Understanding the impact of flooding on trait-displacements and shifts in assemblage structure of predatory arthropods on river banks


  • Kevin Lambeets,

    Corresponding author
    1. Terrestrial Ecology Unit (TEREC), Department of Biology, Ghent University, KL Ledeganckstraat 35, B-9000 Ghent, Belgium;
      *Correspondence author. E-mail: Kevin.Lambeets@UGent.be
    Search for more papers by this author
  • Martijn L. Vandegehuchte,

    1. Terrestrial Ecology Unit (TEREC), Department of Biology, Ghent University, KL Ledeganckstraat 35, B-9000 Ghent, Belgium;
    Search for more papers by this author
  • Jean-Pierre Maelfait,

    1. Terrestrial Ecology Unit (TEREC), Department of Biology, Ghent University, KL Ledeganckstraat 35, B-9000 Ghent, Belgium;
    2. Research Institute for Nature and Forest (INBO), Kliniekstraat 25, B-1070 Brussels; and
    Search for more papers by this author
  • Dries Bonte

    1. Terrestrial Ecology Unit (TEREC), Department of Biology, Ghent University, KL Ledeganckstraat 35, B-9000 Ghent, Belgium;
    2. Wuerzburg University, Field Station Fabrikschleichach, Glashuettenstrasse 5, D-96181 Rauhenebrach, Germany
    Search for more papers by this author

*Correspondence author. E-mail: Kevin.Lambeets@UGent.be


  • 1Species assemblages of naturally disturbed habitats are governed by the prevailing disturbance regime. Consequently, stochastic flood events affect river banks and the inhabiting biota. Predatory arthropods occupy predominantly river banks in relation to specific habitat conditions. Therefore, species sorting and stochastic processes as induced by flooding are supposed to play important roles in structuring riparian arthropod assemblages in relation to their habitat preference and dispersal ability.
  • 2To ascertain whether assemblages of spiders and carabid beetles from disturbed river banks are structured by stochastic or sorting mechanisms, diversity patterns and assemblage-wide trait-displacements were assessed based on pitfall sampling data. We tested if flooding disturbance within a lowland river reach affects diversity patterns and trait distribution in both groups.
  • 3Whereas the number of riparian spider species decreased considerably with increased flooding, carabid beetle diversity benefited from intermediate degrees of flooding. Moreover, regression analyses revealed trait-displacements, reflecting sorting mechanisms particularly for spiders. Increased flooding disturbance was associated with assemblage-wide increases of niche breadth, shading and hygrophilic preference and ballooning propensity for spider (sub)families. Trait patterns were comparable for Bembidiini carabids, but were less univocal for Pterostichini species. Body size decreased for lycosid spiders and Bembidiini carabids with increased flooding, but increased in linyphiid spiders and Pterostichini carabids.
  • 4Our results indicate that mainly riparian species are disfavoured by either too high or too low degrees of disturbance, whereas eurytopic species benefit from increased flooding. Anthropogenic alterations of flooding disturbance constrain the distribution of common hygrophilous species and/or species with high dispersal ability, inducing shifts towards less specialized arthropod assemblages. River banks with divergent degrees of flooding impact should be maintained throughout dynamic lowland river reaches in order to preserve typical riparian arthropod assemblages.


The development of a trait-based ecology provides insight in assemblage-wide functional responses in environmentally variable environments (Van Looy et al. 2006; Violle et al. 2007). Changes in species distribution result from species sorting, mass effects or patch dynamics, leading eventually to community-wide character displacements or community-wide character shifts (Schluter 2000; Marchinko, Nishizaki & Burns 2004). For environments that are affected strongly by natural or anthropogenic disturbance, assemblages of species are expected to be structured by the ability of the species to react upon these disturbances (Plachter & Reich 1998; Ribera et al. 2001; Bonte, Lens & Maelfait 2006a). Because this involves species assimilation, assemblage-wide changes in species diversity are predicted to result from species sorting rather than substantial dispersal per se (Driscoll & Weir 2005). In contrast, when the magnitude of disturbance is higher than tolerated by the potential inhabitants, only highly dispersive species will be able to persist due to repeated colonization events (McAuliffe 1984; Ribera et al. 2001), with mass effects affecting species assemblages (e.g. Schmidt & Tscharntke 2005). Specialized species may be able to survive short-term disturbances, reappearing quickly after the disturbance subsides or benefiting from newly created structural elements (Weigmann & Wohlgemuth-von Reiche 1999; Rothenbücher & Schaefer 2006). However, responses depend upon the type of disturbance and the relation with species, functional traits (Bonte et al. 2006a; Moretti, Duelli & Obrist 2006; Papaik & Canham 2006). When trait variation does not prevail in relation to disturbance regimes, species assemblages can be considered to be functionally equivalent (Ackerly & Cornwell 2007). Consequently, assemblage-wide character displacements rather than trait shifts in response to species sorting take place (Schluter 2000). This can be realized by shifts of taxonomically different species with similar functional traits within assemblages (Marchinko et al. 2004). Which patterns underlie assemblage structure are expected to depend on intrinsic dispersal abilities. Therefore, disturbance may act as an important trigger, affecting assemblage structure in particular ways.

Localized rare disturbance events, irrespective of their magnitude or frequency, are expected to exert a minor effect on regional diversity (Chase 2003; Bonte et al. 2006a). However, spatially restricted disturbance can be important to facilitate the occurrence of specialized species that are able to react rapidly upon changing environment conditions (Bonn, Hagen & Wohlgemuth-von Reiche 2002; Rothenbücher & Schaefer 2006). Observed patterns may, however, vary considerably according to the taxonomic group and with the spatial scale of study (Pollock, Naiman & Hanley 1998; Prinzing et al. 2007; Sanders et al. 2007). Disturbance mechanisms appear to be especially relevant in riverine landscapes in which flooding contributes to strong environmental heterogeneity (Naiman & Décamps 1997; Ward et al. 2002), with subsequent highly structured assemblage patterns and related species diversity (Robinson, Tockner & Ward 2002; Naiman, Décamps & McClain 2005; Van Looy et al. 2005). Unravelling these patterns should be the foundation of riparian ecology (Jensen et al. 2006). As stated by Vannote et al. (1980) and Van Looy et al. (2006), assemblages from harsh riparian environments are assumed to shift constantly in relation to the prevalent disturbance regime, with synchronized species replacements throughout the river system. Therefore, if flooding disturbance affects environmental properties in a homogeneous way as induced by anthropogenic alterations of flooding (either extremely high or low flows), a high similarity in species diversity, assemblage structure and functionality would be expected. However, even if general environmental conditions are spatially similar under disturbance, temporal variation in disturbance will affect the distribution of mobile species, due to the creation of different colonization windows with subsequent species replacements under low frequencies of disturbance (McAuliffe 1984; Death & Winterbourn 1995). Therefore, different aspects of flooding disturbance should be studied simultaneously and in an integrated manner (Langhans & Tockner 2005; Van Looy et al. 2005). In general, differences between local levels of species richness and patterns of species traits reflect the influence of local environmental fluctuations and suggest its possible interference in species interactions, eventually determining the composition of local and regional assemblages.

Whether assemblage composition affected by flooding results from either equivalent or contrastive changes in assemblage-wide traits is virtually undocumented for the riparian fauna (but see Desender 1989a; Plachter & Reich 1998). Given the general idea that sets of traits are related to species abilities to cope with stressful situations, we applied a functional trait approach for predatory arthropods to delineate relevant insights for the restoration and conservation of the vulnerable riparian biodiversity (Kremen et al. 1993. Therefore, we assessed diversity patterns, assemblage-wide shifts and variation in species traits of two well-studied and dominant groups of predatory arthropods, respectively, spiders (Araneae) and carabid beetles (Carabidae), along riparian river banks. We particularly questioned: (i) whether patterns in diversity and species traits are affected by flooding disturbance among and within taxonomic groups; (ii) whether the underlying mechanisms are related to species sorting with congruent assemblage-wide character shifts; and (iii) whether flooding disturbance (dis)favours species with distinct ecological traits.

Materials and methods

study system and sampling protocol

The Common Meuse is the most natural part of the river Meuse and covers approximately 45 km of the total c. 900 km river trajectory. Due to its rainfed character and the rocky soils of the upstream catchment, the watercourse is characterized by strong river flow fluctuations and a meandering pattern of isolated river banks (Pedroli et al. 2002; Van Looy et al. 2006). These banks comprise a top layer of coarse shingle with a sharp sand–gravel or sand–loam fraction in between and related changes in vegetation (Peters, Van Looy & Kurstjens 2000). Only when the river discharge drops below 200 m3 s−1 (from May to September) are gravel banks gradually exposed. At this rather restricted regional scale, no longitudinal downstream variation of gravel structure, vegetation composition or disturbance frequency occurs (all correlations r < 0·24), as reflected by species assemblage structure (Lambeets et al. 2008).

All river banks along a continuous part of the river trajectory (Fig. 1) were sampled from 6 April 2004–2005 to 19 July 2005 with pitfall traps (Φ 9 cm; 6% formaline solution; emptied fortnightly). Each gravel bank contained three to six pitfalls, divided over a maximum of two stations. Pitfalls were arranged parallel with the waterline, situated at an average distance of 6·1 m from the loamy river dyke for higher stations and 21·3 m for the furthest stations on larger banks. As recommended by Topping & Sunderland (1992), pitfalls were spaced 10 m apart in order to avoid interference between the traps. Because unpredictable flood events caused data loss on several occasions, trapped species were interpolated distinctly for each sample date, pitfall trap and sample station. For each species, catches were pooled to total numbers per sample station. It is important to recognize that pitfall trapping has some inherent biases, and catches can be affected by factors including habitat structure, weather conditions and the preservative used (Topping & Sunderland 1992). In this study, standardized pitfall trapping is an appropriate collection method, as we aimed to compare patterns of assemblage-wide (weighted) species traits as affected by flooding disturbance. Contrary to other studies (Andersen 1995), cryptic and smaller-sized individuals made up the majority of the catches (e.g. Bembidiini carabids and linyphiid spiders), by which our sample data are believed to reflect local arthropod composition well, and hence liable for concrete interpretation.

Figure 1.

Map of the River Meuse basin with inset for the Common Meuse river reach and its riparian margin; sampled river banks are indicated as inline image.

characterization of environmental parameters

Flooding is affected by local topography as well as by regional chorological factors (Pedroli et al. 2002; Naiman et al. 2005; Van Looy et al. 2006) and influences both local humidity and vegetation structure, being the most important drivers for habitat quality in the studied arthropod groups (Turin 2000; Entling et al. 2007). Therefore, we recorded parameters related to flooding disturbance, river bank and channel geometry, substrate composition and vegetation structure. The landscape-related parameters measured were sample site location, connectivity along the riparian corridor and surrounding landscape composition. For ease of reading the measured variables, applied field methodology and interpretations of the main principal components are explained in Supplementary material, Appendix S1. Principal components analysis (PCA; Goodall 1954) revealed the prevalence of one ‘disturbance’-axis (PCdyn; eigenvalue 7·102; explanatory value 18·69%), which correlated with flooding disturbance aspects and substrate composition after Bonferroni correction (Table 1). Increasing values of PCdyn indicate a higher frequency of flooding during the sample period, an increased rising speed of the washing water and a substrate composed of less coarse gravel, fine-grained between sediment fraction and increased siltation. Two other axes explained variation related to river bank and channel geometry (PCgeo; eigenvalue 5·166; explanatory value 13·59%) and patch size and vegetation structure (PCveg; eigenvalue 4·284; explanatory value 11·27%). Because we emphasized studying river bank arthropod diversity and assemblage-wide patterns of functional trait distribution in relation to flooding disturbance sensu lato, we retained gravel bank scores from the first principal component for further analyses.

Table 1.  Pearson correlations with the first principal component (PCdyn) of measured parameters of river banks along the Common Meuse river reach. Parameters were transformed accordingly if they did not meet the normality assumption (McCune & Grace 2002). Only significant parameters are shown. Correlation coefficients r > 0·570 are significant after Bonferroni correction. For an overview of the environmental characterization based on the measured parameters and a concise explanation of the applied field methodology see Supplementary material, Appendix S1
ParameterVariable measuredMethodologyPCdyn
  1. River discharge regimes taken from http://www.lin.vlaanderen.be/awz/waterstanden/hydra/ (hourly values). Substrate composition are estimated values based on digital pictures within a 1 × 1 m quadrat surrounding each pitfall taken fortnightly during the field survey.

Flooding disturbanceRSregrRising speed of washing water based on river discharge regimes and fortnightly measured distances pitfalls – water line−0·585
Flooding disturbanceWFRRiver bank water flow rate based on based on river discharge regimes and fortnightly measured distances pitfalls – water line−0·866
Flooding disturbancedayflNumber of days flooded during sampling period based on river discharge regimes and WFR (log) 0·811
Flooding disturbancedayfl5 yearsNumber of days flooded between 2000 and 2005 based on river discharge regimes and WFR (log) 0·843
River bank topographyorientclOrientation eighth of river bank−0·667
Substrate compositiongravAverage gravel size (6 classes ranging from 0 to 10 cm until > 50 cm)−0·782
Substrate compositionsandSediment composition (sand–loam ratio)−0·852
Substrate compositionsiltSiltation class index (none–covering 1/4–half–up to dyke foot) 0·771

species richness and species traits

Species richness (alpha diversity, being the total species richness within one sample station equal to three pitfall traps) was calculated as the total number of species caught in each sample station. As this measure is affected by rare accidental vagrants we used the richness of resident species, i.e. species appearing with at least 10 individuals within one sample station (Bonte et al. 2006a), as a more stringent measure. Riparian diversity was calculated as the species richness of riparian specialists. Thereby, species were defined as ‘riparian’ based on relevant literature dealing with the ecological requirements of spiders (Hänggi, Stöckli & Nentwig 1995; Harvey, Nellist & Telfer 2002) and carabid beetles (Desender et al. 1995; Turin 2000).

Five traits were chosen to represent important life history features of spiders and carabid beetles. Niche breadth was considered as the number of habitat types (related to the species’ geographical rareness) in which spider and carabid beetle species were caught, as derived from Hänggi et al. (1995) and Boeken et al. (2002), respectively. Shading and moisture preference were obtained from habitat type preferences, as calculated by Entling et al. (2007) for spiders (xerophily), ecological group classification, as summarized by Turin (2000), and Boeken et al. (2002) for carabid beetles (hygrophily). Average body size of female spiders was derived from Roberts (1987, 1998), while Boeken et al. (2002) was consulted for the average body size of carabid beetles. Ballooning propensity of spiders, i.e. whether or not aerial dispersal can be performed by a species, was taken from the review by Bell et al. (2005) and extended with new experiments for riparian spiders (Bonte & Lambeets, unpublished data). Flight ability of carabid beetles was assessed by relative wing development in relation to body size, as defined by Desender (1989b). A complete list of trapped numbers and species trait values can be found in Supplementary material, Appendix S2.

data analysis

Our trait-based approach was based on the weighted averages and the variances of trait values of species co-occurring in local assemblages. Average values serve as comparable measures in order to array assemblages along a one-dimensional gradient. The analysis of trait variance is complementary and essential because weighted averages can be the same, despite variation in trait variance, and therefore eases the distinction between prevalent structuring processes (Ackerly & Cornwell 2007) and thus assembly rules (Holdaway & Sparrow 2006). Consequently, we were able to distinguish between assemblage-wide ecological mean values and their amplitudes.

General linear models (GLMs; proc mixed, sas version 9·1) were used to assess the influence of disturbance on species richness and species traits. Number of species, weighted averages and variances of trait values were the dependent variables, whereas the first principal component (PCdyn) was considered as the continuous factor reflecting flooding disturbance sensu lato. Both linear and quadratic functions were modelled. The most reliable model was inferred by Akaike's information criterion (AIC), based on model fit and model complexity criteria (Johnson & Omland 2004). In all cases, normality of residuals was checked (proc univariate, sas version 9·1). Because patterns in life history traits are highly interdependent according to common phylogenetic origin (Bonte et al. 2006a), analyses were performed at the lowest workable phylogenetic level, being the subfamily level for spiders (Erigoninae, Linyphiinae, Lycosidae) and the tribe level for carabid beetles (Bembidiini, Pterostichini). Because the interaction between taxonomic group and traits were highly significant for average values (Araneae: F2,80 > 19·8; all P < 0·0001; Carabidae (F1,54 > 10·2; all P < 0·0023) and most variances (Araneae: F2,77 > 15·3; all P < 0·0001; Carabidae (F1,54 > 21·1; all P < 0·0001 except hygrophily: F1,54 = 1·00; P = 0·3212 and wing development: F1,54 = 0·02; P = 0·8774), we performed trait analyses separately for the differently distinguished taxonomic groups.


species richness

Alpha diversity of carabid beetles (Fig. 2a) and numbers of resident species (Fig. 2b) peaked at an intermediate degree of flooding disturbance, whereas no significant patterns were found for spider species richness. The relation between the richness of stenotopic riparian species and PCdyn revealed a linear decrease for spiders and an intermediate optimum for carabid beetles with increased flooding (Fig. 2c). F-values, significance levels and AIC values are presented in Table 2.

Figure 2.

Relationship between spider and carabid diversity and the degree of flooding disturbance along a lowland gravel river. (a) Alpha diversity; (b) richness of resident species; (c) richness of riparian species. The principal component scores arising from a principal components analysis of site-specific habitat characteristics (PCdyn) are used to indicate the degree of flooding disturbance along the x-axis.

Table 2.  Influence of flooding disturbance on species richness of spider and carabid beetle assemblages of river banks. General linear model (GLM) regression statistics and Akaike's information criteria (AIC) values are shown for spiders and carabid beetles. Degrees of freedom are indicated below each taxonomic group as (numerator degrees of freedom; denominator degrees of freedom)
Diversity measureRegression statisticsSecond-order relationFirst-order relation
Araneae (1, 25)Carabidae (1, 25)Araneae (1, 26)Carabidae (1, 26)
Alpha diversityF0·578·682·890·05
Resident diversityF1·454·940·840
Riparian diversityF0·336·826·121·53

assemblage-wide ecological traits

In the following, we present significant relationships only between flooding disturbance (PCdyn) and assemblage-wide species traits. F-values, significance levels and AIC values are presented in Table 3.

Table 3.  Influence of flooding disturbance on niche breadth, shading preference, drought/moisture preference, body size and flight ability (spider ballooning propensity and carabid beetle wing development) of spider and carabid beetle assemblages from river banks. General linear model (GLM) regression statistics and Akaike's information criteria (AIC) values are shown for weighted averages (Tables 3a and 3b) and variance (Tables 3c and 3d), respectively, for each of the spider (sub)families (Lycosidae, Erigoninae, Linyphiinae) and carabid beetle tribes (Bembidiini, Pterostichini). Degrees of freedom are indicated below each taxonomic group as (numerator degrees of freedom; denominator degrees of freedom)
(a)Regression statisticsSecond-order relationFirst-order relation
Life history traitLycosidae (1, 25)Erigoninae (1, 25)Linyphiinae (1, 24)Lycosidae (1, 26)Erigoninae (1, 26)Linyphiinae (1, 25)
Niche breadthF2·451·240·1414·28·272·84
Shading preferenceF17·023·371·3714·238·135·07
P0·00040·07840·25280·0009< 0·00010·0334
Female sizeF74·4613·120·0142·6220·854·43
P< 0·00010·00130·9433< 0·00010·00010·0456
Ballooning propensityF1·920·320·338·435·830·61
(b) Second-order relationFirst-order relation
Life history traitBembidiini (1, 25)Pterostichini (1, 25)Bembidiini (1, 26)Pterostichini (1, 26)
Niche breadthF1·640·4414·160·05
Shading preferenceF7·710·013·452·24
Body sizeF0·360·451·411·64
Wing developmentF2·361·999·757·68
(c) Second-order relationFirst-order relation
Life history traitLycosidae (1, 25)Erigoninae (1, 25)Linyphiinae (1, 24)Lycosidae (1, 26)Erigoninae (1, 26)Linyphiinae (1, 23)
Niche breadthF1·310·4203·650·050
Shading preferenceF0·050·4100·050·070·26
Female sizeF2·950·583·237·559·580·02
Ballooning propensityF1·545·180·280·230·330
(d) Second-order relationFirst-order relation
Life history traitBembidiini (1, 25)Pterostichini (1, 25)Bembidiini (1, 26)Pterostichini (1, 26)
Niche breadthF0·310·440·285·62
Shading preferenceF0·174·841·230·28
Body sizeF5·252·763·2115·84
Wing developmentF1·382·030·940·75

niche breadth, shading and moisture preference

Assemblage-wide niche breadth increased with increasing disturbance in Erigoninae, Lycosidae (Fig. 3a; Table 3a) and Bembidiini (Fig. 3b; Table 3b). Variance in niche breadth decreased monotonously with flooding for Pterostichini assemblages (Fig. 3c; Table 3d).

Figure 3.

Relationship between spider and carabid beetle niche breadth and the degree of flooding disturbance (PCdyn) along a lowland gravel river. (a) Weighted average Erigoninae, Lycosidae; (b) weighted average Bembidiini; (c) variance Pterostichini.

All spider (sub)families showed an increased preference for shaded conditions with increasing flooding disturbance. Assemblages with, on average, a higher degree of shading preference occurred at more disturbed river banks (Fig. 4a; Table 3a). In contrast, shading preference for Bembidiini was lower at low degrees of flooding and a monotonous increase of shading preference was noted as flooding increased. However, this relation is influenced highly by the prevalence of agrobiont Bembidion carabids on the lowest river banks (skewed distribution at Fig. 4b; Table 3b). Variance of shading preference peaked at intermediate degrees of flooding for Pterostichini (Fig. 4c; Table 3d).

Figure 4.

Relationship between spider and carabid beetle shading preference and the degree of flooding disturbance (PCdyn) along a lowland gravel river. (a) Weighted average Erigoninae, Linyphiinae, Lycosidae; (b) weighted average Bembidiini; (c) variance Pterostichini.

Assemblage-wide xerophily of all spider (sub)families on average decreased with increasing flooding disturbance (Fig. 5a; Table 3a). Variance in xerophily decreased solely for Lycosidae (Fig. 5b; Table 3c). Bembidiini carabids showed a significant decrease in hygrophilic species with increasing disturbance (Fig. 5c; Table 3b).

Figure 5.

Relationship between spider xerophiliy and carabid beetle hygrophily and the degree of flooding disturbance (PCdyn) along a lowland gravel river. (a) Weighted average Erigoninae, Linyphiinae, Lycosidae; (b) variance Lycosidae; (c) weighted average Bembidiini.

body size and dispersal ability

Female size of Lycosidae decreased to a minimum at intermediately disturbed sites, whereas an increase with disturbance was prevalent in Erigoninae and Linyphiinae (Fig. 6a; Table 3a). Significant linear decreases were found with respect to variance in assemblage-wide female size for Erigoninae and Lycosidae (Fig. 6b; Table 3c). Assemblage-wide average size of Pterostichini carabids increased significantly with increasing disturbance (Fig. 6c; Table 3b). Variance in carabid beetle body size was lower at more disturbed river banks for Bembidiini, whereas it increased for Pterostichini (Fig. 6d; Table 3d).

Figure 6.

Relationship between spider female body size and carabid beetle body size and the degree of flooding disturbance (PCdyn) along a lowland gravel river. (a) Weighted average Erigoninae, Linyphiinae, Lycosidae; (b) variance Erigoninae, Lycosidae; (c) weighted average Pterostichini.

Erigoninae and Lycosidae with known ballooning propensity are favoured by increased disturbance (Fig. 7a; Table 3a). Variance in ballooning propensity of Erigoninae peaked at intermediate disturbance (Fig. 7b; Table 3c). On average, assemblage-wide wing development increased for Bembidiini and Pterostichini (Fig. 7c; Table 3b).

Figure 7.

Relationship between spider ballooning propensity and carabid beetle wing development and the degree of flooding disturbance (PCdyn) along a lowland gravel river. (a) Weighthed average Erigoninae, Lycosidae; (b) variance Erigoninae; (c) weighted average Bembidiini, Pterostichini.


Our study contributes to a solid understanding of functional species traits of component predatory arthropods of river banks and their responses to flooding disturbance, thereby affecting species assemblage structure. Species richness of carabid beetles benefits from intermediate flooding disturbance, whereas the richness of stenotopic riparian spiders increases with subsiding flooding. Congruent assemblage-wide shifts in species traits show that species sorting in response to flooding is the underlying mechanism within spider (sub)families and Bembidiini carabids. However, sorting mechanisms appear contrastive in Pterostichini carabid assemblages.

Only the number of riparian spider species decreases with increasing flooding disturbance. This suggests that increased flooding facilitates the settlement of eurytopic species, while specialists tend to disappear. The increase in eurytopic species is reflected in assemblage-wide shifts towards higher dispersal ability, higher shading and moisture preference (lower xerophily) and a smaller body size in Lycosidae. Moreover, lycosid and erigonid spiders with aerial dispersal capacity dominate lower river banks, although both highly mobile and sedentary erigonids are present on banks with an intermediate degree of disturbance, whereas variance in ballooning propensity remained constant for lycosid spiders. This indicates a clear shift towards generally mobile species, but with sorting mechanisms prevalent at high and low flooding for erigonids and species replacements for lycosids. The overall presence of highly dispersive, rather generalist, agrobionts indicates that species from neighbouring arable habitats colonize river banks and dominate assemblages under intensive flooding disturbance. Mass effects, by which a continuous input of species from source habitat is expected (Leibold et al. 2004), are consequently prevalent, comparable with results for spiders from agricultural ecosystems (Schmidt & Tscharntke 2005; Öberg, Ekbom & Bommarco 2007). Dispersal of specialist species might be important on a more restricted spatial scale, adding to subsequent recolonization or escaping flooding successfully (Morse 1997; Kraus & Morse 2005). Generally, spider diversity is related positively to vegetation composition (Perner & Malt 2003; Beals 2006). As previous studies have indicated flooding to homogenize vegetation structure (Peters et al. 2000; Shafroth, Stromberg & Patten 2002), increased flooding can result in a lowered diversity. Nevertheless, studies concerning boreal or upland rivers showed positive relationships between flooding and vegetation heterogeneity per se (Nilsson et al. 1989; Renöfält, Nilsson & Jansson 2005), with concordant effects on riparian arthropod diversity (Bonn et al. 2002). Because vegetation composition is not related to flooding disturbance at our considered spatial scale (see Supplementary material, Appendix S1), it potentially affects species distribution patterns differently to flooding. The decrease in variance of xerophily indicates that assemblages are dominated by only few ecologically similar species, e.g. Pardosa sp. This pattern is reflected similarly by assemblage-wide decreases of both average body size and its variance with increased flooding disturbance. For Erigoninae an opposite pattern was found, with mainly larger species on more disturbed river banks, whereas small linyphiids are replaced by larger species as the variance in body size remained constant. Because larger Erigoninae are the dominant dispersers during early summer, this pattern can be expected to be caused by a replacement of specialist species (often xerophilic species) by highly dispersive agrobionts and hygrophilous species. Agrobionts, however, may not be able to survive flooding events due to the lack of behavioural or physiological adaptations (Suter, Stratton & Miller 2004; Rothenbücher & Schaefer 2006), thereby experiencing river banks as sink habitat.

In contrast to spiders, carabid beetle species richness peaks at intermediate levels of disturbance. Shifts in traits suggest that species sorting is prevalent mainly for Bembidiini species. Interestingly, assemblage-wide changes in dispersal ability are comparable. While Bembidiini species are often considered as inherent elements of the riparian carabid fauna (Turin 2000; Manderbach & Hering 2001), preferring dynamic and moist circumstances, our results demonstrate that specialist species tend to disappear at highly disturbed river banks. On average, shading preference was lowest at higher river banks whereas hygrophily decreased with increased flooding. Variance patterns of body size, however, show that only a restricted subset of Bembidiini species is able to persist on river banks at both ends of the disturbance gradient. These patterns indicate that species tend to be lost as flood pulses rise or at lower degrees of flooding, adding to the prevalence of sorting mechanisms for Bembidiini assemblages, comparable to spiders. Both floods and low flows are often related to anthropogenic alterations of the flooding regime and shown to be detrimental for the invertebrate fauna (Usseglio-Polatera & Beisel 2002; Suren & Jowett 2006). Sorting mechanisms appear less obvious for Pterostichini assemblages. Niche breadth variance is low, especially at the most disturbed river banks, whereas larger species with well-developed wings (cf. Bembidiini) become dominant. Therefore, increased flooding is clearly responsible for the elimination of smaller, more specialized Pterostichiini species from local assemblages, yet they are known to colonize flooded sites quickly by means of epigeal locomotion (Lang & Pütz 1999). Next to it, Pterostichini species tend to profit from intermediate degrees of disturbance as shown by the variance in shading preference. Assemblages of Pterostichini species are structured mainly by changes in dispersal capacity rather than by replacements of species with idiosyncratic ecological needs. Therefore, sorting mechanisms seem to affect Pterostichini assemblages in other ways than Bembidiini, but effects of anthropogenically altered flood regimes are equally prevalent. Floods, in particular, cause shifts towards eurytopic Pterostichini assemblages, hence specialized species are lost. In general, carabid beetle trait patterns in relation to flooding are more variable and specific according to the considered phylogenetic level compared to spiders. This may be caused by conservative traits such as elytra coloration and diurnal activity patterns (related to desiccation tolerance; Desender 1989a). Sorting mechanisms related to local habitat conditions at both ends of the disturbance gradient are in concordance with Bonn & Schröder (2001), who demonstrated incidence patterns to vary in opposite directions for a specialized Agonum and a eurytopic Pterostichus species. Bonn & Kleinwächter (1999) indicated apparent sorting mechanisms for riparian carabid beetle assemblages with specialized species closer to the waterline, shifting to a less specialized carabid fauna further away. In concordance with our results and earlier studies of Desender et al. (1993), they clearly showed wing development to be related to the distribution of Agonum and Bembidion species (increased overall macroptery at sites near the water edge) and Pterostichus species (reduction of hind wings near dykes). Although different flood regimes benefit different species, an optimum in species richness at intermediately disturbed banks is assumed to be maintained by increased microhabitat heterogeneity (Pollock et al. 1998). This allows for a narrow niche separation (Bonn & Kleinwächter 1999), hence benefiting the persistence of species with divergent habitat preferences and interrelated sets of species traits (e.g. dispersal ability) (Ward et al. 2002; Vanbergen et al. 2005). Either low flows or increased flooding would disfavour riparian carabid beetles, leading to constraints on the local assemblages (cf. Vanbergen et al. 2005; Stromberg et al. 2007).

Notwithstanding the prevalence of mobile species on all river banks, sorting mechanisms underlie species assemblage structure. In particular, common hygrophilous species are better represented as flooding increases. However, riparian species with well-developed dispersal abilities (e.g. Bembidiini) are well presented throughout the river system (Desender 1989a; Desender et al. 1993; Lambeets et al. 2008), thereby indicating their efficient movement throughout the system, probably resulting in one patchy population (Bates, Sadler & Fowles 2006). Patterns could, at first sight, be generated by the local landscape structure, but our analysis showed that the latter is independent of flooding regime. Therefore, more intrinsic factors related to, for example, general activity or sediment preference should influence trait patterns. As shown by Bonte et al. (2006b), dispersal mode (passive controlled in carabid beetles vs. predominantly uncontrolled passive in spiders) could additionally underlie differences of the observed species distribution patterns, with stronger resemblance in carabid assemblages due to their better-developed colonization abilities.

In conclusion, flooding disturbance is responsible for variable species sorting in two groups of opportunistic predatory arthropods. Assemblage-wide shifts in species traits were directional for spider (sub)families, with concordant effects regardless of their dispersal abilities. Shifts for carabid beetle tribes were similar for Bembidiini, yet less univocal for Pterostichiini. As eurytopic as well as specialist species are locally present, our results indicate that variation in riparian arthropod assemblages is enhanced by different flood regimes. If we take into account that especially cursorial spider species, with larger body sizes and a higher degree of habitat specialization, and hygrophilous carabid beetles with smaller body sizes, are more vulnerable to extinction (Niemeläet al. 2002; Bonte et al. 2006a), human-driven alterations in flooding disturbance, i.e. either too high or too low, can be expected to have a major impact on arthropod assemblages and the distribution of rare riparian species (Bonn et al. 2002; Lambeets et al. 2008). Moreover, a homogenization of habitat structure as a consequence of repetitive flood events or its absence will result in a more uniform and less specialized species composition (Bonn & Kleinwächter 1999; Vanbergen et al. 2005; Van Looy et al. 2006). Species are added to local communities as disturbance seizes for spiders or at intermediate degrees of disturbance for carabid beetles, thereby increasing alpha diversity (Jonsen & Fahrig 1997; Robinson et al. 2002; Bonte et al. 2006a). Due to generally better-developed dispersal abilities, riparian carabid beetles appear more resilient and able to persist under increased dynamics (Van Looy et al. 2005).


We would like to thank Dr Ir. K. Van Looy (InBo) for providing useful information, I. Lewylle who helped out with spider identification and Dr K. Desender for checking carabid beetle identifications. Hans Matheve rendered assistance during ArcGIS version 9·1 applications. The first author is funded by a PhD grant from the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen). M. L. Vandegehuchte (PhD student) and D. Bonte (post-doctoral fellow) are funded by the Research Foundation – Flanders (FWO-Vlaanderen). The authors are grateful for the useful comments of two anonymous referees.