Navigation – Plan du site

AccueilNumérosvol. 28 - n° 4Statistical approach of contempor...

Statistical approach of contemporary hydrogeomorphological channel changes of the middle Allier River, France: morphostructural controls, human impacts and flow regime

Approche statistique des changements hydrogéomorphologiques récents de la rivière Allier moyen (France) : contrôles morphostructuraux, impacts anthropiques et régime hydrologique
Arfeuillère Anaïs, Johannes Steiger, Erwan Roussel, Stéphane Petit, Catherine Neel et Emmanuèle Gautier
p. 241-256

Résumés

La rivière Allier, France, connue pour ses sections très dynamiques dans son cours inférieur, a néanmoins été soumise à des impacts anthropiques. La présente étude, centrée sur son cours moyen, a pour objectif de comprendre les changements hydrogéomorphologiques récents de l'Allier moyen en relation avec les contraintes morphostructurales, les impacts anthropiques et le régime hydrologique depuis la seconde moitié du XXe siècle. Les changements morphologiques sont analysés à partir de six orthophographies aériennes (1954, 1974, 1992, 2000, 2010 et 2016) et de quatre relevés de ligne d’eau (1934/35, 1999, 2005 et 2020). Des analyses statistiques descriptives et une analyse en composantes principales suivie d’une classification ascendante hiérarchique permettent de définir des types de trajectoires morphologiques. L’influence des contraintes morphostructurales et des impacts anthropiques est testée au moyen d’une analyse factorielle des correspondances et la tendance d’évolution du régime hydrologique par les tests statistiques de Mann Kendall et de Pettitt. Les résultats montrent qu’au cours des soixante à quatre-vingts dernières années, le lit s’est incisé de 0,86 m, s’est rétréci de 25% et que la mobilité latérale du chenal a diminué de 68%. Quatre trajectoires d'évolution morphologiques ont été identifiées pour les sections de plaines alluviales, plus mobiles latéralement que la section naturellement confinée. Ces trajectoires d’évolution ont été caractérisées par des impacts anthropiques d’intensité variable, parmi lesquels l’extraction de granulats et, plus localement, de travaux dans le chenal. De plus, le changement hydrologique enregistré à la fin du XXe siècle, caractérisé par une baisse de la fréquence, de la durée et de l'intensité des crues, a conduit à une augmentation généralisée des surfaces abandonnées annuellement et a contrario à une baisse des surfaces érodées annuellement. En conséquence, ces modifications hydrogéomorphologiques devront être prises en compte dans les futurs plans de gestion de la rivière.

Haut de page

Notes de la rédaction

Received 8 April 2022, received in revised form 29 August 2022, accepted 7 October 2018.

Texte intégral

We acknowledge the financial support of the Auvergne-Rhône-Alpes Region for the PhD thesis of the first author of the research project “RALLIER – Water Resources, Agriculture and Alluvial Forest: Incision and Degradation of the Allier River”, the Agence de l’eau Loire-Bretagne, Établissement Public Loire, SAGE Allier Aval and University Clermont Auvergne CAP20-25. We also thank the two reviewers for their constructive comments and suggestions.

1. Introduction

1Rivers have always been important natural resources for societies. But human pressures on European rivers progressively increased over long timescales (Petts et al., 1989; Gautier et al., 2007; Ejarque et al., 2015). Many authors have documented increased geomorphological changes of European and French rivers during the second half of the 20th century (Wyżga, 1991; Darby and Thorne, 1992; Steiger et al., 1998; Liébault and Piégay, 2002; Surian and Cisotto, 2007; Comiti et al., 2011; Ziliani and Surian, 2012; Grivel et al., 2018). The increased rate of channel changes and fluvial geomorphological responses reflects the mutation of societies during the 20th century. Within upstream river catchments, the fallow of abandoned agricultural lands, reforestation of hillslopes and torrent control works have been carried out since the end of the 19th century, which concomitantly led to the reduction of sediment sources (Liébault and Piégay, 2001 ; Boix-Fayos et al., 2007; Preciso et al., 2012). Concurrently, gravel mining, in the 1950s and more intensely during the 1970s and 1980s, first in river channels and then within floodplains, and the construction of dams in the upper catchments for power and water supply decreased sediment supply to downstream reaches and sometimes also regulated flow regimes (Kondolf, 1997). Moreover, river engineering works, i.e. direct modifications of channel geometry (straightening/realignment, widening and/or deepening) and the construction of riverbank protections, destabilized river equilibrium by increasing their transport capacity (Gregory, 2006). As a response to these direct and indirect human impacts, alluvial rivers seek to adjust their channel geometry through channel incision (Landon and Piégay, 1994; Surian and Cisotto, 2007; Wyżga, 2007), channel narrowing (Surian and Cisotto, 2007; Comiti et al., 2011; Jantzi et al., 2017) and river style change, i.e. river metamorphosis, for example the transition from braided to single channel rivers (Comiti et al., 2011; Salit et al., 2015). These geomorphological channel adjustments often lead to hydrological disconnection of the main channel from its floodplain (Wyżga, 2001) and to planform stability (Dépret et al., 2017). Even though many studies mainly focused on human impacts, several authors pointed out that these strong modifications add to climatic factors which control the flow regime (Gautier, 1994; Warner, 2000).

2The Allier River, in France, known for its lateral mobile and very dynamic river sections in its lower course (Geerling et al., 2006; Petit, 2006; Garófano Gómez et al., 2017), has nevertheless been subjected to human impacts and channelization processes, mainly in its middle section. An overall study (DIREN Auvergne, 1998) from the outlet of its upstream gorges to the confluence with the Loire River observed a generalized channel incision between -0.5 m and -3.5 m in more than 70% of this alluvial river section between 1933/35 and 1980/1995, attesting a bedload deficit. The Allier River is recognized as one of the main sedimentary tributaries of the Loire River (Babonaux, 1970). Its bedload deficit therefore directly impacts morphodynamics of the Loire River downstream of the Allier-Loire confluence, which itself is subjected to a significant bedload reduction (Gasowski, 1994; Nabet et al., 2016). Changes in bedload dynamics, one of the two main control variables of channel adjustments within alluvial rivers at present and contemporary timescales (Schumm, 1977), trigger adjustments of channel geometry as well as channel planform and river style changes. Whereas several studies investigated channel planform changes of the lower Allier River (Geerling et al., 2006; Petit, 2006; Peiry et al., 2009; Garófano Gómez et al., 2017), little information is available for the middle Allier River, downstream of the main sediment production area of the catchment, i.e. the Allier River gorges. The present study therefore focuses on human impacts since the second half of the 20th century and their consequences along the longitudinal continuum of the middle course of the Allier River. Channel planform and vertical channel bed dynamics and adjustments were investigated, also considering the role played by the flow regime and its current evolution in the light of ongoing climate change. More precisely, the aims were (i) to assess planform and vertical bed channel changes of the middle Allier River since the second half of the 20th century; (ii) to determine morphological evolutionary trajectories; (iii) to statistically explore the relationships between channel changes, morphostructural controls and human impacts and (iv) to analyze the flow regime since the 1950s.

2. Study area

3From its source in Lozère (1,485 m a.s.l.) to its confluence with the Loire River downstream of Nevers (167 m a.s.l.), the Allier River has a length of 410 km. It drains a catchment of 14,400 km² in the Massif Central, France, (fig. 1A) and its flow regime is subjected to oceanic, continental and Mediterranean climate influences (Onde, 1923). The longitudinal continuum exhibits several alternating of varying widths such as ‘beads on the string’ (Naiman et al., 2005). The middle course of the Allier River is delineated between Vieille-Brioude, located downstream of the gorges in the upper catchment, and the confluence with the Dore River at Limons, a major right bank tributary (fig. 1B). This study area of 125 km can be subdivided into three distinct morphostructural sections (tab. 1), corresponding to the dislocation of the Hercynian peneplain (mainly constituted of granite and gneiss) into horsts and grabens during the Oligocene (Merle and Michon, 2001), subsequently filled by marl and limestone. The first and third sections within tectonical grabens are characterized by a single sinuous channel with a specific stream power of 59 and 77 W/m2, respectively, whereas Section 2 corresponds to a geological granitic horst incised by the Allier River. In this study area, the flow regime of the Allier River, a pluviometric regime, is characterized by highest flows in spring due to rainfall induced by storm events and a minimum in summer due to anticyclonic atmospheric conditions (fig. 1B).

Fig. 1 – Location maps.
Fig. 1 – Cartes de localisation.

Fig. 1 – Location maps.Fig. 1 – Cartes de localisation.

A: Allier River catchment and location of the study area. B: The middle course of the Allier River between Vieille-Brioude and Limons shows three distinct morphostructural sections. Mean monthly discharges characterize the mean flow regime of each section (calculated from 1919 to 2019 at Vieille-Brioude; from 1984 to 2019 at Coudes and from 1974 to 2019 at Limons; discharge data: Banque Hydro). 1. floodplain limits; 2. limits of morphostructural sections; 3. towns with gauging station ; 4. main towns. French coordinate system Lambert 93.
A : Bassin versant de la rivière Allier et localisation du secteur d’étude. B : L’Allier moyen entre Vieille-Brioude et Limons montre trois sections morphostructurales. Les débits mensuels moyens caractérisent le régime d'écoulement moyen de chaque section (calculés de 1919 à 2019 à Vieille-Brioude ; de 1984 à 2019 à Coudes et de 1974 à 2019 à Limons ; données hydrologiques : Banque Hydro). 1. limites du lit majeur ; 2. limites des sections morphostructurales ; 3. villes avec une station hydrologique ; 4. villes principales. Système de coordonnées Lambert 93.

Tab. – Mean values of main morphological (based on orthophotograph from 2016) and hydrological (Banque hydro) characteristics of the three morphostructural sections within the study area.
Tab. 1 – Valeurs moyennes des principales charactéristiques géomorphologiques (basée sur l’orthophotographie de 2016) et hydrologiques (données de la Banque Hydro) des trois sections morphostructurales du secteur d’étude.

Tab. – Mean values of main morphological (based on orthophotograph from 2016) and hydrological (Banque hydro) characteristics of the three morphostructural sections within the study area.Tab. 1 – Valeurs moyennes des principales charactéristiques géomorphologiques (basée sur l’orthophotographie de 2016) et hydrologiques (données de la Banque Hydro) des trois sections morphostructurales du secteur d’étude.

3. Methods

4The applied methodological framework can be divided into three main stages: (i) identification of morphological channel changes and evolutionary trajectories; (ii) analysis of the role of morphostructural constraints and human impacts on morphological changes; and (iii) analysis of flow regime change (fig. 2).

Fig. 2 – Methodological framework.
Fig. 2 Cadre méthodologique.

Fig. 2 – Methodological framework.Fig. 2 – Cadre méthodologique.

3.1. Channel change analysis

3.1.1. Data compilation of channel planform changes and altitudinal channel bed changes

5The planform analysis was carried out based on six historical aerial photographs (tab. 2, fig. 3) provided by the French “Institut national de l’information géographique et forestière” (IGN) database. The 1954, 1974 and 1992 sets of aerial photographs were orthorectified with ®Agisoft Metashape and georeferenced in Lambert 93 projection (EPGS: 2154). The orthophotographs produced by IGN were used for the years 2000, 2010 and 2016. The georectification process was evaluated according to the method described in Hughes et al. (2006) where the georeferencing error corresponds to the 90th percentile of 71 ground-control points. The active channel, defined as the wetted channel including bare and sparsely vegetated (i.e. herbaceous and pioneer woody vegetation) alluvial channel bars frequently reworked by floods (Petit, 2006; Surian and Cisotto, 2007), was digitized on each date using the GIS software ®QGIS 2.18. The mean error of digitization was evaluated according to Downward et al. (1994). Total errors, i.e. georeferencing plus digitization errors, amount to 10 m (1954), 11 m (1974), 9 m (1992), 7 m (2000), 6 m (2010) and 3 m (2016). A historical meander-belt, which encompasses the active channel boundaries at the six dates, was created and cut into 125 segments of 1 kilometer each, to robustly compare the morphological indices over time (David et al., 2016). Three planform indices were computed. The active channel width was computed every 100 m, perpendicularly to the centerline of the channel and a mean value was computed for each segment. In order to assess lateral channel dynamics, the eroded and abandoned areas between two consecutive dates, expressed in square meter per year, were computed by taking the total error into account (Urban and Rhoads, 2003).

Tab. 2 – Aerial photographs and orthophotographs used for the analysis of planform channel changes.
Tab. 2 – Campagnes photographiques et orthophotographies utilisées pour l’analyse des changements de la forme en plan.

Tab. 2 – Aerial photographs and orthophotographs used for the analysis of planform channel changes.Tab. 2 – Campagnes photographiques et orthophotographies utilisées pour l’analyse des changements de la forme en plan.

Fig. 3 – Examples of the series of photographs used and planform changes between 1954 and 2016.
Fig. 3 – Exemples de chaque photographie utilisée et des changements de la forme en plan entre 1954 et 2016.

Fig. 3 – Examples of the series of photographs used and planform changes between 1954 and 2016.Fig. 3 – Exemples de chaque photographie utilisée et des changements de la forme en plan entre 1954 et 2016.

6To reconstruct altitudinal channel bed change, four longitudinal water surface profiles, dating from 1934/35, 1999, 2005, and 2020, were compared (tab. 3). It is considered that water level change between two dates translates into altitudinal channel bed change. Before the comparison, the elevation datum of the 1934/35 long profile was converted from the original NGF-Lallemand into the currently used altitudinal system NGF IGN69. The altitudinal evolution of the channel bed between 1934/35 and 2020 was analysed according to the previous division of the river channel into 125 segments. For each segment, the mean value of vertical variation was computed when two or more points of comparison were available. Altogether, 117 segments out of 125 could be compared between 1934/35 and 2020, whereas between 1999 and 2005 and between 2005 and 2020, respectively 50 and 44 points for the comparison of water levels were available. For each comparison, vertical changes induced by differences of stage heights during the field surveys were subtracted, in order to robustly quantify altitudinal channel bed changes. For this purpose, stage heights corresponding to the discharges during the different field surveys were determined from the rating curves (available for the period of the fields survey) of the nearest gauging station for every point measured.

Tab. 3 – Longitudinal water surface profiles used for the analysis of vertical channel bed changes.
Tab. 3 – Profil longitudinaux utilisées pour l’analyse des changements altitudinaux du lit.

Tab. 3 – Longitudinal water surface profiles used for the analysis of vertical channel bed changes. Tab. 3 – Profil longitudinaux utilisées pour l’analyse des changements altitudinaux du lit.

3.1.2. Planform and altitudinal channel bed change analyses

7First, descriptive statistical analyses were carried out, combined with a Wilcoxon signed rank test (paired sample) testing the statistical significance of planform indices changes between dates or periods. Second, the changing trend between 1954 and 2016, for each segment and each planform index, was summarized applying a simple linear regression. For the active channel width index, the linear regression coefficient expressed the trend of widening (>0) or narrowing (<0) of the channel and for the two lateral channel dynamics indices (annual eroded area and annual abandoned area), the linear regression coefficient expressed the increase (>0) or the decrease (<0) of the annual area. Finally, in order to determine different morphological evolutionary trajectories, these planform indices describing tendencies of planform dynamics and the mean altitudinal channel bed changes between 1934/35-2020 determined for each segment were employed to perform a Principal Component Analysis (PCA) and Hierarchical Cluster Analysis (HCA; David et al., 2016), using ®Rstudio and the package FactomineR.

3.2. Analysis of control factors

3.2.1. Morphostructural constraints and human impacts

8The more or less intense relationship between morphological evolutionary trajectories and morphostructural constraints on the one hand and human impacts on the other hand was tested through a Correspondence Factor Analysis (CFA). The number of segments which recorded the presence of gravel mining sites, the presence of sills, the presence of in-channel engineering works, the degree of bank protections divided into five classes (0-20%, 20-40%, 40-60%, 60-80%, 80-100%) and the degree of valley confinement divided into five classes, were determined for each distinct morphological evolutionary trajectory. This information was derived from several existing databases and other sources (tab.  4), except for the degree of valley confinement which was calculated using the method of Roux et al. (2015). The degree of bank protections, i.e. riprap in most cases, is expressed, in this study, as the percentage of total length of riverbank protections divided by the total length of banks on both sides in 2016.

Tab. 4 – Data sources used for the determination of human pressures and impacts on channel morphology within the study area.
Tab. 4 – Sources de données utilisées pour la détermination des pressions humaines sur le chenal dans la zone d’étude.

Tab. 4 – Data sources used for the determination of human pressures and impacts on channel morphology within the study area.Tab. 4 – Sources de données utilisées pour la détermination des pressions humaines sur le chenal dans la zone d’étude.

3.2.2. Flow regime changes

9The theoretical bankfull discharge (return period T of 1.5 yrs) was calculated on the basis of the annual maximum flow approach (AMF) and a Gumbel adjustment (Meylan et al., 2008) of the daily mean discharge timeseries from the Vieille-Brioude gauging station located at the upstream end of the study area. Subsequently, all daily mean discharges between 1954 and 2016 equal to or above the theoretical bankfull discharge, defined as the most probable annual flood (Chorley, 1969), were selected to create four indices characterizing flood regime, i.e. the maximum annual daily mean discharge, number of floods per year, number of days of flooding per year and daily mean discharge exceeding the theoretical bankfull discharge (T: 1.5 yrs). The Water Resources Council method (Lang et al., 1999) was applied to distinguish two consecutive flood events, which considers two separate flood events if 5 days + log (area basin in miles²) separate the two peak floods. In addition to the flood analysis, the number of days presenting discharges lower than the inter-annual mean discharge were used to characterize the low water regime. A descriptive analysis was carried out in order to assess trends of change of the five flow indices according to the time periods used for the channel planform analysis. Subsequently, a Mann Kendall test followed by a Pettitt test was carried out in order to detect potential trends and points of change in the time series (Renard et al., 2006) using the Trend package in ®Rstudio.

4. Results

4.1. Channel changes at the study area scale

10Between 1954 and 2016, all sections combined, the median and the interquartile range of the active channel width index showed a decreasing trend with its strongest reduction recorded between 2010 and 2016 (fig. 4). The median of annual eroded area only showed a significant reduction between the two last periods, combined with a decrease of the interquartile range value. The median of annual abandoned areas showed an overall increase trend during the entire study period, which however, was not significant between 1954-1974 and 1974-1992, and showed a temporary decrease during the 2000-2010 period. Between 1934/35 and 2020, channel incision prevailed with an overall mean of -0.86 m and a median of -0.49 m. Maximum values ranged between an incision of -6.27 m and a local aggradation of 0.97 m. For the two most recent periods observed (1999-2005 and 2005-2020), both median values are zero or near zero (0 m between 1999 and 2005; 0.06 m between 2005-2020) with interquartile range values of 0.10 m and 0.22 m, respectively.

Fig. 4 – Morphological indices for different dates and periods.
Fig. 4 – Indices morphologiques pour les différentes dates ou périodes.

Fig. 4 – Morphological indices for different dates and periods.Fig. 4 – Indices morphologiques pour les différentes dates ou périodes.

From left to right: active channel width (m), annual eroded area (m2), annual abandoned area (m2), and altitudinal channel bed change (m). Within the boxplot, the horizontal black line represents the median, the upper edge of the boxplot represents the 75th and the lower edge of the boxplot represents the 25th percentile. Results of the degree of significant of Wilcoxon test’s p-value: * for p-value comprise between 0.05 and 0.01, ** between 0.01 and 0.001, *** between 0.001 and 0.0001, **** between 0.0001 and 0.00001.
De gauche à droite : largeur du chenal actif (m), surface érodée annuellement (m2), surface abandonnée annuellement (m2) et changement altitudinal du lit (m). Dans le diagramme en boîte, la ligne noire horizontale accentuée représente la médiane, le bord supérieur du boxplot représente le 75e percentile et le bord inférieur du boxplot représente le 25e percentile. Résultats du degré de significativité de la p-value du test de Wilcoxon : * pour une p-value comprise entre 0,05 et 0,01, ** entre 0,01 et 0,001, *** entre 0,001 et 0,0001, **** entre 0,0001 et 0,00001.

4.2. Channel changes at the morphostructural section scale

11Whereas Section 1 did not show a significant difference in its median active channel width between 1954 and 2016 (fig. 5), the median active channel width recorded a fluctuation of 8 m during the entire study period in Section 2 and a significant decrease between 1992 and 2000, as well as between 2010 and 2016 occurred in Section 3. Within the three sections, the median annual eroded area showed a significant decrease between the two last periods, and the median annual abandoned area showed significant fluctuations between the periods 1974-1992 and 2010-2016. However, these fluctuations recorded are more important within Sections 1 and 3 than within Section 2. The median altitudinal channel bed change showed no significant variations in Section 2, with a median value of -0.06 m. Conversely, the two others sections recorded a median value of -0.7 m (Section 1) and -0.59 m (Section 3) and an interquartile range of more than 1 m in both cases.

Fig. 5 – Morphological indices for different dates and periods.
Fig. 5 – Indices morphologiques pour les différentes dates ou périodes.

Fig. 5 – Morphological indices for different dates and periods.Fig. 5 – Indices morphologiques pour les différentes dates ou périodes.

4.3. Identification of morphological evolutionary trajectories

12The first and second principal factors of the PCA analysis represent 69.7% of the total variance (38.2% for F1 and 31.5% for F2) (fig. 6A). F1 is mainly positively correlated with channel widening (positive channel width change), and F2 is positively correlated with channel bed aggradation (positive altitudinal channel bed change) and negatively correlated with the annual increase of abandoned and eroded areas. The tendency of the active channel width is correlated with all indices whereas the indices of abandoned and eroded areas, which are strongly correlated with each other, are not correlated with the altitudinal channel bed change. The dendrogram of the cluster analysis was cut into four clusters, according to a Euclidean distance and the Ward method (fig. 6B). Cluster 1 (9 segments) is characterized by the highest incision and active channel width narrowing values, whereas Cluster 2 (22 segments) is characterized by channel narrowing as well as a decrease of annual eroded areas. In turn, Cluster 3 (76 segments) is characterized by the lowest rates of planform and altitudinal channel bed changes during the entire study period, while Cluster 4 (18 segments) is characterized by channel incision in addition to increasing annual abandoned areas.

Fig. 6 – Factor analysis.
Fig. 6 – Analyse factorielle.

Fig. 6 – Factor analysis.Fig. 6 – Analyse factorielle.

A: Variable graph of the PCA analysis; B: Individual graph of the PCA analysis with the classification of the Cluster Analysis; C: Map representation of the 125 segments based on their cluster affiliation and representation of the part of the four clusters in each of the morphostructural sections.
A : Graphe des variables de l’analyse ACP ; B : Graphe des individus de l’analyse ACP avec la classification de l’analyse CAH ; C : Représentation cartographique des 125 segments en fonction de leur appartenance à un cluster et représentation de la part des quatre clusters dans chacune des sections morphostructurales.

13Section 1 is composed of a majority of segments belonging to Cluster 3 (62%) which are located especially in the middle part of this section (fig. 6C). Clusters 2 and 3 represent 17% each and Cluster 1 only 4% of all 47 segments belonging to Section 1. Section 2 is exclusively constituted of segments belonging to Cluster 3. Section 3, just as Section 1, is mainly composed of segments belonging to Cluster 3 (47%), which are located preferentially at the upstream and downstream ends of the section, in association with segments belonging to either Cluster 2 (24%) or 4 (17%). Segments belonging to Cluster 1 (12%) are observed in the middle of Section 3.

4.4. Morphostructural constraints and human impacts

14Valley confinement primarily concerns Section 2 with a confinement index comprised mainly between 20-40% (fig. 7). In the two other sections, the valley confinement degree is less marked, ranging mainly between 0 and 20%. Conversely, bank protections are longer in Sections 1 and 3, especially in the upstream part of Section 3, attaining more than 60% for some of the segments. Among human activities, gravel mining is preponderant with 61% of segments affected by this industry, especially in Sections 1 and 3. In addition to gravel mining, the study area counts nine sills: four in Section 1, one at the downstream part of Section 2 and four in the middle part of Section 3. The in-channel engineering, which refers to channel straightening and river training works modifying the channel cross-section, are only located in Section 3, at three different locations.

Fig. 7 – Degree of valley confinement and human impacts.
Fig. 7 – Degré de confinement dans la vallée et impacts anthropiques.

Fig. 7 – Degree of valley confinement and human impacts.Fig. 7 – Degré de confinement dans la vallée et impacts anthropiques.

Degree of bank protections, gravel mining sites, sills; in-channel engineering works such as channel straightening and river training works, etc., for each of the 125 channel segments within the study area.
Degré de berges protégées, sites d’extractions de granulats, seuils, travaux dans le chenal comme le recoupement de méandres et la modification de la géométrie du chenal, etc., pour chacun des 125 segments du secteur d’étude.

15The first axis of the Correspondence Factor Analysis represents 74.58% of the variance and the second axis 17.85%, i.e., a total variance of 92.43% (fig. 8). The eigenvalues of 0.17 (F1) and 0.041 (F2) indicated the absence of an exclusive relationship between a morphological evolutionary trajectory and morphostructural constraints or human impacts. The first axis on the CFA shows an opposition between Cluster 1 with the three other clusters. The latter are discriminated on the second axis, with Cluster 3 opposite to Clusters 2 and 4. Cluster 1 is linked to the presence of in-channel engineering works and the strongest degree of bank protections, whereas Clusters 2 and 4 are associated to a wide valley but the presence of bank protections (20-40% or 60-80% respectively) and the presence of gravel mining sites. Cluster 3 is associated with different levels of confinement in the valley.

Fig 8 – Correspondence Factor analysis.
Fig 8 – Analyse factorielle des correspondances.

Fig 8 – Correspondence Factor analysis.Fig 8 – Analyse factorielle des correspondances.

4.5. Hydrological regime changes between 1954 and 2016

16Reduction in the number of flood events per year (fig. 9), as well as a decrease in the average number of days of flooding per year, were observed between 1954 and 2016. Concomitantly, an increase of the average number of days below the inter-annual mean discharge occurred throughout the study period. The highest number of high magnitude floods including two 50-yr floods (in 1973 and in 1976) was recorded between 1974-1992, followed by one 10-yr flood event in 1983 and one 20-yr flood event in 1984. Conversely, only low magnitude flood events occurred during the last period (2010-2016), with three 1.5-yr floods and two 2-yr floods. The Mann–Kendall test identified two significant trends in water discharge indices between 1954 and 2016 (tab. 5): the decrease of the maximum annual discharge and the increase of the number of days below the inter-annual discharge. For these two indices, the point of change is statistically estimated for 1984 and 1983, respectively.

Fig. 9 – Daily mean discharge (m3/s) of the Allier River at the Vieille-Brioude gauging station between 1954 and 2016.
Fig. 9 – Débits journaliers moyens (m3/s) de la rivière Allier à la station hydrologique de Vieille-Brioude de 1954 à 2016.

Fig. 9 – Daily mean discharge (m3/s) of the Allier River at the Vieille-Brioude gauging station between 1954 and 2016.Fig. 9 – Débits journaliers moyens (m3/s) de la rivière Allier à la station hydrologique de Vieille-Brioude de 1954 à 2016.

discharge data: Banque Hydro, https://hydro.eaufrance.fr.
données hydrologiques : Banque Hydro, https://hydro.eaufrance.fr.

Tab. 5 – Mann Kendall and Pettitt's test results for five characteristic flow indices.
Tab. 5 – Résultats des tests statistiques de Mann Kendall et Pettitt des 5 indices hydrologiques.

Tab. 5 – Mann Kendall and Pettitt's test results for five characteristic flow indices.Tab. 5 – Résultats des tests statistiques de Mann Kendall et Pettitt des 5 indices hydrologiques.

Qmax/yr = maximum annual daily mean discharge; Q>Q1.5yr = daily mean discharge superior or equal to the bankfull discharge (return period: 1.5 yr); Nb floods/yr = number of floods per year; Nb days of floods/yr = the number of days of floods per year; Nb days<Qmean = number of days with daily mean discharge lower than the inter-annual mean discharge.
Qmax/yr = débit journalier moyen maximal par an ; Q>Q1.5yr = débits journaliers moyens supérieurs ou égaux au débit à pleins bords (période de retour : 1,5 an) ; Nb flood/yr = nombre de crue par an ; Nb day flood/yr = nombre de jours en crue par an ; Nb days<Qmean = nombre de jours où le débit journalier moyen est en-dessous du module.

5. Discussion

5.1. Morphological channel changes within nested spatial scales

17Since the second half of the 20th century, the middle Allier River, between Vieille-Brioude and Limons (fig. 1), has adjusted its vertical and planform dimensions to increased human impacts on its geomorphology and bed load transport (fig. 4). Between 1934/35 and 2020, the Allier River incised its channel bed by 0.86 m on average. Narrowing of the active channel by 25% occurred between 1954 and 2016 and lateral channel mobility decreased by 68% between the first (1954-1974) and the most recent period (2010-2016) considered. These morphological channel changes are consistent with general observed trends for European rivers since the 20th century in response to increased human disturbances and direct impacts on channel morphodynamics (Gautier, 1994; Landon and Piégay, 1994; Steiger et al., 2000; Liébault and Piégay, 2002; Surian and Rinaldi, 2003; Wyżga, 2007) and on the Loire River (Gasowski, 1994; Gautier et al., 2000; Nabet et al., 2016; Grivel et al., 2018), of which the Allier River is a major tributary.

18At the morphostructural section scale, contrasting morphological channel changes were observed between the gorge section (Section 2) and alluvial plain sections (Sections 1 and 3). According to the boxplot analysis and the factorial analyses (fig. 5-6), the gorge section recorded much less temporal changes of channel planform and altitudinal channel bed change than the downstream and upstream alluvial plain sections. These differences in channel changes can be related to the morphostructural context which controls the capacity of the river and its channel to adjust (Fryirs and Brierley, 2013). In its gorge section, the Allier River flows through a confined valley, which inhibits extensive contemporary hydrogeomorphological changes over short timescales. In addition to the natural valley confinement, the railway on the right bank (dating from the mid-19th century) and the motorway on the left bank (dating from the second half of 20th century) further confine the river corridor within the gorges. Conversely, the two alluvial plain sections are constituted by an extensive floodplain with a mean width of 1 km where the channel geometry may adjust vertically and laterally much more freely than in the gorge section (except the reaches where bank protections hinder lateral erosion).

19In addition, four clusters, i.e., four distinguished morphological evolutionary trajectories, were identified within the two alluvial plain sections (fig. 6) and associated with various human impacts (fig. 8). In the plain sections, some areas were less affected by morphological channel changes than others, such as the middle part of Section 1, or, to a lesser extent, the upstream and the downstream ends of Section 3, where a majority of segments belongs to Cluster 3. Cluster 3 groups segments presenting the lowest rates of planform and vertical channel bed changes, mainly due to bank protections. On the other hand, the upstream and downstream parts of Section 1 and the middle of Section 3 underwent more morphological channel changes identified by segments belonging to Cluster 2 (morphological evolutionary trajectory marked by a trend towards an active channel narrowing and a decrease of annual eroded areas) and Cluster 4 (morphological evolutionary trajectory marked by channel incision and a trend to an increase of annual abandoned areas). These two clusters are both associated with the presence of gravel pits and significant degrees of bank protections ranging between 20-40% or 60-80%. Cluster 1, which counts only nine segments out of a total of 125, is less represented than the other morphological evolutionary trajectories, i.e. Clusters 2, 3 and 4. It is characterized by more pronounced morphological changes (highest incision and active channel width narrowing values) associated with engineering works, i.e. direct human impacts on channel and planform geometry. For example, middle Section 3 was impacted by large gravel mining pits and channel straightening works between 1974 and 1992. These were carried out simultaneously to the building of a motorway bridge. The bridge was accompanied by the construction of sills to stabilize the channel bed. A maximum incision of -6.25 m and an active channel width loss up to 122 m were recorded in this area between 1934/35 and 2020. In 1975, Clavel et al. (1977) observed already a loss in diversity of geomorphic and hydraulic channel units as well as an immediate impact upstream of this zone due to an increase of the fall at 1.5 m of a riffle formed by bedrock outcrops.

5.2. Channel changes and gravel mining

20As a response to bedload reduction, for example caused by gravel mining, channels adjust their long profile through progressive and regressive erosion processes in order to recover an equilibrium between their transport capacity and the available bedload (Kondolf, 1997; Rinaldi et al., 2005). Martín-Vide et al. (2010) showed on the Gallego River (Spain) that for 1 million m3 gravels extracted, another million were evacuated indirectly due to channel readjustments. As on most other rivers, detailed quantitative data on the volumes extracted from the Allier River are unavailable. Intensive gravel mining activities did only begin in the second half of the 20th century. Indeed, no gravel pits were identified on the 1954 aerial photographs. However, they are clearly present in 1974. At the beginning, industrial gravel mining took place in the channel itself, on its margins and within the floodplain. According to Bouchardy (1991), 7 million tons were extracted within the Allier River catchment in 1974, 60-65% of which in the Puy-de-Dôme area, which largely corresponds to the study area and includes a right bank tributary, the Dore River. Gravel mining extraction was stopped in-channel and was relocated onto the floodplain in the early 1980s and systematically after the promulgation of the 1994 law (Bouchardy, 1991). According to Dambre (1995), the quantities extracted in the channel between 1981 and 1993 were nearly zero, whereas, according to the same author, gravel extraction in the floodplain represented 41,811,000 tons, of which 79% were extracted in the Puy-de-Dôme area. Even if the total volume extracted from the channel is not known, it can be suggested that, in comparison to other French rivers and the current bed load dynamics observed in the River Allier, the volumes extracted in the channel during the second half of the 20th century (mainly between the 1960s and 1980) exceeded the volumes transported. Consequently, this resulted in a bedload decrease and deficit and the channel incision quantified in this study, with median values of -0.7 m and -0.6 m in the two floodplains sections, respectively Section 1 and Section 3. Bedload deficits caused by human impacts and the consequences caused by “hungry waters”, sensu Kondolf (1997), such as subsequent channel incision, are well studied for temperate rivers and may cause irreversible changes considering human timescales. For example, on the Loire River 58 million tons were extracted between 1981 and 1993 (and probably more than 225 million tons since 1949 (Dambre and Malaval, 1993), which represents 57,200 tons per kilometer of active channel (Nabet, 2013). On the Middle Loire River, Nabet (2013) observed that the incision was continuing more than ten years after the gravel extractions ended. Rovira et al. (2005) estimated that 420 years are needed for the lower Tordera River (Spain) to recover its initial bed level before sediment extraction, which occurred between 1956 and 1987 and caused a mean incision of 1.5 m. This number was evaluated at the minimum at 360 years for a gravel pit extraction site (of about 600 000 m3 extracted in 20 years) on the Figarella River, Corsica, France (Gaillot and Piégay, 1999).

21The results of the present study further indicate that channel incision slowed down or even stopped locally between 1999 and 2020, i.e., at the beginning of the 21st century (fig. 4). Despite the fact of the methodological bias due to the incertainty in data acquisition and the relatively low number of sample points (between 44 and 50 points), a recent deceleration of vertical incision is possible. Indeed, in-channel bedrock outcrops in certain river reaches better resist vertical erosion than a non-cohesive sediment layer. On the Garonne River, Jantzi (2018) observed a genuine transformation of certain reaches from an alluvial river to a bedrock river which slowed down in-channel incision rates observed earlier by Steiger et al. (2000), who also discussed the hydroecological consequences of such significant geomorphological river type changes. On the Allier River however, not the entire study area is concerned by massive bedrock outcrops. These are especially located in the upstream part of Section 3. For other reaches, channel incision may have slowed down as a consequence of channel slope adjustments reducing stream power and, concomitantly, an increase of bed load availability since in-channel extraction stopped since the 1980s. In addition, Darby and Thorne (1992) observed that channelization caused bed degradation, which in turn led to increased bank height, bank erosion and mass failures representing major sediment sources and increased bedload availability. In certain cases, the formation of armor layers in impacted river channels, also observed within the study area, may decrease sediment mobility and vertical incision in case of insufficient transport capacity of the stream (Kondolf, 1997).

5.3. Channel changes and flow regime

22When discussing control factors of channel adjustments, it is also necessary to take into account the evolution of the flow regime over the period of time considered. Overall, the analysis at the Vieille-Brioude hydrological station showed that flood frequency, magnitude and duration decreased between 1954 and 2016. In addition, the Mann Kendall test clearly showed a decrease trend of the maximum annual discharge and, conversely, an increase trend of low water after 1983-84 (fig. 9, tab. 5). The changes of the flow regime promoted the establishment and the development of the vegetation on alluvial bars, highlighted during the 1990s by the increase of annual abandoned areas within the three morphostructural sections (fig. 5). These observations are consistent with those from the lower Allier River, where the reduction of high and moderate magnitude floods led to an increase of channel stability and therefore less channel migration and a concomitant clear threshold of trajectory change in alluvial vegetation succession processes since the beginning of the 21st century (Garófano Gómez et al., 2017).

23Riparian vegetation on alluvial bars acts on and modifies fluvial processes with the promotion of an accretion rather than an erosion process and thus the formation of stable fluvial landforms, i.e., islands and floodplains (VanLooy and Martin, 2005; Corenblit et al., 2009; Grivel and Gautier, 2012; Manners et al., 2014). Vegetation establishment ultimately leads to the reduction of the active channel width as well as to its stabilization through a reduction of lateral bank erosion. The ability to erode stable river margins and islands will then depend on low recurrence and high magnitude floods (e.g., Arnaud-Fassetta et al., 2005; Nardi et Rinaldi, 2005; Corenblit et al., 2010; Comiti et al., 2011; Winterbottom, 2000). On the Loire River, the retroactions between vegetation and fluvial bed dynamics have been studied in detail: the rapid alluvial vegetation growth observed since the early 20th century shows the successive impacts of the engineering works (18th – 19th centuries), the reduction of efficient discharge (beginning of the 20th century), followed by the effects of massive gravel extractions and changes in social practices (navigation and agriculture in the fluvial bed). The trapping effect of sediment load by the pioneer vegetation finally contributed to reinforce the sediment deficit in the active bed (Grivel and Gautier, 2012; Grivel et al., 2018; Nabet et al., 2016).

24These channel and vegetation dynamics were clearly identified on the middle Allier River since the early 21st century. In the 2000-2010 period, where two high magnitude floods were recorded, the first in 2003 (20 yr recurrence interval) and the second in 2008 (10 yr recurrence interval, see Figure 9), the channel recorded a decrease of the annual abandoned area compared to the previous period, and a maintenance or even an increase of the active channel width (fig. 5). However, in the following period, which recorded floods with a return period of 1.5 and 2 yrs, the river showed its highest decrease of active channel width and annual eroded areas and, on the other hand, an increase of the annual abandoned area, recorded on the three morphostructural sections (fig. 5). In addition to the morphological consequences induced by human impacts discussed above, the flood regime change also triggers morphological channel changes. The links between hydrological regime, vegetation dynamics and morphological channel pattern will have to be monitored in the coming decades in order to refine the present observations. Future monitoring is all the more necessary in the context of current climate change, which will largely control future channel and vegetation evolutionary trajectories.

6. Conclusion

25This study highlights contrasted vertical and planform changes of the middle Allier River channel observed since the second half of the 20th century. Whereas the gorge section exhibits mainly lowest changes in channel geometry due to natural and anthropogenic structural controls, the two alluvial plain sections present up to four distinct morphological evolutionary trajectories mainly related to human impacts. Especially, the upstream and downstream parts of Section 1 and the middle of Section 3 recorded morphological channel alterations associated with the presence of gravel mining sites and, locally, with in-channel engineering works. In addition, channel planform fluctuations recorded in the early 21st century, the strong active channel narrowing and the decrease of annually eroded areas during the most recent period investigated (2010-2016) are linked to recent changes of the flow regime observed since the end of the 20th century and implying the decrease of flood frequency, duration and magnitude. Since, as is well-known, channel degradation due to bedload deficits, i.e. channel incision, may significantly alter the integrity of fluvial ecosystems and associated ecosystem services, the present study constitutes a stepping stone in establishing a diagnostic of current bedload dynamics of the Allier River in the context of the RALLIER research project financed by the Auvergne-Rhône-Alpes Region. It will thus provide river managers with sound scientific knowledge to support the establishment of sustainable management and restoration strategies that also take into account the critical issue of morphogenetic flood regime changes in a context of climate change.

*Auteur correspondant : Tel : 33 +(0)4 73 34 68 18
anais.arfeuillere@uca.fr (Anaïs Arfeuillère)

Haut de page

Bibliographie

Arnaud-Fassetta G., Cossart E., Fort, M. (2005) Hydro-geomorphic hazards and impact of man-made structures during the catastrophic flood of June 2000 in the Upper Guil catchment (Queyras, Southern French Alps). Geomorphology, 66 (1-4), 41-67.

DOI : 10.1016/j.geomorph.2004.03.014

Babonaux Y. (1970) – Le lit de la Loire - Etude d’hydrodynamique fluviale. Paris, Bibliothèque Nationale, 252 p.

Billet C. (1979) – Inventaire des exploitations de sables et graviers dans le Val d’Allier entre Brassac-Les-Mines et Limons (Puy-de-Dôme). BRGM, Cournon d'Auvergne, 37 p.

Boix-Fayos C., Barberá G.G., López-Bermúdez F., Castillo V.M. (2007) – Effects of check dams, reforestation and land-use changes on river channel morphology: Case study of the Rogativa catchment (Murcia, Spain). Geomorphology, 91 (1-2), 103–123.

DOI : 10.1016/j.geomorph.2007.02.003

Bouchardy C. (1991) – L’Allier. Privat, Toulouse, 188p.

Chorley R.J. (1969) – Introduction to Fluvial Processes. Methuen & Co. Ltd., London, 232 p.

Clavel P., Cuinat R., Hamon Y., Romaneix C. (1978) – Effets des extractions de matériaux alluvionnaires sur l’environnement aquatique dans les cours supérieurs de la Loire et de l’Allier. Bulletin Français de Pisciculture, 268, 121–154.

Comiti F., Da Canal M., Surian N., Mao L., Picco L., Lenzi M.A. (2011) – Channel adjustments and vegetation cover dynamics in a large gravel bed river over the last 200 years. Geomorphology, 125 (1), 147–159.

DOI : 10.1016/j.geomorph.2010.09.011

Corenblit D., Steiger J., Gurnell A.M., Tabacchi E., Roques L. (2009) Control of sediment dynamics by vegetation as a key function driving biogeomorphic succession within fluvial corridors. Earth Surface Processes and Landforms, 34, 1790–1810.

DOI : 10.1002/esp.1876

Corenblit D., Steiger J., Tabacchi E. (2010) Biogeomorphologic succession dynamics in a Mediterranean river system. Ecography, 33 (6), 1136-1148.

DOI : 10.1111/j.1600-0587.2010.05894.x

Dambre J.P., Malaval P. (1993) – Évaluation des conditions de poursuite de la politique de limitation des extractions de matériaux dans le lit de la Loire du Bec d'Allier à Nantes. Service de bassin Loire-Bretagne, Orléans, 69 p.

Dambre J.P. (1995) – Les extractions de matériaux dans le lit mineur et le lit majeur de la Loire et de ses affluents de 1981 à 1993. Conseil général des ponts et chaussées, Paris, 31 p.

Darby S.E., Thorne C.R. (1992) – Impact of channelization on the Mimmshall Brook, Hertfordshire, UK. Regulated Rivers: Research and Management, 7, 193-204.

David M., Labenne A., Carozza J.M., Valette P. (2016) – Evolutionary trajectory of channel planforms in the middle Garonne River (Toulouse, SW France) over a 130-year period: Contribution of mixed multiple factor analysis (MFAmix). Geomorphology, 258, 21–39.

DOI : 10.1016/j.geomorph.2016.01.012

Dépret T., Gautier E., Hooke J., Grancher D., Virmoux C., Brunstein D. (2017) – Causes of planform stability of a low-energy meandering gravel-bed river (Cher River, France). Geomorphology, 285, 58–81.

DOI : 10.1016/j.geomorph.2017.01.035

DIREN Auvergne (1998) – Etude de l’Allier entre Vieille-Brioude et Villeneuve, 135 p.

Downward S.R., Gurnell A.M., Brookes A. (1994) – A methodology for quantifying river channel planform change using GIS. In Olive L.J., Loughran R.J., Kesby J.A (Eds.): Variability in Stream Erosion and sediment transport. Association of Hydrological Sciences Publication, 224, 449–456.

Ejarque A., Beauger A., Miras Y., Peiry J.L., Voldoire O., Vautier F., Benbakkar M., Steiger J. (2015) – Historical fluvial palaeodynamics and multi-proxy palaeoenvironmental analyses of a palaeochannel, Allier River, France. Geodinamica Acta, 27, 25–47.

DOI : 10.1080/09853111.2013.877232

Fryirs K.A., Brierley G.J. (2013) – Geomorphic Analysis of River Systems: An approach to Reeading the Landscape. Wiley-Blackwell, Chichester, 346 p.

Gaillot S., Piégay H. (1999) – Impact of Gravel-Mining on Stream Channel and Coastal Sediment Supply: Example of the Calvi Bay Corsica (France). Journal of Coastal Research, 15 (3), 774–788.

Garófano Gómez V., Metz M., Egger G., Díaz-Redondo M., Hortobágyi B., Geerling G., Corenblit D., Steiger J. (2017) – Vegetation succession processes and fluvial dynamics of a mobile temperate riparian ecosystem: The lower Allier River (France). Géomorphologie : Relief, Processus, Environnement, 23 (3), 187–202.

DOI : 10.4000/geomorphologie.11805

Gasowski Z. (1994) – L’enfoncement du lit de la Loire. Revue de Géographie de Lyon, 69 (1), 41-45.

DOI : 10.3406/geoca.1994.4236

Gautier E. (1994) – Interférence des facteurs anthropiques et naturels dans le processus d’incision sur une rivière alpine. Revue de Géographie de Lyon, 69 (1), 57–62.

Gautier E., Piégay H., Bertaina P. (2000) – A methodological approach of fluvial dynamics oriented towards hydrosystem management: case study of the Loire and Allier rivers. Geodinamica Acta, 13 (1), 29–43.

DOI : 10.1016/S0985-3111(00)00109-1

Gautier E., Bournouf J., Carcaud N., Chambaud F., Garcin, M. (2007) – Les interrelations entre les sociétés et le fleuve Loire depuis le Moyen Âge. In Trémolières M., Schnitzler A., Silan P. (Eds.): Protéger, restaurer et gérer les zones alluviales, pourquoi et comment ? Lavoisier, Éditions Tec & Doc, 83-97.

Geerling G., Ragas A., Leuven R.S.E.W., van den Berg, J. H., Breedveld M., Liefhebber D., Smits A. (2006) – Succession and rejuvenation in floodplains along the river Allier (France). Hydrobiologia, 565, 71–86.

DOI : 10.1007/1-4020-5367-3_5

Gregory K.J. (2006) – The human role in changing river channels. Geomorphology, 79 (3-4), 172–191.

DOI : 10.1016/j.geomorph.2006.06.018

Grivel S., Gautier E. (2012) – Mise en place des îles fluviales en Loire moyenne, du 19e siècle à aujourd’hui. Cybergeo : European Journal of Geography, 615.

DOI : 10.4000/cybergeo.25451

Grivel S., Nabet F., Gautier E., Teman S., Gruwé G., Gardaix J., Lee M. (2018) – Héritages et influences contemporaines des anciens ouvrages de navigation de la Loire moyenne (France). VertigO, 18 (3).

DOI : 10.4000/vertigo.23121

Hughes M.L., McDowell P.F., Marcus W.A. (2006) – Accuracy assessment of georectified aerial photographs: Implications for measuring lateral channel movement in a GIS. Geomorphology, 74 (1-4), 1–16.

DOI : 10.1016/j.geomorph.2005.07.001

Jantzi H., Carozza J.M., Probst J.L., Valette P. (2017) – Ajustements géomorphologiques du chenal de la moyenne Garonne en aval de Toulouse au cours des 200 dernières années (sud-ouest, France). Géomorphologie : relief, processus, environnement, 23 (2), 139-153.

DOI : 10.4000/geomorphologie.11692

Jantzi H. (2018) – Incision fluviale et transition d’une rivière alluviale vers une rivière à fond rocheux : formation et évolution des seuils molassiques de la moyenne Garonne toulousaine au cours du 20e siècle, thèse de doctorat, Université de Toulouse.

Kondolf G.M. (1997) – Hungry water: effects of dams and gravel mining on river channels. Environmental Management, 21 (4), 533–551.

DOI : 10.1007/s002679900048

Landon N., Piégay H. (1994) – L’incision d’affluents méditerranéens du Rhône : la Drôme et l’Ardèche. Géocarrefour, 69 (1), 63–72.

DOI : 10.3406/geoca.1994.4239

Lang M., Ouarda T.B.M.J., Bobée B. (1999) – Towards operational guidelines for over-threshold modeling. Journal of Hydrology, 225 (3-4), 103–117.

DOI : 10.1016/S0022-1694(99)00167-5

Liebault, F., Piegay, H. (2001) – Assessment of channel changes due to long-term bedload supply decrease, Roubion River, France. Geomorphology, 36, 167186.

DOI : 10.1016/S0169-555X(00)00044-1

Liébault F., Piégay H. (2002) – Causes of 20th century channel narrowing in mountain and piedmont rivers of southeastern France. Earth Surface Processes and Landforms, 27 (4), 425–444.

DOI : 10.1002/esp.328

Manners R.B., Schmidt J.C., Scott M.L. (2014) Mechanisms of vegetation-induced channel narrowing of an unregulated canyon river: results from a natural field-scale experiment. Geomorphology, 211, 100–115.

DOI : 10.1016/j.geomorph.2013.12.033

Martín-Vide J.P., Ferrer-Boix C., Ollero A. (2010) Incision due to gravel mining: Modeling a case study from the Gállego River, Spain. Geomorphology, 117 (3-4), 261–271.

DOI : 10.1016/j.geomorph.2009.01.019

Merle O., Michon L. (2001) The formation of the West European Rift: a new model as exemplified by the Massif Central area. Bulletin de la Société Géologique de France, 172, 213–221.

DOI : 10.2113/172.2.213

Meylan P., Favre A.C., Musy A. (2008) – Hydrologie fréquentielle: une science prédictive. Presses polytechniques et universitaires romandes, Lausanne, 173 p.

Nabet F., (2013) – Étude du réajustement du lit actif en Loire moyenne, bilan géomorphologique et diagnostic du fonctionnement des chenaux secondaires en vue d’une gestion raisonnée. Thèse de doctorat, Université Paris 1, 416 p.

Nabet F., Grivel S., Gautier E. (2016) – Le rôle des aménagements sur la réponse topo-sédimentaire d’un cours d’eau à différents événements hydrologiques, la Loire moyenne. Géomorphologie : Relief, Processus, Environnement, 22 (2), 211–225.

DOI : 10.4000/geomorphologie.11398

Naiman R.J., Décamps H., McClain M.E. (2005) – Riparia. Ecology, Conservation, and Management of Streamside Communities. Academic Press, London, 446 p.

Nardi L., Rinaldi M. (2015) Spatio-temporal patterns of channel changes in response to a major flood event: the case of the Magra River (central-northern Italy). Earth Surface Processes and Landforms, 40 (3), 326–339.

DOI : 10.1002/esp.3636

Onde H. (1923) – Les crues de l’Allier. Revue de Géographie Alpine, 11 (2), 301–372.

Peiry J.L., Petit S., Lepicek P. (2009) – Dynamique hydrogéomorphologique et paysagère de la rivière Allier dans la Réserve Naturelle du Val d’Allier (1840-2005). Revue d’Auvergne, 592, 73–94.

Petit S. (2006) – Reconstitution de la dynamique du paysage alluvial de trois secteurs fonctionnels de la rivière Allier (1946-2000), Massif central, France. Géographie physique et Quaternaire, 60 (3), 271–287.

DOI : 10.7202/018000ar

Petts G.E., Möller H., Roux A.L. (1989) – Historical change of large alluvial rivers: Western Europe, Wiley, New York, 355 p.

Preciso E., Salemi E., Billi P. (2012) – Land use changes, torrent control works and sediment mining: effects on channel morphology and sediment flux, case study of the Reno River (Northern Italy). Hydrological Processes, 26 (8), 1134–1148.

DOI : 10.1002/hyp.8202

Renard B., Lang M., Bois P., Dupeyrat A., Mestre O., Niel H., Gailhard J., Laurent C., Neppel L., Sauquet E. (2006) – Evolution des extrêmes hydrométriques en France à partir de données observées. La Houille Blanche, 92, 48–54.

DOI : 1051/lhb:2006100.

Rinaldi M., Wyżga B., Surian N. (2005) – Sediment mining in alluvial channels: physical effects and management perspectives. River Research and Applications, 21(7), 805–828.

DOI : 10.1002/rra.884

Roux C., Alber A., Bertrand M., Vaudor L., Piégay H. (2015) – “FluvialCorridor”: A new ArcGIS toolbox package for multiscale riverscape exploration. Geomorphology, 242, 29–37.

DOI : 10.1016/j.geomorph.2014.04.018

Rovira A., Batalla R.J., Sala M. (2005) Response of a river sediment budget after historical gravel mining (the lower Tordera, NE Spain). River Research and Applications, 21, 829–847.

DOI : 10.1002/rra.885

Salit F., Arnaud-Fassetta G., Zaharia L., Madelin M., Beltrando G. (2015) The influence of river training on channel changes during the 20th century in the Lower Siret River (Romania). Géomorphologie : Relief, Processus, Environnement, 21 (2), 175–188.

DOI : 10.4000/geomorphologie.11002

Schumm S.A. (1977) The Fluvial System, Wiley, New York, 338 p.

Steiger J., James M., Gazelle, F. (1998) – Channelization and consequences on floodplain system functioning on the Garonne River, SW France. Regulated Rivers. Research and Management, 14, 13–23.

Steiger J., Corenblit D., Vervier P. (2000) – Les ajustements morphologiques contemporains du lit mineur de la Garonne, France et leurs effets sur l’hydrosystème fluvial. Zeitschrift für Geomorphologie, 122, 227–246.

Surian N., Rinaldi M. (2003) – Morphological response to river engineering and management in alluvial channels in Italy. Geomorphology, 50 (4), 307–326.

DOI : 10.1016/S0169-555X(02)00219-2

Surian N., Cisotto A. (2007) – Channel adjustments, bedload transport and sediment sources in a gravel-bed river, Brenta River, Italy. Earth Surface Processes and Landforms, 32 (11), 1641–1656.

DOI : 10.1002/esp.1591

Urban M.A., Rhoads B.L. (2003) – Catastrophic Human-Induced Change in Stream-Channel Planform and Geometry in an Agricultural Watershed, Illinois, USA. Annals of the Association of American Geographers, 93 (4), 783–796.

DOI : 10.1111/j.1467-8306.2003.09304001.x

VanLooy J.A., Martin C.W. (2005) – Channel and Vegetation Change on the Cimarron River, Southwestern Kansas, 1953–2001. Annals of the Association of American Geographers 95, 727–739.

DOI : 10.1111/j.1467-8306.2005.00483.x

Warner R.F. (2000) – Gross channel changes along the Durance River, Southern France, over the last 100 years using cartographic data. Regulated Rivers: Research & Management, 16 (2), 141–157.

DOI : 10.1002/(SICI)1099-1646(200003/04)16:2<141::AID-RRR574>3.0.CO;2-2

Winterbottom S.J. (2000) – Medium and short-term channel planform changes on the Rivers Tay and Tummel, Scotland. Geomorphology, 34 (3-4), 195–208.

DOI : 10.1016/S0169-555X(00)00007-6

Wyżga B. (1991) – Present-day downcutting of the Raba River channel (Western Carpathians, Poland) and its environmental-effects. Catena, 18 (6), 551-566.

DOI : 10.1016/0341-8162(91)90038-Y

Wyżga B. (2001) – Impact of the channelization-induced incision of the Skawa and Wisłoka Rivers, southern Poland, on the conditions of overbank deposition. Regulated Rivers: Research & Management, 17 (1), 85–100.

DOI : 10.1002/1099-1646(200101/02)17:1<85::AID-RRR605>3.0.CO;2-U

Wyżga B. (2007) – A review on channel incision in the Polish Carpathian rivers during the 20th century. In Habersack H., Piégay H., Rinaldi M. (Eds): Developments in Earth Surface Processes, 11, Elsevier, 525–553.

DOI : 10.1016/S0928-2025(07)11142-1

Ziliani L., Surian N. (2012) – Evolutionary trajectory of channel morphology and controlling factors in a large gravel-bed river. Geomorphology, 173, 104–117.

DOI : 10.1016/j.geomorph.2012.06.001

Haut de page

Annexe

Version française abrégée

D’important changements morphologiques des systèmes fluviaux européens, en lien avec l’augmentation des pressions anthropiques, ont été enregistrés depuis la seconde moitié du XXe siècle. La rivière Allier, France, connue pour ses sections très dynamiques dans son cours aval, a été cependant soumise à des impacts anthropiques, aboutissant à une incision du chenal et un déficit sédimentaire lors de seconde moitié du siècle dernier (DIREN Auvergne, 1998). La présente étude a pour objectif de comprendre les changements hydrogéomorphologiques récents de la rivière Allier dans son cours moyen, à travers (i) la quantification des changements verticaux et en plan du chenal depuis la seconde moitié du XXe siècle ; (ii) la détermination des trajectoires d’évolution morphologiques ; (iii) l’analyse des liens statistiques entre les changements morphologiques enregistrés, les contraintes morphostructurales et les pressions anthropiques et (iv) l’examen de l’évolution du régime hydrologique entre 1954 et 2016.

La rivière Allier parcourt 410 km de sa source en Lozère jusqu’à sa confluence avec la Loire près de Nevers, drainant un bassin versant de 14 400 km² situé dans le Massif Central (fig. 1A). Cette étude se focalise sur l’Allier moyen, de sa sortie des gorges (Vieille-Brioude) jusqu’à sa confluence avec l’un de ses affluents majeurs, la Dore (Limons). Le long de ce secteur, l’Allier traverse successivement trois sections morphostructurales (fig. 1B), deux sections de plaines alluviales encadrant une section en gorges (tab. 1).

L’évolution morphologique du chenal a été mise en évidence à partir (i) d’un jeu de six photographies aériennes historiques de 1954, 1974, 1992, 2000, 2010 et de 2016 (tab. 2, fig. 3) pour retracer l’évolution de la forme en plan du lit à travers l’évolution des largeurs du lit actif (m), les surfaces érodées et abandonnées annuellement (m²), et (ii) de quatre relevés de lignes d’eau de 1934-35, 1999, 2005 et de 2020 (tab. 3) pour retracer l’évolution altimétrique du lit. Les évolutions morphologiques ont été analysées à partir d’analyses statistiques descriptives s’appuyant sur un découpage du secteur d’étude en 125 segments. Puis une analyse en composantes principales suivie d’une classification ascendante hiérarchique ont permis d’identifier différentes trajectoires d’évolutions morphologiques. Les liens entre les trajectoires d’évolutions morphologiques et les contraintes morphostructurales (degré de confinement du chenal dans la vallée) ainsi que les pressions anthropiques (présence de sites d’extraction de granulats, de seuils et de travaux dans le chenal et degré de berges protégées) (tab. 4) ont été analysés à partir d’une analyse factorielle des correspondances. Enfin une analyse du régime hydrologique a été réalisée à partir des tests de Mann Kendall et Petitt (fig. 2).

Á l’échelle du secteur d’étude et entre 1954 et 2016, le chenal actif s’est rétracté, la quantité de surfaces érodées annuellement est stable, exceptée sur la dernière période, et la quantité de surfaces abandonnées annuellement a augmenté (fig. 4). Une incision généralisée a également été enregistrée entre 1934/35 et 2020, tandis que les médianes des différences altitudinales du lit entre 1999 et 2005 et entre 2005 et 2020 sont proches de zéro. Á l’échelle des sections morphostructurales (fig. 5), les évolutions de la forme en plan, pour les trois sections, sont identiques à celles observées à l’échelle du secteur d’étude, hormis pour l’évolution de la largeur du chenal actif dans la section 1, qui ne présente pas de changement significatif. Néanmoins, ces évolutions de la forme en plan sont moins marquées dans la section 2. Également l’incision domine entre 1934/35 et 2020 dans les sections 1 et 2. Les deux premiers axes de l’analyse en composantes principales représentent 69,7% de la variance totale et la classification ascendante hiérarchique a permis de distinguer quatre groupes (fig. 6). Le groupe 1 (9 segments) est caractérisé par les valeurs les plus élevées d'incision et de rétraction du chenal actif, le groupe 2 (22 segments) est caractérisé par les tendances à la rétraction du chenal actif et à la diminution des surfaces érodées annuellement, le groupe 3 (76 segments) est caractérisé par pas ou peu de changements morphologiques, et le groupe 4 (18 segments) est quant à lui caractérisé par une incision du chenal et une tendance à l’augmentation des surfaces abandonnées annuellement. Tandis que la section 2 est exclusivement composée du cluster 3, dans les deux autres sections morphostructurales, chacun des quatre groupes est représenté. L’analyse factorielle des correspondances montre un lien étroit entre les clusters 2 et 4 et la présence de sites d’extraction de granulats, ainsi qu’un lien entre le cluster 1 et la présence de travaux dans le chenal comme des recoupements de méandres (fig. 8). Entre 1954 et 2016, le régime hydrologique de l’Allier montre une tendance à la baisse de la fréquence, de la durée et de l’intensité des crues, et au contraire une augmentation annuelle du nombre de jours avec des débits en-dessous du module (fig. 9). Le test de Mann-Kendall montre une tendance à la baisse des débits annuels maximaux et au contraire une augmentation du nombre de jours avec des débits en-dessous du module, pour lesquels le test de Pettitt marque deux ruptures significatives dans les séries en 1984 et 1983 (tab.  5).

Le cours moyen de l’Allier a enregistré une incision moyenne de son lit de 0,86 m entre 1934/35 et 2020, ainsi qu’une diminution de 25% de sa largeur et une baisse de sa mobilité latérale de 68% entre 1954 et 2016, similaire à l’évolution d’autres lits fluviaux européens depuis le XXe siècle en réponse à l’augmentation des pressions anthropiques (Steiger et al., 1998 ; Liébault and Piégay, 2002 ; Surian and Rinaldi, 2003 ; Wyżga, 2007). Cependant, cette évolution globale est à nuancer (i) entre les sections : la section de gorges enregistre moins de changements morphologiques que les deux sections de plaines alluviales car elle est largement contrainte latéralement (gorges confinées, chemin de fer river droite, autoroute rive gauche), laissant moins d’espaces de liberté, (ii) au sein des sections de plaines alluviales où les parties amont et aval de la section 1 et le milieu de la section 3 ont enregistré davantage de changements morphologiques du chenal en lien avec la présence de sites d’extraction de granulats et plus localement de travaux dans le chenal. Également, l’analyse du régime hydrologique conduit à prendre en considération son rôle dans l’évolution morphologique du chenal, notamment de la forme en plan depuis le début du XXIe siècle. En réponse à des crues moins fréquentes, plus courtes et avec des intensités plus faibles, et au contraire une augmentation du nombre de jours avec des débits en-dessous du module, la quantité de surfaces annuellement érodées a baissé de manière généralisée sur l’ensemble du secteur et au contraire, celle des surfaces abandonnées annuellement a augmenté, ce qui a conduit à une rétraction du chenal et à sa stabilisation latérale dans sa plaine alluviale. En conséquence, les futurs plans de gestion de la rivière devront prendre en considération ces ajustements morphologiques hérités des anciennes exploitations anthropiques et le changement climatique en cours qui induit des modifications du régime des crues, lequel représente également une variable de contrôle majeure de la morphodynamique des rivières.

Haut de page

Table des illustrations

Titre Fig. 1 – Location maps.Fig. 1 – Cartes de localisation.
Légende A: Allier River catchment and location of the study area. B: The middle course of the Allier River between Vieille-Brioude and Limons shows three distinct morphostructural sections. Mean monthly discharges characterize the mean flow regime of each section (calculated from 1919 to 2019 at Vieille-Brioude; from 1984 to 2019 at Coudes and from 1974 to 2019 at Limons; discharge data: Banque Hydro). 1. floodplain limits; 2. limits of morphostructural sections; 3. towns with gauging station ; 4. main towns. French coordinate system Lambert 93.A : Bassin versant de la rivière Allier et localisation du secteur d’étude. B : L’Allier moyen entre Vieille-Brioude et Limons montre trois sections morphostructurales. Les débits mensuels moyens caractérisent le régime d'écoulement moyen de chaque section (calculés de 1919 à 2019 à Vieille-Brioude ; de 1984 à 2019 à Coudes et de 1974 à 2019 à Limons ; données hydrologiques : Banque Hydro). 1. limites du lit majeur ; 2. limites des sections morphostructurales ; 3. villes avec une station hydrologique ; 4. villes principales. Système de coordonnées Lambert 93.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-1.jpg
Fichier image/jpeg, 183k
Titre Tab. – Mean values of main morphological (based on orthophotograph from 2016) and hydrological (Banque hydro) characteristics of the three morphostructural sections within the study area.Tab. 1 – Valeurs moyennes des principales charactéristiques géomorphologiques (basée sur l’orthophotographie de 2016) et hydrologiques (données de la Banque Hydro) des trois sections morphostructurales du secteur d’étude.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-2.jpg
Fichier image/jpeg, 99k
Titre Fig. 2 – Methodological framework.Fig. 2 Cadre méthodologique.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-3.jpg
Fichier image/jpeg, 388k
Titre Tab. 2 – Aerial photographs and orthophotographs used for the analysis of planform channel changes.Tab. 2 – Campagnes photographiques et orthophotographies utilisées pour l’analyse des changements de la forme en plan.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-4.jpg
Fichier image/jpeg, 187k
Titre Fig. 3 – Examples of the series of photographs used and planform changes between 1954 and 2016.Fig. 3 – Exemples de chaque photographie utilisée et des changements de la forme en plan entre 1954 et 2016.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-5.jpg
Fichier image/jpeg, 240k
Titre Tab. 3 – Longitudinal water surface profiles used for the analysis of vertical channel bed changes. Tab. 3 – Profil longitudinaux utilisées pour l’analyse des changements altitudinaux du lit.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-6.jpg
Fichier image/jpeg, 88k
Titre Tab. 4 – Data sources used for the determination of human pressures and impacts on channel morphology within the study area.Tab. 4 – Sources de données utilisées pour la détermination des pressions humaines sur le chenal dans la zone d’étude.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-7.jpg
Fichier image/jpeg, 69k
Titre Fig. 4 – Morphological indices for different dates and periods.Fig. 4 – Indices morphologiques pour les différentes dates ou périodes.
Légende From left to right: active channel width (m), annual eroded area (m2), annual abandoned area (m2), and altitudinal channel bed change (m). Within the boxplot, the horizontal black line represents the median, the upper edge of the boxplot represents the 75th and the lower edge of the boxplot represents the 25th percentile. Results of the degree of significant of Wilcoxon test’s p-value: * for p-value comprise between 0.05 and 0.01, ** between 0.01 and 0.001, *** between 0.001 and 0.0001, **** between 0.0001 and 0.00001.De gauche à droite : largeur du chenal actif (m), surface érodée annuellement (m2), surface abandonnée annuellement (m2) et changement altitudinal du lit (m). Dans le diagramme en boîte, la ligne noire horizontale accentuée représente la médiane, le bord supérieur du boxplot représente le 75e percentile et le bord inférieur du boxplot représente le 25e percentile. Résultats du degré de significativité de la p-value du test de Wilcoxon : * pour une p-value comprise entre 0,05 et 0,01, ** entre 0,01 et 0,001, *** entre 0,001 et 0,0001, **** entre 0,0001 et 0,00001.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-8.jpg
Fichier image/jpeg, 125k
Titre Fig. 5 – Morphological indices for different dates and periods.Fig. 5 – Indices morphologiques pour les différentes dates ou périodes.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-9.jpg
Fichier image/jpeg, 291k
Titre Fig. 6 – Factor analysis.Fig. 6 – Analyse factorielle.
Légende A: Variable graph of the PCA analysis; B: Individual graph of the PCA analysis with the classification of the Cluster Analysis; C: Map representation of the 125 segments based on their cluster affiliation and representation of the part of the four clusters in each of the morphostructural sections.A : Graphe des variables de l’analyse ACP ; B : Graphe des individus de l’analyse ACP avec la classification de l’analyse CAH ; C : Représentation cartographique des 125 segments en fonction de leur appartenance à un cluster et représentation de la part des quatre clusters dans chacune des sections morphostructurales.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-10.jpg
Fichier image/jpeg, 192k
Titre Fig. 7 – Degree of valley confinement and human impacts.Fig. 7 – Degré de confinement dans la vallée et impacts anthropiques.
Légende Degree of bank protections, gravel mining sites, sills; in-channel engineering works such as channel straightening and river training works, etc., for each of the 125 channel segments within the study area.Degré de berges protégées, sites d’extractions de granulats, seuils, travaux dans le chenal comme le recoupement de méandres et la modification de la géométrie du chenal, etc., pour chacun des 125 segments du secteur d’étude.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-11.jpg
Fichier image/jpeg, 170k
Titre Fig 8 – Correspondence Factor analysis.Fig 8 – Analyse factorielle des correspondances.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-12.jpg
Fichier image/jpeg, 141k
Titre Fig. 9 – Daily mean discharge (m3/s) of the Allier River at the Vieille-Brioude gauging station between 1954 and 2016.Fig. 9 – Débits journaliers moyens (m3/s) de la rivière Allier à la station hydrologique de Vieille-Brioude de 1954 à 2016.
Légende discharge data: Banque Hydro, https://hydro.eaufrance.fr.données hydrologiques : Banque Hydro, https://hydro.eaufrance.fr.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-13.jpg
Fichier image/jpeg, 178k
Titre Tab. 5 – Mann Kendall and Pettitt's test results for five characteristic flow indices.Tab. 5 – Résultats des tests statistiques de Mann Kendall et Pettitt des 5 indices hydrologiques.
Légende Qmax/yr = maximum annual daily mean discharge; Q>Q1.5yr = daily mean discharge superior or equal to the bankfull discharge (return period: 1.5 yr); Nb floods/yr = number of floods per year; Nb days of floods/yr = the number of days of floods per year; Nb days<Qmean = number of days with daily mean discharge lower than the inter-annual mean discharge.Qmax/yr = débit journalier moyen maximal par an ; Q>Q1.5yr = débits journaliers moyens supérieurs ou égaux au débit à pleins bords (période de retour : 1,5 an) ; Nb flood/yr = nombre de crue par an ; Nb day flood/yr = nombre de jours en crue par an ; Nb days<Qmean = nombre de jours où le débit journalier moyen est en-dessous du module.
URL http://journals.openedition.org/geomorphologie/docannexe/image/17312/img-14.jpg
Fichier image/jpeg, 61k
Haut de page

Pour citer cet article

Référence papier

Arfeuillère Anaïs, Johannes Steiger, Erwan Roussel, Stéphane Petit, Catherine Neel et Emmanuèle Gautier, « Statistical approach of contemporary hydrogeomorphological channel changes of the middle Allier River, France: morphostructural controls, human impacts and flow regime »Géomorphologie : relief, processus, environnement, vol. 28 - n° 4 | 2022, 241-256.

Référence électronique

Arfeuillère Anaïs, Johannes Steiger, Erwan Roussel, Stéphane Petit, Catherine Neel et Emmanuèle Gautier, « Statistical approach of contemporary hydrogeomorphological channel changes of the middle Allier River, France: morphostructural controls, human impacts and flow regime »Géomorphologie : relief, processus, environnement [En ligne], vol. 28 - n° 4 | 2022, mis en ligne le 09 décembre 2022, consulté le 09 juin 2023. URL : http://journals.openedition.org/geomorphologie/17312 ; DOI : https://doi.org/10.4000/geomorphologie.17312

Haut de page

Auteurs

Arfeuillère Anaïs

Université Clermont Auvergne, CNRS, GEOLAB UMR 6042, 4 rue Ledru, 63000 Clermont-Ferrand, France.

Johannes Steiger

Université Clermont Auvergne, CNRS, GEOLAB UMR 6042, 4 rue Ledru, 63000 Clermont-Ferrand, France.

Articles du même auteur

Erwan Roussel

Université Clermont Auvergne, CNRS, GEOLAB UMR 6042, 4 rue Ledru, 63000 Clermont-Ferrand, France.

Articles du même auteur

Stéphane Petit

VEODIS 3D, 1 rue des Moulins, 63400 Chamalières, France.

Catherine Neel

CEREMA Centre-Est, 8-10, rue Bernard Palissy, 63000 Clermont-Ferrand, France.

Emmanuèle Gautier

Université Paris 1, CNRS, LGP UMR 8591, 2 rue Henri Dunant, 94320 Thiais, France.

Articles du même auteur

Haut de page

Droits d’auteur

Tous droits réservés

Haut de page
  • Logo Groupe français de géomorphologie
  • Revue soutenue par l’Institut des sciences humaines et sociales du CNRS
    CNRS - Institut national des sciences humaines et sociales
  • OpenEdition Journals
Search OpenEdition Search

You will be redirected to OpenEdition Search