1 Introduction

This article analyses some of the factors that currently limit the human benefits of flood forecasting, warnings and response systems (FFWRS), in order to suggest ways of enhancing these benefits. We focus specifically upon warning response which is a key benefit determinant. Research results for two interrelated human dimensions of flood warnings are presented. First, we examine human response to flood warnings which can reduce damage which floods would otherwise cause, and which generates economic benefits of enhancing those systems. In Britain the economic ‘worthwhileness’ of investments in FFWRS must be demonstrated through benefit-cost analysis. Second, we evaluate some of the non-economic impacts of flood warnings, including reduction of ill health effects and loss of life. We argue that the public’s health, security and safety are important warning benefits, and that investment decision-making should be based upon more than economic arguments. Response to warnings is currently limited. We argue that risk communication designed to increase response should not be based solely upon economic perspectives. A trans-disciplinary approach is required which takes account of research findings in the social and behavioural sciences. Our research results suggest that a number of issues need to be addressed regarding current risk communication strategies in order to enhance public response to flood and other slow-onset warnings.

2 Changing public expectations and flood policy developments

In the flood field in Britain and internationally, as the technological sophistication of FFWRS increases, both people and warnings are becoming more important. The potential is growing for better flood warnings and improved warning response (Andryszewski et al. 2005). Public expectations are growing, with increased demands for better warnings, especially where loss of life occurs (Johnson et al. 2005). In response a multi-media approach is being developed in England and Wales by the Environment Agency—the lead institution in the flood risk field—to connect more people to warning systems, and to deliver more targeted messages to those at risk and to enhance warning response (Environment Agency 2003). In addition, the policy framework for flood management in England and Wales has moved away from flood defence towards flood risk management (Defra 2005) bringing the human dimension more to the fore.

This is illustrated by two examples. First, the Foresight ‘Future Flooding’ project concluded that changing societal characteristics had a greater influence than climate change on the rate of risk enhancement (Evans et al. 2004). It also concluded that a portfolio approach to flood risk management—combining traditional flood defence with non-structural alternatives such as flood warning systems—was the best means of countering increasing flood risk. Second, the new UK government policy for flood risk management articulated by the Department for Environment, Food and Rural Affairs (Defra)—termed Making Space for Water (Defra 2005)—sees human behavioural change as equally important as flood defence measures in developing sustainable policies. Flood risk management is becoming people-centred rather than water-centred, with a greater emphasis on FFWRS as ‘an important risk management tool’ (Defra 2005, p. 36).

We acknowledge that flood forecasting and warning services are not just about ‘event specific’ warnings, which are the focus of the data below, but also about year-round awareness raising and information provision which, if successful, may create a context in which event specific warnings are less crucial and make less difference because residents of flood risk areas are more flood aware and prepared. It is also important to recognise that flood warning systems are not just about warning people as individuals, but also about providing information for agencies and institutions so that they can close flood barriers, begin emergency response and mobilise the necessary resources. These effects of warnings are not the focus of this article, important though they are.

3 Economic benefits of flood warnings

Flood forecasting provides a basis for warning and for more informed decision-making by those in the path of floods to reduce their flood damages which may be measured in economic terms. Since the Bye Report on the Easter 1998 floods (Bye and Horner 1998), which was critical about the performance of FFWRS in England and Wales, the Environment Agency has developed a flood warning investment strategy (Environment Agency 2003). Estimations of economic loss savings underpin the Environment Agency’s latest strategy, while recognising that there are uncertainties in estimating flood warning benefits. The Agency has embarked upon an investment of some £250 millions over 10 years up to 2012/13 to deliver its current targets and an effective flood warning service. This investment was estimated to yield benefits in flood damages avoided of £1,261 millions over 10 years with a ratio of benefits to costs of 4.82–1.0.

3.1 Flood damages and the impacts of warnings

In theory, all of the adverse impacts of flooding may be reduced by flood warning. There is now a significant body of research for England and Wales on tangible flood damages (Penning-Rowsell and Chatterton 1977; Parker et al. 1987; Penning-Rowsell et al. 1992). The most recent work (Penning-Rowsell et al. 2005a) reveals that potential flood loss values have been rising rapidly in Britain, largely owing to increased affluence, and are almost certainly also rising similarly in other European countries (Messner et al. 2005). This research and the resulting maximum potential flood damage data (i.e. assuming no warnings) provide an important starting point for estimating flood warning benefits.

Research evidence on the damage reducing effects of flood warnings in Britain is sparse, but began accumulating in the 1970s (e.g. Penning-Rowsell et al. 1978) and was summarised by Parker (1991). This produced flood damage reduction relationships for four flood warning lead times, five flood depths, and two flood durations (short and long, although differences in savings are small). Most research focussed upon flood damage savings to houses and their contents; more limited data are available for non-residential properties (Penning-Rowsell et al. 2005a).

The summary equation used by Parker (1991), has since formed the basis of assessing the economic impact of warning arrangements for the residential sector in Britain (Table 1):

$$ {\hbox{FDA}}\,{\hbox{ = }}\,{\hbox{PFA}}\, \times \,{\hbox{R}}_{\hbox{1}} \, \times \,{\hbox{PRA}}\, \times \,{\hbox{PHR}}\, \times \,{\hbox{PHE}} $$
(1)

See Table 1 for explanation of symbols.

This formula was adapted by the Environment Agency to cover the performance factors and targets used for developing and evaluating its new flood warning investment strategy for England and Wales (Environment Agency 2003):

$$ {\hbox{FDA}}\,{\hbox{ = }}\,{\hbox{(AAD}}\, \times \,{\hbox{DR}}\, \times \,{\hbox{C)}}\, \times \,{\hbox{(R}}_{\hbox{2}} \, \times \,{\hbox{RA}}\, \times \,{\hbox{PR}}\, \times \,{\hbox{RE)}} $$
(2)

Strategies for increasing tangible flood warning benefits in England and Wales (Fig. 1) have focussed upon improving C, R2, RA, PR and RE. Social surveys of flood warning recipients have been designed to determine who was available to respond to a warning (PRA and RA in Table 1), actions that they were able to take, whether actions were effective (PHE; RE), the damage that was thereby averted (FDA), and the factors that affect this including flood warning lead time, the availability of assistance with moving vulnerable household goods, etc. Prior to 2004 we undertook some 1200 interviews and, since the mid-1990s, the Environment Agency has commissioned regular surveys from the British Market Research Bureau (BMRB) to reassess PHR or PR and PHE or RE.

Fig. 1
figure 1

Flood Warning Investment Strategy Performance Targets 2003/4 to 2012/13 for England

Full size image
Table 1 Flood warning equation parameters
Full size table

3.2 Parameters from earlier studies (see Table 1) and recent survey results

Based on research between the late 1970s and 1991 (e.g. Chatterton et al. 1979; Penning-Rowsell et al. 1983; Parker (1991) proposed values of 0.55, 0.75 and 0.70 for PRA, PHR, PHE factors, respectively, leading to a combined value of 0.29 to compare with the very optimistic combined value of 0.70 published in 1977 (Penning-Rowsell and Chatterton 1977).

Table 2 The saving of household inventory flood damage by those given a warning and those not warned
Full size table

Environment Agency post-event surveys gauge flood warning dissemination reliability (R2), and also the availability of residents to receive warnings (RA). Developments such as the Automatic Voice Messaging (AVM) direct telephone service mean that residents no longer need to be at home to receive warnings and, indeed, those registered on AVM at the time of flooding were twice as likely to receive warnings as those not on AVM (Tunstall et al. 2005). However, nationally, AVM is only taken up by about 35% of at-risk residents, despite its promotion (Environment Agency 2005). As yet, post-event surveys do not indicate any clear improvements in overall warning dissemination rates (R2 × RA) such that surveys since 1997 suggest that the percentage in receipt of a warning is currently unlikely to be more, on average, than 40% (Tunstall et al. 2005). A concern in placing reliance upon new technologies is their penetration and acceptability to people at risk (Tapsell et al. 2004). However, the new Floodline Warnings Direct systems (Andryszewski et al. 2005) and other changes under the Environment Agency Investment Strategy are expected to yield improvements in both R2 and RA (Fig. 1).

Not all people can respond to warnings and PHR and PR values reflect this: not all those receiving a warning will be able to understand it and be physically able to take action. Post-event surveys show that English (the principal language in warnings) is not the first language for about 1% of the population at risk, and that 15–23% of residents have some form of disability (mean = 18%). Our view is that PHR or PR is not likely to be more than approximately 80–82%. Research for the Environment Agency should enhance our understanding of whether and how specific vulnerable groups heed and respond to warnings (Thrush et al. 2005) and perhaps suggest means for improvement here.

Past research has suggested that response effectiveness (PHE and RE) is generally low. Sandbagging and putting flood boards in place aim to prevent flood waters entering a property and are common where flooding is very frequent. Alternatively, damages can be reduced by raising or moving valuables out of the reach of flood waters. Post-event surveys provide information on the proportions of all those affected by floods taking these actions. For example, in the BMRB survey of properties affected by the Autumn 2000 floods, only 30% of those with property flooded moved valuables or personal belongings upstairs or to a safe place. About 28% moved vehicles to a safe place, and 28% blocked doorways and/or airbricks with sandbags. Where flood waters reached property, significantly more damage reducing action was taken but, amongst those with property flooded, the receipt of a warning did not appear to spur action. Indeed, in the case of moving valuables and keeping water out, the warned did significantly less than those not warned in this particular flood event.

Passive response to warnings might be influenced by the availability of flood insurance or government compensation for losses, but this does not appear to be the case in England and Wales. In results from a new survey (Tunstall et al. 2005), the details of which are discussed in 3.3 below, 78% of respondents possessed ‘new-for-old’ insurance. However, those with this insurance were as likely to take some action to save their property as without such insurance, and proved as successful in their efforts as those without. Unlike in many other European countries, in England and Wales there is very limited flood compensation. Organisations active in flood response and recovery (e.g. local authorities, fire services) are able to claim some costs for designated flood events under the ‘Bellwin’ system, but individual flood victims are normally unable to recoup their flood damage costs from government. In particularly severe flood events, a public appeal for money from local people may be successful in providing some compensation funds, but such cases are exceptional.

3.3 The latest survey results

Continuing uncertainty concerning the degree of response to and damage saving from ‘event specific’ flood warnings has led to a major new survey (Tunstall et al. 2005) which reveals that economic benefits are limited. The methods used were as rigorous as possible. Focus groups were first held in five locations to test the survey methods, including the questionnaire. Interviews were then conducted in all of the Environment Agency regions in England and Wales, organised in two phases. The first used the BMRB’s regular post-flood event survey, yielding 168 responses (residential and non-residential), 60% of which were in the Thames region, but also including Wales (22 interviews) and the North West (43). The second phase used Market and Opinion Research International (MORI) interviewers, covering the Anglian (81), Midlands (43), North East (43), North West (15), South West (35), Southern (24) and the Thames (37) regions. The two phases covered a total of 408 residential properties. Of these, the 341 who had experienced property flooding, and thus had suffered potential or actual damage, were the main focus of the study.

Residents were asked which of a list of 100 moveable household inventory items they had moved from a flood-prone location to ‘safety’ prior to or during a flood. The term “warned” (Table 2) refers to those who themselves had a warning, broadly leaving respondents to define what constituted a warning. The reported damage saving was costed using standard inventory values as used in the Middlesex University Flood Hazard Research Centre flood damage database (Penning-Rowsell et al. 2005a).

The results show that those warned act to prevent over £2,300 of potential damage but that this damage saving is extremely variable, with a large standard deviation across the sample of 128 cases (Table 2). Moreover, even those “not warned” manage to make significant damage saving amounting to nearly two thirds of the saving accomplished by those warned, simply by taking some carefully targetted actions to move particularly valuable items quickly to safety. The difference between the damage savings is, however, statistically significant.

This research also shows that vulnerable households (those with residents aged 75 and over, or ill, disabled or living alone) are slightly, but not significantly, less likely to take damage saving action (74% compared with 80%). They also do not save a significantly different proportion of property at risk compared with other households (48% vs. 53%). The vulnerable thus appear to make as much effort and achieve almost as much savings as non-vulnerable households (Tunstall et al. 2005, p. 31). Therefore, there appears to be no basis for including the ‘capability/ability’ factor (PHR or PR) in the equations above (Table 1), over and above other factors that may explain damage savings. It is therefore excluded from the calculations shown in Table 3.

Table 3 Residential flood damage reductions with warning: summary results
Full size table

Table 3 sheds light on how potentially large economic benefits are whittled downwards in practice. First, inventory items only amount to some 52% by value of total potential damage (which includes damage to the building’s fabric), and the movable items only constitute some 41% of inventory value. This means that potential savings will inherently be low in relation to total potential damage. Given that only 38% of our sample of 341 flood-affected households received what they considered to be a warning, and that only 71% of these made some form of effective response even with an 8 hour warning (55% with a shorter warning lead time), then the average result in terms of damage saving falls to only some 8–11% of inventory value (or between 4 and 6 % of total potential damage).

Table 4 Self-reported health effects of those warned and not warned of flooding
Full size table

3.4 Understanding why economic benefits of flood warnings are low

This evidence gathered in the latest surveys suggest that the economic impact of flood warnings is limited, partly for the reasons explained in the previous paragraph. However, it is necessary to make a number of important qualifications about these results. First, the aggregation of individual responses to flood risk and warning may result in poor economic benefits, which may subsequently influence investment decisions. Disaggregation of response to the individual or household level may cause benefits to be missed or under-reported. Household survey responses, from which the above results are derived, may also miss the economic benefits of warnings accruing to non-floodplain residents visiting, travelling or otherwise active throughout a flood risk area. Second, investment decisions about warnings should not be based upon solely economic benefits. There are also important collective, public safety and security benefits which are rarely measured in economic terms and may go unreported at the disaggregated, individual level. Significant public safety benefits—in the form of the safeguarding of life and limb—may accrue to both those occupying and those otherwise active in the floodplain, and these benefits may well go uncounted. Therefore, although the economic benefits of flood warnings are low from the survey data reported above, economic and non-economic flood warning benefits at the collective level are likely to be larger, and in some cases significantly larger, than the survey results may suggest. This is an important consideration in investment decisions.

Economic benefits are also low partly because of the relatively small proportion who receive warnings, and partly because of other important factors which influence warning response and its effectiveness. Here we need to turn to the social and behavioural sciences, other than economics, for an understanding of why public warning response is limited. Socio-psychological and behavioural research emphasises that warning is a social process and that in risk response people do not behave as linear thinkers and responders, but get involved in complex cognitive processes. For example, disbelief in warnings is a pervasive initial reaction, which is most probably followed by a variety of behaviours that may confirm, or neutralise, the warning, if indeed there is any response at all (Perry et al. 1981). Warnings received by an individual are often, in reality, received by the social group (e.g. a household) of which the recipient is a member. Members of the household become involved in decisions about how to respond which may either reinforce or reduce belief in the warning (Drabek 1986, p. 84). These factors become important in considering the central issue of how to go about motivating more people to act which in turn underpins the economic effect of warnings. Warning recipients must also be capable of acting to reduce potential losses, and so risk communication must also focus upon people being made aware of what to do to protect themselves and their property. Last, people must understand the need to respond effectively within the available time window.

Warning recipients do not respond in a pseudo-mechanical manner to warnings. Research indicates that people’s images of the future are shaped by the experience of the past, and a major limitation of human ability to use hazard information, such as a flood warning, is a basic reliance upon experience, or what Kates (1962) termed the ‘prison of experience’. The literature on risk communication and psychology shows that people often make judgements in terms of how easily they can recall past examples, or how easily they can imagine such occurrences, known as the ‘availability heuristic’ (Tversky and Kahneman 1973; Morgan et al. 2002, p. 12). The availability heuristic explains why and how some of the systematic errors appear in people’s quantitative estimates and responses to risk. Further, Fischhoff et al. (1998, p. 666) indicate that people have difficulty correcting their risk estimates for systematic biasses arising from past exposure to examples, or even in thinking spontaneously about the possibility of such biasses.

Insecurity and stress may contribute to a limited behavioural response to warnings. In a study of the influence of risk perception on stress reactions and coping strategies in Mexican people who had experienced either natural or industrial catastrophes, and who remained exposed to them, Lopez-Vazquez and Marvan (2003) found that amongst those exposed to industrial risks, there were significant correlations between feelings of insecurity and both the stress level and passive coping strategies. Those who were more highly insecure and stressed exhibited passivity.

4 Other human dimensions of flood forecasts and warnings

There are many other useful benefits of ‘event specific’ flood warnings for individuals and communities, many of which cannot be measured in monetary terms and thus are termed ‘intangibles’. These need to be taken into account in decisions about investing in FFWRS. Some of these are addressed below.

4.1 Warnings and health impacts

The wide range of physical and psychological health impacts of flooding worldwide has been reviewed recently by Few et al. (2004), but only a limited number of studies provide evidence of these impacts in Britain (Green et al. 1985; Tapsell et al. 1999; Tapsell and Tunstall 2000, 2001; Reacher et al. 2003; RPA/FHRC 2004; Penning-Rowsell et al. 2004, Tunstall et al. 2006)). Even less is known about how flood warnings may or may not modify these effects.

A recent large-scale survey carried out in 30 locations in England and Wales (RPA/FHRC 2004) investigated the impact of warning on the health effects of flooding (23% of respondents were warned in some way prior to flooding). As to these effects, some 59% of those with houses flooded attributed some physical health effects to flooding. Self-reported physical health effects experienced during or in the immediate aftermath of an event were more common (reported by 54%) than those suffered in the weeks and months following the flooding (reported by 33%). Self-reported psychological impacts, especially anxiety when it rains, were much more commonly reported than physical ones. Receiving warnings or otherwise has a complex effect on these results. Simply receiving a flood warning of some kind made hardly any significant difference to the incidence of our respondents’ self-reported physical or psychological symptoms. Furthermore, warning recipients reported experiencing as many effects as those who had no warning (Table 4). Nor does the receipt of a warning per se affect short term and long term mental health impacts as measured by the well-established General Health Questionnaire (GHQ-12) (Goldberg and Williams 1988) and the more recent Post-Traumatic Stress Scale (PTSS) (Scott and Dua 1999) (Tables 5 and 6). A GHQ-12 score of 4 or more (on the 12 item scale) is generally taken as indicative of impaired mental health. Warning recipients were as likely as those unwarned to have GHQ-12 scores of 4 or more for their “worst time” in the aftermath of flooding (60% compared with 64%). At the time of survey, many months after the flood, the proportions for their “current (time)” GHQ-12 scores of 4 or more were also similar for those warned and unwarned (23% and 25%, respectively).

Table 5 The effect of warnings on the mental health of flood victims at their “worst time” after flooding, as measured by the GHQ-12 scale
Full size table
Table 6 The effect of warnings in influencing their “current” stress levels of flood victims (i.e. at the time of the interview)
Full size table

However, we did find some evidence that a longer warning lead time has a mitigating effect on the mental health of flood victims at the time of flood and later. This did not emerge in the bivariate analyses shown in Tables 5 and 6 which identified no significant differences in the health measures according to the length of warning time. Using multiple regression analyses, a wide range of variables was examined as possible explanatory factors for the mental health and stress effects experienced by flood victims. Warning lead time then emerged as one of ten factors that had an influence on mental health at the worst time of flooding as measured by the GHQ-12 Likert scores. A longer warning time has a small but significant effect in reducing “worst time” scores (standardised beta coefficient: −0.09). It was also one of ten factors influencing the “current” stress levels of flood victims (i.e. when interviewed), as measured by the PTSS Intensity score. A longer warning lead time was again associated with a reduction in mental health and stress effects (standardised beta coefficient: −0.10) (see also Tunstall et al. 2005).

Flooding in your home can be deeply disturbing (Thrush et al. 2005). But simply receiving a warning did not afford significant protection from experiencing symptoms of ‘shock’ during or immediately after the flooding (reported by 30% of the warned compared with 35% of those unwarned). However, when those who had no warning were excluded, a longer warning time was associated with significantly less reported shock (Fig. 2). Similarly, although those who received a warning were as likely as others to report experiencing ‘fear, helplessness and horror’ as a result of the flood (40% for the warned compared with 46% for the unwarned), the trauma again lessened significantly with a longer warning lead time (Fig. 2). Therefore warning lead time, rather than receiving a warning or not, appears to be the crucial factor in reducing the adverse psychological effects of flooding. Alternatively the characteristics of floods where earlier warnings could not be given may produce more adverse psychological effects.

Fig. 2
figure 2

The effect of warning lead time on self-reported shock (Chi-square = 9.043; df = 3; p = 0.029) and other reported mental health effects (Chi-square = 15.268; df = 3; p = 0.002)

Full size image

4.2 Warnings and loss of life in floods

Reduction in loss of life and physical injury are important rationales for warning and response systems. Warnings can advise people on avoiding risky behaviour (Gruntfest and Ripps 2000). They can allow people to evacuate, as in the 1995 Netherlands floods in which nearly 250,000 people left their homes in advance of potentially life-threatening flooding.

However, loss of life in floods remains a hazard, and few recent major European floods have occurred without fatalities (European Environment Agency 2003). Research by Jonkman and Kelman (2005) sheds some light on the causes of such fatalities. Drowning in vehicles, for example, is significantly less important in Europe than in the USA. There has been an increase in the coverage and technical effectiveness of flood warning systems over the last three decades. At the same time the number of floods has also increased markedly, and although the trend in the number of fatalities per flood event remained somewhat static during the 1980s and 1990s, since 2000 the number of floods has markedly increased, whereas the number of fatalities has markedly decreased (Fig. 3). It is too early to draw certain conclusions, but flood warnings and other measures may be becoming more effective.

Fig. 3
figure 3

Relationship between flood incidence and flood deaths: Europe 1975–2005 (Source: Emdat)

Full size image

A British-based model of the factors likely to influence loss of life in floods has been developed by Penning-Rowsell et al. (2005b), taking the form:

$$ {\hbox{N(I)}}\,{\hbox{ = }}\,{\hbox{N}} \cdot {\hbox{X}} \cdot {\hbox{Y}} $$
(3)

where N(I) is the number of deaths/injuries, N is the floodplain population, X is the proportion of population exposed to a chance of suffering death/injury (for a given flood) and Y is the proportion of those at risk who will suffer injury. The model defines the characteristics of the area (effectiveness of warning; speed of onset; land use) and of the population (age profile; presence of very infirm people) as affecting fatality rates. The model appears to achieve good calibration results, albeit for just three floods (Lynmouth in 1952; Norwich in 1912; Gowdall in 2000); the actual death toll in the three events analysed was 38 compared with the model's prediction of 35 fatalities (see Penning-Rowsell et al. 2005a, b). Separating out the effects of warning and response systems, Table 7 gives the reduction in fatalities with an increasingly effective flood warning system, again as predicted by the model. Warnings appear to be less effective in saving lives than one might suppose, with a reduction in loss of life of only some 12% for the difference between a situation with no warnings compared with an "effective tried and tested flood warning and emergency plans". This small predicted difference is because the determinants of the fatality rate in the model are many and varied (as they are in reality) and the floods used to calibrate the model are characterised by a rapid rate of rise, high depths and velocities, which mean that the events are predicted by the model to be hazardous even if warnings are given. The model has limitations, not least of which is that it is based on data from just three flood events, all of which are extreme events (with estimated return periods of 100, 750 and 800 years). Further research is required to refine this model so that it satisfactorily addresses types of flooding which are more the norm in Britain.

Table 7 The reduction in fatalities with an increasingly effective flood warning, as predicted by the model developed by Penning-Rowsell et al. (2005b). The actual death toll in the three events analysed was 38
Full size table

Although we may rely on warnings to reduce the hazard to life, and Fig. 3 suggests that in aggregate this may be a sensible conclusion, the result from the existing, unrefined loss of life model suggests that this reliance may well be misplaced for all but the most benign floods. Indeed it is the combination of variables that is used by the model that most affects fatality rates, and to change just one parameter (in our case the warning system) has only a marginal effect on the predicted result, thus reinforcing the need for a portfolio approach to flood risk reduction as advocated in the Foresight project (Evans et al. 2004).

4.3 Public confidence in flood warnings

Relatively poor public response to warnings may reflect the public’s uncertainty over the reliability and credibility of the information that the warnings bring. Certainly the performance of warning systems in the past has often been poor (Parker 2004). To pursue this and other questions we have surveyed at-risk householders within the 1:50 year return period flood outline between Datchet and Teddington on the River Thames west of London (McCarthy et al. 2006). Here the available warning lead time is measured in terms of several days rather than hours. The research focussed on the river from Walton Bridge to Teddington, where there are approximately 5,000 flood-prone properties. About 13 survey sites give a spread of localities and a sufficiently large sample of households.

Using a quantitative survey with a supplementary qualitative research component the study aimed to assess public attitudes to risk and risk reduction to inform decisions regarding the ‘social feasibility’ of introducing community level flood risk reduction measures (e.g. demountables and temporary defences, and individual household measures). The survey sought to gauge current levels of resident awareness and engagement with flood risk, the acceptability of these approaches and possible constraints at a later public consultation stage (our survey was, in effect, a pre-consultation survey). The results show that even with limited experience of flooding of their homes, 53% of residents had received some form of flood warning in the past. On average respondents mention 1.6 sources of warning communication (Table 8), with the bulk being official sources, although some 13% of the residents claimed that their warning was taken from their own observations of rising water. Residents were also asked to rate on a five-point scale their reliance on their own judgement for assessing if the Thames would flood. The results (Fig. 4) show that only just under one-third (30%) of residents rely ‘mainly’ or ‘completely’ on the authorities for that judgement. More respondents rely mainly or completely on their own judgement, and many combine both sources of information. Interestingly there was no significant difference in this pattern of response between residents living on or somewhat away from the riverbank.

Table 8 Cited sources of flood warnings (lower Thames survey (McCarthy et al. 2006))
Full size table
Fig. 4
figure 4

The public’s reliance on authorities for flood warnings (sample = 206); Question: “How much, if at all, do you currently rely on the authorities or your own judgement of when the River Thames is going to flood?” (McCarthy et al. 2006)

Full size image

The results overall suggests a level of scepticism concerning the official sources, even though the accuracy of flood forecasts and warnings at this location is likely to be relatively high. Residents often rely on a number of sources of information including local visual cues and knowledge of the watercourse to make a personal assessment of the flood risk. The behavioural literature suggests that, at the local community level, people often take a broad perspective (Lopez-Vazquez and Marvan 2003) and emphasizes trust and credibility issues ahead of specific types of information, such as flood warnings in this case. Some professionals assert that trust may be more fundamental to citizen’s capacity for understanding and weighing environmental risks than any strategy of risk communication (Slovic 1993). Trettin and Musham (2000) report that the absence of withholding or distorting information, or giving people sufficient facts to generate their own opinions, may be important in building trust. They also report that the erosion of public trust in corporate management and government agencies, in the USA in particular, is a major stumbling block to success in risk communication programmes.

5 Conclusions: enhancing the benefits of warnings

Our research suggests a number of ways of enhancing the benefits of warnings. The research reveals that the economic benefits of flood warnings are low, and lower than we had previously estimated. An important reason for this is that moveable inventory items are a relatively small proportion of total potential damage. Therefore, flood warnings which allow temporary demountable or other physical community measures to be put in place are likely to be effective in increasing warning benefits because they eliminate reliance upon the portability of household inventory items. Importantly, our data are derived from a disaggregated household approach which is likely to underestimate economic and collective warning benefits. At the same time, there are many other intangible warning benefits, including public safety and security (e.g. avoidance of loss of life and physical injury) and possible mental health benefits which our research data above demonstrate. All of these human benefits need to be fully recognised and taken into account in investment decision-making in future.

The research suggests that improving warning lead time will reduce stress and mental effects. Our research and the discussion above illuminates some of the reasons for the relatively low response rate to warnings, apart from the relatively small proportions who receive flood warnings which must be addressed, and is currently being addressed by the Environment Agency’s comparatively new Flood Warning Direct service. In England and Wales low response rates do not appear to be influenced by the potential for compensation of flood losses. The psychology of risk perception and communication indicates the complex processes involved in human response to risk and warnings. Warning message content and factors such as trust are likely to influence whether systematic or heuristic risk information processing occurs and whether warning recipients are likely to respond in the desired manner. Issues of experience of flooding, belief in warnings, trust in the warning agency and authority in general, insecurity and stress, uncatered-for special needs, and the acceptability of warning technologies to people, may all be implicated, and are likely to need more attention to motivate more to respond. Most of these variables do not currently figure directly in Eqs. 1 and 2. To an inevitably limited extent, lack of flood experience may be countered through the Environment Agency acting as the community’s collective memory and making high-quality flood information available to community members, as is already largely the case. Belief in warnings and trust in the Environment Agency is likely to be a crucial underlying and pervasive influence, because public distrust of governmental and other large organisations appears to have been growing in Britain, as in the USA. Trettin and Musham (2000) suggest that the best way to build public trust in environmental risk information, is not to attempt to build trust in the institutions themselves, but as a by-product of procedures that genuinely involve the public in decision-making. In this context, the pre-consultation surveys in the Lower Thames referred to above are perhaps a small step forward in the right direction towards building public trust and confidence, but the Agency’s overall public trust profile needs to be maintained at a high level. Higher levels of trust may help to counter insecurity and stress which appear to neutralise response and action, and reinforce passivity. Catering for special needs in the flood warning service, and more generally in flood management, may help to build trust and confidence and to reduce insecurity. Finally, making sure that new flood warning communication technologies are acceptable to people will be important.

Currently, poor public response to flood warnings, and a limiting economic perspective on warning benefits, could undermine the progress that has been made in seeing the future UK vision for flood risk reduction formulated in terms of a portfolio approach to developing multiple flood risk measures, with flood forecasting and warning at its core. This means that much more attention needs to be given to ‘people factors’ and to ways of influencing human behaviour.