The Mont Saint-Michel bay (NW France) is an extensive littoral zone (500 km²) located between the regions Brittany and Normandy (48°40’N, 1°40’W). Two sites have been studied in salt marshes: “la Ferme Foucault”, on the western part of the Mont St.-Michel (coded F; 48°37’N, 1°32’W) and “la Rive” on the eastern part of the Mont St.-Michel (coded R; 48°37’N, 1°29’W) (Fig. 1).

For the diachronic approach, ground beetle populations were compared at seven stations (A to G) located along the same land-sea transect at the “Ferme Foucault” site between 1983–1984 and 2002. During the study of 1983–1984, Elymus athericus was restricted in this salt marsh to the dyke (station A) and to the upper marsh (station B), but absent from stations C-G. Invasion by Elymus athericus modified the plant cover of the sampling stations between 1984 and 2002. The middle marsh and lower marsh stations (station C till F), dominated in 1984 by Atriplex portulacoides (Chenopodiaceae), were dominated by Elymus athericus in 2002.

Secondly, natural (dominated by Atriplex portulacoides), and invaded (dominated by Elymus athericus) stations were studied at different marsh levels in the synchronic approach. Comparisons of paired stations (natural and invaded – coded N and I, respectively) were spatially replicated three times for avoiding pseudo-replication (Hulbert 1984). Paired stations were located at the same distance from the dyke because of the existence of a salinity gradient influencing both species richness and abundance (Pétillon et al. 2004): stations 1 (350m), stations 2 (800–900m; both couples of stations at the “Ferme Foucault” site) and stations 3 (1000 meters from the dyke; “La Rive” site). Because of the clonal progression of the invasive species, all Elymus populations (stations I1, I2 and I3) formed a uniform and continuous plant cover. The natural areas sampled were either patch-like formations (in case of stations N1 and N2) or strip-like formations (station N3). Mean salinities did not significantly differ between invaded and natural stations at each salt marsh level (Pétillon et al. 2005) and elevations were similar between compared stations (J.C. Castel and J. Huet, 1999, unpublished data). More details on the sampling stations can be found in Fouillet (1986) and Pétillon (2005).

For both the synchronic and diachronic approaches, ground beetles were sampled with pitfall traps, consisting of polypropylene cups (10 cm diameter, 17 cm deep) with ethylene-glycol as preservative. Traps were covered with a raised wooden roof to keep out rain. They were emptied weekly when tides permitted (i.e., about three weeks per month). Pitfall traps were grouped by four and spaced 10 m apart, this being considered to be the minimum distance for avoiding interference between traps (Topping and Sunderland 1992). Before the Elymus invasion, Fouillet (1986) sampled the transect with one trap per station from May to September in 1983 and 1984, for a total of 16 five-day samples. In 2002, four traps were installed at each station in both study sites, from April to November 2002. Sampling time was comparable between both periods (90 days in 1983–1984 and 96 days in 2002). Because of the differences in sampling efforts, we only compared the two studies on the basis of species presence / absence (i.e., distribution range along the land-sea transect).

Ground beetles were preserved in 70% ethanol and identified using Jeannel (1942) and Trautner and Geigenmüller (1987). Nomenclature follows Lindroth (1992) as far as possible, and Fauna Europaea otherwise (http://www.faunaeur.org/).

Statistics on the abundances of halophilic species were performed only for species represented by at least 10 individuals in couples of stations. Catches in pitfall traps were related to trapping duration and pitfall trap perimeter, which calculates an “activity trappability density” (number of individuals per day and per meter – Sunderland et al. 1995). Mean species richness and mean abundances were compared using a two-way mixed model (habitat × station) with habitat type as fixed factor, station (1, 2 and 3) and interaction habitat*station as random factors. In case of non-significant interaction between habitat type and station, the interaction was removed from the model and a new model was performed for detecting significant effects of habitat type and/or station. In case of significant interaction between habitat type and station, parameters were analysed station by station (one-way ANOVA). Statistical analyses were performed using the Statistica-7 software.

A total of 24 species (represented by 7,774 individuals) and 35 species (represented by 8,588 individuals) were caught in 1983–1984 and in 2002, respectively. Five species were exclusive to the first sampling period and 16 to the second one. All the species that were only recorded in 1983–1984 were caught in very low numbers (max. 2 individuals), four species on the dyke (Clivina colaris, Dromius linearis, Harpalus rufibarbis and Harpalus rufipes) and only one in the salt marsh (Dyschirius chalceus). As the sampling effort was quite different between 1983–1984 and 2002 (see Material and Methods), it cannot be concluded that the ‘appearance’ of species between the two studies can be related to the invasion by Elymus athericus. The comparison in distribution was thus restricted to the 19 shared species (Table 1).

In terms of distribution ranges, two groups of carabids were distinguished: species with constant distribution range in the salt marsh or on the dyke and species with an increased distribution range between 1983–1984 and 2002. Eight species were caught on the dyke in 1983–1984 and in 2002, and seemed not to have progressed with Elymus athericus in the salt marsh (Amara equestris, Anisodactylus binotatus, Bembidion tetracolum, Harpalus anxius, Leistus fulvibarbis, Nebria brevicollis, Pterostichus melanarius and Pterostichus niger: Table 1). Eight other species had a similar habitat range in the salt marsh, extending from the upper to lower marsh or from the dyke to the lower marsh (halophilic species: bold in Table 1), plus two high-marsh living species (Badister bipustulatus and Pterostichus vernalis), one low-marsh living species (Dyschirius salinus) and one species with a discontinuous range along the land-sea transect (Loricera pilicornis).Only three speciesshowed an extension of their distribution in the salt marsh, both to the upper and lower marsh (Bembidion iricolor, Bembidion lampros and Pterostichus cupreus).

The Elymus athericus invasion led to a decrease in the percentage of halophilic species in invaded salt marshes (Fig. 2).

A total of 505 individuals belonging to 17 species were sampled in the three pairs of natural and invaded stations. The synchronous comparison of natural and invaded habitats revealed the existence of eight shared species. Two species were exclusive to natural habitats (Pogonus littoralis and Pogonus luridipennis) and six to invaded habitats (Anisodactylus poeciloides, Bembidion obtusum, Harpalus anxius, Harpalus distinguendus, Pterostichus cupreus and Pterostichus versicolor). Total species richness was higher in invaded habitats than in the natural ones (Table 2). Significant interactions between habitat type and station were found for species richness and two species Pogonus chalceus and Dicheirotrichus gustavii. Mean species richness was significantly higher in an invaded station compared to its adjacent natural one (one-way Anova, F-ratio=22.04, p=0.003, d.f.=7). More Pterostichus chalceus were caught at a natural station than at the paired invaded one (one-way Anova, F-ratio=14.68, p=0.009, d.f.=7). Dicheirotrichus gustavii was significantly higher in an invaded station compared to the natural one (one-way Anova, F-ratio=6.89, p=0.039, d.f.=7) and the opposite pattern was found in another couple of stations (one-way Anova, F-ratio=11.94, p=0.014, d.f.=7). Bembidion minimum was significantly higher in invaded habitats compared to natural ones (Factorial Anova, F-ratio=5.91, p=0.025, d.f.=20). No difference between habitat types was found for Dicheirotrichus obsoletus and Bembidion normanum(Table 2).