Assessing the effects of hydrological and chemical stressors on macroinvertebrate community in an Alpine river: The Adige River as a case study

1 Institute of Agricultural and Environmental Chemistry, Università Cattolica del Sacro Cuore di Piacenza, Piacenza, Italy Department of Civil, Environmental and Mechanical Engineering, University of Trento, Trento, Italy Department of Evolutionary Biology, Ecology and Environmental Sciences, Universitat de Barcelona, Barcelona, Spain Current address: Department of Zoology and Animal Cell Biology, University of the Basque Country, Leioa, Spain Water and Soil Quality Research Group, Department of Environmental Chemistry, Institute of Environmental Assessment and Water Research (IDAEA‐CSIC), Barcelona, Spain Water Quality Area of Catalan Institute for Water Research ICRA, Parc Cientific i Tecnologic de la UdG (Edifici H2O), Girona, Spain Correspondence Monica Giulivo, Institute of Agricultural and Environmental Chemistry, Università Cattolica del Sacro Cuore di Piacenza, Via Emilia Parmense 84, Piacenza, Italy 29122. Email: monica.giulivo@unicatt.it Funding information European Union 7th Framework Programme, Grant/Award Number: 603629‐Globaqua


| INTRODUCTION
Agricultural, industrial, and domestic activities exert pressures on freshwater ecosystems, in some cases, impairing their ability to provide essential services (EFSA, 2016). Threats to freshwater biodiversity are grouped under a number of interacting categories such as water overexploitation, water pollution, flow alteration, destruction or degradation of habitat, geomorphological alterations, land use changes, and invasion by exotic species and pathogens (Arthington, Naiman, McClain, & Nilsson, 2010;Dudgeon et al., 2006;Ormerod, Dobson, Hildrew, & Townsend, 2010;Vörösmarty et al., 2010). Diffuse (e.g., agricultural activities and intensive animal farming) and point (e.g., from urban areas due to the increase in the human population density) pollution are the main sources of contaminants entering freshwater ecosystems. In particular, concerns have been raised regarding pesticides (insecticides, herbicides, and fungicides), pharmaceutical products (PhACs), and personal care products (PCPs) (Ippolito, Carolli, Varolo, Villa, & Vighi, 2012).
Alpine rivers are part of the essential freshwater reservoir in Europe (Alpine Convention, 2009), since they provide freshwater for human consumption and for productive activities such as agriculture, livestock, and industry (Viviroli et al., 2011;Viviroli, Weingartner, & Messerli, 2003). In addition, the rough topography of their watersheds creates favourable conditions for hydropower production, which however alters the hydrological regime, thereby impacting the freshwater ecosystem (Liebig, Cereghino, Lim, Belaud, & Lek, 1999;Moog, 1993). Moreover, with the expected reduction of glacial runoff due to the retreat of Alpine glaciers, sediment loads will decrease, thereby driving potentially significant shifts in the biological communities of glacier-fed rivers (Ilg & Castella, 2006).
However, to the best of our knowledge, studies on the combined effects of a multiplicity of stressors are still lacking in the Alpine region.
In this regard, the application of a comprehensive approach that allows the effects of multiple stressors to be investigated at the catchment level may provide essential information to better understand and assess biological responses to this multiplicity of stress factors.
Given the wide range of activities conducted in its catchment, resulting in a multiplicity of stressors, the Adige River was selected in the EU FP7 project GLOBAQUA (Navarro-Ortega et al., 2015) as a case study representative of the Alpine region. In the present work, specific attention was given to the middle course of the Adige River, in the province of Trento, and to one of its main tributaries, the Noce River. The predominant pressures affecting the Adige River are: (a) streamflow and water temperature alterations caused by hydropower production (Zolezzi, Bellin, Bruno, Maiolini, & Siviglia, 2009;Zolezzi, Siviglia, Toffolon, & Maiolini, 2011); (b) land use (mainly agriculture) and industrial activities (Cassiani et al., 2016), which relevance increases from upstream to downstream; and (c) nutrients and pollutants released by waste water treatment plants (WWTPs); that is, effluents, which are expected to show significant seasonal variations due to tourism (Chiogna et al., 2016). All these pressures may negatively impact the benthic invertebrate communities, which, thanks to their capacity to respond to both chemical and physical alterations, can be used as indicators for bioassessment.
This work aims to identify the relationships between multiple pressures and the response of the invertebrate community at the investigated sites, which are representative of a number of scenarios encountered in Alpine rivers. We hypothesised that (a) seasonal and spatial patterns of hydrological and chemical parameters are observed not only according to the natural seasonal hydrological regime and the different water uses (e.g., hydropeaking), but also according to the activities in the basin (e.g., tourist activities upstream in winter and agriculture downstream in spring-summer) (Hypothesis H1); (b) the richness, diversity, and invertebrate community composition change as a consequence of the temporal and spatial pattern of water pollution and hydrological alterations (Hypothesis H2).

| STUDY AREA
The Adige River, with a total length of about 410 km, is the second longest river in Italy after the Po River. It rises near Lake Resia at the elevation of 1,586 m a.s.l. (46.834444, 10.514722), and it then flows through the southern-east Alps, and reaches the Adriatic Sea at Rosolina Mare, south of Venice (45.149722, 12.320278; Autorità di bacino del Fiume Adige, 2008). Glaciers cover a total surface area of 128 km 2 , although this extent is reducing at a relentless pace due to the observed trend for increasing temperature (Lutz et al., 2016). The flow regime has a typical Alpine character, with peaks in summer due to snow melting, and in autumn when cyclonic storms hits the catchment from the south. At the gauging station of Ponte San Lorenzo in Trento, the long-term mean annual streamflow is 203 m 3 /s, with a contributing surface area of 9,763 km 2 .
The majority (68.7%) of the territory of the Trento Province is covered by forest, and the remainder by rocks (11.5%), agriculture (16.5%), urban areas (2.8%), and water (lakes and rivers 0.05%; TERNA, 2011). Land use percentages for the study area are reported in Table   S1. The main water use is for hydropower. For this purpose, 28 reservoirs, 15 in the Bolzano and 13 in the Trento provinces, are in operation with a total operational storage of 560.59 × 10 6 m 3 . Another important activity is tourism, which leads to a larger increase of presences in both the winter and summer seasons, with the largest increment in winter.

| Sampling
Sampling was performed in two campaigns: The first (referred to as 1) was held in February, and the second (referred to as 2) in July 2015, in  Table S1. Water temperature, pH, dissolved oxygen (DO), and electrical conductivity were measured using a multiparameter probe (Aquatroll 200), while turbidity was measured using an optical turbidimeter (Ponsel IR). River velocity was measured using a radar gun (Decatur Electronics Europe Inc., Welber et al., 2016), except at Sites 2 and 3 where mean water velocity was determined by tracer tests using bromine (in February 2015) and NaCl (in July 2016).
At each site, water samples were collected at 50 cm depth at three points (left, right, and center of the river section) and mixed immediately after sampling. Water samples for the analysis of PhACs, PCPs, and pesticides were stored in 1 L grey PE bottles and within a few hours were transported to the laboratory in a refrigerated isothermal container and stored at −20°C until extraction and analysis. Water samples for ion analyses were collected in triplicate. The samples were filtered immediately through glass fibre filters (Whatman GF/F) and frozen at −20°C until analysis.
Macroinvertebrate communities were sampled using a pond net (0.32 m width and 500-μm mesh size) along the wadable zone of the river. Six samples were randomly collected at each site after disturbing the streambed 1-m upstream of the net by kicking. More than 90% of the river bed was mainly stones and cobbles in all sites. We used the same number of sampling actions at each site, six times, approximately 0.32 m 2 of surface sampled and a duration of 3 min each action. This procedure provides semiquantitative data; however, as we always used the same procedure, patterns between sites were comparable.
Samples were preserved with 4% formaldehyde.

| Determination of hydrological stressors
The hydrological regime was characterised by means of suitable statistical indicators of water discharge variation: annual mean, standard deviation, and coefficient of variation (FCV), 10th, 25th, 75th, and 90th quantiles (Q10, Q25, Q75, and Q90, respectively). Streamflow records (both daily and hourly) were obtained from the Ufficio Dighe of the Province of Trento (www.floods.it). As streamflow measurements were not available at Sites 1, 2, and 4, reliable estimates were extracted from the simulations performed by Bellin, Majone, Cainelli, Alberici, and Villa (2016). The natural regime (i.e., in the absence of water use) was reconstructed by excluding all water uses within the catchment . Statistics were also computed for the time series of streamflow (Q) increments between two successive time periods, t i + 1 and t i, defined as follows:

| Macroinvertebrate analysis
In the laboratory, samples were sieved through a 500-μm mesh, and macroinvertebrates were sorted, counted, and identified under a dissecting microscope (Leica Stereomicroscope). Identification was at the genera or species level for nearly all groups of taxa with the exception of the Oligochaeta and Diptera, which were identified at the family level.
Moreover, in order to assess the biological status, the extended biotic index (IBE; Italian biotic index, Hilsenhoff, 1982) was calculated. The IBE is based on the presence of invertebrates representative of groups of varying sensitivity to pollution and number of taxa (Ghetti, 1997).

| Statistical analysis
Organic pollutants included in the analysis were grouped into three families, based on their mode of action: pesticides (including herbicides and insecticides), PCPs, and PhACs. If the concentration was below the detection limit (mLOD), a value equal to one-half of the limit was assigned (Clarke, 1998), while the average of mLOD and quantification limit (mLOQ) was assigned when the concentration was in between these two values. Principal component analysis (PCA) was applied to the hydrological and environmental data. To diagnose autocorrelation and colinearity between environmental data, draftsman plots were used. When the determination coefficient was higher than 0.90, one of the variables forming the pair was removed. Variables included in the dataset analysed by PCA were standardised (the variable values were divided by the total for that variable) and inspected for normality, and when necessary log transformed using decimal logarithms. This resulted in the selection of the coefficient of variation of water discharge (FCV), water temperature (temp), nitrate concentration, water conductivity (cond), water turbidity (turb), urban and agricultural land use percentages (% urb, % agr), PCPs, PhACs, and pesticides ("Pest") as variables to be used in the PCA analysis.
With the aim of finding temporal and spatial patterns in the community, composition and density data (individual/m 2 ) were used.
Taxa present at less than 1% of the total density or only present at one site were excluded. Taxa densities were log transformed to reduce the influence of extreme observations on the subsequent ordination procedure (Siddon, Duffy-Anderson, & Mueter, 2011). Species richness (S) and Shannon diversity were calculated for each site and sampling period. These measures were contrasted between samplings and between up and downstream sites using a general linear model (GLM, sampling and site group as fixed factors).
A non-parametric distance-based redundancy analysis (dbRDA) was performed to determine the correlation between taxa composition and the environmental variables. RDA is a direct ordination analysis that selects a set of variables (predictors) that best explains the variability of a biological community (Borcard, 1992). Additionally, a PERMANOVA test was used to analyse differences in the macroinvertebrate community between samplings and site groups. Spearman correlations between some biological parameters and environmental characteristics were also calculated.
Analyses were performed using PRIMER 6 (version 6.1.6, Primer-E Ltd, Plymouth U.K.) and SPSS (IBM) for the GLM.

| Hydrological characteristics
At all sampling locations, water discharge was higher in the summer (July) than in the winter (February) sampling campaign ( Figure S1), except at Site 5, where the natural hydrological regime is altered by hydropower, this section being located downstream, and at short distance from the restitution of the Mezzocorona hydropower plant. Based on the analyses of the time series and their statistics, greater variations in discharge between summer and winter seasons were observed for small streamflows (i.e., the 10th and 25th quantiles, Q10 and Q25) compared with high streamflows (Table 1). This was due to the alterations caused by hydropower, which are particularly evident at low flow (seee.g., Zolezzi et al., 2009).

| Physical and chemical parameters
As expected, water temperature was higher in summer than in winter (Table 2). In both sampling campaigns, a similar spatial gradient of water electrical conductivity and turbidity was observed, with higher values observed at downstream sites (Table 2) In summary, our data showed a spatial pattern of chemicals (upstream, Site 2, urban pollution, downstream pesticides), and, as suggested by Hypothesis 1, hydrological seasonality determines the level of dilution at the most polluted sites.

| Macroinvertebrate community
The highest species richness was detected at Site 1 in both sampling campaigns, while a gradual decrease was observed at Sites 2 and 3, corresponding to the absence of several sensitive species (Plecoptera, Trichoptera, and Coleoptera groups) and an increase in other taxa (e.g., Chironomidae). At all sites, richness was significantly lower in the summer with respect to the winter sampling campaign (GLM, F = 24.63, P = 0.001; Table 2). The Shannon diversity index ranged from 1.3 to 2.4, and the most obvious decrease between winter and summer was observed at Sites 1 and 6; however, seasonal differences were not significant at all sites (GLM, P > 0.05).
None of the two metrics showed significant differences between upstream (Sites 1, 2, 3, and 4) and downstream sites (Sites 5, 6, and 7), where differences in flow variability and chemical concentration were observed. Higher richness and diversity relative to its upstream site were observed only at Site 4 in both samplings. This site located between the Mollaro reservoir and the restitution of the Mezzocorona hydropower plant is affected by a significant alteration in the natural streamflow, since the reservoir discharges a constant amount of water without any seasonal modulation, but it is not affected by hydropeaking, which instead impacts Site 5. In addition, the constant release of water reduces seasonal temperature variations (the release causes warming in winter and cooling in summer) that may favour the presence of some species (Maiolini, Silveri, & Lencioni, 2007;Ward, 1994). Accordingly, we found higher densities of some taxa, such as Baetis, Simuliidae, Chironomidae, and some species of Coleoptera, Trichoptera, and Gasteropoda, while other species (e.g., Capnia sp. and Capnioneura sp.) that are adapted to colder waters were less abundant. The mean densities of the most abundant species are reported in Table S2.
The first principal component of the dbRDA analysis (Figure 5a In particular, most sites occupied the upper part of the graph in summer and were characterised by poorer community composition (less taxa) compared with the winter sampling (located in the lower portion of the axis). This axis was positively correlated with water temperature and negatively with PhAcs and nitrate concentrations, which were both higher in winter at Site 2. These changes related to human perturbation at headwaters have been observed in previous studies in other Alpine rivers (Lencioni & Rossaro, 2005 with respect to other species, in tolerating rapid and periodic changes in the river flow due to hydropeaking (Mondy, Muñoz, & Dolédec, 2016).
As suggested by Hypothesis 2, the present study provides evidence for the seasonality in invertebrate community composition.
The two samplings show differences according to taxonomical community composition and density. A general decrease in richness and abundance was observed in the summer season, although some taxa (e.g., Serratella and Helodidae) showed higher densities in this period.

Seasonal distribution of invertebrates was also identified in
Apennine rivers (Bottazzi et al., 2011;Fenoglio, Bo, Cammarata, López-Rodríguez, & de Figueroa, 2014). These works suggest that the major forces shaping invertebrate communities seemed to be related to the Alpine climate and especially to snow accumulation and melting with the consequent substantial discharge variations.
In addition, a number of studies on glacial river ecosystems highlighted that water temperature is a key factor influencing biological communities (Brown & Milner, 2012;Milner, Brown, & Hannah, 2009). Therefore, most of the seasonal changes in taxa abundance observed in this study would be strictly related to species life cycle (Maitland, 1965;Milner & Petts, 2006), while the spatial patter is most likely related to stressors. Hydropeaking increased flow variability and determined a shift in the community at the downstream sites, but not in the diversity, partially according with our hypothesis. Dickson, Carrivick, and Brown (2012) highlighted that regulated flows may exert stronger effects on Alpine catchments than natural changes because they are active during winter, when river discharge and temperatures vary little. Pollution effects in the studied river appeared pointwise, were closely related with specific activities (i.e., urban and agricultural pollution) and were more evident in winter with lower flow. Such disturbances (i.e., hydropeaking and chemical inputs) produce discontinuities along the river, which influence the spatial distribution of organisms such as in this, as well as in other studies concerning glacial rivers (Knispel & Castella, 2003).

| CONCLUSIONS
This study shows that the composition of the macroinvertebrate community responded to seasonality and to changes in the main stressors along the Adige River. The inputs from WWTPs (already detected in headwaters) and a general increase in pollution downstream were the factors associated with chemical stressors, and these had more influence in winter when river discharge was lower.
Water flow variability due to hydropower seemed to favour some taxa (e.g., Gammarus) at sites located downstream, the restitution of a large hydropower plant. Richness and diversity did not change significantly between upstream and downstream sites. This research also highlights the importance of the spatial and temporal patterns of stressors in this Alpine river. The ecological status of impacted Alpine rivers cannot be improved further without considering the combined effect of these drivers, as discussed in the present work.

ACKNOWLEDGMENTS
This study has been funded by the European Union 7th Framework Programme (No. 603629-Globaqua).