Fig 1.
The two-step hierarchical model of community assembly and the effects of the ecological filters on functional trait distribution [14, 15].
Fig 2.
Location of the 13 fish sampling sites in the Gironde estuary.
FDM, BDM1, BDM2 and BDM3 are dyked sampling sites. IM1, IM2, IM3, IM4, IM5 and ICH are natural sampling sites. TRM1 is the sampling site of the 12-year-old tidally restored Mortagne marsh. TRM2a and TRM2b are located in the 1-year-old restored part of the Nouvelle island.
Table 1.
The single-trait indices of functional structure.
Fig 3.
The morphological measurements on the fish ecophases.
Table 2.
The 10 selected functional traits and their ecological interpretation.
Fig 4.
Two-dimensional non-metric multidimensional scaling (2D-NMDS) of the fish community-weighted means (CWMs) among five groups of habitats clustered with respect to their naturalness.
The points on the plot correspond to the communities at each site × sampling time. Abbreviations for habitat groups: FDYK, Freshwater dyked marshes (one site); BDYK, Brackish dyked marshes (three sites); TRYO, Young tidally restored marsh (two sites); TRMA, Middle age tidally restored marsh (one site); NINT, Natural intertidal habitats (six sites).
Table 3.
Monotonic response of the functional structure of fish communities to increasing intertidal naturalness after factoring out salinity.
Fig 5.
Principal component analysis (PCA) biplot of the response of functional structure indices (other than functional identity) to increasing naturalness of intertidal habitats.
The response of functional structure indices was measured using Spearman semipartial correlation (ρsp). The top and right axes (grey color) show the loadings of the variables. Ellipses correspond to the groups of functional traits identified from the clustering analysis (Fig 6). See Table 2 for the abbreviations of functional traits. Codes of functional indices: Tr, functional richness; FRO, functional evenness; FDis, functional dispersion; wMNND, functional originality; wFSpe, functional specialization.
Fig 6.
Clustering of the functional traits according to the response (Spearman semipartial correlation, ρsp) of their functional structure indices (richness, evenness, dispersion, originality and specialization) to increasing naturalness of intertidal habitats.
The names of the functional traits were in Table 2.
Table 4.
Effect sizes of functional richness (TrES) and dispersion (FDisES) along the gradient of intertidal habitat naturalness.
Fig 7.
Patterns of fish community assembly and monotonic trends in community-weighted means (CWMs) along the naturalness gradient of intertidal habitats.
The X-axis represents the gradient of intertidal habitat naturalness (DYK, dyked marshes; TRM, tidally restored marshes; NINT, natural intertidal habitats) and the Y-axis represents the CWM. The names of the functional traits were in Table 2. Trends related to the CWMs were corrected for salinity using Spearman semipartial correlation. Ascending and descending straight lines represent, respectively, significant positive and negative correlation between the CWMs and intertidal naturalness. Horizontal straight lines indicate lack of significant correlation. Grey areas correspond to parts of the gradient where environmental filtering (i.e., lower-than-expected trait range) was detected under the null model 2. Striped areas indicate higher-than-expected trait range. Double arrows indicate either trait convergence (convergent arrows) or divergence (divergent arrows). Grey arrows indicate both lower significavity (0.01 < p ≤ 0.05) and lower effect sizes (−0.5 < medianES < 0.5). Adapted from [14].