Risk Perception – Issues for Flood Management in Europe

17 Public perception of flood risk and flood risk information is often overlooked when developing 18 flood risk management plans. As scientists and the public at large perceive risk in very different 19 ways, flood risk management strategies are known to have failed in the past due to this 20 disconnect between authorities and the public. This paper uses a novel approach in exploring the 21 role of public perception in developing flood risk communication strategies in Europe. Results 22 are presented of extensive quantitative research of 1375 questionnaire responses from thirteen 23 communities at risk across six European countries. The research forms part of two research 24 projects funded under the 2nd ERA-Net CRUE Funding Initiative: URFlood and FREEMAN. 25 Risk perception is conceptualised as a pillar of social resilience, representing an innovative

definition and defines perception through the relationship of a specific set of risk characteristics: 1 awareness, worry and preparedness. Increasing any or a combination of these raises the 2 perception of risk and in so doing can contribute to enhanced flood resilience. Over time, worry 3 and hence awareness will decrease (Figure 1). 4

Figure 1 5
Awareness is therefore a vital element for a community to effectively adapt to a flood risk and 6 according to Shen (2009) it is diminished when the provision of appropriate information is low 7 or when memories of previous events diminish. Although differences in how best to raise 8 awareness are reported in the literature (Shidawara, 1999;Poortinga et al., 2011), it is generally 9 recognised that levels can be raised through efforts that are focussed on local issues, contain 10 simple solutions to reduce the risk and are repeated on a regular basis (Fischer, 1995;Haggett, 11 1998; Poortinga et al., 2011;Uggla, 2008). Worry in the context of Figure 1 refers to the 12 emotional reactions of individuals exposed to a risk (Sjöberg, 1998). 13 Worry is a necessary risk characteristic that serves as a normative value for awareness. An 14 individual can be aware of a flood risk, however if the individual is not afraid of this risk, he or 15 she will not take any action to prepare. A higher level of worry is more likely to result in a higher 16 level of preparedness (Raaijmakers et al., 2008;Miceli et al., 2008). Reliance on public 17 structural flood protection can however, serve to reduce worry and therefore community 18 preparedness (Terpstra, 2011; Grothmann and Reusswig, 2006). Those who rely excessively on 19 others without taking ownership of the flood risk and responsibility for protecting their own 20 properties are also likely to be less prepared (Lindell and Whitney, 2000; Grothmann and preparedness is not clear and conflicting results are often found in the literature (Siegrist and 23 Gutscher, 2008). Figure 1 shows that worry plays a central role in the relationship between 24 awareness and preparedness; individuals may be aware of a risk but underestimate the 25 consequence of that risk. This paper therefore asks the question "Is worry the link between 26 awareness and preparedness?" 27 3 Methodology 1 Case study research was undertaken in Belgium, Finland, Germany, Ireland, Italy and Scotland 2 forming part of the two-year URFlood (Understanding Uncertainty and Risk in Communicating 3 about Flooding) and FREEMAN (Flood Resilience Enhancement and Management) 2 nd call 4 ERA-Net CRUE projects. The thirteen case studies ( Figure 2) encompassed a range of 5 catchment sizes, flood history and flood risks including fluvial (river flooding), pluvial (rainfall 6 generated flooding) and coastal (inundation of land by sea water) ( Table 1 and Table 2). Some 7 case studies included areas where flood defences have been implemented, facilitating an 8 assessment of residual risk on public perception. The research followed a people-centred 9 approach, collecting quantitative data from the public at risk through self-completion 10 questionnaires. Questionnaires were specifically tailored (in terms of type and number of 11 questions asked) in each case study area to satisfy local needs. However key topics such as 12 awareness and preparedness levels, together with previous flood experiences were assessed using 13 questionnaires that were developed using the key principles of questionnaire design. 14 Questionnaires were therefore of a pre-coded and precise nature and refinement through piloting 15 ensured simple unambiguous language that minimised risk of misunderstanding. Key 16 questionnaire variables are shown in Table 3. 17 The sampling procedure adopted included a mixed method approach of postal, in-person and 18 online surveys, producing a large sample size of 1375 questionnaire returns (Table 1). All case 19 studies only targeted people that were at risk of flooding and included those with (56%) and 20 without (44%) previous flood experience. Individuals living within probabilistic flood envelopes 21 (the 100-year and 200-year envelopes for fluvial and coastal risks respectively) and those 22 residing within designated risk areas (including pluvial risk) were specifically targeted. The 23 SPSS (originally Statistical Package for the Social Sciences) predictive analytics software 24 package (SPSS, 2009) was employed to analyse the data using a set of independent variables 25 including previous flood experience, previous evacuation behaviour, knowledge of evacuation 26 routes, gender, level of education, employment status and age. Statistical significance levels 27 were assessed through chi-square tests, independent samples t-tests, as well as one-way analysis 28 of variance (Anova). The critical significance value (p-value) was set at 0.05 following the usual 29 convention that if there is less than a 5% probability that an observed pattern occurred due only to chance variation in the data, it is accepted that the observed pattern is not due to chance. The 1 results section only reports results that reached statistical significance. 2 Table 1 3   The sample profile as shown in Table 4 was compared with European census data (European 9 Commission, 2011) to investigate whether the sample was representative of a larger population 10 in terms of gender, age, attained education level and employment status. The research sample 11 was found to have a slight over representation of males (54% in sample compared with 49% in 12 census), older respondents (23% of sample over the age of 65 compared with 17% in census) and 13 higher education (45% of the sample reached a third level of education or higher (using the 14 ISCED (2011) definition of tertiary education), which is slightly higher than the census average 15 of 32% for the partner countries). The high retirement rate (24%) reflects the age bias, while the 16 sample is under represented in terms of unemployment (4.5% compared with 8.5% in census). 17 However, a wide range of age groups, education levels and employment statuses were 18 represented, which when combined with the large sample size allowed the sample to be 19 considered sufficiently robust for generalisation of findings. Results were initially analysed in 20 terms of the Raaijmakers et al. (2008) concept of risk perception ( Figure 1) and therefore 21 involved an exploration of risk awareness, preparedness and worry. In addition, and because of 22 its importance in flood risk management strategies, the role of these three indicators of 23 perception in implementing evacuation orders for extreme events was investigated. 24 Table 4 25

Risk perception and awareness 1
Prior to individuals and communities taking measures to improve their resilience to floods, they 2 must be aware that they are indeed at risk of flooding. Awareness is therefore an integral 3 component to effective flood risk management. Regardless of previous flood experience, all 4 participants in this study were at risk. Awareness of this risk amongst respondents was generally 5 good, with approximately 80% of respondents accepting they were at risk. Awareness was 6 subsequently found to be strongly correlated to previous flood experience (chi-square = 6.7, df = 7 1, p = 0.006) (Figure 3 actions that should be taken (Wilson, 1990) and frequent inundations ensure that the perception 12 of risk and associated resilience levels remain high. Conversely, long periods without floods will 13 serve to diminish awareness (Burn, 1999  consequences can potentially generate greater concern amongst individuals to empower them to 20 seek more information for use in the planning of effective mitigation strategies. This problem 21 can also be addressed by the continued provision of information that highlights the existence of a 22 risk and prescribes appropriate preventive and protective measures. Drawing attention to local 23 flood management plans is therefore important in this regard. 24 From a flood risk management perspective, those that are at risk but don't consider themselves to 25 be (approximately 20% of respondents in this study), are potentially of more concern. The lack 26 of awareness of flood risk reported in this study could be explained, perhaps even justified, if 27 those (n = 238) that were unaware had not been previously flooded. However, results indicate 28 that a significant number of respondents (n = 118) were unaware of any risk despite having direct 29 experience of previous floods. This issue was further investigated in the Irish case study areas 30 where questionnaire returns were referenced to geo-spatial data. It was found that respondents who fell into this category resided in areas adjacent to structural flood protection works that had 1 been constructed since the last reported flood. Results suggest therefore that residents in the 2 vicinity of structural flood defences are instilled with a false sense of security and believe 3 themselves to be immune from future flood risks. As one respondent quoted "With the flood 4 defences now in operation, I could not imagine being flooded again". Similar results of low 5 public risk awareness in structurally protected areas have been shown in the literature (Ludy and  6 Kondolf, 2012; Terpstra, 2010; Lassa, 2011). However, it is unclear whether residents in these 7 studies had been flooded previously. Further to this, it has been shown that government policy 8 also overlooks residual risk, with lands behind structural defences not considered at-risk areas, identified a disconnect between the language used by the engineering community and that 17 understood by the public at large (Bradford et al., 2011). This failure to understand key terms is 18 likely to be a contributing factor to the perceived immunity to risk of those protected by formal 19 flood protection structures. As such structures are designed to provide a specified level of 20 protection quantified in terms of recurrence intervals or return periods, public understanding of 21 these terms will lead to recognition that these structures will be exceeded for events greater than 22 the design capacity. A lack of awareness of and a 'switching off' to a prevailing residual risk will 23 adversely affect the ability of residents to respond to a flood, should it arise. This presents 24 additional challenges to flood managers in these areas that need to be overcome by continued 25 education and communication of these remaining risks. 26 Other explanations for the poor awareness amongst those who were previously flooded may 27 include a denial of risk, a reluctance to accept risk or a lack of concern regarding the risk. These 28 factors can be attributed to unrealistic optimism, psychological attachment to the home or 29 economic interests in not wishing to devalue the home by accepting and acknowledging risk These factors represent barriers to community resilience and need to be specifically addressed in 1 flood risk management communication plans. A role exists for the responsible agencies together 2 with local and national media sources in this regard. Examples could include discussing risk in 3 the media as well as at the local, regional and national level to normalise the concept of flood 4 risk and help reduce risk denial. Publishing risk maps will identify at-risk areas and allow 5 discussion to take place within the community, facilitating further acceptance of risk. Levels of 6 risk awareness were shown to be unrelated to the demographic factors assessed in this study, 7 indicating that efforts to raise awareness need to remain focussed on all social groupings. 8 Furthermore, no obvious differences between those of different nationality or from different case 9 studies were observed. 10 11

Risk perception and preparedness 12
High levels of preparedness will contribute to individual and community resilience and facilitate 13 an improved response to a flood, thereby reducing adverse consequences (Paton et al., 2006). 14 Respondents in Belgium, Finland, Germany, Ireland and Italy (n = 1231) were requested to self-15 assess their preparedness levels, with just 34% of these respondents feeling prepared for future 16 floods. While it may be expected that an awareness of residing in a flood risk area would 17 promote higher levels of preparedness, an independent samples t-test found this not to be the 18 case (t = 0.41, df = 1100, p = 0.68). This suggests that increasing awareness on its own will not 19 necessarily result in increased levels of preparedness. Although risk awareness had no direct 20 influence on flood preparation, a significant relationship was found between previous flood 21 experience and preparedness levels (t = 6.6, df = 1182, p <0.001) (Figure 3(c)). This higher level 22 of preparedness may in part be explained by the desire of individuals to avoid the often 23 unanticipated negative emotions of insecurity, fear and helplessness experienced during a flood 24 event (Terpstra, 2011). The positive impact of this experience may however, be relatively short 25 lived (Weinstein, 1989;Mulilis et al., 1990). In addition, it is common for individuals to assume 26 that future flood events will be of the same magnitude as those previously experienced and may 27 therefore only take mitigating actions that would have been appropriate for these prior events 28 (Burn, 1999;Weinstein, 1989; Green et al., 1991; Howe, 2011). Furthermore, those that 29 experience extreme events where mitigation measures taken were of little use can be overcome with a sense of helplessness that can lead some individuals to conclude that damage will result 1 regardless of actions taken to protect property (Slovic, 2000). Findings from this study indicate 2 therefore that a strong case exists for disseminating preparedness information specifically to 3 those in vulnerable communities that have no direct experience of floods or those whose 4 experience is based on events from some time in the past. The importance of being prepared and 5 methods of improving preparedness levels needs also to be continuously highlighted in focussed 6 flood risk communication strategies. 7 Gender differences in perceived preparedness levels were observed within the sample with males 8 reporting higher levels than their female counterparts (t = -4.9, df = 1154, p <0.001). However, 9 the reasons for this are somewhat unclear and the finding may reflect higher confidence levels of

Risk perception and worry 18
Worry, fear or concern is reported as being either a direct or indirect motivator for promoting 19 preventive behaviour amongst those at risk (Weinstein, 1989;Grothmann and Reusswig, 2006;20 Tanner et al., 1991). Self-assessed levels of worry were therefore investigated in the Belgian, 21 Finnish, German and Italian case studies. The majority of respondents (65% of 795) stated that 22 they frequently worry about being flooded. This increased significantly for that fraction of the 23 sample that had either direct experience of flooding (Figure 3(b)) or were aware of their 24 exposure to a flood risk (t = 7.9, df = 729, p <0.001 and t = 7.2, df = 227, p <0.001 respectively). 25 Those with lower levels of education were also found to worry more about floods (1-way Anova, 26 F = 5.9, p <0.003) supporting research by Sjöberg (1998). Given that a relationship exists 27 between education and income levels (Hansen, 1970), it may be inferred that those in higher 28 income brackets worry less about the consequences of floods. While this social grouping may 29 experience greater financial loss than those less well off, better insurance cover together with the 1 resources to more easily repair damaged properties and replace lost belongings ultimately 2 reduces the impacts of floods. 3 Results contrast somewhat with other studies which found no relationship between worry and 4 risk awareness or education (Pagneux et al., 2011;Poortinga et al., 2011). Moreover, women are 5 often found to worry more about risks in the literature (Poortinga et al., 2011;Sjöberg, 1998), 6 however a strong relationship was not found between worry and gender in this study (t = 1.8, df 7 = 750, p = 0.07). Furthermore, no significant relationship was found between worry and 8 preparedness levels within this study (1-way Anova, F = 1.4, p = 0.234) suggesting that worry on 9 its own is not enough to motivate preventive behaviour for flood risks. Fear arousal is often 10 advocated in order to increase risk perception (Kievik and Gutteling, 2010;Witte, 1992). assessed. Higher levels of risk awareness were shown to be correlated to those who chose to 29 evacuate properties (chi-square = 5.05, df = 1, p = 0.015) but this awareness was not shown to 1 translate to knowledge of evacuation routes. The findings support results from Section 4.2 that 2 awareness alone is not sufficient to cause those at risk to take positive action. Furthermore, 3 independent samples t-tests indicated that no significant relationship was found between levels of 4 worry and previous evacuations. A relationship was however, found between worry and 5 knowledge of evacuation routes (t = 2.3, df = 209, p = 0.023) suggesting perhaps that while 6 worry may not lead directly to increased preparedness levels, it may lead to an increased 7 understanding of the local impacts of floods. Those that evacuated properties in previous floods 8 were shown to have higher preparedness levels (t = 2.8, df = 364, p = 0.006) indicating that the 9 upheaval and emotional stress expended during evacuations may instil a willingness amongst 10 some to implement preparatory measures. The inclusion of personal accounts from flood victims 11 in ongoing flood communications can serve to highlight adverse impacts of floods with a view to 12 reinforcing the need to take mitigating measures. Knowledge of evacuation routes, associated 13 with higher levels of worry, also led to respondents feeling more prepared for floods (t = 7.6, df 14 = 181, p <0.001). This shows the need for locally tailored information that identifies safe routes 15 and appropriate actions in times of flood. 16 17

4.5
General discussion 18 The findings of this study advance the understanding of how risk awareness, worry and 19 preparedness contribute to the concept of risk perception. Figure 1 shows awareness 20 contributing to worry and worry contributing to preparedness levels with the effect, that 21 elevating any of the three indicators, raises perception and resilience. However, a similar 22 framework (Figure 4) for the current study shows no correlation between worry or awareness and 23 preparedness levels. Worry is therefore found not to be the central characteristic in the link 24 between awareness and preparedness. The findings do however reinforce the significance of 25 direct flood experience in raising these three elements of perception. Findings show that while 26 risk awareness is linked to concern of this risk, awareness and worry are not correlated to high 27 levels of preparedness which would serve to enhance resilience. 28

Figure 4 29
Given that significant numbers without direct flood experience continue to reside in areas 1 exposed to flood risks, consideration from flood risk managers must be given to methods of how 2 best to provide flood related information that serves to raise preparedness levels in the absence of 3 the experiential learning developed when dealing with flood events. Based on the findings of the 4 Irish and Scottish case studies in the URFlood project, the challenge is likely to be compounded 5 in some communities where there is a belief that emergency responders and flood risk agencies 6 are solely responsible for implementing preventative measures to protect properties and that self-7 protection is less important (Bradford et al., 2011). Recommendations from this study can be summarised as follows: 26 2. Providing understandable statements on risk will lead to recognition that structural 1 protection measures will be exceeded for events greater than the design capacity, thus 2 reducing the issue of residual risk 3 3. Preparedness information needs to be tailored to those who are at risk but have no direct 4 experience of floods or whose experience is based on events from some time in the past 5 4. Including personal accounts from flood victims in ongoing communications can serve to 6 highlight adverse impacts of floods, reinforcing the need to take alleviation measures 7 5. Locally tailored information that identifies safe routes and appropriate actions in times of 8 flood should be provided 9 6. Providing specific information on easily implementable mitigation measures will increase 10 confidence, especially in women, in personal ability to protect property 11 7. As worry does not increase preparedness, communication strategies should not aim to 12 evoke fear in vulnerable communities 13 The practical findings in this study gathered through an innovative approach in utilising a social 14 resilience model will facilitate flood managers in developing management plans that allow for 15 the complexity in public perceptions of this risk. The paper will be useful in preparing risk 16 communication strategies that will increase flood resilience in at-risk communities.  Table 1. Characteristics of case studies undertaken 2 Table 2. Characteristics of previous flood events by case study area 3 Table 3. Key variables investigated in questionnaires in Belgium (B), Finland (F), Germany (G), 4 Ireland (IR), Italy (IT) and Scotland (S) 5 Table 4. Gender, attained education levels, employment status and age of respondents 6