France: Restoration of Floodplain Wetlands in the Mediterranean Region

Project Stage:
Implementation

Start Date:
1995-09-01

End Date:
2008-09-01

Primary Causes of Degradation

Degradation Description

The area was embanked in the 1950s, drained, and subsequently cultivated with rice, wheat, sunflower, and hay. Two polders were built, separated by a narrow strip of reedbed that remained non-modified. The polders were split into more than 50 plots separated by irrigation and drainage channels ranging from 0.5 to 1.5 m deep. In the last years before the start of the project, land use was rice field in the western polder, and dry annual crops and pasture in the eastern polder. A riverine forest dominated by Fraxinus angustifolia and Salix alba covered some of the large dikes surrounding the polders. The elevations of the plots ranged from −0.45 to +0.15 m ASL, except the dikes and circa 2 ha where dredged sediment from the navigation canal were deposited.

Project Goals

In the Mediterranean region, restoration of wetlands in a restricted sense appears unrealistic because disturbance is very old, measured in centuries or millenniums, and hardly reversible. The main hydrological disturbances will generally not be removed (e.g. navigation canals, major rivers, embankments), and very little ecological information exists on the status before disturbance. Moreover, even small patches of natural areas are seen very much in an anthropocentric, utilitarian way for economical, social and ecological reasons. High social and economic pressures will allow the return towards less intensively managed ecosystems only when well-identified and highly demanded (desired) functions or services can be expected. Within this context, the aim of the present project was to create the conditions for the development of an ecosystem that would maintain itself with little continuous management effort. Placed on a new trajectory of natural succession (Aronson et al., 1993), the ecosystem would perform the different functions that are expected by the local users, and that may be expected from such a wetland.

Further objectives were to provide suitable feeding habitats for fish-eating herons and waterfowl during the breeding season, limited waterfowl hunting, and, after complete expansion of the reedbed, to allow extensive grazing by domestic herbivores during the summer drawdown. Thus, project objectives were not only determined by the need for restoring natural processes, but also by social demand.

Monitoring

The project does not have a monitoring plan.

Stakeholders

As soon as the area returned to public ownership, a strong social pressure developed around the project. On top of the initial objective of flood control, objectives related to hunting, thatching, potential grazing, fishing, and nature conservation were superimposed. They correspond to different hydrological regimes, which are not necessarily all compatible. The priority objective of flood control led to the decision for the level of sill; it is not, in all cases, optimal for other uses.

Description of Project Activities:
The reference ecosystem was chosen as a riverine reedbed that should be dry by the end of summer in order to be able to receive the autumn and winter floods. The internal ditches and levees were not removed and were thus supposed to contribute to the diversity of the site, providing a wider range of water depth and sources for further colonization (i.e. plants and fish). The flora and fauna were expected to colonize the area without specific intervention, and the economic use of the area by the local population (e.g. grazing and thatching) was dependent upon this colonization. Development of wetland vegetation was not expected to be limited by distance for colonization, as many species were already present as weeds on the agricultural land and along rivers and canals in, or connected to, the study site. Completely removing the dikes would have resulted in permanent flooding and the objectives would not have been met. A summer drawdown was therefore chosen to enhance the flood control function, the vegetation dynamics, and the nutrient mineralization, and to allow summer grazing. The level at which dikes were lowered was calculated in a preliminary hydraulic study using an 8-year series of daily water levels in the navigation canal and meteorological information (Grillas, unpublished report). The level of the sills was calculated as a compromise between the flooding frequency in spring and the storage of water inside the polders. A higher level would have decreased flooding frequency but increased storage of water. On the basis of the preliminary hydraulic study, two openings (150 m wide, elevation +0.4 m ASL) were made in 1995 in the surrounding dikes of each of the two polders. Water level was monitored continuously (via limnigraphs) at three locations: (1) at the connection of the river and the canal, (2) in former east- and (3) west-polders. The preliminary study of the site included soil analysis, vegetation map and history of land use. A 5-year ecological monitoring program of the site started in 1996, 1 year after opening the dikes. A detailed vegetation map was developed in 1996. In each plot, abundance of species was recorded according to a semi-quantitative scale (0, 0% cover; 1, cover below 20%; 2, cover below 40%; 3, cover below 60%; 4, cover below 80%; 5, cover below 100%; 6, cover 100%) in 1-m2 quadrats regularly spaced along three parallel transects. A second visit in late summer was made in the plots that had dried up during summer. Different censuses of target species were made: trees for their importance for birds (structure), an exotic species (Ludwigia peploides (Kunth) P.H. Raven, invasive species that constitutes an important threat for the Mediterranean wetlands), and a protected species (Leucojum aestivum L.). In ten plots, eight permanent transects 1 m wide were installed for monitoring the encroachment/regression of the fringes of reed that surrounded each former plot. Along each transect, the location of the fringe was calculated as the mean distance from a reference point to the five furthest living shoots. In two plots, large numbers of seedlings of Phragmites australis (Cav.) Steudel were found. In these plots, the density of seedlings was mapped in 1997. Viable seedbank was measured using the germination method in a greenhouse. On three former plots (ricefield, annual dry crops and pasture), 90 regularly distributed samples of sediment (4 cm diameter, 4 cm depth) were collected in spring 1996 and brought to the laboratory. The sediment of each sample was dispersed as a 1-cm layer on sand in a 15-cm diameter tray. Samples were sprayed with tap water every hour from April to November 1997. After being dried during winter in order to break possible dormancy of seeds, half of the samples per plot were submitted to the same spraying conditions described previously between March and July 1998. During the same period, the other half of the samples were flooded in tanks for hydrophytes which may not have germinated during the first trial. Seedlings were counted weekly and removed when identified. Unidentified plantlets were transplanted and cultivated until they could be identified. In the flooded treatment, only the hydrophytes were taken into account; other species could usually not grow through the water column and be identified. Water birds using the site were counted fortnightly during the migrating and wintering period (late summer to spring). During the reproductive season, the number of breeding pairs of colonial water birds were counted each year. The density of feeding herons was counted in 1997 in 17 visits between 30 April and 8 July along a fixed itinerary (line transect) representing a constant sample of both sectors. To establish a preliminary list of species, and determine the relative proportions per size class as an indication of food availability for fish eating herons, fish populations were evaluated during one field campaign in September 1997, combining electric fishing and passive fishinggears. At this date, most of the plots had dried, and water and fish concentrated in the ditches and canals. In large canals, nets with various mesh size, ranging from 10 to 60 mm, and traps (1.1 m long, 0.4 m diameter, mesh size 12 mm) were used during one night. Three nets and three traps were used in a large canal inside the site and in a branch of the neighboring navigation canal. Electric fishing was performed in five sites with an Albatros apparatus (Dream Electronic) for brackish water. Electric fishing was applied during 10 - 16 min for exhaustive capture on surface areas ranging between 65 and 150 m2 (intensity 4 - 10 A, Power 0.5 - 1.6 KW, according to electric conductivity of water 5 - 13 mS/cm). All large fishes, and a sub-sample of smaller ones, were counted and their length measured. Total biomass was evaluated per species.

Ecological Outcomes Achieved

Eliminate existing threats to the ecosystem:
A 5-year monitoring program was launched in order to determine which processes were most important within such a project, and to measure ecological change at the site. Results presented here were obtained after the first 3 years. The opening of the dikes and sill creation were followed by two winters with very high rainfall and floods. Floods expanded in the site, which served as a flood retention area. The water level in the sites was identical to that in the river when higher than the basal elevation of the openings (e.g. January - February 1996). When water level in the river dropped below this elevation, the wetland was disconnected from the river. The water level decreased progressively by evapo-transpiration and outflow through a valved drainage outlet. Decrease in water level in the former polders was almost linear as a function of the time elapsed since the last filling event, and the frequency of those events is hence a key parameter for the hydrological functioning. A total of 160 plant species were found in the plots and surrounding ditches and levees in 1996, the second year after opening the dikes. Six species were found in more than 40 plots out of a total of 51: Potamogeton pectinatus L., P. australis, Ceratophyllum demersum L., Paspalum paspalodes (Michaux) Scribn, Scirpus maritimus L. and Zannichellia pedunculata Reichenb. Except for the exotic species P. paspalodes, the most frequent species are submerged macrophytes and tall helophytes which can cope with the high water levels that followed the opening of the dikes. P. paspalodes is a C4 late growing species, requiring high temperature (Pearcy et al., 1981; Mesléard et al., 1993), and therefore, it grows late in the season when water levels are low. The species richness per plot exhibited a significant positive correlation with its elevation. A total number of 31,200 seeds of 38 species germinated during the first phase and 6800 seeds of 16 species during the second phase. Only five new species were found during the second phase, three terrestrial species and two hydrophytes. Among these species, only Lotus corniculatus L. produced a significant number of seedlings (160). The spatial distribution of species was very heterogeneous and 70 samples of sediment were necessary to get 90% of the species per plot. However, all dominant species were found within the first 30 samples. The seedbank differed significantly between the three studied plots. In the former grassland, the seedbank was characterized by the largest mean number of species per sample, including more terrestrial species than in other plots. However, the total number of seedlings per sample was much lower than in the other plots. Conversely, the lowest mean number of species and number of terrestrial seedlings per sample characterized the former ricefield. The most abundant species that germinated from the seed bank in former grassland were Chara sp., Cyperus fuscus L., Aster squamatus (Sprengel) Hieron., L. corniculatus, Typha angustifolia L. and S. cf alba L.. In the dry crops, C. fuscus, A. squamatus and Polypogon monspeliensis L. (Desf.) dominated the seed bank, while in the former ricefield, C. fuscus contributed to more than 90% of the total number of germinations. For all species except Atriplex prostrate DC. and S. alba, the differences between plots were significant. The most striking feature of the dynamics during the studied years was the gradual shift from terrestrial/amphibious vegetation to amphibious/aquatic communities. The grassland was dominated in 1994 by terrestrial grasses such as: Festuca arundinacea Schreber, Poa pratensis L., and leguminosae (Trifolium, Vicia, and Lotus). Two years after the project started, these species had disappeared and submerged macrophytes dominated (Chara sp., Z. pedunculata). Agriculture had been abandoned 1 year earlier in former dry crops than in other plots, and in 1995, wetland vegetation was already abundant with A. squamatus, A. prostrata, J. gerardii Loisel, P. australis, P. paspalodes, P. monspeliensis, S. maritimus, Spergularia rubra (L.) J. and C. Presl. The shortest and least flood tolerant of these species (e.g. Spergularia, Aster, Atriplex, and Polypogon) disappeared, while the tallest remained abundant (Phragmites, S. maritimus, T. angustifolia) and submerged macrophytes appeared (Z. pedunculata, Chara spp.). In former ricefields, many wetland plants--such as: S. maritimus, P. australis, T. angustifolia, and Echinochloa crus-galli (L.) P. Beauv.--were present as weeds. Some of them maintained significant populations (P. australis, P. paspalodes, S. maritimus), although submerged macrophytes dominated in 1996 (P. pectinatus, C. demersum). In 1997, the plots were all permanently flooded, and no terrestrial plant was found. Although the specific composition of the seedbank was different from the actual vegetation in 1996, many terrestrial and amphibious species were still present as seeds. An additional soil sampling realized in 1998 will show whether those seeds are still viable after 3 years in flooded soils. Vegetative growth from the edges was expected to be the dominant process that would contribute to rapid colonization of P. australis. Aerial photographs and field surveys showed that reedbeds did not expand during the first years of the project. From the ten plots where the location of the fringe of the reedbed was measured, eight showed a regression of the reedbeds ranging between 0.7 and 3.4 m. In only one plot did the reedbed colonize 0.5 m, and on another there was no change. The depth and duration of flooding are probably responsible for this regression. There was no significant correlation between changes and elevation in the first year, but a significant trend to colonization on highest plots was observed over the 3 year period 1996 - 1998. Other factors, such as grazing by coypu and sediment characteristics, may have interfered with hydrological factors. In contrast, P. australis rapidly colonized the deposits of sediment where elevation is about 1 m higher than in the other plots and was very rarely flooded. Although seeds were found in the seed bank of the three sampled plots, abundant seedlings of P. australis were found in only two plots of rather high elevation (−2 and −3 cm ASL). In these two plots, density was high (up to 5 per m2), and most resulted from germination in 1995 and 1996. Those seedlings were deeply flooded in 1996 and 1997, and their growth was reduced. In better conditions, such colonization may lead to the very rapid development of a reedbed. Birds used the site during summer as a feeding and/or breeding site depending on the species. A colony of tree nesting herons established right after the start of the project and is currently increasing. In 1996, 17 couples of night herons (Nycticorax nycticorax) nested, and the following year, the colony had 454 nests, among which were 222 Bubulcus ibis, 190 N. nycticorax, 25 Egretta garzetta and 17 Ardeola ralloides. Nesting success ranged from 2.3 (S.E. 0.45, n=18) for E. garzetta and 3.4 (S.E. 0.5, n=12) for A. ralloides. Fish eating birds were mainly herons plus ten couples of grebes (eight Tachybaptus ruficollis and two Podiceps cristatus). Bird counts on the fixed itinerary in 1997 showed that Ardea purpurea started using the area in May and had a stable number of individuals (15 to 20) till mid July. They used mainly the western half with about 2/3 of the individuals foraging in the plots and 1/3 foraging in the surrounding channels. A. cinerea was the most abundant, with a peak of 45 birds at the end of June, and the E. garzetta numbers were variable, ranging from ten to 30 individuals. Complete counts performed in September 1997 showed that total numbers were much higher, with 750 E. garzetta and more than 160 A. cinerea on the same day. At that date, A. purpurea had already left the site. The establishment of a colony of tree nesting herons was difficult to predict, although the expected structure of the site was recognized as suitable and therefore, had not been identified as an objective of the project but rather considered as a possibility. The occurrence of breeding colonies of water birds is also strongly influenced by external factors such as the distribution of colonies in the region, the distribution of feeding habitats, the reproductive success and possible disturbances (Fasola and Alieri, 1992; Hafner and Fasola, 1992). The colony of tree nesting herons is large and it has now become a site of regional importance. This was probably promoted by the combination of several factors such as the low disturbance by visitors during spring, the suitable structure of the breeding habitat (and particularly the protection from the wind), and the abundance of fish at short distance. A general increasing trend for these fish populations further favored the establishment of the colony (Hafner and Fasola, 1997). The colony may persist even if the surrounding marshes dry up during the breeding season since two deep canals protect the site from predators. Similarly, after colonization by reed, the establishment of a colony of purple heron is possible on the site but would depend on similar external factors. A total number of 16 species of fish were captured in the internal channels, and three additional species were found only in the navigation canal connected to the marsh during floods (Silurus glanis, Stizostedion lucioperca, Micropterus salmoides). The fish community was typical of littoral ponds and lakes or rivers with a slow current. The connection to the sea is highlighted by species that breed at sea (Anguilla anguilla, Mugil cephalus and Liza ramada). A remarkable feature is the abundance of exotic species, contributing to more than 90% to total biomass and 60% of species. Furthermore, two exotic crayfish species were also captured (Procambarus clarkii and Orconectes limosus). All the exotic species are considered naturalized, that is populations are maintained without permanent input of new individuals. The proportion of exotic species is much higher than usually found in other sites in the same region (Rosecchi et al., 1997). Due to late floods between the reproductive season and fish sampling, it cannot be demonstrated that some of these species reproduced on the site. However, it is very likely for some of them, according to the characteristics of the site and the species requirements (Gambusia, Cyprinus.). The size distribution of fish corresponds to the objectives of the project in providing suitable food for herons. Large numbers of fish are small enough to be caught by E. garzetta, and the range of size covers the requirements of the most abundant bird species.

Factors limiting recovery of the ecosystem:
The summer drawdown expected from the preliminary study and surveys was not obtained during the first years. The date of complete drying out calculated from the preliminary study was very sensitive to the date of the last flood, as drying out was a gradual process where evapo-transpiration dominated. Hence, the land was not dry in autumn, when the probability for flooding was the largest. However, at the end of September 1996, the water levels were low and the area still had 80% of its storage capacity. In 1997, water levels did not drop below 0 cm ASL, and storage capacity was only half of the maximum in late September. In 1996, there was enough time between the last spring and the first autumn filling events (about 3 months) to reach an "˜almost dry' state. The decrease in water level after flood was enhanced by the installation of the valved drainage outlet. The exceptionally high rainfall in 1996 (more than 1000 mm compared to a 620-mm mean) was not the only reason for the observed results. Early-autumn floods contributed to reduce the duration of the dry period, and unauthorized manipulation of sluices resulted in summer flooding in both 1996 and 1997. Erosion promoted by an exotic rodent, coypu (M. coypus), lowered the sill level and caused water inflow to start when river level was 10 to 20 cm lower than the 0.4 m planned. The sluices were subsequently blocked, and a drier year in 1997/98 promoted a total dry out of the area like that expected from initial hypotheses. The high water levels had multiple consequences on biota, and especially on vegetation dynamics. A slower than expected establishment of the amphibious vegetation was observed, along with a rapid dominance of submerged vegetation. To a large extent, the vegetation that developed in each plot was independent from the seed bank and the species previously present in the former agricultural land. The correlation between the composition of the seed bank and the vegetation is often weak in wetlands (Van der Valk, 1981; Wilson et al., 1993). Moreover, in the context of restoration projects, this correlation is expected to be even weaker as many species in the seed bank are agricultural weeds rather than wetland species. Most of the species that were found in the seedbank were terrestrial or short-lived amphibious species unable to withstand the high level and long duration of flooding. Among the species that were found in the site before rehabilitation, only a few were still present in 1996. Some species, such as the hydrophytes C. demersum or P. pectinatus, were abundant in 1996 but were absent from both the 1994 inventory and from the seed bank. They were, however, present in 1994 in the permanent internal canals. Those canals formed a network of wetland vegetation patches among the former polders, which facilitated their colonization. Very few species that had not been found before in the site were observed during the 1996 inventory (e.g. Najas marina L.). They probably have colonized during floods. As hypothesized initially, the abiotic conditions, mainly flooding stress, determined the dynamics of the vegetation. This result highlights the impact of the abiotic factors on the species composition and dynamics of vegetation, and their predictive value for restoration of wetlands and planning (Weisner, 1991; Martinez-Taberner et al., 1992; Palmer et al., 1997). Water depth and duration of flooding are often identified as dominant factors controlling the distribution of species in wetlands (e.g. Spence, 1982). Increase in water level can result in changes in the species composition of vegetation (Wallsten and Forsgren, 1989; Van der Valk et al., 1994; Crivelli et al., 1995). Terrestrial species are the most sensitive, followed by emergent species. Submerged species have a higher tolerance for an increase in water level, and they are mainly limited by light attenuation and wave action. In this case, the impact of increased water depth was probably enhanced by the high turbidity which further reduces light availability. The dominance of hydrophytes after 2 years resulted partly from the low cover of helophytes (Grillas, 1992). Reed was affected more than what was expected. The water depth in the site was within the range P. australis can stand in continental freshwater marshes (Haslam, 1972; Squires and Van der Valk, 1992) or in Mediterranean brackish coastal wetlands (Mauchamp, unpublished data). Possible causes for the regression of Phragmites were the anoxia caused by permanent flooding and eutrophication (Weisner and Graneli, 1989; Armstrong et al., 1996). The Vistre River feeding the wetland is heavily polluted by urban wastewater and agricultural runoff. Moreover, slightly brackish water occasionally enters the marsh. This may contribute to an increased richness in sulphate and the production of deleterious sulphide in anoxic conditions (Armstrong et al., 1996). The survival of the rhizomes of P. australis during floods, and their further growth, are favored by the presence of unflooded stems which conduct air to the rhizome and remove metabolic gases (Weisner, 1988; Brix, 1990). Winter flooding and wave action may have damaged old stems and prevented oxygen from reaching the rhizomes (Coops et al., 1994). During the last year, lower water level resulted in new vegetative colonization of P. australis, although it was reduced by higher salinity (Hellings and Gallagher 1992). P. australis should dominate in the future, as hypothesized in the preliminary study, although the process will be much slower than expected.

Socio-Economic & Community Outcomes Achieved

Key Lessons Learned

This project was an opportunity to test the ability to rebuild a wetland ecosystem and make predictions on its dynamics.

About 3 years after opening the dykes, the objectives of the project were only partly met. A number of technical and social difficulties, related to the multi-users character of the site, resulted in higher water level than expected in the initial plans. The level of the sill in the former dyke should be maintained at 0.4 m or slightly higher in order to achieve objectives. These problems seem to be nearly solved and the hydrological situation to be approaching the expected passive functioning, depending on climate and fluctuations of sea level. Although the objective of restoring the flood storage function is widely accepted among the local population, there remain some difficulties to accept the natural fluctuations in water level resulting from the rehabilitation project.

All biological compartments and human-use of the area have been impacted by the hydrological situation. However, the initial hypotheses linking the abiotic situation, the vegetation dynamic and fauna appeared to be verified to a large extent, and project results support the hypotheses that in wetlands, predictions on vegetation structure can be established on the basis of abiotic conditions, assuming that competitive output is correlated with plant maximum height. The establishment of the expected hydrological situation should result in colonization by P. australis. Special attention must be given to this species, as it controls several functions and services that are expected from the wetland. Low water quality might result in reed colonization being limited to shallower water than expected, thus promoting the maintenance of open water in deeper parts of the wetland.

Long-Term Management

The area does not have a sufficient enough size to solve the flood problem at a regional scale, but this is intended to be a pilot project encouraging and facilitating other projects. The flood control function was supposed to be obtained with a minimum permanent cost, and hence a minimum management effort on the long term.

Sources and Amounts of Funding

The project was supported by the Agence de l’Eau RMC, the French Ministry of the Environment (through the program “˜Recréer la Nature’), the Département du Gard, the Fond de Gestion de l’Espace Rural and the Fondation Tour du Valat.

Other Resources

André Mauchamp
mauchamp@tourduvalat.org

MAUCHAMP A., CHAUVELON P., GRILLAS P., 2002. Restoration of floodplain wetlands: opening polders along a coastal river in Mediterranean France, Vistre marshes. Ecol Eng 18 p619-632.

Primary Contact

Organizational Contact