Metacommunity structure of aquatic gastropods in a river floodplain: the role of niche breadth and drift propensity

Introduction

One of the important advances in ecology in recent decades has been the recognition that local communities are products of processes at both local (niche assembly) and regional (dispersal assembly) scales (Ricklefs, 1987). The relative importance and interactions of these processes are addressed by metacommunity ecology (Tokeshi, 1990; Leibold & Miller, 2004; Leibold et al., 2004; Driscoll, 2008). Within this framework, the structure of natural communities is explained through ‘niche-based’ mechanisms, assuming that ecological traits differ among species, and ‘neutral’ theories based on the premise that species are ecologically equivalent, emphasising the importance of dispersal (Hubbell, 2001; McGill, 2003; Chave, 2004; Leibold et al., 2004). Leibold et al. (2004) synthesised empirical and theoretical knowledge and formulated four paradigms that predict patterns in natural communities and that may explain the generation and maintenance of diversity. The models are known as neutral, patch dynamics, species sorting and mass effect. The last three consider differences among species, and the last two additionally consider environmental differences. Logue et al. (2011) and Winegardner et al. (2012) suggest that metacommunity ecology should move on from seperating four discrete paradigms to testing the underlying mechanisms and relative importance of spatial and local processes.

Recent studies have shown that the predominant species traits of the community studied, such as mechanism or speed of dispersal and niche breadth, predicted part of the detected variation in the structure of metacommunities (Van de Meutter, De Meester & Stoks, 2007; Pandit, Kolasa & Cottenie, 2009; Brown & Swan, 2010; Altermatt et al., 2011). To gain a deeper insight into these processes, the characteristics of the individual species should thus be included in analyses; in other words, the metacommunity approach should be supplemented by an analysis of individual species and their traits (Driscoll & Lindenmayer, 2009).

This study focuses on river-floodplain systems, which are especially challenging from a metacommunity point of view. Beside the spatial heterogeneity and the high and often stochastic temporal variability of environmental conditions, they are characterised by the network configuration of their waterbodies and dispersal directionality. These challenges can also be seen as opportunities to expand the field of metacommunity ecology and establish a broader framework that contributes to floodplain ecology (cf. Logue et al., 2011). Recent studies have shown that spatially constrained network connectivity is a key factor for community composition and population persistence (Carrara et al., 2012) as well as biodiversity (Muneepeerakul et al., 2008) and diversity in local communities is strongly influenced by directionally biased dispersal (Altermatt, Schreiber & Holyoak, 2011).

The role of spatiotemporal heterogeneity in determining aquatic communities of rivers and their floodplains has been a central theme of research in recent decades. Most studies have emphasised the significance of the physical conditions of habitats and niche differences of species, while the spatial configuration of habitat patches and dispersal mechanisms of species were largely ignored (see Winemiller, Flecker & Hoeinghaus, 2010). We believe that dispersal must not be disregarded in river-floodplain systems although given the high spatial heterogeneity it may be that niche differentiation is the most important process in generating and maintain diversity. Furthermore, river floodplains are often highly stochastic systems. Stochasticity and the related unpredictability of abiotic conditions may impede adaptations to specific environmental conditions and may thus favour habitat generalists (i.e. species that are neutral with respect to habitat). Such communities can be expected to be structured by regional spatial processes (Pandit et al., 2009).

One characteristic of river-floodplain systems is the network configuration of the waterbodies, which are linked by unidirectional flow from upstream to downstream. The main mode of colonisation under such conditions is dispersal by drift. We therefore assume that the configuration of the network upstream of a site is more influential than its arrangement downstream. We further argue that drift prevention and associated traits are essential because they allow the maintenance of position in a flowing environment. This is crucial for specialist species to decrease the chance of dislocation into unfavourable habitats but less important for generalist species that can deal with a broader range of conditions. Niche breadth and drift propensity should thus be linked. Evidence for a positive relationship between niche breadth and rates of dispersal was found by McCauley (2007) and Bonte et al. (2003) for odonates and spiders, respectively. We expect this relationship to be general and to have a significant effect on the relative importance of local compared to regional processes for metacommunity structure.

To test such ideas, Dumbrell et al. (2010) suggested that an ideal study community should be functionally important, species-rich and show a priori evidence of both niche and dispersal assembly. Aquatic gastropods fit these criteria. They enhance microbial growth and nutrient cycling by their mixing of surface sediments and breakdown of organic detritus (Covich, Palmer & Crowl, 1999). Floodplains host a very rich fauna of aquatic gastropods, and physical conditions have a major bearing on distribution and richness (Reckendorfer et al., 2006). Drift and drift prevention are relevant for juvenile and adult gastropods (Dillon, 2000; Mouthon & Daufresne, 2006; Kappes & Haase, 2012). Since the mobility of gastropods is low, passive drift mechanisms outweigh active movement between patches.

The analysis is based on an assessment of the gastropod fauna of the whole river-floodplain network. We use pooled species data to define an ‘average’ metacommunity. The data set consists of 47 waterbodies differing in their spatial (i.e. spatial connectivity within the network) and environmental (i.e. local physical habitat) characteristics. To assess the relative importance of local (e.g. competition, predation) and regional (i.e. dispersal) processes, we performed a variance partitioning procedure according to Borcard, Legendre & Drapeau (1992). To gain further insight into the significance of species traits (e.g. dispersal ability, niche breadth), their interdependencies and potential trade-offs for community assembly, we also investigated the distribution of individual species.

Methods

Study area: the Donau-Auen National Park in Austria

The Danube River, the second largest river in Europe, is 2900 km long and drains an area of 817 000 km². At Vienna, the Danube is a ninth-order river with a mean annual discharge of about 1950 m³ s⁻¹ and a bankfull discharge (recurrence time of c. 1 year) of 5800 m³ s⁻¹. The average slope in this section is 0.45‰. In 1996, the floodplain east of Vienna was given the status of a national park [Schiemer, Baumgartner & Tockner, 1999; Donau-Auen National Park (DANP), Fig. 1].

Historically braided, the floodplain has been constrained by major regulations that began in 1875. Due to ongoing measures of restoration (Tockner, Schiemer & Ward, 1998; Schiemer et al., 1999; Reckendorfer et al., 2005), the DANP now contains sections of differing connectivity, ranging from predominantly lentic waterbodies to areas of intermediate and high connectivity (Baranyi et al., 2002; Reckendorfer et al., 2006).

Spatial structure

We used network analysis (e.g. Minor & Urban, 2007; Urban et al., 2009; Cumming et al., 2010; Economo & Keitt, 2010) to build a model of our study site. Within this framework, the landscape is represented as a set of nodes (points representing habitat patches or sampling sites) connected by edges between them. An edge between two nodes implies that there is some flux between those nodes, as in the case of dispersal between two patches (Minor & Urban, 2007). We used a network centrality measure to quantify the position of a waterbody in the floodplain (i.e. its connectivity/isolation). Centrality measures allow analysis of connectivity and isolation at different spatial scales (Economo & Keitt, 2010). For the calculation, we divided the network into sections of 500 m, starting at the midpoint of the respective waterbody. We included all waterbodies up to a distance of 5000 m. The connectivity variables (or centrality measure) (C) for each waterbody in the respective channel system were then calculated as

where s is the respective distance step, h is the step length of 500 m and L_s is the sum of the channel lengths in the distance step.

We calculated several variables of habitat connectivity (see Fig. 2 for a graphical representation):

1. All waterbodies were treated as equal, that is, we neglected direction of flow (C_full).
2. Downstream connectivity (C_down) is important and emphasises active upstream movement of organisms. C_down was calculated including only the network downstream of the section.
3. Only upstream connectivity (C_up) is important, placing emphasis on mechanisms of drift:
   1. All upstream waterbodies have equal influence. C_up1 was calculated including only the network upstream of the section.
   2. Only waterbodies showing equal or higher connectivity than the waterbody of interest (i.e. at least equally lotic) have a relevant impact. C_up2 was calculated including only the waterbodies upstream of the section that were at least equally lotic.
   3. To account for the frequency of exchange processes, we calculated C_up3 and C_up4 by weighting network lengths with hydrological connectivity as length (m)*(365-hco)/365; the variables of connectivity where then calculated analogous to C_up1 and C_up2.

Species traits

We selected traits that were related to drift propensity/colonisation (Falkner et al., 2001): (i) the number of eggs or live young (egg) and (ii) ovovivipary or pasting capsulated eggs on conspecifics (brood) (see Table S1). In passive dispersers, the production of large numbers of offspring improves the ability to colonise (Kneitel & Chase, 2004). Parental care (ovovivipary or pasting eggs on conspecifics) associated with a low number of offspring is a good surrogate for competitive ability. High numbers of small eggs relate to a high colonisation ability/drift propensity.

As an estimate of habitat specialisation, an index for niche breadth was calculated (niche) using the table of macrohabitat preferences in the database for European species (Falkner et al., 2001). We chose this method, instead of the direct calculation using our own samples, to get a universal measure of niche breadth and prevent circularity of inferences. The database contains fuzzy-coded preference data for standing and running fresh water, including rivers (vegetated and non-vegetated), brooks, springs, lakes (vegetated and non-vegetated), permanent pools and temporary pools. Preferences range from 0 (no association to a macrohabitat) to 3 (strong association to a macrohabitat), and a species can fall into several categories. The disadvantage of this coding technique is that arithmetic operations cannot be performed. To allow the calculations, the data from the database (Falkner et al., 2001) were transformed into a frequency distribution according to Tachet et al. (2003). Species specialisation was then calculated adopting Levins’ approach (Levins, 1968; see also Pandit et al., 2009) as

where MB_j is the niche breadth of the respective species and A_ij is the proportion of the of species j in the respective habitat state i. As an additional variable for intraspecific diversity, a parameter for polytypism (poly), defined as geographical variability in one species, was extracted from the database (Falkner et al., 2001).

Results

Relationships between species traits and local and regional factors

For single species, the explanatory value of the spatial variables in explaining patterns of distribution increased significantly with niche breadth (Fig. 3a; Pearson's correlation: R = 0.71, P < 0.05) and degree of polytypism (Fig. 3c; Spearman's correlation: R = 0.75, P < 0.05). An inverse relationship was found between the explanatory value of environmental variables and niche breadth (Fig. 3b, Pearson's correlation: R = −0.63, P = 0.05). The relationship for polytypism was not significant (Fig. 3d, Spearman's correlation: R = −0.35, P = 0.33). The explanatory value of the spatial variables was positively correlated with drift propensity, as indexed by egg number (Fig. 4a; Spearman's correlation: R = 0.67, P < 0.05) and presence or absence of brood care (Fig. 4b; Spearman's correlation: R = −0.61, P < 0.1, Mann–Whitney: U = 3, P < 0.1). The distribution of species showing brood care and low fecundity was less determined by regional spatial mechanisms than was the distribution of prolific species without brood care. The importance of the environmental variables decreased with egg number (Fig. 4a; Spearman's correlation: R = −0.45, P = 0.19) and brood care (Fig. 4b; Spearman's correlation: R = 0.57, P < 0.1, Mann–Whitney: U = 3, P < 0.1).

Principal component analysis produced one significant axis (PC axis 1) that explained 67% of the total variance (Fig. 5, Table 3). This axis showed a strong positive correlation with niche breadth, polytypism, egg number and importance of space (configuration of the network) and a negative association with brood care and the importance of the physical environment. Species were sorted in this gradient from specialists with a low ability to colonise and determined by the physical environment to generalists with a high ability to colonise (high drift propensity) and determined by regional spatial structure.

Table 3. Factor loadings of the principal component analysis (PCA); see text for details

	PC axes
	PCA 1	PCA 2
Environment	−0.665	−0.206
Space	0.841	0.245
Niche	0.730	0.646
Poly	0.905	−0.360
Egg	0.911	−0.177
Brood	−0.854	0.384

Discussion

The analysis of the aquatic gastropod metacommunity revealed that processes of both niche and dispersal assembly affect the distributions of species and patterns of richness. Our results are also in accordance with several studies of freshwater ecosystems showing a general trend of communities in rivers and interconnected waterbodies being more strongly related to spatial structure than are those of stagnant, isolated bodies of water (Cottenie et al., 2003; Urban, 2004; Cottenie, 2005; Vanschoenwinkel et al., 2007; Ledger et al., 2008; Muneepeerakul et al., 2008; Brown & Swan, 2010). Our analysis also confirmed that the structure of the aquatic network, and especially the direction of habitat linkages, has a major bearing on the assemblage. This relationship held particularly well for the upstream configuration, whereas the structure of the network downstream of a sampling location had no detectable influence. The analysis explained 17% of the variation in the presence/absence matrix of the gastropod metacommunity, and 83% remained unexplained. Seventeen per cent is low compared to some values from different types of ecosystems previously analysed (see Cottenie, 2005). One possible explanation is the stochastic nature of river-floodplain systems (Winemiller et al., 2010). Additionally, the mode of calculating habitat connectivity and missing environmental variables may partly explain the high proportion of unexplained variation, although this is unlikely since we have used the set of variables known to be of most importance in river floodplains. That we describe the ‘average’ metacommunity, beyond short-term temporal effects, may also blur the relationships and decrease explanatory power. However, we suggest that this is the correct scale to link species traits to metacommunity patterns since these species traits are determined on larger spatial and temporal scales (biogeography, evolution).

The analysis of single-species distributions and species traits revealed that the gastropod metacommunity cannot be described by a single paradigm. The importance of local and regional processes for species distribution is highly dependent on the degree of habitat specialisation (niche breadth). Spatial patterns were predominantly associated with generalist (or neutral) species and species sorting associated with specialists: 33% of species whose distribution was solely determined by environmental variables are predominantly habitat specialists, including riverine species such as Lithoglyphus naticoides, Potamopyrgus antipodarum and Galba truncatula and species bound to stagnant waterbodies such as Viviparus contectus and Lymnaea stagnalis. In contrast, the distribution of generalists is more strongly governed by the spatial configuration. Our findings confirm the result of Pandit et al. (2009), who investigated a rock-pool metacommunity of aquatic invertebrates. This suggests that the proportion of specialists to generalists may significantly determine the outcome of metacommunity analyses.

The inclusion of colonisation traits in the analysis allowed us to explain differences in distribution patterns between species and added further insight into the processes of assembly. In fluvial systems, colonisation is ultimately linked to drift in many animal groups. Under such conditions, species with traits that prevent drift are better able to maintain their position in the floodplain network and are consequently less determined by regional processes.

Our combined analysis revealed that a broad niche width and a high drift propensity are to a large extent interlinked: generalists possess life-history traits favouring drift (i.e. high fecundity and no brood care) whereas specialists have life-history traits that bear on drift prevention (i.e. ovovivipary or pasting egg capsules on conspecifics or hard ground). These relationships may be interpreted as trade-offs between habitat specialisation and colonisation ability/drift propensity. Specialists may succeed locally due to competitiveness and a strategy of risk reduction by preventing drift, because the risk is high that their offspring will drift to an unsuitable habitat, whereas generalists may succeed regionally due to a strategy that spreads risk by high reproductive outputs, high dispersal rates and adaptability to various types of habitats. Theoretical (McPeek & Holt, 1992; Doebeli & Ruxton, 1997; Mathias, Kisdi & Olivieri, 2001; Kisdi, 2002) and empirical (Bonte et al., 2003; McCauley, 2007) studies have already indicated the importance of these two alternative strategies for the coexistence of species. We have shown that they are very important for explaining metacommunity structure.

Our findings have several important implications for future research and applications of metacommunity analysis. The inclusion of species traits into metacommunity analysis may significantly improve our understanding of the underlying assembly processes (see also Pandit et al., 2009). Our study, the first to assess the joint impact of colonisation (drift propensity) and habitat specialisation on metacommunity patterns, also showed that a trade-off between the ability to colonise and habitat specialisation may explain the relative importance of local compared with regional conditions for species composition. Further studies of other species groups and systems will determine the generality of our findings. The inclusion of time series may also add to our understanding of the generation and maintenance of diversity. Floodplains are highly stochastic environments, and the time elapsed since a flood may impact the relative importance of the environment and spatial configuration in supporting local and regional diversity.

Our findings also have serious implications for schemes to restore large rivers where the main emphasis is the enhancement of the lateral integration between the river and its floodplain (Schiemer, 1999; Schiemer et al., 1999; Buijse et al., 2005). The structure of the network and the habitat connectivity of the river-floodplain system are thereby significantly altered, with subsequent effects on the relative importance of spatial and local processes for local diversity. An understanding of the spatial context of river-floodplain networks is thus of central interest for planning measures for the restoration of rivers.