CHAPTER 5 Flood Risk Assessment and Management In

CHAPTER 5

Flood risk assessment and management in France: the case of Mediterranean basins

F. Vinet Department of Geography, University of Montpellier III, France.

Abstract

For several years, Europe has been subject to huge fl oods: central Europe in 2002, Eastern Europe and especially Romania in 2005 and 2007. Southern Europe, considered as countries around the Mediterranean basin, is prone to severe fl ash fl oods that regularly kill people and cause heavy damage. We can remember the fl ood of Biescas (Spain) in 1996 [1]. Southern France was hit in 1999, 2002 and 2003. The toll of these fl oods was 70 fatalities and 3.109 Euros. The consequences of torrential fl oods in Mediterranean regions are caused by heavy rainfall and topo- graphical conditions. Moreover, population and economic activities are growing around the Mediterranean basin, concentrating vulnerability as regards the fl ood risk. As a consequence of the increasing damage of fl oods, a European Directive [2] was drawn up in October 2007 to implement a policy of fl ood-risk assessment and prevention in the European Union. Member States were invited to assess and to map fl ood risk in order to implement fl ood-risk management plans. Flood management has become a relevant issue for many European countries. Most of them have been reviewing their policy and some of them are eager to launch real actions in this fi eld. In France, over the past 15 years, the increasing damage cost has caused the government to react by various laws and incentive measures. The Ministry for Ecology, Energy, Sustainable Development and Spatial Planning (MEEDDAT) wishes to develop integrated risk management but it is faced with diffi culties in the fi eld. Through the case of Mediterranean regions, this chapter proposes a fl ood risk assessment and an overview of fl ood prevention in southern France focusing on the implementation conditions for this prevention on a local scale. We will fi rst recall the characteristics of fl ash fl oods in the French Mediter- ranean zone. Then we will describe the damage caused by those recurrent fl oods, focusing on fatalities. Then we will see how stakeholders and political actors are

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) doi:10.2495/978-1-84564-560-1/05 106 Flood Prevention and Remediation trying to change the ﬂ ood prevention approach. This approach is shifting from ﬂ ood control through engineering works to an integrated management of risk, including the reduction of material and social vulnerability.

5.1 Characterization of ﬂ ash ﬂ ood hazard in Mediterranean basins

The Mediterranean area is getting used to spectacular fl ash fl oods already described in many papers and books [3, 4]. Flash fl oods are admitted to be danger- ous for human life and diffi cult to reduce. They convey strong specifi c peak fl ow and big volumes of water which, added to the sedimentary load and debris, confer on them a strong destructive power. Before considering the damage, let us analyze the conditions of fl ooding that partly determine the fl ood prevention approach in small Mediterranean basins (Fig. 5.1).

5.1.1 Heavy rainfall

The exceptional effi ciency of torrential fl oods is fi rst due to the characteristics of the rainfall, which are the most intense besides tropical regions. Southern France

Figure 5.1: Location of coastal watersheds along the French Mediterranean coast.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) Flood Risk Assessment and Management in France 107 and Corsica, as Liguria or Catalonia [5], are known for their huge rainfall events. Historical records come close to 1000 mm in 24 h: the fl ood of October 1940 in the Pyrenees resulted in nearly 2000 mm in 5 days [6]. Those heavy rains are due to mesoscale convective systems [7–9]. Two recent events (12–13 November 1999 in the Aude department and 8–9 September in the Gard department) are well-known and have been analyzed in-depth [10–12]. Dur- ing both events (Table 5.1), the total runoff volume was about 2000 million m3 over the line of 200 mm. The area which received more than 200 mm was about 5000 km2 in both cases. In 2002, in the Vidourle basin, 390 million m³ of water fell, whereas 216 million m³ ran off for a watershed of 790 km². The volume of runoff water is highly depen- dent on the saturation of the ground and the karsts before the pluvial episode [13]. If the rain falls at the start of the rainy season, the ground and the karst may store 150 mm. The role of this retention is still the object of many research projects. In case of rainfall of 200 mm over 1 km2, the precipitated volume is 200,000 m³. It is important to bear in mind these quantities to keep in perspective the effi ciency of the water retention upstream. Concerning peak discharges, they exceed several thousand m³ s−1 on the watersheds of less than 1000 km². The peak discharge of the Vidourle reached 2500 m³ s−1 in September 2002, which is more than the average discharge of the Rhine in the vicinity of its mouth. The propagation of fl ood waves is fast and, depending on the rivers and the place of heavy rainfall, the fl ood wave arrives in the lowlands 10–24 h after the peak rainfall episode upstream the watershed. In southern France, these rainfall events generally take place in autumn. According to the data of Météo-France, two-thirds of the rain episodes over 190 mm in 24 h took place from September to November and 50% between 15 September and 15 November (Fig. 5.2).

Table 5.1: Rainfall intensity during two major ﬂ ash ﬂ oods in Southern France. Time Générargues Lézignan-Corbières (8 and 9 Sept. 2002) (12 and 13 Nov. 1999) 48 h 620.2 mm 24 h 684.2 mm 551.2 6 h 344 272.2 4 h 209 234 2 h 168 203 1 h 112 112.2 30 min 77.4 (155 mm/h) 61.2 (122.4 mm/h) 6 min 18.8 (188 mm/h) 15.8 (158 mm/h)

Source: Météo-France [14].

50 45 40 35 30 25 20 15 Number of events 10 5 0 J F M A M J Jt A S O N D

Figure 5.2: Monthly distribution of rainfall over 190 mm per 24 h (1958–1994) in southern France (in three regions: Languedoc-Roussillon, Provence- Alpes-Cote-d’Azur and Corsica). Source: Météo-France [14].

5.1.2 The characteristics of ﬂ ooded regions

Mediterranean watersheds are rather small, between 2000 and 10,000 km² wide. Consequently, geo-hydrological conditions are quite similar and linked to topo- graphical conditions. In upper basins, runoff begins on hills of altitudes from 500 to 1500 m. Deep valleys, steep slopes and heavy rainfalls account for the fact that in small watersheds (less than 20 km2), the peak discharges exceeded 20 m3 s−1 km2 at the core of the fl ooded zone in 2002 [13]. Hills are covered with deep brush that protects slopes from landslides. Such upper valleys are sparsely populated (Photograph 5.1). Downhill, the valleys become larger and the fl ood-prone zone is 500 to 1000 m wide. Here, there are vineyard-covered foothills where the population density is higher (50–100 habitants per km2). Going downstream, the rivers run into low plains. The water runs more slowly (less than 1 m s−1) between dikes and the peak discharge usually arrives several hours, sometimes 1 day, after having affected the high parts of the watershed. The plains are wide and density is high (over 150 inhabitants per km2) due to the presence of many towns and tourist activities. Table 5.2 shows the specifi c characteristics of Mediterranean fl ash fl oods.

5.1.3 The concentration of risk in coastal plains

The topography of the lowlands is well-known on the coastal plains around the Mediterranean Sea [15]. The lowlands are ﬂ oodable areas (over 2500 km2 in Languedoc-Roussillon) that are heavily urbanized with a high concentration of transport infrastructures, tourist facilities and huge inhabited areas. Population densities exceed 150 inhabitants per km². In 1999, 2002 and 2003, many dike

Photograph 5.1: Road and networks damaged in the upper valley of the Berre river (F. Vinet, 1999).

Table 5.2: The characteristics of Mediterranean fl ash fl oods. Characteristics Example Consequences for fl ood prevention Unpredictability No hydrological Hydrological forecast forecast is not possible more suffi cient. Need than 3 days for rain forecast. before the event. In huge events several secondary peaks of intense rainfall are possible. Short duration Duration of the event < 48 h Low spatial Small Several 100 km2 Spatial extension watersheds. for small events concentration and 10,000 km2 of damage. for major events. continued

Table 5.2: Continued. Strong energy High speed of Stream over 8 stream due to m s−1 during rainfall fl oods. intensity and slopes. Speed of Fatalities are propagation of rather frequent the fl ood wave (average of up to 30 km h−1 10 per year in southern France). High discharge Peak discharge Peak discharge higher. up to 20 m3 s−1 km2 on small watersheds. High water Volume of rainfall 220 million m3 Reduce the volume is high even in ran off during effi ciency of small watersheds; the Vidourle upstream 200 mm on fl ood in Sept. storage basins. a km2 is 2002 (790 km2). 200,000 m3. Debris and Waters are loaded Debris and branch sediments with numerous jams reduced fl ow debris (branches, conveyance in the cars) and cross-section area. sediments Siltation. (stones, silt). breaches fl ooded hundreds of km² and directly killed at least 10 people. Reducing the risk in these areas is a major prevention issue in Languedoc-Roussillon.

5.1.3.1 A complex hydrography Lowlands are natural watershed containers that stretch from the border mountains towards the Mediterranean. Since the stabilization of the coast line in around 6000 B.P., they have been made up of alluviums that settled during fl oods, by rivers such as the Aude, Hérault or Vidourle. Therefore, fl ooding is part of the normal functioning of the hydrosystem. Moreover, in the past, man has attempted to promote the natural siltation in order to obtain farmland from pond areas. Hydraulic functioning is made complex by the low level of the altitudes, successive diffl uences and the accumulation of anthropogenic interventions. Man has tried to stabilize the river channels. Lot of them (Aude, Agly, Vidourle, Gard, Lez, not to mention the Rhône) are dammed along all or part of their watercourse. Stabilizing

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) Flood Risk Assessment and Management in France 111 the channels, promoted by aggradation, has made it easier for inhabitants to re- appropriate river valleys. Dike system contains rivers for fl ow rates with very vari- able return periods, stretching from 2 years for the Aude river to 40 years for the Agly river. When the fl ow rate exceeds the hydraulic capacity of the river channel, the water invades the plain either by spillways or dike failure. This overfl owing is promoted by the ‘roof layout’ of the lowlands.

5.1.3.2 The ‘roof layout’ of the lowlands The roof layout of the lowlands (Fig. 5.3) results elevating the river banks which thus dominate the low points by a few metres sometimes situated several kilome- tres from the river channel. This functioning explains why many unaware people (new inhabitants, foreigners) are surprised by the arrival of water in a fl ood period because the most frequently and seriously fl ooded areas are not located near the river. Traditional villages are located on the highest sites, either on the natural emi- nences of the plain, or on the natural levees on the banks behind the dikes, or in contact between the fl oodplain and the surrounding hills. The anthropogenic controlling of fl oods is made by weirs. Moreover, near the sea, the fl ood risk can be worsened by the growing sea level and beach barrier erosion. According the MEEDDAT, 3000 km2 in Languedoc-Roussillon are exposed to fl ooding, i.e. 11% of the regional surface area but 25% of the regional population (600,000 people) is living in those fl ood-prone zones. So, Mediterranean coastal areas are clearly at risk (Fig. 5.4). The impact of fl oods on society and livelihoods is strong.

Figure 5.3: The ‘roof layout’ of the lowlands.

Figure 5.4: Flood-prone zones in the Languedoc-Roussillon region.

5.2 The vulnerability of the society faced with the ﬂ ood risk

5.2.1 The increasing cost of ﬂ oods

France is a country that is regularly hard-hit by fl oods. Two general kinds of fl oods can by identifi ed: low-rise fl uvial fl oods and fl ash fl oods in mountainous and Mediterranean regions. The fi rst kind of fl ood mainly concerns northern France, although local urban areas and small basins can be hit by small fl ash fl oods [16]. Flooded regions can be large but damage is limited by the low speed of fl ow. Fatalities are exceptional. Flash fl oods mainly concern the Alps and Pyrenees and Mediterranean regions including Corsica. Even if watersheds are small, powerful streams cause extensive damage and fatalities. Figure 5.5 shows the amount of compensations paid out by insurers for the major fl oods in France over the past 20 years. Even though only 20% of the territory is subject to a Mediterranean climate, the fl oods in this region account for 66% of the cost of damage to property in France. In 1988, the city of Nimes suffered widespread fl ooding, and 10 people died. In 1992, 42 people died during fl ash fl oods that particularly affected Vaison- la-Romaine. The two determining fl ash fl oods occurred in 1999 and 2002. The fi rst event took place in the Aude department and nearby departments on 12 and 13 November 1999. Thirty-fi ve people died, 35,000 houses were fl ooded and the total amount of losses was evaluated up to 650 M€. Less than 3 years later, on 8 and 9 September 2002, the Gard department suffered 23 fatalities and losses exceeding

Figure 5.5: The cost of recent ﬂ oods in France.

1 billion Euros; 640 M€ was dealt with by insurers [17]. In 2003, fl oods in the Rhone basin which were made up of fl ash fl oods and dike failure obliged insurers to pay out 744 M€ in compensation. The MEEDDAT, in charge of natural hazard prevention, estimates that the average annual cost of fl oods is 150 M€ around Mediterranean costal regions.

5.2.2 Fatalities due to ﬂ oods

As well as material damage, Mediterranean fl ash fl oods are known to cause many victims. The International Disaster Database EM-DAT edited by the Centre for Research on the Epidemiology of Disasters (CRED) at the University of Louvain (http://www.em-dat.net/) provides a certain number of examples: fl oods in Alger (Bab-el-Oued) in November 2001, or the 500 victims due to fl ooding in the south of Spain in October 1973, the 750 victims of the Ourika and the neighbouring valleys in Morocco in 1995 …. In France, even if the South-West and the Alps have not been spared, the heaviest human toll is seen in the South-East Mediterranean area. Over 150 people have died in the 15 years since 1992 [18, 19]. However, paradoxically, although the safety of people has been a priority in the prevention of natural risks, there are few general studies based on the databases linking prevention and mortality. The fi rst step towards taking into account mortality in the prevention of fl oods is the death census. Our hypothesis is that in-depth analysis

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) 114 Flood Prevention and Remediation of the circumstances and place of death helps to better deﬁ ne the vulnerability of people faced with ﬂ oods and to better target prevention.

5.2.2.1 Creating a geo-referenced database There are international databases that hold information on natural disasters and the consequences of them. The International Disaster Database EM-DAT is the most well-known one [20]. For France, the CRED database holds information on the 38 fl oods since the end of the 1970s. However, the location of victims is quite imprecise considering the scale of the work. In the USA, the National Weather Service feeds a database covering deaths due to fl ooding [21, 22] but this process is not made general on an international level. This absence is surprising, taking into account the value of better knowing the circumstances of deaths due to fl ooding [23] and the increase in research on the behaviour of disaster victims in case of crisis or on the vulnerability of people exposed to fl ooding [24, 25]. In France, there is a database on the victims of avalanches that can be consulted, in part, on the website http://www.anena.org. This database is fed by the collection of a datasheet after each avalanche. However, although the amount of databases is increasing and there are all sorts of indicators, there is no census and systematic analysis of deaths due to fl ooding in France. Everything would seem to indicate that death due to fl ooding is residual and that the epidemiological study of it is not worth paying attention to. A summarizing study has already been carried out by Antoine et al. [19] on the victims of fl ooding in Languedoc-Roussillon with a historical perspective. The authors listed the victims mentioned by historical sources, from the origin of them to 1999, with relatively good reliability from 1800. The victims are generally located on the commune level. Since 1999, this database has been updated by geo-referencing each death. The information gathered in the press has been con- fi rmed by an enquiry with municipalities and police. We are more interested in the circumstances than the clinical causes of the deaths.

5.2.2.2 Longitudinal analysis of deaths due to fl ooding It is clear that deaths due to natural disasters have dropped in rich countries since the beginning of the 20th century [22]. However, in USA, since 1975, the annual number of victims of fl ooding has levelled off at about 100 per year. Has the progress related to the improvement of forecasting and emergency services been offset by an increase in exposure, in particular due to building in fl oodable areas and traffi c movement? The development of mortality is subject to the occurrence of major disasters including the one in St. Chinian on 12 September 1875 (90 and 125 victims depending on the source), and the one in Bordezac that resulted in 100 victims in 1861. The latter was due to the submergence of a coal mine. However, we would like to point out the reduction in the results of major disasters that marked the 19th century and the fi rst half of the 20th century (350 deaths in Catalonia and Pyrénées-Orientales in 1940). The second half of the 20th century was relatively spared. Antoine et al. [19] did not mention any victims between 1960 and 1970,

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) Flood Risk Assessment and Management in France 115 and only six between the serious fl oods in the Gard in 1958 and the 11 victims in Nimes in 1988. After some decades without a signifi cant event, many severe fl ash fl oods have occurred since the mid-1980s. Since 1988, a certain number of lethal fl oods, mentioned above, have made the number of victims increase (Fig. 5.6).

5.2.2.3 Future developments For the fl ooding risk, the work of Jonkman [26] shows the high correlation between socio-economic development and mortality. In so-called developed countries, we can see a decrease in the number of deaths. However, this trend is not necessarily irreversible, as we saw in 2005, with some 1800 victims of cyclone in Katrina in the USA. Hypotheses about global warming and its consequences give rise to fears concerning the increase in intense rainfall events, although many uncertainties still weigh on the pluviometric response to the global warming. The fi rst IPCC reports [27] suggested that mortality due to natural disasters could increase with global warming, even in Europe, but the fourth report of the IPCC [28, 29] states that ‘to date, no research has been reported that quantifi es the impact on mortality and morbidity risks’ (Section 5.3.1). But apart from simple global warming, an increase in mortality due to fl ooding is not ruled out, if we consider the simple increase in the population, the greater

160

1875 St-Chinian (34) 140

120 1861 Bordezac (Gard) 100

80 1940 Pyr.-Or. 1958 Gard 60

Number of death 2002 Gard 1999 Aude 40

0 1850 1870 1900 1970 1800 1810 1820 1830 1840 1860 1880 1890 1910 1930 1960 1980 1920 1940 1950 1990 2000

Figure 5.6: Chronicle of deaths due to ﬂ ooding in the Languedoc-Roussillon region (Antoine et al. [19], Vinet [10]).

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) 116 Flood Prevention and Remediation exposure of people to risk and the low information of neo-residents quite unaware of the Mediterranean hydrological phenomena.

5.2.3 The recent increase in assets and stakes in ﬂ ood prone zones

5.2.3.1 The spatio-temporal development of vulnerabilities: the migration of areas of vulnerability towards the plains. Of course, the geography of victims is linked to the location of rain cells. However, if forecasting progresses, if the solidity of dwellings is improved, new vulnerabilities appear, linked, in particular, to the occupation of ﬂ oodable areas. What is important is to note the movement of victims towards downstream areas, in accordance with the historical development of the population over two centuries. The major disasters of the 19th century mentioned above took place in the high valleys (Fig. 5.7). The huge number of victims could be explained by the presence of a former habitat in the encased valleys related to activities near to water: textile, mining industry, thermal tourism. The recent disasters took place in the foothills (Aude in 1999, Gard in 2002). The attraction, due to transport and tourism activity, has increased activity and property on the coastal plains. During the last century, the rural exodus pushed people to move seaward (Figs 5.8 and 5.9).

Figure 5.7: Fatalities due to ﬂ oods in the Languedoc-Roussillon region since 1800 (Antoine et al. [19], Vinet [10]).

Figure 5.8: The population of Languedoc-Roussillon region in 1881.

Figure 5.9: The population of Languedoc-Roussillon region in 2005.

5.2.3.2 Increasing vulnerability A consequence of this migration to the plains is that the coastal fl ood-prone areas are three times more densely populated than the non-fl oodable areas (Table 5.3). Many studies have proved the increasing amount of housing and industries in fl ood-prone zones [30]. This is the result of the uncontrolled urbanization which followed WWII. This extension of urban areas was particularly acute between 1975 and 1995 when the model was individual housing that took up a lot of space. Figure 5.10 shows the example of a small town called ‘Cuxac-d’Aude’ near Narbonne. Nearly all the territory of the community is prone to fl ooding. Neverthe- less, for 30 years, house building did not stop while the population rose from 2490 to 4270 between 1975 and 1999. There are many such examples and the case of this community is not so rare. As exposure has grown, some changes in the occupation of houses (especially the ground fl oor) have increased vulnerability. As we can see in Photograph 5.2, the use of the ground fl oor of this house has changed from a garage to housing. Thus, people are now exposed instead of mere material goods such as cars or tools.

5.2.3.3 Increase in risks and global warming Even if the IPCC report [28, 29] foresees an increase in heavy rain in connection with global warming, the longitudinal study of rainfalls [31, 32] and river fl ows data [33] do not show any signifi cant trend. Scientists who are studying climate change on the Mediterranean area are very careful about the consequences of warming on heavy precipitation. The increases detected, when there are some, are not certain ones. In coastal areas, Ullmann and Moron [34] have shown that there is no trend in the increase of storms. Nevertheless, the rising of the sea level associated with sand ridge erosion is more alarming. Beyond simple global warming, considerations should focus on risk development scenarios and maximum crisis scenarios integrating, into the coastal lowlands, a concomitance of fl ash fl oods and fl ooding by marine submersion, not forgetting possible complications by the occurrence of technological incidents such as pollu- tion. Prospective studies should focus on natural aspects and also future social mutations. For example, it is necessary to examine a possible increase in vulnerability by a mutation in tourist practices. Global warming could bring about a lengthening of the tourist period that would exceed the traditional dry summer

Table 5.3: Compared population density in the Languedoc-Roussillon Region. Flood-prone zone Outside Languedoc- ﬂ ood-prone zone Roussillon Region Surface area 3000 km2 24,447 km2 27,447 km2 Inhabitants 585,000 1,912,000 2,497,000 Population density 195 inhabitants 78 inhabitants 91 inhabitants (per km2)

Source: Insee, Diren-LR 2005.

Figure 5.10: The growth of the population in ﬂ ood-prone zones: the example of Cuxac-d’Aude between 1976 and 2001. (A) The commune of ‘Cuxac-d’Aude’ in 1976. (B) The commune of ‘Cuxac-d’Aude’ in 2001.

Photograph 5.2: The increase in vulnerability: a garage has been turned into living space. months (July and August) towards the autumn months that are exposed to heavy rain (Fig. 5.2). This increasing exposure to ﬂ oods is already at work by the settling of populations in mobile homes or precarious shelter due to the increase in the price of rent and property. It has therefore happened that in many regions, structural vulnerability has sharply increased. Undoubtedly, it contributes more to the increase in risk than the simple and hypothetical increase in intense rainfalls due to climate change. The increase in vulnerability that cannot only be seen in Mediterranean regions calls for efﬁ cient actions to reduce the future potential damage of disasters.

5.3 Changes in ﬂ ood management frameworks

Indeed, the increase in damage has highlighted the failure of the current fl ood management policy based on river channel management instead of the mitigation of the fl ood risk. Since the last disasters mentioned above, national and local authorities have increased initiatives to effi ciently reduce risks and modify the way of managing fl ood risk. Even though national authorities are aware that the change is useful, implementing the new guidelines is complicated on the local level.

5.3.1 The need for a changing framework in ﬂ ood prevention

The failure of 20th-century ﬂ ood management policies has been observed not only in France, but also in other European countries, and in the USA. Parker [35] has

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) Flood Risk Assessment and Management in France 121 denounced the ‘escalator effect’ induced by the growing demand for protection responding to the accumulation of material wealth in ‘protected’ fl oodplains. As Werritty [36] states, ‘A seismic shift is taking place in managing fl ood risk in many countries,’ but if current fl ood management frameworks are questioned, the ‘seismic shift’ is slow. There has been a single-minded focus on structural measures. International organizations admit now that one must shift from defensive action against hazards to an integrated management of the risk and alternative methods of risk mitigation [37]. In France, many recent fl ood reports had revealed the failure of fl ood management policies [11, 38, 39]. Two major aspects had been criticized: fl ood warning and dike breaks. After the Gard fl ash fl oods in September 2002, many citizens criticized the mayors (the communes’ elected authorities) before a tribunal, and several mayors criticized in turn the state authorities for lack of warning. As a result, the French government has been reorganizing fl ood warning and crisis management. Another controversial issue has been dike and levee management. The problem of dike maintenance is unfortunately well-known in many countries, e.g. in Germany and in the USA after Hurricane Katrina. Sometimes, fl oods overfl ow dikes, but in many cases failure is due to bad maintenance. Dike owners are not always known, and even when they are identifi ed (e.g. municipalities or private riverside owners), they may have insuffi cient resources to repair the dikes. Another consequence of these fl oods is the demand from citizens for more protection against fl oods. In the Vidourle basin, almost all the communes have been declared to be in a state of ‘natural disaster’ according to governmental decrees. Thus, the fl ooding has reinforced awareness of the risk but also increased the pres- sure on the authorities. People want authorities to protect them from fl oods, and with defi nite and visible results, as soon as possible.

5.3.2 Reforms in the 2000s

The French government has issued a certain number of laws and made some decisions since the fl oods of 2002. Land use control has been reinforced through Prevention of Predictable Natural Hazards Plans (PPNHP) that date from 1995 [40]. Their application is stricter, even if this land occupation control has evoked opposition from local populations. The law dated 31 July, 2003 (‘risks’ law) reinforced information for people living in areas at risk. The French government also modifi ed the fl ood forecasting system in 2006 [10] and reorganized the emergency services by the law dated 13 August 2004. It obliged the communes subject to a PPNHP to have a local emergency management plan. The French tradition of the providence and centralizing State could explain why, for two centuries, the State has been practically the sole manager of risk prevention and recognized as such. However, following the latest disasters, the State admitted that it could no longer be the only guarantor in the fi ght against risks and that local authorities should become involved. Therefore, it has tried to fi nd support from watershed authorities. The aim of the call for project on 1 October 2002, was to reinforce the role of the watershed structures.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) 122 Flood Prevention and Remediation 5.4 Reducing the ﬂ ood risk by watershed structures

5.4.1 The key role of watershed structures

Watershed structures in France are old. Groupings have existed since the 19th century and they have various skills: fi shing, heritage valorisation, water supply, etc. How- ever, these structures were not very effi cient. Until the end of the 1990s, apart from some rare exceptions such as the Vidourle basin authority created in 1989, watershed structures had few technical and fi nancial resources. They lacked the competencies and the political strength to implement fl ood management programmes. As regards fl ooding, each commune managed the fl ood risk as well as it could with few means and without any coherence on the scale of the entire watershed unit. Serious fl ooding obliged authorities to reinforce watershed structures in order to be able to act in a coherent way. We have seen a shift from local risk management at municipal- ity level to a linear vision of the river, with a concern for upstream–downstream coherence, then to the notion of watershed solidarity (Fig. 5.11). The French state offers grants for fl ood prevention if there is a strong basin authority covering the whole catchment.

5.4.2 The ﬂ ood management plans

The MEEDDAT has decided to encourage local river basin authorities to mitigate fl ood risks by launching a call for fl ood management plans (FMP called PAPI in French). The circular of 1 October 2002, that fi xed the characteristics of these plans, was adopted three weeks after the fl oods of the Gard in 9 September 2002. The projects aim to fulfi l four conditions:

• A high level of risk regarding the basin in question • The validity of the plan taking into consideration the above criterion • The credibility of the river basin authorities • An integrated management of ﬂ oods at the catchment level.

Figure 5.11: Three steps of ﬂ ood management in river basins.

The circular stipulates the objectives of these plans: (1) to develop public awareness; (2) to create or to rehabilitate fl oodplain storage areas upstream in order to maintain solidarity between fl ood plains (often struck by fl oods) and upstream watersheds; (3) to reduce vulnerability and protect urban areas. Between 2002 and 2004, river basin authorities throughout France elaborated FMPs. The chosen plans are subsidised 25–40% by MEEDDAT. Further subsidies are provided by regional and local authorities and by the river basin authorities themselves. This call for projects reveals the changing attitude of the French government regarding risks. For many years, state authorities and communities had been the major stakeholders in risk prevention. This pattern, however, was not effi cient, because communities dealt only with their own territory, without coherence in the watershed as a whole. Despite reassurances, state authorities lack the technical and political means to control everything in the fi eld of risk management. The aim of these FMP is to empower river basin authorities, as they have an evident stake in the issue of fl ood management. The principle is to coordinate a hydrological unit (the watershed) with a political and technical unit, i.e. the river basin structure. The political entity is linked to the natural area. The aim of the state authorities is to assure the long-term maintenance of structures, especially dikes. As a result, the French state has decided to stop paying for fl ood defence unless it is handled by river basin authorities.

5.4.3 An integrated approach to ﬂ ood prevention

FMPs must favour a holistic approach to fl ood prevention and must be established in integrated patterns: spatial and technical integration. Spatial integration means that the whole basin has to be affected by fl ood mitigation measures. Technical integration means that fl ood reduction may be achieved by either structural measures (dikes, fl ood storage areas) or non-structural measures (reduction of building vulnerability, development of public awareness of risk, crisis management). Spatially, the call for project recommends reinforcing upstream–downstream solidarity by promoting the ‘Dynamic Flood Slowing Down Strategy’, i.e. the storage of water in the basin upstream in order to slow down and to decrease the peak discharge. The technical integration of the projects supposes the association of traditional structural methods and measures (dams, fl ood storage basins), non-structural measures such as information disseminated to fl ood prone populations, the reduction of vulnerability and land-use control. Here are the fi ve types of measures:

Issue 1: public information and awareness Issue 2: forecast and warning, crisis management Issue 3: vulnerability mitigation, and relocation Issue 4: upstream ﬂ ood control Issue 5: collective protection (dikes and ﬂ oodwalls)

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) 124 Flood Prevention and Remediation 5.5 Success and diffi culties in the fi ght against fl ooding

Although the intentions of the FMPs are to develop an integrated and pluralistic approach to the prevention of fl ooding, the important thing is to note that the plans are largely focused on managing hazards. Apart from crisis preparation and management, the development of non-structural measures is still diffi cult and structural measures take up most of the subsidies for fl ood prevention.

5.5.1 Prevention too focused on structural measures

The fi nancial distribution of fl ood reduction measures in FMP shows that structural measures still prevail in fl ood prevention. Of course, we cannot assess all river basin authorities’ policies on such plans. Nevertheless, the distribution of subsidies by item is signifi cant as regards policy choices, especially between structural and non-structural measures (Table 5.4: watersheds cited can be located in Fig. 5.1). These plans are clearly involved in structural measures. Dike maintenance and building account for 40% of investments and upstream fl ood storage areas for 45%. Therefore, actions on hazard control amount to 85% of fi nancial resources, i.e. 166 of 195 M€. Forecast and warning may be considered apart, because this fi eld depends on State core competencies. Reduction in vulnerability is hardly considered (between 0.1% and 5% of FMP subsidies) except in the Gardons project, where this item is gratifi ed with 8.6 M€ (22%). In fact, the fi rst achievements of FMPs are structural protections already planned before the acceptance of the plans.

5.5.2 Technical and methodological challenges

Unless river basin structures are technically competent, some diffi culties appear when implementing defi nite measures to reduce fl ood risk. The fi rst one is lack of data about the fl ood hazard. Small watersheds have not undergone any fl ooding for a long time, and do not have any stream gauging network. A more profound challenge is the lack of vulnerability assessment: vulnerability of housing, industries and fl ood-defence systems. The Tech and Aude valleys need to check their old embankment systems in the low plains. The heaviest problem is to defi ne the ownership of dikes because, after the 1930 and 1940 fl oods, many dikes were built on private fi elds with public subsidies. River basin authorities have also imple- mented some studies to assess the vulnerability of housing and industries. Such studies have been undertaken for a long time in other countries, but they are more recent in France, having been initiated in the Loire basin. These studies are also an opportunity to make people aware of vulnerability, which is often not known. For example, they have revealed the lack of know-how for campsite emergency plans in the most tourist region of France.

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) Flood Risk Assessment and Management in France 125 Euros) 6 Total (in 10 Total Aude basin (2006–2009) ood management plans. Touloubre basin Touloubre (end 2003–end 2008) ood prevention in fl ood prevention Orb basin (2003–2010) Tech basin Tech (2005–2008) basin Gardons Table 5.4: Planned investments for fl 5.4: Planned investments Table 0.8451.14 0.926.6 2.61 0.399.97 8.60 0.22 16.95 0.3 0.66 2.29 0.7 0.04 0.7 5.9 0.282 0.03 0.72 11.22 1.35 3.25 0.8 41.43 6.28 87.8 17.3 basin 10.57 9.80 10.45 11.9 0 35.21 77.9 Vidourle Vidourle (2003–2006) ood oodwalls) Item Programme chargesPublic information 0.72and awareness and Forecast warning 0.00 Vulnerability mitigation Upstream fl 0.14control Collective protection and (dikes 0.5fl Total 0.312 29.85 0.49 39 2.16 14.16 20 11.88 80 195

5.5.3 Legal and ﬁ nancial uncertainty

The legal status of structures is complicated. Existing structures such as dikes or fl oodwalls belong to the stakeholder and not to granter of subsidies. The long-term status of structures (dikes, fl ood storage areas) is not ensured, and this question is suffi ciently uncertain that fi nancial resources are not warranted for the future. FMP are paid by a consortium of local, regional and national authorities (department councils, regional councils, MEEDDAT and river basin authorities) and the fi nancial backing for maintenance of the structures is unsure. River basin authorities do not have their own tax income. They depend on subsidies.

5.5.4 Political and psychological oppositions

The third kind of diffi culty is of a political nature. The fi rst obstacle is opposition within river basin authorities (led by local political fi gures) or between different regional authorities. This situation results in diffi culties in implementing a coherent fl ood management policy. Among the various diffi culties, psycho-sociological preconceptions are not the easiest to overcome. Mental perceptions of fl oods are deeply rooted in people’s minds. Surveys show that many people believe that keeping a clean rural system of drainage could reduce fl oods, whereas the cost–benefi t ratio of such maintenance is not often interesting. Moreover, the cost–benefi t approach has not been put into practise in France although many references exist and many researches have been undertaken on this issue [41]. When the MEEDDAT called for FMP projects in 2002, the cost-benefi t analysis was not clearly required, although many plans contained an economic aspect. Diffi culties in assessing long-term costs and benefi ts in fl ood management are certainly a handicap, but France suffers from a poor economic culture and the domination of political factors in decision-making. For example, the cost-benefi t ratio problem of upstream fl ood storage basins has never been seriously addressed regarding small Mediterranean basins. While the hydrological benefi ts are small for huge fl oods, the cost of such basins may be assessed not only in fi nancial terms (the cost of building and maintenance) but also in aesthetic terms in a tourist region. Leading authors [24] now think that both land-use planning and warning systems are the most effi cient measures to mitigate fl ash fl oods. The French government and some basin authority think that the reduction in vulnerability is a good way to mitigate risk and to help people to cope with fl oods.

5.5.5 Is reducing vulnerability the new ‘frontier’ for ﬂ ood reduction? The case of reducing building vulnerability

The term ‘vulnerability’ is quite widely accepted. It can be called structural and, in this case, describes the propensity to damage of a building or a structure. It can also be human or social and is, in this case, applied to the abilities of individu- als and social groups to resist and be resilient to risk [42]. Recognizing human

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) Flood Risk Assessment and Management in France 127 responsibility in risk production means taking action as regards the vulnerability of the exposed issues to reduce the potential consequences of a disaster. Faced with uncertainty and the cost of structural measures, the authorities in charge of risk prevention are looking for other approaches. National and local authorities are trying to implement new measures to reduce vulnerability. Let us take as an example the structural vulnerability reduction measures for buildings. For a few years now, the French government has been trying to reduce the vulnerability of buildings as regards fl ooding. The techniques are quite well-known [37] but the implementation of them is often diffi cult. In Europe, there have been several experiments. In France, some measures have been suggested or imposed in PPNHP. They have fi ve aims:

• to reduce ‘encroachment’ by changing the fl oodway capacity • to develop fl oodplain acquisitions (demolition of properties in risk areas) • to promote fl ood proofi ng of buildings • to adapt public buildings and private properties in case of emergency (facilitate evacuation …) • to facilitate the recovery process after fl oods

As regards waterproofi ng, the objective is to modify buildings in order to protect them from submersion. When submersion is inevitable in case of fl ooding, it is necessary to adapt the building to this submersion so as to reduce the impact of the fl ood and to promote a return to normal. Vulnerability reduction measures should be adapted to each type of fl ood in particular to the speed of the fl ow and the turbidity of the water, which are the characteristics of Mediterranean fl ash fl oods. The ministerial circular dated 21 January 2004 listed a certain number of these measures supposed to ensure the safety of people (refuge fl oor and access to eaves, opening in the roof, access footwalk in case of fl ood, etc.) or to avoid damage to goods (fl ood shields, hoisting up, stoplogs, anchoring certain equipment to the ground, placing electrical supply circuits out of the water, etc.). It stipulates that ‘such measures should be integrated into the PPNHP and made obligatory’. The Ministry of Public Works has written guides and many experiments to reduce vulnerability have taken place over the last few years [43]. There are many obstacles to applying vulnerability reduction. The fi rst problems are fi rst of all of a technical and architectural nature. What measures should be made obligatory? Is it necessary to choose depending on the water height? Do we always have suffi ciently precise hazard data to impose good measures? Moreover, the recommended measures are not always compatible with the existing regulations. For example, making an opening in roofs to allow people trapped in their dwellings by the rising water to escape clashes with the existing regulations in sectors that are classed as historical monuments. Not all buildings can be modifi ed. For example, not all the dwellings in a fl oodable area can have a refuge-level built for the safety of people. Experience shows that it is not enough to impose measures in the PPNHP for them to be applied. Therefore, it is necessary to replace the top-down approach by

WIT Transactions on State of the Art in Science and Engineering, Vol 50, © 2011 WIT Press www.witpress.com, ISSN 1755-8336 (on-line) 128 Flood Prevention and Remediation adaptive approach as suggested by Hutter [44]. It is clear that the State does not have the technical and human resources needed to monitor the fi les and a posteriori control of whether the work has been carried out. Insurance companies are not very much involved in reducing the vulnerability of buildings because the French obligatory insurance system for natural disasters is guaranteed by the State. They prefer to concentrate on reducing industrial risks. The problem is also a legal one. The French State is used to be involved in collective protection (dikes, dams, etc.) and not in the individual protection of private homes. A concerted and project-based approach by a local stakeholder is more effi cient for imposing heavy and costly measures. Programmes are planned in the south of France as well as in the Orb basin. The town of Béziers and the Orb basin authority are the works managers. Vulnerability reduction programmes should be well- targeted; enacting general coercive measures in this fi eld has very little sense. A vulnerability diagnosis, at a cost of about 200 Euros, should precede the measures. The fi nancial problem is also a major obstacle, especially in the south of France where property is expensive and where the turnover of owners and tenants is high. How can we impose an increase in expenses that could reach 10% of the item (25,000 Euros for an average dwelling) on often already indebted owners? How do we deal with the problem of second-home owners? Should tenants participate in fi nancing measures since they benefi t from them? If the measures are made obligatory, who fi nances them? It would seem logical to keep a share of self-fi nancing by local authorities and the State. The latter can grant projects by the intermediary of the Fund for the Pre- vention of Major Natural Risks (called Barnier fund). This fund is fi t by a tax on insurance policies and is partly dedicated to subsidies for works to reduce building vulnerability. On the other part, how can authorities involve tenants or owners? Should they operate by tax incentives? Adjustments in insurance premiums? The most pertinent programmes unite actions for reducing vulnerability with urban renewal operations. They are diffi cult to implement because they mobilize several stakeholders but they allow fi nancing to be combined and, in particular, they associate the fi ght against fl ooding with the positive image of habitat improvement. To sum up, the conditions needed for the implementation of a building vulnerability reduction programme are:

i. The involvement of a motivated and powerful local works manager who can rally the technical, human and legal means to carry out the operation. ii. A precise diagnosis of the vulnerability of the existing building should be carried out by an independent research ofﬁ ce. Self-diagnoses are partly possible. iii. Check the compatibility of the recommended measures with other measures such as the protection of heritage, the protection of the environment, etc. iv. Make reducing vulnerability a positive thing by associating it with urban renewal or habitat renovation programmes. v. Mobilize various forms of ﬁ nancing by involving different stakeholders including home owners.

Recent and disastrous fl oods have highlighted the relative failure of the existing risk management system. The analysis of fl ood damage and fatalities focused on the risk production system which has prevailed for three decades. It shows that existing protection measures were not able to prevent the urbanization of fl ood-prone areas. The increase in damage is mainly due to the growth of urbanization in lowlands. Thus, for the future, we think that the evolution of the fl ood risk may mostly depend on social and demographic patterns rather than climate change: growth and change in tourist behaviour, ageing of population, etc. Faced with the growing amount of damage, national authorities have realized that they have to shift risk management towards non-structural methods such as vulnerability reduction. Nevertheless, the change in fl ood management patterns is different depending on the basins. In some basins, it turns out that the ‘hazard paradigm’ [45] still prevails and causes costly expenditure on structural protection. In others, river basin authorities are developing non-structural programmes. The top-down institutional approach to risk prevention generates confl ict and measures that are ill-adapted to the Mediterranean context. However, we do not mean that populations are always lucid in their behaviour and we do not think that local authorities are always right in their decisions for handling fl ood risks. For example, a systematic cost-benefi t approach would be required to avoid unnecessary and costly works. This approach is not used enough in France. At last, fl ood prevention is a perpetual adjustment between general principles and guidelines and local constraints. Constraints are not only natural and technical but also cultural and political. While the new European Directive on Flood Risk is willing to reinforce fl ood risk management in Europe, prevention would benefi t from taking into account the specifi c characteristics of the Mediterranean basins to improve such measures as upstream control, vulnerability reduction or fl ood warning. Stakeholders and policy-makers might implement an adaptive and territorial approach to fl ood prevention. The great challenge will be to consider the characteristics of each region and each watershed to implement appropriate measures.

References

[1] Alcoverro, J., Corominas, J. & Gomez, M., The barranco de Aras ﬂ ood of 7 August 1996 (Biescas, Central Pyrenees, Spain). Engineering Geology, 51, pp. 237–255, 1999. [2] European Community. Directive 2007/60/EC of the European parliament and of the council of 23 October 2007 on the assessment and management of ﬂ ood risks. OJEU L, 288, pp. 27–34, 2007. [3] Belmonte, A.C. & Beltran, F.S., Flood events in Mediterranean ephemeral streams (ramblas) in Valencia region, Spain. Catena, 45, pp. 229–249, 2001.

[4] Gilard, O. & Mesnil, J.-J., La crue de Vaison-la-Romaine du 22 septembre 1992. informations Techniques du CEMAGREF 95, pp. 1–8, 1995. [5] Llasat M.C., Barriendos, M., Barrera., A. & Rigo, T., Floods in Catalonia (NE Spain) since the 14th Century. Climatological and meteorological aspects from historical documentary sources and old instrumental records. Journal of Hydrology, 313(1–2), pp. 32–47. 2005. [6] Soutadé, G., Les inondations d’octobre 1940 dans les Pyrénées-Orientales. Conseil général, Direction des archives départementales, Perpignan, 351 p., 1993. [7] Bechtold P. & Bazile E., The 12–13 November 1999 fl ash fl ood in southern France. Atmospheric Research, 56, pp. 171–189, 2001. [8] Aullo, G., Santurette, P., Ducrocq, V., Jacq, V., Guillemot, F., Sénequier, D., Bourdette, N. & Bessemoulin, P., L’épisode de pluies diluviennes du 12 au 13 novembre 1999 sur le sud de la France. Météo-France, 79 p., 2002 [9] Delrieu, G., Ducrocq, V., Andrieu, H., Gaume, E., Anquetin, S. & Creutin, J.D., The catastrophic fl ash-fl ood event of 8–9 September 2002 in the Gard region, France: a fi rst case study for the Cévennes-Vivarais Mediterranean Hydrometeorological Observatory. Journal of Hydrometeorology, 6(1), pp. 34–52, 2005. [10] Vinet, F., L’épisode pluvieux catastrophique des 12 et 13 novembre 1999 dans l’Aude et les départements voisins: analyse pluviométrique et météorologique. Géocarrefour-Revue de Géographie de Lyon, 75(3), pp. 189–203, 2000. [11] Vinet, F., Crues et inondations dans la France méditerranéenne. Les crues torrentielles des 12 et 13 novembre 1999 (Aude, Tarn, Pyrénées Orientales et Hérault), Editions du Temps, Nantes, 224 p., 2003 [12] Neppel, L. & Desbordes, M., Fréquence de l’épisode pluvieux à l’origine des inondations des 12 et 13 novembre 1999 dans l’Aude. C.R. Acad. Sci. Paris, 332, pp. 297–273, 2001. [13] Gaume, E., Livet, M., Desbordes, M. & Villeneuve, J.P., Hydrological Analysis of the River Aude, France, fl ash fl ood on 12 and 13 November 1999. Journal of Hydrology, 286, pp. 135–154, 2004. [14] Météo-France, Inventaire des situations à précipitations diluviennes sur le Languedoc-Roussillon, la Provence-Alpes-Côte-d’Azur et la Corse, période 1958–1994. Meteo-France, Direction Interrégionale Sud-Est, 190 p., 1996. [15] Segura Beltran, F., Model d’inundacions en ventalls alluvials: el cas de les planes costaneres valencianes. Cuad. de Geogr., 73/74, pp. 207–210, 2003. [16] Douvinet J., Intérêt et limites des données « catnat » pour un inventaire des inondations l’exemple des crues rapides liées aux violents orages. (Bassin Parisien, Nord de la France) Norois, Presses Universitaires de Rennes, 201, pp. 17–30, 2006. [17] Vinet, F., Geographical analysis of damage due to fl ash fl oods in southern France: The cases of 12–13 November 1999 and 8–9 September 2002. Applied Geography, 2008, doi:10.1016/j.apgeog.2008.02.007.

[18] Vinet, F., From hazard reduction to integrated fl ood management: toward adaptive fl ood prevention in Europe, eds D. Proverbs, C.A. Brebbia & E. Penning-Rowsell. Flood Recovery Innovation and Response. FRIAR. WIT Press, pp. 113–122, 2008. [19] Antoine, J.M., Desailly, B. & Gazelle, F., Les Crues meurtrières, du Roussillon aux Cévennes. Annales de Géographie, 622, pp. 597–623, 2001. [20] Hoyois, P. & Guha-Sapir, D., Flood Disasters in Europe: A Short Analysis of EMDAT Data for Years 1985–2004. CRED: Louvain, 6 p., 2005. [21] French, J., Ing R. & Von Allmen, S., Mortality from fl ash fl oods: a review of national weather service reports, 1969–81. Public Health Rep., 98, pp. 584–588, 1983. [22] Pielke, R.A., Flood impacts on society: damaging fl oods as a framework for assessment in Parker D.J. Floods. Routledge: London, vol. 2, pp. 133–155, 2000. [23] Ahern, M., Kovats, R.S., Wilkinson, P., Few, R. & Matthies, F., Global health impacts of fl oods: epidemiologic evidence. Epidemiologic Reviews, 27, pp. 36–46, 2005. [24] Gruntfest E. & Handmer J. eds., Coping with Flash Floods. NATO science series. Kluwer Academic Publishers: Dordrecht, 322 p., 1999. [25] Ruin, I. & Lutoff, C., Vulnérabilité face aux crues rapides et mobilités des populations en temps de crise. La Houille Blanche, 6, pp. 114–119, 2004. [26] Jonkman, S.N., Global perspectives on loss of human life caused by fl oods. Natural Hazards Rev., 34, pp. 151–175, 2005. [27] Downing, T.E., Olsthoorn, A.A. & Tol, R.S.J., Climate Change and Extreme Events - Altered Risk, Socio-economic Impacts and Policy Responses. Institute for Environmental Studies, Vrije Universiteit and Environmen- tal Change Unit, University of Oxford, Amsterdam, The Netherlands and Oxford, United Kingdom, 309 p., 1996. [28] IPCC, Special Report on The Regional Impacts of Climate Change. An Assessment of Vulnerability. http://www.ipcc.ch/, 2007. [29] IPCC, Climate change 2007: the physical science basis. Summary for policy- makers, 21 p., 2007. [30] Pottier N., L’utilisation des outils juridiques de prévention des risques d’inondation: évaluation des effets sur l’homme et l’occupation des sols dans les plaines alluviales (application à la Saône et à la Marne). Thèse de doctorat, CEREVE, Ecole des Ponts et Chaussées, 436 p., 1998. [31] Pujol, N., Neppel, L. & Sabatier, R., Regional tests for trend detection in maximum precipitation series in the French Mediterranean region. Hydrological Sciences Journal, 52(5), pp. 956–973, 2007. [32] Norrant, C., Tendances pluviométriques indicatrices d’un changement clima- tique dans le bassin méditerranéen 1950–2000. Ph.D Thesis, University of Aix-Marseille, 2004. [33] Lang, M. & Renard, B., Analyse régionale sur les extrêmes hydrométriques en France: détection de changements cohérents et recherche de causalité hydrologique. Congrès S.H.F. Variations climatiques et hydrologie, Paris, 8 p., 27–28 mars 2007.

[34] Ullmann, A. & Moron, V., Weather regimes and sea surge variations over the Gulf of Lions during the 20th century. International Journal of Climatology, 28, pp. 159–171, 2008. [35] Parker, D.J., Floodplain development policy in England and Wales. Applied Geography, 15(4), pp. 341–363, 1995. [36] Werritty, A., Sustainable fl ood management: oxymoron or new paradigm? Area, 38(1), pp. 16–23, 2006. [37] Andjelkovic, I., Guidelines on non-structural measures in urban fl ood management IHP-V/Technical documents in hydrology/n°50. International hydrological programme. UNESCO, 81 p., 2001. [38] Lefrou, C., ed. Les crues des 12, 13 et 14 novembre 1999 dans les départe- ments de l’Aude, de l’Hérault, des Pyrénées-Orientales et du Tarn. Rapport au Ministre de l’Aménagement du Territoire et de l’Environnement, 99 p., 2000. [39] Huet, P., Martin, X., Prime, J.L., Foin, P., Laurain, Cl., Cannard, Ph., Retour d’expérience des crues de septembre 2002 dans les départements du Gard, de l’Hérault, du Vaucluse, des Bouches-du-Rhône, de l’Ardèche et de la Drôme. MEDD-IGE, 133 p., 2003. [40] Pottier, N., Penning-Rowsell, E., Tunstall, S. & Hubert, G., Land use and fl ood protection: contrasting approaches and outcomes in France and in England and Wales. Applied Geography, 25, pp. 1–27, 2005. [41] Ledoux, B. & Hubert, G., Le Coût du risque… L’évaluation des impacts socio-économiques des inondations. Presses de l’École nationale des ponts et chaussées: Paris, 231 p., 1999. [42] Gaillard, J.-C., Resilience of traditional societies in facing natural hazards. Disaster Prevention and Management, 16(4), pp. 522–544, 2007. [43] MEEDDAT, Quinze expériences de réduction de la vulnérabilité de l’habitat aux risques naturels RETOUR D’EXPÉRIENCES risques naturels majeurs Quels enseignements? MEEDDAT, CERTU, EP-Loire, 36 p., 2008. [44] Hutter, G., Strategies for fl ood risk management – a process perspective, eds J. Schanze, E. Zeman, J. Marsalek Flood Risk Management. Hazards, Vulner- ability and Mitigation Measures. NATO Science series, IV, Environmental sciences, vol. 67. Springer-Verlag: Berlin, pp. 229–246, 2006. [45] Parker, D.J., ed. Floods. Routledge: London, vol. 2, 748 p., 2000.