Skip to navigation – Site map

HomeIssues108Aléas actuels et futuresProvence pocket beach erosion

Aléas actuels et futures

Provence pocket beach erosion

Érosion des plages de poche de Provence
Cédric Brunel and François Sabatier
p. 77-82


The objective of this paper is to quantify the role of relative sea-level rise in the retreat of the shoreline, and evaluate this retreat in terms of future sea-level rise on 23 pocket beaches in Provence (France). Between 1896 and 1998, all of the investigated pocket beaches underwent erosion with an average retreat of 12.1 (± 3.5) m. Making use of the flooding principle, and assuming a relative sea-level rise of +11 cm between 1896 and 1998, we obtain an average shoreline retreat of 5.8 (± 3.5) m. A comparison between measured average retreat and calculated average retreat suggests that sea-level rise is one of the main factors influencing shoreline retreat (50% of the shoreline retreat can be explained by sea-level rise). The position of the shoreline for 2100 is evaluated not only using the historical shoreline trend and the flooding method, but also by adding the historical shoreline trend and the flooding concept (passive inundation) in the case of an increase in sea-level of about 44 cm (for 2100). On the investigated pocket beaches, the presence of a cliff “blocking” the landward migration of the profile will lead to a significant reduction in the beach width. Thus, we envisage that the pocket beaches will lose 75% to 97% of their present surface area. In other words, we predict the eventual disappearance of pocket beaches in Provence.

Top of page

Index terms

Geographical index :

France, Provence
Top of page

Author's notes

This work was supported by an IGCP-UNESCO programme: Vulnerability and resilience assessment of coastal zones in the Mediterranean and Black Sea areas related to the forecast of sea-level rise for management purposes.
Thomas-Jules Fleury and Philippe Dussouliez (CEREGE) are acknowledged for their technical assistance on GIS. M.S.N. Carpenter was responsible for translation and English style editing of the manuscript.

Full text

1 - Introduction

1Coastal environment managers are increasingly concerned by the current rise in sea level and its possible acceleration during the 21st century due to global warming (Gornitzet al., 1982; IPCC 2001). Thus, there is a crucial need for objective analysis of the potential vulnerability of beaches to sea level because of the unprecedented socio-economic development of the coastal fringe. Although some studies have been carried out on the vulnerability of Mediterranean coasts to Relative Sea-Level Rise (RSLR) (Jeftic et al., 1992; Jiménez and Sanchez-Arcilla, 1997), recent scenarii for 2100 (IPCC, 2001) and sea-level measurements (Church, White, 2006) suggest an acceleration in sea-level rise. This enhances the importance of a new simulation of the impact of rising sea level on shoreline position. In Provence, southern France (Figure 1), the economic stakes of the coastal fringe are considerable and mainly related to the tourism industry. Indeed, seaside tourism in this area accounts for an annual turnover of 4.6 billion euros, that is to say, a fifth of the total turnover of the French tourism industry. Thus, the beaches of Provence represent an extremely important economic stake. Consequently, their disappearance due to an increase in erosion, related to climatic changes, would be very detrimental to the economy of the region. With the exception of the Rhône delta, the beaches of Provence correspond to the “pocket beach” type. Pocket beaches generally display slow shoreline retreat because they are protected from waves, which is their main characteristic. Nevertheless the pocket beaches of Provence are backed up against the cliffs or embankments, and a landward displacement of the beach system is not possible, raising the question of the future evolution of these beaches.

2In this paper, we first evaluate the impact of RSLR on shoreline position during the 20th century (1896 to 1998) by use of the flooding concept. Secondly, we propose to simulate the shoreline position for 2100 using three different methods to estimate the partial or total disappearance of the pocket beaches.

Fig. 1 - Location of study sites.
Localisation des sites étudiés.

Fig. 1 - Location of study sites.Localisation des sites étudiés.

1.1 - Environmental setting

3Dozens of pocket-beaches are located along the rocky coast of Provence, between the Rhône delta and the Italian border, but only twenty-three of them satisfy the criteria selected in this study: i) natural beach evolution without any anthropogenic action influencing the shoreline position (no dredging activities or coastal engineering developments, etc.) and ii) small-sized catchment areas with significantly reduced fluvial inputs that do not influence the sedimentary budget of the beaches. We thus assume no significant river or continental sediment input to the investigated pocked beaches on a 100 year time scale. Moreover, the selected beaches are bordered laterally by rocky points that extend sufficiently seaward (water depth of 8 to 10m) to protect the beaches from storm waves, thus limiting the sedimentary outputs or inputs due to longshore drift. Consequently, this type of beach can be regarded as a relatively closed system, where sedimentary exchanges are only slightly affected by the waves and swell. The beach profile of the twenty-three pocket beaches is characterized by a narrow foreshore (average width between 5 and 47m) and the presence of an active cliff or an embankment “blocking” its landward migration. The submerged part of the beach is partly colonized by solid masses of Posidonia meadows at a depth varying between 2 and 10m. To a certain extent, this sea grass contributes to the reduction of the swell and forms an obstacle to the movement of sediments on the bottom. Sixteen of the selected beaches are made up of medium to coarse sands and the other seven beaches are composed of fine sands, which results in slopes varying from 1 to 3.6 % (average 1.9 %).

4All sites experience a micro-tidal regime (30 cm average tidal range). Waves come from two prevailing directions: the SW (modal and fair-weather waves) and the SE (storm waves). The annual return significant wave height is between 3.2 and 3.9m, with an associated period of 6.5 to 7.5 s. Due to important refraction, the height of the storm waves is reduced by at least 40% in the breaker zone. RSLR has been recorded for more than a century from the tide gauge of Endoume-Marseilles, serving as a reference to evaluate the rate of sea-level rise for the pocket-beaches of Provence. In spite of a strong interannual variability, the tide-gauge readings at Endoume-Marseilles show an upward trend over the century of about +1.1 mm/year (Suanezet al., 1997). The data from the Endoume-Marseilles tide-gauge is comparable with readings from the majority of Mediterranean sites considered as stable on a secular scale (Emeryet al., 1988). There is no indication of any vertical movement of the land on the investigated time scale for the pocket-beaches of Provence, so we can assume that the tide gauge record of Marseilles is representative of the RSLR in the studied area.

2 - Methods

2.1 - Past shoreline changes

5Old maps and aerial photographs were selected to cover the 20th century for the twenty-three selected pocket-beaches. The oldest data was collected by EPSHOM (Etablissement Principal du Service Hydrographique de la Marine) engineers in 1896, using the method of triangulation by theodolite. The most recent surveys, based on aerial photographs, were carried out by the IGN (Institut Géographique National) in 1998. The aerial photographs were treated by means of classical methods used in many studies (Dolan et al., 1991). The data from the 19th century were scanned, digitized and georeferenced in terms of French metric Lambert III co-ordinates. The geometrical correction was applied using ER Mapper© software from a reference document consisting of the BD-ortho 1998©. The BD-ortho 1998© is a very accurate base of aerial georeferenced mosaic photographs produced by IGN in 1998. The complete dataset was compiled in a Geographical Information System (MapInfo 6.5©) and then used to draw the shoreline position (instantaneous limit of the run-up) and calculate the distances of coastline retreat or advance. According to the EPSHOM, the surveying techniques used during the 19th century were highly reliable, so the measuring precision is estimated at +/- 10 m for the 1896 data. All the fixed points (landmarks or invariable features) visible on old maps of 1896 (buildings, works, road crossings, etc.) enables us to define, after superposition, a margin of error of +/- 3.5m. For the 1998 shoreline position, the IGN estimate the errors of BD-ortho at +/- 0.50m.

2.2 - The flooding method

6We did not adopt Bruun’s approach (Bruun, 1962) to determine the impact of sea-level rise on shoreline retreat, because the concepts on which this rule is based show several shortcomings (Pilkeyet al., 1993; Thieleret al., 2000; Pilkey and Cooper, 2004). Moreover, if we applied Bruun’s rule, this would imply we accepted all the necessary conditions for its use (Davidson-Arnott, 2005). Under these conditions, we considered it more advisable to make use of the geometrical relations of the beach profile (slope) and the height of sea level, while basing our study on the flooding principle.

7SR = (RSLR/tanβ) x t

8Where SR is the shoreline retreat; RSLR is the Relative Sea-Level Rise, tanβ is the average slope of the profile measured on the field by means of a theodolite and t is the considered time interval.

9In this study, we chose to use the principle of dynamic submersion, (or flooding method) assuming that the slope of the beach profile remains identical in time and migrates horizontally towards the land when sea-level rises. Even if the flooding method is similar to Bruun’s rule, we do not make use of Bruun’s concepts, except for assuming a constant slope. The approach presented here is based purely on the relationship between sea-level rise and beach slope. These two parameters are clearly correlated, since part of the beach is flooded when the sea-level rises. Nevertheless, the slope of the beach calls into question the morphodynamics of the studied area at the investigated time scale because the major issue is to determine an appropriate slope. Basing our approach on morphodynamics processes, we decided to adopt the average values of two different depths on the profile in order to obtain an accurate slope. The minimum and maximum slopes are taken as +/- bracket boundaries. Firstly, the upper part of the active profile controlled by wave action is defined by the slope between the shoreline and the annual depth of closure. The values of the depth of closure, which range between 2 and 5m, are calculated by taking into account the wave height in the nearshore zone (Hallermeier, 1981, Sabatieret al., 2004). Secondly, the lower part of the profile is defined between the shoreline and the outer limit of the bays on the pocket beaches (4 to 8 m depth). We do not consider the sand-mud limit on these beaches because this boundary is too deep (between 50 to 100m) to play a significant role in beach development. The presence of Posidonia, which can be regarded as a morphological limit of the beach profile, was regularly observed between the two selected depths, thus corroborating our choice of the value for the active slope of the profile.

10The flooding principle defined above thus enables us to simulate the theoretical retreat of the past shoreline, which is based on the slopes measured in the field and the sea-level rise recorded by tide-gauges during the 20th century (Suanezet al., 1997). To define the role of the sea-level rise forcing agent in controlling shoreline retreat, we compared theoretical values with the measured changes in shoreline position between 1896 and 1998.

2.3 - Prediction of future shoreline position taking into account the sea-level rise

11It is difficult to take into account the “sea-level rise” parameter in quantifying shoreline evolution because of the non-linear relationships between the forcing agents and the morphology (Stive and DeVriend, 1995). Each shoreline prediction methodology cited in the literature has its advantages and its drawbacks, so we decided to use three different scenarios to define the shoreline position for 2100.

12Scenario 1 is based on the past shoreline evolution in order to predict its future position. This method is based on the “end-point method” (Dolanet al., 1991) because we only take into account two dates to cover the 20th century. This method focuses on the shoreline position between two dates and considers all the climatic agents (waves, currents, wind and sea level) that influence the morphodynamics of the beach. Scenario 2 is based on the flooding concept (see previous section) and the results do not take into account erosional phenomena related to other forcings (longshore and cross-shore sediment transport). The future RSLR is based on IPCC (2001) prediction assuming a rate of about +4.4 mm/year. Scenario 3 takes into account the historical shoreline trend and the impact of sea-level rise on the shoreline position defined by scenario 2. Since the RSLR is implicitly considered between 1896 and 1998 (+1.1 mm/year), we propose subtracting the historical trend to avoid taking the sea-level parameter into account twice (Ferreiraet al., 2006).

3 - Results

3.1 - Measured shoreline changes during the 20th century

13Between 1896 and 1998, all the investigated pocket-beaches of Provence were undergoing erosion (Figs. 2 and 3). The shoreline of the 23 pocket-beaches examined retreated by an average of 12.1 (± 3.5) m, but this varied from one beach to another, from 2 (± 3.5) to 22 (± 3.5) m. The maximum retreat, 22 (± 3.5) m, was measured on Léoube beach, in the northeastern part of the Rade de Hyères, whereas the nearby beaches retreated by only 13 (± 3.5) to 11 (± 3.5) m (Pelegrin and Estagnole beaches). The investigated beaches lost an average of 40 % (± 10 %) of their surface area between 1896 and 1998.

3.2 - Impact of RSLR on the shoreline position during the 20th century

14For the pocket-beaches of Provence, the flooding method gives an average shoreline retreat of 5.8 (± 0.25) m, between 1896 and 1998, assuming a sea-level rise of + 11 cm over the same period (Suanez et al., 1997). The calculated minimum retreat is 3.3 (± 1.5) m on Rayol beach, while the maximum retreat is 8.6 (± 1.5) m for Pelegrin beach (Fig. 2). On average, these calculated values represent 50 % of the total measured retreat.

Fig. 2 - Shoreline retreat between 1896 and 1998 for the 23 investigated pocket-beaches.
Recul de la ligne de rivage entre 1896 et 1998 pour les 23 plages de poche étudiées.

Fig. 2 - Shoreline retreat between 1896 and 1998 for the 23 investigated pocket-beaches.Recul de la ligne de rivage entre 1896 et 1998 pour les 23 plages de poche étudiées.

3.3 - Simulation of the shoreline position in 2100

15The historical shoreline retreat (scenario 1) measured during the 20th century and extended to 2100 will result in an average retreat of 12.1 (± 3.5) m, with values varying between 2 (± 3.5) and 22 (± 3.5) m. The impact of the RSLR on the position of the shoreline by 2100 (scenario 2) will result in an average retreat of 23 (± 1) m, with values varying between 13 (± 1) and 35 (± 1) m. If we add the effect of the future sea-level position to the historical shoreline retreat (scenario 3), the position of the shoreline by 2100 will result in an average retreat of 29 (± 4.5) m, with values varying between 13.5 (± 4.5) and 46 (± 4.5) m.

16Compared with the present situation (1998), the narrowing of the beaches will result in a disappearance of beaches, with an average loss of 73 %, 95 % and 97 % in the case of scenario 1, 2 and 3 respectively (Fig. 4). Since most of the pocket-beaches are embanked (cliff or scarp above the back-shore), this limits the landward migration of the beach system. The worst situation would occur with the third scenario (RSLR and historical shoreline retreat), whereas the most “optimistic” results are obtained with the first scenario (historical trend of shoreline retreat). In any case, between 12 and 21 beaches will lose at least 75 (± 10) % of their current surface area by 2100.

Fig. 3 - Example of shoreline evolution: Langoustier beach (Porquerolles island). Shorelines indicated for 1896 (with bracket boundaries), 1998 and 2100 (scenario 2).
Exemple d’évolution de la ligne de rivage : plage du Langoustier (Ile de Porquerolles). Ligne de rivage 1896 (avec marges d’erreurs), 1998 et 2100 (scénario 2).

Fig. 3 - Example of shoreline evolution: Langoustier beach (Porquerolles island). Shorelines indicated for 1896 (with bracket boundaries), 1998 and 2100 (scenario 2).Exemple d’évolution de la ligne de rivage : plage du Langoustier (Ile de Porquerolles). Ligne de rivage 1896 (avec marges d’erreurs), 1998 et 2100 (scénario 2).

4 - Discussion

4.1 - Simulation of the shoreline position

17The simulation of future shoreline position is based on the past evolution of shorelines (scenario 1), which assumes that future forcing agents remain identical with time. By contrast, the hypothesis that climate changes will occur does not depend on constant forcing factors. This particular assumption does not give rise to a problem as long as the historical approach yields minimal values for shoreline retreat. The end-point method used here does not allow for any interannual estimation of the shoreline retreat as it was based on only two historical measurements. Moreover, on account of the low values of shoreline retreat, intermediate estimations would be contained within the analytical error brackets of this method.

18We used the basic concept of flooding to estimate the “sea-level rise” factor affecting the shoreline retreat, while distinguishing it from the numerous forcing agents controlling beach morphodynamics. This apparently simple mechanism can be used to estimate an order of magnitude of the relation between shoreline retreat and the sea level. Moreover, the simulations proposed for 2100 in scenario 2 do not take account of the erosion due to other forcing agents (waves and currents) or the feedback effects between morphology and forcing agents (Stive and DeVriend, 1995). Consequently, our results on the shoreline retreat due to sea-level rise represent minimum values, and the shoreline retreat is therefore likely to be even more marked. To compensate for this weakness, we propose a shoreline position for 2100 that includes the historical trend and the estimation of future sea level (scenario 3).

4.2 - The impact of RSLR on shoreline changes during the 20th century

19The comparison between measured (0.1 ± 0.03m/year) and calculated (0.05 ± 0.002m/year) average retreat suggests that sea-level rise plays an important role in shoreline retreat (fig. 2). This role is related to the fact that pocket beaches are protected from wave action, which is their main characteristic. Thus, the cross-shore and longshore processes are limited and the influence of sea-level rise can become evident.

20 The impact of sea-level rise is variable between the studied beaches, explaining 25% of the retreat at Rayol and 120% of the retreat at Pramousquier, with an average of 50 % for all the beaches studied. These differences are due to environmental settings that vary from one beach to another: grain size, swell exposure, slope of the beach, etc. Nevertheless, the effect of RSLR cannot be ignored, especially if we assume that it is accelerating.

4.3 - The future disappearance of pocket beaches in Provence

21On the pocket beaches of Provence, the presence of a break-of-slope “blocking” the landward migration of the profile will lead to a significant reduction in the width of the beach. The acceleration in sea-level rise predicted for 2100 will play an important role in the future position of the shoreline for such beaches because of their narrowness (5 to 47 m in 1998). Considering the rate of sea-level rise, we find that the profile slopes and the historical shoreline retreat data are insufficient to determine the vulnerability of the beaches. The coastal geomorphology, i.e. the morphology of the back-shore, should also be taken into account.

22For these three scenarios, we obtain the slowest shoreline retreat estimates with scenario 1, when only historical retreat is taken into account (implicitly including the sea-level rise of +11 cm during the 20th century). The difference between scenario 1 and scenarios 2 or 3 is not surprising because, in these latter cases, the sea-level rises 4 times faster than in scenario 1 (+ 4.4 mm/yr simulated during the 21th century, IPCC, 2001, as against + 1.1 mm/yr measured during the 20th century; Suanezet al., 1997). Therefore, our results are closely related to the amount of future sea-level rise, and the differences found between scenario 1 and scenario 2 would become lower if the sea-level rise does not reach the predicted value of + 0.44 m by 2100. Effectively, while this value is calculated for the global sea level, some regional differences can exist (Cabaneset al., 2001). Nevertheless, the measurements carried out by the satellites Topex Poseidon and Jason 1 from 1993 to 2006 confirm the global sea-level rise trend and give even higher values (up to 3.1 ± 0.4 mm/yr on average) (Neremet al., 2006). Consequently, we can infer that the actual sea-level rise is likely to be higher than the value used here and the narrowing of the pocket-beaches will be more pronounced.

Fig. 4 - Pocket-beach vulnerability for the three scenarios.
Vulnérabilité des plages de poche pour les trois scénarios.

Fig. 4 - Pocket-beach vulnerability for the three scenarios.Vulnérabilité des plages de poche pour les trois scénarios.

5 - Conclusion

23The pocket-beaches of Provence have suffered an average decrease of 40 (± 10) % in their surface area between 1896 and 1998, so it is reasonable to assume they will undergo significant erosion even at low rates of shoreline retreat (0.1 ± 0.03m/year). Since sea-level rise is an important factor influencing shoreline retreat, the acceleration in rate predicted for 2100 will play an important role in the future position of the shoreline on pocket-beaches that are protected from wave action and with limited back-shore areas (i.e. the beaches of Provence). Due to seal level rise, the width of the beaches is going to decrease significantly. Indeed, if we integrate the low rates of long-term retreat, and take into account the morphology of the back-shore, the pocket beaches appear to be threatened by disappearance due to sea-level rise. Under these conditions, sea-level rise will have an important socio-economic impact on the pocket-beaches of Provence. Our results can be extended to others areas where landward migration of the beaches is not possible. This analysis should enable politicians and environmental managers to make decisions based on a fuller knowledge of the coastal processes.

Top of page


BruunP., (1962), Sea-level rise as a cause of shore erosion, JournalofWaterwaysandHarbors,Division WW1, p. 117-130.

CabanesC.,CazenaveA.&LeProvost, C., (2001), Sea level rise during past 40 years determined from satellite and in situ observations, Science, 294, p. 840-842.

ChurchJ.-A.&White,N.-J., (2006), A 20th century acceleration in global sea-level rise, GeophysicalResearchLetters, 33, L01602.

Davidson-ArnottR.-G.-D., (2005), A conceptual model of the effects of sea level rise on sandy coasts, JournalofCoastalResearch, 21 (6), p. 1166-1172.

DolanR.,FensterM.-S.&HolmeS.-J., (1991), Temporal analysis of shoreline positions: a test of algorithms, JournalofCoastalResearch, 3 (1), p. 723-744.

EmeryK.-O.,AubreyD.-G.&GoldsmithV., (1988), Coastal Neo-tectonics of the Mediterranean from Tide-gauge Records, MarineGeology, 81, p. 41-52.

FerreiraO.,GarciaT.,MatiasA.,TabordaR.&DiasJ., (2006), An integrated method for the determination of set-back lines for coastal erosion hazards on sandy shores, ContinentalShelfResearch, 26, p. 1030-1044.

GornitzV.,LebedeffS.&HansenJ., (1982), Global Sea Level in the Past Century, Science, 215, p. 1611-1614.

HallermeierR.-J., (1981), A profile zonation for seasonal sand beaches from wave climate, Coastalengineering, 4: p. 253‑277.

Intergovernmental Panel on Climate Change, (2001), Climate change 2001: Impacts, adaptation and vulnerability, –Contribution of the working group to the third assessment report of the intergovernmental Panel of Climate Change, World Meteorological Organization, Genève, 124 p.

JefticL.,MillimanJ.-D.&SestiniG., (1992), Climatic change and the Mediterranean: Environmental and societal impacts of climatic change and sea-level rise in the Mediterranean region. Edward Arnold, London, XII, 673 p.

JimenezJ.-A.&Sanchez-ArcillaA., (1997), Physical impacts of climatic change on deltaic coastal systems (II): driving terms, ClimaticChange, 25, p. 95-118.

NeremR.-S.,LeulietteE.,&CazenaveA., (2006), Present-day Sea Level Change: A Review, ComptesrendusGeoscience, 338, p. 1077–1083.

PilkeyO.-H.,Young,R.-S.,RiggsS.-R.&SmithA.-W., (1993), The concept of shoreface profile of equilibrium: a critical review, JournalofCoastalResearch, 9 (1), p. 255-278.

PilkeyO.-H. & Cooper J.-A.-G., (2004), Society and Sea Level Rise, Science, 303, p. 1871-1782.

SabatierF.;StiveM.,&PonsF., (2004), Longshore variation of depth closure on a micro-tidal wave-dominated coast, InternationalConferenceofCoastalEngineering2004, American Society of Civil Engineering, Lisboa, p. 2329-2339.

Suanez S., Prosper-Laget V. & Provansal M., (1997), Variations relatives du niveau marin dans le delta du Rhône et à Marseille, implications tectoniques et/ou climatiques, C. R. Acad. Sci. Paris, 324 (IIa), p. 639-646.

Thieler E.-R., Pilkey O.-H. Jr., Young R.-S.; Bush D.-M. & Chai F., (2000), The use of mathematical models to predict beach behavior for coastal engineering, A critical review, Journal of Coastal Research, 16 (1), p. 48-70.

Top of page

List of illustrations

Title Fig. 1 - Location of study sites.Localisation des sites étudiés.
File image/png, 35k
Title Fig. 2 - Shoreline retreat between 1896 and 1998 for the 23 investigated pocket-beaches.Recul de la ligne de rivage entre 1896 et 1998 pour les 23 plages de poche étudiées.
File image/png, 50k
Title Fig. 3 - Example of shoreline evolution: Langoustier beach (Porquerolles island). Shorelines indicated for 1896 (with bracket boundaries), 1998 and 2100 (scenario 2).Exemple d’évolution de la ligne de rivage : plage du Langoustier (Ile de Porquerolles). Ligne de rivage 1896 (avec marges d’erreurs), 1998 et 2100 (scénario 2).
File image/jpeg, 40k
Title Fig. 4 - Pocket-beach vulnerability for the three scenarios.Vulnérabilité des plages de poche pour les trois scénarios.
File image/png, 28k
Top of page


Bibliographical reference

Cédric Brunel and François Sabatier, « Provence pocket beach erosion », Méditerranée, 108 | 2007, 77-82.

Electronic reference

Cédric Brunel and François Sabatier, « Provence pocket beach erosion », Méditerranée [Online], 108 | 2007, Online since 01 January 2009, connection on 29 June 2022. URL : ; DOI :

Top of page

About the authors

Cédric Brunel

Université de Provence - CEREGE Europôle de L’Arbois - BP 80 – France -

François Sabatier

Université de Provence - CEREGE Europôle de L’Arbois - BP 80 – France -

By this author

Top of page


Tous droits réservés

Top of page
  • Logo Presses Universitaires de Provence
  • OpenEdition Journals
Search OpenEdition Search

You will be redirected to OpenEdition Search