Literature-derived evidence for stream invertebrates
When Townsend & Hildrew (1994) provided their habitat templet predictions for rivers, they did this in association with a synthesis of long-term ecological data from the Upper Rhône River (France), i.e. their predictions were immediately tested across many groups of the plant and animal kingdoms (Statzner et al., 1994a). These comprehensive tests illustrated that the trait patterns of the majority of the assessed organismic groups varied significantly across the different habitat types (superficial to interstitial, main channel to oxbow lakes, permanent to temporary waters) of a relatively natural large river floodplain (Resh et al., 1994). Although significant, this variation of trait patterns was relatively weak in terms of quantities or clarity (e.g. Dolédec & Statzner, 1994; Usseglio-Polatera, 1994), particularly if one admits that the gradients of temporal and spatial heterogeneity across the habitat types of a large river floodplain envelop many of the conditions occurring in freshwater habitats. Subsequent studies on several trait categories expected to respond to physical disturbance in New Zealand streams (Townsend et al., 1997a) or that described reproduction and habitat use of aquatic insects of the world (Statzner et al., 1997b) confirmed the results of the Rhône project. Thus, by 1997, three independent assessments indicated that the trait patterns of natural lotic invertebrate communities are perhaps less variable than previously anticipated.
Assessing potential trait stability through the presence–absence of trait categories, Snook & Milner (2002) illustrated that extremely harsh stream reaches in the French Pyrénées (c. 1 km below a glacier) lacked a few trait categories (e.g. a semivoltine life cycle) that occurred further downstream (c. 1.5–3 km below the glacier), whereas all categories occurring near the glacier also occurred at the downstream sites. In comparison, taxon richness varied much more between these two groups of reaches. Similar patterns have been reported from glaciated parts of the Austrian, French and Swiss Alps (Ilg & Castella, 2006; Füreder, 2007). Likewise, taxon richness in headwater streams varied considerably across three vegetation-defined ecological zones (alpine, spruce-fir, lodgepole pine) in the Rocky Mountains, whereas almost all of the studied trait categories occurred at all sites (Finn & Poff, 2005).
Across Europe, c. 90% of c. 500 almost natural stream sites had less than c. 10% of the 312 genera found across all sites, whereas c. 70% of sites had between c. 75–90% of the 61 trait categories included in the study, and 20 of these trait categories occurred at all sites (Statzner et al., 2007). Among the rarest trait categories (occurring at c. 10–20% of the sites) were very small and very large maximum body size, refuge use to resist desiccation during droughts, almost permanent attachment to the bottom substrate and a parasite or parasitoid feeding habit. Trait category richness at the sites increased rapidly with increasing genus richness and levelled off if more than c. 30 genera occurred (Statzner et al., 2007). Likewise, trait category-accumulation curves levelled off after only a few sites, whereas genus-accumulation curves did not level off for 265 sites each in temperate (Europe) or mediterranean (Europe, North Africa, Asia) climates (Bonada et al., 2007a).
More support for the consistent large-scale occurrence of trait categories is illustrated through an index of trophic completeness of invertebrate communities. This index combines food composition, feeding habit, food size, food acquisition behaviour and food ingestion type to define 12 trophic groups (Pavluk, Bij de Vaate & Leslie, 2000). In least human-disturbed rivers in Russia, Greece and the Netherlands, typically all of these trophic groups occurred regardless of the habitat type, climate zone and season (Bij de Vaate & Pavluk, 2004).
Across even larger scales, however, the occurrence of trait categories can differ. For example, stream invertebrates with very short life cycles, many cycles per year, periods of extended recruitments and overlapping cohorts occur in tropical streams of four continents (Statzner, 1976; Marchant, 1982; Jackson & Sweeney, 1995; Dudgeon, 2000) but are lacking in temperate Europe (Statzner et al., 2004).
Thus, natural filters for stream invertebrate traits at local- (e.g. below glaciers) or large-scales (temperate versus tropical climates) are so efficient that particular trait categories are systematically lacking. However, for many stream types and trait categories in a given climatic zone, such filters scarcely affect the qualitative trait category composition in invertebrate communities. Consequently, it is essential to know whether such natural filters have effects on the quantitative trait category composition of almost natural lotic invertebrate communities. In this context, it would be utopia to assess such large-scale effects through absolute quantities, as the mean annual invertebrate abundance at natural or almost natural European stream sites varies across four orders of magnitude (Statzner et al., 2007). Therefore, descriptions of the relative abundance of the trait categories have typically been used in the articles reviewed here and in subsequent sections.
To assess the potential quantitative effects of trait filters, we start by considering longitudinal (i.e. downstream) patterns, as the longitudinal zonation of the taxonomic invertebrate community structure along running waters is one of the oldest research topics in stream ecology. In particular, stream hydraulics change over shorter distances along headwaters than further downstream, causing similar changes in the taxonomic invertebrate community structure (Statzner & Higler, 1986; Grubaugh, Wallace & Houston, 1996). Correspondingly, the functional feeding group composition and/or other invertebrate traits changed more along (or were more variable among) headwater streams than in downstream sections in North America (Minshall et al., 1983; Grubaugh et al., 1996; Finn & Poff, 2005) and Europe (Snook & Milner, 2002; Statzner et al., 2005; Ilg & Castella, 2006). In contrast to the relatively abrupt taxonomic zonation patterns, traits changed more gradually along running waters.
These differences in the rate of trait change between headwater streams and larger running waters suggest that biomonitoring using the BTI approach should perhaps rely on separate assessment tools for headwaters and larger rivers. Along the latter, the biological trait category composition of benthic invertebrate communities was rather constant across large ecoregions (Dolédec et al., 1999). Furthermore, the trait differences among least human-impacted large river reaches of Europe were so minor that simple descriptions of frequency distributions of trait patterns (i.e. ignoring all environmental differences among the reaches) enabled the correct assignment of 80–90% of independent test sites to least-impacted conditions (Statzner et al., 2005).
Focusing on dimensions other than that along running waters, the stability of trait patterns has been assessed at various spatial scales. Starting at the smallest scale, Finn & Poff (2005) reported relatively similar trait patterns (compared to taxonomy patterns) for communities of different catchments in the Rocky Mountains. Likewise, invertebrate trait patterns were relatively stable over time if compared to taxonomic patterns in mediterranean climate streams of California (Bêche et al., 2006; Bêche & Resh, 2007). Assessments of streams within (Archaimbault et al., 2005) or across (Charvet et al., 2000) biogeographical regions of France illustrated that the biological traits of invertebrate communities in almost natural streams were relatively stable. For example, Archaimbault et al. (2005) reported significant differences for some of the tested biological traits across catchments with different geology (sandstone, granite, clay, schists) but concluded that these differences were quantitatively so negligible that a unique trait reference for streams of similar size could be used. Likewise, the functional feeding group composition of invertebrate communities in Swedish streams was relatively constant across six large ecoregions (Johnson et al., 2004). Scaling up to Europe, this trait stability was confirmed through analyses of the trait composition of 37 most natural regional stream types scattered from West Ireland to the Caucasus and from Central Lapland to Corsica (Statzner et al., 2001a). At an even larger scale, Bonada et al. (2007a) reported statistically significant but quantitatively small differences in the trait category composition of almost natural temperate (Europe) and mediterranean (Europe, North Africa, West Asia) lotic invertebrate communities. These results provide multiple evidence that the relative abundance of trait categories in natural stream invertebrate communities varies relatively little across large spatial scales of Europe and adjacent regions.
This pronounced trait stability has been related to the relative importance of abiotic versus biotic and actual versus historical trait filter action at the scale of sites and landscapes (large biogeographical regions) across Europe (Statzner et al., 2004): (i) actual abiotic filters acted significantly and independently of stream invertebrate taxon richness at the scale of sites and landscapes; (ii) biotic filters (as a result of biotic interactions) had no significant effects and (iii) evidence for the action of historical filters was weak. Thus, only the action of abiotic trait filters produced the relatively few and quantitatively weak (although highly significant) trait category changes observed. Statzner et al. (2004) explained these relatively uniform trait patterns with the action of strong stream system-specific abiotic filters that make the traits relatively similar across large spatial scales (corresponding to the ‘harsh’ condition in Fig. 1).
Thus, the abundance of stream invertebrates likely depends on trait categories that favour (pass through the harsh stream system-specific abiotic filters) or disfavour (are eliminated by the filters) viability in stream systems. Indeed, the mean European abundance of stream invertebrate genera increased with the possession of trait categories favouring this viability in stream systems (e.g. attachment to the stream bottom to resist the flow, aquatic passive dispersal with the flow, exploitation of abundant food sources) and decreased with the possession of trait categories disfavouring this viability (e.g. drag force increase associated with larger body size, flow exposure associated with aerial respiration) (Statzner, Bonada & Dolédec, 2008a). In addition, abundance consistently decreased with specialisation of the genera (e.g. low species richness, oddity of their overall trait profile from an ‘average’ European genus). Using a subset of these traits, it was possible to predict (in crosswise validations) 35% of the observed mean European (ln-transformed) abundance variability of 312 invertebrate genera (having c. 2200 lotic species in Europe) of 27 orders, or 51% of the abundance variability of 121 genera (having c. 1200 lotic species in Europe) of the may-, stone- and caddisflies (Statzner et al., 2008a).
Our last example on the large-scale stability of stream invertebrate trait patterns considers an approach focused on the future biomonitoring of pesticide impacts. Combining information on physiological sensitivity to organic toxicants and life-history traits indicating the recovery potential of invertebrates provided a database for the large-scale assessment of pesticide stress on stream systems (Liess & Von der Ohe, 2005). Von der Ohe et al. (2007) used two criteria (physiological sensitivity to organic toxicants > mean sensitivity of all taxa, generation time ≥ 0.5 year) to identify ‘at risk’ taxa and determined the relative abundance of these taxa in invertebrate communities across European rivers (from Spain to Finland). They applied their approach on all available monitoring data and indicated a ‘high ecological status’ for the reference conditions of all included river basins. In contrast, other national biomonitoring indices applied on the same data did not consistently indicate ‘high ecological status’ for reference sites.
Obviously, we do not believe that a unique global trait reference describing relative category abundances could be used, as we have illustrated through the above example on the occurrence of different life cycles in temperate and tropical regions. Relationships between relative trait category abundance and habitat provide more indications of large-scale differences in trait patterns. For example, grazers and shredders were similarly dominant in Europe and North America (Minshall et al., 1983; Statzner et al., 2004) but not so on a transect from tropical Asia to New Zealand (where shredders were rarer; Dudgeon, 1994; Lake, 1995). Furthermore, changes in the functional feeding group composition along running waters differed between North (e.g. Minshall et al., 1983; Grubaugh et al., 1996) and South (Miserendino, 2004; Tomanova et al., 2007) America. Finally, many invertebrate trait–habitat relationships (with flow, bottom roughness and benthic organic matter) differed between neotropical and European running waters (Tomanova & Usseglio-Polatera, 2007; see below for some examples). Thus, there is evidence that filters acting at very large spatial scales define a certain level of possible trait responses across continents.
In summary, the literature published over the last decade provides considerable evidence for the large-scale stability of natural (or almost natural) trait patterns in lotic invertebrate communities. However, some traits vary so much in certain circumstances (harsh environment below glaciers, along headwaters, across very large spatial regions) that it would be difficult to define a unique trait reference for all these circumstances in future applications of the BTI approach.
Data-derived evidence for stream invertebrates
Beyond the literature-derived evidence reviewed in the previous subsection, we analysed original data assembled for other purposes (an ongoing joint project) to assess the large-scale stability of stream invertebrate trait patterns. Given that abundant taxa dominate the trait patterns in communities, we examined whether abundant taxa in Europe and North America have similar traits that would result in similar trait patterns among natural stream invertebrate communities in both continents.
Using databases described in Appendix 1, our comparison of the mean trait profile of the common genera (25% of the genera having the highest mean abundance on each continent) illustrates a high intercontinental similarity for most trait categories (Fig. 2a). The clearest differences occurred in the food and feeding habit traits, which were partly caused by different definitions of these traits. In the European database, shredders could feed on coarse detritus but also on other coarse material (e.g. living macrophytes, living animals). In the U.S.A., the shredder habit was originally subdivided into herbivore-chewers/miners of living macrophytes and detritivores-chewers/wood borers of coarse detritus (e.g. Merritt & Cummins, 1978), but the subsequent use of the term was often limited to shredders of coarse detritus (e.g. Minshall et al., 1983). Consequently, shredding was more represented among the common European than the common North American invertebrate genera, whereas coarse detritus food was similarly represented on both continents (Fig. 2a). Among the other trait differences, common genera in Europe had (in comparison to North America) more frequent longer life cycles, once-per-year reproduction, aquatic imagines, aquatic active dispersal, cocoons to resist unfavourable conditions, aerial respiration, swimmers, interstitial forms, living macrophyte or microinvertebrate food, piercers and/or less frequent small size (≤2.5 mm), shorter life cycles, refuge use against desiccation, gill respiration, fine detritus food and/or deposit-feeders (Fig. 2a).
Corresponding to the previously reported relationships between abundance and trait categories in Europe (see above), common genera of both continents were characterised by trait categories favouring viability in flowing water (e.g. attachment of eggs to the stream bottom to resist the flow, aquatic passive dispersal with the flow, exploitation of abundant food sources). In contrast, traits disfavouring this viability were scarcely represented in common genera (e.g. larger body size and associated drag force increase, aerial respiration and associated flow exposure, swimming locomotion) (Fig. 2a).
Expectedly, the mean trait profile of the common genera governed the mean trait profile of the natural and almost natural invertebrate communities in both continents (Fig. 2b). The major difference between the mean trait profile of the common genera and the mean trait profile of the communities was the frequency of significant intercontinental differences, which was related to the number of replicates in the data sets (European-North American genera: 78–96; European-North American communities: 527–424). As a result, the standard errors around the community means were extremely low.
Plotting the mean trait category values of North America against those of Europe provides statistical confirmation that the overall trait patterns were very similar in both continents (Fig. 3). For both the category mean of the common genera and the category mean of the communities, R2-values were c. 0.9, regression slopes equalled c. one and intercepts equalled c. zero.
Figure 3. Plot of the overall mean trait profile of 61 trait categories in the U.S.A. versus Europe for (a) common invertebrate genera and (b) invertebrate communities (see Fig. 2 for further details). We indicate the line y = x (dashed) and the regression line (solid) of the models (parameter estimates ± 1 SE): (a) y = −0.304 ± 1.138 + 1.02 ± 0.05 x, R2 = 0.895, P < 10−15; (b) y = −0.449 ± 0.905 + 0.975 ± 0.035 x, R2 = 0.931, P < 10−15.
Download figure to PowerPoint
In summary, this example provides so far the strongest support for the idea that a unique trait reference for almost natural or natural conditions could be applied across very large spatial scales, because ecological significance (i.e. the percentage composition of categories per trait and per community) matters more than statistical significance. Thus, significant but quantitatively small (i.e. ecologically insignificant) trait differences across landscape units should not be used as an argument for small-scale, regionally differing trait references (for such pleas, see Heino et al., 2007).
Evidence of trait stability from other groups
The spatial scales across which trait patterns of other community types vary more or less are currently not well known. For lotic fish of Europe, more trait metrics were incorporated in assessment methods as the spatial scale increased (e.g. from ecoregion to continent), because traits provided more general information on the fish assemblages than, for example, taxonomic information (Schmutz et al., 2007). Predicting community characteristics of fluvial fish from hydraulics, biological trait compositions were more easily predicted (i.e. less variable across zoogeographical regions) than the relative abundances of the species in the French Rhône River basin (Lamouroux et al., 1999). Comparing trophic guild and body morphology composition of fish communities between hydrologically variable and stable stream sites of the U.S.A., Poff & Allan (1995) reported significant differences in some of the categories of these traits, although these differences were quantitatively weak. Expanding these comparisons to Europe and North America, Lamouroux, Poff & Angermeier (2002) found intercontinental convergences in the biological traits of fluvial fish (reflecting morphological and behavioural adaptations) in relationship to hydraulics and geomorphology, despite considerable phylogenetic and historical differences between the continents. Although not consistently supported, most of these results suggest that the trait composition of fluvial fish communities is relatively stable across large spatial scales.
Furthermore, the BTI approach was resistant to large-scale biogeographical variation in marine benthic invertebrates (Bremner et al., 2003b) and forest birds (Hausner et al., 2003). In contrast, different taxonomic groups may vary more in their trait responses to natural environmental characteristics in fluvial floodplains than in running waters (Henle et al., 2006). Overall, it would be interesting to assess potential system-specific differences in the large-scale variation of trait patterns through a comparative approach.
In summary, research in the last decade considerably changed our (at least of the authors of this review) perception of the action of filters for lotic invertebrate traits. Starting from the idea that temporal and spatial variability of natural stream habitats act as filter for invertebrate traits and thus cause clear trait responses, we are now convinced that running water habitats are physically so harsh that natural invertebrate communities scarcely differ in their trait composition across large spatial units. Consequently, reference conditions using the BTI approach and lotic invertebrates (and perhaps other organisms) could be defined for large legislative units.