Structural and functional responses of invertebrate communities to climate change and flow regulation in alpine catchments

Abstract Understanding and predicting how biological communities respond to climate change is critical for assessing biodiversity vulnerability and guiding conservation efforts. Glacier‐ and snow‐fed rivers are one of the most sensitive ecosystems to climate change, and can provide early warning of wider‐scale changes. These rivers are frequently used for hydropower production but there is minimal understanding of how biological communities are influenced by climate change in a context of flow regulation. This study sheds light on this issue by disentangling structural (water temperature preference, taxonomic composition, alpha, beta and gamma diversities) and functional (functional traits, diversity, richness, evenness, dispersion and redundancy) effects of climate change in interaction with flow regulation in the Alps. For this, we compared environmental and aquatic invertebrate data collected in the 1970s and 2010s in regulated and unregulated alpine catchments. We hypothesized a replacement of cold‐adapted species by warming‐tolerant ones, high temporal and spatial turnover in taxa and trait composition, along with reduced taxonomic and functional diversities in consequence of climate change. We expected communities in regulated rivers to respond more drastically due to additive or synergistic effects between flow regulation and climate change. We found divergent structural but convergent functional responses between free‐flowing and regulated catchments. Although cold‐adapted taxa decreased in both of them, greater colonization and spread of thermophilic species was found in the free‐flowing one, resulting in higher spatial and temporal turnover. Since the 1970s, taxonomic diversity increased in the free flowing but decreased in the regulated catchment due to biotic homogenization. Colonization by taxa with new functional strategies (i.e. multivoltine taxa with small body size, resistance forms, aerial dispersion and reproduction by clutches) increased functional diversity but decreased functional redundancy through time. These functional changes could jeopardize the ability of aquatic communities facing intensification of ongoing climate change or new anthropogenic disturbances.

Global circulation models project an increase in air temperatures between 0.3 and 4.8°C before the end of this century depending on regions and climatic scenarios (IPCC, 2013). A stronger increase in air temperature is expected in high-altitude areas, primarily due to clouds (latent heat release), snow-albedo and water vapor-radiative feedbacks and low concentrations of aerosols (Pepin et al., 2015).
Furthermore, mountain regions are particularly sensitive to global warming because even a slight increase in air temperature (1-2°C) can cause drastic changes in their biological communities (Cannone, Sgorbati, & Guglielmin, 2007). In this context, glacier-and snow-fed rivers are considered as sentinel systems of climate change and could provide early warning of wider scale changes (Hotaling, Finn, Giersch, Weisrock, & Jacobsen, 2017;Jacobsen, Milner, Brown, & Dangles, 2012;Milner, Brown, & Hannah, 2009;Woodward, Perkins, & Brown, 2010). Alpine catchments are sensitive to the acceleration of glacier retreat and snowpack shrinkage which result in altered meltwater contributions Milner et al., 2017). Although most significant hydrological changes usually occur in upper areas, seasonal water releases from snow and ice are critical for maintaining water flow, sediment and nutrient transport, as well as ecosystem structure and function, thereby providing vital ecosystem services much farther downstream (Huss et al., 2017).

| Climate change effects in alpine catchments
Climate change induces modifications in sediment supply and transport as well as in thermal and hydrological regimes, thereby affecting hydrogeomorphological and physico-chemical conditions downstream (Cauvy-Fraunié et al., 2013;Hannah et al., 2007;McGregor, Petts, Gurnell, & Milner, 1995). These changes are likely to have cascading effects on the structure and function of aquatic communities (Milner et al., 2017). Reduced sediment loads, warmer water temperature and stabilized channels associated with climate change can lead to significant modifications in species distribution range and contribute to local extinctions (Jacobsen et al., 2014;Milner et al., 2009Milner et al., , 2017Milner & Petts, 1994). In fact, elevation shifts in species distributions toward higher river reaches, colonization of thermophilic species, increase in the incidence of temperature-dependent diseases and phenological shifts have been reported for fishes (Hari et al., 2006) and invertebrates (Domisch et al., 2013;Vittoz et al., 2013).
Increased temperature and reduced meltwater decreases environmental harshness in upper reaches surrounded by glaciers and snowpacks, which can lead to local taxonomic (alpha diversity) increases Milner et al., 2017). Otherwise, these changes can reduce alpha diversity in downstream alpine valleys where glacial and nival influences diminish (Jacobsen et al., 2012). Such changes in alpha diversity can be coupled with decreased between-site taxonomic differences (beta diversity) due to community homogenization (Cauvy-Fraunié, Espinosa, Andino, Jacobsen, & Dangles, 2015;Milner et al., 2017), ultimately resulting in a lower regional taxonomic richness (gamma diversity) and loss of endemic species Domisch et al., 2013;Jacobsen et al., 2012).
Although functional approaches are still scarce in the context of climate change, changes in taxonomic composition (primarily driven by the different responses of taxa to changes in temperature) could result in shifts in biological traits related to physiology, behavior or dispersal (Bonada, Dolédec, & Statzner, 2007;MacLean & Beissinger, 2017). Climate change (or other anthropogenic disturbances as flow regulation) could alter the organization of functional space through the loss of species with traits poorly adapted to the new environmental conditions and the colonization by better-adapted species (e.g. more competitive and/or productive species; Mouillot, Graham, Villéger, Mason, & Bellwood, 2013), or by homogenizing the functional structure of aquatic communities (Buisson, Grenouillet, Villéger, Canal, & Laffaille, 2013;Clavel, Julliard, & Devictor, 2011) due to the replacement of specialists by generalist species (Hotaling et al., 2017) (Daufresne, Lengfellner, & Sommer, 2009), while the ability to find appropriate food and habitat influences their persistence (MacLean & Beissinger, 2017).
Accordingly, generalist species should be more likely to find suitable resources, given their smaller body size, greater diet and habitat breadth (Angert et al., 2011). Such functional homogenization combined with the expected decrease in species richness might involve decreases in FD components and FR, and consequently in ecosystem resilience and resistance (Sonnier, Johnson, Amatangelo, Rogers, & Waller, 2014). In particular, FRic could show a delayed response to climatic change because its decrease requires the loss of species with extreme combinations of traits. By contrast, FDis could diminish as specialist species become locally extinct. Finally, climate change (and other anthropogenic disturbances as flow regulation) is expected to increase the importance of trait filtering (e.g. when original trait modalities are affected in a greater extent than common ones), potentially causing co-occurring species to become more clustered in functional space, thus decreasing FEve progressively (Mouillot et al., 2013).
This study addressed the structural and functional effects of climate change and flow regulation on aquatic communities in two alpine catchments in south-eastern France, one being influenced by dams and intakes of various capacities. Biological responses to climate change and the interaction between both stressors were explored using environmental and biological data collected at 15 sampling sites in the 1970s and compared with those recollected in the 2010s. According to climate change models (IPCC, 2013), both catchments have become drier and warmer in the last few decades.
As ecological consequences of these environmental changes between the 1970s and the 2010s, we hypothesized: 1. Decreases in local and regional taxonomic richness (alpha and gamma diversities, respectively), and in the dissimilarity among sites (spatial beta diversity) due to biotic homogenization in both catchments.

2.
The replacement of psychrophilic by eurythermic and thermophilic taxa, leading to a high temporal turnover and changes in taxonomic composition patterns.
3. Shifts in functional traits to cope with climate change: reductions of body size and aquatic dispersion, but increase of taxa with resistance forms and several reproductive cycles per year.

4.
Declines in community functional redundancy and diversity indices due to shifts in functional traits and biotic homogenization.
We expected such changes to be more pronounced in the regulated catchment than in the free-flowing one, due to additive or synergistic effects between flow regulation and climate change on aquatic communities.  Durand, Giraud, Brun, Merindol, & Martin, 1999). The study area is located in the alpine hydroclimatic region of Southern Alps, which is characterized by glacier, snowmelt and mixed hydrologic regimes (Renard et al., 2008). Natural and seminatural land uses are predominant representing more than 80% of the area for both catchments as estimated from CORINE land-cover

| Invertebrate sampling design
Benthic invertebrates show globally consistent responses to climate change effects (Brown et al., 2018) so they were used as model aquatic communities. Samples were collected in late winter (February) and late summer (August-September) on 15 sites within two catchments in the late 1970s (1977)(1978)(1979) and recollected in the 2010s (2013-2014) following the same sampling protocol. Sites were distributed longitudinally along both river networks covering an altitudinal range from 800 to 1600 m.a.s.l (sub-alpine; Figure 1, Table   S1.1 in Appendix S1). Five sites were located in the upper Verdon They were sorted, counted and identified to the lowest practical taxonomic level. Most taxa were identified to genus-or specieslevel except for Diptera, which were identified to family, subfamily or tribe (Chironomidae) and genus when possible. Abundances were standardized by sampling effort to obtain density (calculated as the total number of individuals observed divided by the sampled area) in order to make both catchments comparable. Finally, we averaged seasonal data to obtain annual taxonomic composition and density data for each sampling site and for each time period (data available in Table S1.2 and Appendix S2).

| Functional traits
The functional features of invertebrate communities were characterized using 11 biological traits describing life-history (body size, life-cycle duration, number of cycles per year and types of aquatic stages), resilience or resistance potentials (dispersal, locomotion and substrate relationship, resistance forms), reproduction, respiration, food preference and feeding habits (Tachet, Richoux, Bournaud, & Usseglio-Polatera, 2010). These traits are likely responsive to climate change (Bonada et al., 2007). Furthermore, food preference, feeding habits, body size, dispersal and locomotion can be also considered functional effect traits since they represent biological features that directly influence a specific ecosystem function (Hevia et al., 2017).
In addition, an ecological trait, called water temperature preference (Tachet et al., 2010), was also analyzed to track potential changes in psychrophilic, eurythermic and thermophilic taxa. Each taxon was coded according to its affinity with each trait category using a fuzzy coding approach (Chevenet, Dolédec, & Chessel, 1994). Fuzzy coding data were then converted to percentages of affinity for each trait. This procedure standardizes the potential differences in the codification scores (i.e. different row sums for each taxon and trait).

| Climate and discharge data
For all sampling sites, daily climatic data were obtained for the preceding 20 years including the collection dates, 1960-1979 for the 1970s and 1996-2015 for the 2010s, using the SAFRAN model (Durand et al., 1999; a detailed description of SAFRAN and its applications over France can be found in Quintana-Seguí et al., 2008). We selected and compared these periods as an acceleration of climate change has been observed globally (NOAA, 2015) and particularly in the Alps from the 1980s (Hari et al., 2006). More specifically, total (mm) and solid (mm) annual precipitation, daily mean (°C), maximum (°C) and minimum (°C) air temperature, annual evapotranspiration (mm), soil humidity (dimensionless index ranging from 0 to 1) and snow water equivalence (mm) were gathered for each sampling site and averaged for each period of 20 years. Similarly, daily and hourly flow data were collected from Eaufrance (the French public service on water information; http://www.eaufrance.fr) and Électricité de France (EDF) databases. There were only five sampling sites corresponding to gauging stations with enough flow records to hydrologically characterize both time periods (1960-1979and 1996-2015Kennard, Mackay, Pusey, Olden, & Marsh, 2010), four located in RC: Fontenil (FON), Briançon (BRI), l'Argentière (ARG) and Embrun (EMB), and one in FC near Saint-André-les-Alpes (STA; Figure 1).

| Changes in climatic and flow conditions between the 1970s and 2010s
Changes in each climatic variable between both periods and across catchments were tested using linear mixed-effect models (LMEs), considering date and catchment (and the interaction between them) as fixed factors and site as random factor (hereafter LME procedure).

| Structural changes in invertebrate communities between the 1970s and 2010s
To test spatial and temporal differences in taxonomic diversity (Hypothesis 1), indexes describing the different components of aquatic invertebrate diversity (α, β and γ) were computed. First, α-diversity was calculated as the taxonomic richness of each sampling site. "LME procedure" was performed to check differences in α-diversities between dates (1970s and 2010s) and catchments (RC-regulated and FC-free flowing). Second, β-diversity (βSOR, Sørensen's dissimilarity) and spatial turnover in taxonomic composition (βSIM, Simpson's dissimilarity) were calculated using occurrence data (Baselga & Orme, 2012). Then, whether sampling sites had higher βSOR and βSIM in FC than those located in RC were explored through null models for each period, which accounted for the different number of sites between catchments Finally, γ-diversity (total taxonomic richness) was estimated for each catchment and period, and compared considering that RC had more sampling sites than FC (null models were applied like those for β-diversity).
To test our second hypothesis (i.e. replacement of psychrophilic by eurythermic and thermophilic taxa, high temporal turnover and changes in taxonomic composition patterns), changes in water temperature preference between periods (1970s and 2010s) and catchments (FC and RC) were analyzed using LMEs on the relative abundance of psychrophilic, eurythermic and thermophilic taxa.
For this, "taxon × temperature preference" (Tachet et al., 2010) and "taxon × site" matrices were crossed to obtain water temperature preference per site for each period. LMEs were analogous to those performed for climatic variables (i.e. considering date and catchment as fixed factors and site as random factor). Nonmetric multidimensional scaling (NMDS) was performed to describe taxonomic composition patterns in the 1970s and 2010s using Bray-Curtis dissimilarity on log-transformed density data. NMDS was also done separately for summer and winter biological datasets to discard notable differences in species composition patterns between seasons, to be able to focus structural and functional analyses on inter-decadal changes by using composite samples. Multivariate dispersion was calculated for both basins and periods. Procrustes analysis (least-squares orthogonal mapping) was used to examine changes in community composition between the 1970s and 2010s for each site. This analysis shows the temporal displacement of each site in multivariate space. Finally, the temporal replacement of taxa was estimated in both catchments using temporal β-diversity (tβSOR, Sorensen's dissimilarity) and temporal turnover (tβSIM, Simpson's dissimilarity) following Baselga and Orme (2012). The extent to which their compositional divergence between periods was due to different tβSOR, tβSIM or rates of temporal turnover (tβSIM/tβSOR; i.e. the rate of temporal dissimilarity explained by the temporal turnover component) between catchments was assessed through one-way ANOVAs.

| Functional changes in invertebrate communities between the 1970s and 2010s
Trait community-weighted means (TCWM) were calculated to quantify the proportion or weight of each trait modality in each sampled community. Fuzzy correspondence analysis (FCA) was applied on TCWM to explore the functional structure of invertebrate communities (Chevenet et al., 1994). Subsequently, Pearson correlation analysis was performed to look for significant relationships between the first two axes of FCA and functional traits. Five functional indexes widely used to characterize freshwater invertebrate communities (Schmera, Heino, Podani, Erős, & Dolédec, 2017) were also quantified using TCWM for each site and period to account for the changes in functional space at the whole-community level: FD (Rao's quadratic entropy, Botta-Dukát, 2005), FRic (Villéger, Mason, & Mouillot, 2008), FEve (Villéger et al., 2008), FDis (Laliberté & Legendre, 2010) and FR (Rosenfeld, 2002). Previously, functional spaces were built (following Maire, Grenouillet, Brosse, & Villéger, 2015) using Gower dissimilarity matrices (adapted for fuzzy-coded traits, Pavoine, Vallet, Dufour, Gachet, & Daniel, 2009) on either all the functional traits (for the estimation of FRic, FEve and FDis) or only the effect traits (i.e. body size, dispersal, locomotion, food and feeding habits for FD and FR). FR was estimated as the difference between taxonomic diversity (using the Gini-Simpson diversity index) and FD (Pillar et al., 2013). Temporal (between dates) and spatial (between catchments) changes in each functional trait category (using TCWM) and functional indices were analyzed using the "LME procedure" to test hypotheses 3 and 4 (i.e. shifts in taxa functional traits to cope with climate change and decline in functional redundancy and diversity indices, respectively). Given the high number of tests performed for functional traits, Bonferroni correction was applied to p-values.

| Changes in climatic and flow conditions between the 1970s and 2010s
Climatic variables experienced substantial changes between 1960-1979 and 1996-2015 in both catchments (Table 1, Figure S1.1 in Appendix S1). In particular, mean air temperature (average increase of +0.77°C), maximum air temperature (+2.98°C) and evapotranspiration (+40.07 mm) significantly increased, whereas minimum air temperature (−0.46°C), soil humidity (−0.02), snow-water equivalence (−17.62 mm) and annual total precipitation (−33.34 mm) significantly decreased in the study area. None of the climatic variables showed different temporal changes between FC and RC (i.e. non-significant interaction between date and catchment for any of them, Table 1).
Mann-Whitney U tests pointed that there were no significant differences between years with or without biological samples in terms of total precipitation and air temperature (p-value > 0.05) within each 20 year period (i.e. 1960-1979 and 1996-2015). Complementarily, boxplots showed that biological samples were taken during humid years both in the 1970s and the 2010s ( Figure S1.2 in Appendix S1).
There was a generalized decrease in flow magnitude between the 1970s and 2010s in both catchments (Table S1. (Table S1.5 in Appendix S1).

| Structural changes in invertebrate communities between the 1970s and 2010s
Across all sites, about 70 and 76 taxa were collected in the 1970s and the 2010s, respectively. In the 1970s, 53 taxa were observed in FC, whereas 62 taxa in RC. In contrast, 65 taxa were recorded in FC, while 56 taxa were collected in RC in the 2010s. Moreover, according to null model results, γ-diversity values of FC did not differ from those estimated for a number of equivalent random sites from RC in the 1970s (z-score = −0.22, p-value > 0.05), but they did in the 2010s (z-score = 4.35; p-value = 0.001; Figure S1.3 in Appendix S1), when FC hosted more taxonomic diversity than RC.
Although α-diversity did not differ between the 1970s and the 2010s when considering all the stations together (term "Date" in LMEs, p = 0.13, F = 2.43, R 2 = 0.11), there were differences between catchments (term "Catchment" in LMEs, p = 0.005, F = 9.17, R 2 = 0.43) and a significant interaction between date and catchment (term "Date:Catchment" in LMEs, p = 0.004, F = 9.92, R 2 = 0.46). Although both catchments had similar α-diversities in the 1970s (Figure 2), RC experienced a strong reduction in taxonomic richness between the 1970s and the 2010s, whereas it increased in FC.
TA B L E 1 Results of linear mixed-effect models (LMEs) on climatic variables and water temperature preference of taxa. Marginal R 2 (R 2 m) and p-values for the whole model and the different terms (date, catchment and the interaction between them) are shown. The sign or trend of the relationship for each term is also displayed that is, temporal (date), spatial (catchment) and spatiotemporal trends (Date: catchment) Overall spatial β-diversity (βSOR) and spatial turnover ( Table 1). Psychrophilic taxa decreased similarly in both catchments (LME, term "Date", p-value = 0.001, F = 17.02, R 2 = 0.88, Figure S1.4 in Appendix S1), while thermophilic taxa increased more in FC than in RC (LME, interaction "Date:Catchment", p = 0.013, F = 7.06, R 2 = 0.1; Figure S1.4). There was no change in eurythermic taxa between the 1970s and the 2010s (LME model, Community composition patterns were more similar between FC and RC in the 1970s (with more shared taxa between them) than in the 2010s ( Figure 5). RC experienced a homogenization in terms of composition with a reduction in multivariate dispersion, whereas FC showed a slight increase in multivariate dispersion and some compositional diversification ( Figure S1.5 in Appendix S1). The same composition patterns were observed when considering summer and winter samples independently ( Figure S1.6 in Appendix S1).

| Functional changes in invertebrate communities between the 1970s and 2010s
Fuzzy correspondence analysis revealed temporal (1970s/2010s samples distributed from left to right along the first axis which explained 62% of the total variability) and spatial (FC/RC distributed along the second axis which explained 18%) effects on the functional structure of aquatic invertebrate communities (Figure 7).
Both axes were significantly correlated (R > 0.38; p < 0.05) with a wide range of functional traits (Figure 7, Table S1.6 in Appendix S1). Similarly, LMEs detected a severe shift in functional traits related to life cycle, physiology, behavior, resistance and resilience between the 1970s and the 2010s in both catchments (Table 2; see detailed results in Table S1.7 in Appendix S1). With the exception of life cycle duration, all functional traits experienced changes between both periods, some of them being more intense in FC than in RC (LME, interaction term "Date:Catchment" significant for body size, food preference and reproduction, Table 2). Regarding life cycle, the main temporal changes included reductions in body size and in the abundance of taxa with aquatic larva and eggs, as well as the increment of multivoltinism (>1 reproductive cycle per year).
In relation to resistance and resilience traits, reduction of crawlers, burrowers and taxa with aquatic dispersal was observed. On the other hand, increases in taxa with aerial dispersal and temporary attachment were detected. A proliferation of taxa with resistance forms (eggs or staboblasts and diapause or dormancy) to the detriment of those without resistance forms was also observed. With respect to behavioral traits, taxa reproducing by egg clutches increased, whereas those using cemented, isolated eggs decreased.
The main temporal changes in food preference and feeding habits  whereas FR decreased in the same period (term "Date", pvalue = 0.004, F = 10.14; R 2 = 0.94, Figure 8). None of them showed significant spatial differences between catchments or interactions between date and catchment. FRic (LME model, p-value = 0.061), FEve (p-value = 0.3) and FDis (p-value = 0.165) did not change through time or across catchments (terms "Date", "Catchment" and the interaction between them showed p-value > 0.05).

| Changes in climatic and flow conditions between the 1970s and 2010s
Climate change affected both alpine catchments. One of the main shifts was a noteworthy rise of mean daily maximum temperature

| Structural and functional responses of aquatic invertebrate communities
Changes in taxonomic composition, structure and functioning of invertebrate communities have been detected between the 1970s and the 2010s in our study area. Climate change reduces the habitat of psychrophilic species, especially in alpine ecosystems (Hotaling et al., 2017;Vittoz et al., 2013). In particular, rivers are experiencing northward range shifts of warm-water species while some cold-water species may go locally extinct, particularly in high altitudinal and latitudinal areas, where species distributions are most obviously limited by elevation and climate (Domisch et al., 2013;Heino, Virkkala, & (Hotaling et al., 2017;Jacobsen et al., 2012). However, hydrological alterations could limit colonization and dispersal of thermophilic species able to establish under the new climatic conditions in regulated rivers, due to habitat homogenization (Belmar, Bruno, Martínez-Capel, Barquín, & Velasco, 2013;Heino et al., 2009) and disruption of longitudinal connectivity associated with hydropower production (Ward & Stanford, 1983). Consequently, our results (hypothesis 1) showed that temporal patterns of α-and γ-diversities diverged between catchments, with increases in the free-flowing (as also found by Brown et al., 2007) but decreases in the regulated one associated with biotic homogenization. Similarly, although both catchments exhibited high temporal (hypothesis 2) and spatial turnover (βdiversity), these patterns were stronger in the free-flowing one. The absence of hydroelectric infrastructures enables greater spatial environmental variability, availability and diversity of mesohabitats (Rahel, 2002), and ecological integrity (Bunn & Arthington, 2002), promoting higher colonization rates in free-flowing catchments. In a context of climate change, a more intense replacement of taxa over time and space in free-flowing versus regulated rivers could allow the establishment of more heterogeneous, diverse and resilient communities.
However, this pattern may be temporary since some of the colonizing species might outcompete the native ones as climate change intensifies in alpine rivers (Hotaling et al., 2017;Vittoz et al., 2013).
The observed increases in taxonomic diversity were related to decreases in abundance and occurrence of some taxa with par-  (Bonada et al., 2007), which could ultimately affect primary and secondary production, biogeochemical processes and nutrient cycles (Wilby, 2008).
The observed pattern in functional traits is closely related to the increase of FD and the decrease of FR (hypothesis 4). Contrary to expected trends in more temperate areas, colonization of climatically restricted ecosystems by functionally different taxa may produce an increase in FD in the context of climate change (Brown et al., 2018). FRic, FEve and FDis (estimated using all traits) followed a similar, but not significant, pattern than FD (estimated only with effect traits), which could be due to a greater influence of climate change on effect traits (those used for FR and FD estimation This biological transition entails a temporarily increase in FD (now there is a greater variety of effect traits in the study area) but a decrease in FR (less taxa sharing the same effect trait combination).
Functional redundancy, which represents the number of species contributing similarly to an ecosystem function, relates positively to the ecosystem stability, resistance and resilience (Hooper et al., 2005). Thus, the loss of individuals and species contributing to the ns. ns. ns. p < 0.01 p < 0.01 same ecosystem functions increases the risk of ecosystem failure related to climate change intensification or other anthropogenic disturbances (Bruno, Gutierrez-Cánovas, Sanchez-Fernández, Velasco, & Nilsson, 2016). In the future, as climate change and the subsequent replacement of species intensify, as initially hypothesized, a decrease of FD could be observed. Finally, although observed effects of climate change on aquatic invertebrates are clear and supported by robust outcomes, sampling effect could have exerted certain influence given the limited dataset and temporal replication. Therefore, the biological responses to climate change and flow regulation found here should be followed by empirical research incorporating additional information such as hydropeaking frequency and magnitude, environmental flows and additional invertebrate samples in order to reach more extensive conclusions and improve river management.

| Management implications
Climate change can negatively affect dam management since reduced flows jeopardize the ability of alpine rivers to satisfy current water uses (van Vliet et al., 2016). Reservoirs fed with water coming from glaciers and snowpacks will not be able to maintain the same water levels in summer and autumn, when demand is greatest (Barnett et al., 2005). In fact, a transition from ice-and snow-fed to more rain-fed rivers has been recently detected as a consequence of climate change in the region (Van Vliet et al., 2013). Furthermore, decreases in snow cover between −24% and −55% (depending on the climatic scenarios) are expected for the second half of this century in the study area, which will be coupled with a decrease in total runoff (Etchevers, Golaz, Habets, & Noilhan, 2002). In this climatic context, reduction in snow-and ice cover can enhance sediment transport and deposition in alpine rivers with strong effects on water quality and quantity, aquatic habitat, flooding, infilling of hydropower reservoirs, turbine abrasion and agricultural and infrastructural development (Schaefli, Hingray, & Musy, 2007). Thus, under climate change, a major challenge is to satisfy multiple water uses as hydropower production or water abstraction for irrigation, while preserving biodiversity, ecological functions and associated ecosystem services (Arthington, Naiman, Mcclain, & Nilsson, 2010).
The survival of many species under climate change will depend on their ability to disperse and colonize new favorable sites (Root, Price, Hall, & Schneider, 2003), but could be limited by the recurrent fragmentation along alpine river networks. The cumulative effect of flow regulation entails a reduced discharge and the disruption of longitudinal connectivity, which may imply a greater isolation of alpine aquatic communities and reduced resilience (Maiolini & Bruno, 2007). Thus, free-flowing river communities can be more resilient to climate change than regulated ones, for which proactive and adaptive environmental flow management, as well as restoration measures at catchment scale, seem necessary to maintain and increase longitudinal and lateral connectivity, which are essential to the viability of populations of many aquatic species (Arthington et al., 2010;Bunn & Arthington, 2002;Palmer et al., 2008).