Synthesising the trait information of European Chironomidae (Insecta: Diptera): Towards a new database

Chironomidae are among the most conspicuous and ecological diverse group of freshwater invertebrates. They may dominate unimpacted communities in abundance and biomass, accounting for more than 50% of macroinvertebrate species in standing and flowing waters. In deep zones of eutrophic lakes and highly human-impacted streams, they are often the only family of aquatic insects remaining. In bioassessment programmes, Chironomids are often identified at the family and subfamily levels, due to difficulties in the taxonomic identification of larvae resulting from a high intrinsic morphological similarity. This may potentially result in bias as, similarly to Ephemeroptera, Trichoptera or Plecoptera, Chironomidae species, which are replaced along natural and human-impacted gradients due to differences in their ecological requirements. Recently, multiple trait-based approaches have been proposed to complement taxonomicbased assessment of streams and rivers using macroinvertebrates. However, the lack of specific trait information for Chironomidae prevents their use in the functional assessment of communities. Therefore, here, we aimed to: (1) develop a trait database for European Chironomidae genera that can be used in future bioassessment and ecological studies; (2) evaluate, by multivariate analyses, whether our new database provides additional information on Chironomidae compared to the trait information provided in the commonly used European trait database (Tachet et al., 2010); and (3) determine whether the new information on Eltonian traits (proxy to biological traits) translates the most accepted phylogenetic relationships among Chironomidae subfamilies. We gathered information on 744 species and 178 genera, for 37 traits covering 186 trait categories, and found substantial differences between our database and the commonly used European trait database. In addition, available information on traits was not always in agreement with phylogenetic relationships among subfamilies. Orthocladiinae and Chironominae which are considered sister groups in evolutionary terms actually showed confident trait relatedness based on Eltonian traits tree while the remaining relationships between subfamilies are questionable. In addition, different traits can occur in closely related taxa depending on the environmental drivers operating on their habitats. Our study reveals that the usually accepted redundancy within the Chironomidae family and subfamilies must be a product of averaging the information from finer taxonomic resolution added to the substantial lack of information for this insect group. © 2015 Elsevier Ltd. All rights reserved. ∗ Corresponding author. Tel.: +351 916635553. E-mail addresses: sonia.rqs@gmail.com (S.R.Q. Serra), fernando.cobo@usc.es (F. Cobo) jf@ci.uc.pt (M.J. Feio). ttp://dx.doi.org/10.1016/j.ecolind.2015.09.028 470-160X/© 2015 Elsevier Ltd. All rights reserved. , mgraca@ci.uc.pt (M.A.S. Graç a), sylvain.doledec@univ-lyon1.fr (S. Dolédec),


a b s t r a c t
Chironomidae are among the most conspicuous and ecological diverse group of freshwater invertebrates. They may dominate unimpacted communities in abundance and biomass, accounting for more than 50% of macroinvertebrate species in standing and flowing waters. In deep zones of eutrophic lakes and highly human-impacted streams, they are often the only family of aquatic insects remaining. In bioassessment programmes, Chironomids are often identified at the family and subfamily levels, due to difficulties in the taxonomic identification of larvae resulting from a high intrinsic morphological similarity. This may potentially result in bias as, similarly to Ephemeroptera, Trichoptera or Plecoptera, Chironomidae species, which are replaced along natural and human-impacted gradients due to differences in their ecological requirements. Recently, multiple trait-based approaches have been proposed to complement taxonomicbased assessment of streams and rivers using macroinvertebrates. However, the lack of specific trait information for Chironomidae prevents their use in the functional assessment of communities. Therefore, here, we aimed to: (1) develop a trait database for European Chironomidae genera that can be used in future bioassessment and ecological studies; (2) evaluate, by multivariate analyses, whether our new database provides additional information on Chironomidae compared to the trait information provided in the commonly used European trait database (Tachet et al., 2010); and (3) determine whether the new information on Eltonian traits (proxy to biological traits) translates the most accepted phylogenetic relationships among Chironomidae subfamilies. We gathered information on 744 species and 178 genera, for 37 traits covering 186 trait categories, and found substantial differences between our database and the commonly used European trait database. In addition, available information on traits was not always in agreement with phylogenetic relationships among subfamilies. Orthocladiinae and Chironominae which are considered sister groups in evolutionary terms actually showed confident trait relatedness based on Eltonian traits tree while the remaining relationships between subfamilies are questionable. In addition, different traits can occur in closely related taxa depending on the environmental drivers operating on their habitats. Our study reveals that the usually accepted redundancy within the Chironomidae family and subfamilies must be a product of averaging the information from finer taxonomic resolution added to the substantial lack of information for this insect group.

Introduction
Chironomidae is the most widely distributed dipteran family; its larvae have colonised terrestrial habitats, as well as marine habitats, and fresh waters. The family can tolerate a wide range of environmental conditions, and some taxa can be found in extreme environments including ice-cold glacial trickles, hot springs, and rather unusual environments such as sub-desert steppes, aquatic hygropetric habitats and leaf axis of plants or rot-hole of trees Cobo and Blasco-Zumeta, 2001;Vallenduuk and Pillot, 2007;Pillot, 2009Pillot, , 2013.
Chironomidae richness worldwide is estimated at 20,000 species, but the lack of adequate description and identification difficulties at finer taxonomic resolution such as genus or species suggest that this number is underestimated Coffman and Ferrington, 1996).
In fresh waters, the Chironomidae family can account for ∼50% of the macroinvertebrate community Coffman and Ferrington, 1996), it is particularly abundant in reservoirs, lakes and in lowland rivers and urban streams, and may be the only insect remaining in highly human-impacted water bodies (Coffman and Ferrington, 1996;Raunio et al., 2011;Andersen et al., 2013). Chironomidae play a key role in organic matter processing by consuming fine particles of organic matter and transferring energy and nutrients to upper trophic levels since they represent prey for an array of organisms, including other invertebrates, fish and birds. They thus have a great influence over productivity and population dynamics of top consumers. Finally, Chironomidae assemblages change along the river continuum similarly to EPT taxa (Ephemeroptera, Plecoptera and Trichoptera) (e.g. Prat et al., 1983;González, 1990, 1991;Lindegaard and Brodersen, 1995;Puntí et al., 2009) and according the lake typology (Saether, 1979;Brodersen and Lindegaard, 1999;Mousavi, 2002).
Historically, Chironomidae family played an important role in lake and running water classification based on its trophic level and saprobity, which reflected the production and decomposition of organic material (Saether, 1979). Fossil chironomid assemblages also provided insights on past environmental conditions (Walker, 2001;Brooks, 2006) whereas abnormalities in body parts, mostly mouthpart deformities, have been used as indicator of contaminant effects in both water and sediments (Rosenberg, 1992). Therefore, the bioassessment potential of Chironomidae is great, being of particular importance in environments where other invertebrate groups are not present. This family includes taxa tolerant to different water salinity, pH, depth, temperature, organic carbon, nutrients and oxygen concentration (e.g. Laville and Vinç on, 1991;Schmidt et al., 2010;Servia et al., 2004) among other environmental variables. Some Chironomidae occur in good quality waters (e.g. Rheopelopia spp., Conchapelopia pallidula, Orthocladius thienemanni and Zavrelimyia melanura; Vallenduuk and Pillot, 2007;Marziali et al., 2010;Pillot, 2013), whereas others are rather tolerant to high organic contamination and high trophic degrees (e.g. Chironomus riparius, Rheocricotopus fuscipes and Rheocricotopus chalybeatus; Brodersen and Quinlan, 2006;Marziali et al., 2010;Prat et al., 2013) or low levels of dissolved oxygen (e.g. Procladius sp. and Eukiefferiella claripennis; Bazzanti and Seminara, 1987;Marziali et al., 2010). Despite the wide range of responses to the environmental gradients, the bioassessment of running waters generally use a coarse taxonomic resolution for depicting Chironomidae assemblages (Rosenberg, 1992;Coffman, 1995;Hawkins and Norris, 2000) because of the difficulties associated with the morphological identification of larvae beyond family and subfamily.
Besides the usual taxonomy-based approaches, trait-based approaches are being increasingly used as an alternative to assess stream biological integrity (Dolédec and Statzner, 2010). Traits may help to reveal the cause of impairment and give an indirect insight into which ecosystem functions may be affected by human disturbance (Archaimbault et al., 2005;Culp et al., 2011;Feio and Dolédec, 2012). Since traits are indicators of function, community trait composition allows a better understanding of stream functioning . However, few researchers have attempted to quantify trait information for Chironomidae, with some of them achieving a trait database at the subfamily and tribe levels (see Tachet et al., 2010 for Europe andPoff et al., 2006 for North America). Few works gathered information at the species or genus level considering a reduced number of traits and/or taxa (see Franquet, 1996 for France; for North America).
Here, we had three objectives. First, we aimed to categorise the European Chironomidae genus characteristics into a set of 21 traits and 110 categories used in Tachet et al. (2010) for all aquatic macroinvertebrates and a set of 16 additional traits specific to Chironomidae. Secondly, we investigated the distribution and variability of trait patterns within Chironomidae subfamilies, using the new trait database. Given the great variability reflected in trait heterogeneity within each Chironomidae subfamily, we expected that trait information gathered at higher or lower level of taxonomic resolution would determine differences in traits patterns gathered within each subfamily. To determine whether our database was actually providing additional assessment information, we contrasted trait patterns given by our database at the genus level with that obtained at the subfamily-level in the trait database of Tachet et al. (2010), which is commonly used in bioassessment studies. Finally, assuming that heritable traits (Eltonian) of organisms could disclose evolutionary processes operating among taxa, Chironomidae subfamily traits relatedness was expected to reflect their phylogenetic distances across subfamilies. Therefore, we compared the subfamily Eltonian trait relatedness with the most accepted Chironomidae phylogeny found in literature (Saether, 2000;Cranston et al., 2010Cranston et al., , 2012.
Our database includes the most widespread European Chironomidae species, covering a wide geographic area, different categories of water bodies at different altitudes and latitudes. Lotic and lentic freshwater systems were given equal importance, being mentioned in at least 20% and 19% of total references used, respectively. References covering temporary freshwater systems and hygropetric habitats were also included. Brackish habitat references were also used to support the trait salinity preferences. Other references used did not focus on a specific type of aquatic habitat but addressed ecological, physiological, morphological and/or life history characteristics of specific taxa. Whenever possible, the references for which species were first described in Europe were exploited. Information gathered from publications between 1931 and 2013 (ca. 150), including articles, books and a few PhD theses, were used to describe the species traits.
The initial list was composed of 21 traits and 110 categories of biological, physiological traits and ecological requirements, as used in Tachet et al. (2010); some traits and categories were adapted given the type of information available for Chironomidae (Table 1). A set of 16 additional traits specific to Chironomidae larvae included respiration (number of tracheas), tube construction, number of eggs per egg mass, flight period, duration of emergence, distance of aquatic and/or aerial dispersion, hibernation stages, length of larval development, oxygen saturation preferences, presence/absence of haemoglobin, migration type, depth preferences, optimal temperature interval for emergence, chlorinity, and general/gross habitat ( Table 1). Traits that differed among Chironomidae life stages were gathered for the fourth larval instar (except for number of eggs per egg mass, flight period, and others) and categorised into Grinnellian or Eltonian traits according to the terminology of Devictor et al. (2010) and Mondy and Usseglio-Polatera (2014). Grinnellian traits are related to taxon requirements and performance over a range of environmental conditions considering biotic and/or abiotic resources (e.g. pH, temperature, and food preferences), whereas Eltonian traits focus on the impact of the species on its environment, emphasising their functional role in the ecosystem rather than their response to particular resources (e.g. body size, voltinism, feeding habits).
Following Franquet (1996), the affinity of species or genera to trait category was quantified using the number of references citing this category for a given taxon. The higher the number of references associating a taxon to a trait category, the greater the affinity of that taxon to that particular trait category. Taxa with no available information on a trait were scored 'zero' for all categories, and were treated as missing values, being replaced by the mean of all taxa having information for a given trait category. Trait-affinity scores were further treated as frequency distributions and standardised to sum 1 for a given taxon-trait combination, to give the same weight to each taxon and to each trait in further analyses. This procedure is known as fuzzy coding (Chevenet et al., 1994).
Total number of genera described per trait was estimated to define the best described traits, i.e., with information gathered for more than 50% of the European genera. The genus trait database is provided as Supplementary data with the list of the references used to extract trait information and the list of species used to describe each genus.

Comparison between the two databases
To determine whether our trait database built at the species and genus levels involved different distributions of taxa (subfamily, tribes) compared to the database of Tachet et al. (2010), we used Fuzzy Correspondence Analysis (FCA) that enables the joint ordination of taxa and trait categories (Chevenet et al., 1994). FCA uses a matrix (n × p) to interpret the relationships between trait categories (p) and resemblances among individual taxa (n). The affinity profile of each trait category among taxa enables the positioning of each trait category at the weighted average of taxa that uses this category. The variance of these positions corresponds to a correlation ratio (i.e. the highest the correlation ratio the highest the separation of taxa across trait categories) and FCA maximises the average correlation ratio across traits when FCA was performed separately on Grinnellian and Eltonian traits. For comparison with the European trait database of Tachet et al. (2010) (hereafter TDB -Tachet DataBase), the fuzzy information of our database at the genus level (hereafter GDB -Genus DataBase) was averaged at the subfamily and tribe levels. Afterwards, these average affinity scores were rescaled so that their sum, for each of these coarser taxonomic groups for a given trait, equals one. Thereby, traits were described at the same scale for all different taxonomic levels of resolution. While the biological information in Tachet et al. (2010) describes only the Podonominae, Tanypodinae, and Orthocladiinae subfamilies and the Chironomini and Tanytarsini tribes, our database included Table 1 Traits and their categories and codes used in the European Chironomidae database. Eltonian and Grinnellian traits are ordered with (1) traits coded by the first author of this paper, (2) traits adapted from the European database of Tachet et al. (2010) and (3)   additional tribes: Pseudochironomini (Chironominae), Diamesinae, Telmatogetoninae, Buchonomyiinae, and Prodiamesinae.
Finally, to assess the variability in community trait composition explained by the difference between GDB and TDB, we computed between-class variance (with class as type of database; see Dolédec and Chessel, 1987;ter Braak, 1988) and tested its significance against simulated values obtained after 999 permutations of the rows of the trait-composition arrays.

Chironomidae subfamily trait relatedness
FAC was performed on Eltonian traits of genera averaged at the subfamily level. The resulting FCA coordinates of the 8 subfamilies along the 7 axes (n − 1; in which n is the smallest rank of the trait matrix; here, the number of subfamilies) was used to yield the Euclidean distance matrix among subfamilies. Finally, neighbourjoining (Saitou and Nei, 1987;Studier and Keppler, 1988) allowed estimated a tree among subfamilies. Bootstrap procedure was used to assess tree's accuracy and the 'confidence' of each tree bipartition (Efron et al., 1996). This representation was visually compared with the most accepted evolutionary relationships of Chironomidae subfamilies derived from cladistics analysis (Saether, 2000) and molecular analysis (Cranston et al., 2010(Cranston et al., , 2012. Statistics and graphical outputs were computed with the ade4 (Thioulouse et al., 1997;Chessel et al., 2004;Dray et al., 2007) and ape libraries (Paradis et al., 2004;Paradis, 2012) implemented in R freeware (R Development Core Team, 2013).

European freshwater Chironomidae trait database
Our final list contained 178 Chironomidae genera and 744 species distributed among 8 subfamilies. Biological information on species and genera was found in the literature for ∼59% of the most widespread European species, and 92% of the European genera for 37 traits (Table 2 and see in the information supplied as Supplementary data). From all of the gathered trait information, 11 Grinnellian and 4 Eltonian traits had information for more than 50% of the European Chironomidae genera present in the database (indicated in Table 1). The best described Grinnellian traits were: transversal distribution in streams and general/gross habitat preferences (>90% of genera present in the database). Food types, pH tolerance, salinity preferences, longitudinal distribution along streams, altitude, substrate preferences, current velocity preferences, oxygen and depth preferences were described for 53-74% of the genera. The best described Eltonian traits (for more than 95% of genera) were: maximal size of the fourth larval instar and type of aquatic stages. Potential number of generations per year (voltinism) and flight period were described for 50-53% of genera.
Chironomidae subfamilies with less information were the Buchonomyiinae, Podonominae and Telmatogetoninae. Tanypodinae, Orthocladiinae and Chironominae subfamilies had also  Fig. 1a-c), whereas for other traits, all Chironomidae behaved in a very similar way (e.g. altitudinal preferences, pH tolerance, number of generation per year, Fig. 1d-f). Tanypodinae, Orthocladiinae and Chironominae that contained the highest diversity of genera described (ca. 89% of all Chironomidae) covered wide ecological amplitude. Other less diversified subfamilies were associated with specific environments: Buchonomyiinae (Buchonomyia thienemanni) were recorded at low altitudes and in lotic habitats, whereas Diamesinae had a higher affinity for upper reaches (e.g. kryon, reaches fed by ice-melt) with higher current velocities and water temperature <15 • C. Telmatogetoninae were well represented in brackish and marine habitats but there is a substantial lack of information about their species traits. Prodiamesinae were generally recorded in sites with a heavy load of organic pollution. Podonominae were mostly represented in ponds and pools, temporary habitats, and marshes and bogs.
Projecting the genus trait information against the taxonomic tree showed a great variety of traits within each Chironomidae subfamily, exemplified by two subfamilies and two traits in Fig. 2. For example, Chironominae (Chrn in Fig. 2) and Diamesinae (Dmsn in Fig. 2) had genera with affinities for contrasted trait categories. For example within Chironominae, Axarus sp. and Chironomus sp. had high affinities for the large size categories (SIZE 4 and 5, Fig. 2) whereas Kloosia sp. and Lauterborniella sp. had high affinities to small size categories (SIZE 2; Fig. 2). Similarly, for food type, within the Chironominae subfamily Demeijerea sp. and Demicryptochironomus sp. had high affinities for animal food (MICINV and MACINV; Fig. 2)) whereas Paratendipes sp. and Pagastiella sp. had high affinities for plant debris (DEBRI1 and 2; Fig. 2) and live microphytes (MICPHY; Fig. 2). Within the Diamesinae family, despite its lower  Table 1 for acronyms) with (a) food type, (b) salinity preferences, (c) longitudinal distribution, (d) altitudinal preferences (e) pH tolerance, and (f) number of generations per year. species richness in comparison to other subfamilies, affinities could also vary within the same trait. For instance, the Diamesa sp. larvae showed affinities from small to large size categories (SIZE 2 to SIZE 4; Fig. 2), whereas Protanypus sp. showed larger sizes (SIZE 4; Fig. 2). Considering food types, Diamesa sp. generally consume living microphytes such as diatoms (MICPHY; Fig. 2) whereas Potthastia sp. feed on detrital particles (DEBRI1 and 2; Fig. 2).
Diamesinae subfamily also includes genera with a wider spectrum of food preferences (Protanypus sp., Boreoheptagyia sp.). Similarly, Chironominae subfamily includes opportunistic genera able to feed on almost any food item (e.g. Chironomus sp. and Glyptotendipes sp.) and genera that live in woody microhabitats and introduce wood in their diets, being true wood miners with the ability to digest wood fibres (e.g. Stenochironomus sp.; WOOD).  Table 1 for acronyms) for two subfamilies (Chironominae and Diamesinae). The information is presented against a taxonomic tree with two taxonomic levels: l1 (tribe) and l2 (subfamily). Chironominae (I2 Chrn) are represented in the top leaf of the cluster by two tribes (Chironominii, I1 Chrn; and Tanytarsinii, I1Tnyt). Diamesinae are represented in the below leaf by three tribes (Boreoheptagyiini, l1 Brhp; Diamesini l1 Dmsn; and Protanypodini, l1 Prtn). The size of each square is proportional to the frequency of the corresponding trait category (on top) for a given genus (at right).

Comparison between the two databases
FCA performed on Grinnellian traits showed low but significant differences between the two databases (15% of variance explained; simulated p-value = 0.001; Fig. 3a, Table 3). Transversal distribution along the stream channel, pH, and to a lesser extent food type, were more important contributors for the difference among databases. Substrate preferences and transversal distribution each explained more than 30% of variance over first axis (32% and 40% respectively), Table 3 Correlation ratios of Grinnellian traits for Chironomidae subfamilies/tribes (from two databases, the new developed in this work and existing in Tachet  whereas altitude preferences, food type and longitudinal distribution explained 13%, 14% and 18% of the variance, respectively. Transversal distribution also had a high contribution to explain distribution over the second axis (explaining 27% of the variance), whereas temperature and pH preferences were also relevant (13% and 11% variance explained, respectively). FCA performed on Eltonian traits likewise revealed significant differences between databases (32.7%, simulated p-value = 0.002; Fig. 3b, Table 4) to which reproduction type, resistance form and, Table 4 Correlation ratios of Eltonian traits for Chironomidae subfamilies/tribes (from two databases, the newly developed in this work and the existing one from Tachet et al., 2010), on the first-two axes of the fuzzy correspondence analysis and respective eigenvalues.

F1 F2
Size (  Eltonian traits of the two databases, considering the information existing at the subfamily level and few tribe levels for Chironominae. The information is grouped by database with GDB, the European database developed using the genus information compiled at the subfamilies and tribe levels; and TDB, the database from Tachet et al. (2010) at the subfamily/tribe level. Subfamily or subfamily/tribe level is identified by their 5 first letters (Bucho -Buchonomyiinae; Chiro -Chironominae; ChirC -Chironomini; Diame -Diamesinae; Ortho -Orthocladiinae; Podon -Podonominae; Prodi -Prodiamesinae; Tanyp -Tanypodinae; ChirT -Tanytarsini; Telma -Telmatogetoninae); followed by a G for GDB plot, or a T for the TDB plot. Ellipses include 80% of the points for readability.
to a lesser extent, life cycle duration had the highest contributions. Resistance forms and reproduction type explained 79% and 76% of the variance, respectively, considering the first axis, followed by life cycle duration, which explained 33% of the variation. The variance along axis 2 was explained by voltinism (43% variance explained) and to a lesser extent by the substrate relation, explaining 10% of the variance.

Chironomidae subfamily trait relatedness
The neighbour-joining analysis performed on the 20 Eltonian traits revealed the subfamilial trait similarity among Orthocladiinae and Chironominae on one hand, and Podonominae with Tanypodinae on the other hand (Fig. 4). The analysis also showed trait similarity between Diamesinae and Prodiamesinae with Tanypodinae and Podonominae segregating them from Orthocladiinae and Chironominae.
The accuracy of the tree assessed through the bootstrap analysis give confidence to the group formed by Orthocladiinae and Chironominae, with 100% of trees showing the same combination. All other nodes and bipartitions do not reveal a strong confidence, with confidence values below 0.44 revealing no trait relatedness signal.

Discussion
Most studies in which Chironomidae were used at higher taxonomic resolution than subfamily or tribe are historically associated to lakes, either considering subfossil Chironomidae assemblages for paleolimnological studies or the analysis of communities in deeper zones (Raunio et al., 2011). In running waters, the extensive use of Chironomidae in bioassessment is still a matter of debate. Some authors have suggested that assessments may be more efficient by eliminating Chironomidae from the protocols and by using resources for analysing additional sites (Hawkins and Norris, 2000;Rabeni and Wang, 2001). Some authors fully agreed with the family-level and its ability to detect impairment (Móra et al., 2008), while others have even strongly recommended the use of a finer level of taxonomic resolution for Chironomidae, showing that family level yielded much weaker assemblage-environment relationships, which emphasised the risk of reducing accuracy in bioassessment (King and Richardson, 2002).
Here, we defend the hypothesis that Chironomidae could be appropriate indicators of environmental conditions, as the same taxonomic group includes tolerant (e.g. Chironomus) and sensitive (e.g. Diamesa spp.) taxa to human impacts King and Richardson, 2002;Lencioni et al., 2012). One main problem for bioassessment purposes arises from the difficulties Fig. 4. Trait relatedness tree estimated among Chironomidae subfamilies using neighbour-joining, given the Euclidean distances of their FCA coordinates. Values associated to each node represent the percentage of partitions present in bootstrap trees. of taxonomic identification and the poor knowledge on traits at the genus or species level, contrary to other groups of freshwater invertebrates (e.g. Poff et al., 2006;Tachet et al., 2010). Aiming to fill this gap, information on 37 Chironomidae traits and 184 trait categories was compiled in this study. The number of species and genera covered by our database (59% and 92% respectively) highlights the great effort that is still needed to understand the behaviour, physiology and ecological tolerances of Chironomidae species. Considering the Eltonian traits only, the development of the database clearly showed the poor information available in the literature as only 4 of this biological type of traits were characterised for more than 50% of the European genera. Our database thus identifies the genera and species to which more attention should be given in future studies due to reduced available information. One of the reasons for the reduced and uneven information on Chironomidae traits is the fact that morphological and physiological studies typically focus on Chironomus species (e.g. C. tentans and C. riparius), because they are easy to keep in the laboratory and use in routine ecotoxicological tests (Ankley et al., 1994;Armitage et al., 1995;Penttinen and Holopainen, 1995).
A total of 16 European genera belonging to Diamesinae, Orthocladiinae and Chironominae subfamilies lacked information (Supplementary data). The inability to characterise these genera may be due to their limited distribution so far (e.g. Molleriella, Neobrillia) and their small number of species in Europe (e.g. Baeotendipes noctivagus, Nilomyia aculeata) or by their relatively recent or very recent discover (e.g. Olecryptotendipes Zorina 2007; Arctosmittia Zelentsov 2006). One advantage of our database is that any additional information available on references not used in the original dataset can be simply added to the information provided in Supplementary data.
Compared to pre-existing information (i.e., from Tachet et al., 2010), the data that we compiled at the genus level resulted in significant differences in the separation of Chironomidae subfamilies. This suggests that differences in specialisation among Chironomidae occur primarily at higher levels of taxonomic resolution (genus and species). Even at the genus level, generalisation should be carefully considered since environmental requirements, life history traits, and sensitivity to anthropogenic pressure may vary considerably within a genus (Rossaro et al., 2006;Lencioni et al., 2007). A high number of species per family in many aquatic environments usually suggests an extensive adaptive radiation by diversification of ancestral species into several ecologically different species by adaptive morphological, physiological and/or behavioural divergence in those environments limiting the utility of the family level in bioassessment (Hawkins and Norris, 2000;King and Richardson, 2002).
It is common to find autoecological information at the subfamily level mentioning faunistic patterns along environmental gradients (e.g. longitudinal, altitudinal) such as the greater abundance of Diamesinae and Orthocladiinae upstream, giving place to Tanypodinae and Chironominae downstream (Prat et al., 1983;Bitušík et al., 2006;Lencioni et al., 2007). Averaging trait affinities of Chironomidae subfamilies showed that they were distinct from each other considering some traits (e.g. maximal body size, food type); although the great trait diversity within each subfamily suggests that the subfamily level operating in the database of Tachet et al. (2010) is not appropriate. The latter database points out the ecological redundancy in the Chironomidae family and subfamilies, which may be simply due to the averaging operation, that masks the real trait diversity of Chironomids. Such false redundancy was highlighted by others (e.g. Lenat and Resh, 2001) and may compromise the results of studies that attempt to recognise Chironomidae faunistic patterns using a lower taxonomic level.
The differences between the subfamily/tribe trait patterns gathered at low and high levels of taxonomic resolution (TDB and GDB respectively) were clear for pH tolerance, transversal distribution in the river channel, reproduction types, resistance forms, and, to a lesser extent, food types and life-cycle duration. Food type is often considered a key factor in the distribution of Chironomidae species along with temperature. Additionally, flow regime and pH also have indirect influence on their distribution by regulating food availability, quantity and quality (Lencioni et al., 2007;Vallenduuk and Pillot, 2007). Consequently, differences in these traits can compromise multiple trait-based assessments.
A given set of traits in the organisms of the same species result from the process of evolution and adaptation to specific environmental conditions. Thus, it is generally accepted that there is a link between taxa phylogenetic relatedness and the traits they possess (Kraft et al., 2007). Usseglio-Polatera et al. (2000) recognised that traits related to morphology, physiology and life history (Eltonian traits) appeared to be more constrained by phylogeny than traits related to behaviour and habitat preferences (Grinnellian traits). Therefore, we expected that the most recently diverged subfamilies would tend to share more Eltonian trait categories among their taxa than subfamilies that diverged a long time ago from the Chironomidae common ancestor. The tree estimated by neighbourjoining revealed a small trait distance between Chironominae and Orthocladiinae subfamilies, which is in agreement with cladistics (Saether, 2000;Fig. 5a) and molecular phylogeny studies (Cranston Fig. 5. Subfamily relationships among Chironomidae subfamilies given by: (a) cladistic analyses using parsimony of morphological characters of adults, pupae and larvae (Saether, 2000) and (b) and (c) molecular phylogenies by Cranston et al. (2010Cranston et al. ( , 2012Cranston et al. ( ), respectively. et al., 2010Cranston et al. ( , 2012Fig. 5b and c). In fact, Orthocladiinae retain some ancestral traits (e.g. respiration type given by the number of tracheas), which slightly differentiates the two subfamilies, whereas the presence of other much more recent traits (e.g. presence of haemoglobin) bring together the two subfamilies.
The fact that Eltonian trait patterns do not necessarily reflect phylogenetic relationships of the subfamilies means that environmental drivers are operating differently in species leading to higher functional diversity, exposing the labile nature of traits. Several authors argued that more closely related taxa may not be ecologically similar since the same ecological function can evolve through different pathways depending on the environmental drivers operating in the habitats, often called trait lability through evolutionary time (Webb et al., 2002;Poff et al., 2006;Cavender-Bares et al., 2009). According to Poff et al. (2006), multiple trait-based approaches should precisely take advantage of the selection of traits relatively unconstrained by phylogeny (i.e. more evolutionary labile), with low statistical and phylogenetic correlations, and more responsive to local selection, such as voltinism, which tell more about the drivers and environmental filters that operate in the systems than about the history of the taxa.

Conclusions
Among lacustrine macroinvertebrates, Chironomidae have been well studied and pointed as a powerful paleo-environmental indicator when using preserved subfossil assemblages collected from lake sediments. The value of Chironomidae as an indicator is not only associated to its wide distribution or community composition, but also to potential morphological responses to changes in environmental conditions such as exposure to contaminants. Despite its demonstrated importance and ecological role, in many freshwater studies (e.g. springs, streams, littoral of lakes) Chironomidae are still disregarded or neglected with their identification kept at family/subfamily levels. This has been limiting a more extensive use of Chironomidae in biomonitoring and the knowledge about autecology of taxa therein. Our study shows that Chironomids are indeed a quite diverse group with different ecological requirements and characteristics, and if used at the genus or species level, they have the potential to improve the signals provided by ecological assessment tools, either in taxonomic-based structural assessments or in indirect functional assessments using multiple-traits-based approaches. To further prove these insights, tests comparing both types of assessments based on sub-family level and genus/species level are needed. Our database, which is the first comprehensive European database for Chironomidae traits at the genus level to the best of our knowledge, can be used for that purpose, as well as in ecological studies on functional patterns of freshwater systems, especially those including habitats that are traditionally considered less diverse.