When the going gets tough, the tough get going: The enigma of survival strategies in harsh glacial stream environments

Funding information TWF Tiroler Wissenschaftsfonds, Grant/ Award Number: UNI0404/1910 Abstract 1. Glacier retreat is a key component of environmental change in alpine environments, leading to significant changes in physico-chemical characteristics and biological communities in glacier-fed rivers. While the overall effects of the environment on community structure of invertebrates are largely understood, its influence on functional strategies such as feeding habits of same species are not. 2. The aquatic larvae of the species-rich family Chironomidae, or nonbiting midges, are the first invertebrates colonising alpine headwaters, and the first macroinvertebrate consumers in these harsh environments. Species composition in the two subfamilies, Diamesinae and Orthocladiinae, is diverse and is strongly affected by the changing habitat conditions upon glacier retreat. Here, we show that Diamesinae have extremely flexible feeding strategies that explain their abundance, high body-mass and predominance in glacier-fed streams. 3. Along a multifactorial ecological gradient from benign to harsh, based on water temperature, sediment transport and degree of glacial influence, Diamesinae expanded their trophic niche area and covered more trophic levels when conditions harshened. In contrast, niche areas of Orthocladiinae remained small and were not related to this gradient. In Diamesinae, mean body-mass increased with harsher environmental conditions, but no such effects were found in Orthocladiinae. 4. As facultative predators and able to feed on diverse food sources, Diamesinae have evolved survival mechanisms that allow them to thrive and successfully reproduce in glacier-fed streams, which likely explains their predominance in these habitats. 5. Climate change-induced glacier retreat affects the global water balance, with many downstream effects, including on irrigation and domestic use, and our study deepens our understanding of its effects on animals that depend on glacier-melt.


| INTRODUCTION
Alpine headwaters are of importance for downstream aquatic networks by providing multiple ecosystem services, including water, nutrients, organic matter or invertebrate prey for higher trophic levels (Milner et al., 2017). The heterogeneity of alpine streams is also vital for the biodiversity of invertebrates (Finn, Bonada, M urria, & Hughes, 2011) and bacteria . However, glaciers continue to retreat in alpine environments almost worldwide as a consequence of human-induced warming (IPCC, 2014), leading to alterations of the physical conditions in alpine glacier-fed stream ecosystems (e.g., altered water temperature, suspended sediment concentration and electrical conductivity) (Milner, Brown, & Hannah, 2009;Milner et al., 2017). These modification of freshwater habitats seem to occur accelerated and more obvious in the alpine life zone compared to lower altitudes and will increase over time as glaciers retreat (Zemp, Hoelzle, & Haeberli, 2009). Currently, biodiversity losses of adapted 'glacial species' and shifts of species (mainly invertebrates) in alpine stream ecosystems around the world are expected (Brown, Hannah, & Milner, 2007;Cauvy-Frauni e et al., 2016;F€ ureder, 2007, 2012Jacobsen, Milner, Brown, & Dangles, 2012).
However, whilst significant advances during the last decades have led to a good estimate of the structural patterns of invertebrate communites and their relationship with environmental conditions in alpine streams, understanding of their behavioural adaptations such as the feeding strategy of these animals is still limited (e.g., Clitherow, Carrivick, & Brown, 2013;F€ ureder, Welter, & Jackson, 2003a).
Benthic invertebrates living in alpine streams are either primary consumers and/or processors of autochthonous and/or allochthonous material, or predators (F€ ureder, Welter, & Jackson, 2003b;Niedrist & F€ ureder, 2017). Available classifications of alpine stream invertebrates to certain functional feeding groups (scrapers, grazers, shredders, collector-filterers or predators) are mostly based on studies from lowland streams. Understandably, in harsh environments like alpine streams, the suitability and accuracy of this approach, as well as the degree of feeding on a restricted range of particle sizes among species are not known (Niedrist & F€ ureder, 2017). As input of terrestrial organic matter is generally low in glacial streams , the invertebrates seem to depend on in-stream primary production. These basal food resources underlie considerable seasonal variations mediated by dynamics of disturbance and light regimes (Biggs, 1996;Hieber, Robinson, Rushforth, & Uehlinger, 2001;Uehlinger, Robinson, Hieber, & Zah, 2010). Consequently, in streams with glacial influence (the intensity depends on glacial extent and its melting activity), similar producer communities (dominance of filamentous chrysophyte Hydrurus foetidus, diatoms, epilithic bacteria and a lower abundance of cyanobacteria) represent the main food resources for grazing invertebrates (Cantonati, Corradini, J€ uttner, & Cox, 2001;Lods-Crozet et al., 2001;Rott, Cantonati, F€ ureder, & Pfister, 2006). Differences in abiotic conditions are thought to be of minor importance for the fatty acid composition within groups of benthic algae and cyanobacteria (Twining, Brenna, Hairston, & Flecker, 2016). Consequently, the uptake of 'highly unsaturated fatty acids', called HUFA's, might mainly depend on how invertebrates can consume trophic resources in harsh environments (F€ ureder et al., 2003b;Milner, Brittain, Castella, & Petts, 2001;. As it is known for soil organisms (Gongalsky, Persson, & Pokarzhevskii, 2008), environmental stressors might affect invertebrates' trophic niches, a specific aspect of an invertebrate's fundamental niche (Hutchinson, 1957). Taxa living in streams with high glacial influence (high turbidity, low water temperature) might possess trophic strategies for coping with the environmental harshness such as omnivory (F€ ureder et al., 2003b), predatory or even cannibalistic activity (Clitherow et al., 2013). General characteristics of food webs in glacial streams have recently been illustrated (Clitherow et al., 2013;F€ ureder et al., 2003b). However, the role of key-environmental conditions in affecting the feeding strategies and trophic niche breadth of dominant invertebrate taxa remain unclear in alpine glacial streams.
Here, we present results from a study of six glacier-fed streams across a gradient of glacial influence in remote areas of the Austrian Alps. We studied the effects of in-stream habitat conditions on abundance and body mass of macroinvertebrates, and estimate food source preferences, and degree of feeding plasticity of dominant taxa. In particular, the objective of this study was to examine whether invertebrate feeding performance (in terms of preferred food sources) affects their trophic niche extension and their trophic height, over a gradient of glacial influence and associated living conditions (defined as environmental harshness and based on differences in water temperature, sediment load and degree of catchment glaciation) in alpine and subalpine glacial streams in the Hohe Tauern National Park. We analysed trophic properties of four taxa belonging to the dipteran family of nonbiting midges (Diptera: Chironomidae), which are known to be the first inhabitants (e.g., Lods-Crozet et al., 2001;Niedrist & F€ ureder, 2016), and thus the first connection between producers and consumers, in glacial headwater regions (Niedrist & F€ ureder, 2017). In particular, individuals of Diamesa steinboecki and Diamesa latitarsis-group I are among the first and dominant inhabitants of glacial headwater regions in the Alps (Lods-Crozet et al., 2001), but understanding of their survival strategies in these harsh environments is limited, but see Lencioni, Jousson, Guella, and Bernab o (2015) for potential molecular adaptations.
Resource use by species and populations in different environments can be quantified and compared using naturally occurring stable isotopes of carbon and nitrogen and trophic niche metrics . Exploring new statistical techniques (Bearhop, Adams, Waldron, Fuller, & MacLeod, 2004;Jackson, Inger, Parnell, & Bearhop, 2011), these metrics allow a clear refinement of feeding modes and niche extensions of consumers.
We used novel Bayesian modelling to disentangle feeding preferences and to reliably quantify the trophic niche breadth (isotopic niche area) among four chironomid taxa. Based on earlier evidence from alpine stream studies (F€ ureder et al., 2003a,b), where NIEDRIST AND F € UREDER | 1261 autochthonous food sources seemed to be a significant contributor in invertebrates' diets, we employed the isotopic niche area to identify the diversity of resources assimilated by consumers. We hypothesised that i) independent from taxonomic identity, autochthonous food sources are more important than allochthonous material, and ii) to successfully sustain dominating populations in streams with harsh environmental conditions, taxa are forced to feed opportunistically.
With this first application of trophic niche metrics combined with Bayesian modelling in alpine stream habitats, we provide insights into the enigma of survival strategies in harsh environments. This is ecologically significant as the ongoing retreat of glaciers causes important abiotic changes that can affect the stability and structure of invertebrate food webs in glacier-fed stream ecosystems (Niedrist & F€ ureder, 2017;Tunney, McCann, Lester, & Shuter, 2012 Brown, Hannah, & Milner, 2016;Milner et al., 2001;Niedrist & F€ ureder, 2016). Such an approach decouples environmental conditions from spatial gradients and simulates multifactorial environmental change in glacial streams (Walker, Wardle, Bardgett, & Clarkson, 2010) and its effect on the functional performance of aquatic organisms.

| Characterisation of stream environments
All streams are continuously monitored in the long-term project monitoring alpine rivers (F€ ureder & Sch€ oner, 2013

| Sample collection
To characterise benthic chironomid communities, three multihabitat samples were taken at each site using a Surber sampler (mesh size 250 lm, area: 30 9 30 cm, 0.09 m²). We immediately preserved samples in 75% ethanol and stored them for further analyses in the laboratory. For stable isotope analysis, additional multihabitat samples were collected using a Surber sampler and spread on a white board. We manually removed Chironomid larvae, stored them separately in filtered stream water for 24 hr to allow evacuation of their digestive tracts, and rinsed them in distilled water before freezing (F€ ureder et al., 2003a,b). Additional chironomids were collected by hand searching. The rest of the multihabitat samples such as sand, algae, macrophytes, detritus and other allochthonous material (e.g., detritus and the gold alga Hydrurus foetidus) was frozen in the field and then separately sorted for every stream site in the laboratory.
T A B L E 1 Stream reaches sampled and their locations in the Hohe Tauern National Park, Glaciated catchment in % indicates the ratio of glacial ice in the respective catchment and assumes the degree of glacial influence on stream ecosystems Epilithic biofilm was collected by scraping and washing the entire surface of submerged rocks (n = 3) using a toothbrush and stream water. The material was concentrated onto pre-combusted Whatman GF/F filters and frozen in the field. We also collected additional benthic algae (esp. Hydrurus foetidus) by hand and froze them in the field.

| Chironomid community characterisation
We identified sampled chironomids to species or species-group level using published and unpublished keys (e.g., Ferrarese & Rossaro, 1981;Rossaro & Lencioni, 2015;Schmid, 1993) by mounting head capsules on slides using Euparal and examining them with 400-9 magnification. We measured the length of each individual from the antennal base to the procercus (a short structure where cerci originate) according to Nolte (1990) using an optical ocular ruler and calculated the body-mass of each taxon at each site using the available length-mass relationships (Benke, Huryn, Smock, & Wallace, 1999;Johnston & Cunjak, 1999;Nolte, 1990).

| Analysis of stable carbon and nitrogen isotopes
All target consumers and potential food sources from each stream Orthocladius luteipes) were dried before separately weighing individuals (0.2-1.2 mg) into tin capsules. The analysis of individual larvae enabled us to assess intraspecific variation and species' isotopic niches. Potential food sources (mixed detritus including allochthonous sources, Hydrurus foetidus and epilithic biofilm on filters) were homogenised using a mortar and pestle before weighing them into tin capsules (2 mg for detritus and Hydrurus, >20 mg for biofilm onto filters) to ensure best representation of the sources. All samples were analysed for stable isotope ratio and percentage of carbon and nitrogen on a Delta V advantage (Thermo Scientific, Germany) continuous flow isotope ratio mass spectrometer (CF-IRMS) interfaced with an elemental analyser at the Stable Isotope Unit at the Free University of Bozen, Italy.
Analytical standard deviation is <0.15% for both, d 15 N and d 13 C, assessed using in-house standards (IAEA-600, IAEA-CH3, IAEA-NO3). Stable isotope values were calculated using the following equation: where R is the ratio of heavy and light isotopes of the element (X) in samples (R sample ) and standards (R standard ). Data were reported as d values, defined as the per mil (&) deviation from the isotope standards (Pee-Dee Belemnite for carbon and atmospheric nitrogen for nitrogen).

| Gradient of environmental harshness and glacial influence
We combined the following environmental variables to order all sites gradually according to their abiotic conditions by performing a noncentred principal component analysis (PCA), similar to Ilg & Castella, (2006) and Niedrist, Cantonati, & F€ ureder (2018). In the analysis, we included monthly mean of the maximum daily water temperatures one month prior to sampling in July (mean water temperature), suspended load (sediment load) and relative glaciated area within the catchment (% glaciated catchment). The variable sediment load was log-transformed due to a skewed distribution. We used PC1 as the final multifactorial environmental harshness gradient, describing most of the environmental differences between the glacial stream sites (86.1%, Supporting Information Figure S1).

| Abundance and mean body-mass of target species along gradient of environmental harshness
We calculated the relative abundance (% of all chironomids) and the mean body-mass (lg/individual) of chironomid taxa groups analysed for stable isotope ratios (A-D) and related them linearly to the degree of environmental harshness of the stream sites. We illustrated the degree of significant or nonsignificant changes of both, relative abundance and body mass, for all taxa groups along the whole gradient. (one-way ANOVA) and included them separately into the model, as isotopic signals slightly differed between stream sites. Consumers C: N ratios exceeded 3.5, ranging between 3.8 and 8.8. As high lipid contents (>5% for aquatic animals) are known to deplete isotopes d 13 C signals, we made lipid normalisation of consumers' d 13 C signatures as suggested by :

| Identifying taxon-specific niche breadth and variability between stream sites
The isotopic niche space size (& 2 ) and its variability of each consumer group were evaluated for each site along the gradient of environmental harshness with the analysis of Bayesian Standard Ellipses using SIBER. This is an innovative method to quantify the size (diversity) of the niche in terms of "trophic area" based on the variability of a population's stable isotope (C and N) data. The ellipse corrected isotopic niche area is less influenced by extreme values or small samples and thus, represents more reliable estimates of niche extensions than the use of convex hulls, describing the smallest possible area in which all data points are enveloped (Jackson et al., 2011).
We included 130 consumer samples in this analysis. Posterior estimates (analogous to maximum likelihood in classical statistics) of isotopic niche areas for the four consumer groups at each site were modelled using Bayesian Standard Ellipses. The estimated areas (480,000 different simulations for each taxon in each glacial stream, a total of >10 7 simulated isotopic niche areas) were related to degree of environmental harshness using linear models, loess smoothing and by indicating confidence intervals.

| RESULTS
The six glacial streams were separated mainly along the first component of the PCA, significantly explaining 86.1% of differences in water temperature, sediment load and degree of glaciation in the catchment between the sites, whereas the remaining two components described 8% and 5.95% (Supporting Information Figure S1).    , and Orthocladius luteipes) in studied glacial stream sites 1-6 (see Table 1). Isotopic signatures of consumers are corrected for their lipid content. Primary producer data are presented as means AE SD, boxes represent min-max area for the respective elemental isotopic niche  Table 2). In contrast, the niche area of D. cinerella-gr. and O. luteipes did not vary across investigated stream sites (Figure 7c,d). Their niche area remained small

| DISCUSSION
We applied stable isotope analysis coupled with novel Bayesian mixing modelling to disentangle differing resource use and trophic niche variability of the dominant chironomid taxa in glacial streams with differing environmental conditions. Taxa already known to dominate the harshest glacial stream habitats in the Alps (e.g., F€ ureder et al., 2001;Lods-Crozet et al., 2001;Niedrist & F€ ureder, 2016) were found to depend on epilithic biofilm as a food source but also on plastic feeding behaviour as strategies to survive cold, dynamic and resource poor living conditions (=harsh conditions). In streams with less harsh conditions, the same taxa showed reduced feeding plasticity, a potential consequence of higher competition and resource partitioning .

| Body-mass of dominant chironomids in alpine and sub-alpine glacier-fed streams
The observed high densities of D. steinboecki and D. latitarsis-gr. I, dominating the invertebrate fauna in the investigated harsh glacial streams, support the general longitudinal patterns of invertebrate taxa according to the general model for glacier-fed streams (Milner et al., 2001) and previous studies in the European Alps (F€ ureder et al., 2001;Lods-Crozet et al., 2001). In addition to their dominance in terms of abundance, our results show that populations of the genus Diamesa found in harsh environments have a mean bodymass, multiple times higher than the body-mass of the same taxa in more moderate glacial streams. Most pronounced was the mean difference in body-mass of D. steinboecki, the species known to dominate glacier-fed headwaters in the European Alps (Lods-Crozet et al., 2001), which was more than 125 lg between larvae in the most benign and the harshest glacial streams (equals a mean increase of 226%). In contrast, no increase in relative abundance and mean body-mass was found for Orthocladius luteipes, a generalist species.
The implication of this is that withstanding harsh conditions in dynamic glacial stream habitats is likely allowing the pioneer invertebrate species to benefit from low competition in that they can acquire the available food sources in these environments. In less harsh environments with higher diversity (Milner et al., 2001) and higher densities of other invertebrates, higher competition leads to enhanced resource partitioning within the invertebrate community (Chase & Leibold, 2003;Tilman et al., 1997), but also results in a more balanced body mass of present taxa.

| Primary producers in glacier-fed streams
The isotopic composition of benthic food resources for stream invertebrates are reported to be highly variable between sites , potentially depending on current velocity and water temperature, which are important factors for the physiological discrimination of carbon isotopes by plants (MacLeod & Barton, 1998 also highly variable in our sites, but differentiable thanks to the analysis of samples from several catchments and the same stream type. The d 13 C signatures of epilithic biofilm, which usually is a consortium of autotrophic and heterotrophic bacterial cells and algal assemblages (Gesierich & Rott, 2012;Robinson & Kawecka, 2005;Wilhelm, Singer, Fasching, Battin, & Besemer, 2013), were significantly related with the gradient of environmental harshness between our sites. We observed a depletion of d 13 C in harsher environments (streams experiencing higher glacial influence). This potentially indicates the incorporation of glacial carbon, enriched in 13 C through microbial fractionation (Anesio, Hodson, Fritz, Psenner, & Sattler, 2009;Anesio & Laybourn-Parry, 2012), into heterotrophic bacterial cells within the epilithic biofilm, as reported by several authors (Bardgett et al., 2007;Fellman et al., 2015;Singer et al., 2012). Glacially derived organic carbon is highly bioavailable (Singer et al., 2012) and thus stimulates and sustains food webs in glacier-fed streams.

| Consumers in glacier-fed streams
Generally, there is little connection between riparian vegetation and the water of glacier-fed streams, which mainly depend on the interplay between sediment loads and conditions related to glacial melting dynamics (Milner et al., 2009;Smith, Hannah, Gurnell, & Petts, 2001;. Besides the direct effect of flow pulses and glacial runoff variability on benthic communities (Brown, Dickson, Carrivick, & F€ ureder, 2015;Cauvy-Frauni e et al., 2016) (F€ ureder, 2007;F€ ureder et al., 2001;Lencioni et al., 2015;Lods-Crozet et al., 2001). In contrast, the non-Diamesa taxa Orthocladius luteipes, a common species in alpine streams (Niedrist & F€ ureder, 2016 (Clitherow et al., 2013;F€ ureder et al., 2003b), we assume that omnivorous feeding is a widespread survival strategy for pioneer invertebrates in harsh glacier-fed streams. Such behaviour leads to body-masses of individual larvae exceeding the estimations from larvae living in benign streams. In benign glacial riverine ecosystems, together with a higher diversity of benthic communities (Brown et al., 2007), isotopic niche areas are small, indicating a specialisation/restriction of invertebrate species to certain food particles.
Thus, our results suggest that in less harsh glacial streams diversity and isotopic niche area are associated with increasing trophic niche differentiation (Tilman et al., 1997) triggered by the availability and the higher input of allochthonous food sources to the stream ecosystem. Consequently, higher competition for the available resources limits the body-mass of larvae.
Here, we show that feeding plasticity is a strategy to acquire enough of the biomass of patchily distributed producers in dynamic glacier-fed streams. In analysing the isotopic niche breadth (food source diversity) along streams with differing environmental conditions, our work demonstrates that considerable flexibility in invertebrate' feeding activities (this includes predatory activity) is required to dominate harsh glacier-fed streams. In less harsh sites, our results indicate that the same species are restricted to specific food sources.
However, how invertebrate predation and cannibalism (Clitherow et al., 2013) contribute to the diet of Diamesa taxa in harsh glacial streams was not quantifiable using stable isotope modelling due to technical limitations and remains to be studied. We suppose that predation or cannibalism could have contributed to the increased niche area and the higher mean trophic level of species in harsh sites. However, Clitherow et al. (2013) detected such feeding links directly at a glacier snout but they were rarely observed in sites further downstream (Niedrist & F€ ureder, 2017).
We hypothesise that as glaciers recede in mountainous regions (Zemp et al., 2009) and do not further contribute to stream flow, the food web of more diverse invertebrate assemblages will be mainly sustained by seasonal primary production and less by bacteria within the epilithic biofilm fed by glaciers organic carbon (Fellman et al., 2015). Additionally, we suppose that the expected decrease of nutritional quality due to shrinking glacial activity (Niedrist, Cantonati & F€ ureder, 2018) will further affect the overall production of invertebrate grazers in alpine streams.

ACKNOWLEDGMENTS
This work is part of the doctoral dissertation of GHN. The study was supported by a studentship to GHN from the society supporting higher education of citizens from the Autonomous Region South Tyrol (Verein zur F€ orderung der wissenschaftlichen Ausbildung und T€ atigkeit von S€ udtirolern an der Landesuniversit€ at Innsbruck), two research grants to GHN by the D. Swarovski KG (project "FEACH") and the Tiroler Wissenschaftsfond (project "FEED-STREAM"), and a research fund to LF (project "Monitoring Alpine Rivers") by the Nationalparkrat Hohe Tauern, the Ministery BMLFWU and the European Union. We thank the Administration of the National Park for permissions to work in these remote areas and are grateful to Saskia Amann, Bernhard Kofler, Simon Leinfellner, and Stefan A. Sch€ utz for their help in the field. We also appreciate the comments by two anonymous referees and the editorial advice by Associate Editor Belinda Robson that improved the manuscript.

CONF LICT OF I NTEREST
The authors declare no conflict of interest.