Introduction

The community structure (taxa distribution and abundance) of benthic invertebrates is commonly used in the ecological assessment of rivers (Oliveira and Cortes 2006; Dolédec and Statzner 2010; Hawkins et al. 2010; Chang et al. 2014). Within benthic macroinvertebrates, the importance of Ephemeroptera, Plecoptera, and Trichoptera (EPT) taxa is often emphasized on the premise that high-quality streams generally host high EPT richness. However, in some rivers, EPT richness can be naturally reduced due to specific hydrological conditions such as those prevailing in Mediterranean temporary rivers (Borer et al. 2006, Bonada et al. 2007). In contrast, Chironomidae are usually disregarded mainly due to taxonomic challenges, in spite of constituting an important fraction of the macrozoobenthos in most of freshwater systems and often the only insects present in naturally poorly diverse streams (Hawkins and Norris 2000). In Mediterranean streams, Chironomidae assemblages are known to shift quickly, due to their high resistance and resilience, high fecundity and relatively short life cycles, and dispersion capabilities; as a result, Chironomidae are often the first colonizers after periods of drought or flood (Langton and Casas 1999; Calle-Martínez and Casas 2006; Puntí et al. 2007, 2009; Marziali et al. 2010). Due to the presence of hemoglobin in several species (e.g., Chironomus plumosus, Chironomus riparius), which confer them high tolerance to low oxygen concentrations, the family is often used as an indicator of bad water quality (Moller Pillot 2009). Consequently, Chironomidae indicating clean water have not been incorporated in Saprobic systems or received low scores (poor sensitivity) in biological indices such as the BMWP (Hawkes 1998) despite their known sensitivity to anthropogenic disturbance (Saether 1979; Wiederholm 1981; Seire and Pall 2000; Lencioni et al. 2012) and attempts to develop a Chironomidae-based index (Lindegaard 1995).

The growing need for tools that provide not only a structural assessment of aquatic communities but that also link patterns to processes gave rise to multiple trait-based approaches (Statzner et al. 2001; Statzner and Bêche 2010; Menezes et al. 2010). Consequently, the ability of macroinvertebrate communities’ traits to detect of various kinds of impairment has been evaluated in aquatic ecosystems including multiple interacting stressors (e.g., Charvet et al. 1998; Vieira et al. 2004; Feio and Dolédec 2012; Lange et al. 2014; Mondy and Usseglio-Polatera 2014; Greenwood and Booker 2016). In parallel, studies defining states or strategies as a combined product of evolution emphasized that traits are not evolutionary independent (Poff et al. 2006; Verberk et al. 2008a, 2013). To date, few studies have paid attention to the Chironomidae traits at a fine identification level in the context of stream environmental monitoring (but see Franquet 1996; Van Kleef et al. 2015; Serra et al. 2016). In the study of Van Kleef et al. (2015), Chironomidae were assigned to some life history strategies (LHS), based on previous works establishing interrelationships between macroinvertebrates traits and their adaptive values in a variety of types of water bodies (Verberk et al. 2008a,b; Verberk et al. 2010). Solutions presented by taxa to different ecological problems should be reflected in the abundance of taxa exhibiting distinct LHS, but in a same water body it is not expected to find only one type of solution (Verberk et al. 2008b). This approach allows for the comparison of water bodies geographically distant (Verberk et al. 2008a).

In the present study, we hypothesize that (1) Chironomidae taxonomic and trait composition can be a useful indicators of anthropogenic disturbance in typical Mediterranean streams (naturally poor in other taxonomic groups) and that (2) traits can provide insights on the changes in the ecosystem functioning promoted by stressors. To test these hypotheses, we assessed the ability of Chironomidae taxonomic and trait composition to discriminate least-disturbed from disturbed samples. For that, we used data collected in six sites (three disturbed and three least-disturbed) in a Mediterranean stream over 3 years. We established and tested also a priori hypotheses on the direction of changes in the frequency of individual Eltonian traits in the presence of multiple stressors (Table 1). Eltonian traits include biological traits related to organisms’ functional role and their impact in the ecosystems including life cycle aspects and physiological and behavioral characteristics (Devictor et al. 2010; Mondy and Usseglio-Polatera 2014). In addition, we tested the utility of combination of traits, aggregated in LHS (as proposed by Van Kleef et al. 2015), to discriminate disturbed and least-disturbed sites starting from predictions on which LHS would be more affected by the stressors affecting our sites. This approach is expected to reflect how the individual traits are linked and which combinations favor Chironomidae, in the presence of disturbance.

Table 1 Predictions of changes in Chironomidae trait proportions in disturbed sites affected by multiple stressors (nutrient increase, oxygen depletion, sedimentation and hydromorphologic changes, habitat homogenization)
Full size table

Methods

Study area, sites, and environmental characterization

The study sites are located in the south of Portugal, Alentejo, a typical Mediterranean-climate area characterized by dry summers, irregular but intense rainfall in winter, and total annual precipitation below 600 mm year−1 (Fig. 1). Stream sites (three least-disturbed, L, and three disturbed, D) were selected from two similar temporary streams at altitudes between 200 and 250 m. Disturbed sites, which are located downstream from a discharge of the effluent of draining mining waters product of water treatment plant that processes the wastewater, have high conductivity, nitrate and sulfate concentrations, and low dissolved oxygen concentrations but it contained no metals (Feio et al. 2010). In addition, disturbed sites are subjected to hydromorphological alterations including fine sediment deposition whereas least-disturbed sites are characterized by a high diversity of habitats (sand, cobles, and macrophytes; Feio et al. 2010). In summer, all sites have low flow to intermittent, according to the typical Mediterranean streams. Yet, the flow in the first sites located downstream to the mine may receive slightly more flow due to the effluent discharge. Water physical and chemical parameters (pH, dissolved oxygen, conductivity, temperature, nitrates, and sulfates) were recorded in early spring of 2012, 2013, and 2014 at the six stream sites.

Fig. 1
figure 1

Location of the study area (square) located in southern Portugal

Full size image

Chironomidae sampling and identification

Benthic macroinvertebrates were collected in early spring of three consecutive years, from 2012 to 2014, using a kick-net (500 μm mesh size; 0.25 m × 0.25 m opening). Sampling followed a multi-habitat approach protocol, which covers all habitats (riffles, macrophytes beds, pools) proportionally to their abundance, and comprised a 6 × 1-m kicking (INAG 2008). Samples were fixed with formalin (4%) and Chironomidae were sorted in the laboratory and preserved in ethanol (70%). Chironomidae identification was performed sequentially. In a first step, individuals were grouped into morphological types and counted under a stereomicroscope. Morphotypes are smaller groups of taxa (usually groups of genera) within Chironomidae that share similar morphological features related, e.g., to body setae, head capsule color, antenna length, position, and number of eye spots (Prat and Rieradevall 2014). In a second step, Chironomidae were mounted on microscopic slides under the stereomicroscope. To do this, larval specimens were digested in caustic potash (KOH, 10%) at 85 °C in a water bath for 15–20 min; digested tissues were then washed in distilled water for about 5 min. Specimens were then dehydrated: 3 min in 70% ethanol, followed by 3 min in 96% ethanol. Heads and bodies were separated and mounted using Euparal medium (Andersen et al. 2013). When a morphotype gathered more than 200 individuals, only a representative portion was mounted (≥50%). In a last step, Chironomidae larvae were identified under the microscope (magnification ×400–1000) to the highest possible taxonomic resolution, frequently to genus but also species and species-groups level, following Paleartic/Holartic guides for Chironomidae taxa (e.g., Cranston 1982; Lencioni et al. 2007; Andersen et al. 2013; Prat and Rieradevall 2014).

We used genus level identification in data analyses because the available information on traits was predominantly available at that level. Exceptions included the genera of Cricotopus and Orthocladius, which represent particular morphological groups (groups of species) inside each genus rather than the genus itself. In this study, Cricotopus included taxa belonging only to the sylvestris, trifascia, and bicinctus groups, whereas Orthocladius included Orthocladius (Euorthocladius) rivulorum and Orthocladius (Symposiocladius) lignicola. All the other Cricotopus and Orthocladius were gathered in a wider morphological undistinguished group of three genera Cricotopus-Orthocladius-Paratrichocladius (CriOrtPar).

Trait composition

Chironomidae Eltonian traits were taken from Serra et al. (2016) genus level database (Table 1). Trait categories were standardized to sum 1 in each trait, following the fuzzy coding procedure (Chevenet et al. 1994), ensuring that all taxa have the same weight in further analyses. Trait-by-site arrays were further computed as the cross-product between the abundance (ln[x + 1]) of Chironomidae taxa and the standardized trait profile (see, e.g., Gayraud et al. 2003). As a result, trait-by-site arrays contained the relative abundance of each trait category in each D and L sites.

Life history strategies

Six Chironomidae LHS (A–F, see Table 2) have been proposed by Van Kleef et al. (2015) based on Eltonian traits that are expected to be relevant under environmental changes: hemoglobin, tube building, and larval development. These traits in turn interact with other traits such as feeding habits, voltinism, spring emergence (spring synchronization), and maximal body size. These are also relevant set of traits with adaptive value for Chironomidae in other environments. Therefore, Chironomidae taxa were assigned to each LHS based on individual traits coded in Serra et al. 2016 for European Chironomidae (Table 2). Based on these LHS, we made a priori predictions on the expected changes in their proportions in the D sites, compared to the L sites (see Table 2). We expected an increase in D sites of Chironomidae having hemoglobin (in D, E, and F strategies), constructing tubes (E), and a higher number of generations (A and B) because (1) the presence of hemoglobin allows Chironomidae to live in water with poor oxygen content and (2) the oxygen concentration inside the tube may increase significantly, compared with surroundings, due to the larval body movements and their opening above the level of sediments (Armitage et al. 1995). In other cases, the higher number of generations may compensate the absence of other adaptations to stress (e.g., hemoglobin) by the high number of individuals present in a site, resulting in a better chance of species survival and allowing organisms to attain adult size and dispersing to more favorable environments earlier, avoiding adverse conditions. According to this rationale, we expected that, in D sites, the proportions of individuals with (1) LHS-E and A should increase and (2) the proportions of individuals with LHS-C should decrease. Genera that could not be assigned to a LHS category were included in an unknown category (LHS-U) due to the absence of data about their biological traits, the existence of various affinities to different categories (high variability to given trait), or because the combination of traits in the genus did not fit in any defined strategy.

Table 2 Chironomidae life history strategies (LHS) based on different combination of biological traits and predictions made for disturbed sites in comparison to least-disturbed sites
Full size table

Data analysis

Environmental characterization

To confirm the discrimination of sites regarding their levels of anthropogenic disturbance (i.e., D vs. L sites), we performed a normalized principal component analysis (PCA) on the abiotic data. To assess the degree of discrimination between D and L sites, a between-class analysis was used (BCA-PCA, class: disturbance level, L vs. D; see, e.g., Dolédec and Chessel 1987; ter Braak, 1988; Lebreton et al. 1991; Chessel et al. 2004). The significance of the explained variance was tested against simulated values obtained after 9999 permutations of the rows of the normalized environmental arrays (Monte-Carlo test; see, e.g., Romesburg 1985; Romesburg and Marshall 1985).

Taxonomic and trait composition

Patterns of richness and abundance in D and L sites were explored and significant differences were accessed by a Kruskal-Wallis rank sum test. Permutational Multivariate Analyses of Variance (PERMANOVA) were initially performed on multivariate data to assess significant dissimilarities (Bray-Curtis coefficient) in taxonomic (ln[taxa abundance + 1]), trait (trait-by-site arrays containing the relative abundance of each trait category), and LHS composition (ln[LHS taxa abundance + 1]) between disturbance level and years (two factors) to evaluate possible temporal differences in communities.

To analyze the segregation of sites according to their level of disturbance based on species abundance, we performed a PCA on the abundance (ln[x + 1]) of Chironomidae taxa with a between-class analysis (BCA-PCA, class: disturbance level, L vs. D) followed by a Monte-Carlo test to assess the significance of the variance explained by the difference (simulated values obtained after 9999 permutations of the rows of the taxa composition arrays).

In a following step, we used a SIMPER analysis (Bray-Curtis similarity) to determine which Chironomidae genera were representative of L and of D sites, i.e., contributed the most to the within-group similarity (up to 90% cumulative contribution), assuming that those taxa should be best adapted to one or the other level of disturbance due to their traits. As abundance data did not add more information, we selected those taxa contributing the most to groups similarity based on presence/absence data.

We computed Rv coefficient (the multivariate equivalent of a R 2 for tables; Robert and Escoufier 1976) to measure the correlation between each pair of trait-by-taxa array. This allowed eliminating possible redundancy among traits and testing for potential correlations between biological and morphological traits. Biological traits with correlation ≥0.85 were removed from further analysis. Based on those selected taxa, we performed a PCA on trait-by-site arrays considering the transformed relative abundance of trait categories (ln[x + 1]) to yield multivariate scores of D and L sites (Cundari et al. 2002; Sârbu and Pop 2005). To test for the degree of discrimination between D and L sites considering Chironomidae taxonomic and trait composition, a between-class analysis (BCA-PCA, class: disturbance level, L vs. D) was used. The significance of the explained variance was tested against simulated values obtained after 9999 permutations of the rows of the trait composition arrays (Monte-Carlo test).

Finally, considering trait categories that contributed more to the differentiation among D and L, differences between D and L sites were assessed for each trait category using a Kruskal-Wallis rank sum test. Trait categories’ individual differences of traits that contributed more to the differentiation among sites with different level of disturbance in the BCA-PCA were assessed by performing a Kruskal-Wallis test.

Life history strategies

Taxa of LHS-U, corresponding mainly to the group CriOrtPar (which account for taxa inside the genera Cricotopus, Orthocladius, or Paratrichocladius), were not further analyzed, given the diversity enclosed in this group. Taxa in LHS-U corresponded to 13% of the total abundance.

For the remaining six LHS, we tested our predictions by performing a PCA on Chironomidae transformed abundance (ln[x + 1]) in each LHS in disturbed and least-disturbed sites, followed by a between-class analysis (BCA-PCA) and a Monte-Carlo test on the between-groups inertia (as in previous section). The LHS contributing the most to the difference among levels of disturbance in the BCA-FPCA were assessed by performing a Kruskal-Wallis test.

All statistics and graphical outputs were computed using R freeware (R Core Team 2015). Taxa and trait compositions were specifically analyzed with vegan (Oksanen et al. 2015) and ade4 packages (Thioulouse et al. 1997; Chessel et al. 2004; Dray et al. 2007a, b), while graphical outputs were obtain using ggplot2 and factoextra packages.

Results

Environmental characterization

The first-two axes of a PCA performed on environmental data explained 73.8% of the total variance (56.9 and 16.9% of the total variance, fort the first and second axis, respectively, Fig. 2). PCA axis 1 segregated disturbed from least-disturbed sites. Disturbed sites had higher SO4 2− concentrations (701.3 ± 164.8 vs. 33.3 ± 4.5 mg l−1; mean ± SEM), higher conductivity (1682.9 ± 320.4 vs. 356.9 ± 71.2 μS cm−1), and higher NO3+ concentrations (10.4 ± 2.6 vs. <0.2 mg l−1), whereas percentage of dissolved oxygen was higher in the least-disturbed sites (90.1 ± 3.7 vs. 63.6 ± 10.5%). A Monte-Carlo test revealed a highly significant variance of environmental data explained by the disturbance (variance explained = 48.0%, simulated P = 0.001).

Fig. 2
figure 2

First-two axes PCA biplot according to abiotic conditions in disturbed (D; black circles) and least-disturbed (L; gray triangles) sites and eigenvalues associated to each component. The largest triangle and circle show the center of each group of sites

Full size image

Taxonomic composition

A total of 2738 Chironomidae individuals, from 35 taxa, were identified, constituting 26% of the total abundance in macroinvertebrate communities. The only group more abundant than Chironomidae was the Hemiptera (44% of the total abundance) but mainly due to the high abundance of one taxon (Micronecta sp.). Chironomidae were more abundant in disturbed sites (P = 0.016; Fig. 3), but richness was not significantly different from least-disturbed sites (P = 0.807; Fig. 3). Taxonomic composition was significantly different between disturbance levels (PERMANOVA; P = 0.004, based on 9999 permutations; Table 3) but not between years nor in the interaction between disturbance level and year. A total of seven taxa (Ablabesmyia, Cricotopus, CriOrtPar, Dicrotendipes, Paratanytarsus, Rheocricotopus, Sympotthastia) showed a higher abundance in L sites compared than the simulated expected value (if the abundance registered in L sites was uniformly distributed by all taxa). In D sites, only four taxa are above this theoretical value (Chironomus, Cricotopus, CriOrtPar, Paratanytarsus).

Fig. 3
figure 3

Chironomidae abundance and richness in disturbed (D, black) and least-disturbed (L, gray) sites. Kruskal-Wallis rank sum test results are shown as letters “A” and “B” associated to each box (different letters indicate significant differences)

Full size image
Table 3 Results of PERMANOVA analysis (factors: disturbance and year) for Chironomidae taxonomic composition (abundance ln[x + 1]), Chironomidae trait composition of the most contributive taxa (SIMPER: cut-off 90%), and Chironomidae LHS composition (ln[x + 1])
Full size table

The first-two axes of a PCA performed on Chironomidae composition explained 55.9% of the total variance (PC1 38.1% and PC2 17.8%; Fig. 4a). A Monte-Carlo test showed that 23.2% of the variance of the Chironomidae composition was significantly explained by the level of disturbance (simulated P = 0.001).

Fig. 4
figure 4

First-two axes (PC1, PC2) of a PCA performed on disturbed (D; black circles) and least-disturbed (L; gray triangles) sites and eigenvalues associated to each component, considering: a Chironomidae taxonomic composition (ln[x + 1]); only the taxa (N = 18) with higher contributions were projected as variables; b Chironomidae trait composition of the most contributive taxa (SIMPER: cut-off 90%) to the within similarity of L and D sites; only the 13 trait categories with higher contributions were projected as variables; and c Chironomidae LHS abundance (ln[x + 1]) of taxa from each strategy. The largest triangles and circles show the center of each group of sites

Full size image

A total of 11 taxa contributed to the similarity within least-disturbed sites (up to 90% of cumulative contribution), whereas seven taxa contributed to the within disturbed sites similarity (Table 4): five taxa (Potthastia, Cricotopus, CriOrtPar, Sympotthastia, Tvetenia) were common to L and D sites and six taxa (Corynoneura, Polypedilum, Rheocricotopus, Paratanytarsus, Ablabesmyia, and Tanytarsus) contributed exclusively to least-disturbed sites similarity, whereas two taxa (Chironomus and Eukiefferiella) were representative only of disturbed sites.

Table 4 Average abundance [ln(x + 1)] and percentage of taxa that contribute the most to the similarity within least-disturbed (L; with average similarity = 29.7%) and disturbed (D; average similarity = 66.2%) sites, considering Chironomidae presence/absence generated by a SIMPER analysis (cut-off level for cumulative contributions up to 90%)
Full size table

Trait composition

Trait composition differed between levels of disturbance (PERMANOVA; P = 0.028, based on 9999 permutations) but not between years or the interaction of the two factors (Table 3).

For the 13 selected taxa (Table 4), none of the traits were correlated (≥0.70), so all traits were kept in further analyses. The first-two axes of a PCA performed on the trait composition of the selected taxa explained 71.2% of the total variability (PC1 50.5% and PC2 20.7%; Fig. 4b). A total of 12 trait categories (from seven traits) showed contributions to the first-two axes higher than the expected: resistance form (RESDIA, RESNO), hibernation stages (HIBL2, HIBM2), hemoglobin (HBPRES, HBNONE), life cycle duration (LESS1Y, MORE1Y), locomotion (BURROWER), and tube construction (TUBNON, TUBPRE), voltinism (PLURIV). The BCA-PCA performed on the trait-by-site arrays showed low but significant differences among sites with different levels of disturbance (18.6% of variance explained by disturbance level, simulated P = 0.008), and the traits but contribute more to this differences (in decreasing order of importance), were: resistance form, life cycle duration, hibernation stages, and voltinism. The Kruskal-Wallis tests indicated which trait categories within these traits that significantly differed among the D and L sites (Fig. 5 and Table 5). L sites had more individuals that live less than 1 year (MORE1Y), have more than one generation per year or plurivoltine (PLURIV), without resistance forms (RESNO), and hibernation in less than two stages (HIBL2); contrarily, in D sites, there were more taxa living more than 1 year (MORE1Y), semivoltine (SEMIV), and with hibernation in more than two stages (HIBM2). Patterns of life cycle duration and voltinism were against our initial predictions (Table 1).

Fig. 5
figure 5

Chironomidae trait categories of selected taxa (SIMPER: cut-off 90%) showing significant differences between disturbed (D, black) and least-disturbed (L, gray) sites

Full size image
Table 5 Results of Kruskal-Wallis rank sum test used to test differences in Chironomidae trait categories considering traits that contribute more to FCA-BCA L and D sites segregation
Full size table

Life history strategies

LHS composition were significantly different between disturbance levels (PERMANOVA; P = 0.011, based on 9999 permutations) but not between years or the interaction between factors (Table 3c).

The first-two axes of the PCA on LHS composition explained 72.4% of the total variability (PC1 43.9% and PC2 28.5%; Fig. 4c). The BCA-PCA showed low but significant differences between D and L sites (Monte-Carlo test: R 2 = 17.5%, simulated P = 0.018). LHS that mostly contributed to these differences were LHS-B, followed by LHS-U and LHS-A, which were significantly different when assessed by the Kruskal-Wallis test: LHS-A significantly better represented in L sites and LHS-B and LHS-U significantly better represented in D sites (Fig. 6).

Fig. 6
figure 6

Abundance (ln[x + 1]) of taxa from each LHS, in disturbed (D, black) sites and least-disturbed (L, gray). Asterisk: LHS is significantly different between D and L sites (LHS-A, LHS-B, and LHS-U)

Full size image

Discussion

This study showed that both taxonomic (genus level) and trait compositions of Chironomidae are able to discriminate the degradation of a typical Mediterranean stream ecosystem. In our sites, it was more abundant in disturbed sites, and despite a similar number of taxa in least-disturbed and disturbed sites, there were compositional differences in Chironomidae assemblages: Diamesinae genera (e.g., Potthastia and Sympotthastia), Orthocladiinae (e.g., Corynoneura, Cricotopus), and Chironominae (e.g., attending to the Tanytarsini Tanytarsus and Paratanytarsus but also Chironomini Polypedilum) abundances differed between disturbed and least-disturbed sites and Chironomus sp. (Chironominae) was almost exclusive of D sites (only represented by one individual in one of the 9-L samples). This result agrees with previous studies that indicate the sensitivity of Chironomidae taxonomic composition to nutrient enrichment (Camargo et al. 2004; Maasri et al. 2008; Stewart et al. 2014). Chironomus (C. riparius) and Cricotopus (e.g., C. bicinctus) are also known to be tolerant and dominant in Mediterranean systems (Calle-Martínez and Casas 2006; Chaib et al. 2011). C. riparius has also been used as bioindicators of freshwaters through the evaluation of anatomical deformities (Servia et al. 2000; Stuijfzand et al. 2000; De Haas et al. 2005; De Haas and Kraak 2008; Moller Pillot 2009) and the life history traits of C. riparius (e.g., larval development time, hatchability of the eggs) have been shown to change in the presence of metal contamination, rapidly evolving in response to external conditions (Postma et al. 1995).

Indeed, the trait composition of the selected Chironomidae was also able to separate disturbed from least-disturbed sites (e.g., Dolédec & Statzner 2008, Van kleef et al. 2015). In agreement with our predictions, we found a higher proportion of taxa with no resistance form in least-disturbed sites and a relatively higher proportion of taxa with more than two stages with hibernation/diapause in disturbed sites. Diapause is a dormant state during Chironomidae development, which enables individuals to deal with the reduction in the amount of water in summer (drought) but also with temperature extremes or reduced food availability. The fact that taxa present in disturbed sites have more larvae stages undergoing through this dormant or inactive stages reveals the unfavorable conditions in these sites but also reveals that the combined effect of stressors will specially affect life cycles. At the ecosystem level, a larger proportion of animals in dormancy may result in a more reduced secondary production.

However, the direction of change was sometimes contrary to our predictions. We expected Chironomidae requiring more oxygen to be less abundant in disturbed sites. However, the presence of hemoglobin was not significantly different between the two levels of disturbance. Feeding traits were not sensitive to the disturbance. Thus here, depletion of dissolved oxygen and alteration of the primary food sources were not the main stressors affecting the Chironomidae roles in the ecosystem, as result of human activities. Also, Chironomidae with more than one generation per year and with shorter life cycles were counterintuitively favored in the better quality sites, whereas those with longer life span were more frequent in disturbed sites. This pattern might be caused by taxa with a high representativeness in disturbed sites, such as Cricotopus (55% of the total abundance of D sites), as some of the relevant traits are phenological, showing a higher plasticity from region to region depending on the environmental conditions. For example, the life cycle duration of this genus may have few or many but also one or less generations per year (Bazzanti et al. 1997; Casas and Langton 2008; Tokeshi 1995). Also the emergence season can change to confer a better adaptability to the external conditions to the genus Chironomus and Cricotopus, which colonize temporary habitats (Bazzanti et al. 1997; Tokeshi 1995). Depending on external factors (see, e.g., Moller Pillot 2009), the use of own phenological data would be ideal in trait studies, but given the scarcity of information, trait studies largely rely on the use of collections extended to relatively large areas like the one used here.

Regarding the trait combinations in life history strategies (Verberk et al., 2008a, b, 2013; Van Kleef et al. 2015), we observed that medium- and small-sized multivoltine taxa with a fairly rapid development, lacking strong spring synchronization and not investing in hemoglobin or tube (LHS-A and B), were generally predominant in all our Mediterranean sites In fact, these organisms are well adapted to highly dynamic environments, such as those found in Mediterranean streams, which justifies their high abundance. Yet, the B strategy (medium size animals) was better represented in disturbed sites than in better quality sites, while A strategy (small animals) decreased in disturbed sites. This indicates that disturbed sites are less environmentally dynamic than least-disturbed sites favoring the presence of distinctly larger final body sizes with a slower juvenile development, when compared to those of A strategy. In fact, the discharge (although small) of the industrial effluent may result in a more continuous flow over the year, disrupting the natural pattern of summer droughts and allowing the existence of organisms with longer life cycles and few generations per year. This along with the decrease in water quality and in habitat diversity leads to a functional homogenization of the Chironomidae community. Such trend was also found for the global invertebrate communities in a previous study in the same study area (Feio et al. 2010). Finally, the E strategy, unexpectedly, was not favored in disturbed sites. Contrary to strategies A and B, these organisms possess hemoglobin but that did not confer them a clear advantage in their ability to support the degraded conditions in disturbed sites. This may indicate that oxygen depletion is not the most disruptive factor for the communities functioning of these disturbed sites. In addition, a large investment in secondary production (through increase of body mass) by carnivores is also not supported by the system in disturbed sites, inhabited mostly by medium chironomids.

This study showed that the trait-based approaches allow the definition and assessment of their biological integrity while encompassing the functional integrity, contrary to the strictly taxonomic approaches, which only address structure. In spite of the similar results, traits have the potential to improve the mechanistic understanding of species-environment relationships, linking organizational structural patterns to processes, since the species are only present in a certain place if possessing biological traits that confer them the ability to cope with the environmental constraints guaranteeing their individual survival and reproduction (Poff 1997; Poff et al. 2006). In addition, and although not analyzed in our study, traditional biomonitoring tools are very dependent on the region for which they were developed while trait-based approaches could be used in large spatial scale comparisons because traits are expressed by many species being theoretically comparable among distant regions (Statzner et al. 2001; Bêche and Statzner 2009; Dolédec & Statzner 2010). Finally, predictability and the ability to discriminate between different types of changes are important criteria in the establishment of trait-based approaches as unequivocal ecological tools for functional assessment of the ecosystems (Bonada et al., 2006b; Statzner and Bêche 2010). As we could see in this study, some traits and even strategies showed some predictability under the anthropogenic disturbances. However, in a complex multiple interactive stressor scenario, the selection of traits and combination of traits that should have adaptive value is hindered. Yet, it still holds the potential of exposing trait patterns that disentangle cause-effect relationships, such as the life history traits and strategies which were important in the segregation between least-disturbed and disturbed sites, pattern also evident in other multiple stressor scenarios (Lange et al. 2014).

In this study, the abundance of specific taxa like Chironomus and Cricotopus and the undifferentiated group CriOrtPar had a great influence in the trait patterns of disturbed sites. Their relative large body sizes compared with other Chironomidae resulted in a pattern of larger body sizes in disturbed sites, against the initial predictions. This is also contrary to the lower proportions of resistance traits (large body sizes usually associated to more stable environments) and higher proportion of resilience traits (small body sizes generally associated to unstable environments) found by Diaz et al. (2008) in Mediterranean streams under anthropogenic disturbance. Our counterintuitive results may have resulted from the use of trait databases that can be seen as collections of taxa “fundamental” traits (paralleling with the fundamental niche), where trade-offs, plasticity, and intraspecific plasticity are not represented. A way to avoid this could be the use of morphological traits measured at the individual level in a continuous scale. These traits that can be clearly related to organisms’ performance and functions in the system should be further tested in different environments and under different stressors (Serra et al. 2017).

In conclusion, our study showed that Chironomidae taxonomic and trait assemblages at genus level were able to distinguish disturbance levels in Mediterranean streams. The most representative taxa in least-disturbed sites and in disturbed sites offered a set of traits that supports the use of life history strategies with Chironomidae. Therefore, independent analyses of traits and the use of combined-trait methods may complement each other: the first that may be important in the identification of a set of traits (regardless the direction of their individual response) that respond to a multiple stressor scenario, whereas the second reflects the linkage between those traits and others with which they interact, offering an integrated perspective of functional changes occurring in the system. Finally, due to the high abundances of Chironomidae, the changes in trait proportions as a result of anthropogenic disturbance may mirror changes in ecosystem functions.