Species traits predict assembly of mayfly and stonefly communities along pH gradients
Much recent ecological research has centred on the interrelations between species diversity and ecological processes. In the present study, I show how species traits may aid in comprehending ecology by studying the link between an environmental variable and functional traits. I examined the composition of species traits with a theoretically underpinned relationship to ecological processes along a pH gradient. I focused on body size, reproductive output, life cycle length and feeding habit of mayflies and stoneflies. In mayfly assemblages, I found smaller body size, greater reproductive output, faster life cycles and a larger proportion of gathering collectors and scrapers with increasing pH. In stonefly assemblages, I found smaller body size, greater reproductive output and faster life cycles at sites with a history of long-term natural acidification, but no clear trends in feeding habits and in most traits where acidification is anthropogenic. The results suggest that mayflies and stoneflies exhibit different ecological functions following different ecological strategies. Mayflies follow an opportunistic strategy relative to stoneflies, likely facilitating high rates of ecological processes with respect to the autotrophic resource base at neutral sites. Relative to mayflies, stoneflies follow an equilibrium strategy contributing to ecological functioning in heterotrophic ecosystems and likely maintaining heterotrophic processes despite the erosion of species diversity in response to acidification. The rules governing an ecological community may be more readily revealed by studying the distribution of species traits instead of species diversity; by studying traits, we are likely to improve our understanding of the workings of ecological communities.
KeywordsBiodiversity Ecological functioning Ephemeroptera Plecoptera Trait diversity
Traditionally, researchers have tried to comprehend the workings of ecosystems in two different ways (Loreau 2010): Community ecologists have attempted to unravel the functioning of ecosystems by studying ecological structure but solely describing the species composition of an assemblage is insufficient, if we want to comprehend the functioning of an ecosystem and understand how ecological processes, including production, consumption rates and decomposition, are controlled (Cummins 1974; McGill et al. 2006; Loreau 2010). Ecosystem ecologists, on the other hand, have tried to disentangle the mechanisms by focusing on ecological processes at the expense of structural detail and ignoring the idiosyncratic nature of species (Loreau 2010). However, the mechanics of an ecosystem may be more readily revealed by examining the trait composition of a community, as the trait that is manifested in a particular species, rather than the species identity, mediates a particular ecological process (McGill et al. 2006). In addition, trait composition appears to be structurally more stable than species composition (Fukami et al. 2005; McGill et al. 2006; Bonada et al. 2007; Schmera et al. 2009). The larger similarity in trait than species composition among communities may reflect similarities in the traits between species, i.e. some, often related, species may share the same traits, whereas others may have developed different traits (Felsenstein 1985; Bonada et al. 2007; Cadotte et al. 2008). Hence, the smaller variation in the trait than species composition between communities and the intimate bond between trait and function may aid in revealing links and relationships that have remained unnoticed or poorly understood (Fukami et al. 2005; McGill et al. 2006).
In freshwater ecosystems, macroinvertebrates have been identified as a functionally vital component stimulating a number of studies on their functional traits (Dolédec et al. 1999; Bonada et al. 2007; Statzner et al. 2007). However, a large proportion of the previous work has focused on traits in general regardless of whether and how they relate to ecological processes (e.g. Lamouroux et al. 2004; Bonada et al. 2007; Heino 2008; but see Schmera et al. 2009). In addition, in many studies, a traditional approach was chosen by selecting or developing trait diversity metrics––for instance, metrics that are related to Shannon’s diversity index, or measures of community (dis)similarity––to explore the effects that different variables have on trait diversity (e.g. Heino 2008; Schmera et al. 2009) rather than directly studying the effects that the environmental variables have on the magnitude of the traits or the proportional composition of the trait categories (cf. Lamouroux et al. 2004).
Much previous research has centred on the effects of pH on freshwater species diversity (Otto and Svensson 1983; Townsend et al. 1983; Rosemond et al. 1992), but a lot of the variation in ecological processes along an acidity gradient remains unaccounted for. For instance, it is still unclear why decomposition is impaired by low pH under some circumstances but not under others (Petrin et al. 2008a). Concurrently, little is known about how pH affects the composition of species traits that may control ecological processes (e.g. Ledger and Hildrew 2000, 2005). Furthermore, the acidic sites in most of the previous studies were selected to address the effects of anthropogenic acidification (Otto and Svensson 1983; Townsend et al. 1983; Rosemond et al. 1992), whereas few studies have included naturally acidic locations (Collier et al. 1990; Dangles et al. 2004; Petrin et al. 2007a, 2008b). Some studies suggest evolved adaptation to naturally low pH levels that may cause different functional trait responses than anthropogenic acidity (cf. Collier et al. 1990; Dangles et al. 2004; Petrin et al. 2007a, 2008b). None of these latter studies have included data on changes in trait composition along the pH gradients.
In the present study, firstly, I examined how pH affects species traits that are characterised by a theoretically underpinned relationship to ecological processes. Traits with an unclear relationship to ecological processes, such as body form, dispersal, locomotion, aquatic stages, resistant life stages and where and how eggs and egg masses are deposited, are discarded from the analyses. Secondly, I accounted for the differential effects of a pH gradient with a history of long-term natural acidification compared to a pH gradient characterised by recent anthropogenic acidification. Therefore, thirdly, I examined how the traits specifically varied along the environmental gradients rather than analysing a metric of trait richness or diversity that, by ignoring trait identity, may be difficult to interpret functionally.
Materials and methods
I studied the distribution of traits that affect ecological processes, such as consumption rates, as these traits are most likely to link ecological structure to functioning. The traits include body size, fecundity, voltinism and feeding habit. Generally, the metabolic rate of the whole organism scales as ¾ power of its body mass (Brown et al. 2004). Hence, given a constant population biomass, larger organisms tend to be more efficient requiring less food per unit biomass, whereas smaller organisms, in need of more energy per unit biomass, tend to interact more strongly with their resource base (Brown et al. 2004). Fecundity is a proxy of the potential population growth describing how quickly the abundance of an organism may change when the supply of resources is unlimited. Abundance may then greatly influence ecological functioning. The potential population growth rate––the change in the number of individuals over time––may be approximated as the product between fecundity and voltinism, the number of generations per year (Clifford 1982). Lastly, the feeding habit directly affects the trophic position of an organism and its relationships to other components of the community through the type and size of food items consumed (Cummins 1973; Merritt et al. 2008).
I assembled data on the maximum and mean body size, fecundity, voltinism and feeding habits of Swedish mayfly and stonefly species from the literature (Online Resource 1). The biology and ecology of mayflies and stoneflies have been intensively studied in Scandinavia during the past decades and are therefore well understood. In contrast, other diverse aquatic invertebrate taxa, for instance caddisflies, have received less attention, both ecologically and taxonomically. To facilitate the interpretation of the results, I preferred focusing on the better researched mayflies and stoneflies. Hence, other taxa were disregarded in the present study.
Where multiple trait measurements were available, I calculated the average value. Fecundity measurements were unavailable for 43% of the species. For the remaining species, the reported values are based on one or several publications. The authors variously reported fecundity values either individually for several females, or as an average, species-specific value. When calculating species-specific average fecundity values for the present study, I treated each dataset reported in the literature as one replicate. The life cycle categories included bivoltine, univoltine, semivoltine and merovoltine, meaning that the organisms completed two, one, half and less than half a generation per year (Clifford 1982). Flexible life cycles have been documented for 29% of the studied species, and they were therefore classified in multiple categories. None of the stoneflies were bivoltine. The feeding habit categories included filtering collectors, gathering collectors, scrapers, shredders and predators (Cummins 1973; Merritt et al. 2008). Collectors consume fine detrital particles and microscopic algae, scrapers feed on periphyton largely consisting of algae and microbes, shredders process coarse particulate organic matter, and predators capture living prey (Cummins 1973). Feeding habits were flexible in 71% of the species. They were classified in multiple categories. None of the stoneflies were filtering collectors. The present approach disregards intra-specific variation in trait values among sites due to limitations in the available data. Hence, I was unable to examine whether intra-specific variation among sites may have affected the observed ecological patterns.
I analysed data from two Swedish surveys of 700 streams in 1995 and 2000 (Wilander et al. 1998, 2003). The databases comprised species abundances for macroinvertebrates, including mayflies and stoneflies, and hydrochemical measurements, including pH, that emerged as the major chemical variable directly controlling species composition (Sandin 2003). The mean pH values changed by 0.1–0.2 pH units across latitudes and longitudes suggesting that pH varied mainly within rather than between regions. At each stream, riffles were sampled once during autumn following Swedish standard procedures (Wilander et al. 1998, 2003; Sandin 2003). Macroinvertebrates were sampled using a kick net (500 μm mesh, 1 m2 sampling area, 1 min sampling time), preserved in the field and sorted and identified by professional taxonomists at accredited and inter-calibrated laboratories (Wilander et al. 1998, 2003; Sandin 2003). The hydrochemical analyses were done at the Department of Environmental Assessment, Swedish University of Agricultural Sciences, Uppsala, following international standards (Wilander et al. 1998, 2003).
Previous research indicated that the hydrochemistry and ecology differ between streams in south-western and northern Sweden. The differences reflect variation in hydrochemical and biogeochemical processes (Laudon and Bishop 1999, 2002; Bishop et al. 2000), dissimilarities in acid deposition (Warfvinge and Bertills 1999), disparities in natural acidification following the past glaciations across the Scandinavian peninsula (Renberg et al. 1993; Korsman 1999) and likely diverging biogeographic and evolutionary processes (Petrin et al. 2007a, 2008b). To address the effects of the history and causes of acidification, the dataset was divided into northern Sweden, comprising the northern half of the country, and southern Sweden, comprising the south-west (see Petrin et al. 2007a for a map illustrating the locations of the regions). In addition, limed sites and sites affected by agricultural activities were removed from the dataset (see Petrin et al. 2007a for further details). The remaining dataset comprised 1,008 different sites.
For data analysis, I employed linear mixed models with generalised least squares (GLS) and multinomial logistic regression models. Separate mixed models were fitted for the dependent variables maximum body size, mean body size, fecundity and potential reproductive rate. For each assemblage, I first calculated the average of each dependent variable using the species-specific values weighted by the respective species abundances. To account for unequal relatedness between species and hence statistical non-independence between taxa (Felsenstein 1985; Harvey 1996), I employed a phylogenetic GLS method (PGLS) when calculating the means of the dependent variables for the different assemblages (Garland et al. 2005). The evolutionary relationships among the taxa were derived from published phylogenies (e.g. Zwick 2000; Ogden et al. 2009). Disregarding statistical non-independence between taxa from the statistical analyses changed the effect sizes by up to 21%, on average by 10%. I therefore report phylogenetically corrected model results. For the species lacking fecundity data, fecundity was set equal to the estimate for the next higher taxonomic level. For instance, no fecundity data was available for Amphinemura borealis. Hence, the fecundity of A. borealis was set equal to the fecundity of Amphinemura that was estimated as the average of A. standfussi and A. sulcicollis. The potential reproductive rate was calculated as the product between fecundity and voltinism. I then used mixed models to study the effects of pH and region on the dependent variables across assemblages. In these models, pH constituted a covariate, region a fixed factor with two levels (northern and southern Sweden) and study year a random factor with two levels (1995 and 2000). I assessed the significance of the variables pH, region and their interaction with a conditional F test (Pinheiro and Bates 2000). Similarly to the linear mixed models, multinomial logistic regression models were fitted to study the effects of pH and region on the dependent variables voltinism and feeding habit. For each assemblage, I calculated the proportions of the different voltinism and feeding habit categories again, using PGLS to account for differential relatedness among species as disregarding statistical non-independence between taxa changed the model results by up to 79%, on average by 8% (Garland et al. 2005). To avoid repeatedly counting individuals with flexible life cycles and feeding habits, I weighted the individuals’ contributions to each category by the inverse of the number of categories recorded for the respective species. I analysed the dependent variables voltinism and feeding habit separately and assessed the significance of the independent variables and their interaction with log likelihood ratio tests (Pinheiro and Bates 2000; Venables and Ripley 2002). The mayfly and stonefly data were analysed separately for all dependent variables.
I also included a spherical model of the semi-variogram by implementing the GLS method (S-GLS) for the dependent variables of maximum body size, mean body size, fecundity and potential reproductive rate to account for spatial autocorrelation (Beale et al. 2010). Due to the lack of multinomial logistic regression models that allow for specifying a spatial autocorrelation structure, I fitted surrogate GLS models to study the effects of pH and region on the dependent variables voltinism and feeding habit and at the same time account for spatial autocorrelation (cf. Venables and Ripley 2002; Beale et al. 2010). Comparison of the models that accounted for spatial autocorrelation with the models that disregarded spatial autocorrelation revealed very similar trends and suggested that spatial autocorrelation affected the results only marginally. I therefore only report the more parsimonious models disregarding spatial autocorrelation.
To examine whether physicochemical variables may have confounded the effects of pH, I employed redundancy analyses (RDA; Jongman et al. 1995). RDA is similar to canonical correspondence analysis (CCA) relating a set of response variables (species) to a set of environmental variables. Hence, RDA allows the assessment of the proportion of the variation in the response variables that is explained by environmental variables. However, the assumptions between RDA and CCA differ; and as preliminary analyses suggested that the response curves were linear, rather than unimodal, I preferred RDA to CCA. Results from RDA are presented and interpreted similarly to results from CCA. In the present study, I constrained the ordinations to a set of 20 environmental variables including pH, study region, water temperature and different chemical variables. The complete set of environmental variables was available for 427 sites for mayflies and for 421 sites for stoneflies. As response variables, I used the trait data. I employed an ANOVA-like permutation test with the ratio of the constrained and unconstrained total inertia (F) as test statistic to assess the significance of the complete set of the constraints (Oksanen et al. 2008). I separately analysed the data for mayflies and stoneflies. pH emerged as the major environmental variable affecting trait composition (Online Resource 2). Further important predictors were the acidity related variables absorbance (at 254 nm, a measure of the dissolved organic carbon content), total organic carbon content and phosphorous content. Phosphorous content is correlated with the dissolved organic carbon content in Sweden, a major variable controlling pH (Laudon and Bishop 1999, 2002; Bishop et al. 2000; Fölster et al. 2004). Hence, trait composition was largely related to pH and other acidity-related variables.
Most of the study sites were characterised by pH values ranging from 5.5 to 8.0. To examine whether the results may reflect the influence of outliers beyond either side of that range, I repeated all analyses limiting the data set to sites with pH levels from 5.5 to 8.0. Most results of the new analyses were similar to the original ones using the complete dataset, and I therefore only report the new results based on the limited dataset when the analyses indicated that the original results may not be robust.
Where macroinvertebrate abundances were overall lower, the variation in the trait composition between sites would be larger, which would increase uncertainty masking any ecological patterns rather than generating spurious results. Hence, the analyses are conservative and likely to be robust with respect to potentially low macroinvertebrate abundances.
The statistical models were fitted using the software package R 2.9.2 including the nlme, nnet and vegan packages (Pinheiro and Bates 2000; Venables and Ripley 2002; Oksanen et al. 2008; Pinheiro et al. 2009; R Development Core Team 2009). All tests were performed at a probability level of 5% for type I errors.
Effect of pH and region on the maximum body size, mean body size, fecundity, potential reproductive rate, voltinism and feeding habit of mayfly and stonefly assemblages
F1,739 = 17.6
F1,694 = 29.4
F1,739 = 20.6
F1,694 = 16.2
pH × region
F1,739 = 0.5
F1,694 = 5.4
F1,739 = 22.7
F1,694 = 30.1
F1,739 = 22.5
F1,694 = 19.0
pH × region
F1,739 = 0.5
F1,694 = 4.8
F1,739 = 72.0
F1,694 = 18.2
F1,739 = 1.2
F1,694 = 0.7
pH × region
F1,739 < 0.1
F1,694 = 8.3
Potential reproductive ratea
F1,739 = 138.9
F1,694 = 11.7
F1,739 = 19.1
F1,694 = 0.2
pH × region
F1,739 < 0.1
F1,694 = 12.1
LR = 9.6
LR = 0.9
LR = 5.6
LR = 1.5
pH × region
LR = 0.4
LR = 6.0
LR = 28.9
LR = 7.3
LR = 12.8
LR = 0.5
pH × region
LR = 2.1
LR = 2.2
Fecundity and potential reproductive rate of mayfly assemblages increased with increasing pH by 296 and 412 eggs per unit of pH, respectively (Table 1; Fig. 2), corresponding to an average effect size of 19 and 24% for assemblages in the circumneutral range. The reproductive output was slightly larger in northern than southern Sweden. In stonefly assemblages, the reproductive output decreased by 149 eggs per unit of pH in the north, but remained similar along the whole acidity gradient in southern Sweden (Table 1; Fig. 3). The average effect sizes equalled 29 and 3%, respectively, for assemblages at circumneutral sites.
Shredders constituted a relatively large proportion of mayflies at low pH values in both regions, while gathering collectors and scrapers dominated at higher pH levels (Table 1; Fig. 4; Online Resource 3). Filtering collectors and predators were only recorded at neutral sites in the south albeit at small proportions. There were no clear regional differences in the patterns. In stonefly assemblages, no clear patterns were revealed except for a trend towards different proportions of feeding habits with changing pH (Table 1; Fig. 5; Online Resource 4). However, restricting the analysis to sites with pH levels ranging from 5.5 to 8.0 suggests no effect of pH (LR = 6.0, p = 0.111).
For mayflies, the results were in agreement with the expectations. The mayflies constituting assemblages at acidic and presumably less productive sites were slightly larger than those at neutral, more productive localities possibly reflecting differences in metabolic efficiency (cf. Otto and Svensson 1983; Maurice et al. 1987; Meegan and Perry 1996; Brown et al. 2004). The reproductive output of mayfly assemblages and the proportion of fast life cycles increased with increasing pH as predicted (cf. MacArthur and Wilson 1967; Roughgarden 1971; Pianka 1972). Mayfly assemblages at acidic sites also comprised a larger proportion of shredders relying on allochthonous organic material (Cummins 1973; Cummins and Klug 1979; Wallace et al. 1997), while the assemblages at neutral sites constituted a large proportion of gathering collectors and scrapers depending more strongly, though not exclusively, on autochthonous primary production (Cummins 1973; Cummins and Klug 1979). Low macroinvertebrate abundances at some sites should reduce resource scarcity relaxing the expected relationships. Hence, finding relationships with respect to pH suggests robustness of the results. Similarly, the flexible classification of mayflies in different life cycle and feeding habit categories is conservative as it may mask ecological patterns (Rawer-Jost et al. 2000).
Albeit in disagreement with the expectations, the results for stoneflies were revealing. Whereas the trait composition of mayfly assemblages changed along the pH gradient in the same way in both regions, in stoneflies the outcome for reproductive output and voltinism depended on the region. For changes in body size and feeding habit along the pH gradient, no unequivocal regional differences were detected. Most abundant stoneflies are shredders, especially at acidic sites, although many exhibit flexible feeding habits alternatively functioning as gathering collectors or scrapers (Cummins 1973; Hynes 1976; Lillehammer 1988). Hence, except for predators, many stoneflies are independent of autochthonous, primary production. Instead, they may rely on the influx of allochthonous organic material that should be largely unrelated to pH (Cummins 1973; Cummins and Klug 1979; Wallace et al. 1997). The allocation of numerically dominating stoneflies to multiple categories (Cummins 1973; Hynes 1976; Lillehammer 1988), their independence of autochthonous primary production (Cummins 1973; Cummins and Klug 1979; Wallace et al. 1997) and the observation that many stoneflies are acid-tolerant (Brinck 1949; Lillehammer 1988; Malmqvist 1999) should together contribute to masking any patterns implying limited systematic change in the composition of feeding habits along an acidity gradient.
Contrary to the expectations, stonefly body size was larger at neutral sites in northern Sweden, and the reproductive output was lower. However, macroinvertebrate abundance often is larger at neutral sites (e.g. Townsend et al. 1983; Rosemond et al. 1992). Thus, competition for resources should also be larger there, and a larger body size, entailing a more efficient metabolism and hence greater competitiveness (cf. Brown et al. 2004), would prove advantageous. This would also explain the lower reproductive output and slower life cycles of stoneflies at neutral sites, as competitiveness tends to inversely covary with fecundity and voltinism (cf. MacArthur and Wilson 1967; Roughgarden 1971; Pianka 1972). In the south, where tolerance to anthropogenic acidity is less widespread (Petrin et al. 2007a, 2008b), extinction due to low pH levels may have been random rather than having covaried with particular traits. A random order of extinction would explain the absence of a clear relationship between body size or reproductive output and pH in the south. However, in the north, the distribution of stoneflies in response to pH may have covaried with functional traits as at least some of the stonefly taxa are likely adapted to acidic water in northern Sweden (Dangles et al. 2004; Petrin et al. 2007a, 2008b). The finding for voltinism in southern Sweden, that life cycles are faster at neutral sites, corresponded to the expectation.
The available data did not allow examining whether intra-specific variation in the trait levels among sites may have contributed to the reported results. However, a comprehensive study of plant traits revealed that changes in trait means along environmental gradients largely reflect species turnover rather than intra-specific variation (Cornwell and Ackerly 2009) and may suggest that the patterns reported in the present paper are largely unaffected by intra-specific variation. The insects’ relative mobility should also favour species turnover rather than intra-specific variation as the mechanism controlling trait levels. Although plausible, the suggested mechanism that species turnover affects trait levels more strongly than intra-specific variation remains to be demonstrated for aquatic insects.
Disregarding differential relatedness and hence statistical non-independence between taxa may result in reduced or inflated effect size estimates. The differences between the phylogenetically corrected and uncorrected results were on average small in the present study, but larger differences between the results do occur and may affect the conclusions (Felsenstein 1985; Harvey 1996). We can only be confident that the results reflect underlying ecological principles if the findings are supported by data comprising several unrelated taxa that represent different evolutionary lineages. Closely related organisms, due to the inheritance of similar traits from a common ancestor, may constitute little more than one replicate, independent observation. If the pattern is driven by closely related taxa, the result may be spurious (Felsenstein 1985; Harvey 1996). Phylogenetic corrections should thus be especially important when studying assemblages that include many related taxa such as the often numerically dominant and relatively diverse mayfly Baetis spp. in fluvial ecosystems.
At neutral sites, the larger proportion of herbivorous mayflies, faster life histories, higher reproductive output and smaller body size all would likely contribute to higher rates of ecological processes and hence a stronger interaction with the autotrophic resource base. Most stonefly assemblages in northern Sweden––where the stoneflies may be adapted to conditions of low pH (Petrin et al. 2007a, 2008b)––comprise smaller organisms, a higher reproductive output and probably also faster life cycles at acidic sites. This, in addition to the stoneflies’ reliance on allochthonous detritus (Cummins 1973; Cummins and Klug 1979; Wallace et al. 1997), should contribute to relatively higher rates of ecological processes and therefore a stronger interaction with the heterotrophic resource base at acidic sites (Petrin et al. 2007b). The attributes of mayfly and stonefly ecology discussed here may explain why ecological functioning in heterotrophic ecosystems is maintained at low pH levels despite reductions in species diversity (Petrin et al. 2007b, 2008a, b). Furthermore, the finding that the trait composition in mayfly assemblages varied in the same way in both northern and southern Sweden is in agreement with the widely accepted view that mayflies are generally acid-sensitive regardless of the causes of acidity (Otto and Svensson 1983; Townsend et al. 1983; Rosemond et al. 1992). In contrast, the trait composition of stonefly assemblages along a pH gradient changed in different ways in the two regions in accordance with the hypothesis of an evolved tolerance to low pH levels in stoneflies that inhabit naturally acidic freshwater ecosystems including those in northern Sweden (Dangles et al. 2004; Petrin et al. 2007a, 2008b). Finally, the results suggest that mayflies and stoneflies follow different ecological strategies in addition to having different ecological functions. Mayflies exhibited relatively fast life cycles, a high reproductive output and a tendency towards exploiting abundant food resources, suggesting they primarily followed an opportunistic strategy (Brittain 1982; Winemiller and Rose 1992). Stoneflies, characterised by slower life cycles, a lower reproductive output and a tendency towards exploiting partly limited food resources, appeared to rely more on an equilibrium strategy relative to mayflies (Hynes 1976; Winemiller and Rose 1992; Wallace et al. 1997).
In conclusion, where researchers have failed to derive more general laws in community ecology, species traits may comprise the component where the laws that rule an ecological community become evident. Studying species traits of putatively functional importance may provide novel insights into the workings of an ecosystem because species traits provide information complementary to measurements of species diversity and ecological functioning (McGill et al. 2006). When neglecting species traits, data comprising lists of species identities and abundances may indeed remain silent about the species’ functional significance even when combined with data on the rates of ecological processes. For instance, drawing inferences on the differential functional importance of mayflies and stoneflies along a pH gradient may be difficult if not impossible when solely based on studies of species diversity and ecological functioning (cf. Dangles et al. 2004; Petrin et al. 2007a, 2008a, b). Whether the goal of a research project is to test hypotheses and ecological theory (Townsend et al. 1997; Fukami et al. 2005; McGill et al. 2006), assess anthropogenic impacts on biodiversity and ecological functioning (Haybach et al. 2004; Bracken et al. 2008), monitor environmental and land use change (Fortunel et al. 2009), or manage environmental resources and ecosystem services (Merritt et al. 2002; Díaz et al. 2007), species traits will improve our comprehension of the principles that govern life-sustaining ecosystems.
I thank Edwige Bellier, Núria Bonada, John Edward Brittain, Grégoire Certain, Ola Diserud, Richard Hedger, Ingeborg Palm Helland, Frank Johansson, Odd Terje Sandlund, Ann Kristin Schartau and Maxim Teichert for discussing data analysis, the results and data presentation. Joel Trexler’s, the anonymous referees’ and the editor’s helpful comments are gratefully acknowledged. This paper is a contribution to the BIOCLASS-FRESH project (VANN: Biological indicators for classification of ecological status in freshwater, 184002) funded by the Norwegian Research Council (the MILJØ2015 programme), the Norwegian Energy Directorate (NVE), the Climate and Pollution Agency (KLIF, formerly SFT) and the Norwegian Directorate for Nature Management (DN).
- Bonada N, Dolédec S, Statzner B (2007) Taxonomic and biological trait differences of stream macroinvertebrate communities between mediterranean and temperate regions: implications for future climatic scenarios. Global Change Biol 13:1658–1671. doi:10.1111/j.1365-2486.2007.01375.x CrossRefGoogle Scholar
- Brinck P (1949) Studies on Swedish Stoneflies [Plecoptera]. In: Opuscula Entomologica Supplementum XI. Entomologiska Sällskapet i Lund, LundGoogle Scholar
- Clifford HF (1982) Life cycles of mayflies (Ephemeroptera), with special reference to voltinism. Quaest Entomol 18:15–90Google Scholar
- Fölster J, Sandin L, Wallin M (2004) A suggestion to a typology for Swedish inland surface waters according to the EU Water Framework Directive. Department of Environmental Assessment, Swedish University of Agricultural Sciences, report 13, Uppsala, SwedenGoogle Scholar
- Jongman RHG, ter Braak CJF, van Tongeren OFR (eds) (1995) Data analysis in community and landscape ecology. Cambridge University Press, CambridgeGoogle Scholar
- Lillehammer A (1988) Stoneflies (Plecoptera) of Fennoscandia and Denmark. Brill, LeidenGoogle Scholar
- MacArthur HH, Wilson EO (1967) The theory of island biogeography. Princeton University Press, PrincetonGoogle Scholar
- Malmqvist B (1999) Lotic stoneflies (Plecoptera) in northern Sweden: patterns in species richness and assemblage structure. In: Friberg N, Carl JD (eds) Biodiversity in Benthic Ecology, NERI Technical Report No. 266. National Environmental Research Institute, Denmark, Silkeborg, DenmarkGoogle Scholar
- Merritt RW, Cummins KW, Berg MB (eds) (2008) An introduction to the aquatic insects of North America, 4th edn. Kendall/Hunt, DubuqueGoogle Scholar
- Otto C, Svensson BS (1983) Properties of acid brown water streams in South Sweden. Arch Hydrobiol 99:15–36Google Scholar
- Pinheiro J, Bates D, DebRoy S, Sarkar D, the R Core Team (2009) nlme: Linear and nonlinear mixed effects models. R package version 3:1–93Google Scholar
- Renberg I, Korsman T, Anderson NJ (1993) A temporal perspective of lake acidification in Sweden. Ambio 22:264–271Google Scholar
- R Development Core Team (2009) R: A language and environment for statistical computing. R Foundation for Statistical Computing, http://www.R-project.org/, Vienna, Austria
- Venables WN, Ripley BD (2002) Modern applied statistics with S, 4th edn. Springer, New YorkGoogle Scholar
- Warfvinge P, Bertills U (1999) Recovery from acidification in the natural environment: present knowledge and future scenarios. Swedish Environmental Protection Agency, report 5034, StockholmGoogle Scholar
- Wilander A, Johnson RK, Goedkoop W, Lundin L (1998) Riksinventering 1995––En synoptisk studie av vattenkemi och bottenfauna i svenska sjöar och vattendrag. Swedish Environmental Protection Agency, report 4813, UppsalaGoogle Scholar
- Wilander A, Johnson RK, Goedkoop W (2003) Riksinventering 2000––En synoptisk studie av vattenkemi och bottenfauna i svenska sjöar och vattendrag. Department of Environmental Assessment, University of Agricultural Sciences, report 2003:1, UppsalaGoogle Scholar