Rethinking flood risk communication

Flooding is a serious hazard across Europe, with over 200 major floods documented in the last two decades. Over this period, flood management has evolved, with a greater responsibility now placed on at-risk communities to understand their risk and take protective action to develop flood resilience. Consequently, communicating flood risk has become an increasingly central part of developing flood resilience. However, research suggests that current risk communications have not resulted in the intended increase in awareness, or behavioural change. This paper explores how current risk communications are used by those at risk, what information users desire and how best this should be presented. We explore these questions through a multi-method participatory experiment, working together with a competency group of local participants in the town of Corbridge, Northumberland, the UK. Our research demonstrates that current risk communications fail to meet user needs for information in the period before a flood event, leaving users unsure of what will happen, or how best to respond. We show that participants want information on when and how a flooding may occur (flood dynamics), so that they can understand their risk and feel in control of their decisions on how to respond. We also present four prototypes which translate these information needs into new approaches to communicating flood risk. Developed by the research participants, these proposals meet their information needs, increase their flood literacy and develop their response capacity. The findings of the research have implications for how we design and develop future flood communications, but also for how we envisage the role of flood communications in developing resilience at a community level.

at risk, the research presented in this paper looks to circumvent this framing by allowing research participants to determine what information is important to them for understanding their risk and increasing their resilience.
2 Current approaches to communications in flood risk management Across Europe, the 2007 EUFD established common standards for the preparation of flood hazard and flood risk maps (EXCIMAP 2007a). The UK provides a good example of these products, with the Environment Agency (EA) publishing a well-developed suite of different mapping types available online (de Moel et al. 2009), alongside an array of supporting communications (Table 1). These include communication of real-time river levels, and flood alerts and warnings intended to highlight the short-term potential for flooding.
The EA's prime purpose for flood risk communications is to encourage participation in local FRM and develop community resilience (Environment Agency 2011). EA research on resilience has previously focused on generating trusted, long-term relationships with atrisk communities (Twigger-Ross et al. 2011, 2014 (Table 1). As a result, communications have traditionally been supported by a network of local flood groups and wardens, tasked with working alongside the EA to prepare local communities for flooding (Gilissen et al. 2016). However, recent high-profile floods have caused a shift of focus towards infrastructure and property-based resilience programmes (McBain et al. 2010;Chatterton et al. 2010;Environment Agency 2011). As a result, community-based resilience has become somewhat of a secondary objective and the potential for risk communications as an enabler of resilience has taken on a much greater level of importance (Environment Agency 2010). However, the existing research suggests that current communications are having limited impact on driving risk awareness or resilient behaviours. O'Sullivan et al. (2012) examined the impact of flood risk communications across Europe and identified low levels of information penetration and personal preparedness, often accompanied by a high level of These online maps indicate areas of potential flood hazard and differentiate high-, medium-and low-risk categories. Intended to raise awareness of the risk of flooding of those living in at-risk areas Common to the majority of countries in the EU (EXCIMAP 2007a, b;de Moel et al. 2009) Real-time water level information Hydrographs of 'real-time' river levels monitored at river gauging stations provided online. During flooding conditions these records are updated at 15-min intervals. These hydrographs also display the level over which flooding can be expected and the highest level ever recorded. Intended to allow local people to monitor local river levels and decide when to take action in response to potential flooding Flood warnings (Flood Information Service) A flood warning system is also implemented across England (Fielding et al. 2007). Three alert levels are provided, the intention being that those at risk should begin to monitor local river levels at the Flood Alert Stage and begin to implement flood-resilient actions at a Flood Warning Stage. Intended to instruct those at risk when to take action in response to a potential flood Nat Hazards (2018) 92: 1665-1686 1667 distrust in communications and management organisations. In the UK, a 2016 EA poll indicated that only 45% of people living in at-risk areas appreciate their risk and only 7% identify any risk to their own property (Environment, Food and Rural Affairs Select Committee 2016). Similarly, independent polling by the 'Know Your Flood Risk' campaign (Davies 2015) reported that 31% of at risk households surveyed had no flood plan and would not know what to do in the event of flooding.

Risk communication approaches and the adoption of resilient behaviours
Research has therefore identified that the existing UK model of flood risk communication is not functioning as intended, with communications failing to meet user needs or match their experiential knowledge (Environment Agency 2010; Meyer et al. 2012;Fisher 2015). It has also been argued that by centralising and professionalising the production of risk information, local communities lose their ability to properly understand their local risk situation (Lane 2012;Bubeck et al. 2012). The outcome is that both the practice of communicating risk information and how information receivers interpret information may not actually be aligned with the stated purpose of flood risk communications in the UK. In this section we explore the fundamental research that underpins risk communication and examine why this might be the case. Callon (1999) and Demeritt and Norbert (2014) have both proposed models for how risk is communicated, considering the direction of communication, the roles of the communicator and the receiver, and the purpose of the communication (Table 2).
Flood risk communications have a joint purpose of both transmitting information and also altering behaviour (Hagemeier-Klose and Wagner 2009) and can therefore be seen as a hybrid of the risk message model (RMM) or the public education model (PEM), and the risk information model (RIM). Research driving RIM-focused communications has explored a wide range of potential factors which influence the translation of risk information into behaviours: examples include previous experiences of a threat (Fielding et al. 2007;Hopkins and Warburton 2015); cultural, geographical, and socio-economic factors (Burningham et al. 2008;Bubeck et al. 2012); reliance on public flood protection (Terpstra and Gutteling 2008); trust/distrust in communications from a management authority (Terpstra 2011;Wachinger et al. 2013); or a need to protect an individual's sense of personal security against high levels of future uncertainty (Harries 2008;Willis et al. 2011).
An alternative approach to examining individual variables is proposed by Rogers (1975), who presents the protection motivation theory (PMT) model (Fig. 1). PMT explains and provides an overarching framework for the interplay between the disparate variables which may contribute to triggering behavioural responses from risk information. Rogers argues that individuals make their decision by appraising the severity and likelihood of their exposure (the threat appraisal) against the potential efficacy of potential protective behaviours (the coping appraisal), with their protection motivation representing the intervening stimulus which determines their actions. Grothmann and Reusswigg (2006) and Bubeck et al. (2012) build upon Roger's work by expanding the sub-components of the threat and coping appraisals (Fig. 1), as well as identifying the potential for non-protective responses such as denial or wishful thinking, in situations where threat and/or coping appraisals are negative. This concept is supported 1668Nat Hazards (2018) 92:1665-1686 by research on 'learned helplessness' (Paton and Johnston 2001), where individuals see disaster events as uncontrollable and therefore assume that their impacts are in turn uncontrollable, triggering feelings of despair or helplessness (Paton and Johnston 2006). PMT demonstrates the complex, contested, and highly personal nature of the linkage between communication and the adoption of protective behaviours. Comparison against the models of communication reveals the likely limitations of current communication approaches based on the RIM or PEM. These approaches, which assume a rational response from the receiver, are unlikely to address the complex nature of the threat and coping appraisals. 4 Using participatory approaches to developing new ways to communicate flood risk We suggest that participatory working (Kindon et al. 2007) offers an opportunity to rethink how information can be communicated to those at risk by positioning people at the heart of flood risk communications. Participatory working re-imagines the traditional roles of experts and lay people (Bucchi and Neresini 2008;Landström et al. 2011;Lane et al. 2011) and considers circulation of different forms of expertise (Whitman et al. 2015), with participants working together as equals to co-produce shared knowledge and outputs (Mees Table 2 A comparison of the defining characteristics of risk communications models proposed by Callon (1999)  . Participatory working approaches have been applied to a variety of environmental problems, including the co-production of options for managing local flood risk , the breaking down of borders between different organisations, professionals, and lay people involved in catchment-scale land management to manage floods (Bracken et al. 2016), and developing end-user specific research outputs regarding agricultural pollution (Whitman et al. 2015). To date, however, participatory practices have not been applied to flood risk communications, with recent research concluding only that participation was a useful approach for raising awareness or communicating flood risk complexity (Environment Agency 2012), or as a way of providing limited feedback on current communication approaches (Fisher 2015). These limited approaches to participation fail to exploit the potential of participatory working to open up the debate on what risk information is important and why. Here therefore we look to expand the participatory approaches demonstrated by previous studies into exploring the efficacy of current flood communications and, working together with a flood group of flood-affected locals, to coproduce alternative communications better suited to driving resilient behaviours.

The Corbridge study area
Corbridge (Fig. 2) has a long history of flooding; approximately 70 properties in Station Road and The Stanners are situated on the floodplain and are vulnerable to flooding. River level records date back to the 1700s (Archer et al. 2007a), and the area has a long history of

The research approach
In the summer of 2016, we undertook research to explore local knowledge about flooding in the Tyne Valley based on working together with local people to develop new approaches to communicating risk. Our aim was to blend academic research expertise with the experiences of Corbridge residents to re-imagine what flood risk information could be communicated and how it might be best presented. Figure 3 shows the multi-methods participatory approach developed.

Understanding local knowledge and flood experience: workshops with the Corbridge Flood Action Group
In Phase 1 we conducted several group mapping and discussion workshops with members of the local Corbridge Flood Action Group (CFAG). The purpose of these meetings was to assess local knowledge and experiences of flood risk. Using a grounded theory approach, following Charmaz (2011), the material produced by these workshops, the maps, the researchers' notes and the group discussions were integrated to identify key themes arising from the local experience of flooding. Using this approach we developed an understanding of the level of knowledge about, and engagement with, the flood risk problem, as well as developing a trusting relationship between the researchers and the CFAG.

Adopting a participatory approach to developing new flood risk communications: the Corbridge Flood Research Group
The relationship developed between the researchers and the CFAG during Phase 1 was instrumental in developing the Corbridge Flood Research Group (CFRG), which was developed through Phase 2 of the research. The CFRG was a group loosely modelled on the Environmental Competency Group used by Lane et al. (2011). Similar to Lane et al. the CFRG was set apart from traditional focus or consultation groups by its focus on the practice of knowledge creation as a collaborative process and the integration of local 'non-experts' into a practice of flood management usually carried out far removed from the local scale. The CFRG consisted of six, self-selected members of the wider CFAG who had personal experiences of (five members), or interest in (one member), flooding at Corbridge. One of the researchers (Rollason) also took an active role in the group as a member, as opposed to a more traditional role as facilitator or group leader. Local members of the group contributed their experiential knowledge of flooding and flood communication, whilst Rollason (as an academic specialist and former professional flood manager in industry) brought expert technical knowledge and experience. By blending these two perspectives, the group was able to consider both the CFRG's communication desires and the practicalities of what could be achieved.
CFRG meetings were framed specifically to explore flood risk communications. The group met three times; only the theme of the first, 'how flood risk is currently communicated?' was predetermined. Subsequent meetings were driven by the group discussions and were predominantly unstructured, with participants determining what should be discussed and how. Meetings were audio-recorded, and field notes taken. After each meeting key discussion points were summarised, notes circulated to the group; all members thus Analysis of the material was undertaken throughout the process by adopting a flexible, mixed-method approach, situated within the principles of grounded theory (Knigge and Cope 2006; Charmaz 2011), for identifying and linking key areas of discussion. The discussions held with the group during CFRG1 and CFRG2 allowed Rollason to prepare a series of prototype interfaces for communication. The technical skills employed involved flood risk mapping using GIS software and running two-dimensional flood models to capture and present information. The prototypes were presented for group deliberation at CFRG3, with the group jointly choosing four concepts and then working together to produce a final, shared version which represented the agreed outputs from the group.

Testing the prototypes outside the CFRG
In the final stage of the research the prototypes were presented to a larger focus group consisting of eight members of CFAG (new to the research), Rollason and one original member of the CFRG. The design and purpose of the concepts were outlined, and the focus group discussed what they thought of the ideas, how they might be used, and any alternative ideas. The key aspects of this discussion were recorded during the focus group. No further amendment of the concepts was deemed necessary following discussions.
5 Current flood risk communications: Do current approaches meet users' needs?

Understanding local knowledge and experiences of flooding
The initial CFAG workshops and CFRG1 revealed that local participants had a wealth of experiential knowledge about flooding. Many also had an understanding of wider catchment processes developed through hobbies, such as fishing, or work. Despite this, few participants had expected the flooding to occur despite the receipt of an official flood warning (see Sect. 5.2), with many assuming that the recently completed flood defences would protect them, as one participant stated: To be honest I didn't really believe it, because we had such faith in the flood defences that I actually didn't think we'd flood (Participant GW44) Based on this commonly held belief, several participants made the decision not to evacuate, even when contacted by the emergency services (Oliver 2016). That participants were surprised by the flooding, and unsure of how to react to it, demonstrates that flood communications had not developed the resilience of the Corbridge community to respond to flooding after the 2005 flood. These findings highlighted the need to examine in more detail how current flood risk communications were used by participants and how they might be redesigned to better develop resilience.

Reflections on current methods of flood risk communication
The CFRG members were familiar with the principle communications provided by the EA and several had used them before the 2015 flood. Table 3 summarises the group's attitudes towards the current flood communications, and these are expanded upon below.
Nat Hazards (2018) 92: 1665-1686 1673 The CFRG felt that current approaches did not provide them with enough information to understand their flood risk or make an informed decision of what to do when they received a flood warning. In the case of the passive communications, the simple presentation of a flood risk extent, lacking any information on how floods occur, provided them with no information that they could actually use to understand what the stated flood risk meant. Participants stated they only used these maps for buying their homes or negotiating insurance; other than this participants thought that the maps told them nothing that they did not know already: For me, I know I'm in a high risk area so all it [the flood map] would tell me is what I know already (Participant GW44) Some participants also expressed a lack of trust in how the maps had been produced, as one participant explained: Originally, when they did the first online extreme flood map they drew the lines through the centre of the church […], and I said ''if it's getting to that level, it's coming down my chimney'' (Participant AK97) The church at Corbridge sits approximately 16 m above the floodplain. The group member still linked this experience and his distrust of those original maps to the current flood maps which appear superficially the same. The advancements in modelling and data since the production of the early maps are not evident in the way the maps are communicated, and information on how they are actually produced is not publicly available.
Participants were much more engaged with the active communications, particularly the online availability of real-time river levels. Several CFRG members noted that they watched gauges upstream of Corbridge to try and judge how river levels might change at Corbridge in the near future. However, all participants expressed frustration with the lack of forecast river levels, which did not allow them to judge when flooding might occur, or how severe flooding of their homes might be in comparison with past events. This was a particular problem when participants received flood alerts, preliminary warnings that flooding might occur in the near future. These alerts, issued some time before formal flood warnings, are intended to prompt people to begin monitoring local river levels and prepare to take protective action. However, participants felt that the lack of forecast information left them unable to judge what to do and when. Fundamentally, participants felt that the information they were being provided currently told them when to act, but did not provide them with enough information to judge what it was feasible for them to do, or to what extent they should take action. As one participant noted regarding the 2015 floods: When we put things up high, not thinking that when the river comes over the water was going to be so high it would upend all those things, so everything I put up high to save we lost (Participant GW44)

What information do users want in flood communications?
CFRG2 focused on the information that people actually wanted from flood communications to allow them to understand their risk and take action, setting aside for the moment the practicalities of whether or not such information could be provided. The discussions reflected their initial criticisms of existing communications, focusing particularly on understanding the severity of the risk, and therefore what degree of action they could and should take (Table 3). Ultimately, group members wanted flood levels to be forecast, and a specific linkage between what these flood levels meant for their properties and what they could do in response, for example how high they needed to lift valuables: What you need is the starkest information, […]  These discussions encompassed both passive and active communications, with participants generally agreeing that active communications, such as the river level graphs, should be more specifically linked to the passive communications, which could provide more indepth and detailed information on property-level impacts.
Some group members were concerned that providing more complex information would be confusing and potentially undermine responses to flooding. As a result, the group discussed how it was necessary to communicate flooding dynamics, for example how, when, and where flooding might occur, in order to be able to effectively interpret local flood risk. Participants referred to this type of information as contextual information, examples of which included where and when flood defences might be over-topped and how flood water might flow across the floodplain in order to flood their properties. This potentially reflects the relatively complex dynamics of the 2015 flood, where the principle areas of defence over-topping were out of sight of participants properties, and therefore flooding occurred from an unexpected direction.
6 Working together on new approaches to communicating flood risk Between CFRG2 and CFRG3 a series of draft prototypes of alternative passive and active flood risk communications were developed. Six prototypes were originally produced, exploring different types of information that could be communicated and different ways of Nat Hazards (2018) 92: 1665-1686 1675 communicating it (Table 4). Although the CFRG2 discussions had considered participants' information aspirations without considering the practicalities of implementing them, the group felt that it was important, in producing the prototypes, to consider how these ideas might be implemented in practice. Thus, where possible, proposals draw inspiration from existing examples of flood risk communications in other countries, proposed methods drawn from the literature, or examples of communications drawn from other fields ( Table 4). The prototypes were the focus of CFRG3. From the suite of initial concepts developed, the group considered four to be particularly useful (Fig. 4). These four were considered by the local participants to give them the information they felt they needed to understand the risk of flooding, but also to make informed decisions about what action to take, and when, for future floods. These four prototypes were further developed by the group during and after CFRG3, and those shown in Fig. 4 represent the final, agreed outputs from the group. The four prototypes adopted by the group reflect the CFRG's two core desires: 1. to be able to take responsibility for effectively monitoring their flood risk and judge, through forecast information, how significant any flooding might be (Fig. 4a, b). 2. to have a detailed understanding of how flooding might occur based on a knowledge of past flooding dynamics (Fig. 4c) and to be able to link forecast flooding information with the potential impacts on their own properties, allowing them to judge what action they could take in response. Figure 4a shows the catchment-wide overview of the river gauging station status, enabling users to quickly assess how the catchment is responding to rainfall. This prototype map is linked specifically to individual gauge records allowing users to select and explore specific sites in more detail. Inspired by the online angling tool 'Fishpal' (Fishpal.com) used by one of the CFRG participants, the group members felt this tool allowed them to easily monitor catchment-scale river response, using their knowledge of how rainfall in different areas of the catchment translated into flood risk at Corbridge. Figure 4b outlines a prototype gauge dashboard which allows users to 'pin' multiple gauges of their choice into a single place for rapid review of how river levels upstream are responding to rainfall. This prototype answers group members annoyance with only being able to view one gauge at a time using the current system. This prototype also reflects different options for how gauged levels should be displayed, which were also discussed by the CFRG: (1) current approach adopted by the Environment Agency; (2) current approach with an overlay comparing current levels with a historical hydrograph; (3) current approach including predicted future water levels based on the method proposed by Leedal et al. (2012); or (4) current approach with both (2) and (3). CFRG participants felt that forecast water levels (as in 3) allowed them to plan protective actions in advance of flooding occurring by anticipating when they would need to take certain actions, whilst historical comparisons (such as that shown in 2) allowed them to contextualise the significance of predictions and therefore judge what level of protective action was necessary. Figure 4c presents a detailed assessment of historical flood dynamics (in this case the December 2015 flood event). In this prototype users would be able to select different elements to be provided with a detailed account of how flooding occurred, and what action might have been taken in response. The hydrograph allows users to identify water levels at which different flooding mechanisms begin to operate. CFRG members felt this map developed their understanding of how flooding occurred and when, allowing them to understand the significance of local gauged river levels. Figure 4d shows the simulated extent and depth of potential flooding based on a userselected water level for the Corbridge gauge. This prototype was inspired by the United States Geological Survey 'Flood Inundation Map Library' (United States Geological Survey 2016). Current and predicted water levels at Corbridge are displayed on the gauge display to allow users to link current and predicted water levels with their evaluation of potential impacts. CFRG participants felt this simple linkage between river levels and potential floodplain impacts was important for correctly interpreting what forecast river levels might mean and for demonstrating what degree of protective action was needed in different situations. The four updated prototypes were presented to the wider CFAG at a group meeting and also at a smaller focus group. The prototypes were well received by the focus group (Table 5), with the underlying themes of understanding flood dynamics, flood impacts and future prediction being reflected in the discussions. All participants saw the potential for the active communications to enable them to take action to reduce the impact of future floods. The prototype in Fig. 4c provoked a different response to that of the CFRG. The focus group members thought that this map was not for them to use in preparing for flooding, but instead was as a tool for them to use to engage more effectively with the EA about ongoing FRM on a more even information footing: I think that information is important for us to see, so that we can have intelligent conversations with the Environment Agency (focus group participant) Instead, to prepare for a flood in the near future, the focus group participants preferred the simple water level model shown in Fig. 4d.
This discussion highlights the complex interactions between individual users and the different approaches to communicating flood risk and the difficulties of presenting only a small variety of information in order to represent a complex and dynamic threat such as flooding. • Very useful for understanding flood impacts and allowing users to link gauged information with potential flood depths • More useful than the historical pathways and timings idea

Discussion
In this section we bring together the experience of the CFRG experiment with theories of risk communication. We argue that participation such as that demonstrated by the CFRG must play a role in developing future flood communications, especially in light of the shift from flood defence to flood management and the resulting distribution of FRM responsibilities onto those at risk (Butler and Pidgeon 2011). Such involvement will enable responsible agencies to better communicate flood risk in new ways, empowering those at risk to apply their local knowledge and experience to improve their resilience in the face of flood events.

Implications for current flood risk communications
The research undertaken has shown that there are severe limitations to current flood risk communication approaches which prioritise simple threat messages. The PMT model (Rogers 1975) can be used to analyse the responses of the CFRG to the existing flood communications and their desire for alternative approaches, focusing particularly on the ideas of the threat and coping appraisal (Grothmann and Reusswig 2006). The CFRG saw no useful information in the existing passive maps, which suggests that this approach does not support the development of threat appraisal. The lack of information on flood dynamics also provides no basis on which users can judge for themselves how communicated risk information might translate into an impact on their own property. In this context, threat appraisal is reliant on previous experiences, whether personal or vicarious. In the Corbridge context this wholly underestimated the threat, resulting in a non-protective response based on 'wishful thinking' regarding the recently completed flood defence works. Hopkins and Warburton (2015) refer to this paradox as the 'prison of experience', in which infrequent or unrepresentative events imprint themselves into subjective knowledge as representative experiences to be drawn on in the future. CFRG participants' desire for detailed information on past local flooding characteristics or the simple flood depth simulator can be seen as an attempt to place their experiences in a wider context, breaking out of the prison of experience and establishing a more holistic understanding.
Both the passive and active communications assessments also suggest a failure of current approaches to establish a meaningful coping appraisal, particularly in relation to the judgement of how much time participants in this study had to react and what degree of action they should, or could, take. Several participants expressed surprise at the prototype flood map showing flood dynamics, which highlighted over-topping of upstream defences approximately 4 h prior to property flooding occurring. These participants had no understanding from the current flood maps (which do not show information such as areas of potential over-topping) that flooding might either be inevitable or occurring, or that they potentially had several hours in which to prepare or act. Neither were participants able to accurately judge the degree to which they should prepare based on the live gauged information, since this online information does not currently offer information on predicted water levels. This led to negative coping appraisals and the adoption of non-protective behaviours, where participants either ignored what might be happening or took ineffective action. In this context, the Group's desire to see whole-catchment-scale information, which incorporates future predictions, can be seen as building not just their personal appraisal of the threat, but also their coping appraisal. Understanding the threat allows them to feel in control of their own flood risk situation and to make their own decisions, rather than Nat Hazards (2018) 92: 1665-1686 1681 reacting blindly to flood warnings; a situation that participants said left them very stressed and uncertain.

Future flood risk communications: participation as a vehicle for developing resilience through flood literacy
Viewed in the context of the PMT, current flood risk communications could therefore be judged to be counterproductive; they attempt to provoke a heightened perception of flood risk, without providing the information required by users to establish strong, positive threat and coping appraisals. Without developing coping appraisals, users adopt the kinds of nonprotective behaviours proposed by Bubeck et al. (2012), ignoring, rejecting or misinterpreting official risk information to make them feel more secure in the face of extreme uncertainty (Harries 2008). These behaviours reduce community resilience by increasing the shock of events when they occur unexpectedly or do not match individuals previous experiences; increase individual hazard through refusals to evacuate; or foster learned helplessness when believed protective behaviours fail or have no effect.
To encourage positive threat and coping appraisals future flood communications need to move away from the simplistic flood threat messages that are currently cascaded to people at risk. Instead, and as the four prototypes created here demonstrate, communications should provide more detailed, holistic hazard information. This type of information, rather than relying on raising risk perception alone, seeks to develop a local 'flood literacy' by fostering local knowledge about flooding. Flood literacy repositions those at risk as active agents in managing local flood risk, able to make their own judgements and decisions on risk and protective behaviour, rather relying on expert knowledge (Willis et al. 2011). By empowering people in this way, flood literacy develops local resilience in a way in which simple, threat-based communications cannot: it provides at-risk individuals and communities with the information necessary to (1) assess their personal level of risk and how they might be affected, (2) determine when a flood might be about to occur and how it might affect them, and (3) determine appropriate actions by which they might mitigate potential flood impacts.
To encourage effective flood literacy through improved flood risk communications, there is a need to re-establish resilience as a process grounded in relationships, social learning and dialogue (Twigger-Ross et al. 2011, 2014Benson et al. 2016), rather than 'hard' infrastructure or property (McBain et al. 2010). Participatory approaches offer a potential avenue through which the reinvigoration of resilience in this fashion might occur. The results of our research demonstrate the importance of working together with end-users in developing new solutions to flood risk problems, similar to the findings of previous participatory research Lane et al. 2011;Whitman et al. 2015;Bracken et al. 2016). The practices of participatory working help to unify local and official perspectives on flood risk and develop local capacity to understand and take action (Pain 2004) in ways that established approaches to communication have been shown not to be able to achieve.

Conclusions
The last three decades have seen rapid changes in our approaches to addressing flood risk, and a professional acceptance that flooding cannot be prevented and must instead be managed. Societal resilience to floods has emerged as a key pillar of this new approach to 'living with floods'. Changes in policy have increasingly focused on the resilience of critical infrastructure, and developing community resilience has increasingly been undertaken through an educational model of risk communication. However, research suggests that this approach is failing to develop individual and community capacities for understanding and responding to floods in a resilient manner.
The research presented here has demonstrated the application of participatory approaches to exploring the linkage between flood risk communication, individual behaviour and generating resilience. We have worked together with a competency group drawn from a flood-affected community to understand how they use and interpret current flood risk communications, what information is important to them in understanding and responding to floods in a resilient manner and how could information be better communicated. Our conclusions are as follows: 1. Current approaches to flood risk communications fail to meet user needs in understanding flood risk or allowing personal judgements of how and when to act. Through a reliance on communicating simple, threat-based messages rather than developing in-depth understanding, current communications heighten threat appraisal, but diminish coping appraisal. This promotes non-protective behaviours, either through wishful thinking and over-reliance on management organisations, or through denial and learned helplessness. 2. Users desire a greater range of information about floods, including locally specific information on flood dynamics, which would allow them to understand their personal flood risk situation and how floods will affect them. Delivering this information is vital to enable those at risk to judge what protective actions they can take, and when they should take action. Our results demonstrate that users desire forecast information beyond what is provided currently. Without forecasts of river levels or flood extents, users are unable to judge the potential severity of future flooding, which means they are reluctant or unwilling to take action blindly. 3. There are a great variety of different perspectives on how flood risk should be communicated and the purpose of these communications, even within a small area. The complexity of the risk message-behaviour interface means that one message cannot be tailored to all perspectives. We propose a communications model which is instead focused on the development of 'flood literacy', where communities and individuals are empowered to develop their own knowledge about local flood risk and how they can act to manage it. 4. Flood literacy can reinvigorate flood communications as a tool for developing flood resilience by establishing flood communications as a two-way dialogue focused on the development of shared, locally grounded knowledge. Participatory working approaches represent a vehicle through which communications and resilience can be linked. Resilience and participation are both grounded in the principles of trust, the development of relationships, and the co-production of knowledge and solutions. Participation therefore has the potential to offer a solution, re-imagine our approaches to communication, integrate alternative perspectives and place 'knowledge consumers' at the heart of the process.
Nat Hazards (2018) 92: 1665-1686 1683 5. We propose four co-produced prototype user interfaces which can deliver the information needed to help those at risk develop flood literacy.
The challenge of quantifying how new and innovative modes of knowledge creation, communications and relationship-building can provide valuable opportunities for bettering flood risk management remains. However, the approaches described here have important implications for how we communicate flood risk and how we work alongside those living with risk to develop more flood-resilient communities.