Ocean & Coastal Management 53 (2010) 679e690

Contents lists available at ScienceDirect

Ocean & Coastal Management

journal homepage: www.elsevier.com/locate/ocecoaman

Use of high water marks and eyewitness accounts to delineate flooded coastal areas: The case of Storm Johanna (10 March 2008) in ,

Jean-Marie Cariolet*

Géomer e UMR 6554 CNRS LETG (Université de Bretagne Occidentale), Institut Universitaire Européen de la Mer, Technopôle Brest e Iroise, Place Nicolas Copernic, 29285 Plouzané, France article info abstract

Article history: Mapping flooded coastal areas can be carried out using different methods and can promote a better Available online 20 October 2010 understanding and management of coastal flood risks. The delineation of the coastal areas in western Brittany inundated during a storm that came through on 10 March 2008 was determined based on eyewitness accounts and physical marks noted in situ. Using this methodology, 25 sites were mapped, representing an overall flooded area of more than 30 ha. The delineation of the flooded areas was compared with the official French (PPR-SM) flood zones, revealing some discrepancies. Finally, two case studies illustrate how coastal flood mapping can be useful for validating hydrodynamic models. Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction When these markers cannot be used (or are no longer visible), sediment deposits can provide other clues. The analysis of sediment There are many methods for mapping areas that are occasionally deposits from tsunamis or storm surge is based on grain sorting and subject to coastal flooding. Large surfaces of flooded areas can be many studies have demonstrated the link between sorting and the mapped using satellite images [1,2]. Based on a resolution of several limits of the flooded area. A horizontal gradient can be distin- tens of meters, this technique can only be used when flooding guished by a change in grain size from coarse to fine as the limits of affects a large surface area. For example, satellite image mapping is the inundated areas are reached; likewise, vertical limits can be not appropriate for the coast of Brittany where most low-lying determined from the sediment layer deposited by swash and the coastal areas comprise relatively small land surfaces. An alternative sediment layer deposited by backswash [4,7,11e16]. However, the method involves aerial surveys of flooded areas during inundation, upper limits of sediment deposits are often lower than the actual which generally occurs during high tide in a macrotidal environ- highest water level reached by run-up, because backswash tends to ment [3]. However, this technique is entirely dependent on weather erase them. Sediment deposits thus only rarely correspond to the conditions that are rarely clement during coastal flood events. limits of the flooded area, and this method of delineation may When neither of these two methods can be used, the remaining thereby underestimate the extent of flooding [11]. solution is to collect evidence in the field. This method has been Mapping flooded areas is useful for several reasons. First, many effectively used after tsunami and hurricane events d in particular authors stress the fact that information on the extent of flooding Hurricane Katrina d and provides useful information. For example, must be collected quickly to assess damage, to assist authorities and some studies compile eyewitness accounts to delineate the areas insurance companies in determining damage and to produce flood that had been flooded [4e7]. Others use in situ physical marks or risk maps [2,6,14]. Second, the use of geographic information markers to determine horizontal and vertical high water levels. systems (GIS) makes it possible to compare flooded areas with Storm deposits and visible water marks left on the walls of build- official flood zones d created by regulations such the French ings are excellent markers of the elevation of flood waters Coastal Flood Risk Prevention Plans (Plans de Prévention des Risques- [5,6,8e10]. Identifying damage to vegetation (injury due to salt Submersion Marine, PPR-SM) in France or the Federal Emergency water and broken branches on trees and shrubs) can also help Management Agency (FEMA) in the United Statesdto estimate delineate areas that were inundated for several hours [5e11]. their veracity [8,17]. Third, mapping flood data can help validate hydrodynamic models [10]. Fletcher et al. [8] and Tsuji et al. [5] have used mapping to validate their estimates of extreme water levels obtained using numerical approaches. Other studies have fl * Tel.: þ 33 664 03 12 91; fax: þ33 02 98 49 87 03. mapped ood water marks to validate wave overtopping models E-mail address: [email protected]. that simulate the extent of flood zones [3,18,19].

0964-5691/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.ocecoaman.2010.09.002 680 J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690

Fig. 1. Physical marks used in this study. A. Wrack line, Tredrez-Locquémeau (Côtes-d’Armor). B. Water mark on a wall, Tredrez-Locquémeau.

On 10 March 2008, a storm hit the northwestern coast of France This study presents the method that was employed to delineate during a spring tide, causing coastal flooding in many areas. The the areas that underwent coastal flooding in Brittany on 10 March extreme water levels observed on that day were due to the rare 2008. The results are then illustrated by three specific examples combination of several marine weather phenomena: storm surge, that demonstrate the usefulness of mapping low-lying coastal areas rough seas and spring tide. Accordingly, flooding on Brittany’s that were flooded during this storm event. macro- and megatidal coasts took place during two successive high tides. During the morning high tide, when the storm eye was located near Ireland, a strong southwesterly flow caused many 2. Materials and methods floods on the southern coast of Brittany. During the evening high tide, the storm eye was located on the eastern coast of the United In macrotidal and megatidal environments, coastal flooding Kingdom, causing a strong northwesterly flow and flooding generally only lasts several hours, and usually occurs during high occurred mainly on the northern coast of Brittany. The wind and tide. No aerial surveys could be undertaken on 10 or 11 March 2008 waves observed during this storm have a two-year return period. in Brittany. Thus, any evidence that could help to delineate areas For the observed tides, the return periods range from 2 to 100 that were flooded during the storm was collected directly in the years [20]. field. This evidence included in situ physical marks (wrack lines,

Fig. 2. Diagram summarizing the different layers incorporated in the GIS for mapping the flooded coastal areas. J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690 681

Fig. 3. Spatial distribution of flooded coastal area at Gâvres (Morbihan). high water marks on walls) or was based on eyewitness accounts accounts more vague), the affected coastal municipalities were from people present during the flood event (Fig. 1). visited as quickly as possible. On 11 March 2008, an inventory of all sites that were flooded in Brittany was drawn up based on local newspaper articles and 2.1. Collecting data in the field telephone calls to town halls to confirm whether or not town had been flooded. Since the longer the time after a flood, the more any Field investigations were based on a series of informal inter- evidence is attenuated (e.g. physical marks erased and eyewitness views with local residents as well as on an inventory of the physical

Fig. 4. Spatial distribution of flooded coastal area at Tredrez-Locquémeau (Côtes-d’Armor). 682 J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690

Fig. 5. Spatial distribution of flooded coastal area at Beniguet Island (Finistère). marks that indicated the (vertical and horizontal) limits of flooding information provided in each interview was compared with (Fig. 2). During each informal interview, the following information neighbors’ answers and vertical marks to confirm (or invalidate) was recorded: street address (house number and street name) to the previous interview. Many eyewitnessesdwho had filed insur- map the interview, the degree of flooding, the time and duration of ance claims for damages incurreddoften exaggerated the facts. flooding, the high water level, the direction of current flow, and Untrue accounts were easy to identify and were eliminated from where the seawater came from. Each eyewitness was also asked to the database. Ultimately, all interviews, which were mapped using indicate the limits of the flooded area. The identified limits of the their street address, contributed to a grid of points of gathered data; flooded areas were mapped using a GPS or by noting the address of these data points were then ‘interpolated’ to delineate the flooded the residence closest to the indicated limit. area. Each interview was assigned a code (interview1, interview2, Each piece of evidence/physical mark (e.g. wrack lines and water etc.) and recorded both in a field notebook and on a map. The marks on walls) was also coded and photographed.

Fig. 6. Location of the built-up coastal areas that were flooded during Storm Johanna in Brittany, France on 10 March 2008. J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690 683

Fig. 7. Delineation of the areas that flooded on 10 March 2008 and the official PPR-SM flood zones for the town of Penmarc’h (Finistère) (source: Baillet [22]).

Whenever possible, these marks were mapped with a differen- were conducted, delineations were obtained from the contour line tial GPS (DGPS) to the nearest centimeter to determine the eleva- determined from the height of debris lines, using Surfer 8 software tion1 of the flood water levels. Regarding swash marks, their (Fig. 2). The flooded areas were digitalized on the French National elevation sometimes indicated the maximum level of run-up. This Geographic Institute (Institut Géographique National, IGN) 2005 Ò was equal to the maximum run-up (Rmax) reached during inunda- orthophotographs (BD ORHTO ) or on IGN SCAN25 digital maps by tion. This information is very useful because it helps to validate the overlaying surface elements (polygons). Surface areas were calcu- theoretical and mathematical approaches that are often used to lated from these polygons. The swash marksdwhich delineate the calculate extreme water levels. These levels also delineate flood- area that was floodeddmeasured using DGPS were incorporated prone zones (PPR-SM zoning) and define crest levels for coastal into the GIS database using Surfer 8 software. They were repre- structures (breakwaters, etc.). The data collected at Tredrez-Loc- sented by line elements and were used to delineate the flooded quémeau (Côtes-d’Armor département) presented in the Results areas in certain areas more precisely (Fig. 2). Each field measure- (Section 3.2) will address this point in particular. ment of alleged water levels was symbolized by a point element (information related to the point, such as water level, were incor- porated as point attributes). Finally, wave overtopping and/or 2.2. Incorporation of data into a GIS database overflow areas were represented by a line element (Fig. 2).

All the information gathered in the field was then incorporated into a GIS database, using ArcGIS 9.2 software. Each interview was 3. Results mapped based on the street address of where the interview took place. The delineation of the flooded areas was deduced from the Extracts from the GIS database illustrate how the interviews grid of interviews and physical marks. Where topographic surveys (symbolized by points on the maps) and physical marks made it possible to map the areas that were flooded (Figs. 3e5). At Gâvres (Morbihan département), a highly built-up site and where no 1 All elevations are relative to French Datum (NGF), unless otherwise noted. horizontal marks could be detected, interviews and vertical marks 684 J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690

Fig. 8. Delineation of the areas flooded on 10 March 2008 and the official PPR-SM zones for the town of Ile Tudy (Finistère) (source: Baillet [22]).

Fig. 9. Port of Tredrez-Locquémeau on 11 March 2008 (photo: Joel Le Jeune). The shingle barrier that protects the low-lying area located landward was leveled and pushed landward by rollover. By comparing aerial IGN orthophotographs from 2005 and DGPS coordinates, the barrier migrated landward by an estimated 9 m. J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690 685

calculated by the French Naval Hydrographic and Oceanographic Service (Service Hydrographique et Océanographique de la Marine, SHOM). Since the storm event on 10 March 2008 can be considered as a reference for coastal flooding, the objective was to compare official PPR-SM flood zones with areas that were actually flooded on 10 March 2008.

3.1.1. Methods The GIS layer representing the regulatory flood zones laid out in the PPR-SM (Source: DDEA 292) was superimposed on the GIS layer representing areas that were flooded on 10 March 2008 in Finistère. This approach identified areas that were inundated during the storm and classified as part of the official PPR-SM flood zone.

3.1.2. Results Results show that for the Finistere département, only 2.7% of areas flooded on 10 March 2008 were located in an official PPR-SM flood zone. Two observations can be drawn from this comparison. Fig. 10. Wrack line deposited at the Tredrez-Locquémeau site at an elevation of more First, many municipalities that were flooded in March 2008 do have than 6.5 m. This physical mark shows the maximum run-up reached on the evening of not designated flood-prone zones. These municipalities include 10 March 2008. Sein Island, Ushant Island, Molène Island and also , Camaret-sur-Mer and . Second, of the municipalities were the main indications that could be used to delineate the that have defined flood zones and that were flooded on 10 March flooded area (Fig. 3). At Tredrez-Locquémeau (Côtes-d’Armor) and 2008, the flooded areas did not correspond to the official PPR-SM on Beniguet Island (Finistère département), where the degree of flood zones. The examples of the towns of Penmarc’h and Ile Tudy human influence is low or practically nil, swash marks were (southern Finistère) are particularly instructive. At Penmarc’h, the recorded with a DGPS, and then used to precisely map flooded Saint-Guénolé port and the low-lying coastal area located landward surfaces (Figs. 4 and 5). were inundated on the evening of 10 March 2008 (Fig. 7). This same The sites that were inundated during the morning high tide on coastline had already been flooded on 13 December 1978 (source: 10 March 2008 were mainly located on the southern coast of Le Télégramme, regional newspaper, 14 December 1978). However, Brittany and oriented in a south-southeastern direction (Fig. 6). none of the official PPR-SM flood zones (red, blue and green zones) These low-lying areas, whose shorelines directly face onshore ratified in December 1999 include the port (Fig. 7). winds, experienced many overflows after the active cold front came At Ile Tudy, the area that underwent flooding on 10 March 2008 through [20]. During the evening high tide, after surging seas is located at the tip of the peninsula (Fig. 8). This area has already developed in the Atlantic Ocean, the coastlines of Finistère and the been flooded several times, notably on 27 October 2004 (source: Le Tregor region (Côtes-d’Armor) were the most affected (Fig. 6). Télégramme, 28 October 2004). Nevertheless, the official PPR-SM Overall, nearly 25 cases of flooding in built-up areas were inven- flood zone ratified in June 1997 does not include the tip of the toried in Brittany on 10 March 2008 alone. peninsula (Fig. 8). Based on the constituted GIS database, the total surface of built- These comparisons show that the official PPR-SM flood zones up areas that were flooded on 10 March 2008 was determined to be that were defined during 1997e2007 are inconsistent with areas equal to 24.4 ha (Fig. 6). This value probably underestimates the that were actually flooded during the 10 March 2008 storm. surface actually flooded because many sites were less drastically Nevertheless, these results should be interpreted with caution flooded and could not be mapped. Inundated surfaces covered an because they are based on only one storm event, which was average of 1.1 ha. The largest flooded area was located at Gâvres exceptional in character [20]. (Morbihan) where nearly 6.7 ha were submerged (Fig. 3). As mentioned above, these results are useful in more than one aspect and can be utilized in different ways. The following three 3.2. Case study 2: using swash marks to validate estimates of cases illustrate how mapping areas flooded on 10 March 2008 can extreme water levels be useful for better management and understanding of coastal flooding. Many studies concur that the elevation of the high water mark corresponds to the highest level reached by swash and thereby 3.1. Case study 1: comparison of flooded areas with regulatory indicates wave run-up, theoretically occurring during high tide seawater flood zones (Finistère) [23]. This can be expressed by the following equation: ¼ þ T The French Coastal Flood Risk Prevention Plans (Plans de Elevationswash mark Elevationobserved high tide Rmax (1) Prévention des Risques de Submersion Marine - PPR-SM) were where RT corresponds to the highest level of run-up reached at instituted as Law no. 95e101 of 2 February 1995 (also known as the max a given location on the beach. Barnier Law), and can be considered analogous to actions under- Nott and Hubbert [10] thus compare elevation of storm deposits taken under the auspices of FEMA. The PPR-SM regulates land use with extreme water levels estimated from storm-surge models. The according to the flood risk of the land area in question. The high water deposits left by the 10 March 2008 storm were used to objective of the PPR-SM is to guarantee human safety, reduce the vulnerability of property and activities in flood-prone areas and limit insurance claims in case of a flood event [21]. 2 Department of Public Works and Agriculture of the Finistère département fi fl In Finistère, 22 municipalities have rati ed a PPR-SM ood zone. (Département de l’équipement et de l’agriculture du Finistère); see www.finistere. These zones are based on 100-year return extreme water levels developpement-durable.gouv.fr 686 J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690

Fig. 11. A: Aerial photograph of Beniguet Island taken before the storm of December 1989 (source: Priser [29]). B: Aerial photograph of Beniguet Island taken on 25 July 2009 (source: Tiphaine Franz).

2 2 validate the estimates of extreme water levels calculated from level, calculated from the equation h ¼ 2.28 He 68.5 (He/gTe) where numerical formulas. Taking the example of Tredrez-Locquémeau He and Te are the height and the associated period, respectively, of (Côtes-d’Armor), this town was suddenly flooded on 10 March waves that exceed 12 h in one year, and dx ¼ fetch of the center of 2008 around 19:00 UTC (Fig. 4) by waves that flooded the shingle the depression to the coast (m); and barrier which protects the low-lying coastal area located landward s ¼ 2 (Fig. 9). Storm deposits indicated the highest level reached by s raCDW (4) swash (Figs. 1, 9 and 10). This physical mark, recorded using DGPS, 3 where, ra ¼ density of air (1.21 kg/m ), W ¼ wind velocity (m/s) culminated at 7 m NGF in elevation with an average elevation of 10 3,C ¼ drag coefficient. 6.5 m NGF. For this portion of coastline, the 100-year return D The meteorological data necessary for these formulas come from extreme water level calculated by SHOM, using the method the data recorded at the Météo-France weather station at Lannion- developed by Simon [24], is 5.5 m, or more than 1 m lower than the Servel, located 8 km to the northwest of Tredrez-Locquémeau. estimates obtained from the field survey3. Simon’s method [24] is Maximal wave run-up was estimated from the equation devel- based on historical tide gauge records. The tide gauges are only oped by Holman [27]: located in harbors, sheltered from wave action. Therefore, SHOM calculations [24] are based only on storm surge and does not take RT ¼ 1:07H x (5) into account the effect of storm waves on water levels on the coast, max o o T which is determined by wave set-up and wave run-up. The 100- where Rmax represents the value of maximum run-up, Ho repre- year return extreme water levels established by SHOM were used to sents the height of offshore waves, and xo is the Iribarren number define the reference elevations for delineating official PPR-SM flood [28],or zones [25]. 1=2 x0 ¼ tan b=ðH0=L0Þ (6) 3.2.1. Methods where L is the offshore wavelength (1.561 T2). The extreme water level reached during the 10 March 2008 o event at Tredrez-Locquémeau can be numerically calculated using 3.2.2. Results equation (1). To obtain the elevation of storm surge during high tide On 10 March 2008 at 19:00 UTC, i.e. during the evening high at 19:00 UTC, values were taken from the tide gauge located at tide, the Lannion-Servel weather station recorded a pressure of , the closest representative tide gauge, located 31 km to the 986.8 hPa and a westerly wind (direction 270) of 15 m/s. At the west. However, Tredrez-Locquémeau’s coastline directly faces any same point in time, the storm eye was located at 450 km from onshore winds coming from the west, suggesting that the storm Tredrez-Locquémeau. surge was higher here than at the port of Roscoff, which is orien- Calculated from mathematical models (equations (2)e(4)), tated northward. Therefore, storm surge was estimated using the storm surges reached 0.74 m during the evening high tide, or following equation: 0.26 m (elevation due to the inverse barometer effect) plus 0.48 m S ¼ dA þ dx (2) (elevation due to wind set-up). The storm surge recorded by the Roscoff tide gauge at the same point in time was 0.29 m, or 0.45 m where dA represents the increase in water level due to a drop in lower. The calculation of maximum run-up at 19:00 UTC was based atmospheric pressure - a 1 hPa decrease in pressure leads to a 1 cm 4 on simulated wave data (Hmo ¼ 10.2 m and Tpic ¼ 16 s) and increase in water level - and dx represents the water level generated morphological data (beach slope ¼ 0.026). The maximum run-up by wind set-up. Wind set-up can be estimated using the expression reached at 19:00 UTC was estimated at 1.78 m. developed by Bowden [26]: Adding the height of high tide predicted at 19:00 UTC (4.356 m), the calculated value of storm surge (0.74 m) and the maximal run- dx ¼ðs =grhÞ$d (3) s x up value (1.78 m) give an estimated water level of 6.876 m. This where, g ¼ acceleration due to gravity (9.81 m/s2), r ¼ density of value roughly corresponds to the elevation of the storm deposits seawater (water at 12 C ¼ 1026 kg/m3), h ¼ depth of wave-base that had a mean elevation of 6.5 m and a maximum elevation of

4 Swell data obtained by simulation using the WAVEWATCH IIIÔ wave model 3 Source: Map of extreme water levels (northern Brittany), SHOM 2007. (SHOM). J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690 687

Fig. 12. Digital elevation model of the area flooded on 10 March 2008 on Beniguet Island (Finistère).

7 m. This shows the usefulness of an empirical field approach for calculated based on the surveyed points, according to a 50 50 cm calculating extreme water levels. grid (Fig. 12). The volume of water that flooded the island during the storm was then estimated, assuming that the substrate is impermeable. Using the reference elevation, data were recorded 3.3. Case study 3: using swash marks to calculate incoming water regarding the morphological characteristics of the flooded area, volumes such as the altitude of the two breaches and the shingle barrier which waves overtopped. The eastern part of Beniguet Island, located in the Molène archipelago (Finistère), consists in a gravel barrier surrounding 3.3.1.2. Calculation of the volume of water that overtopped natural “ ” low-lying areas inland, called the Tahiti barrier. This part of the island structures. The overtopping discharge rates were calculated island was particularly invaded by the sea on 10 March 2008 for Breaches 1 and 2 and the Tahiti barrier using the equation (Fig. 5). Seawater entered through two breaches created during developed by Van der Meer and Janssen [30] and widely used in e Winter 1989 1990 (Fig. 11). Many pebbles were dumped on the civil engineering [19,31]: other side of the Tahiti barrier at the northeastern part of the island ! and prove that seawater overtopped the barrier, flooding one of the x qqffiffiffiffiffiffiffiffiffiw pffiffiffiffiffiffiffiffiffiffiffiop Rc 1 1 two affected low-lying areas. ¼ 0:06$ $exp 5:2$ $ $ for xop < 2 3 tan b Hs xop grgbghgb As the island is classified as an integral nature reserve, swash gHs marks remained intact after the 10 March 2008 storm. This site, free (7) from any human-related disturbances, was an exceptional study 3 site for studying a coastal flood event. where, qw ¼ mean overtopping discharge rate (m /s/m), xop ¼ Iribarren number or surf similarity parameter [28], see 3.3.1. Methods equation (6), tan a ¼ beach slope, Rc ¼ height of coast at the point of 3.3.1.1. Estimation of incoming water volumes in the low-lying areas overtopping - observed tide, Hs ¼ significant wave height, based on measurements of swash marks and topographical sur- gr gb gh gb ¼ reduction factors for slope roughness, presence (or veys. First, swash marks were surveyed using DGPS. The data were absence) of a berm, shallow water and angle of wave approach, linked to the French geodetic system and to the Lambert II g ¼ acceleration due to gravity. projection system through an IGN terminal located on the island.5 Data on storm waves were obtained from measurements taken Swash marks could therefore be georeferenced to determine their by the (directional Datawell) wave gauge located near Pierres elevation and determine the limits of the area that was flooded on Noires, 5 km southwest of Beniguet Island. The tide data come from 10 March 2008. The whole flooded area, the shingle barriers, the the tide gauge at located 5 km east of Beniguet Island. two breaches and the foreshore were thus surveyed. More than To obtain the total volume of water that overtopped the 5300 points were recorded during two visits to the island (20e21 breaches and the Tahiti barrier, the mean discharge rates, expressed May 2008 and 24 June 2008). A digital elevation model (DEM) was as m3/s/m, were multiplied by the length of the coastline over- topped and the number of waves counted during the period considered (period in seconds divided by the wave period (Tpic); 5 Site no. 2904003 (source: http://geodesie.ign.fr/). e.g. 30 min with Tpic ¼ 15 s, or 1600 s/15 s ¼ 120 waves). 688 J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690

Table 1 Morphological features of wave-overtopped areas and data on the storm surge of 10 March 2008 on Beniguet Island.

Overtopped areas Length of overtopped coastline Elevation at the entrance of the breach Beach slope (tan b) Affected Low-lying area Breach 1 92 m 4.7 m (NGF) 0.1 1 Breach 2 104 m 4.7 m (NGF) 0.081 2 Tahiti barrier 90 m 6.73 m (NGF) 0.15 1

Low-lying area Elevation of swash mark Volume of incoming water on 10 March 2008 Low-lying area 1 (Loc’h) 4.7 m (NGF) 54 791 m3 Low-lying area 2 3.7 m (NGF) 19 073 m3

3.3.2. Results gathered in this study and specific analysis of the study region thus 3.3.2.1. Estimation of the volume of overtopping water and morpho- provide valuable information for coastal flood management in logical characteristics of the areas that were overtopped. Both Brittany. The quality control of the verbal interviews constitutes breaches have identical morphology and the same elevation: 4.7 m one limitation of the data utilized in this study; however, false (8.2 m on marine charts). For comparison, the highest astronomical accounts could easily be identified by cross-checking with neigh- tides reach 4.24 m (7.74 m on marine charts) in this part of the boring residents’ accounts and vertical marks. Molène archipelago. Fletcher et al. [8] and Dobosiewicz [17] compared official zones The mean elevation of the swash mark deposited by the storm defined as flood-prone in the by FEMA with areas that tide in the trough of the low-lying area (Loc’h, Breach 1) was 4.7 m. were actually flooded during hurricanes. These studies assessed the It can be thus assumed that the trough was “filled” up to the rim. pertinence of the official flood zones. In the present study, mapping Most of the water entered by the breach, even though the projec- helped to show that there were numerous inconsistencies between tion of pebbles behind the Tahiti barrier proves that waves also official PPR-SM flood zones and zones that were actually flooded. In overtopped the opposite side of the breach (Fig. 8). The calculated Finistère, many municipalities that were flooded such as Sein water discharge, assuming that the ground is impermeable, was Island, Audierne and Dournenez have not defined official flood 54 791 m3 (Table 1). However, the sandy substrate is porous, sug- zones. Moreover, in municipalities where flood zones have been gesting that the volume of discharged water was greater. defined, zones that were flooded did not necessarily correspond to Regarding the network of troughs in Breach 2, the mean those that are indicated on official flood maps. The comparison elevation of swash marks reached 3.7 m. The 19 073 m3 of water between 100-year return flood zone and Storm Johanna-flooded entered the breach (Table 1). Here, water only entered through the areas is limited because the storm-surge response varied, reaching breach, which may explain the lower water volume compared to from 2- to 100-year return surge levels depending on the location. Breach 1. Nevertheless, coastal flooding with a two-year return surge level in an area that is not even referenced as a 100-year return flood zone 3.3.2.2. Calculation of the volume of overtopping waves. Mean shows that other parameters, such as wave run-up, have to be take overtopping discharge was calculated for the three areas that were into account to estimate extreme water levels and thus PPR-SM overtopped and whose morphological characteristics are compiled flood zones. in Table 1. The results show that Low-lying area 1 (Loc’h) was flooded The present study highlights the importance of obtaining and by overtopping at breach 1 in the morning and in the evening. The archiving spatial data that indicate the actual extent of flooding. At “Tahiti” barrier was overtopped only in the evening. Low-lying area the request of government authorities, municipalities declare 2wasflooded by overtopping at Breach 2 in the morning and in the damage incurred within municipal limits to file insurance claims evening (Fig. 13). Calculated overtopping volumes matched rela- and obtain monetary compensation. Most declarations of natural tively well those measured in the field (Table 2). disasters are not accompanied by any georeferenced data. Setting up a GIS database would serve as a “risk memory” and better 4. Discussion localize the areas that are at flood risk. This type of work has been carried out in the United States by FEMA [9,32,33]. In France, While the combined use of eyewitness accounts, high water mapping flooded areas may, for example, supplement and marks and topographic surveys for delineating flood zones is not contribute to the French national natural disaster database (known new in coastal flooding studies [4e10], this is the first study to as the CatNat/GASPAR database) by adding a fine-scaled spatial employ this [three-pronged] approach in a macro- and megatidal approach. This database has some shortcomings [34] and mapping environment, where inundations are very short-lived. The data would help enhance the information contained in it. Work carried out in Tredrez-Locquémeau shows the importance of revising theoretical estimates of extreme water levels in France. The approach used here, which takes the effect of waves into account in estimating water levels on the coast (run-up), proved to be more accurate than the method used by SHOM [24], which only considers storm surge and underestimates extreme water levels (Fig. 14). Many PPR-SM flood zones have been delineated based on 100-year return levels calculated by SHOM, such as the PPR-SM

Table 2 Comparison of overtopping water volumes and volumes calculated using the Van der Meer and Janssen [30] equation.

Observed volume Calculated volume Low-lying area 1 (north) 54 791 m3 50 774 m3 Fig. 13. Average wave-only overtopping discharge rate calculated for the two breaches Low-lying area 2 (south) 19 073 m3 22 155 m3 and the Tahiti barrier and total incoming water volume on 10 March 2008. J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690 689

Fig. 14. 100-year return extreme water levels calculated by SHOM (2007) according to the method developed by Simon [24] for the northern coast of Brittany.

flood zones in northern Finistère and in the town of Guisseny [25]. reassess the delineation of future PPR-SM flood zones in some The official PPR-SM flood zones are inaccurate, because the refer- coastal towns at risk. ence water level used does not account for rough seas and thereby underestimates real water levels. Accordingly, more and more Acknowledgments studies favor an approach that integrates the effects of both set-up and run-up [19,25,35e37]. In this study, the swash marks were I would like to thank Serge Suanez for his invaluable advice. used to validate estimates of extreme water levels at only one Thanks also go to Louise Baillet for her work and her help in location. Most of the swash marks surveyed after this storm were comparing PPR-SM zones with flooded areas in Finistère. Thank located in low-lying areas and cannot be used to determine run-up you to Joël Le Jeune, mayor of Tredrez-Locquémeau for his help and elevation. Only at Tredrez-Locquémeau was a swash mark located his interest in the study, to Pierre Yesou and the whole staff of the above a coastal defense structure, and actually attests to run-up natural reserve of Beniguet Island for facilitating my visits. Warm elevation. thanks to Ruddy Magne at SHOM for providing offshore swell data, Finally, the study conducted on Beniguet Island shows that the to Philippe Huguet at the Risk Prevention Bureau of the Finistère wave overtopping model, based on Van der Meer and Janssen [30], DDEA for data on PPR-SM. I am grateful to Laurence David for her used for designing coastal structures such as breakwaters or jetties help and to all the people I met in the course of this field study and [19], is also useful for natural barriers such as dunes and shingle who gave an account of the 10 March 2008 storm. barriers.

5. Conclusion References

[1] Wang Y. Using Landsat 7Ô data acquired days after a flood event to delineate This paper presented a method used for mapping seawater fl fl fl the maximum ood extent on a coastal oodplain. International Journal of ooded areas during Storm Johanna that hit Brittany on 10 March Remote Sensing 2004;25(5):959e74. 2008. The combined use of eyewitness accounts and physical marks [2] Kiage LM, Walker ND, Balasubramanian S, Badin A, Barras J. Applications of led to a relatively precise delineation of flooded areas that was Radarsat 1 synthetic aperture radar imagery to assess hurricane-related flooding of coastal Louisiana. International Journal of Remote Sensing 2005;26 sometimes further improved with DGPS surveys. This method can (24):5359e80. be applied rather well to a large range of environments, whatever [3] Benavente J, Del Rio L, Gracia FJ, Martinez-Del-Pozo JA. Coastal flooding hazard their degree of human impact. In built-up areas, field work should related to storms and coastal evolution in Valdelagrana spit (Cadiz Bay Natural Park, SW Spain). Continental Shelf Research 2006;26:1061e76. include interviews such as at Gâvres (Fig. 3), although in pristine [4] Keating B, Whelan F, Bailey-Brock J. Tsunami deposits at Queen’s Beach, Oahu, areas, such as on Beniguet Island (Fig. 5), only physical markers can Hawaii e initial results and wave modeling. Science of Tsunami Hazards be used. This field approach circumvents dependence on image 2004;22(1):23e43. d fi d [5] Tsuji T, Namegaya Y, Matsumoto H, Iwasaki S-I, Kanbua W, Sriwichai M, et al. data which can sometimes be dif cult to obtain or on assertions The 2004 tsunami in Thailand: surveyed runup heights and tide gauge made by authorities in municipalities that have no or little geore- records. Earth Planets and Space 2006;58:223e32. ferenced data. [6] Wijetunge JJ. Tsunami on 26 December 2004: spatial distribution of tsunami The compilation of data in a GIS optimizes the usefulness of height and the extend of inundation in Sri Lanka. Science of Tsunami Hazards 2006;24(3):225e39. spatial data. It was therefore possible to pinpoint some inconsis- [7] Choowong M, Murakoshi N, Hisada K-I, Charusiri P, Charoentitirat T, tencies between PPR-SM flood zones and areas that were actually Chutakositkanon V, et al. 2004 Indian Ocean tsunami inflow and outflow at e flooded in March 2008 in Finistère. Mapping was also useful for Phuket, Thailand. Marine Geology 2008;248:179 92. [8] Fletcher CH, Richmond BM, Barnes GM, Schroeder TA. Marine flooding on the validating hydrodynamic models, in particular those that have been coast of Kaua’I during hurricane Iniki: hindcasting inundation components developed to estimate extreme water levels and the volume of and delineating washover. Journal of Coastal Research 1995;11(1):188e204. overtopping water. [9] FEMA. High water mark collection for Hurricane Katrina in Alabama. FEMA- 1605 e Task order 414 and 421; 2006. 69 p. Overall, mapping can be a valuable tool for managing seawater [10] Nott J, Hubbert G. Comparisons between topographically surveyed debris flood risks in Brittany. It may also help government authorities to lines and modelled inundation levels from severe tropical cyclone Vance and 690 J.-M. Cariolet / Ocean & Coastal Management 53 (2010) 679e690

Chris, and their geomorphic impact on the sand coast. Australian Meteoro- [23] Moore LJ, Ruggiero P, List H. Comparing mean high water and high water line logical Magazine 2005;54:187e96. shorelines: should proxy-datum offsets be incorporated into shoreline change [11] Dawson AG. Geomorphological effects of tsunami run-up and backwash. analysis? Journal of Coastal Research 2006;22(4):894e905. Geomorphology 1994;10:83e94. [24] Simon B. Détermination des hauteurs d’eau extrêmes pour la délimitation du [12] Dawson AG, Shi S. Tsunami deposits. Pure and Applied Geophysics 2000;157: domaine public maritime. Annales Hydrographiques 1996;20(767):17e43. 875e97. [25] Cariolet J-M, Suanez S. Approche méthodologique pour une cartographie du [13] Nanayama F, Shigeno K, Shimokawa K, Koitabashi S, Miyasaka S, Ishii M. risque de submersion des côtes basses. La Houille Blanche 2009;2:52e8. Sedimentary differences between the 1993 Hokkaido-nansei-oki tsunami and [26] Bowden KF. Physical oceanography of coastal waters. New York: Wiley; 1983. the 1959 Miyakojima typhoon at Taisei, southwestern Hokkaido, northern 302 pp. Japan. Sedimentary Geology 2000;135:255e64. [27] Holman RA. Extreme value statistics for wave run-up on a natural beach. [14] Jaffe B, Gelfenbaum G. Using tsunami deposits to improve assessment of Coastal Engineering 1986;9:527e44. tsunami risk. In: Solutions to coastal disasters ’02, Conference Proceedings. [28] Battjes JA. Surf similarity. In: Proceedings of the 14th International coastal ASCE; 2002. p. 836e47. engineering Conference, vol. 1. American Society of Civil Engineers; 1974. [15] Nanayama F, Shigeno K. Inflow and outflow facies from the 1993 tsunami in p. 466e80. southwest Hokkaido. Sedimentary Geology 2006;187:139e58. [29] Priser L. Iles de Bretagne, Edition Libro-Sciences; 1991. 110 pp. [16] Paris R, Lavigne F, Wassner P, Sartohadi J. Coastal sedimentation associated [30] Van der Meer JW, Janssen W. Wave run-up and wave overtopping at dikes. In: with the December 26, 2004 tsunami in Lhok Nga, west Banda Aceh (Sumatra, Kabayashi Demirbilek, editor. Wave forces on inclined and vertical wall Indonesia). Marine Geology 2007;238:93e106. structures. American Society of Civil Engineers; 1995. p. 1e27. [17] Dobosiewicz J. Applications of digital elevation models and geographic [31] Hughes SA, et Nadal NC. Laboratory study of combined wave overtopping and information systems to coastal flood studies along the shoreline of Raritan storm surge overflow of a levee. Coastal Engineering 2009;56:244e59. Bay, New Jersey. Environmental Geosciences 2001;8(1):11e20. [32] Bellomo D, Pajak MJ, Sparks J. Coastal flood hazards and the national flood [18] Bates PD, Dawson RJ, Hall JW, Horritt MS, Nicholls RJ, Wicks J, et al. Simplified insurance Program. Journal of Coastal Research 1999;SI 28:21e6. two-dimensional numerical modelling of coastal flooding and example [33] Dean RG, Malakar SB. Projected flood hazard zones in Florida. Journal of applications. Coastal Engineering 2005;52:793e810. Coastal Research 1999;SI 28:85e94. [19] Peeters P, Schoorens J, Le Cornec E, Michard B, Lechat M. Définition de l’aléa [34] Douvinet J. Intérêts et limites des données « CatNat » pour un inventaire des submersion marine sur le site de la Grande Plage de Gâvres (Morbihan). La inondations. L’exemple des « crues rapides » liées à de violents orages (Bassin Houille Blanche 2009;1:45e51. parisien, Nord de la France). Norois 2006;4:17e30. [20] Cariolet J-M, Costa S, Caspar R, Ardhuin F, Magne R, Goasguen G. Aspects [35] Kumar VS, Babu VR, Babu MT, Dhinakaran G, Rajamanickam GV. Assessment météo-marins de la tempête du 10 mars 2008 en Atlantique et en Manche. of storm surge disaster potential for the Andaman Island. Journal of Coastal Norois 2010;215:11e31. Research 2008;24(2B):171e7. [21] Besson L. Les risques naturels: de la connaissance pratique à la gestion [36] Suanez S, Fichaut B, Sparfel L. Méthode d évaluation du risque de submersion administrative. Techni.Cités, collection dossier d’expert. Voiron; 1991. p. 610. des côtes basses appliquée à la plage du Vougot, Guissény (Bretagne). Géo- [22] Baillet L. Etat d’avancement des Plans de Prévention des Risques Littoraux en morphologie: Relief, Processus et Environnement 2007;1:15e32. Bretagne. Mémoire de Master 2 Expertise et Gestion de l’Environnement [37] Garès PA. Predicting flooding probability for beach/dune systems. Environ- Littoral mention recherche. Université de Bretagne Occidentale; 2009. 88 pp. mental Management 1990;14(1):115e23.