and the Ch icago School began the work of recognizing perceptual variables and a pplying them to response in a natural hazards framework decades prior to the adoption of the psychom etric model often associated with risk perception research (White, 1945; 1964; Murphy, 1958; Kates a nd White, 1961). Early psychometric studies in risk perception (Slovic et al., 1980, 1979, 1976; Slovic et al., 1974) focused on identifying, mapping, and expl aining the differences between expert and lay perceptions of risk associated with various technological and natural hazards. While the findings of these studies have greatly influenced subsequent work, the research has been criticized. The main concern has been the aggregati on of the data used in analysis; the research lumped people t ogether and didnÂ’t dis tinguish between groups (Brenot et al, 1998; Marris et al, 1998; Siegrist et al, 2005). Respondents were treated as independent actors, rather th an situated individuals (Cut ter, 1993; Lupton, 1999) and the approach provides only a snapshot rem oved from daily context (Gustafson, 1998). Additionally, it has been argued that the qualitative risk characteristics were treated as hazard attributes rather than a reflection of individual and group characteristics and biases (Marris et al, 1998). These critiques have le d researchers to inve stigate a variety of potential influences on risk perception, includi ng location, experien ce, socio-economic
27 factors, infrastructures, information seeking behavior, world views, attitudes, and mental processing. Since the work of both White and Slovic, there has been an increased emphasis on the idea that risk perception is more related to people than hazards themselves (Marris et al, 1998). As in risk communication, the cons ideration of contex t has become more important. Risk perception is believed to be influenced by both indivi dual characteristics and contextual factors (Wakefield and Ellio tt, 2003). It is argued that, because we experience the world through mental member ship in social communities (Zerubavel, 1997), perceptions are rooted in daily life, fo rmed and mediated through interactions with friends, family and others (Asgary and Willis, 1997). These factors, like those listed in the previous paragraph, are important in a broad conceptualization of risk communication. Tobin and Montz (1997) id entify two categories of co mponents that influence perception: situational factors and cognitive fact ors. Together, they constitute the context and potential coding of communication. Situat ional factors include variables of the physical and socio-economic environmen ts. Tobin and Montz (1997) include psychological and attitudinal variables in the cognitive category. An individualÂ’s coding processes and symbologies, locu s of control, methods for coping, and general outlook all fit under the heading of Â“cognitive factorsÂ”. N one exist outside of situational factors, however. Situational factors will determine whet her or not risk communication is relevant (Slovic, 1986). Cognitive factors will help determine a personÂ’s communicative capabilities and needs, as well as control sh ortcuts to processing information. Cognitive factors work alongside situati onal factors to influence perc eption and behavior, but in
28 cases of extreme marginalization or vulnera bility, situational factors may significantly bind cognitive ones. The relationship is not fully understood. Following is a discussion of research that has related the situational and cognitive factor s used in this project to perception and behavior. Location and Physical Components Physical components include the magnitude, frequency, and duration of individual hazard types. Their technical assessment and comparison was the focus of the first stage of risk communication. Some have argued that neither the physical event nor spatially assessed risk levels are strong predictors of concern and res ponse (Palm and Hodgson, 1992; Tobin and Montz, 1997; Gr asmuck and Scholz, 2005). Other research has indicated that proximity to certain types of hazards (inc luding flooding) is indeed related to attitude and behavior (Greene et al, 1981; Montz, 1982). Gerber (2005) found a significant relationship between place and perceived personal and community risk regarding flooding in two communities in Texas. This may reflect differing education programs, migration patterns, experience levels, or some other community based variable not easily measured. Archival and other qualitative data should make interpretation of community based differences in percep tion and behavior possible. Experience Research has consistently indicated that experience influences perception and response, though the direction is not always the same (Sm ith and Tobin, 1979; Tobin and Montz, 1997; Lindell and Perry, 2000; Mileti and Darlington, 2000). Burton and Kates
29 (1964) found that those with previous experi ence with a hazard had a more accurate perception of risk, while Halp ern-Felsher et al (2001) disc overed that those with less experience had a higher percep tion of negative outcome and argued that risk judgment may reflect behavior experience, rather than the other way round. BanduraÂ’s (1994) research indicated that outcome experience (positive vs. negative) may weigh heavily on perception and behavior. Expe rience tends to bound oneÂ’s knowledge of the event, influencing its imaginability, and thus its categorization and assessment (Slovic et al, 1974). In results related to flooding, previ ous experience was st rongly related to mitigative behavior (Kunreuther, 1978; Bur by, 1988). Mileti and Darlington (1997) also found that those with previous experience were more likely to take action. Experience is a combination of situational and cognitive f actors and stands alone in the conceptual framework of this research. Socio-economic Variables Using the socio-economic environment to predict percepti on and response has been shown to be problematic. Individuals a nd communities with similar social relations may react differently to Â“identicalÂ” hazards (Clifford, 1956). Howeve r, individual traits, such as age, gender, income, and race have been correlated with attitude and response. For instance, while both men and women worry about similar things in structured interviews, women worry more and rate ri sks more seriously (Cutter et al, 1992; Gustafson, 1998). More qualitative approaches flush out different ge nder concerns and show power relations to be a key factor in risk percep tion (Gustafson, 1998). This result is also reflected in LogesÂ’ (1994) work, which indicated that women, minorities, low-
30 income groups and those with less education had a higher threat pe rception. Mileti and Darlington (1997) found that highe r education, middle age, and family in the area were positively related to mitigative behavior. Cu ltural variables like language, religion, and behavioral norms may also influence the categorization of incoming information by providing an existing coding system and socia lly constructed symbol set. The perception of the hundred year flood and its replacemen ts might depend on these socially produced categories. Risk Infrastructure The broad model of risk communication is reflected in the infrastructural approach. Risk infrastructures are descri bed by Heath and Gay (1997) as Â“networks through which people obtain, evaluate, and share information through mediated and interpersonal channels.Â” In research regarding informati on channels, both newspapers (McCallum et al, 1991) and television (H ansen, 1991; Anderson, 2001; Bell, 2004) have been cited as a primary source of hazard info rmation. However, these sources are suited to different messages (Spen cer et al, 1992) and may enc ourage different processing, potentially influencing the perception of flood processes and flood threat as well as behavior. On the other hand, Dow and Cutter (1 998) have shown that before behavioral decisions are made (in their case, hurricane evacuation), a variety of information sources are consulted. Montz (1982) has linked increas ed frequency of a mitigative message with increased likelihood to adopt a measure when combined with proximity. Others argue that the success of education campaigns is heavily influenced by e xperience; those with experience are less affected by media campaigns. The influence of the media on flood
31 perception and behavior needs to be ex amined while controlling for personal relationships. It has been proposed that individuals us e their personal netw orks to interpret mediated reports (Nigg, 1982) and that their pe rceptions are influenced by the ways in which others in their communities perceive (Zerubavel, 1997). Personal contact and the perception of peer approval play a significant role in be havioral change or adoption (Valente and Schuster, 2002); Â“social normsÂ” campaigns capitalize on this and could be used in relation to flooding. Identification wi th those modeling a desired behavior or attitude also makes messages more effectiv e (Andsager et al, 2006). Neither the studies that focus on social cognition nor those fo cusing on media channels have adequately grappled with the increasing role of the internet as part of individuals Â’ risk infrastructure, however. General Outlook Researchers have also sought to explain risk perception thro ugh its relation to individualsÂ’ general outl ook on the world. One method is an adaptation of Mary DouglasÂ’s cultural theory, linking four world vi ews of nature to perception of risk and management strategies (Steg and Siever, 2000). Cultural theory was not used in the conceptual framework of this project. It has been criticized for its inadequate measurement and its lack of complexity (Marris, 1998; Sjoberg, 2000). Additionally, Brenot et al (1998) argu e that the variables used in cult ural theory are strongly linked to socio-economic factors, which were in cluded in the research framework.
32 Also included in the framework were other variables associated with cultural theory; individualsÂ’ outlooks on responsibility for action a nd losses, as well as self estimated knowledge and control. Research has linked those who think they know a lot about a hazard and those with an internal locu s of control to lower perceptions of threat (Loges, 1994; Grasmuck and Scholz, 2005). Following SjobergÂ’s (2000) suggestion, general risk sensitivity wa s also accounted for; a pe rson who worries a lot about everything perceives a greater threat from a specific hazard as well (Grasmuck and Scholz, 2005). Information Seeking and Knowledge People both receive and actively seek risk information through their risk infrastructure. Information seeking activity ha s been linked to higher levels of preventive behavior (Mileti and Fitzgerald, 1992; Griffi n et al, 2000), but not necessarily with increased perception of risk (Johnson, 2005). In stead, information avoidance (a potential coping method for reducing ambiguity) and info rmation insufficiency have been related to higher risk perception (Johnson, 2005). A pe rceived information sufficiency gap is believed to be linked to more systematic processing of info rmation (Griffin et al, 2000). However, the desire for more information doe s not appear to be re lated to what people actually donÂ’t know, but by emotional and cogni tive involvement, affective response, and normative pressures (Griffin et al, 2000; Gr asmuck and Scholz, 2005) and perhaps the desire to reduce uncertainty (Heath and Gay, 1997). Perception, though, may be dependent on pre-existing knowledge (W ildavsky and Dake, 1990) and heuristic processes. Recognizing information seeking patterns and describi ng these patterns in
33 relation to perception, behavior and other cognitive and situat ional factors may allow risk communicators to better understa nd their role. Many have advo cated the need for more research in the areas of information seek ing and processing (e.g. Heath and Gay, 1997; Grasmuck and Scholz, 2005; Johnson, 2005). This research helped situate seeking in a larger context of communicat ion, perception, and behavior. Information Processing and Cognitive Setting Heuristics are mental short-cuts. Li ke coding, they make mental tasks like processing information more manageable. The patterning of uncertainty discussed previously is a heuristic process. Two additi onal types of heuristics pertinent to risk perception have been identifie d through psychometric research : availability and affect (Slovic et al, 1976; Finucane et al, 2000). Availability has to do with the ease with which something is recalled. The more easily a hazard is imagined, the more frequent it is assumed to be (Slovic et al., 1976). Risk communication in both the broad and narrow senses will have a significant bearing on availability (Slovic, 1976, 1986; Kasperson et al., 1988). Targeted information campaigns or me dia attention may incr ease the visibility, and thus the perceived frequency and threat of hazards, especially (perhaps primarily) in those without first hand experience. Non-targ eted, informal risk information will also increase the perception of fr equency and, potentially, threat. However, in the absence or failure of risk communication, emotion, repr esented by the affect heuristic, becomes more important (Gerber, 2005; Lee et al, 2005). The affect heuristic descri bes the knee-jerk like-it-o r-donÂ’t-like-it reaction to information. As perceived benefit increases, th e perceived risk decreases. Similarly, as
34 dread increases, perceived risk increases. Levels of dread may be based on knowledge of potential consequences, the pe rceived ability to control the hazard, and the perceived impact on future generations. In general, th e more familiar a hazard is, the less it is dreaded (Daggett, 1987; Slovic, 1987). Most natural hazards like flooding are not dread, while nuclear energy, in any form, is (Slovi c, 1987). Negative reac tions (fear, worry, dread) appear to have a bigger effect on per ception than positive one s (Lee et al, 2005). Researchers have also tried to measure heuristic versus systematic processing of hazard information in relation to perception. Pe ople use systematic processing when they reflect on, discuss, and connect and compar e incoming information to what they know. Systematic processing has been associated w ith stronger, less transitory evaluations, attitudes, beliefs, and behavior regarding risk (Griffin et al, 2000) as well as with specific communication channels (Spencer et al, 1992). It is also argued that processing type depends on emotional involvement (Johnson, 2005), cognitive involve ment (relation of subject to self interest an d others) (Heath and Gay, 1997) and information sufficiency (Griffin et al, 2000). Johnson (2005) has found heur istic-systematic measurem ent scales of general information to be unreliable, but heuristic and systematic differences in processing may be very important to the interpretation of flood risk messages. Because people often use shortcuts, communicators need to better unders tand how these shortcuts relate to specific messages and settings. In this project, specific flood risk messages were examined in settings conducive to both he uristic (questionnaire survey) and systematic (focus groups) processing, but focus group numbers were t oo small to make a useful comparison between the two.
35 Conceptual Framework Based on the literature cite d above, two models were developed to guide this research. Figure 2.1 illustrates the conceptual framework used to explore the general context of flood perception and behavior. Fi gure 2.2 models the relationships between specific flood risk messages, specific cogni tive settings, and perception. Dotted lines indicate weak relationships or those with c ontradictory evidence in the literature. These models were used with the understanding th at the assumptions relating understanding, threat perception, and behavior needed testi ng. Clarification of cont extual associations should aid managers, policy makers, and comm unicators in identifying key constraints and influences as well as appropriate measures. Figure 2.1. General Model of Perceptu al and Behavioral Influences BEHAVIOR (Persuasion) PERCEPTION I Understanding of Flood Related Uncertainty PERCEPTION II Perception of Threat (Persuasion) SOCIO-ECONOMIC Education Age Income Race/Ethnicity Gender Home Ownership Residence EXPERIENCE Frequency of Impact Severity of Impact RISK INFRASTRUCTURE Information Sources Information Type Information Channels Credibility Frequency COGNITIVE FACTORS Information Seeking Information Sufficiency Knowledge General OutlookSITUATIONAL AND COGNITIVE FACTORS LOCATION Distance Floodplain Status Community BEHAVIOR (Persuasion) PERCEPTION I Understanding of Flood Related Uncertainty PERCEPTION II Perception of Threat (Persuasion) SOCIO-ECONOMIC Education Age Income Race/Ethnicity Gender Home Ownership Residence EXPERIENCE Frequency of Impact Severity of Impact RISK INFRASTRUCTURE Information Sources Information Type Information Channels Credibility Frequency COGNITIVE FACTORS Information Seeking Information Sufficiency Knowledge General OutlookSITUATIONAL AND COGNITIVE FACTORS LOCATION Distance Floodplain Status Community
36 Figure 2.2. Model of Specific Flood Risk Messages, Settings, and Perception COGNITIVE SETTING Conducive to Systematic or Heuristic Processing? R I S K M E S S A G E PERCEPTION Size Likelihood Uncertainty Threat SOCIO-ECONOMIC Education Age Income Race/Ethnicity Gender Home Ownership Residence EXPERIENCE Frequency of Impact Severity of Impact RISK INFRASTRUCTURE Information Sources Information Type Information Channels Credibility Frequency COGNITIVE FACTORS Information Seeking Information Sufficiency Knowledge General OutlookSITUATIONAL AND COGNITIVE FACTORS LOCATION Distance Floodplain Status Community COGNITIVE SETTING Conducive to Systematic or Heuristic Processing? R I S K M E S S A G E PERCEPTION Size Likelihood Uncertainty Threat SOCIO-ECONOMIC Education Age Income Race/Ethnicity Gender Home Ownership Residence EXPERIENCE Frequency of Impact Severity of Impact RISK INFRASTRUCTURE Information Sources Information Type Information Channels Credibility Frequency COGNITIVE FACTORS Information Seeking Information Sufficiency Knowledge General OutlookSITUATIONAL AND COGNITIVE FACTORS LOCATION Distance Floodplain Status Community SOCIO-ECONOMIC Education Age Income Race/Ethnicity Gender Home Ownership Residence EXPERIENCE Frequency of Impact Severity of Impact RISK INFRASTRUCTURE Information Sources Information Type Information Channels Credibility Frequency COGNITIVE FACTORS Information Seeking Information Sufficiency Knowledge General OutlookSITUATIONAL AND COGNITIVE FACTORS LOCATION Distance Floodplain Status Community RESEARCH OBJECTIVES The primary objective of this research wa s to make a comparative evaluation of the efficacy of terms commonly used to desc ribe policyÂ’s benchmar k flood (hundred year flood, one percent chance flood, and flood w ith 26 percent chance of occurring in 30 years). Efficacy was judged through both understanding and persuasion. In order to evaluate these codifications of flood risk, however, a better cont extualization of the factors related to the perception and behavi ors associated with flooding was necessary. To that end, the relationships of the factors outlined a bove to the understanding of flood processes, perceived threat, and flood related mitigative behavior were quantitatively modeled. Additionally, this research modele d the interaction of specific flood risk messages (hundred year flood, one percent chance flood, and flood with 26 percent chance of occurring in 30 years) with key situational and cognitive factors and the
37 understanding of flood related uncertainty, thr eat perception, and behavior. The final goal of this research was to identify potentia l improvements to flood risk communication. RESEARCH QUESTIONS The research objectives and conceptual mode ls were examined through four sets of research questions: 1. Which situational and cognitive factors are most highly related to varying perceptions of flood pr ocesses and uncertainty when relationships between the factor s are controlled? To a general perception of flood threat? To mitigative behavior? How are these outcomes related to each other? 2. When relationships between them ar e controlled, whic h situational and cognitive factors are most highly related to varyi ng perceptions of size, likelihood, uncertainty, and concern asso ciated with specific flood risk messages? Messages addressed in this project include the hundred year flood, a flood with a one percent chan ce of occurring in any year, and a flood with a 26 percent chance of occurring in 30 years. 3. Which of these flood risk messages are comparatively most effective with regards to understa nding and/or persuasion? 4. How do people describe floods and what worries them about flooding? How might flood risk communication be improved?
38 CHAPTER 3: DATA COLLECTION In order to answer the research ques tions, data were collected through both archival and field work. Archival resear ch was conducted thr oughout the project and relied primarily on contemporary and historic al written sources in both hard copy and digital form. Collections of pictures and maps were also consulted. Spatial and other archival data provided a star ting point for field work and are more thoroughly referenced in Chapter 4. This chapter focuses on the methods used to collect data in the field and is broken into three sections. The first addresse s study site selection. The second covers the collection of data using a structured questionn aire and the third disc usses gathering data by means of focus groups. Specific methods us ed to analyze the collected data are presented in subsequent chapters. SITE SELECTION Study sites were selected based on si ze and location, National Flood Insurance Program participation, floodplain propor tion, and flood experience. Comparable populations and location within the same c ounty or region would limit some differences in political structures and other systemic variables. Na tional Flood Insurance Program (NFIP) participation indicates a community has a designated Special Flood Hazard Area (SFHA), a recorded hundred year flood elevatio n, and available maps. In theory, an NFIP
39 community will also have been exposed to the terms used to describe flood risk through regulatory processes and requirements. Becau se this project looke d at perception and behavior in areas designated both high and medium risk by FEMA, study sites required sufficient numbers within both the hundred and five hundred year floodplains. Flooding needed to be a legitimate public and persona l concern. Because of the types of problems identified with hundred year flood termi nology, recent experience with major flooding was desired. It was also thought that recent community experience might improve response rates. In order for personal expe rience to be a useful model component, however, impact levels had to vary from little or no impact to high impact; flooding could not extend over the whole of both floodplains. The Towns of Union and Vestal, New York were chosen as study sites based on the above considerations. They are both loca ted in Broome County, in the south central portion of the state, and are separated by the Susquehanna River (see Figure 3.1). The population of the Town of Vestal in 2000 was approximately 26,500; about 27,700 people lived in the unincorporated portion of the Town of Union (US Census Bureau, 2000). The Town of Union also includes Endi cott and Johnson City, incorporated villages which were not used in this study in order to maintain equa l levels of government. Both Towns participate in the NFIP and have flood maps designati ng hundred and five hundred year floodplains. A spatially wei ghted analysis of year 2000 block group populations showed that approximately 15 perc ent of Vestal reside nts and 18 percent of Union residents live in one of the tw o designated floodplains. Both communities sustained heavy damage as a result of reco rd flooding in June of 2006, but individual
40 impact levels varied considerably. Site vis its and Census data s howed that neighborhoods located in the floodplains of both communities are economically diverse. Figure 3.1. Location of Union and Ve stal in Broome County, New York The above commonalities controlled for va riation to some extent. There are differences that might be represented when Town designations ar e included in models, however. VestalÂ’s median income was about $9500 higher than unincorporated UnionÂ’s in 2003 (US Census Bureau, 2006) and a larg er proportion (about ten percent more) of adults in Vestal had BachelorÂ’s degrees or higher. While both communities have Town governance, public service provision is more centralized in Vestal, both generally and during emergencies. Additionally, Union participates in the National Flood Insurance Union Vestal
41 ProgramÂ’s Community Rating System (CRS), while Vestal does not (FEMA, 2007a). More detailed information on the study area will be presented in Chapter 4. STRUCTURED QUESTIONNAIRE The bulk of the data analyzed for this project was collect ed by means of a structured questionnaire survey, included as Appendix A. Face to face administration provided both qualitative and quantitative da ta and allowed interviewers to collect location information. These data enabled a di stance calculation from the nearest mapped creek or river to be made using GIS soft ware and improved cla ssification of floodplain status. Additionally, recording addresses pr evented overlap in survey and focus group recruitment as well as repeat contact after rejection. A strati fied random sample was used to emphasize location; subgroups were created geographically, delineated between those in the hundred year floodplain and those in the five hundred year floodplain in each Town. Floodplains were identified using FE MA digital Flood Insurance Rate Maps (DFIRM). The sampling frame consisted of English speaking adult occupants of single family homes located in either the hundr ed year or five hundred year floodplain. Conducting surveys solely in English could li mit the generalizability of the results. However, in both Towns, percentages of non-English speakers were small, and only one of the households contacted could not pa rticipate due to a language barrier. Field work was conducted duri ng three ten to 12 day peri ods from late October, 2006 to mid January, 2007. Though generally re moved from the study areaÂ’s designated floodplains, locally severe overland flow and tributary flooding occurred in November, 2006. In order to maintain reference consis tency in the survey, areas impacted in
42 November were not sampled on subsequent tr ips. Affected areas included neighborhoods in the hundred and five hundred year floodplai ns surrounding Choconut Creek in Vestal. As a result, the Vestal sample is perhaps th e least spatially representative of the four subgroups. Data collection was designed to incl ude all neighborhoods in Vestal and unincorporated Union that partia lly lay in either the official hundred or five hundred year floodplains. Site visits were made to determ ine whether specific streets fit the sampling criteria (residential single family homes) and were accessible. Individuals living on private streets with gates or no trespassing si gns were not contacted for the survey, but were recruited by mail for the focus groups. Su rvey sampling was concentrated in seven neighborhoods: two in Vestal and five in unincorporated Union. Choconut Creek neighborhoods are not included in this count. A dditionally, substantial portions of streets in both towns remained uninhabited throughout field work; the availa ble pool of hundred year floodplain respondents was thus reduce d. Focus group recruitment postcards were sent to these addresses, but there wa s no response. Regardless, these seven neighborhoods represent most of the townsÂ’ mapped flood risk. Prior to survey recruitment, random st reets in each neighborhood were selected for use in focus group mailings. Residents of these streets were not contacted during the survey process. In the five hundred year floodplain, an nth door random sampling technique was used on remaining streets. Sp ecific Â“nÂ” depended on the size of the development and proportion in the five hundred year floodplain, but was generally every second house. Because the number of hundred year floodplain residents was smaller and
43 a larger proportion of these homes were uni nhabited, every single-family residence on the selected streets that was thought to be in the SFHA was approached. Surveys were conducted from approxi mately 10:00 AM to 5:00 PM on both weekdays and weekends. If there was no res ponse, the house was cont acted at least once more at a different time and day. Introductions were made 167 times. In nine cases, age, illness, or language barriers prevented a me mber of the household from participating. Approximately 72 percent of the remaini ng contacts completed the survey. No one stopped part way through. Interviews took place in respondentsÂ’ homes or on their properties and lasted fr om ten to 45 minutes. The survey was broken into five sections: flood experience and loss mitigation activities; general perception of flood risk and cause; flood information infrastructure; perceptions associated with specific flood ri sk descriptions; and basic demographic data. Questions were generally closed and made use of interval, ordinal, and modified Likert scales as well as nominal responses (see Appe ndix A). Most Likert-t ype scales had a six or seven point spread and were analyzed as interval, rather than ordinal scales. Answers to multiple response questions were selected through a literature revi ew and pre-test and, in most cases, were printed on cards given to the participants. Â“OtherÂ” was always listed as an option. Defined interval and ordinal scales were also printed on cards to aid response. Responses to open-ended questions were analyzed, grouped and coded for use in statistical analyses after data collecti on was completed. Any participant comments and interviewer observations were written on the survey sheet at the point of mention. Face validity and content validity were assessed throughout the development of the questionnaire. A pre-test was conducted to improve clarity, c ontent, and flow and
44 identify potential problems. Completed survey s and any notes taken during the interviews were read over and clarified at the end of each field day. A graduate student in the Binghamton University Department of Ge ography assisted me with survey data collection on several days. In order to better ensure consiste ncy, we had a training session and at the end of each day went through the completed surveys question by question to clarify both coded responses and observati ons or commentary. We worked on opposite sides of the same street, so any difficulties or questions could be addressed quickly. Testing showed no statistical differences in the response patterns of participants based on interviewer, though his rejection rate was somewhat higher. FOCUS GROUPS The projectÂ’s research design included fo cus groups for two main purposes: 1) to help evaluate the potential effects of cogniti ve setting on the percep tion of specific flood risk messages, and 2) to uncover potentia l improvements to flood risk communication. Focus groups were moderately structured and made use of a questioning route rather than a topic guide. Focus group materials are in cluded as Appendix B. Each session lasted approximately two hours, with discussion limite d to about an hour and a half to prevent fatigue. Structured conversati on time ranged from one hour and twelve minutes to one hour and thirty seven minutes. Focus groups were used to collect data related to flood experi ence; perception of flood threat and causes; perceived mitigati on options and responsibility; information networks and preferences; pe rceived meanings of and pref erences for descriptions of flooding; and suggestions for im proving flood risk messages. Th e topics were intended to
45 overlap somewhat with those covered by the su rvey in order to f acilitate comparison. Prior to the beginning of di scussion, individuals were al so asked to provide basic demographic data and brief answers to two questions regarding the hundred year flood and loss mitigation responsibility. All furt her conversation was taped and notes were taken by the moderator. Tapes were then tran scribed. Questions were reviewed and tested prior to use and a final summary question wa s asked in each of the sessions to improve validity. Sessions were scheduled for weekday evenings and weekend mornings on the second and third field trips (mid-December and mid-January) at Binghamton University in Vestal. About three weeks before the sessi on blocks, recruitment postcards were sent out describing the research, session options incentives (refreshm ents and a gift certificate) and a contact email address and phone number. Groups of six or eight are recommended for non-commercial research (K rueger and Casey, 2000) and the target for this project was four groups of approximate ly six people each. The recruitment goal was seven people per group. Focus group recruitment, like survey recruitment, concentrated on adult residents of single family homes living in one of th e two officially designated floodplains. A modified stratified random sample that again emphasized location was used for the first stage of recruitment. Site visits during the first trip identified streets practically closed to door to door survey work. These addresses were included in focus group mailings. In addition to these streets, random streets in each identified neighborhood were pulled from the survey pool to be included in focus groups The addresses of residential parcels with more than half the property in a designate d floodplain located on these streets were
46 determined using a 2004 Broome County pa rcel data layer overlain with FEMA floodplain data in ArcGIS. An nth listing strategy was then employed (n depended on eligible number on particular street, but was usually three). Only recruitment was spatially stratified; the focus groups themselves were mixed. In order to better mimic the social and information networks potentially used by participants, a snowball techni que was used as a second phase of recruitment. Upon first contact, recruits were asked if they knew of anyone else that might want to participate and were encouraged to talk to their nei ghbors about the group and pass out my contact information. Though some have cautioned agains t the use of married couples in groups (Krueger and Casey, 2000), spouses were not discouraged, as they are presumably integral parts of an individualÂ’s in formation and communication network. A total of 317 postcards we re sent to Union and Vest al floodplain residences. Only six people responded directly, well below the anticipated eight to ten percent. These first contacts recruited five additional participants (thr ee spouses and two friends). Weekday sessions did not elicit any interest so three Saturday sessions running from 10:00 AM to 12:00 PM were finalized with expe cted numbers of three, three, and five. Reminder e-mails and calls were made the da y before the scheduled session. The first two sessions ran as planned. On the day of the third and final meeting, however, a storm hit, and only one of the five expected participants attended. A demographic profile of the seven individuals who took part in the focus groups is included as Table 3.1. Focus group data were collected for tw o purposes: 1) to help evaluate the potential effects of cognitive setting (systema tic versus heuristic) on the perception of specific flood risk messages, and 2) to poi nt out possible improvements to flood risk
47 communication. Because of the small numbers of both groups and individuals, the first purpose and the associated model (Figure 2.2) we re not addressed in the analysis of focus group results. Analysis was instead concentrat ed on identifying consistent themes related to improving communication. These results wi ll be presented in Chapter 6. Further presentation of focus group data is limited to substantiation or contradiction of the results of survey analysis. Table 3.1. Demographic Profile of Focus Groups Group 1 12/19/06 (3 people) Group 2 1/13/07 (3 people) Group 3 1/20/07 (1 person) # Women 2 2 0 # White, non-Latino 3 2 1 # From Vestal 2 0 0 Mean Age 67 73 59 # with Bachelors 1 1 1 Median Income $20K Â– $35K $20K Â– $35K Over $100K Included Married Couple? Yes Yes No
48 CHAPTER 4: THE STUDY AREA Every year in the United States, hundreds of thousands of people are affected by flooding. Katrina is still in the national news, but other events fade more quickly from the public conscience. There is certainly an issu e of scale, but we might also look at the cumulative effects of flooding over time. Th e Susquehanna River, which runs between the Towns of Union and Vestal and thr ough New York, Pennsylvania and Maryland, is one of the most flood prone rivers in th e country. The Susquehanna causes average annual flood damages of $150 million per year, in part because over 80 percent of basin communities have residents living and work ing in the floodplain (Susquehanna River Basin Committee, 1998). Chapter 3 outlined the cr iteria used to select study sites and methods of data collection in the field. This chapter provides info rmation on the physical and social context in which data were gathered. PHYSICAL CONTEXT: THE SU SQUEHANNA RIVER AND BASIN The main branch of the Susquehanna Ri ver begins as the outflow from Lake Otsego, close to Cooperstown in northeas t New York State. The river travels approximately 444 miles and empties into Chesapeake Bay near Havre de Grace, Maryland. The Susquehanna is the BayÂ’s larges t tributary and supplie s 50 percent of its freshwater influx (Susquehanna River Basin Commission, 1998), as well as most of its
49 nutrient and pollution loads (Boyer et al, 2002) At the last United States Geological Survey (USGS) gage at Conowingo, MD, the SusquehannaÂ’s discharge ranges from an average 14,800 cubic feet per second (cfs) in August to 79,500 cfs in April (USGS, 2006a). The major tributaries include the West Branch Susquehanna and the Juniata, respectively draining 6847 mi and 3354 mi in Pennsylvania, and the Chemung, which drains 2506 mi in New York and Pennsylva nia (US Army Corps of Engineers, 1999). The Chenango is a somewhat smaller river, but is the largest tributary in the SusquehannaÂ’s Upper Sub-basin; its wate rshed covers 1610 mi (US Army Corps of Engineers, 1969). Figure 4.1 illustrates the 27,50 0 mi extent of the Susquehanna Basin and the locations of its six sub-basins. Basin Physiography and Geology The Susquehanna Basin is part of the A ppalachian Highlands and includes pieces of five physiographic provinces. The vast majo rity of the Basin is contained within the Appalachian Plateau, the Va lley and Ridge and the Pied mont provinces, though very small portions of the Blue Ridge and Coasta l Lowlands provinces are also found within the watershed (Susquehanna River Basi n Study Coordinating Committee, 1970a). Province delineations are shown in Figure 4.2. The remainder of this section will address only the Plateau, Valley and Ridge, and Piedmont Provinces.
50 The Appalachian Plateau Over half of the Susquehanna Basin lies within the Appalachian Plateau (SRBSCC, 1970b). The Plateau is primarily made up of sedimentary rocks including sandstone, shale, and limestone, as well as conglomerates and bituminous coal (SRBSCC, 1970a). Most limestone is found in the western basin. These Devonian layers are saucer shaped (Hunt, 1967), and the forma tion dips slightly, forming a cuesta (Van Diver, 1985). The strata, for the most part, rema in horizontal, but they have been uplifted and dissected, so the formations appear heav ily folded. Folding is actually relatively gentle, but becomes more pronounced as it meet s the Valley and Ridge province. In the western portion of the province, elevati ons are about 1000 feet but in the Basin, elevations reach approximat ely 3000 feet at the Allegheny Front (Hunt, 1967). The Front marks the border between the Appalachian Plateau and the Valley and Ridge Province and the escarpment can be up to 1000 feet high. The northern portion of the province was glaciated during the Wisconsin an (Eisenstadt, 2005). The Valley and Ridge Province Just over a third of the Basin is contai ned by the middle section of the Valley and Ridge Province (SRBSCC, 1970b). The sedimentary la yers of this province were part of a coastal plain and were pushed into folds by multiple mountain building events in the Paleozoic Era (Hunt, 1967; Van Diver, 1985). In the Susquehanna Basin, most formations are sandstone, shale and bituminous coal, though limestone is also present. The Valley and Ridge province is heavily disse cted, with drainage routes forced into more angular patterns than the more dendri tic systems of the Plat eaus (see Figure 4.2).
51 Figure 4.1. Susquehanna Basin and Sub-basins Source: Susquehanna River Basin Commission, 2001. http://www.srbc.net/gis/map_gallery.html
52 Figure 4.2. Physiographic Provinces of the Susquehanna Basin Source: Susquehanna River Basin Commission, 2001. http://www.srbc.net/gis/map_gallery.html
53 The Piedmont Province Unlike most of the formations in the Appalachian Plateaus and Valley and Ridge provinces, those in the Piedmont tend to be heavily metamorphosed. Slate and other lessmetamorphosed or non-metamorphosed rock s can be found, but in the Susquehanna Basin, gneiss and schist are more common. Quartz ite, marble and granite are also present (Hunt, 1967). Below the surface, the Piedmont has characteristics of severely folded mountains, a result of Taconian, Acadian, and Alleghanian mountain building events (Van Diver, 1985), but it resembles a rolli ng plateau; both elev ation and relief are moderate. In Pennsylvania, elevation ranges from one hundred feet to about one thousand feet (Voigt, 1972). Hunt (1967) juxtaposes th e Piedmont and the Plateaus, saying that Â“the Appalachian Plateaus are mountainous with a plateau structure, whereas the Piedmont Province is a low plateau with the kind of structures that generally produce mountainsÂ” (pg. 166). The Piedmont has older, d eeper, and more evenly spread soils than the other provinces of the Susquehanna, maki ng it ideal for farming. Drainage patterns, and thus settlement patterns, are less confined than in th e Valley and Ridge or Plateau provinces. Climate The BasinÂ’s climate is generally humid c ontinental, but there are differences in temperature and precipitation. Orographic pr ecipitation can result from systems coming from the east, south, or west. In the summe r and fall months, tropical systems can move up from the Gulf of Mexico or in from the Atlantic and produce intense rainfall (SRBHOS, 2004). Average precipitation ranges fr om 30 inches to 50 inches per year.
54 Central Pennsylvania tends to get the most rain, while the northwestern portion of the Appalachian Plateau and the Chemung sub-ba sin receive the least (SRBC, 2001). In the southern portion of the Basin, average ma ximum temperatures range from 39.6 degrees Fahrenheit in January to 86.7 degrees in July. Seasonal snowfall is approximately 20 inches (National Weather Service, 2006c). At Cooperstown, the January high is 30.5 degrees and the July high is 79.7 degrees. Coope rstown receives over four times as much snow as southern Pennsylvania (Nat ional Weather Service, 2006a). Vegetation Like most of the eastern United States the Susquehanna Basin was at one time covered by a continuous deciduous forest (Yahner, 2000). By, 1900, however, only 30 percent of the Basin was forested (SRBHOS, 2004). New York and northern Pennsylvania were tapped for timber as supplies in New England declined in the first half of the 19th century and most of the lower Sus quehanna watershed was converted to agriculture (Yahner, 2000). Through regula tion, reforestation, and afforestation, approximately two thirds of the Basin is now covered by trees (Boye r, 2002). Forests are most dense in the western Plateaus and the ri dges of the Valley a nd Ridge province, but little of the fertile land in the Piedmont has been let reve rt to forest. About 30 percent of the Basin remains dedicated to agriculture. The eastern deciduous forest contains over 110 species of trees (Yahner, 2000). Broad communities can be identified, but specific species composition varies depending on elevation, slope, soil, etc. Major forest t ypes within the basin in clude the oak-hickory, northern hardwood configurations, as well as a bit of southern oak-pine (Thompson,
55 1966). Cedar and cypress swamps can be f ound at either end (Brubaker, 2002). Red maple is ubiquitous, sugar maple, yellow bi rch, basswood, elm, and northern red oak are common, and hemlock stands can be found in mo ister valleys. Other oaks and the pignut species are found together and do not mix with red spruce, red pine, or aspen (Yahner, 2000). PHYSICAL CONTEXT: THE UPPER SUSQUEHANNA BASIN About 23 percent of the 27,500 mi the Sus quehanna drains are within New York State (SRBSCC, 1970b). These 6300 mi are encompassed by two sub-basins: the Chemung and the Upper Susquehanna. This section will highlight the geology, climate, and vegetative communities of the Upper Susquehanna sub-basin and the Binghamton area in Broome County, NY. Figure 4.3 illust rates their location within the basin. Binghamton is the population ce nter of the Upper Basin, th e seat of Broome County, and developed at the confluence of the main st em Susquehanna and the Chenango River. The Towns of Union and Vestal are pa rt of the greater Binghamton area. The Upper Basin covers approximately 4944 mi. At the USGS gage site near Waverly, upstream of the Chemung, the Susque hannaÂ’s median long-term discharge is 8000 cfs. Like the Susquehanna as a whole, di scharge from the Upper Basin is lowest in August (averaging 2020 cfs) and highest in Ap ril, when it averages 18,500 cfs (USGS, 2006a). The long term median discharge at Conklin, located before the Chenango meets the Susquehanna, is 3160 cfs; the ChenangoÂ’ s median discharge at Chenango Forks is 2100 cfs (USGS, 2006a).
56 Figure 4.3 The Upper Susquehanna Sub-Basin Source: Susquehanna River Basin Commission, 2005. http://www.srbc.net/gis/map_gallery.html Physiography and Geology of the U pper Basin and the Binghamton Area As can be seen in Figure 4.2, the Upper Susquehanna Basin lies entirely within the Appalachian Plateau physiographic provin ce. The river drops approximately 2.5 feet per mile towards the southwes t in this region (USACE, 1969). At Waverly, the riverÂ’s elevation is 744 feet above sea level, while th e gage at Vestal is situated at 821 feet (USGS, 2006a). The surrounding hills rise to between 500 and 800 feet above river level throughout the basin (SRBSCC, 1970a); in th e Binghamton area, elevations can exceed
57 1800 feet (topozone.com). Figure 4.4 illustra tes VestalÂ’s topographical variation and related settlement patterns. The ga ge site is marked by an arrow. Figure 4.4. Vestal Gage Site and Topography Source: USGS The steep sided hills and flat valleys ar e indicators of glaciation, and set the Upper Susquehanna and Chemung basins apart from those lying in the unglaciated plateaus and provinces to the south. The lowlan ds are full of till, ice contact deposits, and outwash gravel and sand. Many of the Uppe r BasinÂ’s populated areas depend on locally thick, connected deposits of coarse glacial ma terial for their water needs (Randall, 1986; Yager, 1986), as the shale that underlies the population centers in the southern basin does Vestal Ga g e
58 not store or transmit enough water for public supply (Yager, 1986). Most till originates from the shale and sandstone that dominate the Upper Susquehanna Basin; the sedimentary formations found in other regions of the Appalachian Plateau (i.e. limestone) are patchy. The Devonian strata of the norther n mountains and platea us were formed by the mud, silt, and sand of the former Catskill Delta (SRBSCC, 1970a; Van Diver, 1985), though the valley fill does contain unrelated ups tream materials and er ratics have been found in many places. Climate Like the Susquehanna Basin as a whole, the climate of the Upper Basin is predominantly humid continental (Thompson, 1966) It is, however, more susceptible to cold air masses coming from the west and north than the Lower Susquehanna. Winters are colder and the sub-basin is somewhat removed from maritime influences. Summers are cooler as well, though similarly wet. Temperatures recorded at the Binghamton Regional Airport range from an average low of 15 degrees Fahrenheit in January to an average high of 78 degrees in July. The growing season is April through September (National Weather Service, 2006a). Rainfall averages just over 38 inches per year (SRBSCC, 1970b; NWS, 2006a), though the northeast basin receives slightly more than the south. Rainfall is rather evenly distributed temporally, peaking in June w ith 3.8 inches (NWS, 2006a). Most rain comes with warm air masses traveling from the Gulf of Mexico, though tr opical systems are not uncommon. An estimated 54 percent of rain fall becomes runoff (SRBSCC, 1970a). The Binghamton area receives the majority of its estimated 81 inches of snow from December
59 to March (NWS, 2006a), contributing to hi gher discharges in March and April as temperatures increase and precipitation shifts to liquid form. Average monthly rainfall and snowfall are included in Figure 4.5. Figure 4.5. Average Monthly Precipitatio n at Binghamton Regional Airport Month12 11 10 9 8 7 6 5 4 3 2 1 Inches of Rain4.00 3.00 2.00 1.00 0.00 Month12 11 10 9 8 7 6 5 4 3 2 1 Inches of Snow25.00 20.00 15.00 10.00 5.00 0.00 Vegetation The Upper Susquehanna Basin falls almo st entirely within U.S. Level III Ecoregion 60, the Northern Appalachian Plateau and Uplands. In this region, low
60 glaciated hills and wide va lleys are covered with hard wood forests and dispersed agriculture. Two vegetative communities dominate the Upper Susquehanna Basin: oakhickory and northern hardwood (Thompson, 1966) The northern hardwood configuration is more prevalent, but ther e is significant ove rlap in tree species (Yahner, 2000), and specific forest compositions do not persist over large areas. According to Kachmor and Goeller (2005), over 90 percent of the Upper Susquehanna Basin had been cleared by the early 20th century. Early settlers cut trees to make room for agriculture and pasture, and the BasinÂ’s forests suffered more when the lumber industry moved from New England to New York in the 19th century. In 1929, the state stepped in and passed the State Reforestation Law, followed by the Hewitt Amendment in 1931. This legislation allowed the state to buy large tracts of land; reforestation began in the 1930Â’s with the help of the Civilian Conservation Corps (Kachmor and Goeller, 2005). While the specie s planted were not those that had been there before, the plantings helped stab ilize the soil, promoted infiltration, and reestablished forests in the physical world as well as in the public consciousness. Peak recovery has passed. Today, appr oximately 36 percent of the basin is devoted to agriculture (Stoe, 1999). About 60 pe rcent is covered with forest, though it is becoming more fragmented as people move out from the Upper BasinÂ’s few populated areas (New York State Division of Envi ronmental Conservation, 2005). The number of people in the region has not increased, but they are taking up and paving over more space, reducing habitat and altering hydrol ogy. Figure 4.4 shows VestalÂ’s expansion to the ridges and slopes surrounding the river vall ey. This upward migration is primarily
61 residential, but many new reta il and commercial sites have been developed at lower elevations (Town of Vestal, 2004). SOCIAL CONTEXT Though the footprints of the populated pl aces within Broome County are growing, the population itself is shrinking. As Figure 4.6 indicates, the county experienced rapid growth through the middle of the 20th century. The population peaked in 1970 at 221,815 and has been declining since. The estimated population in 2005 was 196,947, a 2.1 percent loss from 2000. The to tal population decreased by approximately 5.5 percent from 1990 to 2000. Binghamton, the largest city in Broome County, experienced an 11.8 percent loss over the same period. The populat ion trends reflect the areaÂ’s economic history, which is covered in the next section. Figure 4.6. Population of Broom e County, New York: 1900-2000 Census Years2000 1980 1960 1940 1920 1900 Population250,000 200,000 150,000 100,000 50,000
62 A Brief History of the Study Area Prior to white settlement and conflic t, the Binghamton area was inhabited by tribes belonging to the Lea gue of Iroquois, who drove out or assimilated the Algonquin who had farmed the valleys before them (Smith, 2006). An Oneida village was located at the confluence of the Susquehanna and Chen ango rivers; Mohawks lived to the west (Gordon, 1966). Two additional villages occupied what are currently Castle Gardens and the lower Choconut Valley in Vestal (Sm ith, 2006). In 1768, the Iroquois and the English entered into the Treaty of Fort Stanwix, wh ich granted land to the east and south of the Delaware to whites, and land north and west of the river to the Iroquois. What is now Broome County lies west of the Delaware. In 1784, a second meeting was held at Fort Stanwix, and the Iroquois were forced to gi ve up land in Pennsylvania and New York, including what is now Broome County (Gordon, 1966; Smith, 2006). Settlers and developers moved in shor tly thereafter. In 1785, 2.3 million acres north of the Susquehanna between Chenango and Owego was sold for 12 cents per acre and divided between 60 investors (Mered ith, 1999). In 1786, two land patents were granted to Robert Hooper and William Bingham and James Wilson (Fiori, 1990; Meredith, 1999). Broome CountyÂ’s development began with these patents and the first permanent settler arrived in 1787 (Smith, 2006). Tioga County was formed in 1791 and Broome County split from Tioga in 1806. Th e Town of Union, like Tioga County, was created in 1791 and covered 700 square miles and parts of three cu rrent counties (Fiori, 1990; Smith, 2006). Approximately 600 people oc cupied the Town at the time (Fiori, 1990). Vestal was carved from Union in 1823 an d, when the first census was taken in 1825, consisted of 784 residents (Learner, 1989). In the early years, Broome County was
63 rather isolated and agriculture and timber were the primary economic activities (Fiori, 1990; Avery, 1973). Growth increased in the mid 19th century when canals and railroads made the county more accessible to both goods and people. By 1880, a wide variety of products were made in the Binghamton area (Smith, 2006). Two companies were pivotal to the areaÂ’s 20th century development: Endicott Johnson Shoe Company and IBM. The Lest er Brothers Boot and Shoe Company incorporated in 1890 and a factor y located in what is now Jo hnson City helped shift the focus of Union from farming to industry and commercial ventures (Fiori, 1990; Smith, 2006). Henry Endicott bought the company and retained the superintendent, George Johnson; the two formed the Endicott-Johnson Shoe Company in 1899. The Village of Endicott developed as housing for E-J work ers, as did West Endicott. The company constructed 40,000 homes in 40 years, which it sold at cost to its employees. At its peak in 1951, the Endicott-Johnson Shoe Comp any employed approximately 21,000 people (Smith, 2006). Competition from foreign-made shoes undermined the companyÂ’s success in the late 1950Â’s and it was sold to the McGowan Corporation in 1968, which began letting go of company and associated community assets (under some conditions). The last plant closed in 1995. In 1914, Thomas Watson became president of the International Time Recording Company, the precursor of IBM. The first IBM plant was located in Endicott in 1924 and business increased in the late 1930Â’s and 40Â’s. World War II was bene ficial to all three major employers (E-J Shoes, IBM, Link Aviation). IBM continued to do well after Endicott-Johnson faltered and was the ar eaÂ’s largest employer in 1970 with 14,000 workers. Other companies like Link Aviation, Lockheed Martin, Remington and GE also
64 contributed to economic and population growth through the middle of the century (Fiori, 1990; Smith, 2006). However, area business d eclined in the 1980Â’s and 90Â’s and in 1994, IBM began laying people off, part of the wors t job loss in Broome County history (Smith, 2006). While Vestal was originally called CraneÂ’ s Ferry and was the s ite of a trading post in 1782, the Susquehanna River isolated the Town from the early economic trends in Union (Learner, 1989; Smith, 2006). Vestal de veloped later, remaining predominantly rural until the mid 20th century (Avery, 1973; Smith, 2006), growing with EndicottJohnson and IBM. Some manufacturing came to Vestal in the 1940Â’s, but most companies were gone by the early 80Â’s and only a few remain today (Town of Vestal, 2004; Smith, 2006). In 1960, Harper College, a four year university, moved from Endicott to Vestal and drew bot h students and employees to the area. The college became SUNY Binghamton in 1965. It is now the larg est employer in Broome County and is viewed as a key part of Ve stalÂ’s economic strategy (Town of Vestal, 2004). The townÂ’s population is relatively stable, but unlike much of the area, Vestal has strengthened its retail sector and attracted development, adding 69 comme rcial businesses since 1994 (Town of Vestal, 2004). Neighborhoods As discussed in the previous chapter, da ta collection was concentrated in seven neighborhoods with identified one hundred and/or five hundred year floodplains. Figures 4.7 through 4.11 illustrate the locations of these neighborhoods with regards to the Susquehanna River as well as Nanticoke and Choconut Creeks. The five neighborhoods
65 of unincorporated Union are addressed firs t, followed by the two in Vestal. A brief description of each is included. West Corners and West Endicott West Corners straddles Nanticoke Cr eek and contains both commercial and residential land uses. The land was first se ttled in 1816 or 1817 by Orman West, a farmer, and development was relatively slow (Fiori 1990; Smith, 2006). The biggest residential developments were established in the 1950Â’ s by Endicott-Johnson at the peak of the companyÂ’s employment levels (Fiori, 1990). West Endicott (labeled in Figure 4.7) served as both a residential and manuf acturing base for E-J; the Fa irplay factory operated from 1921 to 1987 (Fiori, 1990). Figure 4.7. West Corners and West Endicott West Endicott
66 Endwell Endwell was the site of the Hooper land claim and was called Hooper until 1921, when it was renamed after an Endicott-John son shoe (Fiori, 1990; Smith, 2006). Endwell is not incorporated, but is th e largest of the areas treate d as a Â“neighborhoodÂ” in this project and spans much of region between E ndicott and Johnson City. For the purposes of this research, IÂ’ve included the River Road area as well, which lies just to the east and is not shaded in Figure 4.8. The Town of Union also links River Road to Endwell (Town of Union, 2006a). While named after a shoe, Endwe llÂ’s growth is primarily associated with IBM rather than Endicott-Johnson; in the 1940Â’s, much of the area remained farmland (Fiori, 1990). Figure 4.8. Endwell
67 Fairmont Park and Westover Fairmont Park is a small residential neighborhood located between Endwell and Johnson City and labeled in Figure 4.9. It di d not develop till the 1920Â’s (Fiori, 1990). GrayÂ’s Creek is located to the east and north east. Much of the region north of Fairmont Park is currently forested, but a developer has bought a portion of the elevated land to the northwest (personal communications). Westove r had seven houses in 1908 (Fiori, 1990), but was one of two planned developments of farmland (the other was Westunder) mentioned in a Broome County plat book of the same year (Smith, 2006). The neighborhood was settled at about the same time as Fairmont Park (Fiori, 1990). In 1942, however, a main road was run through Westover and Remington Rand moved in, providing jobs and boosting other development. Figure 4.9. Fairmont Park and Westover Fairmont Park
68 Twin Orchards/Ideal Terrace Though these two areas cover qui te a bit of space, the Town of Vestal treats them as one neighborhood (Town of Vestal, 2004). Idea l Terrace is located to the west of the area shaded as Twin Orchards in Figure 4.10 and just east of the highway cloverleaf. Also included in this neighborhood is the Eldr edge Drive area west of the cloverleaf and east of Pumphouse Road and the USGS gage. Along the riverfront, residences form smaller clusters than in Twin Orchards its elf. The western edge has some commercial development. Twin Orchards/Ideal Terrace is one of the oldest developments in Vestal, dating to the 1920Â’s and 30Â’s (Town of Vestal, 2004). Figure 4.10. Twin Orchards/Ideal Terrace and the Riverfront
69 Castle Gardens Most of Castle Gardens was a Planned Development District and the housing is both denser and newer than that of Twin Or chards. The average home price is also higher and the population is older (Town of Vestal, 2004). The neighborhood is primarily residential, but some parcel s are devoted to industry. Al so visible in Figure 4.11 are Choconut Creek and Choconut Creek Valley. Th is neighborhood was not included in data collection due to flooding in November, 2006. Figure 4.11. Castle Gardens Current Demographics This project focused on residents of si ngle family homes in the one hundred and five hundred year floodplains of unincorporated Union and Vestal. The most densely Castle Gardens
70 populated areas of Union (Endicott and Johnson City) were not included in the study area in order to maintain equal levels of gove rnment. Populations of unincorporated Union and Vestal were similar in 2000; Vestal had 26,535 people, while 27,725 lived in unincorporated Union. Table 4.1 includes population data and other demographic information for both towns and Broome County. Estimated data pertaining to the one hundred and five hundred year floodplains in each town are also included, as these residents better represent the target p opulation. Census 2000 spatial block group data were areally weighted and summarized in ArcGIS based on FEMA floodplain status. This method assumes equal distribution and the results are not exact, but they provide a better approximation than town data alone. Patterns may have changed somewhat since 2000. Table 4.1. Census 2000 Demographic Data for Unincorporated Union*, Vestal and Broome County Union: 100 and 500 Year Floodplains** Union Vestal: 100 and 500 Year Floodplains Vestal Broome County Population 5,030 27,725 3,950 26,535 200,526 % Female 51.6 51.8 54.0 52.5 51.8 % White, Non-Hisp. 93.3 94.9 94.7 86.0 90.4 % Age 65+ 16.7 17.2 24.8 15.8 16.4 % Bachelors 20.1 28.5 22.7 38.6 22.7 % Own Home 70.3 74.2 81.1 78.7 65.1 Median Income Not Provided $41,628 Not Provided $51,098 $35,347 % Housing Multi-Unit Not Provided 24.1 Not Provided 17.7 33.3 Data for Endicott and J ohnson City are not included ** Results for floodplains were calculated by areally weighti ng and summarizing 2000 Census block group data in ArcGIS based on FEMA floodplain designations
71 Differences were most pronounced between Vestal as a whole and the Vestal floodplain grouping. The percentages of white non-Hispanics and those 65 or over were much higher in the floodplains. A higher pr oportion of Vestal floodpl ain residents than Union floodplain residents were also 65 or older. However, in both Union and Vestal, lower percentages of residents had bachel orÂ’s degrees in the floodplain subset. Key differences between Union and Vestal as a w hole included the lower percentage of white non-Hispanics, higher median income and incr eased rates of higher education in Vestal. All of these patterns likely reflect the influence of Binghamton University. The differences in these variables were le ss pronounced among floodplai n residents, though. HISTORICAL FLOODING IN UNION AND VESTAL In Union and Vestal, development began in the valleys and moved up. In addition to influencing development patterns, the rugged, steep sided hills send precipitation streaming down to the valley floors in small, flashy streams, increasing the chances of high discharges and short lag times. The re gion is susceptible to disturbances and precipitation events related to cold air masse s from the west and north, warm air masses from the Gulf of Mexico, and tropical syst ems from the east and south. Heavy snow is common and the Susquehanna is more prone to ice jams than any ot her river east of the Rockies (USACE, 1999). The rive rÂ’s natural setting and peopl eÂ’s land use choices have created a particularly hazardous environm ent. Figure 4.12 illustra tes the annual peak discharges for the Susquehanna at Vestal for the period of record (1935-2006). Table 4.2 lists the five floods with the highest recorded stages.
72 Figure 4.12. Peak Annual Discharge for the Susquehanna at Vestal: 1935-2006 Table 4.2. Five Largest Floods Record ed by USGS Gage at Vestal Rank Date Stage (feet) Discharge (cfs) 1 6-28-06 33.66 119,000 2 3-18-36 30.50 107,000 3 4-3-05 29.14 97,000 4 1-20-96 27.86 89,100 5 3-22-48 27.73 92,400 Three of the five largest floods in Union and Vestal occurred within a recent span of just over ten years. In 1996, the Blizzard of Â‘96 added to previous snow pack, resulting in the equivalent of four inches of liquid wa ter. Temperatures rose to the 60Â’s and rain followed (SRBC, 2006). At the time, it was th e biggest flood since the record flood of
73 1936; parts of Endwell and Fairmont Park evacuated and suffered damage (Town of Union, 2006a). However, river levels rose even higher in April of 2005, when intense rain and melting snow resulted in extensive damage in the Delaware and Susquehanna basins. Parts of Castle Gardens and Fairm ont Park flooded and Endwell was hit hard (Town of Vestal, 2004; Town of Union, 2006a). Some Endwell resi dents were waiting for buyouts to go through when they were fl ooded again in June of 2006. Additionally, in 2004, remnants of hurricane Ivan caused heavy rainfall and high discharges in surrounding areas. The event ranks fourth larges t at a gage upstream at Conklin (USGS, 2006), but impact was generally less severe in Vestal and Un ion, though flooding did occur in portions of Endwell and a few other areas. Though not included in the top five or ev en top ten events at Vestal, floods associated with Hurricane Agnes are an impor tant reference point for the region. Many of the people I talked with duri ng data collection sp oke of it. In June of 1972, water raged through the Delaware, the Susquehanna, the Potomac, Rappahannock, and the James Rivers, killing 117 and causing 3.1 billion dollars in damage. Every county in Pennsylvania was declared a disa ster area. It was the most co stly disaster in US history until Andrew hit Florida in 1992. The death toll and monetary losses were highest in the Susquehanna Basin, where 72 people were repo rted dead and damages totaled 2.8 billion dollars (NOAA, 2006). In most of the Upper Susquehanna Basi n, precipitation over the five day period associated with Agnes ranged from two to six inches and antecedent moisture levels were moderate. As a result, flood crests around Bi nghamton were considerably lower than the record stages set throughout the Susque hanna Basin in 1936 (USGS, 2006a). Until
74 Agnes, the St. PatrickÂ’s Day Flood of 1936 was the flood to which all others were compared and upon which all structural para meters were based (USACE, 1969). In the Upper Susquehanna, that mind-set persisted until 2006. JUNE, 2006 FLOODS Hydrology Like the floods of 1972, the 2006 floods o ccurred in June and were related to tropical moisture, but not in th e form of a remnant hurricane. A developing system to the southeast and a stalled cold front to the west resulted in heav y rains over eastern Pennsylvania and central New York (NWS 2006b). The Susquehanna and Delaware Basins were primarily affected from June 26th to June 28th. Over the course of the week, 8 to 15 inches fell over the upper basin (Zampogna, 2006). Most rainfall was concentrated within a much shorter period of time and fell on ground already saturated by a month of record rain (NCDC, 2006). These conditions led to a flashy response that gave residents w ho did not receive warn ings (NWS flash flood watches began on the 26th), or did not take warnings seri ously, little time to evacuate (see Figure 4.13). The river was below five feet on June 26th. Figure 4.14 illustrates the 24 hour precipita tion totals ranging from about a quarter inch to seven and eight inches. Unlike Agnes, the highest totals were amassed in the Upper Susquehanna Basin. Very little rainfa ll occurred in the western basins and the lower Susquehanna was able to absorb the excess water with only minor to moderate flooding (Zampogna, 2006). Below the Chemung flooding was generally moderate (SRBC, 2007), though thousands in Wilkes-Barre were ordered to evacuate, and did so,
75 not wanting to see first-hand wh ether the upgraded levees (a result of Agnes) held or not. Of the ten gages in the Upper Susquehanna Basin, eight measured flood stages higher than the past record and bigger than a hundr ed year flood; three recorded discharges estimated to have a recurren ce interval of five hundred ye ars or more (USGS, 2006b). In areas heavily affected by Agnes, records generally stood. Figure 4.13. The Susquehanna at Vestal, NY: June 26th to July 3rd, 2006 Source: National Weather Service Forecast Office, Binghamton, NY. http://www.erh.noaa.gov/bgm/Weath erEvents/Flood/june2006/mpe.shtml On the hydrograph in Figure 4.13, the yellow li ne represents a warning stage, the red line bankfull, the blue line moderate flooding, and the purple major flooding. Major flooding is usually considered to be a flood with a one percen t chance or less of occurring
76 in any year. For the Susquehanna at Vestal, flood stage is 18 feet. At its peak, the river was estimated to have exceeded flood stage by 15 feet. The flat line on the hydrograph indicates that the gage was overtopped and stopped working. Figure 7.14. 24 Hour Precipitation in Northern PA and Southern NY: June 27th-28th, 2006 Source: National Weather Service Forecast Office, Binghamton, NY. http://www.erh.noaa.gov/bgm/Weath erEvents/Flood/june2006/mpe.shtml Major flooding would have been more wi despread had forecastersÂ’ original estimations of the storm track been correc t; the storm ended up 50 to 75 miles east of where it had been expected (Zampogna, 2006). Because most of the Susquehanna Basin lies west of the main stem, this shift likel y reduced flood heights, subsequent damage, and the areal extent of the disaster dram atically. Even Â‘moderateÂ’ flooding can be
77 damaging, though. FEMA paid almost $220 million in claims (FEMA, 2007b) and a total of 25 counties in Pennsylvania and 12 counties in New York were d eclared eligible for some form of disaster assistance (F EMA, 2006b). Broome County received both individual and public assistance. Impacts Figure 4.15 illustrates the areal extent of the June, 2006 floods. Depth, which would be a better indicator of relative impact, is not represented. The image is a screen capture of parcel data overlain with a flood layer in Broome CountyÂ’s excellent interactive mapping site (www.broomegis.c o.broome.ny.us). The flood extent layer was created by New York State Electric and Gas (NYSEG) based on data collected at peak flood stage. The data and representation are rough, but are th e best available approximation of flood coverage. Castle Garden s is represented as completely inundated, though I spoke with several people in th at neighborhood who had no water in or threatening their homes. Egress was cut o ff, however, and they were evacuated. Figure 4.15 represents a general indicati on of impact; most neighborhoods used in this project were at least partially floode d. Figure 4.16 outlines both the one hundred year floodplain (light blue) and the five hundred year floodplain (dark bl ue) and is based on FEMA Q3 data. A comparison of the two imag es shows that flooding occurred in parts of the five hundred year floodplai n, but left some areas in th e one hundred year flood zone relatively dry. Flood Insurance Rate Maps ar e being updated (some date to the 1980Â’s), but even current maps are approximations Neighborhood impacts and flood control structures in unincorpor ated Union and Vestal are described below.
78 Figure 4.15. Areal Extent of Flooding in June, 2006
79 Figure 4.16. FEMA 100 and 500 Year Floodplains
80 West Corners and West Endicott Figure 4.17, part of a set of neighborhood sp ecific brochures sent out to hundred year floodplain residents, illustrates the Sp ecial Flood Hazard Area and levees in West Corners. These brochures were sent out in November, 2006 and are part of an outreach program associated with Uni onÂ’s participation in the Co mmunity Rating System (Town of Union, 2006b). They can be found on Uni onÂ’s website, which contains extensive planning information (www.townofunion.com). Though it sits behind a levee built in th e 1980Â’s, the eastern portion of West Corners was inundated in June and some individuals were st ill living in FEMA trailers during field work. Regardless, this area is c onsidered protected and is not part of the SFHA (Town of Union, 2006a). There is no leve e on the western side of West Corners and homes bordering the Nanticoke also suffere d damage. The southern levee depicted is on the southeast side of the creek and extends past West Endicott. This levee was not overtopped, but the backflow filled up and se veral structures on Frey Avenue were impacted. Endwell Endwell is the most frequently flooded neighborhood in the study area (Town of Union, 2006a). Figure 4.18 depicts the SFHA for southeast and southw est Endwell. There are no levees because it was not considered co st effective to build them (Fiori, 1990). The neighborhood has several repe titive loss properties and Union instituted the River Rd/Argonne Ave buyout program in 1988 (Town of Union, 2006a). A total of approximately eight acres were bought th rough the program by 1993. The Town is trying
81 to access funding for 44 additional homes; FEMA has approved funds for 20 structures damaged in 2005 (Town of Union, 2006b). Many of the houses on Argonne, River Road, Shady Lane, Verdun and Fairmont Avenue remained uninhabited during field work. While homes on Riverview were still undergoi ng repairs, all houses were occupied, perhaps a reflection of differences in resources rather than impact. Figure 4.17. West Corners and West Endicott Flood Control
82 Figure 4.18. Endwell Hundred Year Floodplain
83 Fairmont Park Fairmont Park also suffered severe impact s, though it is somewhat removed from the Susquehanna. GrayÂ’s Creek was part of the problem, but much of the flooding was due to high water levels on the river preventi ng drainage (Town of Union, 2006a). Levees and gates were built in the 1980Â’s and the to wn is once again applying for funding for a formerly approved project that was never im plemented. Figure 4.19 s hows the location of Fairmont Park flood control structures. Floodi ng in June of 2006 extended to near Oak Street and structures on Birch, Woodland, Poplar and Barton were uninhabited for at least part of field work. Some remained abandoned throughout. Figure 4.19. Fairmont Park Flood Control
84 Westover Figure 4.20 illustrates WestoverÂ’s flood control structures, completed in 1958 and 1960 (Town of Union, 2006a; 2006b). The neig hborhood is bordered by both Little Choconut Creek and the Susquehanna, but fl ooding was limited in June, 2006. However, some residents on Onondaga did suffer signifi cant basement and first floor damage. Others in the area had minor water damage. Figure 4.20. Westover Flood Control Twin Orchards/Ideal Terrace Sections of Twin Orchar ds are bordered by levees. The levees do not, however, run the length of the neighborhood to the gage at Pumphouse Road. Additionally, the area
85 suffers from backflow effects. Damage was extremely varied, even within the larger contiguous section shaded in Figure 4.10. The heav iest impacts appeared to be in the Jane and River Rd area, though resi dents of Old Vestal Rd a nd parts of Pearl also had substantial damage. Residents in both the SFHA and the five hundred year floodplain in this neighborhood were asked to evacuate by officials. Castle Gardens Castle Gardens has no flood control st ructures and this higher density neighborhood was developed before a 1986 or dinance limiting building in the hundred floodplain (Town of Vestal, 2004). The north en d of Meadow Lane and portions of North Road flood relatively often. In June of 2006, all of Meadow Lane and almost all the homes in the western section of Castle Gardens had some water damage. Northwest Loretta and Vivian are situat ed on a little bluff and houses located there had no damage, though the whole area was evacuated. Much of North Road and a few homes on Crest, Westview and Greenlawn remained boarded up through January. The Town of Vestal has erected signs delineating hundred year flood depths and extent in Castle Gardens si nce the June, 2006 flood. In the TownÂ’s comprehensive plan, references to flooding emphasize improving th e accuracy of floodplain boundaries as a means to reduce or remove Â“costlyÂ” insuran ce requirements. The plan also appears to equate these boundaries with Â“potentially catastrophic floodingÂ” and encourages better mapping in order to reduce the uncertainty in future waterfront development plans (Town of Vestal, 2004). Maps were updated in 1998 and are again being reworked. The language of the comprehensive plan and the focus on map lines and regulatory
86 boundaries is reflective of the levee effect and, when combin ed with the (understandable) emphasis on development, may facilitate an increase rather than decrease in future damage. However, wetland preservation was al so mentioned as a means to reduce flood impacts and official outlooks may ha ve changed since the 2006 floods.
87 CHAPTER 5: DESCRIPTIVE RESULTS The bulk of the data analyzed for this project were collected by means of a structured questionnaire; questionnaires were completed in each of the seven neighborhoods described in Chapter 4. The que stionnaire was divided into five major themes: 1) flood experience and loss mitigati on activities, 2) general perception of flood risk and cause, 3) flood information infrastructu re, 4) perceptions asso ciated with specific flood risk descriptions, and 5) basic demographic data. For consistency of presentation, however, descriptive results are grouped based on the situational and cognitive factors outlined in Figures 2.1 and 2.2: location, soci o-economic factors, experience, flood risk information infrastructure, and cognitive fact ors. Outcome factors include understanding of flood related uncertainty, threat percepti on, and general and even t specific mitigative behavior. Data presented are in raw form; altera tions made for the sake of analysis will be explained in Chapter 6. Qualitative data co llected during the face to face surveys and focus groups are included in bot h Chapter 5 and Chapter 6. SITUATIONAL AND COGNITIVE FACTORS Location Three variables related to location were used in th is project: town, floodplain status, and distance from nearest major wa terway. Respondents were assigned hundred
88 year floodplain status if half or more of their land parcel fell within the Special Flood Hazard Area (SFHA) specified in FEMA Q3 data, a digital form of the AgencyÂ’s Flood Insurance Rate Maps (FIRMs). Participants w hose parcels were less than half covered by the floodplain were also assigne d hundred year floodplain status if their house fell within the SFHA. This determination was made by personal observation of the property layout. Distance was measured in two dimensions from a parcelÂ’s center to either the Susquehanna or Nanticoke Creek. Location resu lts are presented in Tables 5.1 and 5.2. Table 5.1. Respondents in Floodplain by Town Union (N=60) Vestal (N=54) # % # % In 100 Year Floodplain (N=50) 26 43.3 24 44.4 In 500 Year Floodplain (N=64) 34 56.7 30 55.6 Table 5.2. Distance in Feet from Major Waterway All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 33 33 268 33 268 Maximum 2727 1860 2727 1860 2727 Mean 906.2 798.7 1025.7 831.9 964.2 Std Deviation 526.3 490.1 543.6 507.2 537.4 The sampleÂ’s floodplain composition by town was fairly consistent. The Vestal group is the least spatially representative, as Choconut Creek residents were not contacted after the November floods. Vest alÂ’s mapped residential flood risk is concentrated in the two areas sampled, howev er. The combined effect of levees and
89 regulation can be seen in the distance resu lts; UnionÂ’s largest di stance occurred in the hundred year floodplain of Fairmont Par k. Two neighborhoods of West Corners are equally close to Nanticoke Creek, but a levee exempted much of the area east of Highway 26 from regulation and turned it into what one focus group participant consistently referred to as a Â“Safe Zone.Â” Her home wa s flooded in June a nd she was living in a FEMA trailer at the time of the meeting. A si milar result is evident in Vestal, where the parcel closest to the river was in the five hundred year floodplain. Socio-Economic Factors In hazards literature, several socio-econo mic variables have been linked to risk perception and behavior. For this project, data were collected on gender, race and ethnicity, age, education, inco me, home ownership, and length of residence. Tables 5.3 through 5.7 include demographic information for the sample. Table 5.3. Gender, Race/Ethnicity and Ownership All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Female 64 56.137 61.7 27 50.0 27 54.0 37 57.8 White, Non-Lat. 110 96.559 98.3 51 94.4 50 100.0 60 93.8 Own Home 108 94.756 93.3 52 96.3 47 94.0 61 95.3 Women made up fifty percent or more of the sample in all spatial categories. The Vestal percentage is the only one that was lo wer than the areally weighted estimate given in Chapter 4. Women were most overrepr esented in Union, which had the lowest estimated percentage of women of any sp atial category (51.6). Only four survey
90 respondents were not white and non-Lati no. One minority household living in the hundred year floodplain was not asked to co mplete the survey because the English speaking member of the family was not yet 18; another man refused. The racial and ethnic make-up of those contacted was not much different than the final sample. Nor was the sample much different than the estim ated composition of the floodplain population, which ranged from 93.3 to 94.7 percent white/non-Latino. Most survey respondents owned, rather than rented, their homes. Sample percentages were much higher than the owne rship rates calculated using census block groups. However, these ownership rates include d apartment renters, while only residents of single family homes participated in th is research. In Union, approximately 24.1 percent of housing stock is multi-unit (US Cens us Bureau, 2006). Vestal has fewer multiunit buildings (17.7 percent). Home owners ma y have been somewhat overrepresented in the sample, but the ownership rate of the ta rget population was lik ely much higher than the ownership percentages listed in Chapter 4. Length of residence was determined re lative to a single date: October 1st, 2006. One person moved into his home the month af ter the floods (fully aware of what had occurred); one unlucky man moved in less than two weeks before the flood and had not yet finished unpacking. Most had lived in th eir homes for much longer, however, and shorter term residents had of ten moved from another part of Broome County. The median length of residence for the total sample was approximately 15 years. The mean was just under 20 years. Many had worked for Endicott-Johnson or IBM and had been in the same house for most of their adult lives. Others had taken over the home of a parent.
91 Table 5.4. Property Residence Time in Years All (N=113) Union (N=59) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 0.25 0.25 0.33 0.25 1 Maximum 55 50 55 50 55 Mean 19.1 19.7 18.4 17.9 20.1 Std Deviation 16.2 15.9 16.6 16.0 16.5 The sampleÂ’s mean age was ten to 15 years higher than the central age calculated for the study area. That estimation in cluded individuals under 18, though. Survey participants tended to be ol der and retired. There were mo re respondents in their 60Â’s than in any other decadal block and 71 percent were fifty years of age or older. This is a common problem in survey research and may skew results somewhat. Table 5.5. Age All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 19 19 20 25 19 Maximum 86 81 86 86 81 Mean 54.9 55.1 54.7 55.0 54.9 Std Deviation 16.3 16.1 16.7 15.6 17.0 The final demographic variables for whic h data were collected were education and household income (Tables 5.6 and 5.7). Income is considered by some to be sensitive information and 14 people chose not to give th eir income level. Two people did not give an education level. Survey participants were generally better educated than the target population and approximately 98 percent of respondents had their High School diploma or equivalent. According to spatially wei ghted Census data, th e percentage of the
92 population over 25 estimated to have a bach elorÂ’s degree or hi gher ranged from 20.1 percent in Union to 22.7 percent in Vestal. Th e percentage of part icipants who reported completing a bachelorÂ’s degree was higher in th ree out of four spatia l subsets as well as the total sample. Union was the only grouping in which the sample rate (16.7 percent) was lower than the estimated percentage of the population. In Vestal, almost 40 percent of respondents had completed higher education. Table 5.6. Completed Education All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % 12th or Less 2 1.8 2 3.3 0 0 1 2.0 1 1.6 H.S. Diploma or GED 24 21.116 26.7 8 14.8 13 26.0 11 17.2 Some College 55 48.232 53.3 23 42.6 22 44.0 33 51.6 BachelorÂ’s 14 12.33 5.0 11 20.4 5 10.0 9 14.1 Graduate 17 14.97 11.7 10 18.5 7 14.0 10 15.6 No Answer 2 1.8 0 0 1 3.7 2 4.0 0 0 Table 5.7. Household Income Levels All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Under 20K 5 4.4 3 5.0 2 3.7 1 2.0 4 6.3 20-35K 35 30.719 31.7 16 29.6 15 30.0 20 31.3 35-50K 22 19.311 18.3 11 20.4 13 26.0 9 14.1 50-65K 13 11.46 10.0 7 13.0 5 10.0 8 12.5 65-80K 16 14.06 10.0 10 18.5 7 14.0 9 14.1 80-100K 3 2.6 2 3.3 1 1.9 2 4.0 1 1.6 100K+ 6 5.3 4 6.7 2 3.7 2 4.0 4 6.3 No Answer 14 12.39 15.0 5 9.3 5 10.0 9 14.1
93 Based on calculations using Census Bureau data, median household income level for unincorporated Union was about $41,600 in 2003. VestalÂ’s median household income was approximately $51,100. Over 95 percent of survey participants made more than the poverty level for a family of four and in all spatial categories, the greatest proportion of respondents indicated that their household income ranged from 20,000 to 35,000 dollars per year. Because the sample group was mostly over fifty, this result was not surprising; many were living on pensions and income is not necessarily a measure of wealth. When combined with family size and age structure, however, it may be an indicator of the ability to refill a savings account emptied by disaster related expenses. There was not a great deal of difference in income distribu tion across spatial groups and the medians of all groups fell in the $35,000 to $50,000 category. The sample appeared to have had more formal education than the target pop ulation, but similar income levels. Experience The first question survey participants we re asked was if they had ever been affected by flooding. If the answer was yes, th ey were requested to think back to the worst flood theyÂ’d been affected by and desc ribe how it affected them. The question was open ended and answers were used to create a five point scale of im pact severity. Most, but not all, described the impacts of the J une, 2006 floods. Results are included in Table 5.8. About 15 percent of the tota l sample reported never havi ng been affected by a flood. Those that had been affected were also as ked how many times thei r home or property had flooded in their lifetime. This question measured frequency of experience and included
94 floods that affected past residences. Table 5.9 contains information pertaining to frequency. The classification of impact severity was based on flood level, prior use of flooded space, evacuation, and work impacts. No reported impact was coded as zero. The next level consisted of individuals who had evacuated and/or had less than one foot of water in an unfinished basement. Impact was categorized as medium if there was more than one foot in an unfinished basement or under a foot in a fi nished basement. Two participants working in the medical field were heavily impacted by the closing of Lourdes hospital; these indivi duals were also assigned a two. Impact was considered Â“HighÂ” if flood waters were more than one f oot in a finished basement or less than one foot on the first floor. Partic ipants assigned a four had more than one foot of water in their primary living area. This scale, and th is research, does not include residents who experienced the most severe damage. Many homes in parts of Castle Gardens, Endwell, and Fairmont Park were still empty in Janua ry of 2007, with debris lines still visible under the eaves. The few who had had this sort of impact and were kind enough to talk were included in the highest category. Table 5.8. Severity of Impact All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % No Reported Impact 17 14.914 23.3 3 5.6 1 2.0 16 25.0 Low Impact 28 24.69 15.0 19 35.2 3 6.0 25 39.1 Medium Impact 25 21.912 20.0 13 24.1 10 20.0 15 23.4 High Impact 22 19.39 15.0 13 24.1 16 32.0 6 9.4 Extreme Impact 22 19.316 26.7 6 11.1 20 40.0 2 3.1
95 Approximately forty percent of partic ipants in the hundred year floodplain experienced what I have called Â“ExtremeÂ” im pact. Almost three quarters of respondents experienced significant damage to their livi ng space. They still have their homes, however, and these terms are relative. Only three Vestal residents reported no impact, though over four times that many in Union were unaffected. Most of the discrepancy can be explained through differences in evacuati on patterns. The majority of five hundred year floodplain residents in Castle Gardens a nd Twin Orchards were ordered to evacuate, and did so. Official evacuati on in Union was less sweeping and was concentrated in the hundred year floodplain. In Table 5.9, frequency of flooding is pr esented categorically, though the variable was measured and analyzed using a continuous scale. The largest number of experienced floods was eight and only six people had lived through three or more. The effect of the June, 2006 floods can be seen in the hundred year floodplain groupi ng; over 90 percent had been flooded, but two thirds of those people had only been flooded once. Over four fifths of individuals who had been flooded more than once lived in Union. All who had been flooded more than twice lived there. E ndwell has been especially flood prone (see Chapter 4). It is also important to note that floods, even multiple floods, were not confined to the Â“High RiskÂ” hundred year floodplain; half the five hundred year floodplain residents had experienced at least one. It is unlikely th at all those floods occurred in another loca tion altogether, given the long residence time of many respondents.
96 Table 5.9. Total Number of Times Home or Property Flooded All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Never Flooded 36 31.617 28.3 19 35.2 4 8.0 32 50.0 One Time 53 46.522 36.7 31 57.4 30 60.0 23 35.9 Two or More Times 25 21.921 35.0 4 7.4 16 32.0 9 14.1 Flood Risk Information Infrastructure The questionnaire was used to gather da ta regarding flood information sources people used or came into contact with, types of flood information looked for or received, the frequency of exposure to flood informati on, and the credibility of flood information sources. Taken together, these factors make up an individual Â’s flood risk infrastructure. Source and type data were collected us ing open ended questions. Frequency and credibility questions were closed. In this section of the su rvey, the June, 2006 floods were used as a reference point. Participants were asked to name sources th ey went to for flood information, sources which provided them information, and the type of information sought or received. This set of questions was asked with reference to Â“during the June floodsÂ” and Â“since the June floods.Â” RespondentsÂ’ conceptualizations of Â“duringÂ” were inconsistent; for several individuals (mostly those who had suffered severe damage), Â“during the floodsÂ” appeared to include all information exchanges related to the event and its impacts, no matter what the timeframe. Additionally, other participants broke the Â“since the floodÂ” questions into two different qualitative time frames: Â“Well, I watched TV a lot right after it happened, but not for a while now.Â” When these patterns became evident, if a person answered that,
97 for instance, he looked online for information on river levels, the inte rviewer asked if he continued to do so. In order to reconcile these varying frames, a third time category was created during data entry. Th e length of time encompassed was somewhat fluid and was one reason that frequency data were not used in further analyses. During Event The sources and types of information pa rticipants searched for and received during the event are listed in Tables 5.10 a nd 5.11. Only eight percent said they did not get information of any kind. The local govern ments provided the greatest proportion of people with information in the total samp le, Vestal and the h undred year floodplain. Almost three quarters of Vestal residents had some contact with the local government. Most contact in all spatial categories consiste d of fire fighters and police officers passing out warning leaflets the evening prior to the flood or conducting evacuations. Some did call local authorities (two people in Union reported no answer) looking for information on berm breaks and shelter locations. Union re sidents had much more limited interaction with local officials than did participants from Vestal. Most of the diffe rence is a result of disparities in pre-event information dispersal and a less wholesale approach to evacuation. Television was the most prevalent source of information in Union and the five hundred year floodplain, though viewership wa s over fifty percent in every grouping but Vestal. People were looking for information on projected flood levels, and as the waters rose, on which areas were sustaining damage whether their own home was safe, and where to go for help. Many were not watchi ng in their own homes, though, having been
98 evacuated or without power. Not everyone was satisfied with the early coverage, however. Table 5.10. Information Sources Searched and Received During Flood All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # %* # % # % # % # % Local Government 68 59.628 46.7 40 74.1 36 72.1 32 50.0 TV 64 56.140 66.7 24 44.4 26 52.0 38 59.4 Friends 53 46.527 45.0 26 48.1 20 40.0 33 51.6 Newspaper 38 33.325 41.6 13 24.1 14 28.0 24 37.5 Family 31 27.218 30.0 13 24.1 18 36.0 13 20.3 Internet 19 16.79 15.0 10 18.5 8 16.0 11 17.2 Radio 19 16.710 16.7 9 16.7 12 24.0 7 10.9 Relief Organization 3 2.6 1 1.7 2 3.7 0 0 3 4.7 None Received or Searched 9 7.9 5 8.3 4 7.4 4 8.0 5 7.8 Table 5.11. Information Types Searched and Received During Flood All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # %* # % # % # % # % General News 72 63.240 66.7 32 59.3 29 58.0 43 67.2 Flood Levels 64 56.135 58.3 29 53.7 26 52.0 38 59.4 Official Evac Order 53 46.517 28.3 36 66.7 29 58.0 24 37.5 Evacuation Warning 28 24.610 16.7 18 33.3 15 30.0 13 20.3 Personal Experience 18 15.810 16.7 8 14.8 6 12.0 12 18.8 What to Do/ Where to Go for Help 17 14.99 15.0 8 14.8 8 16.0 9 14.1 Other Type 11 9.6 6 10.0 5 9.3 8 16.0 3 4.7
99 Friends were also an important source of information, as those who had evacuated called remaining neighbors on cell phones to ch eck on their homes. Others looked to one another for advice and support, as did family members. Groups gathered to discuss the rising waters and government activities and fa ilures, or to check the levels behind the levees and report back. Friends and family ga ve a few residents their first hint that something might be wrong. Brothers and friends in the Midwest or Florida called to ask, Â“Have you evacuated? Is everything okay?Â” Th ey were answered by a confused Â“Why would we evacuate? What are you talking about?Â” In addition to the television and, later, the paper, the radio and internet were used to gather general news. The internet and newspaper were also searched for contact numbers and aid sites. Those whose homes a nd services were unaffected and those who evacuated to hotels or homes accessed real time water levels on the internet, at least until the Vestal gage was overtopped. On average, the sample group used three sources for two types of information. After Impact The type of information sought changed somewhat after the waters receded, but the scope remained limited. The variety of s ources people used to gather information, however, increased. Tables 5.12 and 5.13 include the results. On average, more sources were used (five), especially by hundred year floodplain residents, though again, only two types of information were sought or receive d. Most participants who were impacted contacted several organizations and friends or family for help.
100 Table 5.12. Information Sources Searched and Received After Impact All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # %* # % # % # % # % FEMA 79 69.341 68.3 38 70.4 47 94.0 32 50.0 Newspaper 64 56.136 60.0 28 51.9 33 66.0 31 48.4 Friends 60 52.636 60.0 24 44.4 31 62.0 29 45.3 TV 60 52.632 53.3 28 51.9 30 60.0 30 46.9 New York State 57 50.028 46.7 29 53.7 38 76.0 19 29.7 Local Gov. 53 46.526 43.3 27 50.0 32 64.0 21 32.8 Relief Org. 55 48.229 48.3 26 48.1 32 64.0 23 35.9 Insurance Companies 27 23.716 26.7 11 20.4 25 50.0 2 3.1 Internet 27 23.714 23.3 13 24.1 14 28.0 13 20.3 Family 25 21.916 26.7 9 16.7 13 26.0 12 18.8 SBA 12 10.56 10.0 6 11.1 9 18.0 3 4.7 Mental Health Organization 9 7.9 3 5.0 6 11.1 5 10.0 4 6.3 None Searched or Received 12 10.54 6.7 8 14.8 0 0 12 18.8 Table 5.13. Information Types Searched and Received after Impact All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # %* # % # % # % # % Help 79 69.341 68.3 38 70.4 47 94.0 32 50.0 General News 67 58.838 63.3 29 53.7 32 64.0 35 54.7 Personal Impacts 44 38.625 41.7 19 35.2 21 42.0 23 35.9 Insurance 29 25.417 28.3 12 22.2 25 50.0 4 6.3 Meetings 27 23.74 6.7 23 42.6 18 36.0 9 14.1 Water Level 12 10.56 10.0 6 11.1 8 16.0 4 6.3 Local Involvement or Plans 9 7.9 7 11.7 2 3.7 4 8.0 5 7.8 Other 8 7.0 4 6.7 4 7.4 3 6.0 5 7.8 Everyone looking for help contacted, or was contacted by, FEMA. Fewer reported accessing information from the state. Though 69 percent of the total sample was looking
101 for help, only 11 percent of the whole group and 18 percent of those in hundred year floodplain said they dealt with the SBA. Th e SBA was not overtly listed on the card given to participants and I be lieve that contact with the organization was underreported. Most (but certainly not all) people looking for financial a ssistance went to aid stations housing multiple agencies and organizations. Additionally, as one focus group participant said, the SBA loan Â“seemed to be the kingpin of this whole thing. If you did not apply for an SBA loan, nothing happened. If you did appl y, even if you didnÂ’t want to take it, things started to happen.Â” He was not alone in this sentiment and was directed to the SBA counter by a FEMA representative. FEMA, the state, the Red Cross, church groups and volunteers also went to affected areas with information flyers, direc tions, cleaning kits and instructions, as well as food, clothing and offers to help victim s muck out their homes. Mental health professionals targeted the elderly. Friends a nd family checked on each other and provided assistance. Television and the newspaper provid ed general news to all spatial subgroups, but those who had been impacted also looked to them for information on where to get help, and how. These sources often referred readers and viewers to websites, and the internet was used to find general news and help information, to download forms, as well as to check water levels. Over 69 percent of participants were l ooking for help and almost three quarters had some damage to their living spaces. Less than a quarter of the total sample and only half of the hundred year floodpl ain residents reported contact with an insurance agent, however. This seems very low, but prior to the flood, only 32 percent of the whole group said they had insurance. Approximately 57 percent of those in the hundred year
102 floodplain reported the same. A few policy holders had baseme nt damage, but did not call because they knew it was not covered. Insu rance company contact was probably not under reported by much. In the period shortly after impact, residents of Union and Vestal had similar rates of interaction with the local government. Similar per centages from both groups sought information and both groups of respondents wa nted information about assistance. More than twice as many residents of Vestal were contacted by the government about meetings and help, however. Notice of the daily town meetings was also spread by word of mouth among neighbors. Most exchanges in Vestal appeared to be informational. In Union, there was also heated discussion among nei ghbors and with the government about pump stations not being operational, the reasons behind the seemi ngly quick release of water on the Nanticoke without warning, and government involvement in the (perceived to be) intentional break in a berm in Fairmont Pa rk. Only seven respondents specifically said they contacted the government or talked to fr iends about these issues but a larger number mentioned these things in response to othe r questions. The first two items were also brought up in focus group discussions. Post Response Period As time went on, consumption of flood related information lessened. About 37 percent of the total group ha d neither searched for nor received information since the period they considered the aftermath of the event (see Table 5.14). The average number of both sources and information types wa s one. The types of information sought expanded a bit as the pressing need for help was addressed (Table 5.15). Much of the
103 information obtained through the mass medi a was acquired passively, however, and included general updates, deadlines and information on the b uyout program. After general news, insurance was the most co mmon information type. Part of the group seeking this information was self motivated, but others were pushed into it as a result of conditions associated with their aid packag es. Weather information was both sought and received. Those in the hundred year floodplain paid it particular atte ntion and related it directly to potential flooding. Fewer people reported contact with family and friends during this period; most were checking on othersÂ’ recovery progre ss. Similar percenta ges of respondents continued to monitor river levels on the inte rnet and in the paper as had before. Two consistent community trends were that Union residents appeared to rely on the newspaper more than those in Vestal and that in the time since the event, Vestal had a greater proportion of residents that did not repor t accessing any flood related information. Table 5.14. Sources Searched and Received since Response All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Newspaper 34 29.824 40.0 10 18.5 14 28.0 20 31.3 TV 34 29.819 31.7 15 27.8 16 32.0 18 28.1 Local Gov. 17 14.914 23.3 3 5.6 11 22.0 6 11.1 Insurance Agent 15 13.27 11.7 8 14.8 7 14.0 8 12.5 Family/Friends 13 11.49 15.0 4 7.4 4 8.0 9 14.1 Internet 13 11.48 13.3 5 9.3 8 16.0 5 7.8 Real Estate Agent 3 2.6 1 1.7 2 3.7 2 4.0 1 1.6 FEMA 1 .9 0 0 1 1.9 1 2.0 0 0 None Searched or Received 42 36.819 31.7 23 42.6 16 32.0 26 40.6
104 Table 5.15. Information Types Searched and Received since Response All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % General News 41 36.025 41.7 16 29.6 14 28.0 27 42.2 Insurance 15 13.27 11.7 8 14.8 7 14.0 8 12.5 Weather 15 13.29 15.0 6 11.1 13 26.0 2 3.1 Water Levels 13 11.46 10.0 7 13.0 7 14.0 6 9.4 Personal Progress 10 8.8 7 11.7 3 5.6 4 8.0 6 9.4 Town Plans/ Involvement 9 7.9 7 11.7 2 3.7 7 14.0 2 3.1 Deadlines 7 6.1 7 11.7 0 0 4 8.0 3 4.7 Help 7 6.1 7 11.7 0 0 5 10.0 2 3.1 Buyout 5 4.4 3 5.0 2 3.7 4 8.0 1 1.6 Other 14 12.37 11.7 7 13.0 5 10.0 9 14.1 Total Information Sources and Types Many of the sources and types of information listed in Tables 5.10 through 5.15 overlapped. When the total number of sources and types of informa tion each person had searched or received was calculated, these categories were counted only once. Information types with less than five associat ed responses were lumped into an Â“OtherÂ” category for presentation, but were treated as independent in the tota l count. In addition, if it was different from those listed, the orig inal source of NFIP information was also included. Before they were asked where th ey first heard about the program, however, respondents were asked to rate their familiarity on a scale of 1, Not Familiar at All, to 7, Completely Familiar. Those who were not at all familiar were not asked the source of their NFIP information. Reported sources are presented in Table 5.16. The ranges and means of total source and type c ounts are included in Tables 5.17 and 5.18.
105 Table 5.16. Original Source of NFIP Information All (N=74)* Union (N=37) Vestal (N=37) In 100 (N=47) In 500 (N=27) # % # % # % # % # % Mortgage Lender 25 33.811 29.7 14 37.8 21 44.7 4 14.8 Friends 19 25.710 27.0 9 24.3 10 21.3 9 33.3 Insurance Agent 11 14.93 8.1 8 21.6 6 12.8 5 18.5 FEMA 11 14.98 21.6 3 8.1 9 19.1 2 7.4 In Insurance Business 3 4.1 2 5.4 1 2.7 1 2.0 2 7.4 Real Estate Agent 3 4.1 3 8.1 0 0 0 0 3 11.1 Media 2 2.7 0 0 2 5.4 0 0 2 7.4 *Those answering 1 on NFIP Familiarity were not asked this question. Table 5.17. Total Number of Sources Searched and Received All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 0 1 1 4 0 Maximum 12 11 12 12 10 Mean 6.0 5.9 6.1 7.6 4.7 Std Deviation 2.9 2.9 2.8 1.9 2.8 Table 5.18. Total Number of Information Types Searched and Received All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 0 1 1 1 0 Maximum 12 12 10 12 10 Mean 4.3 4.1 4.6 5.3 3.5 Std Deviation 2.4 2.5 2.3 2.4 2.0 Given the federal requirements surr ounding lending in the hundred year floodplain, it is understandable that the la rgest proportion of re spondents cited their mortgage lender as introducing them to th e program. The exception was the five hundred
106 year floodplain grouping, where friends were the most often mentioned source. This might demonstrate the importance of persona l communication networks in non-regulated areas that perhaps have less flood experience. However, 58 pe rcent of those living in the five hundred year floodplain said they were completely unfamiliar with the program. Over a third of the total sample felt the sa me. There is a lot of room for improvement; intensifying meaningful media coverage ma y help to increase discussion among friends, though these friends may decide collectiv ely that insurance isnÂ’t worth it. The highest number of sour ces and types of information searched or received by any individual was 12. The largest difference in means was between the hundred and five hundred year floodplain subsets. This was primar ily a result of the differential search for help in the immediate post impact period. The re sponse rates stayed fairly stable for the five hundred year floodplain group moving from the event stage to after impact, but were much higher in the hundred year subset. Community means and distributions were similar. Reflective of the patterns found in th e results above, the number of source types was lower than the number of sources. Th e trend was most pronounced in the hundred year floodplain and is likely a result of the search for assistance. Credibility Source credibility is also part of a flo od risk information infrastructure. It is similar, but not identical to trust. In this question, participants we re asked to rate the credibility of flood related information from 11 sources on a scale of one to seven. One indicated that the source was not credible at all and seven was considered completely credible. Responses of DonÂ’t Know were al so acceptable, but were treated as missing
107 values. DonÂ’t know answers were most common for real estate agen ts, insurance agents, and the county government (Â“we never saw themÂ”) and lowest for mass media outlets. Table 5.19 includes the range, mean and standard deviation for the total sample. For most sources, statistics were similar across the subgroups. Exceptions are discussed below. Even though the question centered on ge neral credibility, many participants who were impacted by the June floods turned it into a referendum on the sourceÂ’s perceived performance during the disaster. This was s ource specific, however; family, friends, real estate agents, and to a certain extent, the NWS and insurance agents appeared to be judged by different criteria. Real estate agen ts had the lowest mean rating and were the only source not rated by at least one person as completely credible. The consistent refrain was Â“TheyÂ’re just trying to sell you something. TheyÂ’ll say anything.Â” Table 5.19. Source Credibility for Total Sample* Source (N) ** Minimum Maximum Mean Std. Dev. Family (N=103) 1 7 3.8 1.9 FEMA (N=103) 1 7 4.1 1.8 Friends (N=109) 1 7 4.1 1.7 Town (112) 1 7 4.4 2.0 County (N=92) 1 7 3.9 1.6 Newspaper (N=112) 1 7 4.6 1.5 Real Estate Agent (N=81) 1 6 2.0 1.2 Television (N=113) 1 7 4.7 1.3 National Weather Service (N=98) 1 7 4.9 1.6 New York State (N=101) 1 7 4.1 1.5 Insurance Agent (N=92) 1 7 3.6 1.8 Scale of 1, Not Credible at All, to 7, Completely Credible. **Remainder responded DonÂ’t Know
108 Those who rated the National Weather Service rated it highly. Accompanying comments indicated that people found the or ganization thoroughly cr edible, but wished their information was better (and more qui ckly) disseminated. TV news and the newspaper were also rated fa irly well overall, but there were spatial and temporal differences in perception related to televi sion coverage. Most other sources had mean scores around four and similar deviations. Othe r sources associated with spatial variation were family members, town government, a nd county government. These variations are illustrated in Tables 5.20 through 5.23. Table 5.20. TV Credibility* All (N=113) Union (N=59) Vestal (N=54) In 100 (N=49) In 500 (N=64) Minimum 1 2 1 2 1 Maximum 7 7 7 7 7 Mean 4.7 4.9 4.4 4.2 5.0 Std Deviation 1.3 1.2 1.4 1.1 1.3 Scale of 1, Not Credible at All, to 7, Completely Credible. Table 5.21. Family Member Credibility* All (N=103) Union (N=52) Vestal (N=51) In 100 (N=44) In 500 (N=59) Minimum 1 1 1 1 1 Maximum 7 7 7 7 7 Mean 3.8 3.7 3.8 4.3 3.3 Std Deviation 1.9 1.9 1.9 1.9 1.8 Scale of 1, Not Credible at All, to 7, Completely Credible.
109 Table 5.22. Town Credibility* All (N=112) Union (N=59) Vestal (N=53) In 100 (N=50) In 500 (N=62) Minimum 1 1 1 1 1 Maximum 7 7 7 7 7 Mean 4.4 3.7 5.1 4.6 4.2 Std Deviation 2.0 1.9 1.9 2.2 1.9 Scale of 1, Not Credible at All, to 7, Completely Credible. Table 5.23. County Credibility* All (N=92) Union (N=49) Vestal (N=43) In 100 (N=38) In 500 (N=54) Minimum 1 1 1 1 1 Maximum 7 6 7 7 7 Mean 3.9 3.5 4.2 3.6 4.0 Std Deviation 1.6 1.5 1.7 1.7 1.6 Scale of 1, Not Credible at All, to 7, Completely Credible. TVÂ’s credibility rating was three quarters of a point higher in the five hundred year floodplain than in the hundred year fl oodplain (t-test, p=0.002). In the hours leading up to the flood, most in the hundred year fl oodplain had a greater need for information, but didnÂ’t feel they got it from the local stati ons. As one Vestal reside nt put it, Â“It seemed like the rest of the world was looking at us before we were .Â” Some gave TV a generally high rating, but qualified it with a statement like Â“It was okay later, but in the beginning, it was a 2.Â” There was a full point between mean ratings of family credibility (t-test, p=0.008). In that case, though, hundred year floo dplain residents rated credibility higher. It may be that these people relied more h eavily on their families during the June flood. The largest difference in credibility o ccurred between Union and Vestal in the credibility rating of the Town government (t -test, 1.45, p<0.001). Even if they viewed the
110 state or FEMA negatively, most people in Vestal were impressed by the response and appreciated the consistent meetings and the past responsiveness of the local government. Many in Union gave glowing reports of the utility and garbage crews and firemen, and even a few government officials, but the gene ral impression was one of frustration with a government some did not perceive to be fo rthright about current problems and past allocations and projects. I believe the county sc ores had less to do with any perception of the county and more to do with a lack of clear visibility and a ssociation as Â“local government.Â” Cognitive Factors Seeking and Sufficiency Questions assessing cognitive factors were placed throughout the questionnaire and addressed information seeking tendencie s, perceived sufficiency of information, knowledge, and general outlook. Most cogni tive questions were closed. Seeking tendencies were derived from the data collected for flood risk infrastructures. Because the perceived need for information varied and so me people had less to sa y, calculations were based on individual totals. In each of the th ree time frames, a percentage was calculated by dividing the number of sources sought by th e total number of sources the individual came into contact with. In further analyses, these three percentages were averaged in order to represent a more general tendenc y. Mean results for each of the periods are included in Table 5.24.
111 Table 5.24. Average Percentage of Information Sources Sought All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) % SD % SD % SD % SD % SD During Event 57 38 67 39 47 35 51 37 62 39 After Impact 54 38 58 39 49 37 68 30 42 40 Post Response 31 43 35 44 26 42 44 47 21 38 Seeking activity for the total sample tended to decrease with time. Seeking percentages were much lower in the post response category across spatial groupings. However, percentages were lower in Vest al and the hundred year floodplain during the event than they were in the post impact period and when compared to their spatial opposites. This is primarily a result of pattern s of official warning and evacuation orders. These were sources that coul d not be searched, but did provide information, probably driving down the scores. VestalÂ’s scores we re generally lower than UnionÂ’s and may reflect more accessible information flowing fr om the local government. There was little difference in ratings of overall satisfaction with available flood information between the communities, however (see Table 5.25). Over 70 percent of the sample rated their satisfaction as at least a fi ve on a seven point scale, indicating that most thought information was sufficient. Table 5.25. Overall Satisfaction with Flood Information* All (N=112) Union (N=59) Vestal (N=53) In 100 (N=49) In 500 (N=63) Minimum 1 1 1 1 1 Maximum 7 7 7 7 7 Mean 4.9 4.8 5.0 5.0 4.8 Std Deviation 1.5 1.6 1.4 1.6 1.5 Scale of 1, Completely Dissatisfied, to 7, Completely Satisfied.
112 Knowledge Knowledge questions had to do with both general flood issues and the NFIP. Three out of the four questions were closed and two used self ev aluation. The open ended question addressed factors perceived to c ontribute to flooding and was asked of all participants, regardless of experience. No card containing response possibilities was given to them. One person said that she ha d no idea and another mentioned ten separate contributors. The mean, however, was 3.3 and approximately 82 percent mentioned four factors or fewer. This pattern was consistent across categor ies. Table 5.26 lists specific conditions and the percentages of participan ts that mentioned them in each subset. About half of the respondents mentione d heavy rain, but answers were quite varied and ranged from a shallow water tabl e and antecedent moisture to management and planning issues. Construction and develo pment had the second highest response rate overall, though lack of dredging and flood cont rol was cited as a cause more often in Union and the hundred year floodplain. About ha lf the number that called for structural intervention saw watersheds as connected a nd believed control measures pushed the problems somewhere else. A few mentione d both. One glaring difference is the percentage that cited political decisions as c ontributing to flooding in Union versus those who mentioned politics in Vestal. The issues brought up were identical to those discussed in the information infrastructure section.
113 Table 5.26. Factors Believed to Contribute to Flooding All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # %* # % # % # % # % Heavy/Prolonged Rain 59 51.827 45.0 32 59.3 27 54.0 32 50.0 Construction/Dev. 32 28.117 28.3 15 27.8 14 28.0 18 28.1 Lack of Dredging 28 24.618 30.0 10 18.5 16 32.0 12 18.8 Lack of Flood Structures 28 24.615 25.0 13 24.1 15 30.0 13 20.3 Poor Sewer System 28 24.618 30.0 10 18.5 14 28.0 14 21.9 Climate Change 23 20.212 20.0 11 20.4 9 18.0 14 21.9 Lack of Maintenance 23 20.216 26.7 7 13.0 16 32.0 7 10.9 Political Decisions 20 17.518 30.0 2 3.7 9 18.0 11 17.2 River Change 19 16.79 15.0 10 18.5 9 18.0 10 15.6 Upstream Flood Control 16 14.07 11.7 9 16.7 9 18.0 7 10.9 Ice or Snow 15 13.28 13.3 7 13.0 7 14.0 8 12.5 Landuse Change 15 13.29 15.0 6 11.1 8 16.0 7 10.9 Full Dams 15 13.29 15.0 6 11.1 8 16.0 7 10.9 Physical Characteristics 13 11.47 11.7 6 11.1 7 14.0 6 9.4 Poor Planning 11 9.6 8 13.3 3 5.6 5 10.0 6 9.4 Freak of Nature 10 8.8 4 6.7 6 11.1 4 8.0 6 9.4 Loss of Natural Deterrents 8 7.0 6 10.0 2 3.7 5 10.0 3 4.7 Infill 7 6.1 1 1.7 6 11.1 2 4.0 5 7.8 Respondents used multiple descriptions; percentages may not add to 100. The variety of responses i ndicates the sample as a w hole had a fairly complex understanding of flood processe s and hydrology. There were individuals, however, who described the June 2006 flood simply as a Â“fr eak of nature,Â” a result of conditions unlikely to happen again. After participants were asked to name things that contributed to flooding, they were asked to rate their own knowledge of flooding on a scale that ranged from 1, Not at All Knowledgeab le to 7, Extremely Knowledgeable. There was a moderate
114 positive correlation between the number of contributors cited and the self rating (rho=0.37, p=0.009). Ranges and means are includ ed in Table 5.27. The higher scores in the hundred year floodplain and Union may reflect the importance of experience (severity and frequency) in knowledge and knowledge estimation. Table 5.27. Self Rated Flood Knowledge* All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 1 1 1 2 1 Maximum 7 7 7 7 7 Mean 4.7 4.9 4.4 5.1 4.3 Std Deviation 1.6 1.6 1.6 1.5 1.6 *Scale of 1, Not at all Knowledgeable, to 7, Extremely Knowledgeable. Severity of experience may also explain pa rt of the difference in mean scores for self rated familiarity with the NFIP (see Table 5.28). Those who suffered more damage likely had more sustained dealings with FEMA and insurance agents and had to wade through considerable paperwork. However, mo st of the difference probably resulted from the hundred year floodplain being a regulated ar ea. Distribution for the total sample was right skewed; over a third of the participants rated themselves as completely unfamiliar with the program. Only three people in the hundred year floodplain rated their familiarity as a one; almost all had at least heard of it. However, a quarter of hundred year floodplain participants rated their familiarity a tw o, though this group is the programÂ’s target population. Respondents were also asked whether or not they lived in a Special Flood Hazard Area. Approximately 83 percent of hundred year floodplain residents who answered the question correctly believed that they did. Three said they didnÂ’t know.
115 Table 5.28. Self Rated Familiarity with NFIP* All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 1 1 1 1 1 Maximum 7 7 7 7 7 Mean 3.0 2.9 3.1 4.2 2.1 Std Deviation 2.1 2.0 2.1 2.0 1.6 Scale of 1, Completely Unfamiliar, to 7, Completely Familiar. General Outlook The final cognitive factor addressed in the survey was general outlook. Two closed questions asked participants to rate their agreement with statements related to locus of control and ambient worry. The scale s ranged from 1, Str ongly Disagree to 6, Strongly Agree. Ranges and means ar e included in Tables 5.29 and 5.30. Distributions for control we re fairly normal in all spatial subsets and centered on a rating of three. Participants fe ll into each of the categories wi th regards to worry as well, but over three quarters of the total sample answered the worry question with a rating of one, two, or three. There was no significant di fference between the ratings of any spatial groups and the somewhat lopsided distribution was consistent. Table 5.29. I Have Control over What Happens to Me* All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 1 1 1 1 1 Maximum 6 6 6 6 6 Mean 3.2 3.3 3.1 3.3 3.2 Std Deviation 1.6 1.7 1.5 1.5 1.6 Scale of 1, Strongly Disagr ee, to 6, Strongly Agree.
116 Table 5.30. I Am Constantly Worrying about Something* All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 1 1 1 1 1 Maximum 6 6 6 6 6 Mean 2.7 2.9 2.4 2.7 2.6 Std Deviation 1.5 1.6 1.4 1.5 1.6 Scale of 1, Strongly Disagr ee, to 6, Strongly Agree Participants were also as ked to indicate the level of government, individual or organization that they believed had primary re sponsibility for prot ecting individuals, not against flooding, but flood damages. In the tota l sample, most people believed individuals themselves had primary responsibility (see Table 5.31). When combinations were included, about 44 percent of the sample stat ed that the individual had at least some responsibility for mitigation. The biggest diffe rence was in the responses of the hundred year and five hundred year floodplain gr oupings (38 and 48 percent respectively). Additionally, those in the hundred year flood pl ain and those living in Union were more likely to say that responsibility lay with th e local government. This perception in Union may compound the credibility and comm unication problems discussed above. SFHA residents also looked to the fe deral government to a greater degree than their counterparts and were more consistent in their responses.
117 Table 5.31. Primary Responsibility fo r Protecting against Flood Damages All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Individual 30 26.315 25.0 15 27.8 14 28.0 16 25.0 Local Gov. 25 21.116 26.7 9 16.7 15 30.0 10 15.6 State Gov. 10 8.8 3 5.0 7 13.0 2 4.0 8 12.5 Federal Gov. 13 11.48 13.3 5 9.3 9 18.0 4 6.3 Individual and Local Gov. 9 7.9 4 6.7 5 9.3 3 6.0 6 9.4 All Parties 8 7.0 4 6.7 4 7.4 2 4.0 6 9.4 All Governments 6 5.3 2 3.3 4 7.4 3 6.0 3 4.7 Other Combinations 13 11.48 13.3 5 9.3 2 4.0 11 17.2 GENERAL FLOOD RELATED PE RCEPTION AND BEHAVIOR Outcome factors included understanding of flood related uncertainty, threat perception, and general and even t specific behaviors. Question s pertaining to uncertainty and threat were closed. Questions on mitigative behaviors were open, but participants were given prompts on cards to assist their recollection. Understanding Processes and Uncertainty Understanding of flood related uncertain ty over time and space was evaluated using two questions. In the desc ription specific portion of the survey, participants were asked which of the described floods (hundred year flood, one percent chance, 26 percent chance), if any, could happen more than once in a year. They were also asked to name the floods whose physical size they believed could change over time. Respondents who answered that all of them c ould change and that all of th em could happen more than one time per year were coded as understanding flood related uncertainty. Results are provided
118 in Table 5.32. Approximately 40 percent of th e total sample answered Â“AllÂ” to both questions. The largest gap again occurred between the hundred and five hundred year floodplain groups, but statistical tests showed no distributional differe nces significant at the 0.05 level. Table 5.32. Understanding of Flood Related Uncertainty over Space and Time All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Understands Uncertainty 46 40.423 38.3 23 42.6 24 48.0 22 34.4 Threat Perception: General The idea of threat contains elemen ts of both likelihood and concern. The conceptual framework of this research incl uded both. Participants were asked whether they considered themselves to be at high, me dium, or low risk of future flooding. Policy treats the hundred year floodplain a high ri sk area, but 70 percent of SFHA residents believed their risk to be medium or low (s ee Table 5.33). Half described their risk as medium. Almost 60 percent in the five hundred year floodplain believed their risk of flooding was low, rather than medium. In Vestal, less than ten percent of re spondents considered their risk high, though many more had suffered substantial damage. Over half did say they were at medium risk for future flooding. In Union, where 35 percen t of those interviewed had experienced more than one flood, more than twice as ma ny people rated their risk high as did in Vestal. The greatest proporti on (47 percent), though, belie ved their risk was low.
119 Table 5.33. Perceived Risk Level All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Low Risk 48 42.128 46.7 20 37.0 10 20.0 38 59.4 Medium Risk 48 42.119 31.7 29 53.7 25 50.0 23 35.9 High Risk 18 15.813 21.7 5 9.3 15 30.0 3 4.7 Flood related concern was evaluated using the same agreement scale used for the cognitive variables discussed in the previous section. The statement Â“Flooding is one of my top concernsÂ” situates flood related concer n in relation to other concerns associated with money, family, health and other day to day difficulties. Response distributions were bimodal in the total sample and the two commun ities, in part reflecting the differences in response patterns between the floodplain group ings. SFHA results were left skewed, but seven people chose Strongly Disagree. The mean and median were indeed lower in the five hundred year floodplain, but one quarter of these res pondents rated their relative concern a five or a six. Group m eans are included in Table 5.34. Table 5.34. Flooding Is One of My Top Concerns* All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) Minimum 1 1 1 1 1 Maximum 6 6 6 6 6 Mean 3.5 3.5 3.4 4.1 2.9 Std Deviation 1.8 1.8 1.8 1.8 1.6 Scale of 1, Strongly Disagr ee, to 6, Strongly Agree.
120 Behavior: General The card given to survey participants when they were asked about mitigative activities included thi ngs like raising utilities abov e a designated flood level and purposely buying outside the hundred year floodpl ain as well as checking with neighbors regarding past flood levels. Table 5.35 incl udes a breakdown of speci fic measures taken in consideration of flooding. Insurance purch ase was a separate question; results are included in Table 5.36. Table 5.35. General Consideration of Flooding All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # %* # % # % # % # % Checked with Neighbors 44 38.617 28.3 27 50.0 23 46.0 21 32.8 Personal Experience 19 16.712 20.0 7 13.0 10 20.0 9 14.1 Checked SFHA 8 7.0 5 8.3 3 5.6 7 14.0 1 1.6 Noticed Wall/Levee 7 6.1 5 8.3 2 3.7 0 0 7 10.9 Purposely Live out of SFHA 5 4.4 4 6.7 1 1.9 1 2.0 4 6.3 Modified House or Property 33 28.918 30.0 15 27.8 21 42.0 12 18.8 Prepared for Flood Event 22 19.315 25.0 7 13.0 16 32.0 6 9.4 Modified Use 7 6.1 4 6.7 3 5.6 3 6.0 4 6.3 Nothing Done 27 23.712 20.0 15 27.8 7 14.0 20 31.3 Checking with neighbors before moving in was the most common consideration and was most prevalent in the hundred year floodplain and Vestal. Approximately 17 percent cited personal e xperience with the area as a factor when they purchased or rented
121 a home. By far the most common effort to mitigate losses was the purchase of flood insurance. However, less than ten percent of those in the five h undred year floodplain had it. Many responded when asked if they had insu rance that Â“I was told I didnÂ’t need itÂ” by a lender or a city official or an insurance agen t. Three people said that an agent wouldnÂ’t sell it to them (though this may have been a stretc h of the truth). It may be useful to shift the focus of discussions surrounding the NFIP, and perhaps the program itself, from need to want; how do we get people IN the progr am rather than OUT? Exemptions may be working against incentive programs like the Community Rating System. Table 5.36. Insurance All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # %* # % # % # % # % Has Flood Insurance 44 38.620 33.3 24 44.4 38 74.0 6 9.4 In addition to purchasing insurance, almo st 30 percent of the total sample had modified their home or property in some way. Ab out half had raised their utilities. Others sealed unused drains, installed check valves, or installed resistant materials. About 19 percent of participants had a pump on ha nd and/or had worked out an easy, water resistant storage system so they could quickly move things to higher elevations. About 24 percent, however, had never taken action, and had not given flooding any consideration when moving in.
122 Behavior: Event Specific In addition to answering questions about general measures, participants were asked what they did, if anything, during the Ju ne floods to protect themselves or their property. Evacuation was included on the card used for this question, as were sandbagging and moving belongings. Tabl e 5.37 lists the activities mentioned. Table 5.37. Event Specific Measures All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Evacuated 79 69.333 55.0 46 85.2 46 92.0 33 51.6 Moved Belongings 51 44.726 43.3 25 46.3 33 66.0 18 28.1 Visual Check 16 14.012 20.0 4 7.4 3 6.0 13 20.3 Pumped 12 10.58 13.3 4 7.4 8 16.0 4 6.3 Sandbagged 4 3.5 1 1.7 3 5.6 3 6.0 1 1.6 Other Measures 18 15.85 8.3 13 24.1 8 16.0 10 15.6 Nothing Done 14 12.311 18.3 3 5.6 2 4.0 12 18.8 As noted in previous sections, evacuati on rates were higher in Vestal and the hundred year floodplain than in Union or th e five hundred year fl oodplain. While over 50 percent of participants evacuat ed in all spatial groups, the majority of evacuating five hundred year residents lived in Vestal and most of the Union evacuees lived in the SFHA. Two thirds of SFHA residents moved their be longings to higher ground, just Â“not high enough,Â” as floodwaters swamped fi nished basements and moved into first floor spaces. Since the majority of hundred year fl oodplain and Vestal residents evacuated, most who said they visually checked flood le vels lived in UnionÂ’s five hundred year floodplain. Others called to check on the water height. Very few people sandbagged,
123 perhaps because of the relatively abrupt onset and evacuation orders. Two who did sandbag were not sandbagging their own hous es, but were augmenting flood control structures. People also mentioned gatheri ng important papers a nd keepsakes, taking measures to protect their identity, and sneak ing past barriers to check on their homes. PERCEPTION ASSOCIATED WITH SPECIFIC DESCRIPTIONS In addition to exploring the relationships between situational and cognitive factors and perceptual and behavioral outcomes, th is project examined perception associated with three specific methods of framing polic yÂ’s benchmark flood: th e hundred year flood (return period); a flood with a one percent chan ce of occurring any year (probability); and a flood with a 26 percent chance of occurr ing 30 years (cumulative probability). Participants were given a card with all three full descriptions printed on it. Questions in this portion of the survey dealt with perceive d relative likelihood, re lative size, temporal and spatial uncertainty, and concern. This sec tion includes tables containing the results and brief descriptions; these re sults and pertinent qualitative data will be discussed in more detail in Chapter 6. Relative Likelihood In addition to giving respondents the card w ith the descriptions on it, interviewers also read the descriptions aloud. No further explanation was given, however. Participants were first asked to choose the floods they t hought were most and leas t likely to occur in the next year. They were also told that th ey could answer Â“All,Â” Â“DonÂ’t Know,Â” or with any combination of descriptions. Ta bles 5.38 and 5.39 include the results.
124 Table 5.38. Flood Thought Most Likely to Occur within Year All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % 100 Year Flood 3 2.6 2 3.3 1 1.9 2 4.0 1 1.6 1% Chance Flood 28 24.614 23.3 14 25.9 12 24.0 16 25.0 26% Chance Flood 64 56.132 53.3 32 59.3 26 52.0 38 59.4 1% and 26% Chance Floods 2 1.8 2 3.3 0 0 2 4.0 0 0 All Equal 2 1.8 0 0 2 3.7 1 2.0 1 1.6 DonÂ’t Know 15 13.210 16.7 5 9.3 7 14.0 8 12.5 Table 5.39. Flood Thought Least Likely to Occur within Year All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % 100 Year Flood 82 71.944 73.3 38 70.4 34 68.0 48 75.0 1% Chance Flood 12 10.55 8.3 7 13.0 8 16.0 4 6.3 26% Chance Flood 1 0.9 1 1.7 0 0 1 2.0 0 0 100 Year and 1% Chance Floods 6 5.3 3 5.0 3 5.6 3 6.0 3 4.7 All Equal 2 1.8 0 0 2 3.7 1 2.0 1 1.6 DonÂ’t Know 11 9.6 7 11.7 4 7.4 3 6.0 8 12.5 After hearing the descrip tions and looking at the card, two people said, Â“These are all the same, right?Â” Neither indicated they worked in the insurance business. I then coded the responses as Â“AllÂ” and skipped to the question regarding level of concern. Another six people identified the hundred ye ar flood and the one percent chance floods as equal. These participants were asked all th e questions. Over three quarters of the total sample thought the hundred year flood or both the hundred year and one percent chance
125 floods were least likely to occur. The vote for most likely was split between the two probabilistic descriptions a nd patterns were similar across spatial categories. Relative Size Participants were also asked to name the floods they thought were biggest and smallest in size. For the most part, respons es for the biggest in size were logically consistent with answers given for the leas t likely (see Table 5.40) The number of people answering DonÂ’t Know was identical. Almost 30 percent of the respondents said they didnÂ’t know which flood was smallest, however, more than twice as many as indicated they werenÂ’t sure about the most likely flood (see Table 5.41). Additionally, the rankings of the smallest and most likely floods were not logically consistent. The response rate for the 26 percent chance flood was less than half what it was for most likely to occur. Lack of familiarity may have played a role, but, just as they were for likelihood, rankings were the same in all spatial subsets. Table 5.40. Flood Thought Biggest in Size All (N=113) Union (N=60) Vestal (N=53) In 100 (N=50) In 500 (N=63) # % # % # % # % # % 100 Year Flood 80 70.842 70.0 38 71.7 37 74.0 43 68.3 1% Chance Flood 10 8.8 6 10.0 4 7.5 5 10.0 5 7.9 26% Chance Flood 4 3.5 3 5.0 1 1.9 2 4.0 2 3.2 100 and 1% 6 5.3 3 5.0 3 5.7 3 6.0 3 4.8 All Equal 2 1.8 0 0 2 3.8 1 2.0 1 1.6 DonÂ’t Know 11 9.7 6 10.0 5 9.4 2 4.0 9 14.3
126 Table 5.41. Flood Thought Smallest in Size All (N=113) Union (N=60) Vestal (N=53) In 100 (N=50) In 500 (N=63) # % # % # % # % # % 100 Year Flood 1 .9 1 1.7 0 0 0 0 1 1.6 1% Chance Flood 52 46.025 41.7 27 50.9 21 42.0 31 49.2 26% Chance Flood 25 22.113 21.7 12 22.6 15 30.0 10 15.9 All Equal 2 1.8 0 0 2 3.8 1 2.0 1 1.6 DonÂ’t Know 33 29.221 36.7 12 22.6 13 26.0 20 31.7 Understanding of Uncertainty over Time and Space Questions in this section we re used to link descripti ons to the understanding of flood related uncertainty over time and sp ace. Contrary to previous questions, participants were not asked to compare the descriptions to one another. There was no Â“mostÂ” or Â“least,Â” just what people thought possible or not. Respondents were asked which floods could happen more than once a year and which could change in size over time. Responses of Â“AllÂ” were used to cons truct the variable used for understanding of uncertainty in the general model. In these two questions, answers of Â“NoneÂ” were also acceptable. Results are pres ented in Tables 5.42 and 5.43. Because combinations were not mutually exclusive, each description was coded as a yes even when all of them were men tioned. Thus, the response rates for individual descriptions shown in the tabl es also include votes for Â“All.Â” In Table 5.42, for instance, 51 people said that all of the floods describe d could happen more than one time per year. The number in the cell for the hundred year flood is 52; only one person who did not answer Â“AllÂ” thought that th e hundred year flood could happen more than once. The
127 results for the other two desc riptions had more independent responses. The one percent chance description had the highest proportion of affirmative answers in all groupings. There was little internal variation in the que stion of change over time, however, and the answer appeared to have little to do with descriptions them selves. Almost 90 percent of the survey participants replied wi th either All or DonÂ’t Know. Table 5.42. Flood Could Happen More than Once in a Year All (N=113) Union (N=60) Vestal (N=53) In 100 (N=50) In 500 (N=63) # % # % # % # % # % 100 Year Flood 52 46.024 40.0 28 52.8 27 54.0 25 39.7 1% Chance Flood 89 78.847 78.3 42 79.2 42 84.0 47 74.6 26% Chance Flood 76 67.340 66.7 36 67.9 38 76.0 38 60.3 All 51 45.124 40.0 27 50.9 26 52.0 25 39.7 None 4 3.5 2 3.3 2 3.8 2 4.0 2 3.2 DonÂ’t Know 9 8.0 5 8.3 4 7.5 2 4.0 7 11.1 Table 5.43. Size of Flood Could Change over Time All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % 100 Year Flood 77 67.542 70.0 35 64.8 36 72.0 41 64.1 1% Chance Flood 76 66.742 70.0 34 63.0 36 72.0 40 62.5 26% Chance Flood 76 66.743 71.7 33 61.1 36 72.0 40 62.5 All 74 64.941 68.3 33 61.1 36 72.0 38 59.4 None 7 6.1 2 3.3 5 9.3 3 6.0 4 6.3 DonÂ’t Know 27 23.714 23.3 13 24.1 11 22.0 16 25.0
128 Relative Concern The final questions related specifically to the three descriptio ns had to do with concern and consisted of two parts. First, respondents were asked which of the three floods described on the cards concerned them the most. They were then asked to rate their associated level of concern on a scale ranging from, 1, Not Con cerned at All to 7, Completely Concerned. The process was then repeated for the flood or floods perceived as least concerning. Although over 75 percent picked the hundred year flood as the biggest and about 56 percent chose the 26 percen t chance flood as most likel y, 42 percent of the total sample said that all floods were equall y concerning (see Tabl e 5.44). This response probably had more to do with situational and cognitive factors than perceptions associated with specific descriptions a nd reflected both uniformly high and uniformly low concern levels (see Figur e 5.1). The remaining participants essentially split their votes between the hundred year floo d and the 26 percent chance flood. Table 5.44. Flood Thought Most Concerning All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % 100 Year Flood 33 28.9 17 28.3 16 29.6 13 26.0 20 31.3 1% Chance Flood 6 5.3 4 6.7 2 3.7 1 2.0 5 7.8 26% Chance Flood 22 19.3 12 20.0 10 18.5 10 20.0 12 18.8 100 Year and 1% Chance Floods 4 3.5 2 3.3 2 3.7 1 2.0 3 4.7 1% and 26% Chance Floods 1 0.9 1 1.7 0 0 1 2.0 0 0 All Equally Concerning 48 42.1 24 40.0 24 44.4 24 48.0 24 37.5
129 Figure 5.1. Concern Levels for Equal Concern: All level of concern8 6 4 2 0 Frequency20 15 10 5 0 Individuals who said that all floods were equally con cerning were not asked to name their least concerning flood. Table 5.45 breaks down the answers of the other 66 respondents. As with the most concerni ng flood, two descriptions garnered over 80 percent of the votes. The desc ription with the highest pr oportion was the one percent chance flood, which was also considered th e smallest by the most people. The hundred year flood, overwhelmingly considered the l east likely, received the second largest number of affirmative answers in all spatia l groups. The combined response patterns indicate that both perceive d likelihood and perceived size might influence concern, though perhaps inconsistently.
130 Table 5.45. Flood Thought Least Concerning All (N=66)* Union (N=36) Vestal (N=30) In 100 (N=26) In 500 (N=40) # % # % # % # % # % 100 Year Flood 21 31.814 38.9 7 23.3 8 30.8 13 32.5 1% Chance Flood 34 51.518 50.0 16 53.3 14 53.8 20 50.0 26% Chance Flood 8 12.13 8.3 5 16.7 3 11.5 5 12.5 1% and 26% Chance Floods 1 1.5 1 2.8 0 0 0 0 1 2.5 DonÂ’t Know 2 3.0 0 0 2 6.7 1 3.8 1 2.5 Individuals who said that all floods were equally con cerning were not included In the total sample, mean concern associ ated with the most concerning flood was 4.9, but the distribution was left skewed. In the two communities, means and distributions were similar. The average level of concer n was higher in the h undred year floodplain (5.6) and lower in the five hundred year floodpl ain (4.3). A similar pa ttern occurred with the least concerning flood. The mean for th e whole group was 3.0, while SFHA residents, on average, rated concern 3.6 a nd those living in the five hun dred year floodplain gave it a 2.7. Wilcoxon tests showed signi ficant distributional differen ces in both cases (most, p<0.001; least, p=0.036). For the most concerning level, left skew was substantial in the hundred year floodplain, but the distribution for five hundred year floodplain residents was fairly normal. The opposite was true for the least concer ning levels; the SFHA distribution was relatively normal, while the five hundred year floodplain responses showed right skew. Figure 5.2 i llustrates the difference in leas t concerning levels split by floodplains. Mean range between most and leas t concerning scores was just under two in all spatial groups.
131 Figure 5.2. Distributional Differences in Least Concerning Level: 100 and 500 Year Floodplains 100 Year Floodplain level of concern8 6 4 2 0 Frequency6 5 4 3 2 1 0 500 Year Floodplain level of concern8 6 4 2 0 Frequency12 10 8 6 4 2 0
132 CHAPTER 6: ANALYSIS AND DISCUSSION This project explored four sets of resear ch questions: 1. Which situational and cognitive factors are most highly related to varying perceptions of flood pr ocesses and uncertainty when relationships between the factor s are controlled? To a general perception of flood threat? To mitigative behavior? How are these outcomes related to each other? 2. When relationships between them ar e controlled, whic h situational and cognitive factors are most highly related to varyi ng perceptions of size, likelihood, uncertainty, and concern asso ciated with specific flood risk messages? Messages addressed in this project include the hundred year flood, a flood with a one percent chan ce of occurring in any year, and a flood with a 26 percent chance of occurring in 30 years. 3. Which of these flood risk messages are comparatively most effective with regards to understa nding and/or persuasion? 4. How do people describe floods and what worries them about flooding? How might flood risk communication be improved? Chapter 6 is divided into four sections that correspond with th e sets of research questions above. Each section addresses any changes made to the data for the purpose of analysis, the specific analytic technique s employed, and the results pert inent to the question set at hand.
133 EXPLORING FIGURE 2.1: THE GENE RAL MODEL OF PERCEPTUAL AND BEHAVIORAL INFLUENCES Data and Methods This section addresses the first set of research questions. Figure 2.1 outlines relationships identified in th e literature between five sets of situational and cognitive factors and perceptual and be havioral outcomes. These relationships were modeled and the combined influences of variables were explored through the use of binary logistic regression. Table 6.1 describes wh ich predictor variables were used in analysis and how they were measured. Distance, part of the location factor, was left out because preliminary analysis showed no significant corr elations to outcome variables. This result is likely due to the presence of flood control structures. There was little difference in race or ownership status across the sample (see Table 5.3), and these de mographic variables were also excluded from analysis. Tabl e 6.2 explains the outcome variables. Eight of the measuring variables were de rived from the raw data presented in Chapter 5. Two socio-economic variables we re modified for analysis. First, the measurement for education was changed from an ordinal scale to a binary. Those who had completed a bachelorÂ’s degree or a gra duate degree were assigned a one. Those who had not were assigned a zero. In the income variable, the two highest categories were combined to make the distances between ranks more consistent. One of the questions related to genera l outlook asked what entity had primary responsibility for protecting individuals agai nst flood damages. The variable used in analysis focused on individual responsibil ity. If an answer included Â“individuals themselvesÂ” or individuals in combination wi th another level of government or agency,
134 the participant was considered to hold the individual at least partially responsible for his or her own well-being. Two of the adjusted variables measur ed knowledge in some form. Respondents who lived in the Special Flood Hazard Area (SF HA) and also believed that they did were coded as having knowledge of their location as it relate s to flood risk and policy. A participant who lived in the five hundred year floodplain a nd said that s/he did NOT live in the SFHA was coded the same way. In or der to conceptualize the understanding of flood processes, participants we re asked what kinds of things they thought contributed to flooding. This open ended question led to a va riety of responses outlined in Table 5.45. These responses were then grouped into four broad cause types using factor analysis: natural processes; human alte ration of environment; lack of flood control structures; and planning and management issues. The number of cause types rather than the raw number of causes mentioned was used to construct the measurement scale. Responses of Â“DonÂ’t KnowÂ” were assigned a zero, so the fina l scale ranged from zero to four. The remainder of the modified variables ha d to do with information infrastructure or seeking habits. Flood Risk Infrastruc ture variables were first aggregated by source/channel (these were ofte n conflated) and information type in order to reduce the number of variables. The numbers of s ources and information types were highly correlated and were combined into a singl e measure; CronbachÂ’s alpha was 0.88. Before combination, both scales ranged from zero to twelve. They were weighted equally. Individuals who looked for many kinds of info rmation in a variety of locations scored higher on the scale and were considered to have more well developed flood risk information infrastructures. The general cred ibility of flood risk information sources was
135 assessed by averaging the credibility scores of all sources to which a score was assigned. Sources eliciting response s of Â“DonÂ’t KnowÂ” were not includ ed. As in the original scales, scores could range from one to seven. Ov erall seeking tendencies were measured by averaging the percentage of sources sought in each of the three general time periods. In all cases, data were lost as a result of the a ggregation; future analyses should focus on the associations between individual sources, in formation types and credibility scores, especially as they relate to informa tion satisfaction and s eeking patterns. Three of the binary outcome variables used in the logistic re gressions were also derived from data presented in Chapter 5. Understanding of uncertainty was measured using two questions in the section of the su rvey dealing with flood risk descriptions. Participants were first asked which of the th ree described floods could happen more than once in a year. They were also asked if the described floodsÂ’ sizes could change over time. Respondents were considered to understa nd uncertainty in the timing and size of flooding if they answered that all described floods could happen more than once a year AND that the size of all of described floods could change. Both variables measuring threat percepti on were modified. Scal es or distribution made each inappropriate for linear regression, so binaries were created. In the survey, participants were asked whethe r they thought they were at lo w, medium, or high risk of flooding in the future. Because U.S. flood policy considers the five hundred year floodplain to be Â“mediumÂ” risk and the hundred year to be a Â“highÂ” risk area, these two answers were grouped together as an appropriate response for the entire sample (at least from a policy perspective). For regressions using the SFHA subset alone, only a response of Â“HighÂ” risk was coded as appropriate.
136 Table 6.1. Measurement of Situational and Cognitive Factors Factors Listed Variable Measuring Variable(s) Explanation Education Bachelors or Hi gher Categorical Based on Rank Income 2005 Household Income in (000) Dollars Adjusted 7 Point Ordinal Scale Based Primarily on 15K Increments Age Age in Years Unchanged Scale Gender Female is 1; Male is 0 Unchanged Binary SocioEconomic Residency Years in Current Home Unchanged Scale Floodplain Status 100 Year is 1; 500 is 0 Unchanged Binary Location Community Vestal is 1; Un ion is 0 Unchanged Binary Frequency # Times Property Flooded Unchanged Scale Experience Impact Level Impact Severity from None to Extreme Unchanged Ordinal Scale Information Sources/Channels Information Type Breadth and Depth of FRI Scale Based on Addition of Coded Sources and Types ( = 0.88) Flood Risk Infrastructure (FRI) Credibility Average Credibility Scalar Mean of Source Credibility Scores Information Seeking Average Percentage of Sources Sought Scalar Mean of Seeking Percentages Information Gap Information Satisfaction Unchanged Scale Complexity of Understanding of Flood Processes Scale Based on Categories of Coded Responses Self Rated Flood Knowledge Unchanged Scale Self Rated NFIP Familiarity Unchanged Scale Knowledge Correct Belief Regarding SFHA Residence is 1 Binary Based on Spatial Concordance General Control Unchanged Scale General Worry Unchanged Scale Cognitive Factors General Outlook Individual Responsible is 1 Binary Based on Coded Response
137 Table 6.2. Measurement of Outcomes Factors Listed Variable Measuring Variable(s) Explanation Perception I Understanding of Uncertainty Understanding of Uncertainty in Timing and Size of Floods Binary Based on Responses of Â“AllÂ” Perception of Threat Medium or High Risk Binary Based on Rank Perception II Flooding a Top Concern Binary Based on Median General Has Insurance Unchanged Binary Modified House or Property Binary Based on Coded Response Prepared for Flood Event Binary Based on Coded Response No Mitigative Action or Consideration Binary Based on Coded Response Event Specific Evacuation Unchanged Binary Behavior Protected Home or Property Binary Based on Coded Response General concern regarding flooding was originally measured on an agreement scale of one to six. A binary was created using the median as a cut point. Medians specific to the hundred and five hundred year floodplains were used in subset analysis. Not surprisingly, median concern was higher in the SFHA subset than in the sample as a whole and lower in the five hundred year floodplain. Binary logistic regression was performe d on each outcome variable for the entire sample and for each of the four spatial subset s. This division allowed for the exploration of model stability and subset patterns across outcomes. Potential differences in the effects of variables over space were also easier to identify. L ogistic regression makes fewer assumptions about the distribution of indepe ndent variables than OLS regression and was more appropriate for the data set (Garson, 2007). Ordina l scales or modified Likert scales
138 with two fixed endpoints were treated as interval when there were at least five points. All but one of these scales (damage severity) ha d six or seven points. Regressions were not performed on outcome variables with le ss than ten cases in each category. Prior to each regression analysis, Spearm an rank correlations were run for the outcome variable of interest and all of the situational and cognitive variables. Correlations were performed fo r each of the groupings. Only variables with correlations significant at the 0.05 level were include d in the next step of analysis. Because the research was somewhat exploratory, the number of predictor variables was large, and the goal was pars imonious models, a forward stepwise method employing first order terms and based on Li kelihood Ratios was used. Variables were entered at p=0.05 and removed at p=0.10. After a basic model was created, variables with correlations significant at the 0.10 level were added and the st epwise regression was then run again. Tables 6.3 through 6.10 contain information re lated to the final first order models of each outcome variable. Unless otherwise note d, Beta coefficients were significant at the .05 level, standard error was less than half Beta, and p>0.05 for the HosmerLemeshow goodness of fit test (if the test was appropriate). Several participants did not give their income level. If income was correlated to the outcome variable being investigated, it was included in the stepwise procedure. If income was not included in the final model, the stepwise procedure was run again without it in order to include more cases. Understanding is addressed first, thr eat perception second, and behavior third.
139 Perception I: Understanding of Flood Related Uncertainty In each of the initial regressions, the FRI score helped pr edict variation of understanding of uncertainty. In Vestal and the floodplain groupi ngs, it was the only significant predictor. In Union, FRI oversha dowed the roles of se verity and general control. Using these two variables instead im proved classification and NagelkerkeÂ’s R. In the total sample, general control was a dded to the model in a second regression and improved its usefulness. Final model in formation is included in Table 6.3. Table 6.3. Logistic Regression for Understanding Uncertainty over Time and Space Grouping Model Utility % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=114) 19.307** .210 77.945.764.9FRI .173 1.189 General Control -.281 .755 Constant -1.342 U (N=60) 14.652** .294 81.156.571.7Severity .612 1.844 General Control -.479 .619 Constant -.286 V (N=54) 6.270* .147 87.160.975.9FRI .154 1.166 Constant -1.970 100 (N=50) 4.977* .126 69.258.364.0FRI .172 1.188 Constant -2.313 500 (N=64) 8.003** .162 88.136.470.3FRI .168 1.183 Constant -2.097 p 0.05 ** p 0.01 FRI had a moderate positive correlation w ith severity in all spatial groupings; generally speaking, the higher the level of experienced impact, the more sources and
140 types of information a person came in contact with. This makes sense, but the consistency with which FRI appears suggests that it serves as more than just a surrogate for severity. A better developed information infrastructu re may indeed aid in the understanding of flood related uncertainty, perhaps by making othersÂ’ experiences and knowledge bases more accessible or providing corroboration. Na gelkerkeÂ’s R for FRI alone ranged from approximately 0.13 percent to 0.17. NagelkerkeÂ’s R is an approximation of standard OLS R and is usually somewhat lower (G arson, 2007). Betas and associated exponents were also similar across categories, indicati ng stability and increasing generalizability. The models including gene ral control explained more of the variation in understanding than those consis ting of just FRI. The relatio nship was negative, however. The less control people felt they had over what happened to them, the more likely they were to be categorized as understanding fl ood related uncertainty. A general sense of control had no statistical relationship to sever ity or FRI. Rather than reflecting a cognitive understanding gained through ei ther information networks or direct experience, the Â“understandingÂ” related to control is a result of general attitude Many who rated their personal control low made blanket statemen ts like Â“Anything can happenÂ” when asked about floods occurring multiple times per year or changing size. This type of understanding can not be taught and, if fatalis tic, might discourage action. However, it is not clear that either type of understanding is associated wi th perception of threat or mitigative behavior. Understanding of uncertainty had the lowe st average NagelkerkeÂ’s R and the lowest average percentage of cases categori zed correctly of any out come variable. While FRI, general control and perhaps experi ence may be consiste nt predictors of
141 understanding, most of the variance remained unexplained. A different measure of understanding of uncertainty (i .e. not based on specific descriptions) may lead to better models. However, it also app ears that this research and the existing literature do not address some key factors that may contribu te to the practical understanding of flood related uncertainty over time a nd space. More concentrated wo rk in this area is needed. Perception II: Perc eption of Threat Perception of flood threat was evaluate d through two variables: perception of a medium or high risk of future flooding and the evaluation of flooding as a top concern relative to other life concerns. All situationa l and cognitive factors illustrated in Figure 2.1 were represented in correlations to pe rception of medium-high flood risk and/or flooding as a top concern. However, after logi stic regressions were performed and the effects of other variables controlled for, only variables in three broad categories remained. Final models for both outcome va riables reflect the influences of socioeconomic, cognitive, and experience factors. Table 6.4 includes model information for perception of medium-high risk (or high risk in the case of the SFHA subset). Results overlapped for the sp atial groupings of All, Union, Vestal, and the five hundred year floodplain. The odds of medium-high risk perception increased up to 3.94 times (in the five hundred ye ar floodplain) with each level of impact severity. Of the three group models in which it appeared, severity was least influential in the Union model (odds increased 2.58 times with each level of severity), which, unlike the other two, did not include gender. Gende r appeared in the model for the Vestal subset as well, though no measure of experien ce did. The odds ratio
142 for gender was also the highest in Vestal; be ing a woman increased the odds of mediumhigh flood risk estimation by more than a factor of eight when the effects of information seeking and satisfaction were controlled. Table 6.4. Logistic Regression for Medium-High Flood Risk Grouping Model Utility % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=114) 49.624** .475 72.983.378.9Female 1.370 3.936 Severity 1.212 3.362 Years at Home -.047 .954 Constant -2.995 U (N=60) 38.209** .629 82.181.381.7Severity .949 2.582 NFIP Familiarity .447a 1.564 Age -.055 .947 Constant .034 V (N=53) 21.198** .452 63.279.473.6Seeking .046 1.047 Female 2.120 8.335 Info Satisfaction -.781 .458 Constant 1.948 100 (N=50) 12.101** .305 88.680.086.0Times Flooded 1.559 4.752 Constant -3.076 500 (N=64) 29.607** .500 81.673.178.1Female 1.632 5.116 Age -.061 .941 Severity 1.372 3.944 Constant .060 a SE is .227, slightly higher than half of Beta p 0.05 ** p 0.01
143 Age was a successful predictor of flood risk perception in both Union and the five hundred year floodplain. Age was negatively corre lated to the percepti on of a medium or high risk of future flooding. Researchers often use age as a measure of vulnerability due to decreased mobility, physical and mental health issues, and fixed incomes. However, older individuals may not see themselves as particularly vulnerable to certain hazards because the timeframe in which a disaster could happen seems to shrink with age, especially if an event has recently occurred. Th is attitude was apparent in comments like Â“I wonÂ’t be around for the next one.Â” If attit ude is related to prot ective behavior, this could be a problem, since the elderly may have more trouble coping physically and financially with the impacts of hazards. In the total samp le, length of residence rather than age was part of the final mode l due to a sligh tly higher model , NagelkerkeÂ’s R, and categorization rate. The vari ableÂ’s practical function in the model was very similar to that of age in the others, though, and may reflect an analogous influence. In the SFHA subset, the frequency rather than the severity of flood experience was included in the model. No other variables appeared to be signifi cant predictors of high flood risk estimation. The only model in which the odds of medium-high perception did not increase with experience was that of Vestal, in which cognitive variables were more prominent. In Vestal, the odds of per ceiving the risk of future flooding to be medium or high increased with the percenta ge of sources sought and decreased with higher information satisfaction levels. High s eeking and a perceived information gap may indicate a higher level of involvement a nd relevance (Grasmuck and Scholz, 2005) and their inclusion as predictors of risk percepti on levels is supported by literature. However,
144 gender, age, and the severity and/or frequency of experience ap pear to be more consistent predictors when control ling for other variables. Severity, increased age, and gender were also included in models for flooding as a top concern (Table 6.5). The relationship be tween gender and top concern was similar to the one between gender and risk estimation. Age did not follow the same pattern. Before general worry was added to the Union model, age had been included, and was associated with a decrease in odds. In Vestal, however, every additional year of age increased the odds of perceiving flooding as a top concern by about eight pe rcent. This disparity may result from higher numbers of individuals moving into flood prone areas like Castle Gardens specifically for retirement, but more research is necessary. Of the three main predictors of medium-high risk perception, only seve rity of impact occurred in more than one model related to flooding as a top concer n. Increases in severity levels were most influential in the hundred year floodplain, wh ere impacts were most extreme, but were not significant in either the five hundred year floodplain or Vestal. In these two spatial groupings, higher rates of information seeking led to increased odds of perceiving flooding as a t op concern. Information seeking was the only predictive variable in the five hundred year grouping. In Vest al, higher levels of general worry were also associated with increased odds. The rate of per uni t increase in odds was higher than in the other two groupings in whic h general worry appeared (total sample and Union). Both of those models included severity as well. The remaining variables were cognitive and dealt with sp ecific knowledge and self rated knowledge. The direc tion of self rated knowledge co ntradicts literature (Loges, 1994; Grasmuck and Scholz, 2005). However, in the hundred year floodplain, it is not
145 clear that a correct assessme nt of SFHA status reflected learned and processed knowledge or a more heuristic reaction base d on affect or experience. As a result, it is difficult to say whether knowledge is linked to attitude in this instance. Bo th variables occur only once, however, and do not appear to be as good genera l predictors of relative flood concern as severity, general worry, and, perhaps, information seeking. Table 6.5. Logistic Regression for Flood as a Top Concern Grouping Model Utility % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=112) 45.071** .442 76.366.071.4General Worry .690 1.994 Severity .706 2.026 Self Assessed Flood Knowledge .392 1.480 Constant -5.209 U (N=60) 21.294** .398 73.373.373.3Severity .862 2.369 General Worry .435 1.545 Constant -3.863 V (N=54) 29.896** .569 90.079.285.2General Worry 1.349 3.854 Seeking .041 1.042 Age .074 1.076 Constant -9.320 100 (N=50) 19.919** .439 69.677.874.0Severity 1.129 3.093 Correct SFHA 2.236 9.357 Female 1.642 5.165 Constant -7.028 500 (N=64) 5.969* .119 67.656.762.5Seeking .023 1.023 Constant -1.102 p 0.05 ** p 0.01
146 Behavior: General The final outcome factor addressed in this set of analyses was behavior. Behavior variables were grouped into two categories: general and event specific. Results for general behaviors are presented first. Regressions were run for three general mitigative behaviors as well as for no prot ective action or consideration of flooding at all. In this study, individuals who talked to others in their neighborhood about flooding before moving in were considered to have taken so me action. The three behaviors analyzed were insurance purchase, preparation for a flood ev ent (having a pump, having a plan, etc.) and modification of home or prope rty to reduce the impacts of flooding (raising utilities, using flood resistant materials, installing valv es, etc.). All three behaviors could have been undertaken after the June floods; th is analysis did not account for timing. Insurance Purchase A total of ten variables in four broad factors (locati on, experience, cognitive, FRI) were correlated with insurance purchase. When the effects of other variables were controlled, only three were included in any of the models (Table 6.6). A regression was not run for the five hundred year floodplai n grouping because only six individuals claimed to have insurance. Given the federal requirements regardi ng flood insurance and mortgages in the hundred year floodplain, it is not surprising that SFHA residence was a strong predictor of insurance purchase across spa tial categories. Combinations of residence, severity, and NFIP familiarity resulted in an approximate R ranging from 0.639 to 0.655 and
147 categorization rates of over 85 percent. Th ese models did a good job of categorizing respondents that did not have insuran ce and as well as those that did. The model for the hundred year floodplai n included only severity and was less sensitive; only 25 percent of individuals without insuran ce were classified correctly. NagelkerkeÂ’s R was less than half that of the other three models. Instead of signifying a weak model, however, I believe it may indicate the predictive strength and Â“big pictureÂ” stability of the variables in the remain ing models, especially SFHA residence and severity of experienced impact. A consider able proportion of general variation was explained through the grouping itself. The la rger unexplained varia tion is probably due to much more highly individualized (and more difficult to measure) characteristics. The level of conceptual detail would need to be in creased substantially in order to capture it. As with previous outcomes, experience va riables were absent from the Vestal model; cognitive variables were included instead. Severity of impact and the level of familiarity with the NFIP were relate d (rho 0.521, p<0.001), but both variables were included in the model for the total sample. This indicates that NFIP familiarity has predictive value for insurance purchase outside of its relationship with impact severity. Other research has shown that public familiari ty with the NFIP is generally low. The results of this study may show the practical value of campaigns to increase the visibility of the NFIP and its components. However, familiarity was self rated, and it does not reflect an understanding of the NFIP. While tempting, no conclusions can be made about the relationship of understanding and behavior with regards to insurance purchase.
148 Table 6.6. Logistic Regression for Insurance Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=114) 72.892** .641 90.079.586.0In 100 2.115 8.290 Severity .691 1.995 NFIP Familiarity .364 1.439 Constant -4.311 U (N=60) 38.243** .655 92.580.088.3In 100 2.406 11.088 Severity 1.079 2.941 Constant -4.903 V (N=54) 35.043** .639 83.387.585.2In 100 2.897 18.124 NFIP Familiarity .503 1.654 Constant -3.150 100 (N=50) 11.143** .299 25.097.480.0Severity 1.175 3.237 Constant -2.113 500 (N=64) Only 6 have insurance p 0.05 ** p 0.01 Modification of House or Property Variables from all five broa d factors were correlated with the modification of a home or property. Both SFHA residence and impact severity were correlated to modification in the total sample and in both communities, but only severity appeared in each of the models (see Table 6.7). Since seve rity was not correlated to modification in the hundred year grouping, but was in each of the other spatial groups, it may be that the functions of severity and SFHA residence with regards to property modification overlap. The conceptual model does not capture the re st of the variation in the hundred year floodplain grouping.
149 For the total sample, severity of impact and income were the strongest predictors of modification; in both cases the odds of action increased with higher levels. The same variables were included in the Union model, though their relative power was somewhat different. Income was not correlated to any other outcome variables, including insurance, in any spatial grouping. Both insurance pur chase and property modification represent financial expenditures of vary ing degrees, but regulatory requ irements appeared to trump financial considerations in the purchase of in surance. This result indicates that regulation can induce Â“properÂ” behavior, and works posi tively within regulated space. However, outside the regulated space (the SFHA), specific policy requirements may actually discourage insurance purchase. The adoption of mitigative behaviors other than insurance purchase was more widespread (see Tables 5.35 and 5.36). About three quarters of those in the hundred year floodplain had insurance; less than ten percent of five hundred year floodplain residents had purchas ed it. Double that number had made some modification to their home or property in order to reduce flood impacts, while a much smaller percentage (42 percent) of SFHA residents had modified their property than had purchased flood insurance. Some may consider it a trade off. Severity was also included in the model for Vestal. The increase in odds related to impact severity levels was higher in Vestal th an in either Union or the total sample. The Vestal model did not include demographic vari ables, but FRI and cognitive variables had predictive value. These factors were also re presented in the five hundred year floodplain model, though the specific variables diffe red. The Vestal model had a much higher approximate R than the models of the ot her spatial groupings and more successfully categorized respondents who had made modifications.
150 The odds of modification increased with hi gher levels of sour ce credibility and greater complexity of understa nding of flood processes. From a theoretical perspective, these relationships make sense; if people ha ve confidence in the sources telling them what their options are, theyÂ’re more likely to take advice, whether it comes from friends or organizations (assuming itÂ’s not contradi ctory). The inclusion of complexity of understanding in the model may support the contention that knowledge and understanding can be linked to behavior. Howeve r, it is not clear exactly how that link might function in this context, since comple xity of understanding was not predictive of either understanding of flood related un certainty or threat perception. Table 6.7. Logistic Regression for Mo dification of House or Property Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=100) 25.270** .322 91.728.674.0Severity .947 2.577 Income .432 1.540 Constant -4.691 U (N=51) 13.700** .331 85.750.074.5Severity .788 2.200 Income .570 1.769 Constant -4.621 V (N=54) 28.504** .592 92.373.387.0Severity 1.686 5.396 Credibility 1.368 3.929 Cause Types 1.185 3.270 Constant -15.254 100 (N=50) No Improvement 500 (N=64) 15.292** .343 98.141.787.5FRI .270 1.310 Correct SFHA -1.657 .191 Constant -3.052 p 0.05 ** p 0.01
151 In the five hundred year floodplain, FRI was predictive of both understanding of uncertainty and property modi fication. This may show the influence of information networks on both understanding and behavior, especially in areas less consistently impacted. More focused study of what modifi cations participants made, where they got the ideas for specific modifications and th e perceived credibility of those specific sources, as well as the relations hip to specific types of per ceived flood causes, would help to determine if these predictors were more th an just anomalies. The negative relationship of correct SFHA categorization to modificati on in the five hundred year floodplain could point to either an affective (rather than cognitive) assessme nt and/or show the indirect effects of regulation. Flood Preparation Regressions were also ru n for flood preparation. Preparation might consist of maintaining an emergency kit, storing all im portant papers togeth er in a waterproof container, or simply having a plan. In Vestal only seven people had prepared and in the five hundred year floodplain, only six, so thes e groupings were not included in analysis. Significant correlations were f ound only in the total sample a nd in Union. In both spatial sets, stepwise regression resulted in only one predictive variable, se verity of impact. The models themselves were very similar. However, NagelkerkeÂ’s R was 0.13 in the whole sample and 0.16 in Union. Additionally, th e model did not categorize any of the affirmative responses correctly. Overall, the c onceptual model appeared not to adequately explain preparation for a flood event. The poor performance might also be due, in part, to
152 the fact that flood preparati on activities were not explicit ly listed on the card given participants; these activities may have been underreported. In addition to discovering general predictors of mitigation, it may also be important to explore characteristics associat ed with inaction. Vari ables representing all five broad factors were correlated to a lack of any mitigative action or consideration of flooding. The resulting models were consiste nt in their composition, however (see Table 6.8), and included variables from only three fa ctors. No regression for the hundred year floodplain grouping was run because only two SF HA residents did not take action of any kind. This resulted in a complete separation in the Vestal groupi ng; SFHA residence was not included in VestalÂ’s mode l, but should be considered a predictive variable. Its inclusion in UnionÂ’s model was also somewh at problematic. In both groupings, stepwise regression selected both variab les, but standard errors we re large and at least one exponential confidence interval included one. In Union, using SFHA residence instead of frequency of flooding resulted in a higher a ffirmative categorization rate, but lower model and approximate R. Experience and location are obviously rela ted, but the magnitudes of correlation (rho) in this sample were moderate, ranging from 0.371 (p=0.004) in Union to 0.498 (p<0.001) in Vestal. I believe SFHA residence in this case represents the effects of regulation as well as those of experience. Additionally, while impact severity was also related to flood frequency (and location), it di d not appear in any of the models. Each flood experienced decreased the odds of no ac tion by anywhere from sixty to 72 percent when combined with seeking, and up to 92 percent if seeking was not in the model. In fact, not one person who had experienced mo re than one flood did nothing. Increased
153 seeking behavior also decreased the odds of in action in three out of the four models. The predictive variable was calculated by averag ing the percentages of information sources sought during and after the June floods a nd was intended to represent a general disposition towards seeking. It is not surprising that this va riable would be a consistent and significant predictor, since information s eeking prior to moving in was treated as an action in this analysis. Table 6.8. Logistic Regression for No Action or Consideration of Flooding Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=114) 42.648** .456 90.566.784.2In 100 -2.413 .090 Seeking -.020 .980 Times Flooded -.910 .402 Constant 1.130 U (N=60) 20.326** .413 90.758.881.7Seeking -.026 .974 Times Flooded -1.245 .285 Constant 1.499 V (N=54) 19.203** .448 97.630.881.5 Times Flooded -2.495 .082 Cause Types -1.420 .242 Constant 2.815 100 (N=50) Only 2 took no action 500 (N=64) 17.784** .325 77.875.076.6Seeking -.028 .972 Times Flooded -.903 .405 Constant 1.428 p 0.05 ** p 0.01
154 Behavior: Event Specific Data were collected on specific actions taken during the June, 2006 floods as well as general mitigative behavior for two reasons. First, evacuation behavior is of particular interest to emergency managers. Second, ac tions such as evacuation, sand bagging, etc. become part of experience and may in fluence general flood related understanding, attitude and behavior. Models for evacuati on and the protection of house or property are presented here. Some of the general variables outlined in Table 6.1 were adjusted to be relevant to the specific event and to make the models mo re useful for emergency managers. First of all, severity was not included as a predicto r of evacuation because evacuation itself was incorporated into the impact scale. Second, the number of floods a respondent had experienced was altered to better reflect expe rience levels at the time of the flood; the June flood itself was not included. Third, s eeking percentages a nd FRI totals were adjusted to comprise only information s ources and types gathered during the flood. Additionally, two specific information types we re included in analys is (pre-evacuation warning information and official evacuation orders) in order to gage their impact. Finally, two cases were exclude d from analysis; one person m oved into his current home after the flood and another was out of the state during June. Neith er could decide to evacuate or protect thei r home or property. Evacuation Regressions were run for evacuation in th ree spatial groupings. Only two eligible SFHA residents did not evacuate and only seve n of the 53 respondents from Vestal chose
155 to remain in their homes. The fire department and police issued official evacuation orders to five hundred year floodplain residents in Twin Orchards and Castle Gardens, and in some cases, physically removed individuals in the SFHA to waiting cars or ambulances. Several individuals evacuated prior to official notice, but those who stayed had to sign waivers. Official evacuation orders were more spatially limited in unincorporated Union. Because of the different approaches to evacuation in the two communities, it was expected that odds of evacuation would in crease with Vestal residence. However, correlations between Vestal and official evacu ation orders were moderate, even outside the SFHA. Rho was 0.391 (p<0.001) in the tota l sample and 0.501 (p<0.001) in the five hundred year floodplain. Offici al evacuation orders increas ed the odds of evacuation more than Vestal residence in both the tota l sample and the five hundred year floodplain (see Table 6.9). Official evacuation orders are not in cluded in the Union model presented, but were very effective. Everyone who recei ved them evacuated, which resulted in a complete separation. In the Union model, even t specific FRI is basically standing in for official evacuation information in order to pr esent a model that fits the imposed criteria and can be easily interpreted. When the offici al order was included instead of event FRI, NagelkerkeÂ’s R improved to 0.768 with an 89.8 classification rate. It should be considered a key predictor across all spatial categories. SFHA residence was also associated with large increases in odds in both Union and the total sample. In the Union model pr esented, the odds of evacuation increased by 55 times if an individual lived in the hundred year floodplain. This can be explained by the somewhat targeted evacuation of Uni on and perhaps a higher level of general
156 awareness. The effect of SFHA residence can al so be seen in the drop in approximate R in the five hundred year fl oodplain grouping. Model fit in all spatial categories was good, though. Table 6.9. Logistic Regression for Evacuation Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=112) 64.666** .624 78.891.187.5In 100 3.421 30.607 Vestal 1.584 4.873 Age -.051 .950 Official Evac 2.063 7.866 Constant 1.440 U (N=59) 38.566** .643 80.887.984.7In 100 4.013 55.310 Age -.060 .942 Event FRI .303 1.354 Constant .799 V (N=53) Only 7 didnÂ’t evacuate 100 (N=48) Only 2 didnÂ’t evacuate 500 (N=64) 30.355** .504 83.978.881.3Vestal 1.573 4.821 Age -.057 .945 Official Evac 1.930 6.888 Constant 1.784 p 0.05 ** p 0.01 A possible cause for concern was the cons istent inclusion of age in evacuation models. Increased age was associated with d ecreased odds of evacuation in all three. If we assume that health problems are more likely and mobility and general strength decline with age, older adults are some of the pe ople that might benefit most from evacuating
157 during a serious event. Failure to do so could be related to absence of money or family nearby, attachment to home, lack of a car, informational isol ation, or a number of other factors. In targeting this population, however, care must be taken not to exacerbate the very conditions that might make them vulne rable. One elderly c ouple I spoke with in Vestal was still righteously angry and stil l recovering from the rough treatment they received from police during the evacuation. Protection of House or Property Unlike evacuation, demographic variables were not correlated to the active protection of structures or be longings during the June floods. The other four factors were represented in correla tions. Regression resulted in simila r models for all but one spatial grouping. However, the differences between the five hundred year group and the other models, especially th at of the SFHA subset, are important. Self rated flood knowledge was associated with higher odds of protective action in the total sample, in Vestal, and in th e hundred year floodplain (see Table 6.10). This may reflect a combination of influences, from experience to greater awareness of past flood levels and impacts to more confidence in knowing what to do in the event of flooding. Severity of impact or SFHA residenc e was included in four out of the five models, but never together. If hundred year floodplain residence by itself was a key predictor, one would expect the approximate R to be lower in the SFHA grouping. It was not. While the model was less sensitive th an those including hundred year floodplain residence, NagelkerkeÂ’s R was higher, as it was in the Union model that included severity rather than floodpl ain residence. In Union, the combination of severity and
158 general control resulted in classification rates above 75 percent for both Yes and No categories. Table 6.10. Logistic Regression for Protect ion of Home or Personal Property Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=112) 31.902** .331 75.070.072.3In 100 1.880 6.553 Flood Knowledge .411 1.509 Constant -2.496 U (N=59) 20.306** .389 78.677.478.0Severity .769 2.157 General Control .452 1.572 Constant -3.735 V (N=53) 14.182** .314 70.872.471.7In 100 1.702 5.485 Flood Knowledge .437 1.548 Constant -2.378 100 (N=48) 14.318** .403 30.097.483.3 Flood Knowledge .813 2.255 Severity .821 2.273 Constant -5.521 500 (N=64) 10.535** .210 90.540.973.4 Pre-Evac Warning -2.351 .095 Correct SFHA -1.306 .271 Constant .604 p 0.05 ** p 0.01 The combination of hundred year floodpl ain residence and self rated flood knowledge may in part reflect the indirect e ffects of regulation. For two reasons, though, I think that in these models, SFHA residence and impact severity al so both practically
159 represented flood levels. First, the impact se verity scale incorpor ated evacuation (those who evacuated, but did not suffer flood damage we re assigned an impact level of one, not zero), but widespread evacuat ion outside of the SFHA occurred only in Vestal. Second, most of those within the SFHA did experien ce flooding to some degree. The inclusion of general control in the Union model makes sens e, but it was not correlated to protective action in any other grouping. The model for the hundred year floodplain subset illustrated the inconsistent effects of variables over space. The correct assessment of SFHA status had a negative association with protection of property during an event, ju st as it did with property modification in the analysis of general beha vior. This Â“knowledgeÂ”, whether affective or cognitive, had the opposite association with relative flood concern in the hundred year floodplain. More interesting, how ever, was the decrease in odds of protection when individuals in the five hundr ed year floodplain received pre-evacuation warnings or information. Of the 13 people in the five hundred year floodplain that received this kind of information, only one took any precautions rela ted to their home or possessions. In the hundred year floodplain, on the other hand, A LL who received or sought pre-evacuation information did something to protect thei r belongings, whether it was sandbagging the house or moving boxes and furniture to highe r elevations. Because this circumstance resulted in a complete separation, pre-evacu ation information was not included in the SFHA model. This type of information, t hough, may have given residents time to take action before the water rose too high or an official evacuation order was given. In many
160 cases, those that didnÂ’t receive pre-evacuation in formation said they tried or wanted to do something, but there just wasnÂ’t enough time and they had to get out quickly. In the five hundred year floodplain, respondents appeared not to use the time to protect their belongings or home, but perh aps used it to make arrangements for evacuation. Approximately 69 percent of fi ve hundred year floodpl ain residents who received pre-evacuation information left thei r homes. These individuals did not seem to believe that their possessions were at risk, though over half of them experienced moderate or high impact. In addition to ge neral attitude and the potential effect of regulation on risk perception, it would be in teresting to look at th e specific construction of messages from officials and other sources. What exactly was said? How might it relate to behavior? These data were not collect ed, however, so their influence cannot be assessed. Relationships of Outcome Variables One purpose of this project was to expl ore the relationships between situational and cognitive variables and understanding, at titude, and behavior. These results were presented above. A second purpose was to l ook at connections between the outcome factors themselves. In this sample, was there an asso ciation between understanding of flood related uncertainty, perception of flood threat and general mitigative behavior? Would inclusion of these factors and event specific actions improve the basic models presented above? The steps of analysis were similar to t hose described for previous analyses. First, correlations were run for each outcome variable Individual regressions were then run for
161 each variable with a relationship sign ificant at the 0.05 level. If model was significant and the model guidelines met, then the variable was added to the set of predictors listed above and another stepwise regression run. A linear progression fr om understanding to attitude to behavior was not assumed, but individual general behaviors were not regressed on one another. The final step of analysis identified outcome factors with explanatory power beyond that of related situational and co gnitive variables. In most cases, there was no model improvement. Analyses were run for each of the nine outcome variables in all five spatial groupings. Only three models were clearl y improved by the addition of variables measuring understanding of uncertainty, threat perception, or behavior. The results of two others were mixed. In four out of the five cases, threat per ception helped to better explain variation in behavior. In the other model, risk perception aided in the prediction of relative concern. Table 6.11 contai ns the three improved models. In the original five hundred year floodpl ain model for flooding as a top relative concern, only one significant explanatory variab le was identified. In the improved model, seeking percentage dropped out and was repl aced by perception of medium or high risk. While still low, approximate R increased and both affirmative and negative classification improved. Flooding as a top concer n was also included in step wise regressions for risk perception in all groupings except the SFHA, in which the two were not significantly correlated. None were improved. However, it is not really clear that a perception of medium or high risk leads to a hi gher level of relative concern.
162 Table 6.11. Improved Models by Ou tcome Variable and Grouping Outcome Variable and (Grouping) Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp Flooding as Top Concern (500) 8.978** .175 76.5 60.0 68.8 Perception of Med. or High Risk 1.584 4.875 Constant -.773 Modified Home/ Property (100) 5.360* .137 82.8 47.6 68.0 Perception of High Risk 1.473 4.364 Constant -.780 Protected Home/ Property (500) 14.787* .285 90.5 40.9 73.4 Pre-Evac Warning -2.608 .074 Correct SFHA -1.388 .250 Flood Top Concern 1.213 3.363 Constant .063 p 0.05 ** p 0.01 Perception of high risk in the hundred year floodplain increased the odds of property modification by almost four and a ha lf times. NagelkerkeÂ’s R was not high, but no predictive variables were id entified in the initial analys is. The groupings unrelated to floodplain status included impact severity in th eir models; it is likely that some of this influence was accounted for by SFHA residence. Classification rates for the new model were broadly similar to those of the other groupings. This result supports the contention that threat perception may infl uence behavior, but this spec ific improvement occurred in only one model. Flooding as a top concern, however, impr oved the five hundred year floodplain model for protection of property during th e 2006, June floods. Top concern did not
163 replace any situational or c ognitive variables, but was si mply added. Approximate R increased; classification percentages were iden tical. Stepwise regressi on also resulted in the inclusion of flooding as a top concern in evacuation models for both the total sample and for Union. The extent of model improvement was debatable, though. In the Union grouping, high relative flood concern took the place of age and event FRI and NagelkerkeÂ’s R for this more pa rsimonious model was slightly higher (0.669). Overall classification was the same as in the original model. However, in Union, everyone who received official evacuation orde rs did evacuate, and when this variable was included in stepwise regression, high concern regarding floodi ng had no effect. In the total sample, the addition of flooding as a top concern increased NagelkerkeÂ’s R to 0.653. As with the Union model, overall cate gorization was the same, but the spread between Yes and No percentages was larger. St andard error was less than half Beta. The concern with this model was that the p-va lue of the Wald stat istic was 0.053 and the exponential confidence interval included one, which would indicate no effect. The evacuation models are somewhat ambiguous, but the overall evidence points to an association between higher threat perception and increased odds of mitigative behavior. Summary and Conclusions This set of analyses explored the co mbined effect of socio-economic status, location, experience, flood risk information in frastructure and cogni tive factors on flood related understanding of uncertainty, thre at perception and mitigative behavior. Associations between understa nding, attitude and behavior were also analyzed. The purpose was to identify factors (and specific variables) that best explained variation in
164 perceptual and behavioral outcomes. By split ting the sample into spatial groupings, both general and location specific patterns we re identified. Figures 6.1 and 6.2 illustrate revised models, the first including genera l mitigative behaviors and the second event specific behaviors. They are split to increase legibility and because any effects of event specific behavior on general behavior and vi ce versa appeared to be mediated by or through the other situational and cognitive f actors. Relationships are positive unless otherwise noted. Relationships with stronger, more consiste nt evidence are represented with solid lines; dotted lines illustrate iden tified relationships with weaker evidence. The general conceptual model outlined in Figure 2.1 did an adequate job of identifying key contributing factors for most outcome variables. Models for insurance purchase, evacuation, perception of risk and no mitigative action were fairly strong and consistent. A second tier of models included those describing high re lative flood concern, protection of home or propert y and modification of home or property. Models illustrating the understanding of flood related uncertainty and flood preparation were weak, however, and need substantial improvement. Reconcep tualizing what cons titutes understanding might help to identify key associations with situational and cognitive factors and better evaluate the relationship of understa nding to attitude and behavior. Four areas should be addressed in future research. First, all models would benefit from improved measurement of understandi ng. Second, interactions and higher order functions should be employed where appropriate. Third, flood risk infrastructure needs to be detangled and frequency included. A lot of valuable information is currently hidden in the aggregation. Lastly, the influence of Â“out comeÂ” variables on situ ational and cognitive factors needs to be investigated while taking into account the passage of time.
165 Figure 6.1. Relationships between Situatio nal and Cognitive Factors and Understa nding, Attitude and General Behavior* LOCATION Town Floodplain Status SOCIO-ECONOMIC Gender Age Income EXPERIENCE Flood Frequency Impact Severity FLOOD RISK INFRASTRUCTURE FRI Credibility COGNITIVE General Control General Worry Complexity of Understanding Self Reported Knowledge NFIP Familiarity Correct SFHA Belief Info Seeking UNDERSTANDING Understanding of Flood Related Uncertainty over Space and Time THREAT PERCEPTION Medium/High Flood Risk Flooding is a Top Concern GENERAL MITIGATIVE BEHAVIOR Insurance Purchase Modification of Home or Property Preparation for Event No Mitigation or Consideration LOCATION Town Floodplain Status SOCIO-ECONOMIC Gender Age Income EXPERIENCE Flood Frequency Impact Severity FLOOD RISK INFRASTRUCTURE FRI Credibility COGNITIVE General Control General Worry Complexity of Understanding Self Reported Knowledge NFIP Familiarity Correct SFHA Belief Info Seeking UNDERSTANDING Understanding of Flood Related Uncertainty over Space and Time THREAT PERCEPTION Medium/High Flood Risk Flooding is a Top Concern GENERAL MITIGATIVE BEHAVIOR Insurance Purchase Modification of Home or Property Preparation for Event No Mitigation or Consideration *Relationships are between variables, not factors, and are pos itive unless otherwise noted. Relationships with more consistent evidence are represented with solid lines; dotted lines illustrate identified relationships with weaker evidence.
166 Figure 6.2. Relationships between Situatio nal and Cognitive Factors and Understa nding, Attitude and Event Behavior* LOCATION Town Floodplain Status SOCIO-ECONOMIC Gender Age Income EXPERIENCE Flood Frequency Impact Severity FLOOD RISK INFRASTRUCTURE FRI Credibility Pre-EvacInfo Official Evac COGNITIVE General Control General Worry Complexity of Understanding Self Reported Knowledge NFIP Familiarity Correct SFHA Belief Info Seeking UNDERSTANDING Understanding of Flood Related Uncertainty over Space and Time THREAT PERCEPTION Medium/High Flood Risk Flooding is a Top Concern EVENT SPECIFIC BEHAVIOR Evacuation Protected Home or Property -/+ LOCATION Town Floodplain Status SOCIO-ECONOMIC Gender Age Income EXPERIENCE Flood Frequency Impact Severity FLOOD RISK INFRASTRUCTURE FRI Credibility Pre-EvacInfo Official Evac COGNITIVE General Control General Worry Complexity of Understanding Self Reported Knowledge NFIP Familiarity Correct SFHA Belief Info Seeking UNDERSTANDING Understanding of Flood Related Uncertainty over Space and Time THREAT PERCEPTION Medium/High Flood Risk Flooding is a Top Concern EVENT SPECIFIC BEHAVIOR Evacuation Protected Home or Property -/+ *Relationships are between variables, not factors, and are pos itive unless otherwise noted. Relationships with more consistent evidence are represented with solid lines; dotted lines illustrate identified relationships with weaker evidence.
167 EXPLORING FIGURE 2.2: THE MOD EL OF SPECIFIC FLOOD RISK MESSAGES, SETTINGS, AND PERCEPTION Data and Methods This section addresses the second set of research questions. Figure 2.2 represents specific perceptions of flood irre gularity and threat as the in teraction of th e situational and cognitive factors outlined above, specific flood risk messages, and the cognitive setting in which the messages are introduce d. Because the response rate for the focus groups was so low, the role of cognitive setti ng was not analyzed. During the face to face surveys, participants made comments about th e descriptions (which will be discussed in the next section), but the inte rviewers did not provide additional information or remark on their interpretations. All processing was assumed to be heuristic rather than systematic. The survey contained eight questions pertaining to percei ved size, perceived likelihood, understandin g of uncertainty over time a nd space, and relative concern. Participants were asked to state which of the three described floods they thought was most and least likely, biggest and smallest, and most and least concerning. Additionally, they were asked which floods they thought could happen more than once per year and change in size over time. Multiple res ponses and Â“DonÂ’t KnowÂ” were acceptable answers. Two people said that all three descript ions referred to the same flood; six identified the one percent chance and the hundred year floods as equal. Tables 5.38 through 5.45 show the distribution of responses. This analysis also employed logistic regr ession to identify the factors that best explained the variation in the eight outcom e variables above. Re sponses of Â“AllÂ” or
168 Â“DonÂ’t KnowÂ” were also treated as outcome variables. In the original coding, each description was assigned a one if the answer was Â“AllÂ” or a zero if the answer was Â“DonÂ’t KnowÂ”. However, these cases were not in cluded in most regressions for individual descriptions. If these were included, mode ls would not adequately reflect factors influential in the choice of a cer tain description over the others. Analysis followed steps similar to th ose described in the previous section. Correlated variables were identified and then a forward stepwise regression was run with the same conditions. All the variables listed in Table 6.1 were included. Additionally, the outcome variables associated with understand ing of flood related unc ertainty and threat perception were added to the set of cognitive variables. General mitigative and evacuation behaviors were used as independent variables only in th e regressions related to concern. Regressions were r un for outcome variables with ten or more cases in both Yes and No categories. Responses to some questions concentrated heavily on one description of flood risk; in these circumst ances, models were not created for each description and used only the to tal sample. General patterns were thus somewhat difficult to identify. Results for perceived relative size are presented first, followed by perceived likelihood, uncertainty, and relative concern. Perceived Relative Size When asked to name the described flood they thought was biggest, an overwhelming majority of pa rticipants selected the hundr ed year flood. Only 14 chose another description, ten of whom picked a fl ood with a one percent chance of occurring in any year. An additional 12 said they didnÂ’t know which flood was largest in size.
169 Analyses were conducted for these three options. However, no correlations were significant at the 0.05 level for the one per cent chance flood, and onl y understanding of flood related uncertainty was co rrelated to DonÂ’t Know. Model was not significant, however. General control was th e only variable associated wi th the choice of the hundred year flood as the largest. The model improved with its inclusion, but model fit and power were poor (see Table 6.12). Table 6.12. Logistic Regression fo r Biggest Flood: 100 Year Flood N Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp N=94 4.727* .086 0 100 85.1 General Control -.421 .656 Constant 3.257 p 0.05 ** p 0.01 The results for the smallest perceived flood were somewhat more evenly distributed, though participants had a more difficult time answering the question. About thirty percent responded that they didnÂ’t know which flood was smallest, which was more than twice the rate of the question regarding the biggest flood. This result, combined with the fact that only one person said that the hundred year fl ood was the smallest and the lopsided response for the biggest flood, empha sizes the heuristic power of hundred year flood terminology. Formal and informal co mmunication has succeeded to a certain extent. The hundred year flood is familiar, resona tes, and is thought of as a big flood. It is not clear just how big Â“big Â” is though, or whether peopl e are concerned about it or
170 motivated to mitigate. Additionally, the hundred year flood may be so ingrained as Â“the big oneÂ” that other descrip tions may not register. Approximately two thirds of the partic ipants who did not respond with DonÂ’t Know and selected only one description c hose the one percent chance flood as the smallest. The remainder chose the 26 pe rcent chance flood. Tables 6.13 and 6.14 include model information for all spatial categorie s. Though 33 people said they didnÂ’t know which flood was smallest, there was not a st rong pattern; no correlations were found in the total sample, Union, or the five hundred year floodplain. In Ve stal and the hundred year floodplain, correlated variables (one each) did not result in model improvement. Table 6.13. Logistic Regression for Smallest Flood: 1% Chance Flood Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=78) 9.147** .154 34.698.176.9FRI -.155 .856 Constant 2.387 U (N=39) 5.149* .170 35.788.069.2 Information Satisfaction .236 1.645 Constant -1.738 V (N=39) 6.381* .213 25.096.374.4 Times Flooded -1.695 .184 Constant 2.367 100 (N=36) 9.873** .323 46.7100 77.8Female 1.946 7.000 Understanding of Uncertainty -1.946 .143 Constant .336 500 (N=42) 7.637** .243 36.493.578.6FRI -.219 .803 Constant 2.960 p 0.05 ** p 0.01
171 Table 6.14. Logistic Regression for Smallest Flood: 26% Chance Flood Grouping Model Usefulness % Correct Category Model N R No Yes All Variables Beta Exp All (N=78) 15.391** .251 88.736.071.8FRI .186 1.204 Female -1.105 .331 Constant -2.219 U (N=39) 5.760* .191 96.238.576.9FRI .166 1.181 Constant -2.406 V (N=39) 6.381* .213 96.325.074.4 Times Flooded 1.695 5.449 Constant -2.367 100 (N=36) 9.873** .323 100 46.777.8Female -1.946 .143 Understanding of Uncertainty 1.946 7.00 Constant -.336 500 (N=42) 9.497** .304 93.840.081.0FRI .257 1.293 Constant -3.492 p 0.05 ** p 0.01 Because of the distribution, many of the variables in the models for the smallest flood were similar. In some groupings, the only difference was the direction of association. Only one person gave the hundred year flood as the smallest flood, so after the DonÂ’t Knows and responses of All we re removed, this question essentially represented an either/or choice. Looking at the two sets of models together made the identification of importa nt variables easier. Breadth and depth of FRI was the most c onsistent predictor across the two sets of models. More developed flood risk infrastruc tures increased the odds of choosing the 26 percent chance flood and decreased the odds of choosing the one percent chance flood. Being female, on the other hand, seemed to decrease the odds of choosing the 26 percent
172 chance flood. Frequency of flooding was co rrelated to both answers in multiple groupings, but appeared only in the Vestal model, where it s directional influence was similar to that of FRI. U nderstanding of uncertainty was correlated only in the SFHA. Perceived Relative Likelihood The response patterns for the flood thought le ast likely to occur in the next year reflected the patterns of the flood thought to be the biggest. A similar number responded with DonÂ’t Know (11) and only 13 people di d not choose the hundred year flood. Most of the remainder chose the one percent chance flood. Because the response rates were so skewed towards the hundred year flood, the four spatial subsets were not analyzed. Table 6.15 contains model information for the three ch oices that received more than ten votes. Since most people picked either the one percent chance descri ption or the hundred year flood description, it is not surprising that the models c ontained similar variables with opposing effects. In this sample, it appears th at higher levels of self assessed knowledge (of flood processes or the NFIP) increased th e odds of a person c hoosing the one percent chance flood rather than the hundred year fl ood as least likely. The odds increased by 7.5 times if a person believed that all of the described floods could happen more than once per year and could change in size. This may have to do with the naming of floods experienced in the past or, perhaps, a shif t in terminology in official communications about the NFIP. Most, however, did select th e hundred year descrip tion and models were fairly weak. A choice of DonÂ’t Know was not associated with any type of knowledge or understanding. Instead, higher levels of general worry were associated with a decrease in
173 the odds that a person chose D onÂ’t Know. This may be a result of worriers being more invested in possible dangers or having gr eater general discomfort with uncertainty. Table 6.15. Logistic Regression for Least Likely Flood: Total Sample Response Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp 100 Year Flood (N=95) 16.160** .284 23.193.984.2 Self Reported Knowledge -.515 .598 Understanding of Uncertainty -2.073 .126 Constant 5.603 1% Chance Flood (N=95) 12.770** .237 98.816.788.4 Understanding of Uncertainty 2.016 7.508 NFIP Familiarity .325 1.385 Constant -4.162 DonÂ’t Know (N=114) 10.544** .188 100 0 90.4 General Worry -1.084 .338 Constant -.131 p 0.05 ** p 0.01 No relationship was found between DonÂ’t Know answers and the flood perceived as most likely. These 15 cases were removed fo r analyses of indivi dual descriptions and only three of the remaining participants c hose the hundred year flood by itself as most likely. About 30 percent chose th e one percent flood and two thirds chose the 26 percent chance flood. The models presented in Tables 6.16 and 6.17 consist of similar variables with opposite signs and once again reflect wh at was basically an either/or choice.
174 There were few discernible patterns re garding the most likely flood across the total sample and the two comm unities. Perhaps people are not used to thinking about a flood as being Â“most likelyÂ”. Once the hundred year flood was dismissed because it was considered unlikely, participants might have reacted to the numbers and timeframes in the less familiar terms. The mechanics behi nd this type of heur istic response, and subsequent results, may be difficult to predict within the project framework. Table 6.16. Logistic Regression for Mo st Likely Flood: 1% Chance Flood Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=95) No Correlations U (N=48) No Correlations V (N=47) No Improvement 100 (N=40) 5.995* .197 100 0 70.0 Understanding of Uncertainty -1.897 .150 Constant -.182 500 (N=55) 10.158** .241 94.925.074.5 Self Rated Knowledge -.470 .625 General Control .486 1.626 Constant -.690 p 0.05 ** p 0.01 The lack of correlation in the total sample and communities could indicate a difference in important variables in floodplai n groupings, though patterns may be a result of stepwise regression conforming to th e data. However, both understanding of uncertainty and self rated knowledge also appe ared in models for the least likely flood. Locus of control was a consis tent predictor in the five hundred year floodplain and, like
175 understanding of uncertainty, occurred in models of relative perceived size. The direction of influence for variables in models of bot h size and likelihood was logically consistent. More qualitative research is required to explore the possible reasons behind the relationships of these variab les to the perceived likelihood and size of particular flood descriptions. No literature exists with which to compare these results. Table 6.17. Logistic Regression for Mo st Likely Flood: 26% Chance Flood Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=95) No Correlations .186 1.204 U (N=39) 5.115* .167 60.066.764.1Income .600 1.822 Constant -1.319 V (N=47) No Correlations 100 (N=40) 5.077* .164 0 100 65.0 Understanding of Uncertainty 1.609 5.00 Constant 0 500 (N=55) 12.922** .295 35.397.478.2 Self Rated Knowledge .501 1.650 General Control -.607 .545 Constant .913 p 0.05 ** p 0.01 Uncertainty over Space and Time Participants were asked which of the described floods they thought could happen more than one time per year as well as wh ich floods they believed could change over time. Because these questions were used to evaluate overall understanding of flood related uncertainty over time and space, understanding was not included as an
176 independent variable in the following analys es. Additionally, it was expected that the models for responses of Â“All Could Chang eÂ” and Â“All Could Happen More than Once per YearÂ” would be similar to the ones for unders tanding of uncertainty presented in the previous subchapter. Indeed, FRI was corre lated in all of the groupings examining potential frequency and in four out of the fi ve groupings looking at change in size over time. FRI was a consistent pred ictor across both sets of mode ls. Additionally, three model configurations for Â“More than Once per YearÂ” were almost identical (see Tables 6.3 and 6.18). Table 6.18. Logistic Regression fo r More than Once per Year: All Grouping Model Usefulness % Correct Category Model N R No Yes All Variables Beta Exp All (N=104) 14.976** .179 56.664.760.6FRI .171 1.186 Constant -1.869 U (N=55) 15.246** .325 80.662.572.7 General Control -.486 .615 Severity .683 1.980 Constant -.189 V (N=49) 4.896* .127 40.966.755.1FRI .144 1.155 Constant -1.383 100 (N=48) 6.508* .169 68.273.170.8FRI .206 1.229 Constant -2.480 500 (N=56) 13.288** .283 77.460.069.6FRI .128 1.200 Vestal 1.340 3.818 Constant -2.472 p 0.05 ** p 0.01
177 There was little variation in response to the question of change in size over time; most participants (64 percent) said that all of them could change. Another 24 percent said they didnÂ’t know. None of the individual desc riptions had ten affirmative responses, so analyses were performed only for choices of A ll and DonÂ’t Know. Beca use of the lack of internal variation, all cases were in cluded. Tables 6.19 and 6.20 provide model information for All and DonÂ’t Know. In addition to FRI, complexity of understanding increased the odds of a participant responding that all of the descri bed floods could change over time. Direct experience or the ability to access othe r peopleÂ’s experience through information networks might seed the idea that a flood of any size could happen more than once a year. Those same networks might also provide in formation on management strategies, pump breakdowns, global warming, development plans and environmental conditions, all of which were cited as direct or indirect causes of floodi ng (see Table 5.26). Individuals who believe multiple factors interact to produce flooding may be more likely to understand that, as those factors change, so do hydrologic results. If messages are reinforced, information networks could be pa rticularly important in areas where direct experience with flooding or changing conditions is limited (i.e. parts of the five hundred year floodplain). However, none of the models was particul arly good, and in Union, general control was the only variable correlated to a belief that all the desc ribed floods could change over time. Odds decreased as a se nse of control increased, a pattern found in Table 6.18 as well. The combined results presented in Tables 6.3, 6.18, and 6.19 suggest that access to experience and information, complexity of unde rstanding of flood processes, and locus of
178 control all helped to explain understa nding of uncertainty, though the projectÂ’s conceptual framework did not adequately ad dress most of the va riation in perception. Table 6.19. Logistic Regression for Change in Size over Time: All Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=114) 16.790** .189 37.587.870.2FRI .188 1.125 Complexity of Understanding .264 1.865 Constant -1.779 U (N=60) 5.207* .111 15.890.266.7 General Control -.406 .666 Constant 2.189 V (N=54) 8.442** .196 38.175.861.1FRI .186 1.204 Constant -1.440 100 (N=50) No Improvement 500 (N=64) 9.013** .178 57.776.368.8 Complexity of Understanding 1.013 2.753 Constant -1.470 p 0.05 ** p 0.01 FRI was also correlated to answers of DonÂ’ t Know in four of the five groupings for change in size over time. In the total sa mple, in Union, and in the five hundred year floodplain, a more developed FRI decreased the odds of a DonÂ’t Know answer. Possession of a BachelorÂ’s degree functioned similarly in subset models that did not include FRI. In the hundred year floodplain, no one with a BachelorÂ’s degree or higher gave an answer of DonÂ’t Know. A well develo ped flood risk infrastructure and a degree may both provide access to somewhat specia lized information and serve to increase
179 confidence in oneÂ’s opinions. Though approxima te R was generally higher than for responses of All, model fit was relatively poor, especially in Union and the SFHA. Table 6.20. Logistic Regression for Chan ge in Size over Time: DonÂ’t Know Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=112) 17.419** .215 98.829.682.1FRI -.152 .859 BachelorÂ’s -1.608 .200 Constant .534 U (N=60) 6.129* .147 100 7.1 78.3FRI -.166 .847 Constant .289 V (N=52) 13.551* .340 87.253.878.8BachelorÂ’s -2.735 .065 Medium/High Risk Perception -1.536 .215 Constant .431 100 (N=48) 7.358** .216 100 0 77.1BachelorÂ’s Separation 500 (N=64) 15.926** .326 95.856.385.9FRI -.352 .703 Constant 1.217 p 0.05 ** p 0.01 All cases were included in the analyses of change in size over time. All data were initially retained for analyses regarding fl oods occurring more than once a year as well. After DonÂ’t Know responses were examine d, these ten cases were removed, however. Responses of All, the one percent chance flood alone, the 26 percent chance flood alone, and both the one percent chance and the 26 pe rcent chance floods were then evaluated. For this particular analysis, the above choices were coded as mutually exclusive. Only
180 one person chose the hundred year flood by itsel f as possibly occurring more than once per year, so no tests were perf ormed for this description. Because of the distributions, subset analyses were not conducted for responses of DonÂ’t Know, the 26 percent description, or the one percent and 26 percent flood combination. No correlations were found for either the one percent/26 percent combination or the 26 percent flood. Regressi on results for DonÂ’t Know and for the one percent description are included in Tables 6.21 and 6.22. As it was for responses of DonÂ’t Know re garding change in size over time, FRI was a key predictor of DonÂ’t Know answers for floods happening more than one time per year. FRI had a similar relationship to the one percent chance de scription once DonÂ’t Know responses were removed, however. Of those who had an opinion about this question (even if it was Â“NoneÂ”), people with less well developed information infrastructures were more likely to say th at ONLY the one percent chance flood could happen more than one time per year. The eff ect was most pronounced in the five hundred year floodplain grouping. These results suggest that the one percent description may be more effective than the others in conveying flood re lated uncertainty to people with little direct or indirect expe rience and information. Table 6.21. Logistic Regression for Mo re than Once per Year: DonÂ’t Know Grouping Model Usefulness % Correct Category Model N R No Yes All Variables Beta Exp All (N=114) 15.228** .279 100 10.092.1FRI -.296 .744 Constant -4.456 p 0.05 ** p 0.01
181 Table 6.22. Logistic Regression for More than Once per Year: 1% Chance Flood Grouping Model Usefulness % Correct Category Model N R No Yes All Variables Beta Exp All (N=104) 17.361** .236 100 30.484.6FRI -.234 .791 Constant .919 U (N=55) 12.397** .303 97.638.583.6FRI -.277 .758 Constant 1.244 V (N=49) 5.274* .160 100 20.083.7FRI -.189 .828 Constant .529 100 (N=48) Only 7 Yes 500 (N=56) 20.527** .440 95.050.082.1FRI -.427 .653 Constant 2.094 p 0.05 ** p 0.01 Relative Concern The final outcome factor addressed in this section is relative concern. Participants were asked which of the three floods describe d on the cards concerned them the most and which concerned them the least. They were th en asked to rate their associated concern levels on a scale of one to seven. Approxima tely 42 percent said that the floods were equally concerning. Responses of Â“Equally C oncerningÂ” did not differentiate between those with high levels of concern and thos e who thought that none of floods were of much concern. However, 18 of the 48 individu als (38 percent) who said the floods were equally concerning rated their concern as seve n out of seven points; the median was five. Table 6.23 includes model information for undifferentiated responses of Â“Equally Concerning.Â”
182 Table 6.23. Logistic Regression for Equally Concerning Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=114) 14.015** .156 74.260.468.4 Understanding of Uncertainty 1.481 4.339 Constant -.947 U (N=60) 4.236* .092 72.254.265.0 Understanding of Uncertainty 1.123 3.073 Constant -.860 V (N=54) 28.396** .548 86.775.081.5 Understanding of Uncertainty 2.302 9.989 Individual Responsibility 2.186 8.900 Evacuated 3.235 25.396 NFIP Familiarity -.523 .593 Constant -3.429 100 (N=50) 6.585* .165 69.266.768.0 Understanding of Uncertainty 1.504 4.500 Constant -.811 500 (N=64) 16.447** .309 97.537.575.0 Understanding of Uncertainty 1.791 5.998 Individual Responsibility 1.861 6.431 Constant -2.176 p 0.05 ** p 0.01 Only one variable, understanding of uncerta inty, occurred in the model for every spatial grouping. This may indeed indicate an intellectual or affec tive understanding of the uncertainty of all flood risk estima tions and environmental conditions. Based on participant commentary, though, I believe that in this model the variable also represented a tendency to lump all flooding into one mental box and treat it as a condition that either affected the respondent or didnÂ’ t. Flooding itself, rather than the specific description of flood risk (since most thought they were different floods), became the important
183 reference. However, NakelkerkeÂ’s R was low and the Hosmer-Lemeshow test did not show good fit for the models in which understanding was the only variable. Additional patterns were identified when the group was di vided between those with relatively high levels and relatively low levels of concern. Individuals were assigned to the high c oncern group if they rated concern a six or a seven. The low concern group consisted of th ose whose concern level was a one, two or three on the seven point scale. Two separa te logistic regressions were run after determining significant correlations. Flooding as a top concern was not included as an independent variable because of conceptual overlap. Model inform ation for high and low concern is presented in Table 6.24. Perception of medium or high risk of future flooding was included as a strong predictor in both models. The odds responde nts rating specific con cern a six or seven increased by over 15 times if they believed them selves to be at medium or high risk of flooding in general. A person who thought th eir risk of future flooding was low was much more likely to be relatively uncon cerned about all three specific floods. This supports the argument made above that the choice of all descri ptions as equally concerning may, at least in part, be a result of participants treating flooding as a condition without gradation. In addition, the association of risk perception to concern levels here validates linking risk percepti on and flood related concern in the general model (Figure 6.1). General control was also predictive of flooding as a top relative concern.
184 Table 6.24. Logistic Regression for Equally Concerning: High and Low Concern Response Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp High Concern (N=48) 18.584** .482 68.082.675.0 Med/High Risk Perception 2.735 15.408 General Control -.668 .513 Constant .125 Low Concern (N=48) 27.145** .600 93.881.389.6 Med/High Risk Perception -2.265 .104 NFIP Familiarity -1.049 .350 Constant 2.766 p 0.05 ** p 0.01 In Vestal, NFIP familiarity had a negativ e relationship to a personÂ’s choosing all floods as equally concerning. NFIP familiar ity also appeared in the model for low concern levels within that group. It may be th at individuals more familiar with the NFIP are more concerned about the description they most associate with the area the NFIP regulates (hundred year fl oodplain). Among those that ra te all floods as equally concerning, lack of NFIP familiarity may represent a lack of information and/or experience. The 48 participants who responded that al l floods were equally concerning were not included in analyses of least concerni ng and most concerning floods. The four people who identified the hundred year flood and the one percent chance flood as identical were also removed. All voted this flood as the most concerning. All but six of the remaining 62 responses were split between the hundred ye ar flood and the 26 percent chance flood as the most concerning. Regression results are included in Tables 6.25 and 6.26.
185 Table 6.25. Logistic Regression for Most Concerning: 100 Year Flood Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp V (N=28) 4.972* .218 66.775.071.4 Flooding a Top Concern -1.792 .167 Constant 1.099 500 (N=37) 4.712* .160 58.870.064.9Age -.043 .958 Constant 2.530 All Other Groupings No Correlation p 0.05 ** p 0.01 Table 6.26. Logistic Regression for Most Concerning: 26% Chance Flood Grouping Model Usefulness % Correct Category Model N R No Yes All Variables Beta Exp V (N=28) 5.828* .258 83.350.071.4Age .071 1.074 Constant -4.794 500 (N=37) No Improvement All Other Groupings No Correlation p 0.05 ** p 0.01 No correlations were found for either choice in the total sample, Union or the SFHA. Only two variables were included in models for Vestal and the five hundred year flood plain, flooding as a top c oncern and age. Flooding as a top relative concern was also positively correlated to the 26 percent chance flood as most concerning in the five hundred year floodplain, but there was no model improvement. Though of limited scope, these results are interesting, especially give n the opposite associations of age to flooding as a top concern in Union and Vestal found in the exploration of the general model. Age
186 was negatively correlated in Union and positiv ely correlated to flood concern in Vestal. These results suggest that individuals, even those who are relatively concerned about flooding in general, might be less concerne d about the hundred year flood than floods described in probabilistic terms. This conclu sion is further supported by analyses of the least concerning flood, the results of which are found in Tables 6.27 and 6.28. Two individuals responded Â“DonÂ’t KnowÂ” when asked which flood was of least concern to them and were not included in analyses. Only eight judged the 26 percent chance to be least concerning, so regressi ons were run for the hundred year and one percent descriptions alone. Ta ble 6.27 shows that while approximate R was low, models were consistent. Age was the only listed vari able and increased the odds of choosing the hundred year flood as the least concerning. Th e effect was most pr onounced in the five hundred year grouping. Table 6.27. Logistic Regression for Least Concerning: 100 Year Flood Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=64) 4.647* .098 93.014.367.2Age .039 1.040 Constant -2.957 U (N=36) 4.301* .153 81.842.966.7Agea .046 1.048 Constant -3.033 V (N=28) Only 7 said Yes 100 (N=25) Only 8 said Yes 500 (N=39) 6.003* .198 84.623.164.1Age .056 1.057 Constant -3.980 p 0.05 ** p 0.01 aSE is .024, just over twice Beta
187 Table 6.28. Logistic Regression for Least Concerning: 1% Chance Flood Grouping Model Usefulness % Correct Category Model N R No Yes All Variables Beta Exp All (N=64) 4.579* .092 44.880.064.1BachelorÂ’s -1.179 .564 Constant .560 U (N=35) 4.824* .172 76.555.665.7 Info Satisfaction -.455 .634 Constant 2.117 V (N=28) 6.125* .264 66.775.071.4FRI -.225 .798 Constant 2.659 100 (N=25) 4.975* .242 72.771.472.0Female 1.897 6.667 Constant -.693 500 (N=39) 6.309* .200 66.771.469.2Age -.052 .949 Constant 3.107 p 0.05 ** p 0.01 The odds of a person responding that the one percent chance flood was the least concerning decreased with age in the five hundred year floo dplain. The remainder of the variables included in the models for the one percent chance flood as the least concerning appeared inconsistent and unrel ated. However, they were very similar to the variables used to model the perception of the one per cent chance flood as the smallest of the three described floods (see Table 6.14). Models of concern related to the hundred year flood may illustrate the role of perceived relative likelihood in determinations of concern, which could be more influential in older pe ople. The models for the one percent chance flood may indicate a link between perceive d size and relative c oncern. Relationships between likelihood, size, and con cern will be further examined in subsequent sections.
188 Summary and Conclusions There is no current literatu re regarding situational and cognitive factors and their relationships with the perception of specific fl ood risk messages or frames; this section of the analysis was by far the most exploratory. Models were generally weak, indicating that the conceptual framework was inadequate. It might be useful to further break down messages (and use more of them) in order to identify what aspect s of the flood risk descriptions (numbers, timeframe, concepts words) people are r eacting to and what characteristics are associated with specific reactions. This will require much more indepth qualitative analysis. Some general trends were identified, however. Perhaps not surprisingly, many of the variables that consistently appeared in the models represented cognitive or informational factors. Location and experien ce were not included in any of the models, but might underlie some of variables th at were included. Figure 6.3 is a broad representation of relationships supported by this analysis. It depicts variables that may generally contribute to the formation of differing perceptio ns (including DonÂ’t Know) of size, likelihood, uncertainty and concern a ssociated with specific messages.
189 Figure 6.3. Relationships of Situational and Cognitive Factors to Perceptions of Specific Descriptions* PERCEPTION OF UNCERTAINTY PERCEPTION OF SIZE PERCEPTION OF LIKELIHOOD CONCERN FRI Breadth and Depth of FRI SOCIO-ECONOMIC Education Age COGNITIVE FACTORS Self Assessed Knowledge Complexity of Understanding Understanding of Uncertainty General Control General Worry Risk Perception PERCEPTION OF UNCERTAINTY PERCEPTION OF SIZE PERCEPTION OF LIKELIHOOD CONCERN FRI Breadth and Depth of FRI SOCIO-ECONOMIC Education Age COGNITIVE FACTORS Self Assessed Knowledge Complexity of Understanding Understanding of Uncertainty General Control General Worry Risk Perception *Relationships are between variables and variation in nominal outcome choices. Relationships with more consistent evidence are represented with solid lines; dotted lines illustrate identified relationships with weaker evidence.
190 JUDGING RELATIVE EFFECTIVENESS Data and Methods This section addresses the third research question. Which of th e three descriptions of flood risk is comparatively most effectiv e? Effectiveness as understanding was judged through the understanding of flood related uncer tainty over time and space. Effectiveness as persuasion was evaluated based on the le vel of concern associated with each term relative to the others. No absolute measure of effectiveness was used in this analysis. As described above, participants were gi ven cards with the descriptions on them and were instructed that they could also an swer the questions with a combination, Â“AllÂ”, or Â“DonÂ’t KnowÂ”. Understanding of flood re lated uncertainty over time and space was assessed through two questions: 1. Which of these floods, if any, do you think could happen more than once in a year? 2. Do you think the size of any of the floods described on these cards could change over time? If yes Which? Two questions were also used to measure relative concern: 1. Which of the floods described on the cards concerns you most? 2. Which of the described floods concerns you the least? If more than one description was give n as an answer to a question, each affirmative response was coded as one. If a participant said he or she didnÂ’t know, all descriptions were coded as zero. Â“DonÂ’t K nowÂ” responses were also recorded as a separate variable. CochranÂ’s Q was used to detect differences betw een response rates for the three descriptions. CochranÂ’s test is a non-parametric repeated measures test of
191 variance for more than two dic hotomous variables. It is an expansion of McNemarÂ’s test, which was used for post hoc comparisons. A scale of relative concern was also constructed for each description. If a description was perceived as most concerning, it was assigne d a positive one; if perceived as the least concerning, it wa s assigned a negative one. Desc riptions elicit ing no response were treated as neutral and assigned a zero. Poin ts were then added together and adjusted to a range of zero to two. Frie dmanÂ’s test, a non parametric repeated measures test based on ranks, was used to distingui sh statistically significant differences between the three descriptions. Higher numbers signaled greater effectiveness. FriedmanÂ’s test was also used to look for variance in overall scales of effectiveness th at included both the understanding and concern fact ors. Wilcoxon tests were used for post hoc comparisons. All statistical tests were conducted at a 0.05 level of significance and the Holm method was used to adjust for multiple comparisons. The questions in this section of th e survey were closed. However, many respondents had quite a lot to say; their commentary helped clarify some of the comparative quantitative results as well as the regression models presented in the previous subchapter. Interviewer observati ons and participant comments, if provided, were recorded for each question. The comments were written up and themes within and across questions were identified. As they were throughout this project, analyses were conducted for the total sample and for four spatial subsets in order to illustrate pattern c onsistency and identify possible areas of further investigation. Qualitative themes were summarized by question or question set. Commentary was meant to assi st in the practical understanding of general
192 quantitative trends and was not separated into spatial groupings. Results for effectiveness as understanding are presented first, followe d by those for effectiveness conceptualized as persuasion (measured by concern). A comb ined measure of effectiveness is also addressed. Understanding of Flood Related Uncertainty When asked which of the described floods they thought could happen more than once per year, about nine percent of the whole sample said th ey didnÂ’t know and approximately 45 percent said all of them could happen more than once. These figures became neutral in analyses of variance. Affi rmative response percentages and the results of comparisons are included in Table 6.29. It sh ould be noted again that only two people said that all three descriptions referred to the same flood; six identified the one percent chance and the hundred year floods as equal. Description rankings were consistent acr oss spatial subsets. The one percent chance description always had the highest res ponse rate and the hundred year description the lowest. Post hoc testing showed the hundr ed year description performed significantly worse than both the one percent chance a nd 26 percent chance de scriptions in all categories. Differentiation between the two prob ability based descriptions was less clear, showing significance only within the total sample. The outcome appeared to be more heavily affected by group size than the othe r comparisons and shoul d be treated with more caution.
193 Table 6.29. Variation in Possibility of Occurring More Than Once per Year: CochranÂ’s Q Description All (N=114) Union (N=60) Vestal (N=54) 100 Year (N=50) 500 Year (N=64) % Yes % Yes % Yes % Yes % Yes 100 Year 46 40 52 54 39 1% Chance 78 78 78 84 73 26% Chance 67 67 67 76 59 Q Q Q Q Q Omnibus 43.14** 28.76** 14.80** 18.10** 25.31** Post Hoc Sig. a Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. 100 and 1% ** ** ** ** ** 100 and 26% ** ** ** ** 1% and 26% NS NS NS NS a McNemarÂ’s test used in post hoc comparisons of two responses; significance adjusted using Holm method. p 0.05 ** p 0.01 Unlike the ranking of descriptions, pe rcentages varied. The hundred year description was most sensitive to changes in spatial category, though tests showed no significant differences at = 0.05. The one percent was most consistent. Every description had a higher response percentage in the hundred year floodplain. This was not unexpected, given that 13 percent of those in the five hundred year floodplain answered the question with Â“DonÂ’t Know Â” versus only four percent of those in the SFHA. The result may be a function of relative experience. Experience was one of the themes that came up in participant comments on this question (see Table 6.30). Experience commen ts were associated with the one percent chance and hundred year descriptions as well as responses of Â“AllÂ” and made reference to concrete examples, general personal experi ence and general commun ity experience. If a specific description was chose n, the choice seemed to depend on what label a respondent had assigned to recent floods and if that labe l was consistent. This would suggest that the
194 naming of experience, in addition to experien ce itself, may be important to individualsÂ’ understanding and assessment of risk. Includ ing the naming of past experience might improve the models presented in the previous subchapter. In the third comment, the tag Â“anythi ngÂ’s possibleÂ” hints at the potential interaction of experience and general ou tlook. This phrase or something like it was usually accompanied by a response of Â“All.Â” While it, like references to climate change, came up more often when discussing change in size over time, the phrase was used in this context as well. The linking of climate ch ange only to the overtly probabilistic descriptions may demonstrate the strength of association between the hundred year description and ideas of a strict cycle. Table 6.30. Participant Comments on Fl ooding More than Once per Year Themes Representative Comments Our car almost got wiped out in June AND November. The 1% HAS happened more than once a year. Experience I say they could happen more th an once only because of personal experience Â– anythingÂ’s possible. Outlook Anything can happen. Likelihood Just had the 100 year, wonÂ’t happen anytime soon. Climate Change The 1% and 26% could happen more than once. With climate change, you never know. Difficulty Whether they happen more than one time depends on the size. ItÂ’s hard to tell which floods could happen more than once by the descriptions. In the course of the survey, two peopl e stated that the 26 percent chance flood occurred every seven years and a few said that the one percent chance flood happened every year. Many, many people made statements similar to the comment associated with likelihood in Table 6.30. These t ypes of statements were made when discussing the
195 possibility of multiple floods per year, relati ve concern, and when asked directly about likelihood of occurrence in the next year. About 72 percent of respondents chose the hundred year flood alone as the least likely to occur. Only 11 percent of participants chose the one percent chance flood, the descrip tion with the second hi ghest response rate. These statements again show the potential relationship between experience, the naming of experience, and perception, and give furthe r credence to critiques of the hundred year flood description. Another theme running through commentary on each of the questions was dissatisfaction with the descriptions. In some cases, it was a general sense of difficulty or confusion. Others pointed to a particular problem. The participant who said that, Â“Whether they happen more than one time depends on the size,Â” ultimately answered Â“DonÂ’t KnowÂ” and indicated that a piece of in formation vital to making a differentiation (the physical size of a flood) was missing. Wh ile some difficulty with the questions themselves might be expected, this man may have articulated specifically a frustration others expressed more generally. This co mment and the associated frustration, does support the linking of size and likelihood in Figure 6.3, however. A separate individual gave a somewhat ex asperated response de crying the lack of size markers in the descriptions when asked about change in size over time (see Table 6.31). A much higher proportion of participants (24 percent) answered Â“DonÂ’t KnowÂ” to this question and made comments about having trouble. This perhaps reflects the level of difficulty of the question as we ll as the perceived lack of pertinent information in the descriptions. It may also indicate that a num ber of people are unfamiliar with the kinds of conditions and data constraints that impact flooding and flood estimation.
196 However, those who believed at least one of the described floods could change often explained their reasoning, and most cited either climate change or human alteration of the environment as a factor in their res ponse. Some linked the two. Participants relied on their own interpreted experience as well as the experiences of their social contacts to estimate change, reflected in the inclusion of FRI in the above models. One person specifically mentioned Â“An Inconvenient Trut hÂ” (Al GoreÂ’s presentation on climate change) as an influence. Informal communica tion appeared to have made a difference in some peopleÂ’s decisions, as did a general exp ectation of change or lack of control. Control was also included in models of understanding of uncertainty and its component variables. Table 6.31. Participant Comments on Change in Size over Time Themes Representative Comments Outlook Time can change anything. Experience Floods are getting bigger and bigger. My neighbors say there are big time increases in flooding. Climate Change With global warming, all of them could change. I suspect they can change, espe cially due to human activity. Human Influence All of them could change when we Â’re fooling with Mother Nature. Difficulty But these descriptions donÂ’t GIVE the size! This is confusing. None of the comments listed in Table 6.31 contain references to the descriptions. None of the recorded comments did either. St atistical tests showed no difference between any of the terms in any spatia l category. As seen in Table 6.32, the largest difference was only four percentage points. In fact, 65 percen t of the total sample said that all could change over time, 24 percent answered Â“D onÂ’t KnowÂ”, and six percent believed none
197 could change. That left possible variation among only six people. Similar patterns were found in the subsets, though a ll six who chose only one or tw o descriptions lived in the five hundred year floodplain. Relative experien ce may be a factor, but thought processes associated with this question appeared unrelated to the terms presented. Table 6.32. Variation in Possibility of Change in Size over Time: CochranÂ’s Q Description All (N=114) Union (N=60) Vestal (N=54) 100 Year (N=50) 500 Year (N=64) % Yes % Yes % Yes % Yes % Yes 100 Year 68 70 65 72 64 1% Chance 67 70 63 72 63 26% Chance 67 72 61 72 63 Q Q Q Q Q Omnibus .333 .667 2.00 0.00 .333 p 0.05 ** p 0.01 Relative Concern Relative concern was used to approximat e persuasion. When asked about the most and least concerning of the floods described, about 43 percent said they were equally concerning. The comments in Ta ble 6.33 indicate that a resp onse of Â“AllÂ” could reflect either a general position that, as one partic ipant put it, Â“Any flooding equals concernÂ”, or an opposite contention that flooding isnÂ’t really a problem for them and they donÂ’t worry about it much. For these 43 percent, the desc riptions were somewhat irrelevant; flooding was flooding, and other situational and cognitive factors seemed to have more effect on response. Like general outlook (perhaps associated with experience or lo cation), actions to mitigate physical or financial impact influe nced some participantsÂ’ responses. Though
198 individuals who made action related comments chose Â“All,Â” the actors involved and levels of concern differed. Comm ents involving Â“theyÂ” were a ttached to both relatively high and relatively low levels of concern; va riation might be connect ed to the perceived likelihood of governments taking Â“properÂ” actio n. Trust and/or credibility may play a part. Neither mitigative actions nor credibility appeared in models of concern, however. Table 6.33. Participant Comments on Concern Themes Representative Comments Outlook Any flood is bad. IÂ’m not personally in danger. ItÂ’s pretty safe if they do something. Mitigation IÂ’ve insured the hell out of the house. Likelihood This (100 year) already happened. It (100 year) makes you leery, but I wonÂ’t see it again in my lifetime. If you have a disaster (100 year), youÂ’re probably okay. 30 years is a short time. Wow, that concerns me. Size IÂ’m least concerned about the 26% because itÂ’s the smallest. Difficulty IÂ’m most concerned that one will exceed the last. These arenÂ’t very good indicators. While responses of Â“AllÂ” may have ha d little to do with the descriptions, statistical results included in Tables 6.34 and 6.35 show significant variation in the remainder of the responses acros s all spatial categories. If a participant said that all descriptions were equally concerning, all of them were assigned a one in both questions, thus neutralizing the effect in analyses of variance while retaining information for other assessments. The rankings of response rates we re internally consistent for both most concerning and least concerning. In all subsets, more people chose the hundred year flood than either of the other two descripti ons as the most concerning. The 26 percent description ranked second, but post hoc test s showed no significant difference between it
199 and either the hundred year or one percent chance descript ions. The percentage range between rankings was most variab le in the floodplain groupings. Table 6.34. Variation in Most Concerning Flood: CochranÂ’s Q Description All (N=114) Union (N=60) Vestal (N=54) 100 Year (N=50) 500 Year (N=64) % Yes % Yes % Yes % Yes % Yes 100 Year 75 72 78 76 73 1% Chance 52 52 52 54 50 26% Chance 62 62 63 70 56 Q Q Q Q Q Omnibus 15.39** 6.00* 9.87** 7.46* 9.05* Post Hoc Sig. a Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. 100 and 1% ** ** * 100 and 26% NS NS NS NS NS 1% and 26% NS NS NS NS NS a McNemarÂ’s test used in post hoc comparisons of two responses; significance adjusted using Holm method. p 0.05 ** p 0.01 Table 6.35. Variation in Least Concerning Flood: CochranÂ’s Q Description All (N=114) Union (N=60) Vestal (N=54) 100 Year (N=50) 500 Year (N=64) % Yes % Yes % Yes % Yes % Yes 100 Year 61 63 57 64 58 1% Chance 73 72 74 76 70 26% Chance 50 47 54 54 47 Q Q Q Q Q Omnibus 15.86** 9.72** 7.36* 7.28* 8.67* Post Hoc Sig. a Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. 100 and 1% NS NS NS NS NS 100 and 26% NS NS NS NS NS 1% and 26% ** ** NS * a McNemarÂ’s test used in post hoc comparisons of two responses; significance adjusted using Holm method. p 0.05 ** p 0.01
200 The one percent chance flood description had the lowest response rate for the most concerning flood; it ranked highest in all subsets for the least concerning. One might expect to see a similar inversion with the hundred year description, but instead, the 26 percent chance description had the lowest response rate. In post hoc comparisons, however, only the differences between the 26 percent and the one pe rcent descriptions were significant. The comments in Table 6.33 hint at an e xplanation for both the lack of logical consistency in the ranking of most concerni ng and least concerning descriptions and the weakness of the specific concern models. Par ticipants did not rela te the hundred year flood to concern in a consistent manner. Al most identical statements were accompanied by very different responses and ratings on the seven point concern scal e. Several justified their answers with variations on Â“It alr eady happened.Â” One might rank it as least concerning; another might ha ve said most concerning, but rated it a one. Similar discrepancies emerged with more cautious individuals. The person who wouldnÂ’t Â“see it again in my lifetimeÂ” was still Â“leeryÂ” and ranked the hundred year flood as most concerning, but rated it a f our. The individual who equate d the hundred year flood with disaster, but thought that Â“you Â’re probably okayÂ” for the future if you experienced it, ranked the description as leas t concerning and also rated it a four. Elsewhere in the interview, she said, Â“Every time it rains, I worry. I think about it constantly.Â” This combination illustrates a sort of wishful thin king underlain by apprehension that was not uncommon. The above comments, along with the sheer proportion of statements dealing with the hundred year flood, indicate that this phr ase elicits strong res ponses and has power.
201 While a repeated measures ANOVA identified no significant differenc es in mean concern ratings, words mattered. Unfortunately, they mattered in different ways to different people. The hundred year description, perhaps because it is more frequently used and seemingly straightforward, appeared more prone to inconsistent interpretation. The irregularity was reflected in the resu lts of the combined concern analyses presented in Table 6.36. The one percent chan ce description had the lowest mean ranking in all spatial categories. Contra ry to previous analyses, howev er, the relative rankings of the other two descriptions did not hold acros s subsets. Post hoc comparisons showed a pattern consistent enough to conclude that the one percent descrip tion was significantly less effective in inducing con cern than the other two descri ptions. There was not enough evidence in any grouping to suggest that the hundred year description was more persuasive than the 26 percent chance descri ption, or vice versa. As they did for the hundred year flood, comments in Table 6.33 sh ow contradictory conclusions regarding the 26 percent chance description. These we re limited, though. Unlike the hundred year description, the 26 percent chance flood ranked highly because participants generally did not have strong reactions to it vis--vis concern.
202 Table 6.36. Variation in Relative Concern: Friedman Test Description All (N=114) Union (N=60) Vestal (N=54) 100 Year (N=50) 500 Year (N=64) Mean Rank Mean Rank Mean Rank Mean Rank Mean Rank 100 Year 2.12 2.07 2.18 2.11 2.13 1% Chance 1.77 1.78 1.75 1.75 1.78 26% Chance 2.11 2.15 2.07 2.14 2.09 Omnibus 17.27** 7.82* 10.69** 9.71** 8.05* Post Hoc Sig. a Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. 100 and 1% ** NS ** * 100 and 26% NS NS NS NS NS 1% and 26% ** * * a Wilcoxon test used in post hoc comparisons of two distributions; significance adjusted using Holm method. p 0.05 ** p 0.01 Combined Measure of Effectiveness An ideal flood risk message would contribute to both understanding and persuasion. A combined scale was created fo r each of the descriptions and then compared; results are presented here. Because answers to the question about change in size over time appeared to have little to do w ith the descriptions themselves, the question was not included in either a scale of overall understandi ng or a scale of combined effectiveness. The concern scale ranged from zero to two. In order to equally weight understanding and persuasion components in a scal e of overall effectiveness, responses to the question regarding multiple floods per year were multiplied by two and added to the concern score. The resulting scale ranged fr om zero to four and variation was assessed using the Friedman test. Results are included in Table 6.37.
203 Table 6.37. Variation in Combined Effectiveness: Friedman Test Description All (N=114) Union (N=60) Vestal (N=54) 100 Year (N=50) 500 Year (N=64) Mean Rank Mean Rank Mean Rank Mean Rank Mean Rank 100 Year 1.83 1.75 1.92 1.79 1.86 1% Chance 2.04 2.08 2.01 2.00 2.08 26% Chance 2.13 2.18 2.07 2.21 2.06 Omnibus 9.12** 9.48** 1.217 8.17* 3.01 Post Hoc Sig. a Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. Post Hoc Sig. 100 and 1% NS NS NS NS NS 100 and 26% ** ** NS NS 1% and 26% NS NS NS NS NS a Wilcoxon test used in post hoc comparisons of two ditributions; significance adjusted using Holm method. p 0.05 ** p 0.01 Relative rankings were consistent in f our of the five groupings, with the 26 percent chance description at the top and the hundred year description at the bottom. The scaleÂ’s calculation gives a slight advantage to descriptions that performed well in the understanding component; descriptions received either a zero or a two, whereas in the concern portion, a score of zero, one, or two wa s possible. This may explain, in part, the one percent chance descriptionÂ’s relatively high ranking, given its poor performance in the concern section. However, percentage ranges between the hundred year and one percent descriptions were closer in the concer n results than they were in the results for understanding. The hundred year description wa s also hurt by the split response with regards to concern. Variation between the descriptions was si gnificant in only thr ee of the groupings. In each case, the result was driven by the disparity between the 26 percent and hundred year descriptions. Post hoc comparisons show ed no significant differences between either of these and the one percent de scription. While the rank order of descriptions was fairly
204 consistent, the somewhat ambiguous results ma y point to a problem in combining the two conceptions of effectiveness. Summary and Conclusions Which description is best? The answer might depend on an addendum to the question. Best for what? Is public understand ing or persuasion th e goal? Research has shown that these outcomes may not be n ecessarily linked (Sjoberg, 2000; Beehler et al ., 2001; Bell and Tobin, 2007). The results of this analysis support that conclusion, as do the regression results. Response rates for both the hundred year and one percent chance descriptions were very sensitive to whethe r the question asked d ealt with understanding or concern. Risk managers and communicators may have to decide which is ethically and/or practically more important in the current system of flood loss mitigation and distribution. The 26 percent chance description appeared to be the most effective overall, given the shortcomings of the other two. It also received the highest response rate when participants were asked which of the floods de scribed was most likely to occur within the next year. There are reasons to be cautious however. The 26 percent description scored well, in part, because it did not stand out. Th is may be due to its unfamiliarity to most respondents. Familiarity contributes to the pr oblems associated with the hundred year flood description, and lack of it could be seen as a benefit, but si milar research in a different location showed strong negative reac tions to the 26 percen t description (Bell and Tobin, 2007).
205 Comments in Table 6.33 also touched on a potential difficulty that is more obvious with other descripti ons. In a setting that enc ourages heuristic processing, individuals react to different aspects of th e descriptions and come to contradictory conclusions. Resonance might be related to a number, or a quickl y interpreted likelihood or size. Future research to clarify relationships between perceived size, likelihood, and concern, and specific descriptions would be useful. Many individuals admitted trouble attaching a physical size to the descriptions, however, and voc alized a general frustration with all of the terms. Possible improvement s to flood risk communi cation are the subject of the next section. IMPROVING FLOOD RISK COMMUNICATION This subchapter addresses the final set of research questions. First, how do people describe floods? What worries them about flooding? The answers to these questions provided a starting point for the next: how might flood risk communication be improved? This research has shown that one message may not fit all conceptualizations of effectiveness. One goal of this analysis was to identify trends in the data that relate to suggestions made in the literature for both improved persuasion and/ or understanding of uncertainty. A second goal was to identify what was important to the people of Union and Vestal with regards to flood risk co mmunication and come up with specific improvements based on their own experiences. It was not assumed that these suggestions would fit with the conceptualizations of effectiveness presented thus far.
206 Participant Descriptions of Flooding Official flood risk information is gene rally communicated using return periods, probabilities, or cumulative probabilities (e.g. hundred year flood, flood with one percent chance in any year, flood with a 26 percent chance in 30 years). These risk frames were a focal point of this research However, thes e methods of communicating risk may not be effective if message recipients talk about floods and flood ri sk using different words or different frames. In order to explore how lay people talk about flooding in this area, survey participants were asked to describe the size, in their own words, of the largest flood they had experienced. Experience need not have been direct. Approximately 15 percent of respondents reported never havi ng been impacted by flooding and were not asked this question. The given descriptions fell into five gene ral categories: a generic description like Â“HugeÂ”; the reference floodÂ’s re lationship to other floods; th e reference floodÂ’s relation to some specific landmark (restaurant, home, etc. ); return period; and stage. The accuracy of participant descriptions was not evaluated. Four peopl e indicated they didnÂ’t know how big the flood was; all lived in VestalÂ’s five hundred year floodplain. The results, broken into spatial groupings are included in Table 6.38. Generic descriptions were the most frequent ly used overall, but in forty percent of the cases, they were accompanied by an addi tional description. In the Union and five hundred year floodplain groupings, however, a greater percentage of people used a relationship to past floods to describe the size of the re ference flood. A much larger proportion of SFHA residents used a generic term than did those living in the five hundred year floodplain; the percentage using onl y a generic term was higher as well. It
207 may be that SFHA residents have more direct experience, are more invested in flooding, and less likely to not describe their experience at a ll, even if they donÂ’t know the size. Generic terms were also usually animated ex clamations, a response perhaps less likely in those less severely impacted. Generic descriptions are not particularly useful because there is no common reference, but they ma y reflect lack of knowledge or of being overwhelmed. Table 6.38. Descriptions of the Largest Experienced Flood All (N=96) Union (N=48) Vestal (N=48) In 100 (N=49) In 500 (N=47) # % # % # % # % # % Related to Other Floods 42 43.8 24 50.0 18 37.5 21 42.9 21 44.7 Landmark 14 14.6 9 18.8 5 10.4 5 10.2 9 19.1 Return Period 12 12.5 5 10.4 7 14.6 10 20.4 2 4.3 Stage 4 4.2 3 6.3 1 2.1 3 6.1 1 2.1 Generic Term 45 46.9 22 45.8 23 47.9 29 59.2 16 34.0 Only Generic 27 28.1 13 27.1 14 29.2 16 32.7 11 23.4 DonÂ’t Know 4 4.2 0 0 4 8.3 0 0 4 8.5 The most common non-generic description referred to past floods. For most, the June, 2006 flood was the largest they had e xperienced; comparison floods included the 1936 flood, Agnes, and the 2004 and 2005 floods. Comparison words included bigger and smaller, worse, and higher. The evalua tion of a flood as Â“worseÂ” or bigger did not necessarily reflect an estimation of stage, but in some cases indicat ed perceived severity of total impact, personal im pact, or total area impacted. With more precise language, comparisons to past floods could make flood risk communication consistently relevant in communities with fairly stab le populations and long residence times, like Union and
208 Vestal. Fewer people gave this description in Vestal, which appeared to have a higher proportion of people who retired to the town. A larger percentage of people in Uni on and the five hundred year floodplain used a local landmark to describe flooding than either return period or stage. In Vestal and the SFHA, however, a greater percentage us ed the return period. The difference was particularly pronounced between the hundred ye ar and five hundred year floodplains; in the five hundred year floodplain, only two people used the return period and only one used stage. Four people said that they purposely bought a home outside the SFHA, but most residents of the five hundred year fl oodplain probably have little reason to know their specific elevation relative to the river. They may also have less exposure to the more formal and abstract descriptions of floodi ng than people who have had to deal with regulation and regulatory boundaries. Those w ho live inside the hundred year floodplain may be better able to link the return period to area, and indirectly, size. It is important to note that no one, whether in or out of th e SFHA, used probability or cumulative probability to frame flood size. These results reflect the frustrations regarding size discussed in the prev ious subchapter. Individuals who used landmarks to descri be flood size usually mentioned relative flood level, but not an absolute stage. Inst ead of Â“The flood was 32 feet,Â” participants said Â“The flood went up to the roof of the Drive InnÂ” or Â“It was four feet up JanetÂ’s house.Â” Using estimated flood stage, it would not be difficult to calculate a range of levels relative to well known community la ndmarks for use in bot h general and event specific communication. A larger percentage of participants described flood size in terms of past floods than described them relative to a landmark. Both are more concrete, but
209 reference to past floods depends to a gr eater extent on collective memory; landmark references might be more widely useful. Participant Concerns about Flooding Exploring Concern through Direct Questioning Participant concerns were evaluated in two ways. First, survey respondents were asked directly what concerned them most about flooding. Options given included: The level of possible flooding; The frequency of flooding of any level; A combination of flood frequency and flood level; and Other. Table 6.39 lists re sponses mentioned five or more times. Though participants were asked what concerned them most, several gave more than one answer. All answers were r ecorded, so percentages do not add to 100. Responses in the Â“OtherÂ” category were vari ed, but included, for example, loss of life, speed of onset, loss of community, and duration. The largest percentage of participants in all spatial groupings cited a combination of frequency and level as the most con cerning aspect of flooding. Similar proportions answered that level alone concerned them mo st. There was little pattern variation across the spatial sets. There was a bigger gap be tween communities and floodplain designations in the proportion that res ponded frequency was most concerning. The higher affirmative response rate from those in the SFHA makes sense, but results might have changed had the 500 year floodplain residents impacted in November been included. The more obvious pattern is the magnit ude of difference between th e proportions citing frequency versus the other choices. A similar question in another study garnered zero responses for frequency (Bell and Tobin, 2007).
210 Table 6.39. Most Concerning Thing about Flooding All (N=114) Union (N=60) Vestal (N=54) In 100 (N=50) In 500 (N=64) # % # % # % # % # % Level 41 36.023 38.3 18 33.3 18 36.0 23 35.9 Frequency 10 8.8 4 6.7 6 11.1 6 12.0 4 6.3 Combination Level/Freq. 44 38.625 41.7 19 35.2 20 40.0 24 37.5 Damage 23 20.211 18.3 12 22.2 3 6.0 20 31.3 Evacuation Issues 6 5.3 1 1.7 5 9.3 0 0 6 9.4 Health Issues 5 4.4 4 6.7 1 1.9 2 4.0 3 4.7 Other 25 21.910 16.7 15 27.8 9 18.0 16 25.0 While calculations of return period, probability and cumulative probability include size, participant comment s indicated that the relation to size is not obvious in the finished product. These methods of frami ng flood communication appear to emphasize timing, probability and frequency, and may not ge t at what people are really interested in. More survey respondents chose damage than chose frequency, though damage was not overtly listed. It is likely th at an even greater percentage would have cited damage as most concerning had it been clearly given as a choice. Future surveys should include it. Damage may be a touchstone for many pe ople, but an interes ting response pattern can be seen in Table 6.39. While approximate ly twenty percent of the total sample mentioned damage, only six percent of thos e living in the SFHA did the same. Since residents of the hundred year floodplain gene rally suffered more damage, this doesnÂ’t seem to make sense. However, it may be th at SFHA residents more easily link potential damage to flood level, being more familiar with stages or return periods and having perhaps had more frequent experience. Further re search is needed to explore this result.
211 Results for evacuation and health issues were also spatially skewed, though the number that mentioned them was low. Responses may have reflected personal experience. Four out of the five people c oncerned about health lived in Union, where multiple people mentioned receiving letters about or witnessing soils being sampled for contaminants. Concerns about the physical and mental impacts on the elderly were brought up as well. All who cited evacuation as a concern lived in the five hundred year floodplain; five out of six lived in Vestal. Almost al l the five hundred year floodplain residents in Vestal were asked to evacuate, and, for many, th is represented the full severity of impact. Evacuees in Union often had serious damage they were still dealing with; evacuation issues were forgotten, or of lesser concern. Se veral Union participants living in Fairmont Park were upset, though, that the fire road had not been maintained, remained closed, and could not be used for vehicle evacuation. Castle Gardens had experienced a similar problem in 2005, but respondents who brought it up indicated that th e Vestal government had been very responsive and that the problem was fixed prior to the June floods. Their action likely accounted for some of the disparit ies in credibility discussed in Chapter 5. Exploring Concern through Perceptions of Specific Flood Risk Descriptions Data gathered through questions looking at perception of specific descriptions were also used to explore participant concer ns. Correlations and regressions were run for the two descriptions that rece ived more than ten votes as most concerning and the two descriptions with more than ten votes as least concerning (see Tables 5.44 and 5.45). The analyses were performed using description sp ecific variables (plus DonÂ’t Know) for the
212 biggest, smallest, most likely and leas t likely floods. Because size and likelihood estimates were only moderately correlate d, results would indicate whether size or likelihood (or neither) was more strongly predictive of concern. Participants that said the described floods were equally concerni ng were not included in the analysis. Perceived relative size was predictive of the one percent chance flood being chosen as the least concerning in all spa tial groupings. If the one percent flood was described as the smallest, the odds that it was chosen least concer ning increased between 4.5 and nine times (see Table 6.40). The associ ation was strongest in the SFHA, a result which reflects comparative concern patterns di scussed in the previous subchapter. Only one other correlation was consis tently significant. In the to tal sample, in Union, and in the five hundred year floodplain, a response of DonÂ’t Know was positively correlated to a person choosing the hundred year flood as least concerning. In the reduced sample, though, only four people said they didnÂ’t know which was biggest, and no regressions were run. However, in the total unabridged sample, every single person who didnÂ’t know which of the described floods was largest chose the hundred year flood as the least concerning. These combined results could in dicate that perceived size is the more important determinant of concern; when size in formation is not readily discernible, other factors, including perceived likelihood, may become more important. However, many who cited the hundred year fl ood as least concerning did so specifically because they thought it was unlikely to occur. No correlations were found in analyses of the most concerning floods (hundred year flood, 26 percent chance flood). Additionally some participants stated outright that aspects of timing were what concerned them The lack of clear association between
213 concern and likelihood or size may reflect the problems discussed in the previous subchapter regarding quantitative analysis of divergent interp retations and ratings. Perhaps a better representation and evaluation of the combined perception of flood level and flood frequency or likeli hood is needed. Additionally, concern triggers may simply be too individualized for this framework. Rega rdless of the reason, re sults of this portion of the analysis were inconclusive. Table 6.40. Logistic Regression for 1% Chance Flood as Least Concerning: Size or Likelihood? Grouping Model Usefulness % Correct Categorization Model N R No Yes All Variables Beta Exp All (N=66) 11.226** .209 77.462.969.71% Smallest 1.758 5.802 Constant -.613 U (N=36) 4.961* .172 82.452.666.71% Smallest 1.646 5.185 Constant -.442 V (N=30) 6.709** .268 71.475.073.31% Smallest 2.015 7.500 Constant -.916 100 (N=26) 6.363* .290 83.364.373.11% Smallest 2.197 9.000 Constant -.693 500 (N=39) 5.240* .164 73.761.967.51% Smallest 1.515 4.550 Constant -.560 p 0.05 ** p 0.01 Improving Flood Ri sk Communication Survey respondents were not asked to provide suggestions for improving flood risk communication. Focus group participants, however, were specifically requested to talk about it. After a discussion of the thr ee descriptions, members of the groups were
214 asked if they would use any of the descriptions if they we re trying to convince a friend that a flood was a real threat to him or her. They were then asked what else they would say. Other methods of conveying flood risk we re brought up as we talked. Some key comments or exchanges related to each of the questions are presented below. Would you use the descriptions? No one in the focus groups suggested using the hundred year flood to convey threat, though one person mentioned it as mo st concerning. Individua ls in the second two groups answered this question rather quick ly. In the middle group, one person suggested that the 26 percent chance should be used to convey threat; th e other two concurred. Earlier in the discussion, the 26 percent ch ance description had been cited as most concerning and Â“the most likelyitÂ’s more r ealistic.Â” The lone participant of the third group answered the question by saying that Â“t he bottom two (1 percent and 26 percent) are a little wordyyou need something in be tween these, a major flood.Â” He viewed the hundred year flood as purely cyclical and se parated timing from size: Â“The hundred or the one percent concerns me the most because itÂ’s implied that theyÂ’re more severe, though they may not be.Â” The 26 percent wasnÂ’t a concern because it was the smallest. These conversations further emphasize the inco nsistent relationships between perceptions of size, likelihood and relative concern. In the first focus group, discussion on the topic was longer and more involved. Though a description was suggested, the group as a whole was less convinced, as the exchange below illustrates. In answering a brief questionnaire prior to the start of the session, one member of this group had define d the hundred year flood as having a one in
215 a hundred chance of occurring in any year. When the descriptions were presented to the group, he informed the other part icipants that all three referred to the same flood. This information framed the discussion somewhat di fferently than the ot hers. Additionally, the 26 percent chance description was sort of dismissed, since Â“it doe snÂ’t really register with you that it could happen every year.Â” All three groups focused on a different aspect of the description and came to different conclusions. What Else to Say? After the exchange above, in which part icipants struggled to find a concrete threat, conversation shifted to address what might be added. A key statement emphasized two different conceptualizations of effectiveness as understanding. A: TheyÂ’re all horribleÂ… C: IÂ’d probably use the middle oneÂ… M: The middle one? B: But that isnÂ’t very threatening, one percent. C: No, but I would Â… B: ThatÂ’s more threatening than a hundred year Â… C: YeahÂ… B: I donÂ’t know, I mean, I thin k they all are PRETTY vague. M: Okay. C: But, I mean, the middle one tells exactly what it is, and a 500 year is, yÂ’know, even less. C: Well, I guess it depends on what youÂ’re tr ying to do. I mean, if you tell em that the flood is a 1% chance every year, where you live, thatÂ’s explaining when a flood could occur. Now, if you want to exp lain what a flood do es to you, thatÂ’s a separate subject, really. And, you ha ve to explain that it can be very devastating, you could lose everything.
216 Explaining when a flood can happen and what a flood can do were perceived as two distinct communication goals which might re quire substantially di fferent approaches. Both were framed in terms of explaining ra ther than persuading, though. It is the second type of understanding that those pushing to include damage estimates in flood risk communication seek to address. However, two ou t of the three participants were skeptical that communication of either kind would actually make a difference in attitude, behavior or decision making. Situational and cognitive f actors were seen as more influential. However, all three groups emphasized this second type of understanding. The second group focused primarily on communi cating potential financial losses and generally assumed a narrow model. The othe r two groups had a somewhat more broad conception of what information enhanced an understanding of wh at a flood could do and in what forms it should be exchanged. C: Well if youÂ’re talking to somebody that Â’s a thousand feet a bove flood level, itÂ’s not gonna do any good to talk to them period. Uh, I think we know where some of the danger spots are, I mean, all th e way from Conklin, up along the Chenango, and along the Susquehanna. If you talk to those people, theyÂ’re gonna understand already. If they woulda sa id our zone was a hundred year flood zone and we needed flood insuran ce, being that the property had never been flooded, I probably woulda bought it an yway. Because weÂ’d looked for 6 months and this was the first piece of pr operty that met all of our parameters. YÂ’know? B: I donÂ’t know if thereÂ’s any thing that you could say. I mean, well, at least personally, I tend to be an optimisti c person, and I wanna be optimistic. It makes for a better life. And so, I donÂ’t th ink you could, other than to say, well, every five years, this property has flo oded. Well, no thanks, I donÂ’t want it.
217 The man in the third session believed th at Â“getting somebody to worry about or plan for a flood or anything, you need to give an example of what HAS happened.Â” Few of the suggestions for providing these ex amples had to do with narrow official communication. Like the one above, they inst ead suggested ways in which to bring personal experience to the public and vice ve rsa. The suggestions were essentially attempts to build a collective memory. Th ey included the book described above, other publications or coverage of lo cal flood history, family histor ies, integrating local events into the school curriculum, and putting peopl e into events through modeling or museum exhibits. Several of these ideas were mentioned by th e participant of the third session. With some reservations, he also suggested conveying this type of information with the express purpose of instilling fear, moving from a frame of understanding to a frame of persuasion. Eventually, he said, knowledge would replace fear. Understanding became the ultimate goal, achieved by first in stilling attitude and behavior. In Castle Gardens, Vestal is trying to create a collectiv e knowledge base, though it is not clear whether the goal is understanding or pe rsuasion. I noticed signs demarcating hundred year flood levels on my second trip, in November of 2006. They were discussed in the first focus group. A: I think what might be good, too, is, um the counties might put out, um, books that people could either see, or sell, or in the library, um, pertaining to different areas that has flooded. This is what the devastation of a flood could be, what itÂ’s caused, and what itÂ’s done to the pe ople, and it will affect your life, mentally, physically, and most surely financially, cause youÂ’re never gonna get back what youÂ’ve had and paid for things and whateverÂ…Maybe make people more aware and BELIEVE.
218 Underlying this exchange was someth ing brought up in every session: an individualÂ’s right to easily available risk information a nd disclosure, even if somebody loses a sale. The second group saw this as pa rticularly important for new homebuyers and their discussion focused on regulatory pract ices. Suggestions included requiring an appraisal regardless of the property or lend ing status, where you get information not only on official floodplains, but al so how many floods the propert y has experienced and how many homes in the area were impacted. They th ought the early and enforced provision of this information would allow someone to ma ke an educated purchase and to make adjustments before an event if they wanted to. Summary and Conclusions When describing a floodÂ’s size, the greates t number of participants related it to a past event. In the total sample, the next mo st frequently used me thod was relative stage; respondents described the flood level by referr ing to a personal or public landmark. While those in the SFHA used return period c onsistently, none of the survey participants used probability or cumulative probability. Bo th past levels and re lative stages could be B: In our neighborhood, since the flood, th ey came and put on many poles, throughout the whole little area, these red signs designating a hundred year flood level. Now, they were never there before, I donÂ’t know if theyÂ’re keeping them there forever, but I think if I were looking fo r a house and I drove into an area and saw those, I would think twiceÂ…So, I donÂ’t know, but maybe those are good, cause thatÂ’s, thatÂ’s pretty visible, everydayÂ… B: The thing is, people donÂ’t necessarily w ant this information out. If youÂ’re gonna sell your house, you donÂ’t want Â‘em to know itÂ’s a dangerous areaÂ… A: I think people have a ri ght to know, becauseÂ… C: The y do reall y
219 employed to make flood risk more concrete and more relevant, though relative stage may be more widely useful. Using these types of descriptions might relieve some of the confusion surrounding size brought up in the previous section. Participants were asked directly what concerned them most about flooding; level and a combination of level and frequency were cited most frequently. Though the descriptions most commonly used to describe flood risk in official communication focus on issues of timing, less than ten percent of the total group chose frequency alone as the most concerning aspect of flooding. Pattern s were not as clear, however, when description specific responses we re used to evaluate sources of concern. Size appeared to be emphasized, but commentary across multiple questions showed frequency was also important to some people, and no relationshi ps were found between the most concerning floods and size or likelihood. The same issues of variability in in terpretation, reference points and threat perception were reflected in the focus groups. Suggestions to improve communication of flood risk centered on regulatory practices and the creat ion of collective memory through books, visual markers and other methods of sharing experience. All groups ag reed that people needed to understand what could happen as well as when it could happen. The responses of both the survey participants and the focus group members em phasized the concrete rather than the abstract and, when taken as a whole, indicate that current methods of description are not broadly effective, regardless of the criteria used for evaluation.
220 CHAPTER 7: SUMMARY AND CONCLUSIONS This project sought to answer f our sets of research questions: 1. Which situational and cognitive factors are most highly related to varying perceptions of flood pr ocesses and uncertainty when relationships between the factor s are controlled? To a general perception of flood threat? To mitigative behavior? How are these outcomes related to each other? 2. When relationships between them ar e controlled, whic h situational and cognitive factors are most highly related to varyi ng perceptions of size, likelihood, uncertainty, and concern asso ciated with specific flood risk messages? Messages addressed in this project include the hundred year flood, a flood with a one percent chan ce of occurring in any year, and a flood with a 26 percent chance of occurring in 30 years. 3. Which of these flood risk messages are comparatively most effective with regards to understa nding and/or persuasion? 4. How do people describe floods and what worries them about flooding? How might flood risk communication be improved? The final chapter provides a summary of resu lts associated with each question set and links the results to past research. A set of general conclusions and suggestions for future research are also presented.
221 SUMMARY OF RESULTS Exploring Figure 2.1: The General Model of Perceptual and Behavioral Influences The conceptual framework for this research was outlined in Chapter 2. The summary of general results for this set of questions will be organized by the five situational and cognitive f actors illustrated in Figure 2.1. These include location, experience, socio-economic factors, cognitive factors, and flood risk infrastructure. Specific variables are listed in Table 6.1. Rela tionships presented below are not bivariate correlations, but predictive associations that stood out when controlling for the other variables. At least in this data set, these re lationships were the most important, not simply extant. Location Distance was not correlated to any of th e outcome variables, likely due to the presence of flood control struct ures. However, this result cont radicts those of Greene et al (1981) and Montz (1982). Floodpl ain status was associated with both general mitigative activities and event specific behaviors. SF HA residence was strongly associated with insurance purchase and decrea sed the odds of a person not considering flooding at all. The relationship of floodplain status to even t specific behavior was not consistent in direction and depended on othe r variables like community residence, pre-evacuation information and correct estimation of SFHA stat us. While it was not a predictor of threat perception, the relationship of SFHA status to response contrasts the work of Palm and Hodgson (1992) and Grasmuck and Scholz ( 2005). Community re sidence was linked solely to evacuation, a result of the two townsÂ’ differing appro aches to evacuation orders.
222 Experience Severity of experience was the most important factor in the revised model illustrated in Figures 6.1 and 6.2. Increased se verity was clearly linked to appropriate risk perception and higher relative concer n as well as each of the general and event specific behaviors tested. Though the indicators were weak, impact severity is likely generally predictive of unders tanding of flood related uncerta inty as well. These results support the linkage of outcome experience to perception and behavior (Bardura, 1994), to more accurate risk perception (Burton and Kates, 1964), and to increased likelihood of mitigative behavior (Kunreuther, 1978; Burby, 1988; Mileti and Darlington, 1997). Frequency of experience appeared to be le ss important, though it did have a negative relationship to lack of consideration or mitigation. Socio-Economic Factors Three socio-economic variables were predictive of risk perception, property modification, and/or evacuation. Higher inco me was linked to physic ally altering a home or property with the goal of reducing loss potential. However, higher income was not associated with increased rates of insuran ce purchase in any of the groupings; regulatory boundaries appeared much more important. Gender, as it has been in other studies (Cutter et al, 1992; Gustafson, 1998), was predictive of higher percepti on of risk. Age was consistently and negatively associated with both risk per ception and evacuation.
223 Cognitive Factors Grasmuck and Scholz (2005) have associated higher ambient worry with increased threat perception of specific hazards; the same relationship was found in this study area. A more internal locus of c ontrol decreased the likelihood of a person understanding flood related uncertainty as it was measured here. Ge neral control was not linked to any other perceptual or behavioral variables in th is analysis. Increases in the level of self reported knowledge were pr edictive of property protection, but also potentially of higher relative concern, a result that contradicts rese arch by Loges (1994). Analysis also weakly supported drawing a link between higher relative concern and increased seeking. Johnson (2005) found an association between mitigation and seeking behavior, but not between seeki ng and higher risk perception. Flood Risk Infrastructure Depth and breadth of flood risk infras tructure was the only clear predictor of understanding of uncertainty. Flood risk infrastr ucture was also linked to increased odds of property modification. A dditionally, specific types of information (pre-evacuation information and evacuation orders) were linke d to evacuation and protection of property (depending on floodplain status). Event specific FRI was not linked to these behaviors and it seems that most did not look for in formation from multiple sources before evacuating, a result that contrasts with th e experience of Dow and Cutter (1998). This probably has to do with differences in speed of event onset. Local government was the key source.
224 Outcome Variables No clear relationship was found be tween understanding of flood related uncertainty and either threat perception or mitigative behavior, a result that undermines the assumptions of the NRC (2000; 2006). However, there was some evidence that perception of risk influenced relative con cern. Additionally, potential relationships were found between threat perception and behavior. In all cases, associations were positive. Exploring Figure 2.2: The Model of Specifi c Flood Risk Messages, Settings, and Perception There is no current literature regarding situational and cognitive factors and their relationships with the percepti on of specific flood risk messages or frames. This portion of the research was the most exploratory and conclusive results were limited. A few broad general trends were id entified, however. The relations hips listed here simply indicate that a variable or factor may consis tently contribute to th e formation of differing perceptions of size, likeli hood, uncertainty and concern associated with specific messages. While control did not figure heavily in th e general revised model (Figures 6.1 and 6.2), it was related to differing perceptions of both uncertainty and concern, and less clearly, to size and likelihood. General worry ap peared to be weakly related to varying perceptions of likelihood. FRI was linked to perceptions of uncertainty as well as size. Complexity of understanding was associated with perception of un certainty and might indirectly influence concern. Va rying perceptions of uncertain ty were also predicted by education level. Age was rela ted to choices for most c oncerning and least concerning
225 floods. Perceived size and likelihood also a ppeared to influence associated concern levels, but in what ways and to what extent is unclear. Location and experience were not included in any of the regression models, but might underlie some of variables that were included. Judging Relative Effectiveness As did the general regression results, comparative analysis of effectiveness indicated that understanding of uncertainty and persuasion were not necessarily connected. Sjoberg (2000), Beeh ler et al. (2001) and Bell an d Tobin (2007) have also found this to be the case a nd these results support resear chers who have questioned the linear association between understanding, attitude and behavior (e.g. Tierney, 1993; Valente et al., 1998; Mileti and Peek, 2002). Response rates for both the hundred year and one percent chance descriptions were very sensitive to whether the question asked dealt with understanding or concern. Th e hundred year flood description performed significantly worse than both probability based descriptions with regards to understanding uncertainty; the one percent chance description had the highest score. The opposite was true in the comparison of concern, though no significant difference was found between the 26 percent chance descrip tion and the hundred year flood description. Which description is most effective in ge neral risk communication perhaps depends on whether understanding or persuasion is the goal. Risk managers and communicators may have to decide which is ethica lly and/or practically more important in the current system of flood loss mitigation and distribution.
226 The 26 percent chance description appeared to be the most effective when both understanding and persuasion were considered. Though this description was rated as most likely by the highest proportion of respondents, the 26 percent description scored well on the combined scale in part because it di d not stand out. It ranked second in the comparative assessments of both understand ing and concern and did not elicit the extreme responses associated with the other two descriptions. However, similar research in another flood prone community recorded strong negative reactions to the 26 percent chance description (Bell and Tobin, 2007). Qualitative data showed that many individuals had trouble attaching a physical size to the descriptions and indicated a ge neral frustration with all of the terms. Additionally, participants reac ted to different aspects of the descriptions and came to contradictory conclusions. The divergent in terpretations and reas oning associated with the 26 percent chance description reduce its vi ability as a widely applicable method of communicating the risk related to policyÂ’s be nchmark flood. Future research to clarify relationships between perceive d size, likelihood, and concern, and specific descriptions would be useful in assessing the generali zability of all flood risk communication. Improving Flood Ri sk Communication When asked to describe flooding, approxi mately 20 percent of those in the hundred year floodplain used a return period to do so. However, in the total sample, the most common descriptors were relation to a pa st event and stage relative to a landmark. This result lends support to the ASFPMÂ’s s uggested use of stages linked to physical
227 markers in flood risk communication (ASFPM, 2007). None of the survey participants used probability or cumulative probability to describe flooding. In response to direct questioning, particip ants most frequently cited flood level or a combination of level and frequency of fl ooding as most concerning. Potential damage was the third most common answer. Less than ten percent chose fre quency alone, a result consistent with other studies (Bell and T obin, 2007). However, qual itative data indicated that issues of timing were a motivating factor in relative concern associated with specific descriptions. Additionally, quantitative analys is showed no clear re lationships between the flood description chosen as most concerni ng and perceived relative size or likelihood. Focus group particip ants made several suggestions for improving communication of both specific and general flood risk. Two se parate conceptualizations of understanding were put forth and groups emphasized the importance of communicating what flooding might do to a person and a personÂ’s community as well as (or instead of) how often flooding could occur. Recommendations focuse d on regulatory practices and the creation of collective memory. A collective memory woul d in turn make it more feasible to use relation to past events as a description of possible or im minent flooding. Their specific ideas reflected suggestions made in the li terature and included emphasizing damage (NRC, 2000; Smith, 2000), employing visual aids in public places (Siegrist and Gutcher, 2006), and instituting early and continui ng education (Frech, 2006). By including interviews and other material on the website www.floodsafety.org, Frech has also tried to bring personal experiences of flooding to the public, something that focus groups recommended when discussing collections of individual and public experiences and family histories.
228 CONCLUSIONS The general conceptual model outlined in Figure 2.1 did an adequate job of identifying important contributing factors for most outcome variables. Each factor was represented in at least one of the models. Se verity of impact was the most influential overall and probably filtered th e effects of other situationa l and cognitive variables as well as outcome variables on understanding, at titude, and behavior. Specific models related to the understanding of flood related uncertainty and flood preparation were weak, though, and need substantial improvement. Reconceptualizing what constitutes understanding might help to id entify key associations with situational and cognitive factors and better evaluate the relationship of understanding to attitude and behavior. Models for specific perceptions of flood ri sk descriptions were also generally weak. A much finer scale model that more h eavily focuses on cognitive variables may be necessary. The weakness of these models is likely related to the inconsistent interpretations of specific messagesÂ’ percei ved size, likelihood, a nd associated concern discovered in multiple analyses and illustra ted in Table 7.1. This example shows the extent of variation between focus group responses to the 26 percent chance flood description. Both the one percent and the hundred year flood descriptions had similar variation. Specific messages should be broken do wn into components in order to identify what aspects of the flood risk descriptions (numbers, timeframe, concepts, words) people are reacting to and what characteristics are associated with specific reactions. Additionally, this type of analysis might allow researchers to identify messages most appropriate to specific and general risk communication. For instance, the hundred year flood was consistently interpreted as the largest and the least likely. Concern
229 rankings and ratings differed, but if people rated it least concerni ng or ranked it low, it was usually because they didnÂ’t think it woul d happen. The description may not be useful for general risk communi cation, but might be very effective in event based communication, especially if combined with landmarks and damage estimates. Timing, the primary source of interpretive difference, would be removed from consideration, and people seemed to think of the hundred year flood as Â“the big one.Â” Reactions to the probabilistic descriptions app eared to be much more complicated and will require indepth qualitative analysis. Table 7.1. Perceptions and Concern: 26 % Chance Description Group Relative Threat Perception Reason 1 DoesnÂ’t Seem Threatening Â“It doesnÂ’t really re gister with you that it could happen every year.Â” 2 Most Concerning Â“ItÂ’s the most likelyitÂ’s more realistic.Â” 3 Least Concerning Â“The 26% flood is minor, like a sewer back-up.Â” This research indicated that current me thods of describing general flood risk are not broadly effective, regardless of the crit eria used for evaluation. The responses of both the survey participants and the focus group me mbers emphasized the concrete rather than the abstract. The descriptions evaluated here do the opposite. Using descriptions based on common experience of the physical world mi ght improve both types of understanding identified by the focus groups as well as persuasion. However, flood policy, communication of flood policy, and general communication of flood risk stress a
230 particular flood rather than the condition of flooding. If the goal is both understanding and persuasion, managers and communicators might be wise to shift the focus from the specific probability of a certain flood to th e constant possibility of flooding in general, promoting a culture of mitigation and resilien ce rather than regulation and response. This approach may also better reflect the wa y many people think of flooding (a condition rather than a gradation). The following are five key ideas that should be taken from this research: 1. Experience, especially impact seve rity, was the most influential variable across outcome variables as well as physical and social space. Naming of past experience may also be important to future assessments of flooding and flood risk. 2. The results of this research do not support the assumption that understanding leads to persuasion, an idea that underlies many risk communication campaigns as well as hi erarchical models of behavior change. 3. Interpretation of the flood risk me ssages addressed in this research (especially the probabilistic descri ptions) was highly individualized and inconsistent. Broadly effective communication strategies may thus be difficult to implement. The NRC (2006) has recommended multiple messages through multiple channels, but this could be detrimental if messages meant to positively reinfo rce each other or reach separate individuals incite cont radictory responses in the same individual. 4. Regulatory practice likely influences perception and behavior in both positive and negative ways. Current approaches to management and development may emphasize the probab ility of a specific flood and deemphasize the possibility of flooding in general. 5. The 26 percent chance description fa red best in comparisons of overall effectiveness in this study site. Ho wever, it may be the best of the worst. Only two of the survey and focus group participants used probabilities to describe floodi ng. Their emphasis on concrete individual and collective referenc es might enhance understanding and /or motivate behavior more effec tively than current abstractions.
231 Contributions and Generalizability This research made both theoretical and practical contributio ns to the existing literature. In the past, much of the res earch regarding flood related perception and behavior has focused on a limited number of isolated situational and cognitive factors. In this project, the broad combined effects of these factors on flood related understanding, attitude and mitigative behavior were examined. This approach contributes to a more situated conceptual understanding of percep tual and behavioral outcomes. However, further qualitative investigat ion is necessary to better ground flood related experience, perception, and behavior in the dail y lives of connected individuals. Additionally, researchers and practitione rs have called for testing of risk messages. Many have voiced concern re garding the use of hundred year flood terminology in public communication of flood policy and flood risk, fearing it masks uncertainty and encourages risk dichotomies. The risk associated with the benchmark flood is now publicly framed in multiple wa ys, but testing of all messages has been limited. This project addressed the gap in the literature in two ways. First, a comparison of the effectiveness of three common met hods of framing flood risk was undertaken. Second, the broad framework used to situate general flood related perception was applied to the three specific flood risk messages in order to identify pot ential patterns of preference and interpre tation over spatial and social gr oups. While analysis provided preliminary answers to these practical inquirie ss, results also raised important questions regarding interpretive variab ility, the ways in which peopl e Â“understand,Â” as well as the value of emphasizing probability versus possi bility and specific events over general conditions.
232 This project focused on flooding, but result s can be extended to other forms of hazard communication and perception. Quantita tive and qualitative analyses supported the adoption of a broad conceptualization of risk communication in both theory and practice. Results illustrated th e tensions between the changing individual construction of meaning and the expectations of message centered risk communication operating under narrow assumptions, as well as its evaluation. Additionally, the possible relationships of regulated space to communication, perception and behavior identified in this research are relevant in other contexts where specific areas are politically delineated as hazardous. These results also contribute to more ge neral debates over public understanding of probability and uncertainty as well as those regarding the connections between understanding, attitude and behavior. This study was relatively small, however, and both the quantitative and qualitative results need to be compared to other data se ts in order to differentiate the general from the particular. The Towns of Union and Vestal were chosen, in part, because of their similarity in size and location. This clearly limits generalizability. Both Towns have recent flood experience and are fairly small, predominantly white and non-Hispanic, with high proportions of home owners and English speakers. Future research in communities large and small, experienced and non-experienced, those quickly growing with high proportions of immigrants and those that are racially a nd ethnically homogenous will help identify components broadly relevant to the development of effective policy and communication. Making research more fully pa rticipatory would also assist in building more effective and flexible policy, communicat ion and enforcement at the local level. The ultimate hope, of course, is that when comb ined, these results might help researchers,
233 practitioners and residents be tter understand the human role in flood hazards and reduce suffering and losses. FUTURE RESEARCH Several areas of future research were iden tified over the course of this project. First, all models would benefit from improved measurement of understanding of uncertainty. Second, interactions and higher order functions should be employed in the regression models where appropriate. Third, flood risk infrastructure needs to be detangled and frequency included. A lot of valu able information is currently hidden in the aggregation. Fourth, future research s hould examine not just how situational and cognitive factors control one another, but how they produce one another. Lastly, the influence of Â“outcomeÂ” variables on situa tional and cognitive factors needs to be investigated while taking into account the passa ge of time. Researchers can help practitioners im prove communication by continuing to examine: 1) what people want to know about flooding, 2) how they talk about flooding, 3) what concerns them about flooding and w hy, and 4) who they talk to about flooding and who they want to talk to them. It is un likely that the answers to these questions will be identical across physical or social space, but common them es might be identified. In this study, participant comments pointed out th e importance of experience, the naming of experience, outlook, mitigative behavior, and information networks in forming understandings of and attitudes toward flood ri sk. Researchers must continue to tease out the complicated relationships between these fa ctors and others identif ied in the literature
234 in order to improve our understanding of not only perception and behavior, but the practical role of risk communication in cr eating resilient individuals and communities.
235 REFERENCES CITED Anderson, C. (2001). The Impact of Ethnicity on Risk Communication in San Marcos, Texas. Unpublished Masters of Applied Geograp hy Directed Research Paper. San Marcos, TX: Southwest Texas State Un iversity, Department of Geography. Andsager, J. L., Bemker, V., Choi, H. L. (2006) Perceived Similarity of Exemplar Traits and Behavior; Effects on Message Evaluation. Communication Research 33(1): 3-18. Arkin, E. (1989). Translation of Risk Communication for th e Public: Message Development. In V. Covello, D. McCallum, and M. Pavlova (eds), Effective Risk Communication: The Role and Responsib ility of Government and Nongovernment Organizations. New York: Plenum Press. ASFPM (Association of State Fl oodplain Managers). (2007). National Flood Programs and Policies in Review Â– 2007, Draft Version Available at: http://www.floods.org Asgary, A., Willis, K.G. (1997). Household Be haviour in Response to Earthquake Risk: An Assesment of Alternative Theories. Disasters 21(4): 354-365. Avery, G.F. (1973). The Town of Vestal: 1823-1973 Vestal, NY: Vestal Sesquicentennial Corp. Beacco, J. C., Claudel, C., Doury, M., Petit, G., Reboul-Toure, S. (2002). Science in Media and Social Discourse: New Cha nnels of Communication, New Linguistic Forms. Discourse Studies 4(3), pp. 277-300. Beehler B. P, McGuinness, B. M, Vena J. E. (2001). Polluted Fish, Sources of Knowledge, and the Perception of Risk : Contextualizing African American AnglerÂ’s Sport Fishing Practices. Human Organization 60(3). Bell, H.M. (2004 ). Efficient and Effective? The Hundred Year Flood in the Communication and Perception of Flood Risk Unpublished Masters Thesis. Tampa, FL: University of South Florida, Department of Geography, pp. 162.
236 Bell, H.M. and Tobin, G.A. (2007) Efficient and Effective? The Hundred Year Flood in the Communication and Pe rception of Flood Risk. Environmental Hazards 30. Belsten, L. (1996). Environmental Risk Co mmunication and Comm unity Collaboration. In S. Muir and T. Veenendall (eds), Earthtalk: Communication Empowerment for Environmental Action. Westport, CT: Praeger. Bernard, H.R. (1994). Research Methods in Anthropology: Qualitative and Quantitative Approaches. Thousand Oaks, CA: Sage Publications. Boisot, M. (1995). Information Space: A Framework for Learning in Organizations, Institutions and Culture. London: Routledge. Boyer, E. W., Goodale, C. L., Jaworski, N. A., Howarth, R. W. (2002). Anthropogenic Nitrogen Sources and Relationships to Riverine Nitrogen Export, NE, USA. Biogeochemistry 57 and 58: 137-169. Brenot, J., Bonnefous, S., Marris, C. (1998). Testing the Cultural Theory of Risk in France. Risk Analysis 18(6). Brubaker, J. (2002). Down the Susquehanna to the Chesapeake University Park, PA: The Pennsylvania State University Press. Burby, R., French, S., Cigler, B., Kaiser, E.m Moreau, D., and Stiftel, B. (1985). Flood Plain Land Use Management. Boulder, CO: Westview Press, Inc. Burby, R.J. (2000). Land-Use Planning for Flood Hazard Reduction: The United States Experience. In Parker, D.J. (ed), Floods: Volume II. London: Routledge. Burger, J., Pflugh, K. K., Lurig, L., Von Ha gen, L. A.,Von Hagen S. (1999). Fishing in Urban New Jersey: Ethnicity Affects Information Sources, Perception, and Compliance. Risk Analysis 19(2). Burton, I. and Kates, R. (1964). The Perception of Natural Hazards in Resource Management. Natural Resources Journal 3. Callaghan, J. (1989). Reaching Target Audience s with Risk Information. In V. Covello, D. McCallum, and M. Pavlova (eds), Effective Risk Communication: The Role and Responsibility of Government and Nongovernment Organizations. New York: Plenum Press. Changnon, S.A. (2000). The Record 1993 Mississ ippi River Flood: A Defining Event for Flood Mitigation Policy in the United States. In Parker, D.J. (ed), Floods: Volume I. London: Routledge.
237 Connelly, N. A., Knuth, B. A. (1998). Ev aluating Risk Communi cation: Examining Target Audience Perceptions about Four Presentation Formats for Fish Consumption Health Advisory Information. Risk Analysis 18(5). Cook, G., Pieri, E., Robbins, P. T. (2004). Â‘T he Scientists Think And The Public FeelsÂ’: Expert Perceptions Of The Discourse Of GM Food. Discourse & Society 15(4): 433-449. Covello, V., McCallum, D., and Pavlova, M. (1989). Priinciples and Guidelines for Improving Risk Communication. In V. C ovello, D. McCallum, and M. Pavlova (eds), Effective Risk Communication: The Role and Responsibility of Government and Nongovernment Organizations. New York: Plenum Press. Covello, V.T., von Winterfeldt, D. and Slovic, P. (1986) Communicating Risk Information to the Public. Risk Abstracts 3: 1-14. Cutter, S.L. (1993). Living with Risk: The Geography of Technological Hazards New York: Routledge. Cutter, S.L., Boruff, B. J., Shirley, W. L. (2003). Social Vulnerabil ity to Environmental Hazards. Social Science Quarterly 84(2). Daggett, C.J. (1989). The Role of Risk Commun ication in Environmental Gridlock. In V. Covello, D. McCallum, and M. Pavlova (eds), Effective Risk Communication: The Role and Responsibility of Governm ent and Nongovernment Organizations. New York: Plenum Press. Dunne, T. and Leopold, L.B. (1978). Water in Environmental Planning New York: W.H. Freeman and Company. Emergency Management Institute. (2007). Community Rating System Resource Center Available at: www.training.fema.gov/EMIWEB/CRS/ Environmental Agency. (2005). Improving Community and Citizen Engagement in FRM Decision Making, Delivery and Flood Response Bristol: Environment Agency. Available at: http://publications.environmentagency.gov.uk/pdf/SCHO1005BJTC-e-e.pdf Eisenstadt, P. (2005). The Encyclopedia of New York State Syracuse, NY: Syracuse University Press. Faber, S. (1996). On Borrowed Land: Public Policies for Floodplains. Cambridge, MA: Lincoln Institute of Land Policy.
238 FEMA. (1995). Managing Floodplain Development in Approximate Zone Areas: A Guide for Obtaining and Developing Base (100-Year) Flood Elevations. Washington, D.C.: U.S. Government Printing Office. FEMA (Federal Emergency Management Agency). (2002). National Flood Insurance Program: Program Description Available at: www.fema.gov/library/vie wRecord.do?id=1480 FEMA. (2003) Florida State Profile. Ava ilable at: http://www .fema.gov/cis/FL.pdf. FEMA (Federal Emergency Management Agency). (2006a). Flood Statistics Available at: http://www.floodsmart.gov/floodsmart/pages /statistics.jsp;jsessionid=CFBC9E0E 291618DD5993AD29BCF29E1D FEMA (Federal Emergency Management Agency). (2006b). New York Severe Storms and Flooding Available at: https://www.fema.gov/news/event.fema?id=6485 FEMA (Federal Emergency Management Agency). (2006c). National Flood Insurance Program: Answers to Questions about the NFIP. Booklet MitDiv-2, F-084 Available at: http://www.fem a.gov/business/nfip/qanda.shtm FEMA (Federal Emergency Management Agency). (2007a). Significant Flood Events Available at: www.fema.gov/business/nfip/statistics/sign1000 FEMA (Federal Emergency Management Agency). (2007b). Community Rating System: Eligible Communities Available at: www.fema.gov/business/nfip/crs.shtm. Finucane, M., Alhakami, A., Slovic, P., and Johnson, S. (2000). The Affect Heuristic in Judgments of Risks and Benefits. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Fiori, J.V. (1990) A History of the Town of Union: Its Settlement, Growth and Development Union, NY: Town of Union. Flooddamagedata.org. Flood Damage in the United Stat es, 1926-2003: A Reanalysis of the National Weather Service Estimates Available at: http:/ /flooddamagedata.org/ Fordham, M. (2000). Participatory Pla nning for Flood Mitigation: Models and Approaches. In Parker, D.J. (ed), Floods: Volume II. London: Routledge. Frech, M. (2005). Flood Risk Outreach and the PublicÂ’s Need to Know. Journal of Contemporary Water Research and Education 130:61-69
239 Garson, G.D. (2007). Logistic Regression. Statnotes: Topics in Multivariate Analysis. Available at: www2.chass.ncsu.e du/garson/pa765/statnote.htm. Gerber, B. J., Neeley, G. W. (2005). Per ceived Risk and Citizen Preferences for Governmental Management of Routine Hazards. The Policy Studies Journal 33(3). GFWNFPF. (2004). Reducing Flood Losses: Is th e 1% Chance Flood Standard Sufficient? Report of the 2004 Assembly of th e Gilbert F. White National Flood Policy Forum, Washington, D.C. Godber, A.M. (2005). Urban Floodplai n Land-Use Acceptable Risk? The Australian Journal of Emergency Management 20 (3): 22-26. Goddard, J. (1961). The Cooperative Program in the Tennessee Valley. In G. White (ed), Papers on Flood Problems: Department of Geography Research Paper No. 70. Chicago, IL: University of Chicago Press. Gordon, J.H. (1966). The Susquehanna Flows on: A Narrativ e of the Development of the Southern Tier Ithica, NY: Wilcox Press. Grasmck, D., Scholz. R. W. (2005). Risk Perception of Heavy Metal Soil Contamination by High-Exposed and LowExposed Inhabitants: The Role of Knowledge and Emotional Concerns. Risk Analysis 25(3). Griffin, R. J., Neuwirth, K ., Giese, J., Dunwoody, S. (2002). Linking the HeuristicSystematic Model and Depth of Processing. Communication Research 29(6): 705-732. Gruntfest, E., Carsell, K., Plush, T. (2002). An Evaluation of the Boulder Creek Local Flood Warning System Colorado Springs: University of Colorado at Colorado Springs. Gustafson, P. E. (1998). Gender Difference s in Risk Perception: Theoretical and Methodological Perspectives. Risk Analysis 18(6). Halpern-Felsher, B. L., Millstein, S. G., Elle n, J. M., Adler, N. E., Tshann, J. M., Biehl, M. (2001). The Role of Behavioral Experiences in Judging Risks. Health Psychology 20(2): 120-126. Hammett, K.M., and DelCharco, M.J. (2005) Estimating the Magnitude and Frequency of Floods for Streams in We st-Central Florida, 2001: U.S. Geological Survey Scientific Investigations Report 2005-5080, 15 p., plus app.
240 Hance, B., Chess, C., and Sandman, P. ( 1989). Improving Dialogue with Communities: A Risk Communication Manual for Government In V. Covello, D. McCallum, and M. Pavlova (eds), Effective Risk Communication: Th e Role and Responsibility of Government and Nongovernment Organizations. New York: Plenum Press. Heath, R. l., Gay, C. D. (1997). Risk Co mmunication; Involveme nt, Uncertainty, and ControlÂ’s Effect on Information Scanning and Monitoring by Expert Stakeholders. Management Communication Quarterly 103: 342-372. Holmes, R. (1961). Composition and Size of Flood Losses. In G. White (ed), Papers on Flood Problems: Department of Geography Research Paper No. 70. Chicago, IL: University of Chicago Press. Hunt, C.B. (1967). Physiography of the United States San Francisco, CA: W.H. Freeman and Company. Johnson, B. (1989). Qualitative Risk Assessm ent: Experiences and Lessons. In V. Covello, D. McCallum, and M. Pavlova (eds), Effective Risk Communication: The Role and Responsibility of Governm ent and Nongovernment Organizations. New York: Plenum Press. Johnson, B. B. (2005). Testing and Expanding a Model of Cognitive Processing of Risk Information. Risk Analysis 25(3). Kachmor, G.A., Goeller, A.J. (2005). Broome State Forests Unit Management Plan, Draft Available at: http://www.dec.state.ny.us/website/dlf /publands/ump/reg7/broomeump.html Kasemir, B., Jaeger, C., and Jaeger, J. (2003) Citizen Participation in Sustainability Assessments. In Kasemir B., Jaeger, J., Jaeger, C., and Gardner, M. (eds), Public Participation in Sustainab ility Science: A Handbook. Cambridge, UK: Cambridge University Press. Kasperson, R., Renn, O., Slovic, P., Brown, H., Emel, J., Goble, R., Kasperson, J., and Ratick, S. (1988). The Social Amplificati on of Risk: A Conceptual Framework. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Kasperson, R.E. and Stallen, P.J.M., eds. (1991). Communicating Risks to the Public: International Perspectives Dordrecht: Kluwer Academic Publishers. Kates, R. and White, G. (1961). Floo d Hazard Evaluation. In White (ed), Papers on Flood Problems: Department of Geography Research Paper No. 70. Chicago, IL: University of Chicago Press.
241 Krimsky, S. and Plough, A. (1988). Environmental Hazards: Communicating Risks as a Social Process. Dover, MA: Auburn House Publishing Company. Krueger, R.A. and Casey, M.A. (2000). Focus Groups: A Practical Guide for Applied Research Thousand Oaks, CA: Sage Publications. Kusler, J. (1982). Regulation of Flood Hazard Areas to Reduce Flood Losses: Volume 3. Washington, D.C.: U.S. Water Resources Council. Larson, L., Plasencia, D. (2001). No Adve rse Impact: A New Direction in Floodplain Management. Natural Hazards Review 2 (4), 157-181. Lave, T.R. and Lave, L.B. (1991). Public Per ception of the Risks of Floods: Implications for Communication. Risk Analysis, 11(2). Learner, L.E. (1989). Four Corners: Vestal, New YorkÂ’s History as Revealed by the Story of Its Main Intersection Vestal: Town of Vestal. Lee, C.J., Scheufele, D. A., and Lewenstein, B. V. (2005). Public Attitudes toward Emerging Technologies: Examining the In teractive Effects of Cognitions and Affect on Public Attitudes toward Nanotechnology. Science Communication, 27(2): 240-267. Lindell, M. K., Perry, R. W. (2000). House hold Adjustment to Earthquake Hazard; A Review of Research. Environment and Behavior 32 (4): 461-501. Loges, W. E. (1994). Canaries in the Coal Mi ne: Perceptions of Thre at and Media System Dependency Relations. Communica tion Research, 21(1): 5-23. Lupton, D. (1999). Introduction: Risk and Sociocultural Theory. In D. Lupton, ed., Risk and Sociocultural Theory: New Directions and Perspectives Cambridge: Cambridge University Press: 1-11. Marris, C., Langford, I. H., OÂ’Riordon, T. ( 1998). A Quantitative Test of the Cultural Theory of Risk Perceptions: Compar ison with the Psychometric Paradigm. Risk Analysis 18(5). Meredith, S.M. (1999) Images of America: Union Charleston, S.C.: Arcadia Publishing. Mileti, D., Farhar, B., Fitzpatrick, C, and He lmericks, S. (1991). Public Perception and Response to the Parkfield California Eart hquake Prediction. In B.J. Garrick and W.C. Geckler (eds), The Analysis, Communication, and Perception of Risk. New York: Plenum Press.
242 Mileti, D.S. and Darlington, J.D. (1997). The Ro le of Searching in Shaping Reactions to Earthquake Risk Information. Social Problems 44(1):89-103. Mileti, D, and OÂ’Brien. (1992). Warnings during Disaster: Norm alizing Communicated Risk. Social Problems 39(1): 40-57. Mileti, D. and Peek, L. (2000). The Social Psychology of Public Response to Warnings of a Nuclear Power Plant Accident. Journal of Hazardous Materials 75: 181Â–194. Mileti, D. and Peek, L. (2002). Understanding Individual and Social Characteristics in the Promotion of Household Disaster Preparedness. In New Tools for Environmental Protection: Education, Inform ation, and Voluntary Measures Washington, DC: National Academy of Sciences. McQuail, D. and Windhal, S. (1981). Communication Models for the Study of Mass Communications. New York: Longman, Inc. Montz, B. (1982). The Effect of Location on the Adoption of Hazard Mitigation Measures. Professional Geographer 34(4): 416-423. Morgan, D.L. (1998). Planning Focus Groups Thousand Oaks, CA: Sage Publications. Murphy, F. (1958). Regulating Floodplain Development. Chicago: University of Chicago Press. National Research Council (NRC). (1995). Flood Risk Management and the American River Basin: An Evaluation. Washington, D.C.: National Academy Press. National Research Council (NRC). (2000). Risk Analysis and Uncertainty in Flood Damage Reduction Studies. Washington, D.C.: National Academy Press. National Research Council (NRC). (2006). Completing the Forecast: Characterizing and Communicating Uncertainty for Better Decisions Using Weather and Climate Forecasts. National Academy Press, Washington, D.C. Newton, D.W. (1987). TVA Practice in Flood Fr equency and Risk Analysis. In V.P. Singh (ed), Application of Frequency and Risk in Water Resources: Proceedings of the International Symposium on Flood Frequency and Risk Analyses. Dordrecht, Holland: D. Reidel Publishing Company. New York State Division of Envir onmental Conservation (NYSDEC) (2005). Comprehensive Wildlife Conservation Strategy for New York Available at: www.nyfoa.org/docs/NYComprehensive WildlifeConservationStrategy.pdf
243 National Oceanographic and Atmosp heric Administration (NOAA), 2006. Hurricane Agnes Available at: www.erh.noaa.gov/er/marfc/Flood/agnes.html National Weather Service (NWS). (2006a). Local Climate Information for Binghamton, New York National Weather Service Forecast Office, Binghamton, New York. Available at: http://www.er h.noaa.gov/bgm/climate/bgm.shtml National Weather Service (NWS). (2006b). Flood of June, 2006 Available at: http://www.erh.noaa.gov/bgm/WeatherEvents/Flood/june2006/ National Weather Service (NWS). (2006c). Now Data Â– NOAA Online Weather Data. National Weather Service Forecast Office, Philadelphia-Mt. Holly. Available at: http://www.weather.gov/climate/xmacis.php?wfo=phi Novotny, V., Clark, D., Griffin, R. J., Booth, D. (2000). Risk Ba sed Urban Watershed Management Under Conflicting Objectives Proc. 1st World water Congress of the International Wate r Association (IWA) Paris, France, July 3-7: 144-151. Palm, R., Hodgson, M. (1992).Earthquake In surance: Mandated Disclosure and Homeowner Response in California. Annals of the Associ ation of American Geographers 82(2): 207-222. Parker, D.J. (2000). Managing Flood Hazards and Disasters: International Lessons, Directions and Future Challe nges. In Parker, D.J. (ed), Floods: Volume II. London: Routledge. Perry, C.A. (2000). Significant Floods in the United States During the 20th Century-Â— Measures a Century of Floods. USGS Fact Sheet 024Â–00 Peters, H. P. (1995). The Interaction of Jour nalists and Scientific Experts: Co-Operation And Conflict Between Two Professional Cultures. Media, Culture & Society 17: 31-48. Pielke, Jr., R.A., Down ton, M.W., Barnard Miller, J.Z. (2002). Flood Damage in the United States, 1926-2000: A Reanalysis of National Weather Service Estimates Boulder, CO: UCAR. Powell, D. and Leiss, W. (1997). Mad Cows and MotherÂ’s Milk: The Perils of Poor Risk Communication. Montreal: McGill-QueenÂ’s University Press. The PresidentÂ’s Water Resources Policy Commission. (1950). Water Resources Law. Washington, D.C.: U.S. Government Printing Office.
244 Randall, A.D. (1986). Aquifer Model of the Su squehanna River Valley in Southwestern Broome County, New York. Water Resources Investigations Report 85-4099 Albany, NY: US Geological Survey. Renshaw, E. (1961). The Relationship betw een Flood Losses and Flood-Control Benefits. In G. White (ed), Papers on Flood Problems: Department of Geography Research Paper No. 70. Chicago, IL: University of Chicago Press. Reuss, M. (1993). Water Resources People and Is sues: Gilbert F. White. Fort Belvoir: USACE Office of History. Rohrmann, B. (2000). A Socio-Psychological Model for Analyzing Risk Communication Processes. The Australasian Journal of Disaster and Trauma Studies 2000(2). Ross, T. and Lott, N. (2006). Billion Dollar U.S. Weather Disasters, 1980 Â– 2005. Ashville, N.C.: National Climatic Data Center. Sandman, P. (1989). Hazard Vers us Outrage in the Public Perception of Risk. In V. Covello, D. McCallum, and M. Pavlova (eds), Effective Risk Communication: The Role and Responsibility of Governm ent and Nongovernment Organizations. New York: Plenum Press. Sauer, B. (1999). Embodied Experience: Re presenting Risk In Speech And Gesture. Discourse Studies 1(3): 321-354. Shannon, C.E. and Weaver, W. (1949). The Mathematical Theory of Communication. Urbana, IL: The Univers ity of Illinois Press. Siegrist, M. and H. Gutscher. (2006). Fl ooding Risks: A Comparison of Lay PeopleÂ’s Perceptions and ExpertÂ’s Assessments in Switzerland. Risk Analysis 26(4):971979. Siegrist, M., Keller, C., Kiers, H. A. L. (2005). A New Look at the Psychometric Paradigm of Perception of Hazards. Risk Analysis 25(2): 211-227. Simpson, R.H., Herbert, P.J. (1973). Atlantic Hurricane Season of 1972. Monthly Weather Review 101 (4): 323-333. Sjberg, L. (2000). Specifying factor s in radiation risk perception. Scandinavian Journal of Psychology 41: 169-174. Slovic, P. (1986). Informing and Educating th e Public about Risk. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd.
245 Slovic, P. (1987). Perception of Ri sk. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Slovic, P. (1993). Perceived Risk, Trust and Democracy. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Slovic, P., Fischhoff, B., and Lichtenstein, S. (1976). Cognitive Processes and Societal Risk Taking. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Slovic, P., Fischhoff, B., and Lichtenstein, S. (1979). Rating the Risks. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Slovic, P., Fischhoff, B., and Lichtenstein, S. (1980). Facts and Fears: Understanding Perceived Risk. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Slovic, P., Kunreuther, H., and White, G. ( 1974). Decision Processes, Rationality and Adjustment. In P. Slovic (ed). (2000). The Perception of Risk London: Earthscan Publications Ltd. Smith, D.I. (2000). Floodplain Management: Problems, Issues and Opportunities. In Parker, D.J. (ed), Floods: Volume I. London: Routledge. Smith, G.R. (2006) Partners All: A History of Broome County, NY Virginia Beach: The Donning Company Publishers. Smith, K. and Tobin, G.A. (1979) Human Adjustment to the Flood Hazard London: Longman. Spencer, J. W., Seydlitz, R., Laska, S.,Trich e, E. (1992). The Different Influences of Newspaper and Television News Reports of a Natural Hazard on Response Behaviour. Communication Research 19(3): 299-325. Steg, L. and Sievers, I. (2000). Cultural Theory and Individu al Perceptions of Environmental Risks Environment and Behavior 32(2): 250-269. Stoe, T.W. (1999). Upper Susquehanna Sub-basin: A Water Quality and Biological Assessment Harrisburg, PA: Susquehanna River Basin Commission. Susquehanna River Basin Commission (SRBC), 1998. Susquehanna River Basin: Everyone Lives in a Watershed Harrisburg, PA: SRBC.
246 Susquehanna River Basin Commission (SRBC). (2001). Susquehanna River Basin Precipitation: Average Annual Precipitation from 1961 to 1990 Map. Available at: http://www.srbc.net/gis/view.asp?img=image/140g.jpg Susquehanna River Basin Commission (SRBC). (2006). June 2006 Susquehanna River Basin Flood Event. Available at: www. srbc.net/docs/Flood2006.pdf Susquehanna River Basin Hydrologic Ob serving System (SRBHOS). (2004). Prospectus Available at: http://srbhos.psu.edu/ Susquehanna River Basin Study Coordi nating Committee (SRBSCC). (1970a). Susquehanna River Basin Study, Appendix C: Economy and Geography Susquehanna River Basin Study Coor dinating Committee (SRBSCC). (1970b). S usquehanna River Basin Study. Tang, Z., B.A. Engel, B.C. Pijanowskib, K.J. Lim. (2005) Forecasting Land Use Change and Its Environmental Impact at a Watershed Scale. Journal of Environmental Management 76: 35Â–45. Tierney, K.J. (1993). Socio-economic Aspects of Hazard Mitigation Newark, DE: University of Delaware Disaster Research Center. Available at: http://dspace.udel.edu:80 80/dspace/handle/19716/577 Thompson, J.H. (ed). (1966). The Geography of New York State Syracuse, NY: Syracuse University Press. Tobin, G.A. (1995). The Levee Love Affair: A Stormy Relationship? Water Resources Bulletin, 31(3): 359-367 Tobin, G. A. and Montz, B. (1997). Natural Hazards: Explanation and Integration. New York: The Guilford Press. Tonn, B., Goeltz, R., Travis, C., and Philli ppi, R. (1991). Risk Communication and the Cognitive Representation of Uncertainty. In B.J. Garrick and W.C. Geckler (eds), The Analysis, Communicati on, and Perception of Risk. New York: Plenum Press. Trettin, L and Musham, C. (2000). Is Trust A Realistic Goal Of Environmental Risk Communication? Environment and Behavior 32(3): 410-426. Town of Union. (2006a). Progress Report Â– Town of Union Recertification into the Federal Emergency Management Agency Community Rating System Available at: townofunion.com/.../Community% 20Recertification%20Activities%20%20Progress%20Report%202006%20w-o%20text.pdf
247 Town of Union. (2006b). Town of Union Floodplain Management Plan Â– 2006 Revision Available at: townofunion.com/.../FLOODPLAIN% 20MANAGEMENT%20PLAN%202006% Town of Vestal. (2004). Town of Vestal Comprehensive Plan Vestal, NY: Town of Vestal. Trumbo, C. W. (1999). Heuristic-Systematic Information Processing and Risk Judgment. Risk Analysis 19(3): 391-400. Trumbo, C. W. (2000). Risk, Rhetoric, and the Reactor Next Door. In N. Coppola and B. Karis (eds), Technical Communication, Delibera tive Rhetoric, and Environmental Discourse: Connections and Directions Stamford: Ablex Publishing Corporation. United States Census Bureau. (2000). Census 2000 Demographic Profiles Available at: censtats.census.gov/pub/Profiles.shtml United States Census Bureau. (2006). State and County QuickFacts Available at: http://quickfacts.census.gov/qfd/index.html US Army Corps of Engineers (USACE). (1969). Broome County, NY Floodplain Information Baltimore, MD: USACE. US Army Corps of Engineers (USACE). ( 1999). Ice events in the Susquehanna River Basin I ce Engineering 21: 1-6. US Geological Survey (USGS). (2006a). USGS Surface-Water Data for the Nation. National Water Informati on System: Web Interface. Available at: http://waterdata.usgs.gov/nwis/sw US Geological Survey (USGS). (2006b). Historical Flood Peaks and Peaks During the Flood of June 28-29, 2006 at Selected U.S. Geological Survey Streamflow Gaging Stations In New York. Available at: ny.water.usgs.gov/htmls/pub/floodpeaks.jun06.08AUG06.pdf United States Water Resources Council. (1971). Regulation of Flood Hazard Areas to Reduce Flood Losses: Volume 1. Washington, D.C.: U.S. Government Printing Office. United States Water Resources Council. (1977). Guidelines for Determining Flood Flow Frequency: Bulletin #17A of the Hydrology Committee. Washington, D.C.: U.S. Water Resources Council. United States Water Resources Council. (1979). A Unified National Program for Flood Plain Management. Washington, D.C.: U.S. Government Printing Office.
248 U.S. Water Resources Council. (1982). Guidelines for Determining Flood-Flow Frequency (Revised): Bulletin 17B Washington, D.C.: U.S. Government Printing Office. Valente, T.W., Paredes, P., Poppe, P.R. ( 1998). Matching the Message to the Process: The Relative Ordering of Knowledge, A ttitudes, and Practices in Behavior Change Research. Human Communication Research 24 (3): 366-385. Van Diver, B.B. (1985). Roadside Geology of New York Missoula, MT: Mountain Press Publishing Company. Voigt, W. (1972). The Susquehanna Compact New Brunswick, NJ: Rutgers University Press. Wakefield, S. E. L., Elliott, S. J. (2003) Constructing the News: The Role of Local Newspapers in Environmental Risk Communication. The Professional Geographer 55(2): 216-226. White, G.F. (1945). Human Adjustment to Floods: Department of Geography Research Paper No. 29 Chicago: University of Chicago Press. White, G. (1964). Choice of Adjustments to Floods: Department of Geography Research Paper No. 93. Chicago, IL: University of Chicago Press. Wilkinson, I. (2001). Social Theories of Ri sk Perception: At On ce Indispensable and Insufficient. Current Sociology 49(1): 1Â–22. Wynne, B. (1992). Science and Social Responsib ility. In J. Ansell and F. Wharton (eds), Risk: Analysis, Assessment and Management. Chichester: John Wiley and Sons Ltd. Yager, R.M. (1986). Simulation of GroundWater Flow and Infiltration from the Susquehanna River to a Shallow Aqui fer at Kirkwood and Conklin, Broome County, New York. Water Resources Investigations Report 86-4123 US Geological Survey, Ithaca, NY. Yahner, R.H., 2000. Eastern Deciduous Fo rest: Ecology and Wildlife Conservation. University of Minneapolis Press, Minneapolis, MN. Zampogna, D.M. (2006). Susquehanna Rive r Basin Report: June 2006 Susquehanna River Basin Event. I n the Flow: The Online Newsletter of the PA-AWRA Fall, 2006.
249 Zerubavel, E. (1997). Social Mindscapes: An Inv itation to Cognitive Sociology Cambridge, MA: Harvard University Press.
251 Appendix A: Questionnaire Introductory Statement for Questionnaire Hello, my name is Heather Bell. IÂ’m a graduate student at the Univer sity of South Florida and IÂ’m conducting a survey on peopleÂ’s at titudes toward flooding. IÂ’d like to ask you some questions about your experience w ith flooding and how you feel about the likelihood of future floods. The study is not funded by any company or corporation, and I am NOT trying to sell you anything. You may c hoose to participate or not to participate and you may stop the interview at any time. You will not bene fit directly from participating and will not receive payment other than my sincere thanks, but your participation will help me finish my dissert ation and might also in fluence the way people talk about flooding and flood policy. The survey takes about 15 minutes to complete. Do you have any questions? May I continue? The results of this study ma y be published. However, your answers will be combined with data from other people and the published results will not include your name or any other information that would personally identify you in any way. Your privacy will be maintained and research records kept confidential to the extent of the law. Surveys will be given ID numbers (no names will be used) an d will be kept locked in a cabinet in my locked office. In addition to myself and my advisor, employees of the Department of Health and Human Services, the USF Institutio nal Review Board, its staff, and any other individuals acting on behalf of USF may inspect the records from this research project. There are no known risks, but if you experience any study related harm, or if you simply have questions or would like more informa tion, please contact Dr. Graham Tobin at the University of South Florida at 813-974-4932. He can also be reached through e-mail at email@example.com. If you have questions a bout your rights as a person who is taking part in a research study, you may contact the Division of Rese arch Integrity and Compliance of the University of South Florida at (813) 974-5638. If you would like to receive an overview of th e results when the research is completed, please e-mail me at firstname.lastname@example.org or send a request to: Heather Bell University of South Florida 4202 E Fowler Ave., NES 107 Tampa, FL 33620
252 Appendix A (Continued) Date_________ Community_____________ Parcel # __________ In 100 Yes (1) No (0) In 500 Yes (1) No (0) Interviewer ________ FIRST IÂ’D LIKE TO ASK YOU SOME GENERAL QUESTIONS ABOUT YOUR EXPERIENCE WITH FLOODING. 1. Have you ever been affected by flooding? Yes (1) No (0) If no, skip to question 5. 2. IÂ’d like you to think back to the worst flood youÂ’ve been affected by. How did it affect you? __________________________________________________________________ __________________________________________________________________ __________________________________________________________________ 3. How would you describe the size of that flood? ________________________________________ 4. In your lifetime, how many times has your home or property been flooded? _____ Begin questions here after skips. 5. In your opinion, what kinds of things contribute to flooding? __________________________________________________________________ __________________________________________________________________ 6. On a scale of 1 to 7, how knowledgeable do you think you are about flooding? Not at All Knowledgeable Extremely Knowledgeable 1 2 3 4 5 6 7 NEXT IÂ’D LIKE TO ASK YOU AB OUT ANY MEASURES YOUÂ’VE TAKEN AGAINST FLOODING, AS WELL AS YOUR FAMILIARITY WITH FLOOD PROGRAMS. Give interviewee general measures card.
253 Appendix A (Continued) 7. In your lifetime, have you done any of th e things listed on th e card in hopes of reducing the possibility of flood damages? Raised House Above Designated Flood Level Yes (1) No (0) Raised Utilities Above Designated Flood Level 1 0 Purposely Bought/Rented Outside the Floodplain 1 0 Checked with Neighbors Regard ing Past Flood Levels 1 0 Â“FloodproofedÂ” Home 1 0 Other (Please describe) 1 0 None 1 0 8. During the June floods, what measures, if any, did you take to protect yourself and your property from flood damage? (Prompt, if necessary, using the following) Sandbagged Property Yes (1) No (0) Moved Belongings to Higher Ground 1 0 Evacuated 1 0 Other (Please describe) 1 0 None 1 0 9. Do you have flood insurance? Yes (1) No (0) DK (555) 10. Do you currently live in a Special Flood H azard Area? Yes (1) No (0) DK (555) 11. On a scale of 1 to 7, how familiar are you with the National Flood Insurance Program? Completely Unfamiliar Completely Familiar 1 2 3 4 5 6 7 If greater than 1, How did you learn about it? ______________________ 12. Do you consider your house to be at low (1), medium (2), or high (3) risk of flooding in the future?
254 Appendix A (Continued) HEREÂ’S A CARD LISTING SOME POSSIBLE SOURCES OF FLOOD INFORMATION. WEÂ’LL USE IT TO ANSWER THE NEXT SET OF QUESTIONS. 13. During the June floods, what sources, if any, did you go to for information? Use table below. 14. What kind of information did you look for from each source? 15. During the June floods, what other sources, if any, provided you with information? What kind? Source (Q13-15) Info Type Searched or Received 16. Since the June floods, have you looked for flood related information? Yes (1) No (0) If no, skip to Question 20. 17. What are the three main sources youÂ’ve gone to? Use table to record answers for 17, 18 and 19. 18. What kinds of information di d you look for from each source? 19. Since the flood, about how often did you look for information from each source? Source (Q17) Info Type (Q18) Freq. (Q19) 1X /wk or more (3) 1X per month (2) < 1X per month (1) 3 2 1 3 2 1 Start here after skips.
255 Appendix A (Continued) 20. Since the June flood, which sources, if any, have provided you with information related to flooding that you werenÂ’t actively looking for? Use table to record information for 20, 21, 22. 21. What types of information did you get from each of these sources? 22. About how often did you get information from each of these sources since the flood? Source (Q20) Info Type (Q21) Freq. (Q22) 1X /wk or more (3) 1X per month (2) < 1X per month (1) 3 2 1 3 2 1 23. Would you say that you now look for flood in formation more often, less often, or about the same as you did before the June flood? More Often Yes (1) No (0) Less Often 1 0 About the Same 1 0 DidnÂ’t Look Before Flood 1 0 Give Interviewee credibility scale card.
256 Appendix A (Continued) 24. Using this scale, how would you rate the credibility of flood related information from each of the following sources? Not Credible at all Completely Credible DK Family 12 3 4 5 6 7 555 FEMA 12 3 4 5 6 7 555 Friends 12 3 4 5 6 7 555 Town Government 12 3 4 5 6 7 555 County Government 12 3 4 5 6 7 555 Newspapers 12 3 4 5 6 7 555 Real Estate Agent 12 3 4 5 6 7 555 TV News and News Specials 12 3 4 5 6 7 555 NWS 12 3 4 5 6 7 555 NY State Government 12 3 4 5 6 7 555 Insurance Agent 12 3 4 5 6 7 555 Other Mentioned 12 3 4 5 6 7 555 25. On a scale of 1 to 7, how would you rate your overall satisfact ion with available flood information? Completely Dissatisfied Completely Satisfied 1 2 3 4 5 6 7 Give interviewee flood description cards. THANKS. THE NEXT QUESTIONS DE AL WITH THE FLOODS DESCRIBED ON THESE CARDS. THE FIRST IS A 100 YEAR FLOOD, THE SECOND IS A FLOOD WITH A 1% CHANCE OF O CCURRING IN ANY YEAR, AND THE THIRD IS A FLOOD WITH A 26% CHAN CE OF OCCURRING IN 30 YEARS.
257 Appendix A (Continued) 26. In your opinion, which of the three floods de scribed on the cards is the most likely to occur in the next year? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 All the Same 1 0 DonÂ’t Know 1 0 27. Which is least likely to occur? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 All the Same 1 0 DonÂ’t Know 1 0 28. Which of the three floods do you think is the biggest in size? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 All the Same 1 0 DonÂ’t Know 1 0 29. Which do you think is the smallest? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 All the Same 1 0 DonÂ’t Know 1 0
258 Appendix A (Continued) 30. Which of these floods, if any, do you think could happen more than once in a year? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 DonÂ’t Know 1 0 31. Do you think that the size of any of the floods described on the cards could change over time? If yes, Which ones? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 DonÂ’t Know 1 0 32. Which of the floods described on the cards concerns you most? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 All the Same 1 0 33. On a scale of 1 to 7, how concerned would you say you are about this flood? Not Concerned at all Completely Concerned 12 3 4 5 6 7 34. Which of the described floods concerns you the least? 100 Year Flood Yes (1) No (0) 1% Chance Flood 1 0 26% Chance Flood 1 0 All the Same 1 0
259 Appendix A (Continued) 35. On a scale of 1 to 7, how concerned are you about this flood? Not Concerned at all Completely Concerned 12 3 4 5 6 7 36. What concerns you the most about flooding in general? The level of possible flooding Yes (1) No (0) The frequency of flooding of any level 1 0 The combination of flood fre quency and flood level 1 0 Other (Describe) 1 0 37. Which of the following do you think has the primary responsibility for protecting individuals against flood damages? Individuals themselves Yes (1) No (0) Local gov. 1 0 State gov. 1 0 Federal gov. 1 0 Other 1 0 Give interviewee agreement card 38. Using the scale on the card, tell me how much you agree with the following statements. Strongly Disagree Strongly Agree Flooding is one of my top concerns. 12 3 4 5 6 Flood maps accurately show areas of flood risk. 12 3 4 5 6 I have control over what happens to me. 12 3 4 5 6 I am constantly worrying about something. 12 3 4 5 6
260 Appendix A (Continued) WEÂ’RE ALMOST DONE. THE LAST QU ESTIONS ARE USED TO GATHER GENERAL INFORMATION ABOUT THE GROUP OF PEOPLE BEING INTERVIEWED. AGAIN, ALL THE IN FORMATION IS CONFIDENTIAL. 39. Gender Female (1) Male (0) Give interviewee race/ethnicity card 40. Which of those listed on this card be st describes your ra ce or ethnicity? African American Yes (1) No (0) Asian 1 0 Latino 1 0 Native American 1 0 White, non Latino 1 0 Other 1 0 41. How many years have you lived at your current address as of October 1st, 2006? _________ 42. Do you own your home? Yes (1) No (0) Give interviewee schooling card 43. Which of the educational levels list ed on this card best describes the highest level of school or highest degree you have completed ? 12th grade or less 1 High School graduate or equivalent 2 Some college 3 BachelorÂ’s degree 4 Graduate or Prof essional Degree 5
261 Appendix A (Continued) Give interviewee income card 44. Which of the categories on this card best describes your household income in 2005? Under $20,000 1 $20,001 Â– 35,000 2 $35,001 Â– 50,000 3 $50,001 Â– 65,000 4 $65,001 Â– 80,000 5 $80,001 Â– 100,000 6 Over $100,000 7 45. Lastly, what is your age as of your most recent birthday? _______ THAT COMPLETES THE SURVEY. IS THERE ANYTHING ELSE YOU THINK I SHOULD KNOW ABOUT YOUR EX PERIENCE WITH FLOOD RELATED INFORMATION? ________________________________________________________________________ ________________________________________________________________________ ________________________________________________________________________ THANK YOU SO MUCH FOR PARTICIPATING. DonÂ’t forget the survey cards.
262 Appendix B: Focus Group Materials Introductory Statement for Focus Groups Welcome everyone; thank you for being here. My name is Heather Bell. As we discussed on the phone, IÂ’m a graduate student at the Un iversity of South Fl orida and IÂ’m working on my dissertation. Today IÂ’d like to discu ss your experiences with flooding and flood related information and get your opinions on some of the ways that floods are described. Additionally, thereÂ’s a brief background questio nnaire that IÂ’d like you to fill out. The questionnaire should take about 5 minutes and our discussion will last approximately an hour and a half. You may still choose to part icipate or not participate and you may leave at any time. While you will not benefit directly from participating, your full participation will help me finish my dissertation and might also influence the way people talk about flooding and flood policy. Feel free to keep th e gift certificate you received upon arrival. The study itself is not funded by any company or corporation. Please let me know if you have any questions before we begin. The results of this study may be publishe d. However, the published results will not include your name or any other informati on that would personally identify you in any way. Our conversation will be recorded and transcribed, but your privacy will be maintained and research records kept confidenti al to the extent of the law. In our notes, the transcripts, the backgr ound questionnaires, and during an alysis, each of you will be identified by code numbers; no names will be used. Transcripts and any notes we take will be kept locked in a cabinet in my locked office. Tapes will be kept in a separate locked cabinet in another locked office. In ad dition to myself and my advisor, employees of the Department of Health and Human Se rvices, the USF Institutional Review Board, its staff, and any other individuals acting on behalf of USF may insp ect the records from this research project. There are no known risks, but if you experien ce any study related harm, or if you have any other questions or would like more info rmation, please contact Dr. Graham Tobin at the University of South Florida at 813-9744932. He can also be reached through e-mail at email@example.com. If you have questions abou t your rights as a pe rson who is taking part in a research study, you may contact the Division of Res earch Integrity and Compliance of the University of South Fl orida at (813) 974-5638. If you would like to receive an overview of the results when th e research is completed, please e-mail me at firstname.lastname@example.org or send a request to: Heather Bell University of South Florida 4202 E Fowler Ave, NES 107 Tampa, FL 33620 Again, thank you for your participation.
263 Appendix B (Continued) FLOOD FOCUS GROUP BACKGROUND SURVEY These questions are simply used to gather some basic information about the group of participants and get you thinking about the topi c. All the informati on is confidential. Demographic Information 46. What is your gender? 47. Which race or ethnicity do you us ually identify yourself as? African American Asian Latino Native American White, non Latino Other (Please describe) 48. Do you consider yourself one of the heads of the household? Yes No 49. Do you rent or own your home? Rent Own 50. Which of the educational levels listed best describes the highest level of school or highest degree you have completed ? 12th grade or less High School graduate or equivalent Some college BachelorÂ’s degree Graduate or Pr ofessional Degree Female Male
264 Appendix B (Continued) 51. Using this card, please indicate which category best describes your household income in 2003? 52. Lastly, what is your age as of your most recent birthday? _______________ Preliminary Questions 1. Please describe what the phrase Â‘100 year floodÂ’ means to you. 2. In your opinion, which organizations and/or individuals should be responsible for trying to reduce th e impacts of flooding in your area? Which organizations or indivi duals do you think should have primary responsibility? Under $20,000 $20,001 Â– 35,000 $35,001 Â– 50,000 $50,001 Â– 65,000 $65,001 Â– 80,000 $80,001 Â– 100,000 Over $100,000
265 Appendix B (Continued) FLOOD FOCUS GROUP QUESTIONING ROUTE 1. To start, letÂ’s have each of you tell us your first name, how long youÂ’ve lived in the area, and how many floods youÂ’ d experienced prior to 2006. 2. How have you been affected by flooding? 3. What do you think might contribute to potential flooding in this area? 4. Do you currently think of flooding as a threat to you or your community? How much do you worry about it, if at all? 5. What kinds of things, if any, have yo ur families or friends done to protect yourselves against the impacts of fl ooding? What else might people do? 6. What role do you think governments should play in reducing the impacts of flooding? 7. Information was mentioned earlier. Wher e did you get information during this yearÂ’s floods? Where do you go for ge neral information on flooding? Who do you talk to about flooding? 8. What do you think about the information you get? What else would you like to know? 9. On the table are cards with three flood de scriptions that you may or may not have come across before today. Do you think they all are equa lly understandable? 10. How would you define a Â‘100 year floodÂ’? Which of the three floods described do you think would be most likely to occur in the next year? Why? Least likely? Why? 11. Which of the three concerns you th e most? Why? The least? Why? 12. If you were trying to convince a friend that a flood was a real threat to him or her, would you use any of these descript ions? What else would you say? 13. (After summary) Did that correctly describe what we talked about today?
266 Appendix B (Continued) Focus Group Recruitment Postcard Example Dear Resident: I am a graduate student working on my dissertation. I am looking at peopleÂ’s experiences with flooding, flood information and communication, and their interpretations of flood descriptions. I will be conducting small group discussions on these topics at SUNY Binghamton on Jan. 13, 16, 17, 18 and 20. Weekday groups will be held from 6-8 PM; the Saturday groups will run from 10 AM to noon. Snacks will be provided, as will a $20 gift certificate of apprecia tion. If you are interested in participating in one of the discus sions or want more information, please contact me at email@example.com or (925) 395-3175. You do not need to have suffered flood damage to participate. I look forward to hearing from you. Heather Bell, University of South Florida
About the Author Heather Marie Bell grew up in north ern California and earned a B.A. in Philosophy from Grinnell Co llege in 1996. After studying in Iowa and Austria, she moved back to the west coast and joined a small cruise ship company based out of Seattle. Heather spent the next four years wo rking as a naturalist a nd cruise coordinator, migrating north to Alaska during the summer and south to Mexico during winter. When not on rotation, she was able to explore more of North America a nd the world. Heather met her husband abroad and the two settled in Hawaii before ending up in Florida. Heather earned her Masters degree in Geography from the University of South Florida in 2004 and was hired by the Univer sity of Miami in 2007. Her work has been published in Environmental Hazards International Journal of Mass Emergencies and Disasters and Papers of the Applied Geography Conferences