Beyond taxonomy: a review of macroinvertebrate trait-based community descriptors as tools for freshwater biomonitoring
Correspondence author. E-mail: email@example.com
1. Species traits have been frequently used in ecological studies in an attempt to develop a general ecological framework linking biological communities to habitat pressures. The trait approach offers a mechanistic alternative to traditional taxonomy-based descriptors. This review focuses on research employing traits as biomonitoring tools for freshwater ecosystems, although the lessons learned have wider application in the assessment of other ecosystem types.
2. We review the support from ecological theory to employ species traits for biomonitoring purposes (e.g. the habitat templet concept, landscape filtering hypothesis), and the subsequent studies that test the hypotheses arising from these theories, and apply this knowledge under real freshwater biomonitoring scenarios. We also include studies that deal with more specific issues such as trait trade-offs and trait syndromes.
3. We highlight the functional trait approach as one of the most promising tools emerging for biomonitoring freshwater ecosystems. Several technical issues are addressed and solutions are proposed. We discuss the need for: a broader unified trait biomonitoring tool; a more accurate understanding of the natural variation of community patterns of trait expression; approaches to diminish the effects of trait trade-offs and trait syndromes; additional life history and ecological requirement studies; and the detection of specific impacts under multiple stressor scenarios.
4. Synthesis and applications. This review provides biologists with the conceptual underpinning for the use of species traits as community descriptors and for freshwater biomonitoring and management. We expect that the functional trait approach will ultimately improve communication to managers and legislators of the importance of protecting freshwater ecosystem functions.
There has been a long publication history examining the relationships between species biological traits and environmental constraints (see Statzner, Hildrew & Resh 2001 for an historical review since the 19th century). Although this interest in the association between habitat affinities and the traits of an organism started with early ecologists, the perception of its significance has changed over time (Statzner, Hildrew & Resh 2001). A trait is defined as a characteristic that reflects a species adaptation to its environment. Traits are usually divided into two categories: biological traits (e.g. life cycle, physiological and behavioural characteristics, such as maximum body size, lifespan, feeding and reproductive strategies, mobility, etc.) and ecological traits (related to habitat preferences, like pH and temperature tolerances, tolerance to organic pollution, biogeographic distribution, etc.). In fact, an organism’s life history is a set of co-adapted traits arising from natural selection to solve particular ecological problems (Stearns 1976), such as limited food resources or the presence of predators. During the 1970s, there was a focus on life-history tactics in general and mechanisms for their evolution in particular (Stearns 1976) in the absence of a global theoretical ecological framework linking habitat pressures and species traits.
Certain traits, by influencing organismal performance, can affect ecosystem functioning (functional traits) (McGill et al. 2006). Therefore, besides the potential to define biological communities and predict changes in those communities due to disturbance, species traits can also be used as measures of community functional diversity (Petchey & Gaston 2006). Both of these aspects can produce tools to predict the functional consequences of biological change caused by human activities.
Biological traits and ecological theory: the habitat templet concept
Research into the predictive potential of using species traits to define biological communities (the traits-based approach sensuBaird, Rubach & Van den Brink 2008) has increasingly appeared in the scientific literature in the last three decades. The first works addressing this issue were in plant ecology and hypothesized that competition, stress and disturbance are interactive determinants of herbaceous vegetation by invoking different strategies on plants (Grime 1974). The value of this approach to plant management was presented by Grime as the ability to detect competition and stress and to predict the intensity of disturbance at particular sites. These ideas and theoretical models concerning plant attributes/traits were the basis for the development of new plant community descriptors and for the functional interpretation of vegetation data.
In 1977, Thomas Southwood formulated a general theoretical model for the classification of ecological strategies in his presidential address to the British Ecological Society (Southwood 1977). Southwood showed that, in theory, the ecological strategies of a species have evolved in response to a non-rigid ‘habitat templet’– the characteristics of the habitat are said to select and favour certain sets of biological characters in the individual. This and other theoretical models for life-history strategies (including Grime’s model) were later compared in their ability to detect patterns and predict the evolution of strategies from the combination of traits under the action of habitat selective forces (Southwood 1988), and similarities were found among the predictions resulting from the application of the different models.
The underlying hypothesis in Southwood’s work has inspired several successful applications of the habitat templet perspective to different areas of ecological knowledge, particularly studies on freshwater systems (including the River Continuum Concept; Vannote et al. 1980), macroinvertebrate communities and biomonitoring.
The river habitat templet
Using the river habitat templet, stream ecologists have made a priori predictions about stream community changes (Townsend & Hildrew 1994) by predicting trends in species traits along gradients of disturbances (i.e. temporal variability) and refugia (i.e. spatial variability). These types of predictions were tested by applying the trait approach to stream communities taxonomically well characterized (e.g. Resh et al. 1994; Townsend, Scarsbrook & Dolédec 1997).
Many examples of this application come from studies in one of the major rivers in Europe, the Rhône. One initial approach linked lists of aquatic coleopteran taxa to environmental and biological variables using correspondence analysis, in order to organize them along a current-substrate gradient (Bournard, Richoux & Usseglio Polatera 1992). In 1994, an entire issue of the Journal of Freshwater Biology (volume 31, issue 3) was devoted to the long-term ecological research carried out on the Upper Rhône and to the attempt to relate theoretical habitat templets, freshwater species traits and species richness. This was the first major test of the river habitat templet concept, by means of a linkage of species traits to a set of temporal and spatial variables, using a fuzzy coding approach of the data (Chevenet, Dolédec & Chessel 1994). This approach codes biological and environmental information on a scale that describes the affinity between two items, from 0 (when no affinity exists between those items, for example a species and a trait modality) to x (strong affinity between the two items). The advantages of this coding methodology are: the standardized coding of information coming from diverse sources or concerning very distinct taxa; the inclusion of trait variation found within a species; and the potential for statistical analysis by ordination methods like correspondence analysis (Chevenet, Dolédec & Chessel 1994).
The overall design of the research strategy of the ‘Rhône team’ was to develop a general tool for ecologically oriented river management, where a priori predictions were derived from available ecological theories (Statzner, Resh & Roux 1994). A physical habitat templet for the Rhône River was scaled using available sets of physico-chemical data, and trait data were structured through fuzzy coding. Alternative mathematical tools allowing the simultaneous analysis of species and environment data as well as their relationships arose, namely a two-table ordination method – co-inertia analysis (Dolédec & Chessel 1994). This allowed the testing of relationships between species traits and habitat utilization, and trait trends were related with the river habitat templet (Townsend & Hildrew 1994).
Based on Southwood’s theory (Southwood 1977), Townsend & Hildrew (1994) predicted that populations in more spatially heterogeneous habitats will be less disturbed by certain temporal variations, stating that the likelihood of an organism removal by disturbance will be influenced by spatial heterogeneity. The authors made autoecological predictions on life-history, morphological and physiological traits, predicting how they would vary with disturbance.
The ‘Rhône team’ analysed 13 animal and plant taxonomic groups of the Upper Rhône River and concluded that spatio-temporal variability did not serve as a templet for the species traits (Resh et al. 1994). The matching between species traits and habitat characteristics appeared to be more complex than initially predicted: alternative trade-offs between trait combinations could have been neglected (Resh et al. 1994) or the set of taxonomic groups was too heterogeneous.
In a New Zealand river, researchers tested the habitat templet with only benthic insects and redefining the templet axes specifically for this taxonomic group, by using the frequency of disturbance as the temporal axis and variation of features associated with refugia as the spatial axis (Townsend, Scarsbrook & Dolédec 1997). The result was that many of the initial habitat templet predictions proved to be correct. This work has supported the idea that the habitat templet analysis had to take into consideration the groups of organisms of interest and their possible relationships with stream disturbance.
Benthic macroinvertebrates constitute a taxonomically well-described, widespread, highly abundant and diverse group of aquatic organisms that respond to a wide range of stressors. They can be good indicators of localized conditions and long-term effects through the application of low cost and easy-to-use sampling methodologies. Their undeniable functional importance in freshwater ecosystems is mostly related to their diverse array of feeding habits (Wallace & Webster 1996).
Macroinvertebrate traits have been used to predict biotic interactions with their predators (e.g. fish). A conceptual model and procedure for ranking invertebrates based on functional relationships with their fish predators has been proposed (Rader 1997) as an important step in predicting disturbance effects on fish populations. Invertebrates were ranked according to traits for drift propensity and availability as food and significant correlations between the predicted ranks and the actual ranks of invertebrates in fish guts were obtained (Rader 1997). Others (Statzner et al. 1997) have focused on reproductive traits of aquatic insects to establish a relationship with the habitat characteristics at different scales, following the approach of the ‘Rhône team’ but using bibliographic searches to extract worldwide insect data. This has introduced bias towards abundant and widely distributed species, which may have reduced the trait range to that of r-strategist species: small-sized species with short generation times, high fecundity and high levels of dispersal, as opposed to the k-strategists (Southwood 1977). A co-structure of the reproductive and habitat traits was detected and the authors positively tested Southwood’s habitat templet theory for reproductive traits of aquatic insects. The agreement with the habitat templet predictions was not true to all traits studied though. The authors conclude that ‘one should not expect that a given habitat acts as a templet in an uniform way for all traits of all its species’ (Statzner et al. 1997), and there seems to exist a certain effect of the habitat scale.
Biological traits and ecological theory: the habitat-filtering hypothesis
The concept of habitats working as environmental trait filters was further developed by several authors due to a growing recognition that there is a need for general predictive models that are able to deal with an increasing number of environmental issues. Resetting the community-level research goal from description to prediction has led to attempts to establish assembly rules that forecast presence/absence of species as well as their abundance according to the available set of environmental conditions, assuming these conditions work as filters (Keddy 1992). Essentially, the least suited sets of biological traits are eliminated in a given environment, and only taxa possessing traits that pass through the habitat filter will be present in the community.
Addressing the interplay of regional and local biotic and abiotic factors that governs freshwater community composition, Poff (1997) presented one possible general framework for understanding and predicting the distribution and abundance of species in those communities, where species are described in terms of their functional relationships to the habitat-selective forces (landscape filters). These filters are said to constrain expression of local selective forces or biotic potential, so that species in a regional pool must possess appropriate functional attributes (species traits) to ‘pass’ through these filters and join a local community. Once more, this means that species distribution and abundance and community composition could be predicted if environmental constrains imposed at different scales are considered (habitat data). The habitat filter scales specified by Poff were basin or watershed, stream valley bottom or stream reach, channel unit (riffle, pool) and microhabitat; the importance of hierarchically integrating different spatial and temporal scales as factors affecting species distribution and abundance is therefore highlighted by Poff’s model. In contrast with the regression techniques used for example in the Rhône River data, this mechanistic approach to community prediction implies explicit, quantitative assumptions about the filtering of species by habitat factors (Poff 1997). Highly detailed, a priori knowledge on both traits and filters is needed in order to establish this predictive model, and thus it is sensitive to data availability.
This importance of habitat properties at different scales in influencing species traits was addressed in a study where reach- and catchment-scale characteristics were used to predict trait patterns of freshwater insects of east-central Michigan agricultural catchments (Richards et al. 1997). The authors successfully predicted life-history and behavioural attributes using the reach-scale characteristics, but found some difficulties in observing species trait patterns at smaller scales. Following these observations, attempts to separate the respective influence of various habitat filters were made (Lamouroux, Dolédec & Gayraud 2004). The complexity of the environmental filter theory development is undeniable (Statzner, Dolédec & Hugueny 2004), in the sense that multiple filters, both biotic and abiotic and working at very different spatial and temporal scales, are expected to affect biological traits, often in antagonistic ways.
Progress in the development of traits as biomonitoring tools
Charvet and co-workers (Charvet, Kosmala & Statzner 1998; Charvet et al. 1998) have directly addressed stream biomonitoring using macroinvertebrate traits, and the need to develop more generic biomonitoring tools. Traditionally used diversity measures, biotic indices and community structure analysis – taxonomy-based tools – do not allow the establishment of causal relationships with stressors and do not integrate natural fluctuations. The authors hypothesized that, based on the habitat templet concept and using a species-abundance table as well as a species-traits table, it is possible to obtain a functional image of the study system and detect pollution impact (in a simple upstream–downstream study design). While comparing the traditional approaches that use physico-chemical or taxonomic data (the latter using the calculations of eight biological indices – e.g. Margalef Index, Shannon Index – and community structure analysis) with the ‘functional approach’ (using biological traits weighted by abundances), a better separation of the upstream and downstream sites was obtained when including trait information. This worked as an initial assessment of the potential of traits to overcome some of the issues raised by traditional biomonitoring methodologies. Dolédec, Statzner & Bournard (1999) evidenced that, at the time, much work had still to be performed in the adaptation of the trait approach to a biomonitoring technique.
In a search for a conceptual framework for trait freshwater biomonitoring, Dolédec, Statzner & Bournard (1999) highlighted that an ideal biomonitoring tool – generic in terms of geographic application, specific in terms of stressor identification, reliable and derived from sound theoretical ecological concepts – is possible to obtain using benthic macroinvertebrate ecological and biological traits, as an alternative to the traditional taxonomy-based approaches. These taxonomic approaches (metrics such as abundance, taxa richness of species or families) are said to fail when generalizations to different types of freshwater systems are needed (regional constraints) and result in losses of ecological information. The authors again tested the river habitat templet, using multivariate analysis to evaluate how patterns of species traits in macroinvertebrate communities of a large river could discriminate differences in overall human impact, in comparison with taxonomy-based approaches (species abundance-by-sites). All tools tested revealed human impact effects; the analysis based on biological traits was less confounded by natural spatial gradients and was the best indicator of stressor impact.
Meanwhile, following the French trait research studies, a subset of the ‘Rhône team’ continued to fill the biological and ecological trait database and published a revised version for the macroinvertebrate fauna of French rivers (Tachet et al. 2000), later complemented with information for other European areas and taxa (Statzner, Bonada & Dolédec 2007). They have also revised the macroinvertebrate trait applications to running water biomonitoring, highlighting that a broader unified trait biomonitoring tool covering not only France but also larger geographical scales was needed, in order to meet current European policies (Statzner et al. 2001).
‘Reference’ state studies in freshwater ecosystems
Charvet et al. (2000) tested whether biological and ecological traits used as biomonitoring tools would yield the same results for semi-natural (non-impacted) streams of different ecoregions. They found that trait-based community structure was in fact stable across environmental gradients, in contrast with the taxonomy-based community structure, which varied significantly with geologic and altitudinal differences. Other authors tested both the spatial stability of the reference state (referring to the example of the ‘reference condition approach’; Wright, Sutcliffe & Furse 2000) and the potential for indication of specific human impacts using data at the European scale (Statzner et al. 2001). A high spatial and temporal stability of the functional composition of non-impacted freshwater communities at the European scale was detected and this was related with a landscape filter (sensuPoff 1997) limiting taxa abundance at the local scale. Although differences in the trait patterns of reference and impacted locations were detected, the human impact examples chosen (dams and sewage input) were not sufficient to highlight clear mechanistic explanations for the patterns found.
Archaimbault, Usseglio Polatera & Bossche (2005) have assessed the influence of geology on trait profiles of macroinvertebrate communities of reference sites belonging to the same biogeographic area, confirming that the functional structure of reference communities was constant across geology types, even when taxonomic variability was observed. Statzner et al. (2005) have shown that the expected variation across ‘reference’ rivers can be predicted by trait patterns, more or less accurately depending on the chosen models of analysis. Bonada, Rieradevall & Prat (2007) concluded that stream flow permanence constrains the invertebrate community both structurally and functionally, relating various trait profiles with the Mediterranean climatic characteristics.
Over the last two decades, the majority of the macroinvertebrate trait applications to freshwater systems have been focused on the European continent but the interest in developing a trait biomonitoring tool applicable also to North American freshwater invertebrates has led to the publication of a trait database with more than 14 000 records (Vieira et al. 2006). North-American studies have also started to examine the relationships between physical variables and benthic trait patterns, particularly under ‘reference state’ conditions. The benthic community structure and functioning of high-gradient mountain streams were studied and the influence of longitudinal and reach-scale variables on those communities was assessed (Finn & Poff 2005). For this particular small spatial extent scenario, defining assemblages functionally provided no greater understanding of community patterns than taxonomic definitions, given several known environmental variables (see also the study of Finnish headwater streams by Heino et al. 2007), and the issue of trait trade-offs and alternative adaptive solutions for the same habitat was once more raised. The results did not agree with previous studies that observed a functional stability of unimpacted communities across environmental gradients (Charvet et al. 2000), but the authors hypothesize that, although human impact was low, the natural stressors associated with more extreme zones (alpine zone, harsher environment, limited human influence) could explain the functional instability detected.
Another recent geographical expansion of the freshwater trait approach was to the neotropical benthic communities (Tomanova & Usseglio-Polatera 2007). Taxonomic and functional structures of reference benthic communities were compared and, after overcoming the difficulty of limited availability of neotropical invertebrate trait information, the authors found that communities that differed taxonomically were functionally similar. They have also found that some of the traits related to the environmental variables, but not sufficiently matching the trait–habitat relationships detected for temperate environments (not enough to confirm the existence of general benthic trait rules that would be applicable over different climatic regions).
As for natural temporal variations, several studies (Bêche, McElravy & Resh 2006; Bêche & Resh 2007) highlight the importance of an accurate understanding of the natural variation of community patterns in ‘reference’ systems (natural ecological filters), so that the biological trait approach can be properly used to detect differences between these systems and the ones affected by human stressors (anthropogenic filters).
Technical aspects of biomonitoring
The way in which taxonomic and spatial resolution influence trait-based predictability is a substantial issue. Dolédec, Olivier & Statzner (2000) combined abundance data obtained at different scales and trait data, and investigated the accuracy of community descriptions expressed at different levels of taxonomic and spatial resolution. They concluded that scale significantly affected the accuracy of community descriptions based on abundance but that, while using biological traits, accurate descriptions were achieved by species, genera or family identifications (Dolédec, Olivier & Statzner 2000). This suggests that species identification may not always be necessary in future stream biomonitoring studies, a contribution for the simplification of the tools under development as the taxonomic expertise required would be significantly decreased.
Under this premise, data of benthic macroinvertebrate traits from French freshwaters were used, at the genus level, to explore issues that limited the successful testing of the river habitat templet, like trait trade-offs (Resh et al. 1994) – variable combinations of traits that organisms may possess that confound the establishment of matches between habitat and trait patterns. Non-taxonomic aggregations of taxa – functional groups of organisms with high relationship similarities among their biological and ecological traits – were defined in order to investigate the mechanisms affecting species distributions (Usseglio-Polatera et al. 2000). The application of these groupings for human impact differentiation was later attempted (Usseglio-Polatera et al. 2001), and functional trait diversity (Shannon diversity index calculated using the non-taxonomic groupings) seemed a better indicator of human impact than taxonomic diversity, with more impacted sites exhibiting higher proportions of organisms with traits assumed to adapt them to frequent disturbances or reduce the impact of certain stressors.
Gayraud et al. (2003) suggest once more (cf. Dolédec, Olivier & Statzner 2000) that functional descriptions of the functional structure of communities more or less affected by human activities at the species level may not be needed. Attention should also be given to the choices of weighting of species or excluding alien species. The authors suggest the development of a low-cost community functional description tool using presence–absence of genera or families of native freshwater taxa, which could be used to discriminate strong human impact gradients. Other tests (Haybach et al. 2004) have also detected high robustness of the trait approach with decreased taxonomic resolution as well as against faunal changes, such as the seasonal ones. In this study, the authors were also able to demonstrate the functional quality of the approach by detecting associations between the trait pattern responses to disturbance of a certain site with the predominating rK-strategies.
Another technical aspect to consider in a trait biomonitoring programme is the effect of different sampling efforts, as these effects on taxonomic variables are well known (Kerans, Karr & Ahlstedt 1992). It is therefore important to test how the sampling effort would affect the functional diversity measures based on biological traits. While confirming the influence of sampling effort on taxonomy-based measures, Bady et al. (2005) concluded that functional diversity indices (using traits) show greater accuracy with less sampling effort and higher precision across season and location gradients.
The inclusion of species traits in ecological studies brings not only a variety of advantages but also large challenges for the simultaneous analysis of three data matrices: environmental, trait and species composition tables. The RLQ analysis, a three-table ordination method, was presented as a solution for this problem (Dolédec et al. 1996). It was followed by the testing procedures proposed by Legendre, Galzin & Harmelin-Vivien (1997) with the intent of establishing a link between an environmental variable and a species trait through a table containing presence–absence data, presented as the fourth-corner problem. Some limitations of the proposed methodologies were highlighted subsequently, namely the lack of consideration for evolutionary linkages among traits, the analysis of a single trait and a single environmental variable at a time, and the impossibility of using abundance data, as both methods imply the use of binary data (Nygaard & Ejrnaes 2004; Poff et al. 2006). To address at least part of these difficulties, an improved fourth-corner method was proposed (Dray & Legendre 2008).
Trade-offs and trait syndromes
Since the first attempts to test the river habitat templet, the issue of alternative trade-offs between trait combinations has been discussed (e.g. Resh et al. 1994; Usseglio-Polatera et al. 2000). It was noted that different possible combinations of traits in an organism, working as adaptive solutions, may confound the establishment of matches between habitat and trait patterns. In fact, trade-offs in species performances of different ecological functions is one of the most common explanations for coexistence in communities (Kneitel & Chase 2004). They are defined as negative functional interactions between traits, with investments in one trait leaving fewer resources available for investments in another (Verberk, Siepel & Esselink 2008b). The need arises for a formal analysis of trade-offs, accounting for phylogenetic relationships and potential confounding effects on trait measures.
Poff et al. (2006) have highlighted that, although the developments of the trait approach in biomonitoring are promising, there is a lack of adequate understanding of how individual traits are intercorrelated and how this dependence among traits reflects phylogenetic constraints. The fact that traits are often linked together by evolution, forming the so-called trait syndromes (Nylin & Gotthard 1998), hinders their treatment as independent entities. The authors have explored this issue for lotic insects of North America by creating a trait database, by demonstrating the importance of trait state linkage within a taxon, by examining trait correlations and using this to provide information on future trait selection for biomonitoring purposes. They have also examined phylogenetic associations of certain traits and suggested that robust (unlinked, uncorrelated) traits would provide better insights to predictive community ecology and multiple trait response along environmental gradients (Poff et al. 2006).
As species traits cannot be viewed independently, one possible solution can be to see them as part of complex adaptations, combinations of co-evolved attributes based on known functional relationships among them, i.e. life-history strategies (Verberk, Siepel & Esselink 2008b). These strategies enable the organism to deal with a range of ecological problems. Verberk et al. present these strategies in four groups (dispersal, synchronization, reproduction and developmental trade-off), based on species traits and their interrelations known from life-history theory. Once more biological knowledge on the different species is available, these strategies should provide the connections between traits and environmental conditions, pursued by those that have been testing the habitat templet theory over the last decades (particularly since Resh et al. 1994). This approach has already been tested for lentic macroinvertebrate species (Verberk, Siepel & Esselink 2008a), where different species are assigned to the same life-history strategy, according to their ability to solve similar ecological problems by employing the combination of species traits. Differences in ‘strategy composition’ of freshwater systems were related with known environmental conditions, and good prospects for environmental quality programmes were presented. This approach illustrates the importance of considering comparative phylogenetic methods as employed in evolutionary ecology (e.g. Freckleton, Harvey & Pagel 2002) where convergent phenotypic solutions to ecological problems across phylogenies can be understood and placed in their correct context, with the potential to contribute towards global analyses of species in relation to their environment (Westoby 2006).
Applications in human impact scenarios
Initial studies dealing with the ability of the biological trait approach to detect the effects of deleterious human impacts on freshwater ecosystems have already been mentioned (e.g. for sewage pollution, Charvet, Kosmala & Statzner 1998; for overall human impact, Dolédec, Statzner & Bournard 1999; Statzner et al. 2001; for agricultural land use, Richards et al. 1997). More specific studies under various anthropogenic impact scenarios followed, attempting to address the issue of human impact differentiation.
In New Zealand streams affected by agricultural development, it was shown that both taxonomy- and trait-based methods were able to discriminate land use practices, but the trait approach worked better for this purpose, accounting for more between-land use variance (Dolédec et al. 2006). Shifts in macroinvertebrate trait patterns with increasing land use were described, proving once more the great potential of this approach for the biomonitoring of stream communities, particularly the ones affected by agricultural stressors. Díaz et al. have suggested a habitat templet for SE Spain streams, including a disturbance axis (natural climatic variation) and an adversity axis (intensive agriculture pressure), and an RLQ analysis was used for identifying species traits that respond to impacts of land use change at different scales (Díaz, Alonso & Gutierrez 2008). Although confounding effects of geology, altitude and climate were detected, the authors were able to link macroinvertebrate community ecological organization with disturbance and human pressure. In the case of this particular type of human impact, it would be important to distinguish, for example, pesticide effects from other particular stressors associated with agriculture or from natural stressors. With this purpose in mind, the concept of classifying species according to their vulnerability towards pesticides (Species At Risk, or SPEAR) defined by certain ecological and physiological traits was tested in Northern Germany streams (Liess & Von Der Ohe 2005) and across European biogeographic regions (Schaefer et al. 2007). This indicator system successfully discriminated reference and pesticide contaminated sites, and it was demonstrated that these pollutants altered both the community structure and function of the studied lotic systems (Schaefer et al. 2007).
Dealing with hydrological alteration at a continental scale (across Canada), Horrigan & Baird (2008) proposed a set of flow-sensitive traits to be used in the development of biomonitoring through flow metrics, addressing the possible effects of trait syndromes. Additionally, they suggest the preliminary establishment of typical ranges of flow-sensitive traits for different types of reference streams (differing in hydrological regime, size, order and slope) and their use as biomonitoring metrics with potential global applicability.
Attempts to use trait profiles to differentiate types of stressors heavily affecting large European rivers (namely heavy metal pollution and cargo-ship traffic) did not result in an adequate discrimination of impacts, highlighting that the detection of specific impacts under multiple stressor scenarios is very complex (Dolédec & Statzner 2007). In order to tackle this problem, individual and combined effects of the principal stressors affecting New Zealand agricultural streams were investigated (Townsend, Uhlmann & Matthaei 2008) by using both taxonomy- and trait-based measures. A change from management solutions based on individual stressor analysis to solutions based on multiple stressor analysis is recommended. Another multiple stressor scenario is the one imposed by wildfires, and the comparison of successional patterns in burned and reference streams has highlighted interactions between functional traits of local taxa and the postfire environmental conditions (Vieira et al. 2004).
Other types of studies dealt with the implications of future climate change scenarios by comparing ecological (trait and taxonomic) differences between Mediterranean and temperate European streams (Bonada, Dolédec & Statzner 2007). It was highlighted that climate change may produce large changes in the taxonomic composition without substantially altering the pattern of functional traits.
One example of another potential of the trait approach is the prediction of the efficiency of restoration management activities by using macroinvertebrate communities to assess the effects of restoration measures (working as trait filters). There are studies recommending phased restoration procedures in order to ensure the survival of species not possessing the traits that would allow them to survive to the future application of such measures (van Kleef et al. 2006). Other studies (Tullos et al. 2009) report that both the taxonomic and the functional-trait approaches showed the effects of restoration measures. Both these works (and also, e.g. Reckendorfer et al. 2006; Paillex et al. 2009) prove that there is an interesting potential for the applications of trait metrics in the establishment of informed management resolutions by proper understanding of the functional implications of such measures.
This review highlights much of the progress towards the development of a theoretical framework for ecologically oriented stream biomonitoring and management. There is still a long way to go before scientists will be able to use community level data to accurately diagnose causes of stream impairment (stressor biodiagnosis), besides the mere identification of degradation presence. Scientists have to be able to convey the importance of protecting freshwater ecosystem functions, particularly to managers and legislators. In the last decades, the functional trait approach has been identified as one of the tools with more potential for biomonitoring and management of stream ecosystems. This potential focuses mainly around the concept of the ideal biomonitoring tool – generic in terms of geographic application, specific in terms of stressor identification, reliable and derived from sound theoretical ecological concepts – and in the fact that traditional taxonomy-based approaches many times fail to work in these ideal ways.
By looking at species trait patterns, significantly affected by human impacts, we can make mechanistic interpretations of the effects of anthropogenic activities by examining how traits covary with specific environmental pressures and drivers. This is possible both in freshwater environments and in other ecosystems where a suitable knowledge-base exists, and environmental management can be substantially improved. The fact that issues with natural spatial and temporal variability can be overcome is a huge advantage for those trying to assess and maintain ecosystem quality over broad spatial scales or in situations of limited availability of taxonomic information (e.g. in less-developed countries, the tropics or the polar regions).
Together, the trait studies reviewed here attest to the ability of the functional trait approach to form the basis of a new generation of biomonitoring tools. However, they also highlight the need for further research in the following areas: (i) a broader, unified trait biomonitoring tool covering larger geographical scales (this implies the need for a confirmation of the existence of general benthic trait rules that would be applicable over different climatic regions); (ii) an accurate understanding of the natural variation of community trait patterns in unimpacted systems; (iii) new trait analysis protocols that diminish the effects of trait trade-offs and trait syndromes; (iv) undertaking additional life-history and ecological niche studies, promoting open access trait databases where the compiled results of these studies can be found (see Appendix S1 in Supporting Information); (v) improved understanding of which traits are functionally important and how to use them as relevant measures of functional diversity at the community level, especially to guide restoration programmes; and (vi) increased effort to apply traits-based approaches to the detection of specific impacts under multiple stressor scenarios.
The valuable contributions from the editor and two anonymous reviewers were greatly appreciated. This work was supported by Fundação para a Ciência e a Tecnologia (Portugal) and European Social Fund (III Quadro Comunitário de Apoio) through the project Grant PTDC/AMB/74346/2006, through a PhD Grant to SM (ref. SFRH/BD/18514/2004) and through a Canadian National Sciences and Engineering Council (NSERC) Discovery Grant to DJB.