Navigation – Plan du site

Vegetation succession processes and fluvial dynamics of a mobile temperate riparian ecosystem: the lower Allier River (France)

Virginia Garófano-Gómez, Meike Metz, Gregory Egger, María Díaz-Redondo, Borbála Hortobágyi, Gertjan Geerling, Dov Corenblit et Johannes Steiger
p. 187-202

Résumés

De nombreux écosystèmes riverains européens des régions tempérées ont perdu leur caractère intrinsèquement dynamique en raison de leur chenalisation et régulation des débits. Une partie du cours inférieur de la rivière Allier est considérée comme l'un des derniers secteurs fluviatiles encore latéralement mobile. Ce secteur constitue donc une opportunité, à l’échelle européenne, pour étudier les processus naturels de succession végétale. Nous avons analysé (i) la dynamique de la succession végétale sur une période de 50 ans (1967-2014) à partir de huit jeux de photographies aériennes, (ii) les trajectoires de succession dominantes et (iii) leur relation avec le régime hydrologique. Nos résultats indiquent que la mosaïque d’habitats du site d’étude était très dynamique entre 1967 et 2005, mais que les crues n’ont pas transformé la composition globale des habitats au sein du corridor fluvial, tout en modifiant néanmoins leur distribution spatiale. Après l’année 2005, les modalités de progression et de rétrogression ont été fortement réduites, avec une augmentation de la stabilité et une réduction concomitante des crues fortes et modérées. Cette étude met clairement en évidence un seuil critique de transition dans la trajectoire de succession végétale depuis le début du XXIe siècle et facilite la compréhension de futures trajectoires d’évolution de l’écosystème riverain dans le contexte actuel du changement climatique global.

Haut de page

Notes de la rédaction

Article soumis le 24 mars 2017, reçu sous sa forme révisée le 31 juillet 2017 et définitivement accepté le 09 août 2017.

Texte intégral

We thank P.A. Dejaifve (conservationist of the Réserve Naturelle Nationale du Val d’Allier) for the information provided about the study area. We are grateful to A. Blaschka, E. Lautsch and K. Dolos for their help in a previous version of this manuscript. V. Garófano-Gómez received a post-doctoral grant from the Université Blaise Pascal (now: Université Clermont Auvergne, France). B. Hortobágyi was funded for her PhD by the French Ministry of National Education, Higher Education and Research. M. Diaz-Redondo was supported by a grant from the Fundação para a Ciência e Tecnologia as part of the FLUVIO Doctoral Programme - University of Lisbon (Portugal). The authors would like to thank the two anonymous reviewers for their constructive comments and suggestions.

1. Introduction

1Natural riverine ecosystems are intimately tied to physical, chemical, and biological processes varying in both space and time. This confers a high dynamism and habitat heterogeneity making them one of the most biodiverse ecosystems in temperate regions (Tockner and Stanford, 2002; Ward et al., 2002). Specifically, riparian and floodplain vegetation are subjected to multiple hydrogeomorphic constraints, namely climate, moisture availability and fluvial disturbance (Gurnell et al., 2016), which determine the distribution of riparian and floodplain habitats (Steiger et al., 2005; Corenblit et al., 2007).

2In riverine ecosystems with a good conservation status, it is possible to identify a shifting habitat mosaic dynamism that fosters the aforementioned high biodiversity (Ward et al., 2002; Tockner et al., 2010) as the result of sediment erosion and deposition, exchanges between surface water and groundwater, and plant succession. This interaction also produces a high turnover rate of habitats and their spatial shift. However, the landscape pattern, i.e. the habitat balance of the riparian corridor, may remain stable over long periods of time: within a given riverine ecosystem, habitats change their distribution in space, but not their mutual proportions which remain unchanged or stable (Stanford et al., 2005; Mouw et al., 2013).

3In Europe, many river systems lost their inherent, highly dynamic character after major human induced impacts, e.g. river training works and flow regulation, which intensified at least since the industrial revolution (Petts et al., 1989; Buijse et al., 2002). In addition to the hydrogeomorphological changes, riparian ecosystems have endured intensive groundwater pumping, land use change, eutrophication or chemical contamination (Nilsson and Berggren, 2000; Meybeck, 2003; Habersack et al., 2014). As a consequence, these alterations have often led to a simplification of habitats and hence to a significant decrease in species diversity (Trémolières et al., 1998; Deiller et al., 2001; Leyer, 2005; Hupp and Rinaldi, 2007), although in some cases the opposite has been observed (Ceschin et al., 2015; Dufour et al., 2015). In this sense, and even though pristine, unimpacted riverine ecosystems do not exist anymore in Europe, the last remaining low impacted and highly dynamic riparian ecosystems constitute a rare opportunity to investigate natural / semi-natural riparian vegetation succession processes to produce guidelines for restoring altered riparian ecosystems (Gurnell and Petts, 2002).

4In the European context, the Allier River (France) is one of the last remaining unregulated rivers with highly dynamic meandering sections and bank erosion, which represents an opportunity to investigate riparian vegetation succession processes of a dynamic, temperate riverine system adjusting to current climate and catchment conditions (Van Looy et al., 2008). Geerling et al. (2006) carried out a multi-spatiotemporal study covering a period of 46 years (1954-2000) in a 6-km dynamic channel-floodplain reach of the Allier River, south of Moulins (France), and suggested the applicability of the “shifting mosaic steady-state” model (Bormann and Likens, 1979) for the second half of the 20th century. Since this study, almost two decades ago, climate change has caused the increase of global temperatures and an increase in drought severity in the southern half of France (Giuntoli et al., 2013).

5It is widely recognized now, that the impact of climate change on floodplain inundation is induced on the one hand by increasing temperatures and on the other hand by spatial and temporal changes in precipitation patterns, and that climate change has an impact on the hydrogeomorphological regime of rivers, and as such can influence the spatial arrangement of floodplain vegetation and related plant succession dynamics (Schneider et al., 2011). The impacts of climate change on the hydrological regime of the Loire River and its tributaries (e.g. Allier River) during the 21st century were investigated by two French research projects, the ICC-Hydroqual (Moatar et al., 2010) and Explore 2070 (Chauveau et al., 2013). Their results project for the lower Allier River (i) an increase in low flow severity during the summer period as well as the duration of low flow periods (Moatar et al., 2010); (ii) an important decrease of mean annual floods from the mid-21st century on (Moatar et al., 2010); (iii) a decline by 10% of higher peak flows (10-yr recurrence interval flood; Van Looy and Piffady, 2017); and (iv) an increase of severe floods (extreme, low frequency floods; Van Looy and Piffady, 2017). However, according to Moatar et al. (2010), uncertainties remain about the evolution of extreme rain and flood events.

6The aim of our research was to study recent (last two decades) plant succession dynamics in relation to observed flood regime changes, based on the baseline study by Geerling et al. (2006) and focusing on a sub-reach of the 6-km stretch studied by Geerling et al. (2006). We hypothesised that after the year 2000, in relation with the flow regime change, (i) a shift in the habitat mosaic has taken place, but that (ii) the study site remained within a “shifting mosaic steady-state”. We also asked the explicit question if this dynamic and relatively low impacted study reach qualifies as a reference of natural plant succession processes in riverine ecosystems. To achieve this, we (i) interpreted the current landscape mosaic and riparian successional pathways based on a large-scale field survey; (ii) analysed the changes in the spatial distribution of succession phases and succession trajectories based on a set of eight aerial images spanning 47 years (1967-2014); (iii) related plant succession processes, contemporary and historical discharge time series and observed channel changes to interpret observed spatio-temporal patterns of the landscape mosaic.

2. Study site

7From its source at 1,485 m a.s.l. in Lozère in the Massif Central, southeast France, the Allier River flows 421 km north until joining the Loire River at the Bec d’Allier confluence (167 m a.s.l.; total catchment area: 14,350 km²) (fig. 1). Its overall climate domain is oceanic temperate - Cfb Köppen-Geiger type (Peel et al., 2007), with a mean annual temperature of 11.4ºC and a mean total annual rainfall of 779 mm (Vichy station - Météo France).

8The pluvial discharge regime of the Allier River has historically been irregular due to its strong intra- and inter-annual variability (Onde, 1923). Sporadically, it can result in severe low water levels during summer, as well as catastrophic floods, which generally occur during winter or spring. High magnitude floods often result from the superposition of a Mediterranean “cévenole” storm from the southeast (upper basin) and a westerly oceanic storm coming in from the Atlantic coast. These “mixed floods” can give rise to the strongest floods in Western Europe (Gautier et al., 2000; FRANE, 2014). Two dams located in the upper basin partially affect the discharge: Poutès (constructed in 1941; height: 18 m, length: 85 m) and Naussac (operational since 1983; height: 50 m, length: 240 m), which mainly store water in winter to support low summer flows. These dams have no regulating effect on large flood events, and probably neither during high flow periods (DREAL, pers. comm.; FRANE, 2014). The flood regime of the Allier River is therefore considered as close to natural or unregulated.

9The study reach (3.7 km long and 0.5-1.5 km wide) is located 8 km south and upstream of the town of Moulins (fig. 1), in the most mobile meandering section of the lower course of the Allier River, where specific annual bank erosion area rates of up to 10 m².m¹ have been observed (Hortobágyi et al., in press). Its average altitude is 210 m a.s.l. and its mean annual discharge is 135 m³.s-¹ (Banque Hydro). The reach is part of the Réserve Naturelle Nationale du Val d’Allier (http://www.reserves-naturelles.org/​val-d-allier), protected since 1994 for its natural patrimony. It is considered as one of the last highly mobile river sections in Europe, making its preservation a key issue for regional and national authorities (Cournez, 2015).

Fig. 1 – Location of the study reach within the Allier River Basin, near the French town of Moulins (A). The red line indicates the study area covered by the vegetation mapping (B).
Fig. 1 – Localisation du tronçon d’étude dans le bassin versant de la rivière Allier, en amont de la ville de Moulins, France (A). Le trait rouge indique la zone d'étude couverte par la cartographie de la végétation (B).

Fig. 1 – Location of the study reach within the Allier River Basin, near the French town of Moulins (A). The red line indicates the study area covered by the vegetation mapping (B).Fig. 1 – Localisation du tronçon d’étude dans le bassin versant de la rivière Allier, en amont de la ville de Moulins, France (A). Le trait rouge indique la zone d'étude couverte par la cartographie de la végétation (B).

Aerial image from July 2013 (source: CRAIG – Centre Régional Auvergnat de l’Information Géographique). The spatial reference is represented by geographic coordinates (latitude / longitude in degrees) relative to the WGS‑84 datum.
Image aérienne de juillet 2013 (source : CRAIG – Centre Régional Auvergnat de l’Information Géographique). La référence spatiale est représentée par des coordonnées géographiques (latitude / longitude en degrés) par rapport au système de référence géodésique WGS-84.

3. Materials and methods

3.1. Changes in succession phases assessed from aerial images and field vegetation mapping

10The study area (376 ha) was defined encompassing the maximum lateral and longitudinal extension that the near-natural riparian and floodplain vegetation has covered during the considered study period (see red line in fig. 1). Based on the most recent and high-quality aerial image (CRAIG – Centre Régional Auvergnat de l’Information Géographique; date: 7/2013, spatial resolution 0.25 m × 0.25 m) the vegetation within the study area was mapped in early autumn 2014. The early autumn (end of the vegetative period) was considered appropriate to map natural and non-natural vegetation types in the field, since no detailed flora inventories were carried out but an identification of the most dominant species within the patches. The area-wide vegetation mapping (Egger et al., 2015; Muñoz-Mas et al., 2017) was carried out at the meso-habitat scale 1:2,000. First, homogeneous polygons (patches or landscape units) were identified on the aerial images from 2013 as a first reference and then these polygons were checked in the field in 2014 and assigned to a certain succession phase. In total, 256 patches were digitised and 90% of them were checked in the field. The minimum patch size was 52 m². For more information about the field survey see Metz (2015).

11Eight sets of aerial images covering a total time span of 47 years (1967, 1978, 1983, 1985, 2000, 2005, 2010 and the aforementioned map obtained in 2014 from the field survey) were used to assess spatio-temporal changes in the landscape mosaic (i.e. succession phases and trajectories). This historical evaluation was based on maps already photo-interpreted by Geerling et al. (2006), unpublished data (Rackusch, 2014; Schwarzmeier, 2015), and the DREAL (2014) – see these references for detailed information about the photo series and their characteristics. In general, the photographic scale varied between 1:14,500 and 1:25,000 and their original photo resolution between 0.25 m and 1 m. Images before 1985 were in black and white and after 2000 in true-colour. All aerial images were taken in the summer (July / August). Therefore, we decided to divide years from August 1st to July 31st, because this data partition provides a better characterization of the hydrological conditions experienced by the vegetation mosaic illustrated in the most recent image of each period under comparison. To minimise errors, the most recent map (i.e. 2014) was produced first and then used as a basis to improve the recognition of the different phases on the majority of the older maps.

12The identified patches of the different maps were classified into seven succession phases based on the colour, texture, density and vertical structure of the vegetation. These phases were: WA (water), BS (bare soil; mainly open gravel and sand bars with vegetation ground cover lower than 5%), PP (pioneer phase; young, sparse vegetation with a ruderal, stress tolerant and disturbance-adapted strategy, like Salicaceae with a vegetation ground cover ~ 25% and ~ 0.5 m high), HP (herb phase; patches composed by annual and biannual short-lived herbs, reed and grassland which can also contain scattered woody pioneer individuals with a vegetation ground cover > 25% and less than 1 m high), SP (shrub phase; woody herbs like Poa pratensis L. or Hypericum perforatum L., and small patchy Prunus sp., Crataegus sp. and Salicaceae shrubs and trees, which may strongly interact with flow and sediment dynamics, with a canopy ground cover less or equal to 60% and generally less than 5 m high), FP (forest phase; dense woody canopy ground cover, typically higher than 60% and vegetation over 5 m high), and AGRI (agricultural land). In subsequent analyses, the phases WA and BS are considered together as they represent the highly flow-disturbed active tract where vegetation cannot establish, and their respective areas depend on the flow stage when the aerial photos were taken.

13In order to analyse the correspondence between the different succession phases across time, a 7 × 8 contingency table between the seven succession phases (rows) and the eight year-maps (columns) was created based on the area information extracted from the vector maps. A Pearson's chi-squared test of independence was used to identify whether there was a significant association between phases and time. A mosaic plot was created to display graphically the contingency table with the R package graphics (R Core Team, 2016). In a mosaic plot, the height of each rectangle or “tile” is proportional to the cell frequency, in our case, to the area covered by each phase within the study area in each year-map. A tile with a solid border indicates that the value is higher than that expected under the null hypothesis of independence between phases and year-maps. Conversely, a tile with a dashed border indicates that the area of that phase is smaller than that expected from random association. Under the null hypothesis of independence, all the tiles are white. The colour code is used to outline the most salient deviations, represented by the magnitude of their standardized Pearson’s chi-squared residuals (Friendly, 1994). A standardized residual of –2 or less indicates that the cell’s observed frequency is significantly lower than its expected frequency (in shaded red), and a standardized residual of +2 or more indicates that the cell’s observed frequency is significantly higher than its expected frequency (in shaded blue).

3.2. Changes in succession processes assessed from raster layers

14Vector layers were converted to 5-m resolution raster layers depicting the seven succession phases using ArcGIS v. 10.3 (ESRI, Redlands, CA, USA, 2014). This pixel size (5 m × 5 m) was considered appropriate from a computational point of view, but also in terms of patches’ spatial accuracy, ensuring at the same time the compatibility between the different available maps. The raster maps were overlain chronologically, obtaining confusion matrices from them. These matrices quantify the detailed changes (“cell-by-cell comparison”) between succession phases from one image to the next, and resulted in seven periods of time for comparison (i.e. 1967-1978, 1978-1983, etc.). Based on these confusion matrices and following Diaz-Redondo et al. (2017), the dominant succession processes were defined and grouped into four trajectories of change. These were progression (prevailing trajectory onwards; involves vegetation growth), retrogression (prevailing trajectory backwards; involves resetting of phases like the destruction of vegetation by lateral channel erosion and sediment deposition on point bars and within side channels which buries vegetation), stability (changeless) and anthropization (change from a natural cover to cultivation or urbanization by human action). The process assigned to each step was supposed to be dominant for that specific period, without knowing exactly what happened in the time in between (classes a and b denote different velocities in natural succession phases).

15All processes were subdivided into more detailed sub-processes depending on the initial and final succession phases identified from the two maps under comparison. For example, three sub-processes were defined within progression depending on the geomorphological processes and developmental stage of vegetation. These were “initial-aggradation”, “colonization” (progressing to pioneer and/or herbs) and “transition” (progression to shrub and/or forest). The change from cultivated land to herb, shrub or forest phase was also considered as a form of progression (classes c of the above-mentioned sub-processes), linked to a decrease of fluvial dynamics. Before the decrease in the flow and flood regime, cultivated areas close to the river banks were eroded more easily, incorporated into the active tract and rapidly colonized by riparian vegetation. Within a retrogression trajectory, “channel-shift” or “initial-aggradation” were considered as complete forms of retrogression, and thus “partial retrogression” was reserved exclusively to the retrogression of natural phases to less developed vegetal formations.

16Ratios between total progression, retrogression and stability were calculated for each time period to detect trends, as well as the most dynamic periods. However, in order to compare between them at the level of specific succession processes, the raw data was standardized into a rate (ha.yr-¹) as the seven time periods differed in length (from 2 to 15 years).

17A hierarchical cluster analysis (average linkage method and Euclidean distance) was carried out with the R packages stats and graphics (R Core Team, 2016) to analyse how the different periods of time differentiated concerning their trajectories of change and associated processes. In this sense, the standardized data allowed to summarize the different periods into groups with a similar pattern. Statistical comparisons between each one of the succession processes for the resultant groups of time periods were done using the robust Welch one-way ANOVA (Wilcox, 2012), implemented in the R package WRS (Wilcox and Schönbrodt, 2014). This robust analysis of variance uses a generalization of Welch’s method. It was chosen since it can be applied to examine groups of unequal and small sizes and does not require the assumption of homoscedasticity and normally distributed samples. If the null hypothesis (H0) holds, this means that there is no statistical difference in the means between the compared groups. Hwas rejected for high values of the statistic Fwe, which follows a Snedecor's F-distribution if His true. Raw p‑values were adjusted using the correction proposed by Benjamini and Hochberg (1995), and implemented in the R package stats (R Core Team, 2016). This correction attempts to limit the probability of false discoveries (i.e. incorrectly rejecting the null hypothesis when there is no real effect) when multiple comparisons are involved. We focused our attention to p‑values ≤ 0.10 as results approaching significance, due to our small sampling sizes.

3.3. River discharges and characterization

18Discharge records for the period 1968-2014 were obtained from the gauging station at Moulins (source: Banque Hydro, station code K3450810). For the missing periods at Moulins (1967-1968, 1993, 1997 and 2006) discharge data was reconstructed using a regression (y = 0.8499x + 9.8357;  = 0.9267) based on the Moulins time series and the one recorded at the station at Cuffy (code K3650810; ~ 55 km downstream from Moulins) since no important tributaries are located between both stations (fig. 2).

19Certain discharges and their associated characteristics, such as magnitude, frequency, duration, timing and rate of change (Poff et al., 1997) are needed to create disturbance and maintain the structural elements that provide the basis for biodiversity in aquatic and riparian habitats (Stanford et al., 1996; Richter et al., 1997). In this study, we defined critical discharge thresholds for maintaining channel and floodplain form and function as the double of the mean annual discharge (Emil Dister, pers. comm.; 2MQ: 270 m³.s-¹). Further characteristic floods are: HQ2 (2-yr recurrence interval flood: 669 m³.s-¹), HQ5 (5-yr recurrence interval flood: 937 m³.s-¹), HQ10 (10-yr recurrence interval flood: 1,110 m³.s¹) and HQ20 (20-yr recurrence interval flood: 1,280 m³.s-¹). These values were calculated with a Gumbel law over a series of 47 years (source: Banque Hydro).

20Using the reconstructed and recorded discharge series at Moulins (1967-2014), two variables were calculated: first, the mean number of independent flood events per year above the discharge thresholds; and second, the mean number of total accumulated days (not continuous) per year when discharge exceeded the different thresholds. Both hydrological variables, mean frequency and duration, were calculated for each one of the seven time periods (fig. 2). Two successive flood events were considered independent if the time interval between them was at least 5 days, following Svensson et al. (2005). This separation, dependent of the catchment area, generally allows for the flow to recede appreciably between peaks.

Fig. 2 – Mean daily discharge data of the Allier River at the gauging station of Moulins: recorded for the period 1968-2014 and reconstructed from discharge data recorded at the Cuffy station for the missing periods at Moulins (source: Banque Hydro).
Fig. 2 – Débits moyens journaliers de la rivière Allier à la station hydrologique de Moulins enregistrés pour la période 1968-2014 et reconstruits à partir de la station de Cuffy pour les périodes manquantes à Moulins (source : Banque Hydro).

Fig. 2 – Mean daily discharge data of the Allier River at the gauging station of Moulins: recorded for the period 1968-2014 and reconstructed from discharge data recorded at the Cuffy station for the missing periods at Moulins (source: Banque Hydro).Fig. 2 – Débits moyens journaliers de la rivière Allier à la station hydrologique de Moulins enregistrés pour la période 1968-2014 et reconstruits à partir de la station de Cuffy pour les périodes manquantes à Moulins (source : Banque Hydro).

The mean frequency of independent flood events and duration of flood days per year in each period are indicated for the different critical discharge thresholds (numbers indicated at the left and right of each cell, respectively). The discharge thresholds correspond to the double of the mean annual discharge (2MQ: 270 m³.s¹), 2-yr recurrence interval flood (HQ2: 669 m³.s¹), 5-yr recurrence interval flood (HQ5: 937 m³.s-¹), 10-yr recurrence interval flood (HQ10: 1,110 m³.s¹) and 20-yr recurrence interval flood (HQ20: 1,280 m³.s-¹). Time intervals are framed between aerial images and the years are defined from August 1st to July 31st.
La fréquence moyenne des crues indépendantes les unes des autres et la durée d’inondation annuelle (en nombre de jours) pour chaque période sont indiquées pour les différents seuils critiques de débits (nombres indiqués respectivement à gauche et à droite de chaque cellule). Les seuils de débits correspondent au double du débit annuel moyen (2MQ : 270 m³/s), et aux crues avec la période de retour de 2 ans (HQ2 : 669 m³/s), de 5 ans (HQ5 : 937 m³/s), de 10 ans (HQ10 : 1 110 m³/s) et de 20 ans (HQ20 : 1 280 m³/s). Les intervalles de temps sont encadrés entre les photographies aériennes. Les années couvrent la période du 1er août au 31 juillet.

3.4. Relation between trajectories of change and hydrology

21A bivariate Spearman's rho rank-order correlation was performed with the R package Hmisc (Harrell Jr, 2016) to find out how the trajectories of change and associated succession processes were correlated with the different flood frequencies and duration (fig. 2). Human-affected transitions (any phase ‑> agricultural land) were discarded for this analysis.

4. Results

4.1. Interpretation of current landscape mosaic and successional pathways

22The field survey conducted in 2014 covered 329 ha and provided a baseline to understand the ecology of the riverine ecosystem of the Allier River (fig. 3) and accordingly, to define a successional pathways scheme (fig. 4). Bare soil (5% of the mapped area) and pioneer (4%) succession phases mostly appeared in the river banks. Grasslands were broadly distributed within the study area (29%). In locations elevated above the channel dominated riparian forest (14%), composed mainly of black poplar (Populus nigra). White willow (Salix alba) forests appeared mostly restricted to silted side channels (3%). Prunus-Crataegus shrubs (11%) constituted the intermediate stage between (partly pastured) grassland and hardwood forest (6%).

Fig. 3 – Spatial distribution of vegetation types, organized by succession phases, mapped in the study site at the Allier River (Moulins, France) in 2014.
Fig. 3 – Répartition spatiale des types de végétation organisés par phases de succession et cartographiés dans le site d'étude sur la rivière Allier (Moulins, France) en 2014.

Fig. 3 – Spatial distribution of vegetation types, organized by succession phases, mapped in the study site at the Allier River (Moulins, France) in 2014.Fig. 3 – Répartition spatiale des types de végétation organisés par phases de succession et cartographiés dans le site d'étude sur la rivière Allier (Moulins, France) en 2014.

For colour figure, see online version of this article.
Pour la figure en couleur, voir la version en ligne de cet article.

Fig. 4 – Successional pathways scheme on dry, fresh and wet habitat conditions (zonation) based on the field mapping carried out in the study site at the Allier River (Moulins, France) in 2014.
Fig. 4 – Chemins de succession dans les conditions xérophiles, mésophiles et hygrophiles (zonation) sur la base de la cartographie effectuée dans le site d'étude sur la rivière Allier (Moulins, France) en 2014.

Fig. 4 – Successional pathways scheme on dry, fresh and wet habitat conditions (zonation) based on the field mapping carried out in the study site at the Allier River (Moulins, France) in 2014.Fig. 4 – Chemins de succession dans les conditions xérophiles, mésophiles et hygrophiles (zonation) sur la base de la cartographie effectuée dans le site d'étude sur la rivière Allier (Moulins, France) en 2014.

For colour figure, see online version of this article.
Pour la figure en couleur, voir la version en ligne de cet article.

23Depending on the habitat conditions, three successional pathways (dry, fresh and wet) were interpreted (fig. 4). The Allier River deposits gravel near the river banks at point bars during floods. These gravelly and highly draining areas are extremely dry, especially in summer (dry series). For that reason, pioneer patches remain open for long periods and are first colonized by Sedum sp. and dry grassland species, such as Corynephorus canescens. In this kind of conditions, P. nigra is one of those few riparian tree species that can establish by suckers, a form of clonal propagation (Corenblit et al., 2014). On bars covered with moist and finer substrate like sand and silt (fresh series), either grasslands or P. nigra establish. If recruitment conditions are favourable (open and moist habitats), P. nigra establishes along dense linear strips or within patches close to the channel (pioneer islands and benches), also with S. alba and S. purpurea in smaller proportions. If poplars do not recruit, those habitats are occupied by grassland. White willow forests, located in silted side channels, are flooded over long periods. Those locations are rich in nutrients and characterized by wet soil conditions (wet series). All series, although with relatively different species’ assemblages, may potentially evolve towards a post-pioneer mixed-forest, composed by a combination of woody and long-lived riparian (willow-poplar) and terrestrial (oak-elm-Robinia) arboreal species. However, only small patches in this mature stage were found in the study area.

4.2. Changes in the distribution of succession phases

24Over the majority of the 5-decade period (mainly before the year 2000) the Allier River at the study site has been characterized by a constant displacement of its channel and a change in the spatial array of succession phases (fig. 5). The correspondence between the succession phases across time was found to be significantly positive (chi-squared = 99.64, df = 42, p‑value < 0.0001), confirming the earlier visual interpretation of the data.

Fig. 5 – Spatial distribution of the seven succession phases in the study site at the Allier River over the last five decades (1967-2014).
Fig. 5 – Répartition spatiale des sept phases de succession au sein du site d'étude de l'Allier au cours des cinq dernières décennies (1967-2014).

Fig. 5 – Spatial distribution of the seven succession phases in the study site at the Allier River over the last five decades (1967-2014).Fig. 5 – Répartition spatiale des sept phases de succession au sein du site d'étude de l'Allier au cours des cinq dernières décennies (1967-2014).

For colour figure, see online version of this article.
Pour la figure en couleur, voir la version en ligne de cet article.

25Trends in the water (WA) and bare soil (BS) phases, represented together in Figure 6 and apart in Figure 7, suggest that the active tract used to cover around 35% of the study reach before the year 2000, and after that date it has experienced a marked drop, representing nowadays hardly 20% of the corridor. In this sense, the residual analysis showed that the bare soil has been significantly under-presented after 2005. It is especially remarkable the low value recorded in 2014, when bare soil (BS) represented only 5% of the study area (Pearson residuals = –2.59), the lowest value recorded in the entire study period. It seems that only the half upstream part of the site keeps nowadays some dynamism, due to its lower elevation and higher hydrological connection. In contrast, a homogenisation of the landscape by the herb phase is observed in the centre and downstream part of the site (fig. 5).

Fig. 6 – Changes in the area (right axis) / proportion of area (left axis) of each succession phase within the study area at the Allier River (1967-2014).
Fig. 6 – Changements de la superficie (axe de droite) / proportion de la superficie (axe de gauche) de chaque phase de succession au sein de la zone d'étude de l'Allier (1967-2014).

Fig. 6 – Changes in the area (right axis) / proportion of area (left axis) of each succession phase within the study area at the Allier River (1967-2014).Fig. 6 – Changements de la superficie (axe de droite) / proportion de la superficie (axe de gauche) de chaque phase de succession au sein de la zone d'étude de l'Allier (1967-2014).

Grey vertical lines denote years with historical photographs. The phases WA (water) and BS (bare soil) are shown together representing the active tract. PP. Pioneer phase; HP. Herb phase; SP. Shrub phase; FP. Forest phase; AGRI. Agricultural land. For colour figure, see online version of this article.
Les lignes verticales grises indiquent les années comportant des images historiques. Les phases WA (eau) et BS (sol nu) sont présentées ensemble ; elles représentent le chenal actif. PP. Phase pionnière ; HP. Phase herbacée ; SP. Phase arbustive ; FP. Phase arborée ; AGRI. Terrain agricole. Pour la figure en couleur, voir la version en ligne de cet article.

Fig. 7 – Mosaic plot of a 7 x 8 contingency table between succession phase (rows) and year-map (columns).
Fig. 7 – Graphique « mosaïque » d'un tableau de contingence de 7 x 8 entre la phase de succession (lignes) et l'année-carte (colonnes).

Fig. 7 – Mosaic plot of a 7 x 8 contingency table between succession phase (rows) and year-map (columns).Fig. 7 – Graphique « mosaïque » d'un tableau de contingence de 7 x 8 entre la phase de succession (lignes) et l'année-carte (colonnes).

The height of the tiles is proportional to the area of each succession phase in each year-map. The border and the colour of the tiles indicate the sign and magnitude of their standardized Pearson’s chi-squared residuals, i.e. they represent if the cell’s observed frequency is higher (solid border; deep blue if significant) or smaller (dashed border; deep red if significant) than that expected from random association between succession phases and year-maps. WA. Water; BS. Bare soil; PP. Pioneer phase; HP. Herb phase; SP. Shrub phase; FP. Forest phase; AGRI. Agricultural land. For colour figure, see online version of this article.
La hauteur des rectangles est proportionnelle à la surface de chaque phase de succession dans chaque année-carte. La bordure et la couleur des rectangles indiquent le signe et la magnitude de leurs résidus Khi-deux de Pearson normalisés, c'est-à-dire qu’ils indiquent si la fréquence observée de la cellule est plus élevée (bordure solide ; bleu profond si significatif) ou plus petite (bordure pointillée ; rouge foncé si significative) que celle attendue d'une association aléatoire entre les phases de succession et les années-cartes. WA. Eau ; BS. Sol nu ; PP. Phase pionnière ; HP. Phase herbacée ; SP. Phase arbustive ; FP. Phase arborée ; AGRI. Terrain agricole. Pour la figure en couleur, voir la version en ligne de cet article.

26In 1978, a significant decrease of agricultural land (AGRI) and increase in bare soil (BS) was observed (fig. 5, 7) that may be caused by the important channel shift and cut-off which occurred the same year. This morphological reconfiguration probably caused the remarkable increase of the pioneer phase (PP) in the subsequent map (Pearson residuals = +4.17). The proportion of forest (FP) has been slightly under-represented since 1967 until the year 2000; the forest area cover recorded in 1983 being especially low (only 6.5% of the reach; Pearson residuals = –2.18). However, since the 1980’s the amount of forest has increased at a steady rate, being over-represented after the year 2005 and covering more than 25% of the study area in 2014 (Pearson residuals = +2.80). On the contrary, the surface area proportions of the herb (HP: ~ 20%) and shrub (SP: ~ 10%) phases, despite certain fluctuations, have been relatively constant during the last five decades (denoted by the white colour of all the cells), showing only a slight non-significant increase and decrease, respectively after the year 2000.

4.3. Changes in the dominant succession processes

27In relation to the trajectories of change, more progression than retrogression was observed in all the periods, as the ratio progression / retrogression was always larger than 1 (tab. 1). This ratio was exceptionally high during the period 2000-2005 when progression tripled retrogression. For the period 1978-1983 there was more succession (progression + retrogression) than stability. For the periods 1967-1978 and 1985-2000 succession and stability were almost equilibrated. Nevertheless, stability was larger than succession for the rest of the periods, being especially notable during the last two periods (2005-2010 and 2010-2014) when stability tripled and doubled succession processes, respectively. In relation to this, the last ratio analysed indicated that all periods showed a stable portion covering between 40% and 55% of the total study area, but in the last two periods this proportion increased remarkably to 76% and 67%, respectively, indicating high similarity between overlaid maps.

Tab. 1 – Ratios of progression / retrogression, succession / stability and proportion of area stable within the study area across time.
Tab. 1 – Rapport de progression / rétrogression, succession / stabilité et proportion des surfaces stables au sein du site d’étude au cours du temps.

Tab. 1 – Ratios of progression / retrogression, succession / stability and proportion of area stable within the study area across time.Tab. 1 – Rapport de progression / rétrogression, succession / stabilité et proportion des surfaces stables au sein du site d’étude au cours du temps.

28The cluster analysis identified three groups of time periods regarding their succession processes’ patterns (fig. 8). Group 1 assembled the time intervals before 2005, i.e. 1967-2005, except for the period 1983-1985, which formed group 2. The time periods after 2005 (i.e. 2005-2014) constituted group 3. As group 2 was formed by only one time period, it was not considered in the Welch test. This statistical analysis confirmed that the mean rate values in certain succession processes of both groups of periods 1 and 3, i.e. before and after 2005, differed significantly (p‑value < 0.10; tab. 2). Differences were found in colonization 2a (WA / BS -> HP), which decreased from 2.724 to 0.741 ha.yr-¹ after 2005. Transition 1c (AGRI ‑> SP) and transition 2c (AGRI ‑> FP) were also significant, showing a decreasing trend after 2005. There were no statistical significant differences in total mean rate values of progression and retrogression trajectories (~ 15 and 10 ha.yr-¹, respectively for the entire study period 1967-2014). However, relevant significant differences were found in stability, whose values after 2005 doubled those observed before 2005, passing from 20 to 50 ha.yr-¹ of new natural areas stabilized. In addition, cultivation virtually disappears after 2005, denoting that no new areas are transformed into cultivated land (fig. 6, 7).

Fig. 8 – Hierarchical cluster analysis (average linkage method and Euclidean distance) of the standardized succession processes (rate ha.yr¹) for the seven time intervals between 1967 and 2014.
Fig. 8 – Dendrogramme (classification ascendante hiérarchique ; méthode d’agglomération « average » et distance euclidienne) des processus standardisés de succession (taux ha/an) pour les sept intervalles de temps entre 1967 et 2014.

Fig. 8 – Hierarchical cluster analysis (average linkage method and Euclidean distance) of the standardized succession processes (rate ha.yr‑¹) for the seven time intervals between 1967 and 2014.Fig. 8 – Dendrogramme (classification ascendante hiérarchique ; méthode d’agglomération « average » et distance euclidienne) des processus standardisés de succession (taux ha/an) pour les sept intervalles de temps entre 1967 et 2014.

Tab. 2 – Classification of trajectories of change and associated processes used in the evaluation of the chronological overlay of raster maps of succession phases.
Tab. 2 – Classification des trajectoires d’évolution et processus associés utilisés pour l’évaluation chronologique des phases de succession basée sur les images rasters.

Tab. 2 – Classification of trajectories of change and associated processes used in the evaluation of the chronological overlay of raster maps of succession phases.Tab. 2 – Classification des trajectoires d’évolution et processus associés utilisés pour l’évaluation chronologique des phases de succession basée sur les images rasters.

The description of each process depicts the change from one phase to another phase (t0 ‑> t+1) or the absence of a phase change. Then, the mean rate values (ha.yr-¹) and statistical value of difference (Welch test) of the different trajectories of change of group 1 (1967-2005, excluding 1983-1985) and group 3 (2005-2014) of time periods are shown. Raw and adjusted p‑values in bold mean statistical significance (p‑value < 0.10).
La description de chaque processus représente le passage d'une phase à une autre phase (t‑> t+1) ou l'absence d'un changement de phase. Ensuite, les taux moyen (ha/an) et résultats des tests statistiques (test de Welch) des différentes trajectoires d’évolution pour les groupe 1 (1967-2005, période 1983-1985 exclue) et groupe 3 (2005-2014) de période de temps sont montrés. Les p-valeurs brutes et ajustées en gras indiquent une significativité statistique (p-valeur < 0,10).

4.4. Relation between trajectories of change and hydrology

29Spearman's rho test revealed positive tendencies for all the significant correlations between succession processes and hydrological variables except for those involving stable processes (tab. 3). Specifically, stability of natural phases was negatively correlated with the duration of the double of the mean annual discharge (2MQ), denoting that the lower the discharge the higher the stability of the ecosystem. On the contrary, transition 1b (PP / HP ‑> SP) was positively related with this critical discharge. Significant positive correlations were found between the frequency and duration of floods of intermediate magnitude (HQ2 and HQ5) and the transition 1c (AGRI ‑> SP) and 2c (AGRI ‑> FP), as well as the initial-aggradation in progression (WA ‑> BS) and retrogression (any phase ‑> BS / PP). Furthermore, a significant positive correlation was also observed between the colonization 1 (WA / BS ‑> PP) and transition 1a (WA / BS ‑> SP) with the frequency and duration of floods of high magnitude (HQ10 and HQ20).

Tab. 3 – Spearman's rho correlations of succession processes and frequency and duration of flood events per year.
Tab. 3 – Corrélations rho de Spearman des processus de succession et fréquence et durée des inondations par an.

Tab. 3 – Spearman's rho correlations of succession processes and frequency and duration of flood events per year.Tab. 3 – Corrélations rho de Spearman des processus de succession et fréquence et durée des inondations par an.

Values in italics mean high correlation (rho > 0.6) and in bold and italics statistical significance (p‑value < 0.10). The discharge thresholds correspond to the double of the mean annual discharge (2MQ: 270 m³.s-¹), 2-yr recurrence interval flood (HQ2: 669 m³.s-¹), 5-yr recurrence interval flood (HQ5: 937 m³.s¹), 10-yr recurrence interval flood (HQ10: 1,110 m³.s-¹) and 20-yr recurrence interval flood (HQ20: 1,280 m³.s-¹).
Les valeurs en italique indiquent une forte corrélation (rho > 0.6) et les valeurs en italique et gras une significativité statistique (p-valeur < 0.10). Les seuils de débits correspondent au double du débit annuel moyen (2MQ : 270 m³/s), et aux crues avec la période de retour de 2 ans (HQ2 : 669 m³/s), de 5 ans (HQ5 : 937 m³/s), de 10 ans (HQ10 : 1 110 m³/s) et de 20 ans (HQ20 : 1 280 m³/s).

5. Discussion

30In the present study, we combined a variety of techniques to understand the vegetation succession within a dynamic meandering section of the Allier River over the last five decades, and in particular to confirm a shift visually observed after the beginning of the 21st century. For this purpose, (i) we conducted a comprehensive field survey to understand the current landscape mosaic and three successional pathways were identified (fig. 3, 4), (ii) we used a set of historical aerial images to identify changes in the spatio-temporal distribution of succession phases (fig. 5-7), (iii) we analysed the trajectories of change at the scale of the riparian corridor and identified a significant change in certain succession processes before and after the year 2005 (fig. 8; tab. 1, 2), and (iv) finally we correlated the observed spatio-temporal landscape changes to the discharge regime (tab. 3).

31Our analysis was constrained by the acquisition dates of the eight sets of multi-temporal aerial images spanning 50 years, which were not placed at equal intervals along the five decades. This made it necessary to standardize the raw data (vegetation and hydrology) to a year basis to ensure that any analytical results were not differentially affected by the length of the studied time periods (from 2 to 15 years). In longer periods, larger changes should be expected between images although this depends on (i) the frequency, magnitude and timing of the disturbance events and if they were captured or not by the precise moment when the images were taken, and (ii) if trajectories were unidirectional or alternated. The standardization necessarily assumed an equal distribution of landscape changes and flood events within the years of each period, which may have produced a certain overestimation of real values for short dynamic periods and an underestimation for long stable ones. Another constraint on our analysis was the resolution of the images that were used to explore the landscape changes. Images with the lowest resolution may have led to less accurate photointerpretation, although given the broad and distinct succession phases employed, it is unlikely that any serious errors in our analysis can be attributed to this factor. Still, due to the small sampling size (i.e. eight maps, seven periods) considered in the statistical analysis, the results must be interpreted with caution.

32Our analysis revealed a general increase in the proportion of the forest phase and a synchronous reduction in the bare soil phase of the riparian corridor of the lower course of the Allier River. Both have been accompanied by a decrease in the frequency and duration of moderate and high magnitude floods, especially after the year 2005 (fig. 2, 6, 7). In the 1978 and 1985 images, the riparian corridor reflects a high proportion of bare sediment (fig. 5-7). Four HQ5 floods took place before these both images were taken (fig. 2). The period 1978-1983 has been identified as the most dynamic one (i.e. with larger succession than stability). In the 1983 image a great increment in the pioneer phase is observed, which can be linked to the high magnitude floods (three HQ10 events and one HQ20 event) which all occurred in the spring time during the early 1980’s. Between 1985 and 2005, two HQ20 events were registered, which both can be related to a high retrogression and consecutive progression (e.g. colonization of bare soil by pioneer species). Quantitative observations could be made especially immediately after the HQ20 event in winter 2003 based on the landscape changes captured by the 2005 image. It seems that the period 2000-2005 has marked a clear state transition, where the width of the active tract has been significantly reduced and the proportion of the forest phase has become the highest recorded since aerial images exist for the Allier River study site (i.e. 1946). After 2005, the system has reduced considerably both progression and retrogression trajectories, and thus stability has turned out to be the dominant process at the corridor scale (tab. 1). Although progression has always been superior than retrogression during the entire study period, we did not find differences in total progression and retrogression on a yearly basis, as their rates of change (ha.yr-¹) were very similar before and after 2005 (tab. 2). This suggests that the system keeps a shifting habitat mosaic steady-state. However, remarkable differences were found in the stability rates, which doubled their values after 2005, passing from 20 to 50 ha.yr-¹. These changes in the landscape mosaic have been associated with the virtual disappearance of moderate (HQ5) and high magnitude floods (HQ10 and HQ20).

33In addition, the statistical analysis showed that the channel becomes less dynamic since a small but significant reduction in the transitions 1c and 2c after 2005 has been identified, which involved the change from cultivated land close to the channel to shrub and forest phases (tab. 2). Although significant, these changes did not represent large areas in the corridor, but this result suggests that river bank erosion rates are smaller than before 2005. Furthermore, both transitions were positively correlated with the frequency and duration of moderate magnitude floods (HQ2 and HQ5), as well as the initial-aggradation, i.e. progression from water to bare soil and the retrogression from any phase to bare soil (tab. 3).

34The progressions from water and bare soil to pioneer and shrub (colonization 1 and transition 1a, respectively) were positively correlated with HQ10 and HQ20, revealing that these high return-interval floods are able to reconfigure the landscape through the turn-over of habitats (Turner et al., 1998; Surian et al., 2015), promoting retrogression but also progression, as they provide potential new areas for vegetation recruitment. Consequently, the absence of these flood events during the last two study periods (2005-2014) may be considered as one of the reasons of the large stability observed.

35The progression from pioneer and herbs to shrubs (transition 1b; rho = 0.69; p‑value = 0.090) and the retrogression from any phase to bare soil and pioneer (initial aggradation-ret.: rho = 0.74; p‑value = 0.058) were positively correlated with the frequency of the double of the mean annual discharge (2MQ). This result suggests that this frequent but relatively low magnitude discharge plays an important role for the maintenance of a balanced ratio between bare and colonized surfaces in the active tract, as well as for the provision of the necessary soil moisture, nutrients and organic matter for the recently recruited vegetation (e.g. seedlings) to grow and persist. In addition, the duration of the 2MQ was negatively correlated with the proportion of the stable area within the study area (rho = –0.92; p‑value = 0.001; data not shown). Hence, the shorter the duration of this 2MQ discharge, the larger the overall stability of the corridor, which underscores the non-linear effect of established plants in increasing rapidly progression and stability modalities.

36Complementary to this result, we observed that on the lower Allier River a significant reduction of the maximum daily discharge (annual maximum flow) from 760 m³.s-¹ (mean value for the period 1967-2005; range: 206-1,390 m³.s-¹) to 555 m³.s-¹ (mean value for the period 2005-2014; range: 377-808 m³.s-¹) occurred. Since the level of lateral channel dynamics is reduced when flows above a critical discharge threshold are less frequent, this significant reduction in the magnitude of annual flood since 2005 (Fwe = 9.510, p‑value = 0.005), highlights the fundamental importance of annual floods and relatively frequent floods as key drivers of morphodynamics (Surian et al., 2015). As summarized by Assani et al. (2006), the alteration of the hydrological characteristics of the maximum annual flow can have other important ecological and morphological repercussions on the fluvial ecosystem. For example, its magnitude can affect the balance of competitive, ruderal, and stress-tolerant organisms in the ecosystem, and its frequency can influence population dynamics for various species (Poff et al., 1997; Nilsson and Svedmark, 2002). Egger et al. (2012) demonstrated that a maintained period of at least three years with very low maximum annual discharges may be enough for pioneer species like poplars and willows to establish and resist uprooting forces later on. In this sense, we consider that the Allier River is reaching a biogeomorphic threshold that may result in a shift towards a domain of biogeomorphic stability that then will require a sequence of large floods (HQ10 and HQ20) or one exceptional flood (e.g. > HQ50) to disrupt the landscape mosaic.

37Our observations about the current spatial arrangement of succession phases seems to be related to the absence of large floods during more than one decade. This absence of large floods could be inherent to the strong intra- and inter-annual variability of the flow regime of the Allier River (Onde, 1923). However, the current and future (21st century) impacts of climate change on the hydrological regime of the lower Allier River have been clearly shown by the recent projects ICC-Hydroqual (Moatar et al., 2010) and Explore 2070 (Chauveau et al., 2013).

38The current and future changes in the flood regime can have direct consequences in other associated factors important for vegetation dynamics, such as the mobilization of sediments within the active tract, and hence, the creation of potential areas for recruitment (Johnson, 2000). At the catchment scale, further changes can modify hydrosedimentary dynamics. The upper Allier River basin has known an intense deforestation in the 19th century for boat building, which produced large water and sediment fluxes at that time (Cournez, 2015). Since the 19th century, the upper basin has progressively recovered its vegetation cover, along with a decrease in the frequency and/or intensity of morphogenic hydrological events and a potential reduction of sediment supply to the river network. Recently, Dépret et al. (2017) hypothesized that these long-term changes in hydrosedimentary dynamics within the Loire river basin reduced the capacity of the Cher River (another major tributary of the Loire River) to erode its banks and thus to cause planform stability.

39Local factors may also be linked to the large landscape stability of the fluvial corridor observed since 2005. For example, formerly, large parts of the floodplain were influenced by grazing (Dejaifve and Duvaut, 2004; Cournez, 2015); now grazing affects fewer areas, and this may be influencing the change from open to more closed vegetation formations. In this sense, Geerling et al. (2006) found a trend towards a more evenly area distribution of succession phases that could be attributed to lower grazing intensities. In addition, during our field survey in 2014 and in different field visits before and after 2014, we observed inputs of fine sediment, organic matter and nutrients in the riparian zone from adjacent agricultural fields, which probably stimulate riparian vegetation growth.

40Considering that a strong link exists between hydrology, geomorphology and ecological processes (Church, 2002), the field survey carried out at the Allier River permitted to characterize the current landscape mosaic, but also to better understand the contemporary spatio-temporal mosaic of vegetation patches and channel dynamics. In particular, our analysis shows that the relation between the disturbance regime (flood frequency and magnitude) and vegetation dynamics (aptitudes of pioneer riparian plants to recruit, establish and resist) is a fundamental component of riparian ecosystems and landscape dynamics.

41The map of vegetation types from 2014 and the general schema of successional pathways and past trajectories (1967-2014) constitute a wealth of information for researchers and river conservationist and managers in the assessment of the evolution and future states of the riparian corridor according to different scenarios. Bino et al. (2015) suggested that shifts in inundation regimes can drive succession and establish new stable states, determined by the magnitude and duration of the hydrological perturbation. Currently, we can expect that under a similar climate and flood regime as the one observed since 2005, the forest phase will continue to increase causing a homogenisation of the fluvial landscape, dominated by dense forest patches. Grassland areas, which are not open (i.e. soil receives low light) and too dry for seeds of woody riparian species to establish, might be progressively colonized by clonal propagation (e.g. root suckering) of Populus nigra (Barsoum et al., 2004; Corenblit et al., 2014), but only if these areas maintain a moderate soil moisture or access to the phreatic zone. This species’ strategy / mechanism has been observed to be naturally potentiated under low hydrodynamic river conditions, for example in regulated rivers (Smulders et al., 2008). Therefore, a large area still exists to be potentially colonized by poplars in the study area. However, if the grassland areas are stable, relatively elevated without access to the phreatic zone and hydrologically disconnected for a long period of time, these areas could be too dry for woody riparian species to establish (even by clonal reproduction). In that case, the succession towards a hardwood forest would proceed, but depending on the slow colonization by strictly terrestrial woody species adapted to dry soil conditions. Such colonization would imply a shift in species composition and a consistent steady increase in the forest woodland area of the riparian corridor.

42Alteration of flow regimes under climate change are believed to be critical to population dynamics in European river ecosystems (Schneider et al., 2011). Disconnection and alteration of flooding have been documented as critical to riparian species survival, such as P. nigra (Villar and Forestier, 2009). In this sense, the predicted climate change impact for P. nigra in the Loire basin is rather strong, a decrease of 25% in its current population is expected, linked to its position in the dynamic riparian zone that shows strongest impacts of predicted hydrological changes (Van Looy and Piffady, 2017).

43The observed change, i.e. reduction of the flood regime since 2005, already induced landscape trajectory changes and probably will have further important ecological consequences for the riverine ecosystem of the Allier River. Depending on the future management of the Natural Reserve, these consequences would include: (i) the reduction in the turnover of riparian habitats and hence the reduction of potential new vegetation recruitment areas; (ii) the increase in hydric stress on established riparian trees located further away from the channel margins (Scott et al., 1996; Auble and Scott, 1998; Katz et al., 2005); (iii) and the likely woody plant encroachment (Huxman et al., 2005), which may eventually drive floodplain structure and composition towards less flood-reliant species, for example promoting the establishment of more late-seral and terrestrial woodland species (Glaeser and Wulf, 2009; Garófano-Gómez et al., 2013), principally because their regeneration is not so dependent on recurrent fluvio-geomorphic events (González et al., 2010). In addition, changes in river channel dynamics caused by geomorphic channel adjustments to the observed reduced flood regime could favour the propagation of invasive plant species within the riparian corridor (Matsubara and Sakai, 2016), like the Japanese knotweed (Fallopia japonica), recognized as potentially problematic for the conservation of the Allier River and other rivers in France and Europe (Schnitzler and Muller, 1998; Haury et al., 2010). All these ecological changes have been largely observed in regulated rivers; however, they have not been so evident in unregulated river systems until now. Our results show that the lower Allier River, despite a largely unregulated flood regime and little river training works, may be going through a comparable situation due to the current and future change in climate conditions and its impact on the flow and flood regimes as observed and predicted by the French research projects ICC-Hydroqual (Moatar et al., 2010) and Explore 2070 (Chauveau et al., 2013), showing a general tendency to decline.

6. Conclusion

44The present study clearly demonstrated that the low level of human impacts and natural high river dynamism permits the lower course of the Allier River (France) to be considered as a contemporary reference system of natural plant succession processes in temperate meandering riverine ecosystems. Furthermore, it showed that during the period 1967-2005, the shifting habitat mosaic of the lower Allier River is modulated by long-term changes in the natural hydrological disturbance regime, thus confirming previous findings of Geerling et al. (2006). However, a transition of the landscape mosaic started after the year 2000 and a clear shift has been detected after the year 2005, which consists in an increase in both, the stability of the floodplain and the forest woodland cover. This shift of the fluvial corridor landscape in the lower course of the Allier River was linked to the reduction in the hydrogeomorphic disturbances (frequency, magnitude and duration of flood events) which in turn was related to observed and predicted impacts of climate change on the flow regime of the lower Allier River (Moatar et al., 2010; Chauveau et al., 2013; Giuntoli et al., 2013). Thus, our results contribute to a better comprehension of future succession trajectories of temperate riparian and floodplain ecosystems within the context of global climate change.

Haut de page

Bibliographie

Assani A.A., Stichelbout É., Roy A.G., Petit F. (2006) – Comparison of impacts of dams on the annual maximum flow characteristics in three regulated hydrologic regimes in Québec (Canada). Hydrological Processes, 20 (16), 3485-3501.
DOI : 10.1002/hyp.6150

Auble G.T., Scott M.L. (1998) – Fluvial disturbance patches and cottonwood recruitment along the upper Missouri River, Montana. Wetlands, 18 (4), 546-556.
DOI : 10.1007/BF03161671

Barsoum N., Muller E., Skot L. (2004) – Variations in levels of clonality among Populus nigra L. stands of different ages. Evolutionary Ecology, 18 (5-6), 601-624.
DOI : 10.1007/s10682-004-5146-4

Benjamini Y., Hochberg Y. (1995) – Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, Series B (Methodological), 57 (1), 289-300.

Bino G., Sisson S.A., Kingsford R.T., Thomas R.F., Bowen S. (2015) – Developing state and transition models of floodplain vegetation dynamics as a tool for conservation decision-making: a case study of the Macquarie Marshes Ramsar wetland. Journal of Applied Ecology, 52 (3), 654-664.
DOI : 10.1111/1365-2664.12410

Bormann F.H., Likens G.E. (1979) – Catastrophic disturbance and the steady state in Northern hardwood forests: a new look at the role of disturbance in the development of forest ecosystems suggests important implications for land-use policies. American Scientist, 67 (6), 660-669.

Buijse A.D., Coops H., Staras M., Jans L.H., Geest G.J.V., Grift R.E., Ibelinds B.W., Oosterberg W., Roozen F.C.J.M. (2002) – Restoration strategies for river floodplains along large lowland rivers in Europe. Freshwater Biology, 47, 889-907.
DOI : 10.1046/j.1365-2427.2002.00915.x

Ceschin S., Tombolini I., Abati S., Zuccarello V. (2015) – The effect of river damming on vegetation: is it always unfavourable? A case study from the River Tiber (Italy). Environmental Monitoring and Assessment, 187, 301, 12 p.
DOI : 10.1007/s10661-015-4521-7

Corenblit D., Tabacchi E., Steiger J., Gurnell A.M. (2007) – Reciprocal interactions and adjustments between fluvial landforms and vegetation dynamics in river corridors: a review of complementary approaches. Earth-Science Reviews, 84 (1‑2), 56-86.
DOI :
10.1016/j.earscirev.2007.05.004

Corenblit D., Steiger J., González E., Gurnell A.M., Charrier G., Darrozes J., Dousseau J., Julien F., Lambs L., Larrue S., Roussel E., Vautier F., Voldoire O. (2014) – The biogeomorphological life cycle of poplars during the fluvial biogeomorphological succession: a special focus on Populus nigra L. Earth Surface Processes and Landforms, 39 (4), 546-563.
DOI :
10.1002/esp.3515

Cournez E. (2015) – Sur les traces de l'Allier. Histoire d'une rivière sauvage. Tomacom, Cesset, 256 p.

Chauveau M., Chazot S., Perrin C., Bourgin P.-Y., Sauquet E., Vidal J.-P., Rouchy N., Martin E., David J., Norotte T., Maugis P., De Lacaze X. (2013) – Quels impacts des changements climatiques sur les eaux de surface en France à l’horizon 2070 ? La Houille Blanche (4), 5-15.
DOI : 10.1051/lhb/2013027

Church M. (2002) – Geomorphic thresholds in riverine landscapes. Freshwater Biology, 47 (4), 541-557.
DOI : 10.1046/j.1365-2427.2002.00919.x

Deiller A.-F., Walter J.-M.N., Trémolières M. (2001) – Effects of flood interruption on species richness, diversity and floristic composition of woody regeneration in the Upper Rhine alluvial hardwood forest. Regulated Rivers: Research & Management, 17 (4-5), 393-405.
DOI : 10.1002/rrr.649

Dejaifve P.-A., Duvaut C. (2004) – Pastoralisme dans la Reserve Naturelle Nationale du Val d'Allier. Cartographie des los amodies (limites, utilisation et pression par les tropeaux). LPO auvergne, ONF, DIREN Auvergne, Conseil Régional d’Auvergne, 18 p.

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

Diaz-Redondo M., Egger G., Marchamalo M., Hohensinner S., Dister E. (2017) – Benchmarking fluvial dynamics for process-based river restoration: the Upper Rhine River (1816-2014). River Research and Applications, 33 (3), 403-414.
DOI : 10.1002/rra.3077

DREAL (2014) – Les cartes de végétation, système d’information du lit de la Loire (SIEL). http://www.centre.developpement-durable.gouv.fr/. Access on September 2014.

Dufour S., Rinaldi M., Piégay H., Michalon A. (2015) – How do river dynamics and human influences affect the landscape pattern of fluvial corridors? Lessons from the Magra River, Central–Northern Italy. Landscape and Urban Planning, 134, 107-118.
DOI : 10.1016/j.landurbplan.2014.10.007

Egger G., Politti E., Lautsch E., Benjankar R., Gill K.M., Rood S.B. (2015) – Floodplain forest succession reveals fluvial processes: a hydrogeomorphic model for temperate riparian woodlands. Journal of Environmental Management, 167, 72-82.
DOI : 10.1016/j.jenvman.2015.06.018

Egger G., Politti E., Woo H., Cho K.-H., Park M., Cho H., Benjankar R., Lee N.-J., Lee H. (2012) – Dynamic vegetation model as a tool for ecological impact assessments of dam operation. Journal of Hydro-environment Research, 6 (2), 151-161.
DOI : 10.1016/j.jher.2012.01.007

Fédération de la Région Auvergne pour la Nature et l'Environnement-FRANE (2014) – Parlons des crues de la rivière Allier, Ceyrat, France, 48 p.

Friendly M. (1994) – Mosaic displays for multi-way contingency tables. Journal of the American Statistical Association, 89 (425), 190-200.
DOI : 10.2307/2291215

Garófano-Gómez V., Martínez-Capel F., Bertoldi W., Gurnell A., Estornell J., Segura-Beltrán F. (2013) – Six decades of changes in the riparian corridor of a Mediterranean river: a synthetic analysis based on historical data sources. Ecohydrology, 6 (4), 536-553.
DOI : 10.1002/eco.1330

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 : 10.1080/09853111.2000.11105360

Geerling G.W., Ragas A.M.J., Leuven R.S.E.W., van den Berg J.H., Breedveld M., Liefhebber D., Smits A.J.M. (2006) – Succession and rejuvenation in floodplains along the River Allier (France). Hydrobiologia, 565 (1), 71-86.
DOI : 10.1007/s10750-005-1906-6

Giuntoli I., Renard B., Vidal J.P., Bard A. (2013) – Low flows in France and their relationship to large-scale climate indices. Journal of Hydrology, 482, 105-118.
DOI : 10.1016/j.jhydrol.2012.12.038

Glaeser J., Wulf M. (2009) – Effects of water regime and habitat continuity on the plant species composition of floodplain forests. Journal of Vegetation Science, 20 (1), 37-48.
DOI : 10.1111/j.1654-1103.2009.05282.x

González E., González-Sanchis M., Cabezas Á., Comín F., Muller E. (2010) – Recent changes in the riparian forest of a large regulated Mediterranean river: implications for management. Environmental Management, 45 (4), 669-681.
DOI : 10.1007/s00267-010-9441-2

Gurnell A.M., Corenblit D., García de Jalón D., González del Tánago M., Grabowski R.C., O'Hare M.T., Szewczyk M. (2016) – A conceptual model of vegetation–hydrogeomorphology interactions within river corridors. River Research and Applications, 32 (2), 142-163.
DOI : 10.1002/rra.2928

Gurnell A.M., Petts G.E. (2002) – Island-dominated landscapes of large floodplain rivers, a European perspective. Freshwater Biology, 47 (4), 581-600.
DOI : 10.1046/j.1365-2427.2002.00923.x

Habersack H., Haspel D., Kondolf M. (2014) – Large rivers in the Anthropocene: insights and tools for understanding climatic, land use, and reservoir influences. Water Resources Research, 50 (5), 3641-3646.
DOI : 10.1002/2013wr014731

Harrell Jr F.E. (2016) – Hmisc: Harrell Miscellaneous. R package version 3.17-4. https://CRAN.R-project.org/package=Hmisc.

Haury J., Hudin S., Matrat R., Anras L. et al. (2010) – Manuel de gestion des plantes exotiques envahissant les milieux aquatiques et les berges du bassin Loire-Bretagne. Fédération des conservatoires d'espaces naturels, 136 p.

Hortobágyi B., Corenblit D., Steiger J., Peiry J.-L. (in press) – Niche construction within riparian corridors. Part I: Exploring biogeomorphic feedback windows of three pioneer riparian species, Allier River, France. Geomorphology.
DOI : 10.1016/j.geomorph.2017.08.048

Hupp C.R., Rinaldi M. (2007) – Riparian vegetation patterns in relation to fluvial landforms and channel evolution along selected rivers of Tuscany (Central Italy). Annals of the Association of American Geographers, 97 (1), 12-30.
DOI : 10.1111/j.1467-8306.2007.00521.x

Huxman T.E., Wilcox B.P., Breshears D.D., Scott R.L., Snyder K.A., Small E.E., Hultine K., Pockman W.T., Jackson R.B. (2005) – Ecohydrological implications of woody plant encroachment. Ecology, 86 (2), 308-319.
DOI : 10.1890/03-0583

Johnson W.C. (2000) – Tree recruitment and survival in rivers: infuence of hydrological processes. Hydrological Processes, 14, 3051-3074.
DOI : 10.1002/1099-1085(200011/12)14:16/17<3051::AID-HYP134>3.0.CO;2-1

Katz G.L., Friedman J.M., Beatty S.W. (2005) – Delayed effects of flood control on a flood-dependent riparian forest. Ecological Applications, 15 (3), 1019-1035.
DOI : 10.1890/04-0076

Leyer I. (2005) – Predicting plant species’ responses to river regulation: the role of water level fluctuations. Journal of Applied Ecology, 42 (2), 239-250.
DOI : 10.1111/j.1365-2664.2005.01009.x

Matsubara Y., Sakai S. (2016) – The role of flood regime on invasive success of exotic species growing in riparian environments. Biological Invasions, 18 (3), 793-808.
DOI : 10.1007/s10530-015-1049-2

Metz M. (2015) – Process-oriented ecological analysis of a dynamic riparian ecosystem: the lower river Allier (France). Master thesis, Karlsruhe Institute of Technology, Karlsruhe, Germany, 84 p.

Meybeck M. (2003) – Global analysis of river systems: from Earth system controls to Anthropocene syndromes. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 358 (1440), 1935-1955.
DOI : 10.1098/rstb.2003.1379

Moatar F., Ducharne A., Thiéry D., Bustillo V., Sauquet E., Vidal J.-P. (2010) – La Loire à l'épreuve du changement climatique. Geosciences, 12, 78-87.

Mouw J.E.B., Chaffin J.L., Whited D.C., Hauer F.R., Matson P.L., Stanford J.A. (2013) – Recruitment and successional dynamics diversify the shifting habitat mosaic of an Alaskan floodplain. River Research and Applications, 29 (6), 671-685.
DOI : 10.1002/rra.2569

Muñoz-Mas R., Garófano-Gómez V., Andrés-Doménech I., Corenblit D., Egger G., Francés F., Ferreira M.T., García-Arias A., Politti E., Rivaes R., Rodríguez-González P.M., Steiger J., Vallés-Morán F.J., Martínez-Capel F. (2017) – Exploring the key drivers of riparian woodland successional pathways across three European river reaches. Ecohydrology, e1888.
DOI : 10.1002/eco.1888

Nilsson C., Berggren K. (2000) – Alterations of riparian ecosystems caused by river regulation. Bioscience, 50 (9), 783-792.
DOI : 10.1641/0006-3568(2000)050[0783:AORECB]2.0.CO;2

Nilsson C., Svedmark M. (2002) – Basic principles and ecological consequences of changing water regimes: riparian plant communities. Environmental Management, 30 (4), 468-480.
DOI : 10.1007/s00267-002-2735-2

Onde H. (1923) – Les crues de l'Allier. Revue de géographie alpine, 11 (2), 301-372.
DOI : 10.3406/rga.1923.5511

Peel M.C., Finlayson B.L., McMahon T.A. (2007) – Updated world map of the Köppen-Geiger climate classification. Hydrology and Earth System Sciences Discussions, 4 (2), 439-473.
DOI : 10.5194/hessd-4-439-2007

Petts G.E., Möller H., Roux A.L. (1989) – Historical change of large alluvial rivers: Western Europe. John Wiley and Sons Ltd., Chichester, 355 p.

Poff N.L., Allan J.D., Bain M.B., Karr J.R., Prestegaard K.L., Richter B.D., Sparks R.E., Stromberg J.C. (1997) – The natural flow regime: a paradigm for river conservation and management. Bioscience, 47 (11), 769-784.
DOI : 10.2307/1313099

R Core Team. (2016) – R: a language and environment for statistical computing. R Foundation for Statistical Computing, Viena, Austria. URL http://www.R-project.org/

Rackusch S. (2014) – Animierte Darstellung von Mäanderverlagerungen am Beispiel der Loire und des Allier in Frankreich. Bachelor thesis, Hochschule Karlsruhe-Technik und Wirtschaft, Karlsruhe, Germany, 50 p.

Richter B., Baumgartner J., Wigington R., Braun D. (1997) – How much water does a river need? Freshwater Biology, 37 (1), 231-249.
DOI : 10.1046/j.1365-2427.1997.00153.x

Scott M.L., Friedman J.M., Auble G.T. (1996) – Fluvial process and the establishment of bottomland trees. Geomorphology, 14 (4), 327-339.
DOI : 10.1016/0169-555x(95)00046-8

Schneider C., Flörke M., Geerling G., Duel H., Grygoruk M., Okruszko T. (2011) – The future of European floodplain wetlands under a changing climate. Journal of Water and Climate Change, 2 (2-3), 106-122.
DOI : 10.2166/wcc.2011.020

Schnitzler A., Muller S. (1998) – Ecology and biogeography of highly invasive plants in Europe: giant knotweeds from Japan (Fallopia japonica and F. sachalinensis). Revue d'Ecologie (la Terre et la Vie), 53 (1), 3-38.

Schwarzmeier J. (2015) – Sukzessionsgeschwindigkeit und Standortsalter der Auenvegetation am Allier (Frankreich). Thesis for teaching degree, Karlsruhe Institute of Technology, Rastatt, Germany, 50 p.

Smulders M.J.M., Cottrell J.E., Lefèvre F., van der Schoot J., Arens P., Vosman B., Tabbener H.E., Grassi F., Fossati T., Castiglione S., Krystufek V., Fluch S., Burg K., Vornam B., Pohl A., Gebhardt K., Alba N., Agúndez D., Maestro C., Notivol E., Volosyanchuk R., Pospíšková M., Bordács S., Bovenschen J., van Dam B.C., Koelewijn H.P., Halfmaerten D., Ivens B., van Slycken J., Vanden Broeck A., Storme V., Boerjan W. (2008) – Structure of the genetic diversity in black poplar (Populus nigra L.) populations across European river systems: consequences for conservation and restoration. Forest Ecology and Management, 255 (5-6), 1388-1399.
DOI : 10.1016/j.foreco.2007.10.063

Stanford J.A., Lorang M.S., Hauer F.R. (2005) – The shifting habitat mosaic of river ecosystems. Proceedings of the International Association of Theoretical and Applied Limnology, 29 (1), 123-136.

Stanford J.A., Ward J.V., Liss W.J., Frisell C.A., Williams R.N., Lichatowich J.A., Coutant C.C. (1996) – A general protocol for restoration of regulated rivers. Regulated Rivers: Research & Management, 12 (4-5), 391-413.
DOI : 10.1002/(SICI)1099-1646(199607)12:4/5<391::AID-RRR436>3.0.CO;2-4

Steiger J., Tabacchi E., Dufour S., Corenblit D., Peiry J.L. (2005) – Hydrogeomorphic processes affecting riparian habitat within alluvial channel–floodplain river systems: a review for the temperate zone. River Research and Applications, 21 (7), 719-737.
DOI : 10.1002/rra.879

Surian N., Barban M., Ziliani L., Monegato G., Bertoldi W., Comiti F. (2015) – Vegetation turnover in a braided river: frequency and effectiveness of floods of different magnitude. Earth Surface Processes and Landforms, 40 (4), 542-558.
DOI : 10.1002/esp.3660

Svensson C., Kundzewicz W.Z., Maurer T. (2005) – Trend detection in river flow series: 2. Flood and low-flow index series. Hydrological Sciences Journal, 50 (5), 811-824.
DOI : 10.1623/hysj.2005.50.5.811

Tockner K., Lorang M.S., Stanford J.A. (2010) – River flood plains are model ecosystems to test general hydrogeomorphic and ecological concepts. River Research and Applications, 26 (1), 76-86.
DOI : 10.1002/rra.1328

Tockner K., Stanford J.A. (2002) – Riverine flood plains: present state and future trends. Environmental Conservation, 29 (3), 308-330.
DOI : 10.1017/S037689290200022X

Trémolières M., Sánchez-Pérez J.M., Schnitzler A., Schmitt D. (1998) – Impact of river management history on the community structure, species composition and nutrient status in the Rhine alluvial hardwood forest. Plant Ecology, 135 (1), 59-78.
DOI : 10.1023/a:1009756428824

Turner M.G., Baker W.L., Peterson C.J., Peet R.K. (1998) – Factors influencing succession: lessons from large, infrequent natural disturbances. Ecosystems, 1 (6), 511-523.
DOI : 10.1007/s100219900047

Van Looy K., Meire P., Wasson J.-G. (2008) – Including riparian vegetation in the definition of morphologic reference conditions for large rivers: A case study for Europe’s Western plains. Environmental Management, 41 (5), 625-639.
DOI : 10.1007/s00267-008-9083-9

Van Looy K., Piffady J. (2017) – Metapopulation modelling of riparian tree species persistence in river networks under climate change. Journal of Environmental Management, 202, Part 2, 437-446.
DOI : 10.1016/j.jenvman.2016.11.019

Villar M., Forestier O. (2009) – Le Peuplier noir en France: pourquoi conserver ses ressources génétiques et comment les valoriser? Reveu Forestière Française, 5, 457-466.
DOI : 10.4267/2042/31523

Ward J.V., Tockner K., Arscott D.B., Claret C. (2002) – Riverine landscape diversity. Freshwater Biology, 47 (4), 517-539.
DOI : 10.1046/j.1365-2427.2002.00893.x

Wilcox R.R. (2012) – Introduction to robust estimation and hypothesis testing. 3rd ed. Academic Press, United States, 608 p.

Wilcox R.R., Schönbrodt F.D. (2014) – The WRS package for robust statistics in R (version 0.24). http://r-forge.r-project.org/projects/wrs/

Haut de page

Annexe

Version française abrégée

De nombreux écosystèmes riverains en Europe ont perdu leur caractère fortement dynamique suite aux impacts anthropiques tels que la chenalisation des cours d’eau et la régulation des débits (Buijse et al., 2002). Le bas Allier (France) est l’un des derniers secteurs de cours d’eau avec des méandres mobiles (Van Looy et al., 2008). Il constitue ainsi une opportunité pour étudier les processus de succession de la végétation riveraine dans un contexte de cours d’eau dynamique de la zone tempérée en cours d’ajustement en réponse aux conditions climatiques.

L’objectif était d’étudier les dynamiques de successions végétales récentes de l’Allier en relation avec le changement observé du régime des crues. L’analyse de Geerling et al. (2006) d’un secteur du corridor riverain de l’Allier au sud de Moulins (France), qui a proposé que le modèle de « shifting mosaic steady-state » (Bormann and Likens, 1979) y était applicable pour la seconde moitié du XXe siècle, a servi comme référence. Nous avons (i) conduit une campagne de mesures de terrain à large échelle et avons interprété la mosaïque paysagère actuelle et les trajectoires de succession riveraine ; (ii) analysé les changements dans la distribution spatiale des phases et des trajectoires de succession à partir de huit jeux de photographies aériennes couvrant 47 années (1967-2014) ; (iii) analysé les processus de succession végétale, les séries hydrologiques contemporaines et historiques, ainsi que les ajustements morphologiques du chenal afin d’interpréter les modalités spatio-temporelles d’ajustement de la mosaïque d’habitats.

Le secteur d’étude (longueur : 3,7 km) est situé 8 km en amont de la ville de Moulins (fig. 1), dans la section méandriforme la plus mobile du bas Allier, au sein de la Réserve Naturelle Nationale du Val d’Allier. Les huit jeux de photographies aériennes (1967, 1978, 1983, 1985, 2000, 2005, 2010 et une carte de la végétation établie en 2014 à partir de relevés de terrain), ont été utilisés pour l’analyses des changements spatio-temporels de la mosaïque d’habitats, c’est-à-dire les phases de succession et les trajectoires de succession, incluant les modalités de progression, rétrogression, stabilité et d’anthropisation, sensu Diaz-Redondo et al. (2017). Les phases de succession sont : eau ; sol nu (gravier et sable avec couverture végétale < 5 %) ; phase pionnière (végétation rudérale adaptée aux perturbations avec couverture ~ 25 % et ~ 0,5 m de hauteur) ; phase herbacée (herbes annuelles, roseaux et prairies avec couverture > 25 % et < 1 m de hauteur) ; phase arbustive (herbes et arbustes ligneux qui peuvent interagir fortement avec le débit et la dynamique des sédiments, avec couverture ≤ 60 % et < 5 m de hauteur) ; phase arborée (couverture ligneuse > 60 % et > 5 m de hauteur) ; terrain agricole.

Les analyses statistiques ont comporté : (i) un test du Khi-deux d’indépendance de Pearson (Friendly, 1994) pour déterminer s’il existe une relation significative entre les phases de succession et les périodes de temps considérées ; (ii) le calcul des rapports entre les modalités de progression, rétrogression et stabilité pour chaque période de temps afin de détecter des trajectoires de changement dans le temps et les périodes les plus dynamiques ; (iii) une analyse hiérarchique ascendante pour comprendre comment les différentes périodes de temps diffèrent par rapport à leur trajectoire de changement ; (iv) une ANOVA robuste de Welch à une voie (Wilcox, 2012) pour déterminer les différents processus de succession résultant du regroupement par période de temps et (v) une corrélation bi-variée des rangs de Spearman pour comprendre comment les trajectoires de changement sont corrélées avec les différentes fréquences et durées des crues (fig. 2).

La campagne de relevés de terrain conduite en 2014 a constitué la référence actuelle et trois trajectoires de succession dépendantes des conditions d’habitat (séries xérophile, mésophile et hygrophile) ont été identifiées (fig. 3-4). Les résultats ont montré que le secteur d’étude a connu un déplacement constant du chenal et un changement dans la disposition des phases de succession, avec le maintien d’une mosaïque d’habitat dynamique (shifting habitat mosaic) entre 1967 et 2005 (fig. 5). La bande active a eu tendance à recouvrir 35 % du secteur d’étude avant l’année 2000, et à la suite de cette date son recouvrement au sein du corridor riverain a diminué de manière importante, avec actuellement une valeur proche de 20 % (fig. 6-7). De plus, les surfaces à recouvrement forestier se sont étendues avec un taux d’accroissement régulier depuis les années 1980, passant de 6,5 % de recouvrement en 1983 à plus de 25 % en 2014. Une légère augmentation de la phase herbacée, en concomitance avec une diminution de la phase « sol nu » (recouvrement de seulement 5 % du site d’étude en 2014), a été détectée après 2005.

Pour l’ensemble des périodes étudiées, la modalité « progression » des trajectoires de changement a été dominante par rapport à la modalité « rétrogression » (tab. 1). Durant les deux dernières périodes (2005-2010 et 2010-2014), la modalité « stabilité » a dominé les modalités de succession (progression + rétrogression). Ainsi, en moyenne, les surfaces stables ont évolué au sein du corridor riverain de 49 % avant 2005 à 72 % après 2005. Dans ce sens, la classification hiérarchique ascendante a permis d’identifier très clairement un seuil critique au cours de l’année 2005 (fig. 8). Des différences significatives ont été observées pour la modalité « stabilité », augmentant de 20 à 50 ha/an de nouvelles zones naturelles stabilisées après 2005 (tab. 2). Ces changements dans la mosaïque paysagère ont été expliqués par la quasi-disparition des crues modérées (HQ5) et des crues de plus grande ampleur (HQ10 et HQ20), en particulier après 2005 (fig. 2). En outre, le chenal a perdu de son dynamisme puisqu’une réduction du taux d'érosion des berges après 2005 a été observée. Cette réduction de l’érosion latérale s’est traduite par une réduction faible mais significative de la transition des terres cultivées vers les phases arbustives et arborées dans la zone proche du chenal, qui a donc été de moins en moins remaniée (tab. 2). Une corrélation forte entre ce changement et la magnitude et fréquence des crues modérées HQet HQa également été mise en évidence (tab. 3).

Les résultats de notre étude suggèrent que la rivière Allier a atteint un seuil critique biogéomorphologique qui pourrait potentiellement résulter en un basculement du système vers un domaine de stabilité biogéomorphologique qui nécessitera une série de crues de très forte intensité pour remobiliser la mosaïque paysagère (Turner et al., 1998). Cependant, non seulement les grandes crues avec un intervalle de retour important (c.-à-d. des événements ponctuels de forte intensité et faible fréquence) ont connu un déclin significatif, mais également les débits maximums annuels, c'est-à-dire la période de hautes eaux de chaque année (de 760 m³/s avant 2005 à 555 m³/s après 2005). Ces résultats soulignent également l'importance fondamentale des crues annuelles dans le maintien d'un rapport équilibré entre les surfaces nues et colonisées par la végétation riveraine (Assani et al., 2006).

Bien que les projections des impacts du changement climatique sur le régime de crues au cours du XXIe restent plus incertaines que ceux sur les étiages (Moatar et al., 2010), de récentes études prévoient une réduction de 10 % des crues décennales (HQ10) sur le bas Allier (Chauveau et al., 2013), avec des impacts significatifs sur la végétation riveraine (Van Looy et Piffady, 2017). Nous suggérons que sous un régime de crues semblable à celui observé depuis 2005, caractérisé par l’absence de perturbations hydrologiques suffisantes pour rajeunir les formes fluviatiles, la phase arborée continuera à augmenter. Cette augmentation de la phase arborée mènera alors à l’homogénéisation du paysage fluvial, dominé par des groupements alluviaux forestiers. Les prairies humides pourraient rapidement évoluer via une propagation clonale de Populus nigra vers de la peupleraie arborée. Cependant, dans d'autres secteurs plus secs, cette trajectoire d’évolution des prairies pourrait être ralentie par la difficulté de leur colonisation par propagation sexuée ou végétative des espèces ligneuses proprement forestières. Au final, notre travail permet une meilleure compréhension des trajectoires de succession futures du bas Allier dans le contexte du changement climatique global, qui modifie le régime hydrologique et en retour les processus hydrogéomorphologiques, les ajustements des formes fluviatiles et la dynamique de succession végétale.

Haut de page

Table des illustrations

Titre Fig. 1 – Location of the study reach within the Allier River Basin, near the French town of Moulins (A). The red line indicates the study area covered by the vegetation mapping (B).Fig. 1 – Localisation du tronçon d’étude dans le bassin versant de la rivière Allier, en amont de la ville de Moulins, France (A). Le trait rouge indique la zone d'étude couverte par la cartographie de la végétation (B).
Légende Aerial image from July 2013 (source: CRAIG – Centre Régional Auvergnat de l’Information Géographique). The spatial reference is represented by geographic coordinates (latitude / longitude in degrees) relative to the WGS‑84 datum.Image aérienne de juillet 2013 (source : CRAIG – Centre Régional Auvergnat de l’Information Géographique). La référence spatiale est représentée par des coordonnées géographiques (latitude / longitude en degrés) par rapport au système de référence géodésique WGS-84.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-1.jpg
Fichier image/jpeg, 2,8M
Titre Fig. 2 – Mean daily discharge data of the Allier River at the gauging station of Moulins: recorded for the period 1968-2014 and reconstructed from discharge data recorded at the Cuffy station for the missing periods at Moulins (source: Banque Hydro).Fig. 2 – Débits moyens journaliers de la rivière Allier à la station hydrologique de Moulins enregistrés pour la période 1968-2014 et reconstruits à partir de la station de Cuffy pour les périodes manquantes à Moulins (source : Banque Hydro).
Légende The mean frequency of independent flood events and duration of flood days per year in each period are indicated for the different critical discharge thresholds (numbers indicated at the left and right of each cell, respectively). The discharge thresholds correspond to the double of the mean annual discharge (2MQ: 270 m³.s¹), 2-yr recurrence interval flood (HQ2: 669 m³.s¹), 5-yr recurrence interval flood (HQ5: 937 m³.s-¹), 10-yr recurrence interval flood (HQ10: 1,110 m³.s¹) and 20-yr recurrence interval flood (HQ20: 1,280 m³.s-¹). Time intervals are framed between aerial images and the years are defined from August 1st to July 31st.La fréquence moyenne des crues indépendantes les unes des autres et la durée d’inondation annuelle (en nombre de jours) pour chaque période sont indiquées pour les différents seuils critiques de débits (nombres indiqués respectivement à gauche et à droite de chaque cellule). Les seuils de débits correspondent au double du débit annuel moyen (2MQ : 270 m³/s), et aux crues avec la période de retour de 2 ans (HQ2 : 669 m³/s), de 5 ans (HQ5 : 937 m³/s), de 10 ans (HQ10 : 1 110 m³/s) et de 20 ans (HQ20 : 1 280 m³/s). Les intervalles de temps sont encadrés entre les photographies aériennes. Les années couvrent la période du 1er août au 31 juillet.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-2.jpg
Fichier image/jpeg, 2,6M
Titre Fig. 3 – Spatial distribution of vegetation types, organized by succession phases, mapped in the study site at the Allier River (Moulins, France) in 2014.Fig. 3 – Répartition spatiale des types de végétation organisés par phases de succession et cartographiés dans le site d'étude sur la rivière Allier (Moulins, France) en 2014.
Légende For colour figure, see online version of this article.Pour la figure en couleur, voir la version en ligne de cet article.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-3.png
Fichier image/png, 2,9M
Titre Fig. 4 – Successional pathways scheme on dry, fresh and wet habitat conditions (zonation) based on the field mapping carried out in the study site at the Allier River (Moulins, France) in 2014.Fig. 4 – Chemins de succession dans les conditions xérophiles, mésophiles et hygrophiles (zonation) sur la base de la cartographie effectuée dans le site d'étude sur la rivière Allier (Moulins, France) en 2014.
Légende For colour figure, see online version of this article.Pour la figure en couleur, voir la version en ligne de cet article.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-4.png
Fichier image/png, 1023k
Titre Fig. 5 – Spatial distribution of the seven succession phases in the study site at the Allier River over the last five decades (1967-2014).Fig. 5 – Répartition spatiale des sept phases de succession au sein du site d'étude de l'Allier au cours des cinq dernières décennies (1967-2014).
Légende For colour figure, see online version of this article.Pour la figure en couleur, voir la version en ligne de cet article.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-5.png
Fichier image/png, 5,8M
Titre Fig. 6 – Changes in the area (right axis) / proportion of area (left axis) of each succession phase within the study area at the Allier River (1967-2014).Fig. 6 – Changements de la superficie (axe de droite) / proportion de la superficie (axe de gauche) de chaque phase de succession au sein de la zone d'étude de l'Allier (1967-2014).
Légende Grey vertical lines denote years with historical photographs. The phases WA (water) and BS (bare soil) are shown together representing the active tract. PP. Pioneer phase; HP. Herb phase; SP. Shrub phase; FP. Forest phase; AGRI. Agricultural land. For colour figure, see online version of this article.Les lignes verticales grises indiquent les années comportant des images historiques. Les phases WA (eau) et BS (sol nu) sont présentées ensemble ; elles représentent le chenal actif. PP. Phase pionnière ; HP. Phase herbacée ; SP. Phase arbustive ; FP. Phase arborée ; AGRI. Terrain agricole. Pour la figure en couleur, voir la version en ligne de cet article.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-6.png
Fichier image/png, 687k
Titre Fig. 7 – Mosaic plot of a 7 x 8 contingency table between succession phase (rows) and year-map (columns).Fig. 7 – Graphique « mosaïque » d'un tableau de contingence de 7 x 8 entre la phase de succession (lignes) et l'année-carte (colonnes).
Légende The height of the tiles is proportional to the area of each succession phase in each year-map. The border and the colour of the tiles indicate the sign and magnitude of their standardized Pearson’s chi-squared residuals, i.e. they represent if the cell’s observed frequency is higher (solid border; deep blue if significant) or smaller (dashed border; deep red if significant) than that expected from random association between succession phases and year-maps. WA. Water; BS. Bare soil; PP. Pioneer phase; HP. Herb phase; SP. Shrub phase; FP. Forest phase; AGRI. Agricultural land. For colour figure, see online version of this article.La hauteur des rectangles est proportionnelle à la surface de chaque phase de succession dans chaque année-carte. La bordure et la couleur des rectangles indiquent le signe et la magnitude de leurs résidus Khi-deux de Pearson normalisés, c'est-à-dire qu’ils indiquent si la fréquence observée de la cellule est plus élevée (bordure solide ; bleu profond si significatif) ou plus petite (bordure pointillée ; rouge foncé si significative) que celle attendue d'une association aléatoire entre les phases de succession et les années-cartes. WA. Eau ; BS. Sol nu ; PP. Phase pionnière ; HP. Phase herbacée ; SP. Phase arbustive ; FP. Phase arborée ; AGRI. Terrain agricole. Pour la figure en couleur, voir la version en ligne de cet article.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-7.png
Fichier image/png, 780k
Titre Tab. 1 – Ratios of progression / retrogression, succession / stability and proportion of area stable within the study area across time.Tab. 1 – Rapport de progression / rétrogression, succession / stabilité et proportion des surfaces stables au sein du site d’étude au cours du temps.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-8.png
Fichier image/png, 30k
Titre Fig. 8 – Hierarchical cluster analysis (average linkage method and Euclidean distance) of the standardized succession processes (rate ha.yr¹) for the seven time intervals between 1967 and 2014.Fig. 8 – Dendrogramme (classification ascendante hiérarchique ; méthode d’agglomération « average » et distance euclidienne) des processus standardisés de succession (taux ha/an) pour les sept intervalles de temps entre 1967 et 2014.
URL http://geomorphologie.revues.org/docannexe/image/11805/img-9.png
Fichier image/png, 528k
Titre Tab. 2 – Classification of trajectories of change and associated processes used in the evaluation of the chronological overlay of raster maps of succession phases.Tab. 2 – Classification des trajectoires d’évolution et processus associés utilisés pour l’évaluation chronologique des phases de succession basée sur les images rasters.
Légende The description of each process depicts the change from one phase to another phase (t0 ‑> t+1) or the absence of a phase change. Then, the mean rate values (ha.yr-¹) and statistical value of difference (Welch test) of the different trajectories of change of group 1 (1967-2005, excluding 1983-1985) and group 3 (2005-2014) of time periods are shown. Raw and adjusted p‑values in bold mean statistical significance (p‑value < 0.10).La description de chaque processus représente le passage d'une phase à une autre phase (t‑> t+1) ou l'absence d'un changement de phase. Ensuite, les taux moyen (ha/an) et résultats des tests statistiques (test de Welch) des différentes trajectoires d’évolution pour les groupe 1 (1967-2005, période 1983-1985 exclue) et groupe 3 (2005-2014) de période de temps sont montrés. Les p-valeurs brutes et ajustées en gras indiquent une significativité statistique (p-valeur < 0,10).
URL http://geomorphologie.revues.org/docannexe/image/11805/img-10.png
Fichier image/png, 163k
Titre Tab. 3 – Spearman's rho correlations of succession processes and frequency and duration of flood events per year.Tab. 3 – Corrélations rho de Spearman des processus de succession et fréquence et durée des inondations par an.
Légende Values in italics mean high correlation (rho > 0.6) and in bold and italics statistical significance (p‑value < 0.10). The discharge thresholds correspond to the double of the mean annual discharge (2MQ: 270 m³.s-¹), 2-yr recurrence interval flood (HQ2: 669 m³.s-¹), 5-yr recurrence interval flood (HQ5: 937 m³.s¹), 10-yr recurrence interval flood (HQ10: 1,110 m³.s-¹) and 20-yr recurrence interval flood (HQ20: 1,280 m³.s-¹).Les valeurs en italique indiquent une forte corrélation (rho > 0.6) et les valeurs en italique et gras une significativité statistique (p-valeur < 0.10). Les seuils de débits correspondent au double du débit annuel moyen (2MQ : 270 m³/s), et aux crues avec la période de retour de 2 ans (HQ2 : 669 m³/s), de 5 ans (HQ5 : 937 m³/s), de 10 ans (HQ10 : 1 110 m³/s) et de 20 ans (HQ20 : 1 280 m³/s).
URL http://geomorphologie.revues.org/docannexe/image/11805/img-11.png
Fichier image/png, 153k
Haut de page

Pour citer cet article

Référence papier

Virginia Garófano-Gómez, Meike Metz, Gregory Egger, María Díaz-Redondo, Borbála Hortobágyi, Gertjan Geerling, Dov Corenblit et Johannes Steiger, « Vegetation succession processes and fluvial dynamics of a mobile temperate riparian ecosystem: the lower Allier River (France) », Géomorphologie : relief, processus, environnement, vol. 23 - n° 3 | 2017, 187-202.

Référence électronique

Virginia Garófano-Gómez, Meike Metz, Gregory Egger, María Díaz-Redondo, Borbála Hortobágyi, Gertjan Geerling, Dov Corenblit et Johannes Steiger, « Vegetation succession processes and fluvial dynamics of a mobile temperate riparian ecosystem: the lower Allier River (France) », Géomorphologie : relief, processus, environnement [En ligne], vol. 23 - n° 3 | 2017, mis en ligne le 18 septembre 2017, consulté le 22 novembre 2017. URL : http://geomorphologie.revues.org/11805 ; DOI : 10.4000/geomorphologie.11805

Haut de page

Auteurs

Virginia Garófano-Gómez

Université Clermont Auvergne, CNRS, GEOLAB – 63000 Clermont-Ferrand, France | Institut d’Investigació per a la Gestió Integrada de Zones Costaneres (IGIC), Universitat Politècnica de València – Paranimf 1, 46730 Grau de Gandia, València, Spain | LTER France, LTSER Zone Atelier Loire – 37000 Tours, France (virginia.garofano_gomez@uca.fr). Tel: +33 4 73 34 68 43.

Meike Metz

Institute of Floodplain Ecology, Institute of Geography and Geoecology, Karlsruhe Institute of Technology – Josefstrasse 1, 76437 Rastatt, Germany (meike_metz@web.de).

Gregory Egger

Institute of Floodplain Ecology, Institute of Geography and Geoecology, Karlsruhe Institute of Technology – Josefstrasse 1, 76437 Rastatt, Germany (gregory.egger@kit.edu).

María Díaz-Redondo

School of Engineering. University of Lisbon – Avenida Rovisco Pais 1, 1049‑001 Lisboa, Portugal (maria.diaz@tecnico.ulisboa.pt).

Borbála Hortobágyi

Université Clermont Auvergne, CNRS, GEOLAB – 63000 Clermont-Ferrand, France | LTER France, LTSER Zone Atelier Loire – 37000 Tours, France (borbala.hortobagyi@uca.fr).

Articles du même auteur

Gertjan Geerling

Deltares – Boussinesqweg 1, 2629 HV Delft, The Netherlands | Institute for Science, Innovation and Society (ISIS), Faculty of Science, Radboud University – Heyendaalseweg 135, 6525 AJ, Nijmegen, The Netherlands (g.geerling@fnwi.ru.nl).

Dov Corenblit

Université Clermont Auvergne, CNRS, GEOLAB – 63000 Clermont-Ferrand, France | LTER France, LTSER Zone Atelier Loire – 37000 Tours, France (dov.corenblit@uca.fr).

Articles du même auteur

Johannes Steiger

Université Clermont Auvergne, CNRS, GEOLAB – 63000 Clermont-Ferrand, France | LTER France, LTSER Zone Atelier Loire – 37000 Tours, France (johannes.steiger@uca.fr).

Articles du même auteur

Haut de page

Droits d’auteur

© Groupe français de géomorphologie

Haut de page
  • Logo CNRS - Institut des sciences humaines et sociales
  • Logo Groupe français de géomorphologie
  • Revues.org