Biotic indices

Thieneman (1914) attributed the birth of biological monitoring to Kolkwitz and Marsson (1902; Fig. 1). These pioneers suggested the use of flora and fauna (i.e., the entire aquatic community) to assess freshwater health in the “Saprobien system” (Saprobien Index [SI]) (see Appendix for abbreviations for bioassessment approaches). This approach is a cornerstone in the development of biological indicators of human alterations of freshwater systems throughout the world, although research on the topic started earlier (e.g., Kolenati 1848, Cohn 1853; Fig. 1).

From this cornerstone arose an array of biotic indices based on formulae or standard tables (e.g., Trent Biotic Index [TBI], Woodiwiss 1964 [Fig. 1]; Tuffery and Verneaux 1968, Andersen et al. 1984, Lang et al. 1995) or on scoring systems, such as Biological Monitoring Working Party (BMWP; Hellawell 1978; Fig. 1) and Average Score Per Taxon (ASPT; Armitage et al. 1983; Fig. 1). Use of such Biotic Indices (BIs) for monitoring ecological conditions of freshwater systems spread all over the world (Fig. 2). Most were based on invertebrates (e.g., Sharma and Moog 1998), but diatoms also were used (e.g., Descy 1979, Prygiel and Coste 1993; Fig. 1). Haury et al. (2006; Fig. 1) designed the Macrophyte Biological Index for Rivers (MBIR), a biotic index based on the concept of trophic state (Thienemann 1925; Fig. 1) that uses macrophyte communities to assess river water quality. Today, most workers in European countries use BIs in their national surveys (e.g., Iversen et al. 2000), and workers in countries of other continents have designed their own BIs (Sharma and Moog 1998; Fig. 2). For example, Stark (1993) derived the New Zealand Macroinvertebrate Community Index (MCI) from the BMWP scoring system, and Chutter (1972; Fig. 1) created the South African Chutter Biotic Index (CBI) from SI (Fig. 2).

The Water Framework Directive (WFD; Fig. 1), which is largely based on the reference condition approach (see Hawkins et al. 2010), arose from a growing need for convergence in environmental policies across the European Union (European Commission 2000). WFD provides a framework to harmonize traditional assessment protocols of individual countries in Europe through intercalibration so that various national assessment systems have comparable integrity scales (e.g., Furse et al. 2006 [Fig. 1], Hawkins 2006). For example, Sandin and Hering (2004) intercalibrated national indices at the European level and showed that ASPT accurately identified good quality sites for all river types (with one exception), whereas SI misclassified many sites. They concluded that the definition of quality classes at the European scale must consider natural differences between river types and must correct for national differences in the perception of ecological quality. The harmonization approach compares national biological assessment procedures with external benchmarking data sets, such as Assessment system for the ecological Quality of streams and rivers throughout Europe using benthic Macroinvertebrate (AQEM) data (Hering et al. 2004b; Fig. 1). National quality-class boundaries are then redefined until differences between national and benchmarking samples become statistically insignificant (Buffagni et al. 2006, Buffagni and Furse 2006).

North American freshwater ecologists were reluctant to use the SI system because it was based on European species, assessed only one stressor (organic pollution) among multiple potential stressors, and relied on the indicator species concept, which overlooks the natural causes of population variations (Cairns and Pratt 1993). This a priori rejection of the indicator species concept probably delayed the development of BIs in North America (Cairns and Pratt 1993), where earlier suggestions for use of BIs in the US (Richardson 1929; Fig. 1) had been ignored (but see Hilsenhoff 1977; Fig. 1).

J-NABS published the Family Biotic Index (FBI, Hilsenhoff 1988; Fig. 1) and the BI for southeastern US streams (US-BI, Lenat 1993; Fig. 1), which strongly depended on organisms' tolerances and indicator values. BIs have been addressed in a significant proportion of J-NABS papers (∼12%; Table 1), but workers in the US preferred to incorporate BIs as a component of multimetric approaches (Hawkins 2006).

Multivariate approach

The increasing calculation power of computers and related statistical developments for synthesizing taxonomic composition or taxonomic completeness led to predictive biological assessment tools in the 1980s. These developments were rooted in multivariate techniques, such as Principal Components Analysis (PCA; Pearson 1901) and Correspondence Analysis (CA; Hirschfeld 1935), that had been designed decades earlier and influenced freshwater research in the 1970s (e.g., Lévêque and Gaborit 1972, Ibanez and Seguin 1972). Subsequent developments in multivariate analyses included Canonical Correspondence Analysis (CCA; ter Braak 1986), which was widely used in ecological (Birks et al. 1996) and freshwater (ter Braak and Verdonschot 1995, Houghton 2007) sciences, and coinertia analysis (COIA), which was developed to relate freshwater species composition to environmental conditions (Dolédec and Chessel 1994) and has a wide range of applications (Dray et al. 2003).

Multivariate analyses are used widely for determining patterns of assemblage structure, inferring species–environment relationships, and assessing human impacts on ecosystems (Cao et al. 2002). Despite initial skepticism related to the complex outputs of multivariate analyses (Norris 1995), ∼30% of J-NABS bioassessment articles were based on the approach (a proportion comparable to that in FWB; Table 1). Some investigators consider these techniques exploratory and useful only in case of limited knowledge of a given ecological system (e.g., Fore et al. 1996; Fig. 1). Others regard them as useful tools in bioassessment because they can be used to determine the most important variables that influence community composition (e.g., Houghton 2007) and to describe and predict the effects of human disturbance on aquatic biota (e.g., Mykra et al. 2008). CCA or COIA currently are used to identify sites with different levels and types of pollution (e.g., Zamora-Munoz and Alba-Tercedor 1996, Truu et al. 2002, Hicham and Lofti 2007), assess the effects of land use on aquatic biota (e.g., Aznar et al. 2004, Kratzer et al. 2006, Couceiro et al. 2007), design new metrics or indices for diagnosing environmental stressors (Griffith et al. 2001, Haase and Nolte 2008), and validate metric performance (Prygiel et al. 1999).

A prominent example of the use of multivariate techniques in biological assessment is the River InVertebrate Prediction and Classification System (RIVPACS, Wright et al. 1984; Fig. 1), which uses reference conditions (Reynoldson et al. 1997 [Fig. 1], Stoddard et al. 2006, but see Chessman and Royal 2004 [Fig. 1]) to compare a predicted minimally disturbed macroinvertebrate assemblage with an observed assemblage (Hawkins et al. 2010). This comparison results in an observed/expected (O/E) ratio of a given biological variable (e.g., BMWP, ASPT), which lies on a scale from impacted (O/E < 0.50; Wright et al. 2000) to undisturbed (O/E  =  1.00) conditions.

RIVPACS was developed for streams and rivers in the UK, but the RIVPACS method has been promoted in the US (Hawkins et al. 2000), implemented in other countries (e.g., AUStralian RIVer Assessment Scheme [AUSRIVAS], Turak et al. 1999, Simpson and Norris 2000 [Fig. 1]; Assessment by Nearest Neighbor Analysis [ANNA], Linke et al. 2005 [Fig. 1]), or extended to lakes (BEnthic Assessment of SedimenT [BEAST], Reynoldson et al. 1995; Fig. 1) (Fig. 2). These methods rely on reference conditions, so their application to other regions requires redefinition of the reference condition for the new geographic area (e.g., Alba-Tercedor and Pujante 2000) because biases can occur in the predictions when test sites are not compared with reference sites of the same region (e.g., Yuan et al. 2008). Separating biological responses to natural and anthropogenic causes at greater spatial scales can be difficult with RIVPACS-like methods because of missing environmental variables in the redefinition of the reference condition, insufficient number of reference sites, and insufficient replication of tested ecological conditions (Hose et al. 2004, Hawkins 2006). Current RIVPACS-like models cannot identify specific stressors (e.g., acidification, logging; Bonada et al. 2006), although combining data from multiple types of organisms (e.g., benthic macroinvertebrates, fish, macrophytes, diatoms) could help solve this problem. This strategy corresponds to the notions of early freshwater ecologists that the entire aquatic biocenosis should be used when assessing ecosystem conditions (Kolkwitz and Marsson 1902, Patrick 1949 [Fig. 1], Hynes 1960). For example, in European rivers, benthic diatom assemblages have a greater sensitivity to eutrophication or organic pollution than do other assemblages (Hering et al. 2006b, Johnson et al. 2006b, Feio et al. 2007), a feature also observed in the Everglades where periphyton was the most sensitive group of organisms to P loading (McCormick et al. 1996). Furthermore, responses of various assemblages to specific stressors are influenced by stream type; e.g., benthic invertebrates were better suited for assessing hydromorphological degradation in smaller streams (Feio et al. 2007), whereas fish or macrophytes were equally suited (with invertebrates) in larger lowland streams (Hering et al. 2006b, Johnson et al. 2006b).

ter Braak and Juggins (1993) developed a weighted averaging partial least squares (WA-PLS) approach for predicting environmental conditions from biota (calibration). This technique has been used in paleolimnological research to reconstruct past environmental conditions from biotic remains (e.g., Gasse et al. 1995). This technique provides an indication of how present reference conditions match historical ones, and can be especially useful when depicting past nutrient loading. Paleolimnology is used primarily in lakes and is little mentioned in J-NABS (<1% of bioassessment articles; Table 1).

Multimetric approach

In his history of water law in North America, Karr (1991) indicated that early legislation (Refuse Act of 1899; Fig. 1) focused on growing problems of disease and oil pollution in navigable waters, whereas awareness of restoration and maintaining the biological integrity of the US waters started with the 1972 Water Pollution Control Amendment (WPCA, Clean Water Act; Fig. 1). Karr (1981; Fig. 1) initiated the first studies based on the multimetric index (MMI) approach in which attributes (metrics) of fish assemblages were used to evaluate human effects on a stream and its watershed. Metrics are biological measures representing aspects of the composition, function of assemblages, or other characteristics of individuals that should vary with increased human impact. The first fish Index of Biotic Integrity (IBI) considered 12 metrics, such as species richness and composition, trophic composition, abundance, and health condition. Metrics are rated relative to what would be expected at an undisturbed site in a stream of similar size in the same region. The sum of these ratings provides an IBI value that is regarded as an integrative and quantitative measure of local biological integrity (Karr 1991) or overall human impact. The original fish IBI was developed for streams in the US Midwest, but it has been adapted for use in other regions by modification or replacement of metrics (e.g., Miller et al. 1988, Steedman 1988).

The multimetric approach is used with lotic invertebrates throughout the US (e.g., Invertebrate Community Index [ICI], Ohio EPA 1987; Benthic Index of Biotic Integrity [B-IBI], Kerans and Karr 1994 [Fig. 1], Barbour et al. 1995), and has been adapted to periphyton (McCormick and O'Dell 1996, McCormick et al. 1996, Hill et al. 2000; Fig. 1) and diatom assemblages (Wang et al. 2005; Fig. 1). The Diatom IBI (D-IBI) is based primarily on community composition metrics (e.g., richness and composition, absolute or relative abundances of dominant taxa, tolerance/intolerance of taxa) and de-emphasizes individual-level measures (e.g., disease, tumors, fin damage, skeletal anomalies). The Periphyton Index of Biotic Integrity (P-IBI) includes periphyton-specific metrics of ecological function, such as % motile diatoms and phosphatase activity (Hill et al. 2000).

Many papers based on MMI developments have been published in J-NABS (∼13% of bioassessment articles; Table 1). Several of these papers, such as Barbour et al. (1996; Fig. 1), Fore et al. (1996), and McCormick et al. (1996) have been among the 100 most-cited J-NABS articles. MMI development provided a common biological approach to assessing the condition of most rivers in the US. By the middle 1990s, multimetric biological assessment was used in 42 US states and was under development in 6 states (Karr et al. 2000). However, regional differences in the fauna or flora and in the intensity of stress at human-impacted sites generally make comparison of MMIs difficult at large spatial scales (Hawkins 2006).

The multimetric approach spread from North America to Europe (Fig. 2) where it currently is under evaluation as a tool for assessing river conditions with invertebrates (e.g., Hering et al. 2004a, Johnson et al. 2006a, Feld and Hering 2007) and fish (Oberdorff and Hughes 1992, Oberdorff et al. 2002, Pont et al. 2006, Schmutz et al. 2007; Fig. 1). However, the approach is not easily transferable because metrics are selected based on their apparent sensitivity to region-specific stressors (e.g., Norris and Hawkins 2000). In contrast to indices used in the US, the European index covers the continent (Pont et al. 2006, Schmutz et al. 2007) rather than regions and includes functional metrics (e.g., reproduction and trophic guild for fish, Pont et al. 2006; feeding habits and ecological preferences for macroinvertebrates, Hering et al. 2006a). Hering et al. (2006a) used a method for designing MMIs that excluded numerically unstable metrics, used correlations between appropriate metrics and a selected anthropogenic impact gradient, and included only the most robust metrics (those with the best response to the anthropogenic impact gradient), given the need to balance the 4 types of metrics. The need for such a selection process was recently outlined by Stoddard et al. (2008) in a large-scale assessment using MMI. The scale of variation of metrics is defined according to reference conditions, and the MMI is generated by combining the selected metrics (e.g., 12 metrics for Böhmer et al. 2004), which represent a small proportion of the metrics tested (165 metrics were computed by Böhmer et al. 2004). This method reflects the strong empirical basis of many European metrics, and is in contrast with the use of theoretical or conceptual understanding of biologically important attributes to select metrics (e.g., Norris and Hawkins 2000, Hawkins 2006).

According to their proponents, multimetric approaches provide biologically interpretable handles (metrics) that better facilitate decision-making than do methods based on indices and that generate reliable indices of the overall impact of human activities (Barbour et al. 1999). However, the meaning of IBIs has been questioned (e.g., no predictability, no diagnostic power, no real reason for high or low index values; Suter 1993), as has the validity of summing heterogeneous metrics into a single measure of river condition (effects on one metric could be blurred by effects on other metrics; Suter 1993, Norris and Hawkins 2000).

Partialling out the effects of natural variation on assemblages should improve the ability of metrics to indicate human impacts on freshwater communities (Pan et al. 2004, Feld and Hering 2007). For example, Moya et al. (2007) elaborated models describing metric responses to natural gradients in the absence of any significant disturbance. They used the residuals of each metric model to select metrics that most effectively discriminated between reference and disturbed sites. In a last step, they converted residuals of selected metrics into probabilities and added the probabilities to obtain an MMI.

The discriminatory capacity of metrics and rates of misclassification of disturbed sites as reference sites (type II error) can vary according to the group of aquatic organisms used. For example, landuse intensity and nutrient enrichment were well detected (classification error < 15%) by metrics derived from diatom and benthic invertebrate assemblages. In contrast, metrics derived from fish and aquatic macrophyte assemblages performed better (lower type II error) as indicators of physical alterations than of nutrient enrichment (Johnson et al. 2006a).

Direct assessment of ecosystem processes (LLB and SP)

Ecosystem-level processes that are potentially useful in stream assessment include SP (e.g., Benke and Huryn 2010), sediment and community respiration, rate of nutrient removal, and LLB (Gessner and Chauvet 2002).

LLB is a central process in the exchange of matter between terrestrial and aquatic ecosystems (e.g., Sponseller and Benfield 2001, Lamberti et al. 2010). LLB coefficients (k values) represent rates of decomposition in an exponential decay model (Petersen and Cummins 1974; Fig. 1). Gessner and Chauvet (2002) suggested that k values from standardized leaf packs could be used in bioassessments as: 1) ratios of k values between impacted and reference sites, 2) absolute k values if no reference is available, and 3) ratios of k values between coarse- and fine-mesh bags to test the relative contribution of shredders and microorganisms to LLB.

LLB is affected by modifications of riparian and catchment vegetation, nutrient enrichment, and urbanization (e.g., Webster and Benfield 1986, Meyer et al. 2005). LLB rates decrease in response to mine pollution (Niyogi et al. 2001, Lecerf and Chauvet 2008) and acidic atmospheric deposition (Dangles et al. 2004, and references therein). When N and P concentrations are low, nutrient enrichment can increase LLB rates via increased fungal biomass and sporulation rates (Grattan and Suberkropp 2001; Fig. 1). However, high organic and inorganic nutrient concentrations can depress conidia production by aquatic hyphomycetes and increase macroinvertebrate density, which can accelerate LLB rates (Pascoal et al. 2003). More research on fungal decomposers is needed to better understand underlying LLB mechanisms (Pascoal et al. 2005).

Agricultural land use can have both positive (nutrients) and negative (sedimentation) effects on LLB (Niyogi et al. 2003). Hagen et al. (2006) found similar LLB rates across landuse categories from forested to heavily cultivated areas. Pascoal et al. (2005) failed to detect differences in LLB rates between a nutrient-enriched site and its upstream reference site, but impairment was obvious when rates were compared between fine- and coarse-mesh bags. These results suggest that metrics based on ratios of k values between fine- and coarse-mesh bags might be more suitable than metrics based on ratios of k values between reference and impacted sites.

One of the first papers showing a strong decrease in LLB in response to acidification was published in J-NABS (Mulholland et al. 1987; Fig. 1), but papers describing use of LLB rates as biological response to human disturbance are infrequent in J-NABS (4.4%, Table 1). For example, Grattan and Suberkropp (2001) is the most-cited (44 citations in the Web of Science® database at the Institute for Scientific Information [ISI] Web of Knowledge^SM site [ISI-Web®];  http://apps.isiknowledge.com/, checked in June 2008) of these articles but is not among the 100 most-cited J-NABS articles. Changing water chemistry (including nutrient loading) and land uses have been the main stressors addressed by LLB assessments, but recent research has addressed more subtle ecosystem modifications. For example, LLB has been used as an early warning indicator of alteration of riparian vegetation by invasive plant species, which indirectly affect instream communities (e.g., Lecerf et al. 2007). An innovative study published in J-NABS used LLB to assess the effect of global change on ecosystem processes (Rier et al. 2002; Fig. 1). This study showed that an increase of atmospheric CO₂ levels might increase litter recalcitrance to microbial activity and decrease LLB rates, which would decrease availability of C and N, thereby affecting food webs (Kominoski et al. 2007).

SP, the formation of heterotrophic biomass over time, is another important measure linking populations (biomass, growth rate, survivorship, development time) and communities to ecosystem-level processes (e.g., energy flow, Benke 1993, Benke and Huryn 2010). Buffagni and Comin (2000) suggested that calculating SP for dominant taxa in the different habitats should be an effective means to assess the ecological integrity of mountain streams. However, annual SP can vary naturally by a factor of ∼4 over short distances in lake outflows (Statzner 1987) or in forested streams in response to the relative composition of litter patches (Kobayashi and Kagaya 2004). Such variability would limit its application at larger scales (see Statzner and Lévêque 2007 for large-scale SP variability).

SP responds to ecosystem degradations, such as river regulation (Raddum and Fjellheim 1993), biological invasions (Statzner and Lévêque 2007), changing riparian forests (Whiles and Wallace 1997), logging (Kedzierski and Smock 2001), stream acidity (Thomsen and Friberg 2002), stream contamination (Carlisle and Clements 2003; Fig. 1), organic pollution (Statzner and Lévêque 2007), and multiple sources (Shieh et al. 2002 [Fig. 1], de Lange et al. 2004). Whole-stream annual macroinvertebrate production was greater in logged than in unlogged stream sections (Kedzierski and Smock 2001) and was lower before than after clearcutting based on changing energy resources (Stone and Wallace 1998). Such changes generally are associated with changes in FFG composition. Invertebrate SP was lower in streams draining pine monocultures than in undisturbed hardwood catchments (Whiles and Wallace 1997). Metal contamination reduced invertebrate abundances and SP (Carlisle and Clements 2003). In contrast, biological invasions or organic pollution caused an increase of SP (Statzner and Lévêque 2007). In cases of multiple sources of impairment, Shieh et al. (2002) attributed increased invertebrate SP to high nutrient availability and low predation pressure in urban and agricultural sites, whereas de Lange et al. (2004) suggested that moderate levels of sediment contamination scarcely affects benthic SP.

Carlisle and Clements (2005) measured several ecosystem functional responses (LLB, microbial respiration, and shredder SP) to metal contamination. The species most sensitive to metal contamination was a functionally dominant species. Thus, loss of a single species might impair ecosystem function. In contrast, de Lange et al. (2004) observed that contamination of sediment by trace metals and other chemicals affected invertebrate community structure, but SP of oligochaetes and chironomids was not correlated with the level of contamination.

Papers published in J-NABS have contributed significantly to our understanding of fundamental aspects of SP (e.g., Benke et al. 1999 with 200 ISI-Web® citations). However, only one paper published in J-NABS has applied SP to bioassessment (Carlisle and Clements 2003). This paper has received 21 citations and is not among the 100 most-cited J-NABS articles.

Indirect assessment of ecosystem processes based on invertebrate FFGs

A significant part of the annual LLB in streams is caused by invertebrate activities, e.g., >25% of LLB in a headwater stream could be attributed to shredder activities (Cuffney et al. 1990). Organic matter cycling and energy flow are basic processes associated with the functional integrity of lotic ecosystems (Vannote et al. 1980). Thus, benthic macroinvertebrates often are assigned to FFGs based on feeding mechanisms, energy sources, and food size (Cummins 1973).

Bioassessments with FFGs can be based on absolute or relative abundance (of individuals, biomass, or SP) of a single FFG. For example, Kedzierski and Smock (2001) reported higher SP of filtering and gathering collectors in a logged than in an unlogged stream section, whereas SP of other FFGs changed little with logging. In contrast, grazers dominated macroinvertebrate communities in catchments with recently logged forests, whereas the proportion of shredders increased with time since logging (Nislow and Lowe 2006). Urbanization and associated changes in stream quality also can affect FFGs. For example, proportions of collectors and gatherers increased, whereas proportions of filter-feeders, scrapers, and shredders decreased with watershed imperviousness associated with urbanization (Stepenuck et al. 2002). In another study, higher proportions of scrapers occurred at urban sites, whereas higher proportions of filter-feeders occurred at agricultural sites (Bacey and Spurlock 2007). Agricultural activities increase sediment deposition and silting, which can decrease absolute densities of all FFGs while favoring relative densities of gatherers (Rabeni et al. 2005) or filter-feeders (Dolédec et al. 2006; Fig. 1). Other stressor types are poorly detected by FFGs. For example, a strong salinity gradient caused only slight alterations of trophic structure (Piscart et al. 2005).

Ratios of abundances of ≥2 FFGs have been used as surrogates of ecosystem processes. For example, Merritt et al. (2002; Fig. 1) suggested using the ratio of total shredders to total collectors as a surrogate for the ratio of coarse particulate organic matter (CPOM) to fine particulate organic matter (FPOM) and the ratio of filtering collectors to gathering collectors as a surrogate for the ratio of suspended particulate organic matter (SPOM) to benthic particulate organic matter (BPOM). Paillex et al. (2007; Fig. 1) used such ratios as surrogates for top-down control of assemblage structure and habitat stability. Ratios declined along a gradient from disconnected to more connected sites in a river floodplain, a result that suggested the ratios might be useful for monitoring conditions in floodplain water bodies after connectivity between main and abandoned channels was restored (Paillex et al. 2007). Compin and Cereghino (2007) reported higher proportions of gathering collectors but lower proportions of shredders and filtering collectors in human-impacted landscapes than in natural ones.

Pavluk et al. (2000; Fig. 1) designed an index of trophic completeness (ITC), based on 12 trophic groups defined by combinations of diet composition, feeding mechanisms, food size, food-acquisition behavior, and energy transfer through prey ingestion. According to Pavluk et al. (2000), all 12 FFGs should occur in undisturbed rivers, and the absence of ≥1 FFGs indicates an effect of anthropogenic activities. Bij de Vaate and Pavluk (2004) confirmed that disturbances cause loss of specific trophic groups. They suggested the ITC might have large-scale applicability because of its independence from distance to source, geographical characteristics, and seasonality, but acknowledged that the ITC was limited in its ability to identify specific stressors. The approach requires more testing because the current reference condition is defined from samples of stony substrate in a few unpolluted river sites.

FFGs have been useful for bioassessments in the northern hemisphere, but generally have failed to detect anthropogenic effects in the southern hemisphere (Palmer et al. 1996, Thorne and Williams 1997). In Argentina, most FFGs (except shredders) occurred along the entire river despite modifications of the hydrological regime caused by an artificial reservoir (Miserendino 2007). FFG-based metrics did not distinguish reference from disturbed sites in Bolivian rivers (Moya et al. 2007) because of a high proportion of generalist feeders that easily switched among food resources (Wantzen and Wagner 2006). Another problem is that the FFG to which many southern hemisphere taxa belong is not known; thus, authors often assign taxa to FFGs based on North American FFG classifications. Analyses of gut contents and mouthpart morphologies of southern hemisphere taxa would solve this problem (e.g., Tomanova et al. 2006).

The evidence that FFGs are surrogates for ecosystem processes has been questioned (Mihuc 1997). A generalist feeding habit apparently is a common strategy among invertebrates in lotic systems. Moreover, FFGs group invertebrates on the basis of the size, rather than the type, of the particles they process as food (Cummins and Klug 1979), and FFGs do not indicate whether processed particles are assimilated (Mihuc 1997).

The FFG approach originated in the US, but the most cited J-NABS bioassessment article based on FFG as a biological response to human disturbance (Angradi 1996; Fig. 1) received 36 citations and does not appear among the 100 most-cited J-NABS articles. Thus, based on citation reports, papers published in J-NABS have contributed relatively little to the development of the topic.

Indirect assessment of ecosystem processes based on multiple biological traits of invertebrates

A recent approach to bioassessment combines FFGs with a wide variety of other biological traits including food types, body size, fecundity, voltinism, and dispersion ability of benthic macroinvertebrates to describe functional community structure (e.g., Corkum and Ciborowski 1988 [2 traits], Charvet et al. 1998 [Fig. 1], Minshall and Robinson 1998, Snook and Milner 2002, Lamouroux et al. 2004 [Fig. 1], Dolédec et al. 2006).

In its simplest definition, a trait is an approximation for organism performance, and the change of this performance might scale up to ecosystem functioning as follows: responses of ecophysiological and life-history traits of individuals (e.g., growth, fecundity, survival) might affect demographic traits of populations (e.g., birth, death, emigration, immigration), which in turn, might affect community structure and dynamics (e.g., species occurrence) and ecosystem functioning (e.g., energy flow, chemical cycling) (Violle et al. 2007).

Multiple-trait based (MTB) assessments are deeply rooted in ecological theory (Statzner et al. 2001b) and rely on predicting ecological responses to selection induced by environmental conditions (i.e., habitat templet concept, Southwood 1977, 1988, Townsend and Hildrew 1994 [Fig. 1], Poff 1997 [Fig. 1]). Use of multiple biological traits has several advantages over traditional assessments based on taxonomic composition. Traits are expressed in many species, so functional community descriptions can be compared among regions that differ in their taxonomic composition and might have large-scale applicability (Statzner et al. 2001a, Horrigan and Baird 2008). Biological trait responses of invertebrate assemblages to similar reference environmental conditions converge at the scale of regions (Archaimbault et al. 2005; Fig. 1), nations (Charvet et al. 2000), continents (Statzner et al. 2004, 2005 [Fig. 1]), and the world (Statzner et al. 1997, see Lamouroux et al. 2002 [Fig. 1] for fish trait response at this scale).

Most traits should be affected predictably by various types of human impact (e.g., Statzner et al. 2005, Dolédec et al. 2006, Dolédec and Statzner 2008). For example, flow increases should select for smaller size, firmer attachment to substrate, and streamlined shape, whereas flow decreases should favor organisms with larger size, good swimming abilities, and spherical shape. Siltation should favor traits that enable penetration of fine substrates (e.g., burrowing), whereas traits that render organisms susceptible to smothering by sediment (e.g., lack of egg protection) should decrease. Organic pollution and resulting O₂ deficits should increase the proportion of organisms with aerial respiration. Heavy metal pollution should decrease organisms with smaller body size (greater body surface–volume ratio) or gill respiration (increase of body surface) and predators (bioaccumulation of toxins). MTB approaches might be able to identify the most sensitive traits for a given impact type and facilitate targeted management actions.

Preliminary studies have shown the responses of traits to various human disturbances. For example, Richards et al. (1997; Fig. 1) proposed models in which catchment- and reach-scale physical characteristics were used to predict the occurrence of specific life-history and behavioral traits as a tool for landuse management. Charvet et al. (1998) reported that effluents from a wastewater treatment plant significantly changed many biological traits of benthic invertebrates in a small stream (e.g., more larger animals with fewer descendants and greater longevity downstream of the effluent). Statzner et al. (2001a) found differences in the trait compositions of communities in near-pristine and sewage-impacted habitats at the European scale. Invertebrate traits discriminated overall human impact on river reaches in the large European rivers (Dolédec et al. 1999 [Fig. 1], Usseglio-Polatera and Beisel 2002) and significantly discriminated among various levels of human impacts (Gayraud et al. 2003; Fig. 1). The frequently debated topic of best taxonomic level for bioassessment (e.g., Bowman and Bailey 1997, Lenat and Resh 2001, Chessman et al. 2007; Fig. 1) also was assessed for MTB approaches. Trait descriptions of genera provided an accurate picture of the functional community structure at different spatial scales in reference situations (Dolédec et al. 2000; Fig. 1), and trait weighting by the presence–absence of genera reliably discriminated various levels of human impacts (Gayraud et al. 2003).

Multivariate techniques have been designed to analyze relationships between biological traits and environmental conditions weighted by taxon presence–absence or abundance (Dolédec et al. 1996, Legendre et al. 1997). Mellado Diaz et al. (2008) recently used the 3-table ordination (RLQ) technique to illustrate that more invertebrates with small size, multivoltinism, diapause, or ovoviviparity were present in frequently disturbed streams in semiarid areas than in more stable streams in undisturbed forested upland areas.

The quality of MTB assessments depends largely on data quality and availability as for other biomonitoring tools (e.g., Hering et al. 2004a, Stoddard et al. 2008). A major problem of MTB approaches is to assemble consistent trait information because knowledge is often lacking, especially for rare taxa. Macroinvertebrate trait databases have been developed in Europe (e.g., Usseglio-Polatera et al. 2000a, b, Tachet et al. 2002, Gayraud et al. 2003, Ilg and Castella 2006, Statzner et al. 2007), the US (Vieira et al. 2006, Bêche and Resh 2007), and in the southern hemisphere (Dolédec et al. 2006, Tomanova and Usseglio-Polatera 2007). Most of these databases use a fuzzy coding procedure (Chevenet et al. 1994) to address trait variability that often is related to differences among life stages of a taxon, differences across the range of a taxon, or absence of solid knowledge for a taxon. Trait databases for biota other than macroinvertebrates have been developed for zooplankton (Barnett et al. 2007), aquatic macrophytes (Willby et al. 2001), and fish species in Europe (Pont et al. 1995, Santoul et al. 2005, Blanck et al. 2007) and in the US (Goldstein and Meador 2004, 2005). MTB assessments potentially are applicable to ecosystems other than freshwaters, including transitional waters (Mouillot et al. 2006) and marine environments (e.g., Bremner et al. 2003, 2006, Frid et al. 2008).

MTB approaches are widely used in Europe, but have received limited attention in other parts of the world (e.g., Merritt et al. 2002, Horrigan and Baird 2008). Less than 2% of J-NABS bioassessment papers address MTB approaches (Table 1), but an often-cited (>200) J-NABS article by Poff (1997) has promoted use of biological traits as responses to natural environmental filters. The most-cited J-NABS article dealing with MTB bioassessment (Merritt et al. 2002) has received 15 citations and is not among the 100 most-cited J-NABS articles.