This paper discusses whether flood hazard in the UK is increasing and considers issues of flood risk management. Urban development is known to increase fluvial flood frequency, hence design measures are routinely implemented to minimize the impact. Studies suggest that historical effects, while potentially large at small scale, are not significant for large river basins. Storm water flooding within the urban environment is an area where flood hazard is inadequately defined; new methods are needed to assess and manage flood risk. Development on flood plains has led to major capital expenditure on flood protection, but government is attempting to strengthen the planning role of the environmental regulator to prevent this. Rural land use management has intensified significantly over the past 30 years, leading to concerns that flood risk has increased, at least at local scale; the implications for catchment-scale flooding are unclear. New research is addressing this issue, and more broadly, the role of land management in reducing flood risk. Climate change impacts on flooding and current guidelines for UK practice are reviewed. Large uncertainties remain, not least for the occurrence of extreme precipitation, but precautionary guidance is in place. Finally, current levels of flood protection are discussed. Reassessment of flood hazard has led to targets for increased flood protection, but despite important developments to communicate flood risk to the public, much remains to be done to increase public awareness of flood hazard.
Floods are one of the most damaging and dangerous natural hazards. This meeting has followed shortly after the inundation of New Orleans by Hurricane Katrina and the vivid television images seen around the world graphically illustrated the devastation that can be caused when flood defences are breached. This paper focuses on fluvial, rather than coastal flooding, but still we only have to reflect on the last few months and years to recall a wide range of flood disasters. On 26 July 2005, 5 million people were affected by the floods of Mumbai (Bombay) and 1000 deaths reported. Nine hundred and forty millimetres of rain fell in this single event—greater than the annual rainfall of London. In August 2005, flooding in Central Europe caused fatalities in Romania, Germany, Switzerland, Bulgaria and Austria. This followed the devastating floods of Central Europe in 2002, with widespread evacuation of towns and cities across the region and damage estimated at 21.5 billion euros (Kron 2005). We may also recall the widespread loss of life in Mozambique in 2000 when half a million people were made homeless and 700 died, only to be followed by further flooding in 2001. Clearly, floods are a recurrent global phenomenon.
The UK has only small rivers by global standards (the Severn and the Thames have catchment areas of 10 000 km2, compared, for example, to the Nile at 3.3 million km2), so that our floods tend to be smaller-scale events, but nevertheless devastating to the communities affected. For example, on 8/9 January 2005, in Carlisle, there was widespread rainfall, of up to 164 mm in 24 h. Flood defences were over-topped, 2000 properties were flooded, two lives lost and an estimated £450 million damage incurred. The flood was estimated to have a return period of 1 in 150 years, whereas the defences were variously designed for 1 in 20 to 1 in 70 year return period events. And on 16 August 2004, there was the flooding of Boscastle, in southwest England. Flash flooding occurred in a small catchment, 60 properties were flooded, some completely destroyed, and more than 100 people were rescued by helicopter. This was a rare event for the UK, with 181 mm rainfall in 5 h (analysis suggests a return period of in excess of 1000 years). And if we look back just a little further in the UK record, we may recall the floods of 1998, 2000 and 2001.
It can be seen that floods occur on a regular basis, both nationally and globally. They range in severity from frequent events, with little more than nuisance value, to catastrophic floods with major loss of life. There are many issues to discuss with respect to flooding, ranging from the science underlying extreme meteorological events to the social and political aspects of flood protection and management, but within the limited scope of a single paper I will focus on principally on UK flood risk and specifically the scientific question of whether flood hazard is increasing due to land use and climate change. The paper concludes with a brief discussion of flood management in England and Wales, with respect to risk assessment, public awareness and public protection. Although the paper focuses mainly on the UK, the issues raised are generic and of relevance to flood management worldwide.
For a more extensive treatment of flooding issues, the reader is referred to Wheater (2002) and, more generally, the 2001 Royal Society Floods Meeting, reported in the 2002 Phil. Trans. R. Soc. A (volume 360, number 1796).
2. Is UK fluvial flood hazard increasing?
Human activities can and do profoundly change the local physical environment in which we live and we are increasingly aware of the effects of human activity on climate. We consider first the effects on fluvial flooding of land use change. Urban development and urban flooding are addressed in §2a and the issues of rural land use change and land use management in §2b. Climate change impacts are considered in §2c.
(a) Impacts of urban development on flooding
The replacement of a green field site with urban development is one of the most dramatic land use changes, and its direct and indirect physical effects have been well understood for 40 years (Hall 1984). Put simply, the construction of impermeable surfaces generates rapid overland flow that bypasses the natural storage and attenuation of the subsurface. This flow is conventionally collected in storm drains and rapidly conveyed to the nearest stream. Storm runoff volumes will therefore increase and the response times decrease, leading to a potentially dramatic local increase in flood peaks (e.g. Wheater et al. 1982).
The magnitude of effects of urban development on stream flow will depend on the natural response of the catchment; effects will be greatest where natural runoff is low, e.g. in catchments with permeable soils and geology. Changes in flood seasonality may also arise. Natural catchments in the UK mainly flood due to prolonged rainfall in winter, when soils are wet and storm runoff is readily generated. Urban catchments are not affected to the same extent by antecedent conditions and respond more rapidly to rainfall, hence intense summer rainfall may become a major cause of flooding, leading to a change in flood seasonality (see Institute of Hydrology 1999).
Regional analysis (Institute of Hydrology 1999) provides simple tools to estimate the impacts for small catchments. These are based on analysis of the hydrological response of large numbers of catchments across the UK and the identification of statistical relationships between both flood peak frequency and hydrograph response and catchment characteristics, including the extent of urban development. It is expected that the relative effects of urbanization will reduce as the severity of event increases, but current design guidance to quantify this (that effects tend to zero at a return period of 1000 years) is highly speculative. For larger catchments, effects are more complex, as the location of development within the catchment will affect response. For example, urban development located near to the outlet of a catchment may generate runoff before the main response of the natural catchment arrives. Clearly the overall effect on the catchment flood peak will depend on the relative magnitude and timing of the constituent responses.
To mitigate the local-scale effects of urban development, engineered solutions have routinely been adopted to reduce flood peaks through the provision of storage. Construction of detention storage, in the form of a reservoir, is a commonly required solution for both small and large-scale developments. For example, Milton Keynes, created as a new town in the 1970s, was designed to use a balancing lake to temporarily retain storm runoff. Currently there is much interest in sustainable urban drainage systems to manage urban runoff and associated problems of water quality. Various design solutions can be implemented, for example, restoring the infiltration of rainfall into the soil by directing storm runoff to engineered soakaways (Verworn 2002). However, there is no clear understanding of the effects of extreme rainfall on the performance of the design measures installed to mitigate urbanization effects.
In an analysis of the Thames catchment, Crooks et al. (2000) investigated the effects of 30 years of urbanization in two catchments, the Cut and the Mole, encompassing the new towns of Bracknell and Crawley respectively. For the Mole, due to the provision of 500 000 m3 of storage, flood frequency showed no increase from the 1960s to the 1990s. For the Cut, flood frequency appeared to increase from the 1960s to the 1970s and then to reduce as additional storage was put in place. For the Thames as a whole, simulation indicated that effects of land use change over 30 years were small. The urban fraction had increased by 40%, but still represented only 6% of the catchment area. Similar conclusions concerning the impact of urban development at large scale have been reached in simulation studies of the Rhine (Bronstert et al. 2005).
We conclude that urbanization can represent a very significant increase in flood risk at small catchment scale, but that the effects are commonly mitigated, to a greater or lesser extent, by design measures. The impacts of effects at larger scales are complex and depend on the relative magnitude and timing of sub-catchment responses and the performance of mitigation strategies. Relative effects of urbanization on flooding are expected to decrease with increasing storm return period, but the performance of mitigation strategies for events rarer than the design criteria adopted is largely unexplored. While detailed modelling is commonly used at small scale to design mitigation measures, adequate modelling methods to represent the larger-scale effects are not yet in place. For example, there is little guidance on how urbanization and mitigation effects can be included, other than by ad hoc empirical adjustments to runoff coefficients and routing parameters. There is a need for a hierarchical modelling approach in which the essence of the detailed model performance can be represented within a distributed or semi-distributed catchment-scale model, but to characterize these local-scale effects requires extensive local information.
(ii) Urban storm water flooding
Part (i) above addressed the effects of urban development on river flooding. There are, however, also issues of flooding due to surface runoff within the urban environment. Indeed, a substantial proportion of insurance claims for flood damage relate to these issues (E. C. Penning-Rowsell 2001, personal communication). Storm runoff is normally channelled, via gully pots, into storm sewers, which have conventionally been designed to accommodate relatively frequent events (with a return period of a few years). Under more extreme conditions, these sewers will start to surcharge (flow full under pressure) and as pressures build up, manhole covers can lift and the sewers discharge to the surface. Such flows combine with surface runoff to generate flooding of roads and properties. The frequency of this surface flooding is not a design criterion, is often not known, and will vary greatly for different systems. There is currently a lack of technical capability to address this problem. Some models used to represent sewer flows can represent discharge to the surface, but there is currently no practically applicable method to represent the surface routing of overland flows and associated storm sewer interactions. However, models to represent these interactions are under development and the high-resolution topographic data needed to support such modelling are becoming available for the urban environment, for example, from light detection and ranging (LIDAR) airborne remote sensing systems. This is an area of current research within the UK flood risk management research consortium (FRMRC).
(iii) Floodplain development
Many major towns and cities are located adjacent to rivers and there are continuing economic pressures to build in river floodplains. However, floodplains have precisely the function indicated by their name; rivers can be expected naturally to flow out of bank at least every few years. The natural functioning of a flood plain is to store and subsequently release flood waters, hence attenuating a flood as it travels downstream. Over the last century or more, floodplains have been increasingly used for urban and agricultural development and the need to protect that development has led to engineered disconnection of the river from its floodplain. The result is a loss of flood attenuation and increases in flood risk downstream. This remains an issue of major concern for the major European rivers, such as the Rhine, where levels of flood protection for some German cities have significantly decreased and active efforts have been made in recent years to recreate floodplain storage. In the UK, the same issues arise, although little work is available to quantify the effects of historic changes. However, there is current interest in the UK in the potential for the return of floodplain land to an active storage role, for example, by reducing the level of flood protection of agricultural floodplain land.
In the UK, economic pressures to permit floodplain development have been severe and until very recently, planning authorities have been able to act against technical advice and permit floodplain development. Once such development is in place, the inevitable flooding gives rise to calls for flood protection and the value of the development may well be sufficient to justify a flood protection scheme. A notable recent example is the Maidenhead–Windsor flood relief scheme for the river Thames. Development in the Thames floodplain at Maidenhead was permitted, flooding occurred and protection was subsequently justified on cost-benefit analysis. The only feasible technical solution was to build a parallel flood channel to the Thames, the Jubilee River, at a cost of some £100 million.
Recent moves have been made by the UK Government to strengthen the role of the environmental regulator (the environment agency (EA) of England and Wales) in the planning process, and also to raise awareness of planners of the risks of flooding. A particular problem is the location of strategic facilities. Currently it is not uncommon for emergency services to be located in the floodplain, and also vulnerable properties, for example, hospitals and residential homes for the elderly.
(b) Rural land use management and flooding
While urbanization clearly represents a dramatic change to the natural environment, effects of other land use changes are more subtle. A classical problem is the impact of afforestation, which has been the subject of long-term experimental research in the UK for 40 years. The effect of afforestation is strongly dependent on climate, due mainly to the importance of interception storage (high rates of evaporation occur for water wetting the surface of leaves), but many studies concur that in the long term, afforestation reduces flows (Bosch & Hewlett 1982). However, in the short and medium term, effects may be very different. Relevant data are limited, but studies by Robinson (1986) show that the drainage practices widely used at that time to establish forests in the UK uplands give rise to an increase in storm runoff and that this effect may last for many years. These observed effects also illustrate some of the associated modelling problems. While the water balance effects of afforestation can be simulated using available and widely used models of forest evaporation, the impacts of drainage cannot be readily predicted. And whether the aggregated effect of local change is discernable at larger scales has been disputed, even in the interpretation of a common dataset (Jones & Grant 1996; Thomas & Megahan 1998).
Current concerns for the impact of agricultural land management practices in the context of UK flooding relate mainly to impacts of changes to agricultural management practices on soil structure. These are echoed by similar concerns elsewhere in Europe (e.g. Brontstert et al. 2002). For example, a review of soils after the 2000 floods pointed to extensive soil degradation (Holman et al. 2002). Recent changes in arable agriculture are associated with changes in cropping and land cultivation practice, the increasing use of heavy machinery, and pressures to work land when soil moisture conditions are unsuitable and to work land unsuitable for purpose. While earlier research studied effects of cultivation on soil structure and physical properties (Goss et al. 1978), these new effects have received limited attention.
In the uplands, changes to grazing patterns are of particular concern and reflect economic pressures to increase animal stocking densities, maintain stock on the land over winter, and hence to keep stock on unsuitable land or under unsuitable conditions of soil wetness. At Pontbren, in mid-Wales, for example, stocking densities of sheep increased by a factor of six from the 1970s to the 1990s, and the weight of individual animals doubled (R. Jukes 2004, personal communication). It is believed that changes in runoff processes have occurred, exacerbated by removal of hedgerows and woodland buffer strips. These effects include reduced infiltration and increased overland flow, higher flood peaks, in some cases accompanied by extensive channel erosion and lower low flows.
The local scale effects of changing land management practice are complex, and depend on soil type, land use, location (in the hillslope context) and the timing of access to the land by machinery and animals. There is concern that such dramatic changes, if of sufficient spatial extent within a catchment, may significantly alter the hydrology of major rivers, but the effects are not known at present. Simulation results from Germany (Bronstert et al. 2002) indicate the potentially significant effects of soil structure on catchment scale runoff, but a high degree of uncertainty associated with the local-scale parameterization of these effects. In the UK, plot studies are limited and have focused on arable crops; information on upland land management, of particular importance for headwater catchments and runoff production, is almost non-existent. There are no data to support understanding of the scale-dependence of these effects. The FRMRC has therefore established a major multi-scale experimental programme at Pontbren to evaluate response at plot and hillslope scale and hence to aggregate to catchment scale. Development of suitable modelling methods to represent impacts of land use and land management change at catchment scale is a research priority.
(c) Impacts of climate change on flooding
Climate change is discussed elsewhere in this issue (Mitchell 2006). The ability of global climate models (GCMs) to represent anthropogenic and other climate forcing and reproduce the historical record of observed temperature provides convincing evidence (to most) of the significance of global warming. What then are the anticipated effects on flooding? We turn to GCMs to provide scenarios of future climate. However, the interpretation of GCM results is not straightforward. While GCM performance with respect to global temperature is impressive, precipitation effects are more problematic. There are serious inconsistencies between GCMs with respect to not only the magnitude of change, but in some cases the direction. Direct outputs of precipitation from GCMs are not generally considered reliable, and hence there has been a considerable research effort to improve the estimation of precipitation under future climate by the development of dynamical and statistical downscaling methods. In the former, GCMs are used to provide the boundary conditions for smaller scale regional climate models. In the latter (e.g. Wilby et al. 1998), statistical relationships are sought between the more reliable outputs of GCMs and precipitation estimates.
For the UK there is a general consensus that climate change will bring warmer, wetter winters and drier summers to the south of England, and, to the north of the UK, wetter winters and also wetter summers. Recent results of the UK Climate Impacts Programme UKCIP02 are presented by Hulme et al. (2002). The obvious interpretation of these scenarios is that flood risk over the UK can be expected to increase and current UK Government (DEFRA) guidance is that when considering flood protection schemes, a potential increase of 20% in flood magnitudes by 2050 should be considered. However, the generation of floods in natural catchments depends strongly on antecedent conditions, i.e. how wet or dry a catchment is when a storm occurs. This requires continuous simulation of hydrological response. Recent outputs from such models (Reynard et al. 2004), based on UKCIP02 scenarios, show a wide range of impacts on peak flows, with both increases and decreases, depending on the location and the characteristics of the catchments. While all but the most extreme increases are within the 20% range of sensitivity testing currently recommended by DEFRA, for a number of catchments, simulations suggested an expected reduction in floods, due to the effects of drier summers on autumn and winter antecedent conditions. Clearly such results are dependent on limitations of the precipitation inputs and the adequacy of the hydrological models used for impacts assessment. Effects on groundwater-dominated rivers (such as the chalk catchments of southern England) are particularly uncertain. The extremely wet autumn of 2000 gave rise to prolonged flooding associated with exceptionally high water tables. Is this indicative of the effects that might be expected from wetter winters, or will the expected drier summers delay winter groundwater recharge and mitigate against this?
An alternative approach to the use of GCMs is to look for a signal of climate change in the observed record, either of precipitation, or of flows. Chandler & Wheater (2002) developed a spatial model for daily rainfall, applied to a flooding problem in Ireland, which was able to represent climatic variability. This application showed clearly that on a decadal time-scale, non-stationary effects were present, but within the period of record available, was unable to discriminate between trend and long-term periodicity. However, an important advantage of the model is that characteristic climate indicators, such as the North Atlantic Oscillation, can be incorporated as explanatory variables, and that effects of climatic variability on rainfall frequency can be explicitly represented. Osborn & Hume (2002) investigated daily precipitation over the UK over the last 40 years were able to demonstrate an increase in the frequency of heavy winter rainfall. However, Robson (2002) when investigating apparent change in river flows, found that some evidence of recent trends of increasing flows could be explained by climatic variation, but there was no statistical evidence of a long term trend in flooding over the last 80–120 years.
We conclude that impacts of climate change are complex and evidence of changes is not detectable in the current observational record. UK Government is currently providing appropriate precautionary guidance on increased flood risk, but that future effects are highly uncertain. Further work is required to quantify the uncertainty associated with such estimates, and to improve methodologies to reduce that uncertainty. Some of this is underway in current DEFRA-funded research (e.g. project FD2113). A particular concern, which is not to the author's knowledge being addressed, is the effect of climate change on severe storms. These events lie beyond the predictive capability of GCMs, but could be investigated using detailed meteorological models. The question of whether a Boscastle event is likely to become more frequent due to global warming remains unanswered.
3. Some issues of flood risk management in England and Wales
The recent sequence of UK floods has raised public and political awareness of flooding. As a result, increased expenditure is being made on flood defences. However, there are difficult issues to be addressed in the allocation and prioritization of resources for flood protection. For example, how does one define an appropriate level of risk for flood protection; are there issues of social equity that should be considered in the prioritization of resources; what is an appropriate balance between the responsibilities of Government and its agents in the provision of flood protection and the insurance industry in the provision of flood insurance? There are also related issues of the public perception of flood risk and issues of the environmental acceptability of flood protection measures.
The EA of England and Wales has recently implemented a number of significant developments in attempting to promote awareness of flood risk. Its web-site home page states that 5 million people (and 2 million properties) are in areas at risk from flooding. An enhancement has been made to publicly available flood mapping. For fluvial floods, the estimated 100 year and 1000 year return period flood inundation areas (excluding effects of defences) are now mapped, and all areas with recent flood protection (at the 100 year return period level) are also indicated. Within flood inundation areas, risks are designated as significant, moderate or low (risk greater than 75 years, between 75 and 200, less than 1 in 200). The system can be interrogated by post code. However, it remains the case that many residents in flood plains are unaware of the risk. For example, it is a common perception that since rivers are managed, floods can be controlled. The fact that during a flood the river discharge rate may exceed river channel capacity by a factor of two or more is not understood. There is also an issue that flood defences may be exceeded (by an event which exceeds the design standard, as in Carlisle), or may fail to give the design protection, due to technical deficiency or deterioration. Residents living behind defences may be unaware of these risks, particularly if they are tenants, rather than property owners, but the risks may be considerable. Although there is increasing awareness within the EA that this is an issue to be addressed and there are moves towards a system where evaluation of extreme events beyond the design criterion of the defences is considered, much work remains to be done to quantify such risks on a national basis.
The current level of design protection in the UK is variable. The Carlisle experience was noted above (currently 20–70 year return period protection), and many major towns subject to flood risk have levels of protection of the order of 50–60 years. The recent Windsor and Maidenhead flood scheme was designed for the approximately 1 in 60 year event (corresponding in that case to the 1947 flood). In exceptional cases, much higher levels of protection have been considered appropriate. The Thames Barrier, for example, is associated with a 1 in 1000 year return period protection, and for reservoir safety, where lives would be at risk from a dam failure, levels of 1 in 10 000 years or more extreme (the probable maximum flood) are applied.
The UK government (DEFRA) has recently set targets of 100 year return period protection for urban areas, but these are currently being prioritized on cost/benefit grounds. This has been the established method to judge acceptance of a scheme and its priority, but cost/benefit analysis has been criticized on the grounds that issues such as social equity are thereby excluded from consideration (Tapsell et al. 2002). The use of alternative criteria is under discussion, as is the appropriate level of protection, for example is a 1 in 200 year return period more appropriate to current attitudes to risk than 1 in 100? However, at national scale even implementation of the 100 year targets has major cost (and budget) implications.
To provide protection for rare floods, such as Boscastle, would not be economically viable, and hence the focus must be on effective flood warning and associated contingency plans. It is of course also commonly the case that conventional flood protection works are seen to be unacceptable on environmental grounds. Residents of Bewdley on the river Severn rejected plans for floodwalls for flood protection—and were then flooded. Interestingly, this has generated a new approach to flood protection; demountable flood defences have been put in place, operating successfully for the first time in 2005. Clearly, a holistic view is needed of flood risk management; public participation is important, environmental aspects must be considered and an integrated approach to managing flood risk on a whole catchment basis is essential. These represent major practical challenges, but are being taken forward in current practice.
4. Concluding remarks
Flooding is a major natural hazard, but, under stationary environmental conditions, the level of risk (hazard) can be estimated. This then raises a set of technical, political and social issues about the management of risk. An integrated approach for a river system is essential and design solutions to mitigate risk must take into account local stakeholder interests and environmental impacts. Historically, ‘hard’ engineering solutions (e.g. flood walls and channels) have been preferred (with associated environmental disbenefits), but alternative approaches may have significant environmental benefits, for example, by returning flood plains and riparian wetlands to an active role in flood storage, or, as in the case of the Jubilee river, to the creation of new riparian and aquatic habitats.
What levels of protection should be adopted is essentially a political question. Recent specification of target levels of protection represents a judgement of what is considered by society to be an acceptable risk, and one that will undoubtedly continue to be the subject of debate and re-evaluation. Cost/benefit analysis has resulted in a national situation where lesser levels of protection have been provided; its use as a criterion for prioritization continues, but there is a perception that other issues of social equity are relevant, and should perhaps be formally recognized.
There will always be extreme events that place people and property beyond economically viable protection. Warning and action to evacuate will be the only solution. Detailed consideration of these issues lies beyond the scope of the present paper. However, the situation of residents living behind defences that might fail or be overwhelmed has been raised. There are technical issues that need to be considered on a national basis and the issue of public perception of risk is relevant. It is argued that public perception of risk is not yet adequate and that much work is needed to raise awareness and preparedness for action in the event of such occurrence.
The above discussion has related to the situation where flood risk is stationary and predictable. The main thrust of this paper has been to consider possible non-stationarity in flood risk. We have seen that there is an expectation that, for the UK, flood risk will increase due to envisaged climate change. A precautionary position has been adopted, but the effects are highly uncertain, and estimates depend on assumptions concerning the translation of climate model outputs to the occurrence of extreme precipitation. It is argued that more work is needed to quantify and refine uncertainties and a particular issue that needs attention is the assessment of low probability extreme events.
Land use change clearly influences flood risk. This has been well understood in the urban context for four decades, and design measures to mitigate effects are in place. The effects of environmental change due to more subtle agricultural practices remain unquantified. There are clearly important local-scale effects of soil degradation, and the term ‘muddy floods’ has been coined to describe flooding from agricultural runoff. However, effects at catchment scale remain to be determined and are the subject of ongoing research. This research has considerable potential importance for land use management—if effects of change are significant, then conversely, improvements in practice could be used to reduce runoff generation and downstream flood risk. An obvious example where agricultural land management practice could reduce flood risk is the restoration of agricultural floodplain land to active use in flood storage.
We conclude that flooding is a complex issue, which involves the prediction of extreme precipitation under climate change, the translation of precipitation into river flow, which must represent effects of land use and land management, and the social and economic issues that determine flood risk to people and property and inform the political process that determines flood protection policy. The UK is relatively well advanced in consideration of these issues, but there is little room for complacency. Five million people in the UK are currently at risk from flooding and flood risk is expected to increase. Flood cannot be prevented, but flood risk can be managed. Effective, equitable and appropriate management of flood risk remains one of the most important challenges for society.
One contribution of 20 to a Discussion Meeting Issue ‘Extreme natural hazards’.
- © 2006 The Royal Society