Impacts of multiple anthropogenic stressors on stream macroinvertebrate community composition and functional diversity

Abstract Ensuring the provision of essential ecosystem services in systems affected by multiple stressors is a key challenge for theoretical and applied ecology. Trait‐based approaches have increasingly been used in multiple‐stressor research in freshwaters because they potentially provide a powerful method to explore the mechanisms underlying changes in populations and communities. Individual benthic macroinvertebrate traits associated with mobility, life history, morphology, and feeding habits are often used to determine how environmental drivers structure stream communities. However, to date multiple‐stressor research on stream invertebrates has focused more on taxonomic than on functional metrics. We conducted a fully crossed, 4‐factor experiment in 64 stream mesocosms fed by a pristine montane stream (21 days of colonization, 21 days of manipulations) and investigated the effects of nutrient enrichment, flow velocity reduction and sedimentation on invertebrate community, taxon, functional diversity and trait variables after 2 and 3 weeks of stressor exposure. 89% of the community structure metrics, 59% of the common taxa, 50% of functional diversity metrics, and 79% of functional traits responded to at least one stressor each. Deposited fine sediment and flow velocity reduction had the strongest impacts, affecting invertebrate abundances and diversity, and their effects translated into a reduction of functional redundancy. Stressor effects often varied between sampling occasions, further complicating the prediction of multiple‐stressor effects on communities. Overall, our study suggests that future research combining community, trait, and functional diversity assessments can improve our understanding of multiple‐stressor effects and their interactions in running waters.


| INTRODUC TI ON
Freshwater ecosystems worldwide are experiencing extreme anthropogenic pressures. Almost all river catchments are influenced directly (e.g., via point-source pollution or physical changes) and/or indirectly (i.e., via global change) by human activities, with reducing freshwater biodiversity and hampering natural ecosystem functioning (Allan, 2004;Davis et al., 2010;Heathwaite, 2010). The large number of simultaneously or sequentially operating stressors renders multiple-stressor studies a necessity in environmental research. However, such endeavors are not dominant in the literature, terminology and approaches variable across disciplines, and a unified framework to mechanistically understand the effects of multiple stressors is yet to be proposed (Côté et al., 2016;Nõges et al., 2016;Orr et al., 2020).
Thus, lack of knowledge how stressors interact to shape ecological processes prevents stakeholders from making efficient short-and long-term managerial decisions for conservation, restoration, or ecosystem services purposes (Lindenmayer et al., 2010).
Worldwide, stressors associated with land use drivers such as urban and agricultural development have become particularly pervasive (Dudgeon, 2019). They have accelerated biodiversity loss and ecosystem functioning declines via changes in the physicochemical parameters of streams, such as nutrient concentrations and their stoichiometric ratios, levels of deposited and suspended fine sediment, flow velocity and water turbidity (Gordon et al., 2008;Horváth et al., 2019;Wu et al., 2019). Fine sediment reduces habitat heterogeneity by infilling the interstitial spaces in the stream bed, smothers the filtering and breathing apparatus of invertebrates and increases water turbidity . Fine sediment also covers biofilm and has direct negative effects on microbial communities Salis et al., 2017). Nutrient enrichment, mainly N and P derived from fertilizers, often produces subsidy-stress response gradients in invertebrate communities (Wagenhoff et al., 2012;Woodward et al., 2012). Finally, flow velocity reduction, often resulting from water abstraction for irrigation, water translocation or water storage by dams (Dudgeon et al., 2006), modifies the physical habitat and the diffusion of material and individuals (Calapez et al., 2018;Lange et al., 2018;Wu et al., 2019). One anticipated interaction between these stressors is that a reduction in flow velocity facilitates sediment deposition and, thereby, local accumulation of chemicals and nutrients while decreasing water re-oxygenation levels (Calapez et al., 2018).
To assess the effects of stressors on ecological stream health, benthic macroinvertebrate communities are often used (Bonada et al., 2006;. Certain groups of invertebrates such as Ephemeroptera, Plecoptera, and Trichoptera are highly sensitive to changes in their physicochemical environment (Bonada et al., 2006).
Invertebrates also connect isolated water bodies across space and time, by dispersing over land and providing an important food source to higher trophic levels in both aquatic and adjacent riparian habitats (Sato et al., 2016). Further, macroinvertebrates are important drivers of stream-wide ecosystem processes, for example by influencing the decomposition rate of organic matter or participating in secondary production (Frainer et al., 2018;Huryn & Benstead, 2019).
One important concept in ecology is the "environmental filtering" theory, which states that environmental conditions select for tolerant species and certain traits (Poff, 1997). While taxonomical assessments have often been used to assess the effects of multiple stressors on communities, they are bound to the regional species pool which reduces their potential for generalization. To overcome this limitation, trait-based assessments, which rely on the compilation of community specific trait databases to characterize community niche breadth, have recently been getting more momentum (Ding et al., 2017). Trait assessments highlight the functional significance of species, that is, what they can do. Such assessments focus on the filtering role that environmental factors have in shaping community characterizations and provide mechanistic insights into community assembly and processes related to ecosystem functioning (Poff, 1997;Statzner & Bêche, 2010;Wu et al., 2019). Trait assessment results are not bound by the identity of species nor their regional pool but rather reflect functions individual can perform, thus facilitating the upscaling of local findings to larger geographical and longer temporal scales. In freshwater macroinvertebrate studies, traits associated with morphological characteristics, mobility, lifecycle, respiration strategy, and feeding habits have been very informative and can be linked to stream-wide processes (Cummins, 2016;Dolédec et al., 2011;Poff et al., 2006). For instance, body shape and breathing apparatus are often associated with flow velocity and water oxygenation levels, two important components of microbial activity regulating stream-wide processes such as nutrient cycling (Calapez et al., 2018;Dolédec et al., 2011). The developmental pace of individuals influences their tolerance to stressors (Dolédec et al., 2011) and their mobility moderates the linkage between habitats and the recovery of communities following stressor application (Guzman et al., 2019;Li et al., 2020;Schäfer et al., 2017). Finally, feeding habits directly link to the metabolic and stoichiometric resources needed by the individuals and thus to the dominant productivity pathways operating in a given stream such as primary or secondary productivity (Cummins, 2016;Frainer et al., 2016).
Despite linking macroinvertebrate communities to stream-wide ecosystem functions such as decomposition and productivity, traitbased assessments in multiple-stressor research tend to be restricted to observational rather than manipulative field experiments (Ding et al., 2017;Dolédec et al., 2011;Mor et al., 2019). Further, mesocosm studies are an invaluable tool to ecologists by giving the ability to control and replicate multiple-stressor treatments (Woodward et al., 2010). Therefore, more data from multiple-stressor experiments conducted at the community and ecosystem levels in environmentally realistic scenarios are needed. To reduce this knowledge gap, we used field mesocosms to investigate the individual and combined effects of nutrient enrichment, flow velocity reduction and increased sedimentation on benthic stream macroinvertebrate communities and their associated functional traits (Figure 1). We aimed to determine whether macroinvertebrate communities are altered by stressor main effects and interactions through changes in functional trait diversity. Based on the findings of previous related research, we tested three specific hypotheses: 1. Sediment addition and flow velocity will have more pervasive stressor main effects than nutrient enrichment on community structure and trait composition because of their direct physical action on macroinvertebrates (Elbrecht et al., 2016); 2. Nutrient enrichment will enhance the biomass accumulation potential, either via an increase in the mesocosms' carrying capacity or through a body-size shift toward larger organisms, due to increased resource availability (Cross et al., 2015;Frost & Elser, 2002;Ott et al., 2014); 3. Interactions between flow velocity reduction and added sediment will be more common than interactions with nutrient enrichment (Matthaei et al., 2010);

| Experimental mesocosm system
The experiment was conducted in autumn from 1 October to

| Experimental design
We manipulated fine sediment cover on the substratum surface, dissolved nutrient concentrations in the water column and flow velocity in 64 flow-through stream mesocosms, with two nutrient treatments (ambient vs. enriched concentrations), two fine F I G U R E 1 Conceptual model of the experiment. The benthic invertebrate community colonizing the stream mesocosms (stream community) is subject to different combinations of three stressors, nutrient enrichment (N), added fine sediment (S), and reduced flow velocity (F) at the mesocosm level. The resulting, filtered communities (local community) possess different densities of traits, which then influence functional diversity of the simulated stream ecosystem within each mesocosm. Functional diversity (FD) is defined by four metrics (FRic, FEve, FDiv, and FR). Nutrient enrichment is anticipated to promote diversity by increasing diversity and abundance of basal resources (e.g., periphyton), while sediment and flow velocity reduction have the opposite effect. We further anticipate antagonistic interactions between sediment addition and flow velocity reduction, whereas nutrient enrichment buffers the antagonistic effect of sediment and flow velocity reduction on functional diversity. Ecosystem functioning (not recorded in our study) is then expected to increase with functional diversity until reaching a plateau (Cardinale et al., 2012) sediment treatments (ambient vs. added), and two flow velocities (fast vs. reduced; fast refers to our control) in a full-factorial design. The experiment ran for 6 weeks with a 3-week precolonization period (Day-21 to Day-1) followed by a 3-week manipulation period (Day 0 to Day 21). Treatments were randomly allocated to 16 mesocosms within four blocks providing four replicates of each treatment combination across two sampling occasions (after 2 and 3 weeks of stressor exposure). Nutrient enrichment and flow velocity reduction started on Day 0 and both were continuously maintained throughout the manipulation phase. Also on Day 0, fine sediment was added in half of the mesocosms where it remained until Day 21.

| Macroinvertebrate sampling
On each sampling occasion, water flow was stopped in two header tanks and the whole substratum and the associated benthic invertebrates of the 32 mesocosms were sieved in the field using a 150-µm metal sieve and stored in 2-L PET containers. These were immediately filled to the top with 95% ethanol and later stored at −18°C in the laboratory until processing. After ~12 hr, one third of the ethanol was replaced with fresh ethanol to account for any dilution caused by the water remaining in the substratum. In the laboratory, the invertebrates were elutriated with a 450-µm sieve to remove the fine sediment and randomly divided into four equal subsamples.
The specimens present in one subsample were counted, measured to the nearest 1 mm (body length excluding cerci and case ; and identified to family using a stereomicroscope (Leica EZ4HD 8-35X, Leica microsystems GmbH, Germany), except for Nematoda, Oligochaeta, and Acari (Brooks et al., 2011). When specimens could not be confidently identified to a family, they were assigned to an order. We adopted this conservative approach across the whole dataset to reduce misassignments associated with the small size and general state of some specimens. Further, previous experiments suggest that family level of identification can be reliably used to examine community-environment and trait-environment relationships in aquatic habitats (Brooks et al., 2011;Tolonen et al., 2017). Adult Coleoptera families and Dipteran pupae were kept as individual taxa as they present different biological characteristics from their larval counterparts. The remaining 3/4 of each sample was scanned for rare taxa, which were added to the total taxon count in each sample. We then extrapolated the total invertebrate abundance for all taxa in each mesocosm by multiplying the subsample counts by 4.

| Species trait data
All invertebrates were assigned into five trait groups, which were subsequently divided into 22 trait categories, using a binary code (Li et al., 2019). Adult beetles (which were rare) and insect pupae were excluded from this classification. Selected traits featured lifecycle, habit, functional feeding groups, morphology, and respiration strategy (Table 1). Together, these traits give an overall description of the ecological characteristics of the community and also represent aspects that are susceptible to having a close relationship with the manipulated stressors. Further, the traits provide information about the resilience and resistance of the community as well as more general biological characteristics (Ding et al., 2017;Dolédec et al., 2011;Li et al., 2019). Trait information was adapted from the literature (Ding et al., 2017;Merritt et al., 2008;Poff et al., 2006) and online databases (Appendix S2) (Schmidt-Kloiber & Hering, 2015). A summary of the different trait categories can be found in Table 1.
We used the dbFD function in the FD R package (Laliberté & Legendre, 2010;Laliberté et al., 2014)  Organic matter decomposition, nutrient cycling, water purification, primary and secondary production (Cummins, 2016;Grimm, 1988;Woodward et al., 2012) TA B L E 1 Functional trait classification of the benthic macroinvertebrates in the mesocosms and ecosystem functions known to be associated with each trait We constructed a site × trait abundance matrix to represent community functional structure for each sampling unit. This matrix is obtained by multiplying a species × trait matrix (see Appendix S1) by a site × species relative abundance matrix (Li et al., 2019). Only widespread trait categories occurring in at least 50% of all mesocosms were retained in this matrix to avoid introducing too many zero values. In total, we measured 24 trait-specific variables: (a) functional richness, (b) functional evenness, (c) functional divergence, (d) functional redundancy, multivariate trait composition in the community (Pillai's Trace statistic), and (e) 19 trait categories.

| Statistical analysis
All statistical analyses were performed using R (version 3.5.2, R Core Team). Where necessary, data were log-transformed to improve normality and heteroscedasticity after exploratory data analysis.
The multivariate equivalent (MANOVA) of this model was used for the 17 common benthic taxa and the 19 widespread trait categories.
The significance level was set at p < 0.05, and all response patterns summarized in the Results were statistically significant unless indicated otherwise. Standardized effect sizes (partial η 2 values, range 0-1; Garson, 2015) are presented for p-values < 0.1 to allow evaluating the likely biological relevance of the results (Nakagawa, 2004). After Nakagawa and Cuthill (2007) smaller than the size of the corresponding main effects (Quinn and Keough, 2002).

| Community-level metrics
Total invertebrate abundance decreased with fine sediment addition. Abundance also decreased with flow velocity after 2 weeks of stressor exposure but not after 3 weeks (velocity × time interaction) (Table 2, Figure S1). Total EPT abundance was likewise negatively affected by sediment addition and flow velocity reduction ( Table 2).
Total invertebrate taxon richness was unaffected by all treatments, whereas EPT taxon richness decreased when sediment was added.
Lastly, Shannon's diversity decreased when sediment was added and Pielou's evenness showed a complex 3-way interaction (nutrients × sediment ×flow velocity), with evenness being highest in nutrient-enriched mesocosms with reduced flow velocity but no added sediment (Table 2, Figure S2).

| Body size metrics
Abundances of invertebrates in all three size categories decreased with sediment addition ( Table 2). The effect of nutrient enrichment on small invertebrates (<1 mm) changed from neutral after 2 weeks of stressor exposure to negative after 3 weeks (Figure 2a), preventing the increase of small individuals with time observed in the ambient treatment. By contrast, nutrient enrichment increased abundance of large invertebrates (>5 mm) on both sampling dates; moreover, fewer large individuals were found after 3 weeks than after 2 weeks (Figure 2b, Table 2). The effects of flow velocity reduction on small invertebrates were negative after 2 weeks but positive after 3 weeks (Figure 2c). Finally, medium-sized invertebrates (1-5 mm) became rarer when flow velocity was reduced (

| Multivariate community composition and individual common taxa
The multivariate results of our analysis indicated that invertebrate community composition changed in response to added sediment and when the three stressors were manipulated together; community composition also changed from week 2 to week 3 (Table 3). Regarding taxon-specific responses, 41% of the abundant taxa (seven of 17) responded to at least one experimental factor as a stressor main effect. All these seven taxa (the mayfly families Heptageniidae, Baetidae and Ephemerellidae, the dipterans Chironomidae and Tipulidae, the caddisfly family Leptoceridae, and the stonefly family Nemouridae) were affected by added sediment, followed by flow velocity reduction (Heptageniidae, Baetidae) and nutrient enrichment (Nemouridae).
These ten stressor main effects were all negative (Table 3). Changes with time (independent of stressor effects) occurred for two taxa; Heptageniidae became generally more abundant after 3 weeks of stressor exposure whereas dipteran pupae became generally rarer.
Temporal changes in stressor main effects affected four taxa and occurred in six cases (three for flow velocity, two for sediment and one for nutrients) (Figure 3a

| Functional diversity and traits
Two of the four functional diversity metrics were affected by sediment as a main effect, while none showed main effects for flow velocity reduction or nutrient enrichment. Functional evenness increased F I G U R E 2 Average numbers per mesocosm of small or large benthic macroinvertebrates on the two sampling occasions, showing the main effects of flow velocity reduction and nutrient enrichment (error bars = ±SE, n = 32 per treatment) when sediment was added whereas functional redundancy declined (Table 4). Nutrients increased functional redundancy after 2 weeks of enrichment but not after 3 weeks (Figure 4a, Table 4), and redundancy also increased with time regardless of the stressor treatments.
Further, functional redundancy showed a complex three-stressor interaction (Figure 4b). Nutrient addition increased functional redundancy at fast flow combined with added sediment but decreased redundancy at fast flow without sediment, whereas the opposite patterns occurred at slow flow.
The multivariate results of the trait analysis indicated that trait community composition (based on the 19 widespread trait categories) changed in response to sediment addition (Table 5); further, trait composition changed from week 2 to week 3 independently of the stressor treatments. Sediment (10 trait categories affected) and flow velocity (9) were the most pervasive stressors, followed by nutrients (2) (Table 5)

| Stressor main effects on the invertebrate community
We found sedimentation and flow velocity reduction to have the most pervasive effects on stream macroinvertebrate community assemblage and trait composition, as predicted in our first hypothesis. All observed effects of both these stressors on invertebrate community-level metrics and abundances of common taxa were negative. This result differs from previous experiments involving the same stressors, perhaps due to the fact that the source community for this experiment came from a "near-pristine" stream compared to agricultural streams in Germany, Ireland, or New Zealand (Beermann et al., 2018a;Davis et al., 2018;Elbrecht et al., 2016;. Fine sediment deposition caused a decrease in total invertebrate abundance irrespective of invertebrate size categories. We attribute this response to habitat homogenization (Petsch et al., 2017), decrease in food availability (Matthaei et al., 2010), and physical damage to the breathing apparatus of gilled invertebrates Wagenhoff et al., 2012;Wood & Armitage, 1997). The likely detrimental effect on brachial respiration is further supported by our univariate trait analysis, which displayed an increase in integumentary respiration concomitant to a decrease in brachial respiration when sediment was added. Further, abundance and richness of EPT taxa decreased with sediment addition overall, which was also reflected in the individual taxon responses.
The lower abundances of these taxa in mesocosms with added fine sediment are likely due to increased emigration rates via drift and/or emergence (Beermann et al., 2018a;.
Flow velocity reduction was the second-most pervasive stressor and displayed the largest number of changes with time in its effects on invertebrate community-level metrics and abundances of common taxa. Thus, the negative effect of flow velocity reduction after 2 weeks of stressor exposure on total invertebrate abundance and abundances of small individuals Chironomidae, Tipulidae, and Caenidae was no longer observed after 3 weeks. A similar situation was also observed in the response of Caenidae to sediment addition.
We suggest that interspecific microhabitat preference differences within these families led to an increased short-term drift response to reduced flow velocity (and sediment addition for Caenidae), which was later masked by recolonization of individuals within the same families that can tolerate or prefer slow flows (Harding et al., 2019;Zhang & Malmqvist, 1997). Previous studies have shown that flow velocity reduction often increases drift propensity, especially of swimming taxa (Beermann et al., 2018a;, which is supported by our finding of fewer swimming taxa at reduced flow velocity (see Stressor main effects on functional diversity and trait categories below). Further mesocosm experiments involving drift sampling should be done to confirm these patterns (Beermann et al., 2018a).
Our second hypothesis-enhancement of the biomass accumulation potential in response to nutrient enrichment either via increased carrying capacity or a shift toward larger-bodied organisms-was largely supported. Abundance of large-bodied individuals increased in nutrient-enriched mesocosms whereas small-sized organisms became rarer after 3 weeks of enrichment, despite total invertebrate abundance and community composition being similar (except for Nemouridae). Because immigration rates by drift into the mesocosms can be expected to be similar for all mesocosms (see Magbanua et al., 2013), this suggests invertebrate in nutrient-enriched mesocosms grew faster, with small individuals becoming medium-sized and medium-sized ones becoming large (Frost & Elser, 2002). Based on the abundances of small, medium, and large

| Stressor main effects on functional diversity and trait categories
In agreement with previous studies, sedimentation and flow velocity reduction were key stressors driving functional diversity and trait category responses (Buendia et al., 2013;Calapez et al., 2018).
However, because the colonizing invertebrate species pool was the same for all experimental units and total taxon richness remained similar across all stressor treatments, it is not surprising that functional richness and dispersion were also unaffected by our treatments. Traits were neither "lost" nor "gained", but rather relative abundances were rearranged to reflect changes in the dominance patterns of the taxa best adapted to the new environmental conditions. Consequently, the reduction in functional redundancy associated with sedimentation can probably be attributed to a smaller density of individuals performing the same functions. This suggests that the stress-induced community may be more vulnerable to further "functional loss" (Cummins, 2016;Pillar et al., 2013).
Additionally, nutrient enrichment increased functional redundancy and seemed to dampen changes with time, probably by allowing species with similar resource needs to coexist via an increase in quantity and quality of resources Sterner et al., 1993).
Shifts in feeding behaviors were also observed, with an increase in the relative abundances of predatory species when sediment was added or current velocity reduced. This result differs from Rabení et al. (2005) who reported a decline in total predator density when fine sediment cover increased, although these authors observed a broad predatory taxon-specific tolerance spectrum linked to their mobility. Our system lacked higher-order predators; therefore, it is possible that sediment deposition favored individuals capable of burrowing or crawling without becoming prey themselves (Ding et al., 2017;Li et al., 2019;Rabení et al., 2005). Further, we speculate that reduced hiding space due to sediment deposition filling interstitial spaces in the mesocosm beds, combined with an increased mobility of crawling predators, most likely facilitated their prey-catching success rate, which could explain their increased density under these conditions. We also observed an increase in the relative abundance of shredders after 2 weeks of reduced flow velocity, which could be related to an increase in CPOM retention with velocity reduction (Death et al., 2009). In our experimental setup, CPOM variations are highly dependent on CPOM load fluctuations in the stream feeding the setup; these were not recorded but could help explain the temporal pattern observed for shredders. When

| Interactive effects on community and functionality
Interactions between two or all three manipulated stressors affected six of 17 common taxa and three of 19 widespread trait categories. Five of these interactions occurred between flow velocity and sediment, two between flow velocity and nutrients, and two were interactions between all three stressors, thus partially supporting our third hypothesis that interactions involving nutrient enrichment should be least common. Previous experiments have highlighted the importance of flow velocity and sediment deposition in shaping stream invertebrate community structure and functionality (Buendia et al., 2013;Dolédec et al., 2011;Elbrecht et al., 2016).
Even though nutrient enrichment appeared to be relatively less important in shaping invertebrate responses in our experiment, past studies have shown that the effects of nutrient enrichment can differ strongly along an increasing gradient of concentration (Wagenhoff et al., 2012). Our enrichment treatment was fairly low compared to some previous similar experiments (Elbrecht et al., 2016); thus, it was perhaps not high enough (or long enough, at only 3 weeks of enrichment) to trigger strong responses of both community structure and trait composition. On the other hand, our enrichment might have already exceeded the subsidy threshold of our mesocosm ecosystem, resulting in a decline in many response variables compared to their peak subsidy enrichment point (Wagenhoff et al., 2012).
Distinguishing between the two outcomes would require further work involving a finer scale of nutrient (N + P) enrichment, to identify which side of the subsidy-stress gradient our results fall into.
In our datasets, interactive effects between stressors occurred most often in the abundance patterns of individual common taxa. Past experiments using the same stream mesocosm setup in Ireland and Germany also observed a similar trend (Davis et al., 2018;Elbrecht et al., 2016). In all three studies, moreover, EPT taxa were more sensitive than other taxa to twoway interactive effects between flow velocity reduction and either nutrient enrichment or sedimentation, as one might expect according to their high sensitivity to environmental changes (Bonada et al., 2006). The only taxon-specific three-way interaction observed in our experiment was a negative response of Chironomidae when exposed to all three stressors simultaneously. This result may seem surprising because this family is usually considered to be fairly tolerant to agricultural and urbanization stressors (Li et al., 2019;Mor et al., 2019). However, we suspect this intricate three-way interaction to stem from the complexity of the Chironomidae family, which encompasses a diverse range of genera and species that vary widely in their microhabitat preferences and tolerance of various stressors. Thus, it is possible that while some midge species were more tolerant to one or two stressors, the overall family responded negatively to all three stressors combined. These results lend more weight to the recommendation of Elbrecht et al. (2016) andBeermann et al. (2018a) that Chironomidae should be studied with a finer taxonomical resolution, for example by using DNA metabarcoding (Beermann et al., 2018b), to fully understand their response patterns to interacting anthropogenic stressors.

| Conclusions
The present study shows how multiple-stressor research can move beyond community assessments to anticipate changes in ecosystem stability and ecosystem processes in response to stressors.
Our experiment demonstrates the complexity of macroinvertebrate community dynamics and individual taxon responses to multiple agricultural stressors. Although traits and functional diversity showed a higher proportion of stressor main effects (74% of functional variables affected compared to 58% for community/taxon variables), invertebrate community and taxon responses were more sensitive to stressor interactions (31% vs. 17%). Thus, taxonomical and trait approaches are highly complementary, even over short spatial and temporal scales (Cummins, 2016). While community abundance patterns can help us investigate macroinvertebrates dynamics, traitbased approaches give a mechanical indication of the reasons why.
Finally, functional diversity facilitates predictions about the stability of a given system when exposed to multiple stressors (Pillar et al., 2013). Further studies, ideally repeated over different seasons, spatial scales and incorporating ecosystem processes such as energy transfer between trophic levels, should be conducted to improve our knowledge of macroinvertebrate community responses to multiple stressors (Kardol et al., 2018).