Real‐time warming of Alpine streams: (re)defining invertebrates' temperature preferences

The accelerating climate crisis intensifies environmental changes in high‐altitude ecosystems worldwide, with rising air temperature among the main stressors. While past research in alpine streams has primarily focused on how retreating glaciers might affect the ecology of glacier‐fed streams on the long run, observations of real‐time alterations of water temperature in such pristine environments are rare. Using long‐term measurements of water temperature (2010–2017) together with datasets on benthic invertebrate communities from 18 glacial and nonglacial alpine and subalpine streams in the European Alps, we illustrate significant ecological relationships of water temperature regimes and the identity of benthic communities and forecast changes thereof due to considerable warming of stream water. Besides reporting multiannual warming of all observed streams during summer with a mean rate of 2.5(±0.6)°C decade−1, this work redefines temperature optima and ranges using robust regression modelling and thereby identifies potential winners and losers among the invertebrate species. We conclude that the various invertebrate taxa in alpine stream networks will respond differently to thermal alterations and that the herein modelled temperature ranges of invertebrates is an essential step towards the understanding of future shifts in species distributions and success.

alterations of water temperature in such pristine environments are rare. Using longterm measurements of water temperature (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) together with datasets on benthic invertebrate communities from 18 glacial and nonglacial alpine and subalpine streams in the European Alps, we illustrate significant ecological relationships of water temperature regimes and the identity of benthic communities and forecast changes thereof due to considerable warming of stream water. Besides reporting multiannual warming of all observed streams during summer with a mean rate of 2.5 (±0.6) C decade −1 , this work redefines temperature optima and ranges using robust regression modelling and thereby identifies potential winners and losers among the invertebrate species. We conclude that the various invertebrate taxa in alpine stream networks will respond differently to thermal alterations and that the herein modelled temperature ranges of invertebrates is an essential step towards the understanding of future shifts in species distributions and success.

K E Y W O R D S
alpine streams, autecology, Chironomidae, cold-tolerant, Diamesinae, stream invertebrates, warming, water temperature

| INTRODUCTION
Freshwater stream habitats at high altitude are strongly influenced by their water source contribution (via melting snow or ice or via springs), but also by local environmental stressors that influence their properties along the longitudinal progression of the rivers. Among others, the water temperature is the most critical factor influencing biota and structuring benthic communities in mountain streams (e.g. Milner, Brittain, Castella, & Petts, 2001;Rott, Cantonati, Füreder, & Pfister, 2006). As the atmospheric temperature continues to rise globally (IPCC, 2018), not only stream sources will change considerably, for example through rapid glacier-retreat (Zemp et al., 2015), but also water temperature will adjust to local environmental conditions, such as local air temperature (van Vliet, Ludwig, Zwolsman, Weedon, & Kabat, 2011). Stream temperatures are rising globally as a consequence of atmospheric warming and other thermal pollutions (Kaushal et al., 2010;Lepori, Pozzoni, & Pera, 2015;Ormerod, 2009), and also temperature regimes of mountain waters are were observed to closely track local air temperatures (Caissie, 2006;Niedrist, Psenner, & Sommaruga, 2018). Since water temperature is crucial not only for the distribution of species in alpine streams, but also for the performance and physiological processes of these species (Füreder & Niedrist, 2020;Schütz & Füreder, 2019), it can be expected that the recorded intensification of atmospheric change in recent and coming years and especially during summer (Christidis, Jones, & Stott, 2014;Hock et al., 2019;Meehl & Tebaldi, 2004) leads to considerable alterations of structural and functional assets of alpine stream ecosystems (Füreder & Niedrist, 2020).
The sensitivity of aquatic invertebrates to key-environmental conditions determines their spatial and stream-type specific distribution and relative abundance in different environments (Bêche & Resh, 2007). The definition into classes (e.g. warm-or cold-tolerant species) generally allows to identify resistant groups to environmental change and provides the basis for scenario modelling (e.g. identify community changes due to anticipated warming). The sensitivity of taxa is usually characterized by its optimal level together with the tolerance range, which allows discriminating between taxa with wide or narrow ranges (e.g. eurytherm or stenotherm). Such autecological characteristics are usually collected in and retrieved via specific databases (such as Schmidt-Kloiber & Hering, 2015). Generally, warming of stream water is supposed to alter the invertebrate community composition (Durance & Ormerod, 2007;Hering et al., 2009;Ward, 1992), and the effect of different temperature regimes on the communities in alpine rivers has been confirmed (Füreder, 1999;Lods-Crozetet al., 2001;Niedrist & Füreder, 2016). Increasing water temperatures in mountain streams are found (a) to force species with limited distribution to upstream migration (Giersch et al., 2015;Sheldon, 2012), where environmental conditions fit their eco-physiological requirements, (b) to facilitate the immigration of warmer-water predators into mountain stream habitats with all its ecological consequences (Khamis, Brown, Hannah, & Milner, 2015), or at worst, (c) to potentially transform river sections into inhospitable habitats for cold-water fish (Eaton & Scheller, 1996;Hari, Livingstone, Siber, Burkhardt-Holm, & Guttinger, 2006). Hence, the understanding of temperature optima and limits of key species in alpine streams is of high importance to anticipate consequences of climate change on the integrity of these headwater stream ecosystems.
Here, we analyzed multiyear water temperature records from distinct stream types in the European Alps and linked them to local air temperature patterns. Further, we assessed the benthic invertebrate communities in diverse alpine and subalpine stream habitats and (re) defined the water temperature preferences (optima and limits) of key species in these ecosystems. The objective of this study was thus to improve the understanding of how vulnerable species are to impacts of climate change, which is the essential basis to forecast multispecies shifts during the warming of mountain waters.
By studying the temperature regime and change together with the benthic invertebrate communities in stream types with differing water source contribution at different altitudes in the European Alps, (a) we quantify the changes in local air and water temperatures in recent years to compare them with known rates from mountain regions, (b) we determine the influence of distinct water temperature regimes on the long-term development of different invertebrate communities and (c) we (re)define existing temperature limits of keyinvertebrate species in alpine streams, which will help to identify potential winners and losers among the invertebrate species in these Central European headwaters. s.l.) and form a gradient of environmental harshness (Niedrist, Cantonati, & Füreder, 2018), where the water temperature is considered to be the main stressor for invertebrate occurrence. The study streams' catchment areas (upslope areas) range from 0.31 to 33.62 km 2 , and the distances to the source (glacier or spring) are between 28 and 6,448 m (Table 1).
Temperature records of water and air range from 2010 to 2017/2018, while the benthic invertebrate communities were surveyed in the summer of 2015 (July).

| Air and water temperature
Air temperature data were obtained from weather stations distributed throughout the study region, located at low (950 and 1,005 m a.s.l.) and high (3,105 m a.s.l.) elevations and at distances between 6.5 and 23 km from studied streams (monthly mean air temperature from the HISTALP database [Auer et al., 2007] Water temperature was monitored continuously with digital loggers (Tidbit; Onset, Bourne, Massachusetts) throughout the years with intervals of 30 min. We averaged these values to daily mean water temperatures, and extracted daily summer water temperature data for each year (all daily mean temperatures between July and August from 2010 until 2017).

| Data analysis
We considered air temperature data from July and August as summer air temperature. This was extracted as daily mean or monthly means from the years 2010 until 2018. From the year-round records of water temperature, we extracted the data from July and August as summer water temperature (WT Summer ). Differences in the overall mean and change (slope) of WT Summer during 2010 and 2017 between stream types (glacial vs. nonglacial) were examined using estimation statistics rather than null hypothesis tests (Halsey, 2019;Ho, Tumkaya, Aryal, Choi, & Claridge-Chang, 2019), presenting all data points, group means and standard deviations and effect sizes with 95% confidence intervals computed from 5,000 bootstrapped samples (nonparametric bootstrap resampling). The corresponding Cumming estimation plots, which illustrate differences as effect sizes were produced using the "dabestr" package in R v.3.5.1. (R Core Team, 2018).
We applied nonmetric multidimensional scaling (NMS) of invertebrate communities at each stream site to visualize the qualitative differences between sites and stream types. For this, all six samples per stream were pooled. Variations in community composition as a function of glacier cover in the catchment (%) and stream water temperatures were tested by applying a two-way PERMANOVA (using Adonis function from "vegan" package in R, Oksanen et al., 2017) with 999 permutations on the Bray-Curtis similarities of averaged total invertebrate communities. The weighing of the variable importance was done based on the partial R 2 and the p-value. The significant environmental vectors (p < .05) were fitted onto the ordination space using envfit function from the "vegan" package (Oksanen et al., 2017). The relationship between WT Summer and taxa richness in the streams was tested and described using a simple linear model together with an F-test and each stream reach as one data point, while deviations per site were also illustrated. We used weighted averaging regression models for optima and tolerance quantifications (using the functions of the "analogue" package, ter Braak & Looman, 1986). Modelled preferences are displayed via univariate forest plots from package "metafor" (Viechtbauer, 2010) for taxa that were found in more than 3% of all samples.

T A B L E 1
Names and abbreviations (used in Figure 1) of study streams with the description and summary of decisive environmental characteristics: dominant water source (water source), distance of the sampling site to the source, catchment area (upslope area from sampled reach), altitude, glacier cover in that catchment (as % of catchment area, GCC) and averaged daily mean water temperature during summer (WT Summer

| Benthic invertebrates and water temperature regimes
The NMS ordination of the invertebrate community compositions converged on a two-dimensional solution (stress = 0.15) and the environmental variables were well and significantly fitted in this ordination space (R 2 = 0.69 for WT summer and R 2 = 0.52 for glaciation, Figure 4). The communities from glacial and nonglacial streams overlapped partially in the ordination space, but the difference in community composition was significant when comparing both groups of streams (PERMANOVA, p < .003). The community composition was related to the prevailing summer water temperature and degree of glaciated catchment (PERMANOVA, p = .007).
Differences in water temperature, however, explained more of the variation in community composition (F 1, 15 = 2.8, R 2 = 0.14, Taxa richness in the streams was significantly related to WT Summer (F 1, 16 = 5.14, p = .038; Figure 5). While the density of invertebrates was not significantly related to WT Summer , the relative abundance of chironomids significantly decreased with increasing water temperature (F 1, 16 = 10.37, p = .005, R 2 = .39).
F I G U R E 3 Summer water temperature (July and August) development in studied glacial (a) and nonglacial streams (b) from 2010 to 2017 with the overall level of summer water temperature (WT Summer , C) and change over time (d) in both stream types. The individual slopes (a and b) are identified with the abbreviations of study sites presented in Figure 1. Cumming estimation plots illustrate the differences in the overall mean temperature (c) and of the overall slopes over time (d) between both stream types (i.e. glacial vs. nonglacial streams). The horizontal lines limit the effect sizes of stream type, the vertical bars limit the 95% confidence intervals of bootstrapped distributions of these (see Methods for details) [Colour figure can be viewed at wileyonlinelibrary.com]

| Thermal optima and tolerances of alpine stream invertebrates
The mean summer water temperature data per site used for optima and tolerance modelling ranged between 1.2 and 10.6 C ( Figure 6c).
Thus, the modelled optima did not exceed these minimum and maximum values. Diamesa spp. (Diptera: Chironomidae) were among the taxa with lowest temperature optima ranging between 3.5 and 6.0 C (Figure 6a). Although the tolerance limits of many taxa overlap, no other taxa within the Chironomidae family, but also not of the Ephemeroptera, Plecoptera and Trichoptera order (Figure 6b), had a lower temperature optimum than these Diamesinae taxa.
While the chironomids showed a clear pattern of taxa with cold and warmer temperature optima (e.g. Diamesinae vs. Orthocladiinae), this was not evident in EPT orders (Figure 6b) Figure 6a).

| DISCUSSION
Stream water temperature varies considerably within alpine stream networks, also depending on the proximity to the water source (Brown, Hannah, & Milner, 2006). Compared to lowlands, water temperature is lower in high altitudes (Milner & Petts, 1994), where it is stable and very low during winter (Niedrist & Füreder, n.d.;Johnson, 1971) but can reach higher levels and variabilities during summer (e.g. Füreder, Schütz, Wallinger, & Burger, 2001). The highest daily mean water temperature in the studied streams was 16.4 C in a glacier-fed stream (site 11-G in Figure 1) at 1530 m a.s.l. and at 5 km distance from the source, while most streams had maximum daily temperatures around 12.0 C (95% CI 10.2-13.9 C). During the summer season, however, the mean daily temperaturesdespite the great variance found between individual streamsdiffered considerably between distinct stream types (glacial vs. nonglacial), with glacierfed streams being on average colder by 2 C (95% CI 0.3-4.4 C) due to the water source contribution by melting glacier-ice.
To show the potential effects of global and local rising air temperature on the temperature of alpine stream water, the slopes of air and stream temperature have been computed using linear regression over the study period data series from 2010 to 2017 (2010 to 2019 for air temperature), but using daily temperature data from summer months (July and August) only. This analysis revealed that summer water temperature has risen in all study streams and considerably at an average rate of 0.25 C per year (95% CI 0.15-0.35 C), but also that the warming rate does not differ between glacial and nonglacial streams (streams with and without glacial meltwater contribution). Although glacier-fed rivers have a unique seasonal temperature regime (Niedrist & Füreder, n.d.;Milner et al., 2001), the studied glacial streams warmed at similar rates as nonglacial streams, which finding confirms the recent report about the missingbut long suspectedsummer cooling effect in glacial streams (Williamson,  F I G U R E 5 Relationship between mean invertebrate taxa richness and the summer water temperature ( C, July-August), n = 6. The bold grey line indicates the significant regression slope (+1.8 taxa/+1 C), horizontal grey lines quantify the standard error of the mean The averaged warming rate of studied streams during the last years (0.25 C per year, 80% CI 0.19-0.32 C) is considerably higher than that of local air temperature and the recently observed and modelled warming of Swiss streams in alpine settings (Michel, Brauchli, Lehning, Schaefli, & Huwald, 2020). This long-term survey over 40 years revealed an average slope of +0.33 C per decade for Swiss streams with diverse catchment topography, elevation and size.
Although focusing on summer temperature only, our results imply that the warming rate has risen considerably in the last years, which was also found in other streams (e.g. accelerated warming of northern German streams in recent years in comparison to previous decades, Arora, Tockner, & Venohr, 2016).
In addition to the observed fast warming of water temperatures, local air temperature during same summer months increased faster (+0.9-1.0 C per decade) than the global average (+0.2 ± 0.1 C per decade, Trenberth et al., 2007). Comparable to water temperature F I G U R E 6 Summer water temperature optima (points) and tolerances (lines, 95% CI) of common invertebrate taxa belonging to the macroinvertebrate groups (a) Chironomidae (Diptera) and (b) Ephemeroptera, Plecoptera and Trichoptera. The columns on the right indicate the weighted mean and the limits of the 95%confidence interval. The temperature lines (panel c) display the range between minimum and maximum daily mean water temperatures in all sampled streams during summers (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017). cin_zer, cinerella/zernyi-group; COP, Cricotopus/Orthocladius/Paratrichocladius group; J, juvenile larvae; Gr. = group [Colour figure can be viewed at wileyonlinelibrary.com] patterns, the warming of air during summer accelerated considerably in recent decades and is known to be higher than the global average in many regions of the world (IPCC, 2018). It can thus be confirmed and concluded that near-surface air temperature in the European Alps are experiencing greater warming compared to other regions and the global average. Although this was previously covered by previous works (Gobiet et al., 2014;Hock et al., 2019), our study revealed that this fast warming of air temperature is also conveyed to other mountain ecosystems like mountain streams. This study thus demonstrates that the generally increasing warming rates expected for the near future around the world (IPCC, 2014) can already be observed in mountain regions (also found in Hock et al., 2019) also in aquatic environments, and thereby confirms these areas as suitable testing grounds for examining ecological consequences due to rapid warming of stream ecosystems, as once claimed by Brittain and Milner (2001).

| Temperature mediates invertebrate community structures
Differences in water temperature levels associated with stream type and the distance from the source drove differences in taxa richness and the community composition of invertebrates in the studied alpine and sub-alpine streams. Many studies have identified water temperature as a critical factor in the formation of distinct invertebrate communities in rivers, including glacial streams Niedrist & Füreder, 2016;Scrine, Jochum, Ólafsson, & O'Gorman, 2017). While most of these studies have shown these relationships on family or subfamily level, our results support and extend this finding on a higher taxonomic level and consider streams with and without glacial influence. In contrast to the often-used proxy for environmental harshness or indicator of glacial influence, that is, the ratio of glacier cover to total catchment area (Brown, Hannah, & Milner, 2007;Brown & Milner, 2012;Fell, Carrivick, Kelly, Füreder, & Brown, 2018;Füreder & Niedrist, 2020;Rott et al., 2006), this study disentangled that the properties of biotic communities were more related to summer water temperatures than to catchment glaciation.
We conclude that the temperature of stream water influences the stream biota more directly and depends on many local and small-scale conditions, such as slope, length, width, current velocity, and exposition of stream reaches (Williamson et al., 2019), which can be derived from a land-use proxy only to a limited extend. Thus, and although water temperature and glaciated catchment area correlated within the studied stream reaches, this work demonstrated that differences in immediate habitat characteristics, such as the temperature of the water itself, are more crucial for the invertebrate communities.
In addition to shifting invertebrate community composition through the exchange of certain taxa, taxa richness is expected to increase with rising summer water temperatures. As expected by previous works and underpinned by this study, warm-tolerant taxa will immigrate, while there will only be a small decline in the number of cold-tolerant taxa. The expected immigrants mainly belong to Ephemeroptera, Trichoptera and Plecoptera, based on their higher optima and tolerances of water temperature. In general, our study confirms the common regularity, that warmer alpine streams are colonized by communities with higher taxa richness, and estimates that warming of 1 C in such cold habitats leads to an overall increasing richness by approximately two taxa. In contrast to the pure tempera- Cold mountain streams have been previously reported to be dominated by the chironomid subfamily Diamesinae (Füreder, 1999;Füreder et al., 2001;Lods-Crozetet al., 2001;Niedrist et al., 2017;Niedrist & Füreder, 2016), with occurrences of additional groups in warmer reaches . Our results support and extend this long-known pattern by defining modelled optimal temperature range and delineated tolerance ranges for a variety of invertebrate taxa. In particular, exclusively six Diamesinae taxa were found to have their temperature optima in cold temperature ranges (3.5-6.0 C), which autecological property corresponds to their general dominance in cold mountain streams and the highest frequency and abundance in the streams studied here. Further, the defined temperature optima of Diamesinae taxa coincide with the identified temperature preferences of the same genus in previous assessments in the Italian Alps (Marziali & Rossaro, 2013 (Füreder & Niedrist, 2020), including behavioural and physiological capabilities to tolerate and survive freezing (Lencioni, Jousson, Guella, & Bernabò, 2015) and growing under harsh conditions , we assume that it is their cold-tolerance rather than their preference for cold temperature that gives them an edge over other invertebrates in harsh glacier-fed rivers. Regarding previously identified temperature preferences of invertebrate groups, we can conclude that the identified optima of Diamesa, of Diamesinae, of other chironomid subfamilies and of other invertebrate orders in European alpine streams roughly correspond to the conceptualized progression of invertebrates in glacier-fed streams across Europe . This consensus confirms the importance of water temperature as a major mediator of invertebrate communities in alpine and subalpine streams and the analogous relationships between invertebrate occurrence and the abiotic conditions in these aquatic habitats.
Our results extend the general understanding of aquatic invertebrate autecology by complementing the conceptual model  with a higher taxonomic resolution and by improving the preference modelling in considering multiple study streams in a broader geographic distribution across the Central Alps than used for previous niche assessments (Niedrist & Füreder, 2016). In modelling requirement-based species preferences in relation to summer water temperature, which environmental condition is demonstrably related to the identity of the invertebrate communities and is expected to drastically increase in the coming years (Hock et al., 2019), this study does not only identify potential winners and losers within the communities but also provides highly needed information for improving available autecological datasets (e.g. Schmidt-Kloiber & Hering, 2015) of invertebrates in alpine and subalpine stream ecosystems in the Alps. Such data sets will be essential for forecasting ecological situations and trends in mountain streams.

ACKNOWLEDGMENTS
This research was performed within the project "Monitoring Alpine Park is acknowledged for the permissions to work in these remote areas. The authors declare that there is no conflict of interests.

DATA AVAILABILITY STATEMENT
The data supporting the findings of this study are available from the corresponding author (River and Conservation Research group, University of Innsbruck) upon reasonable request.