Rock glacier springs: cool habitats for species on the edge

During September 2021, we investigated 15 sampling sites located in five catchments of the Upper Etsch/Adige River basin (Vinschgau/Val Venosta; Autonomous Province Bolzano/Bozen—South Tyrol), Eastern Italian Alps (Fig. 1, Table 1). The headwaters of the basin have already surpassed their peak water, i.e., the long-term period of largest glacier contribution to runoff (Huss and Hoch 2018) and are currently in the late stages of deglaciation (Galos et al. 2022).

In each catchment, we sampled three springs (within the first 50 m from the main water source) with different origin, representing distinct spring types: an ice or debris-covered glacier (termed glacier spring), an intact (i.e., ice-bearing) rock glacier (icy seep), and a spring without any presence of glaciers or permafrost (absent/very unlikely according to Autonomous Province Bolzano/Bozen—South Tyrol, 2023) in the underlain catchment (spring brook). Sampling activities were conducted in the period 8th–10th and 13th–14th September 2021, during non-rainy days and after at least 1 week from rain events. As reference for weather conditions, the data retrieved from high-elevation automatic weather stations (AWS) managed by the Autonomous Province Bolzano/Bozen—South Tyrol (2024) revealed that, during 2021, the snow cover at 2805 m a.s.l. (Teufelsegg AWS) completely disappeared on the 28th of June, and that the average air temperature and total precipitation in the period 15 July/15 September were 6.2 °C (Madritsch/Madriccio AWS, 2825 m a.s.l.) and 223 mm (Sulden/Solda AWS, 1905 m a.s.l.), respectively (Autonomous Province Bolzano/Bozen—South Tyrol 2024).

At each sampling site, we measured discharge (L s⁻¹) by salt dilution or, where this was not applicable, with the timed volume method (Dobriyal et al. 2017). We measured flow velocity (m s⁻¹) with a flowmeter Flowatch (JDC Electronics, CH) at 20 random locations, and used portable probes (portable multiprobes HI9829 and HI98198, Hanna Instruments, USA) to measure water temperature (°C) and turbidity (FNU) in-situ. We collected stream water samples in 50 mL (for water isotopes analysis), 100 mL (for multielement concentrations analysis) and 500 mL (for base chemistry analysis) clean PPE containers with double cap, that were filled to the brim (i.e., lacking air bubbles).

We used a kick-net sampler (100 µm mesh size; Stucki et al. 2019) to collect benthic invertebrates. A total area of 0.25 m² was investigated with five replicates distributed according to the relative abundance of microhabitats (i.e. cobbles/boulders > 250 mm; cobbles 250 > Ø > 25 mm; gravel 25 mm > Ø > 2.5 mm; moss and bedrock > Ø 250 mm) and classes of flow velocity range (Scotti et al. 2022). All five replicates were stored separately in 75% ethanol until further processing.

To derive information about the end-member contribution (stable water isotopes), we collected samples from different hydrological resources during three/six sampling campaigns conducted at each catchment up to 2 months before the invertebrate sampling. We used rainfall containers (Palmex Ltd, Zagreb, Croatia), that are specifically designed to prevent evaporative fractionation (Gröning et al. 2012), to collect samples from summer precipitation at a bi-weekly to monthly scale at each catchment (n = 33 samples). During June-July, we collected samples from snowmelt (runoff beneath—or dripping water from snow patches; n = 18), and August–September samples of glacier melt (glacier rivulets and water dripping from the glacier front; n = 24).

Immediately after the field campaigns, all collected samples were transported at < 10 °C to different laboratories (see section below), where they were stored at 4 °C before the analyses, occurred within 1–4 days (chemistry) or 10–40 days (stable water isotopes) after the delivery.

At the Free University of Bozen/Bolzano, we used a cavity ringdown laser spectroscope (CRDS Picarro L2130i, CA, USA; precision: 0.1‰ for δ ²H, and 0.025‰ for δ¹⁸O) to measure the values of δ²H and δ¹⁸O (‰) in the sampled waters. All isotopic values are referred to the VSMOW2. At the Eco Research Srl laboratory, samples were filtered (acetate membranes, 0.45 μm pore diameter) and acidified (1.5% volume, > 65% HNO3). This method did not change the results (given the accuracy of the outcomes and the type of springs) when compared with the immediate filtration and acidification after collection (Brighenti et al. 2023). Then, concentrations (µg L⁻¹) of dissolved major, minor and trace elements (Na, Mg, Al, K, Ca, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Rb, Sr, Mo, Ag, Cd, Sn, Sb, Ba, Tl, Pb, Th, U) were measured with Inductively Coupled Plasma Mass Spectrometry (ICP-MS ICAP-Q, Thermo Fischer® Scientific, MA). Samples for base chemistry were delivered to a private laboratory where Alkalinity, pH, HCO₃⁻, Ca²⁺, Mg²⁺, Cl⁻, Na⁺, K⁺, total nitrogen (N_tot), NH₄⁺–N, NO₃⁻–N, total phosphorus (P_tot), PO₄³⁻–P, SO₄²⁻ and SiO₂ concentrations (mg L⁻¹) were measured following standard methods (Baird and Bridgewater 2017).

Invertebrates were sorted under a stereomicroscope (50 x) and identified to genus (Ephemeroptera and Plecoptera), family (Diptera except Chironomidae, Trichoptera, Oligochaeta, Coleoptera), or broader taxonomic levels (Tricladida, Nematoda) according to the Italian national assessment method to assess the ecological status of rivers (Buffagni et al. 2006; D.M. 260/2010, 2011). Chironomidae (Diptera) were identified, if possible, at the species level (Lencioni et al. 2007; Rossaro and Lencioni 2015).

We used water isotopes to estimate the relative contribution to runoff from snowmelt, glacier-melt and rainfall hydrological end-members in Bayesian mixing models at each spring. Since there were no significant differences in end-members signatures among catchments, we pooled all samples belonging to different water sources and used the average and standard deviation values of summer precipitation, snowmelt, and glacier melt samples collected during summer 2021 to inform mixing analyses. We used the package Mixsiar v 3.1.12 in R (Stock and Semmens 2016; Stock et al. 2018) and set δ¹⁸O and d-excess (= δ²H–8 * δ¹⁸O) variables to estimate the relative contribution to runoff of glacier melt, snowmelt, and rainfall at glacier springs (Brighenti et al. 2023). Since no glaciers were present in the catchments underlain by spring brooks and icy seeps, and the perennial ice melt proportion can be considered as negligible in rock glacier springs (Jones et al. 2019), we used two-member mixing models with δ¹⁸O to distinguish the rainfall and snowmelt contribution at these spring types (Brighenti et al. 2023). At each site, the total meltwater contribution was calculated as the sum of the glacier- and the snow-melt components. The model uncertainties, calculated as standard deviation for each spring (mixture), were in the range of 9–15%. We used Principal Component Analysis (PCA) to produce two indices representing two environmental gradients summarized in the first principal component, respectively (e.g., Ilg and Castella 2006; Niedrist et al. 2018; Brighenti et al. 2021a). Due to the high number of measured variables, we conducted a first analysis using turbidity, discharge, water temperature and elevation to represent the physical harshness of the habitat. The second PCA included trace element concentrations (Ni, Co, Mn, Al, Fe, Rb, Ti, Sr, As), and it was aimed at producing a gradient of chemical harshness.

Invertebrate data from the five replicates were pooled for each site. Taxa occurring at only one site (Chaetocladius piger gr., Pseudosmittia sp., Psychodidae), meiofauna (e.g., Harpacticoida, Copepoda, Haplotaxidae, Tardigrada), Ephemeroptera, Plecoptera, Trichoptera and Diptera juveniles, and terrestrial fauna (e.g., Collembola) were excluded from the data analyses (Supplementary Information V).

Taxa accumulation curves were computed with the R package iNEXT version 3.0.0 (Hsieh et al. 2016) to compare sampling effectiveness and taxa richness between sites. Accumulation curves were based on taxa abundances and specified on Shannon-diversity (Hill numbers q = 1). To determine differences for diversity among spring types, a non-parametric Kruskal–Wallis test followed by Dunn’s pairwise comparison test was used. We conducted a two-dimensional non-metric multidimensional scaling (NMDS) using the monoMDS function based on Bray–Curtis distances and Wisconsin square root—transformed taxa abundances. We estimated the correlation between the first two NMDS axes and all the collected environmental variables to identify their association with different spring communities. CLAMTEST were set with a majority specialization of 2/3 majority and an alpha value of 0.001 to assess habitat preferences in a pairwise comparison between stream habitats. Minimum abundance for classification varied in each pairwise comparison (spring brook n = 1307 vs glacier spring n = 10; icy seep n = 386 vs glacier spring n = 10; spring brook n = 35 vs icy seep n = 11). NMDS, PCA and CLAMTEST were computed using the vegan R package version 2.6–4 (Oksanen et al. 2022).

We used General Additive Models (GAMs) with gaussian family and identity link function, with the R package (Wood 2011) version 1.8–42, to test the relation between the response variable Diversity (accumulation curves estimates based on Shannon diversity, Hill numbers q = 1), and meltwater contribution, and the indexes of chemical harshness and physical harshness obtained with PCA (smooth functions).

To assess possible species associations in spring brooks, glacier springs and icy seeps we used copulas, implemented relevant covariates based on negative binomial family and sampled 2000 times, from the ecoCopula R package version 1.0.2 (Popovic et al. 2022). The package allows to graphically model and look at possible association between taxa (variables) by jointly modelling the response of each of the two taxa against all other taxa in the community (predictors), accounting for environmental variables, and then looking at the residual correlations. Association plots were based on a lambda of 0.026, 0.067 and 0.03 for icy seeps, glacier springs and spring brooks, respectively.

Environmental conditions differed between the three spring types (for details see Table 2, Supplementary Information I).

Significantly higher water temperatures were measured in the spring brooks than in glacier springs and icy seeps (Kruskal–Wallis χ² = 10.7, df = 2, p < 0.005). Turbidity was significantly higher in glacier springs (Kruskal–Wallis χ² = 7.76, df = 2, p < 0.05). In the PCA summarizing the physical harshness of the habitat (broadly defined by low water temperature, high turbidity, high discharge and high elevation) these characteristics were strongly and positively correlated with PC1 (57% variance explained, see Supplementary Information II).

The meltwater/rainwater contribution was significantly different among different spring types, with glacier springs having higher, icy seeps intermediate, and spring brooks lower meltwater fractions (Kruskal–Wallis χ² = 10.3, df = 2, p < 0.01). This is compatible with significantly different δ¹⁸O values (Kruskal–Wallis χ² = 8.7, df = 2, p < 0.05), increasingly more enriched in heavy isotopes from glacier springs to icy seeps and spring brooks (Table 2). Different types of springs differed also in water chemistry, spring brooks having lower ion and trace element concentrations than glacier springs and icy seeps. Nevertheless, these differences also depended on the geological group of the sampling sites, with glacier springs and icy seeps belonging to the Ötztal unit having lower pH and HCO₃⁻, and higher SiO₂, Al, Mn, Ni, Zn, Y, Rb, and Co concentrations among all sites (Supplementary Information VI). For the same reason, all springs belonging to the Campo fault had larger pH, and concentrations of HCO₃⁻, As, and U than springs belonging to the Ötztal unit. In the PCA summarizing the chemical harshness (Supplementary Information II), Mn, Ni, Co, Rb had a strong negative correlation with PC1 (57% variance explained), Sr had a strong negative correlation with PC2 (28% variance explained).

The total number of taxa found was higher in spring brooks (n = 69; species 50.7%, genus 27.5%, family 18.8%, other 2.9%, Supplementary Information IV), intermediate in icy seeps (n = 38; species 44.7%, genus 34.2%, family 15.8%, other 5.3%), and lower in glacier springs (n = 15; species 60%, genus 26.7%, family 6.7%, other 6.7%). Glacier springs had a significantly lower (p < 0.05) taxa richness (Fig. 2). Diversity in icy seeps was larger at the Saldur/Saldura (SL) catchment, followed by Sulden/Solda (SU), Plima (PL), Lazaun (LZ), and Hochjoch/Giogo alto (HJ) (Supplementary Information III). There was only one specimen of Euorthocladius rivicola gr. in HJ-GL, thus we excluded this site in further analysis.

Differences in community richness and diversity were also related to contrasting composition of invertebrate communities. In the NMDS (Fig. 3), icy seep communities were intermediate between those of spring brooks and those of glacier springs. Physical and chemical variables were significantly associated with the spread of sites according to different spring types (Fig. 3b). Glacier springs were mostly composed by Chironomidae of the genus Diamesa, represented with contrasting abundances of different species (D. goetghebueri, D. zernyi, D. tonsa and D. steinboecki) depending on the spring, and co-dwelling taxa (e.g., Pseudokiefferiella parva, Orthocladius frigidus, E. rivicola gr., Eukiefferiella minor, E. brevicalcor, Rhypholophus sp., Acarina). Notably, D. steinboecki was not found at the glacier springs of Lazaun (LZ) and Hochjoch/Giogo alto (HJ). Differing abundances of few taxa in glacier springs were reflected in the wide spread of sites in the NMDS space (Fig. 3a, stress = 0.12). By contrast, spring brooks were more clustered in the same space. These springs had larger abundances of Ephemeroptera, Plecoptera Trichoptera (EPT) taxa, different subfamilies of Chironomidae and other Diptera. Thirty-two taxa were exclusive for spring brooks. These included two species of Diamesa (D. bertrami, D. dampfi), three Tanypodinae genera, 18 Orthocladiinae species, other Diptera (Simuliidae, Ceratopogonidae, Pediciidae), the Plecoptera, Nemoura mortoni, Trichoptera (Rhyacophila sp., Philopotamidae), and the Oligochaeta, Naididae. Icy seeps had a large variability in community composition, reflecting a gradient between glacier spring and spring brook communities (Fig. 4). Icy seep communities of the Saldur/Saldura (SL), Plima (PL) and Sulden/Solda (SU) catchments were more similar to spring brook communities. There, a relatively large diversity included Diamesinae (including D. latitarsis, Pseudokiefferiella parva and P. branickii), six Orthocladiinae, and Chironominae (Micropsectra sp.), other Diptera (including Muscidae, and Limoniidae genera), Ephemeroptera (Baetis sp.), Plecoptera (Leuctra sp. Dictyogenus sp., Brachyptera sp.), Trichoptera (Limnephilidae, and Philopotamidae), Coleoptera, Oligochaeta, and the Platyhelminthes Crenobia sp. By contrast, the icy seep communities of Lazaun (LZ) and Hochjoch/Giogo alto (HJ) were more similar to glacier spring ones (Fig. 3). The community in Lazaun (LZ-RG) was dominated by the Diamesinae D. goetghebueri, D. zernyi, and P. parva, co-dwelling with low abundances of other EPT and Diptera (EPTD) taxa. The community of the Hochjoch/Giogo alto icy seep (HJ-RG) was composed of low numbers of few taxa (D. tonsa, D. zernyi, Acarina, the Limoniidae Rhipholophus sp., and Enchytraeidae), including D. steinboecki.

According to the CLAMTEST analyses, taxa exclusive of a specific spring type were 32 (44%) for spring brooks, and 3 (4%) for icy seeps. Although no taxa were exclusive for glacier springs, three Diamesa species (D. zernyi, D. goetghebueri, and D. steinboecki) were mostly associated with this type of springs. Pairwise comparisons between spring types revealed a higher amount of both generalist and specialist species between spring brooks and icy seep communities. When compared with glacier springs, both communities had a comparable number of generalists (Fig. 4).

The physical harshness gradient, represented by the PC1 explained 57% variance, correlated with turbidity (0.81), discharge (0.81), elevation (0.7) and decreasing water temperature (− 0.69). The chemical harshness gradient (PC1 53% explained variance) correlated with decreasing Mn (− 0.81), Ni (− 0.8), Co (− 0.87) and Rb (− 0.82) concentrations.

We observed a significant decrease of diversity with increasing chemical harshness (Fig. 5a; PC1_Chem.: F = 7.804, p < 0.05, R² = 0.35, dev. explained = 39.4%) and increasing meltwater contribution (Fig. 5b; F = 12.98, p < 0.01, R² = 0.5, dev. explained = 52%). While an increasing physical harshness (Fig. 5c; PC1_Phys.: F = 2.452, p = 0.05, R² = 0.4, dev. explained 49.6%) showed a decreasing diversity trend.

The gradient of environmental and ecological conditions encompassing spring brooks, icy seeps, and glacier springs was reflected in different patterns of taxa associations found at different spring types (Fig. 6). Glacier springs had the lowest number of possible positive associations (n = 2) and no negative, compatible with a low number of taxa and a large physical and chemical harshness. Icy seeps had intermediate values of possible positive associations (n = 94) and showed possible negative associations (n = 2). This is compatible with a larger number of taxa, a low physical harshness and a large chemical harshness. Spring brooks had the highest amount of both possible positive (n = 472) and negative (n = 354) taxa associations, in agreement with a low chemical and physical harshness. The ratio between associations and taxa richness increased from glacier springs (0.13, only for positive) to icy seeps (2.5 for positive, 0.05 for negative) and spring brook (6.9 for positive, 5.1 for negative) communities.

In this study, we used stable water isotopes to identify the hydrological influence from meltwater/rainwater to spring runoff. End-member mixing models revealed a widespread dominance of the meltwater component, representing more than 70–80% of discharge at all sites. Nevertheless, the three types of springs had different meltwater contribution with lower values in spring brooks (77–84%), intermediate values in icy seeps (80–88%), and higher values in glacier springs (96–98%). Our outcomes are in line with previous studies on alpine hydrology based on stable water isotopes, highlighting a dominance of the snowmelt and/or (for glacier springs) the glacier-melt components (e.g., Penna et al. 2013, 2017; Marchina et al. 2020b), and a rainwater fraction generally larger (up to 30–50%) at springs not originating from glaciers or rock glaciers (Lucianetti et al. 2020; Brighenti et al. 2023).

Macroinvertebrate communities of the investigated springs were strongly associated to this meltwater gradient. Indeed, community composition and diversity were negatively associated with meltwater contribution at different spring types. By contrast, the relative abundance of Diamesa spp. in the Chironomidae community was positively related to meltwater contribution. These results agree with those of a previous study in the Pyrenees and in the Italian Alps, where a strong correlation was found between the abundance of Diamesa spp. (and Empidiidae), and the meltwater contribution (Khamis et al. 2014; Lencioni et al. 2021). This genus comprises the Chironomidae species with strongest preference of high glacier influence: D. steinboecki, D. goetghebueri, D. zernyi, D. latitarsis (Lencioni 2018). Thus, not surprisingly, these species represented the majority (91–100%) of Chironomidae specimens found at glacier springs, except at Hochjoch/Giogo alto (HJ), where we found only one specimen of Orthocladiinae (Euorthocladius rivicola gr.) among Chironomidae. The same four Diamesa species accounted for less than 4% of the total Chironomidae specimens at spring brooks and at the icy seeps in the Plima (PL), Saldur/Saldura (SL) and Sulden/Solda (SU) catchment and represented the majority (64–66%) of Chironomidae in icy seeps of the Schnals/Senales valley. The contextualization of community gradients under the prevalence of different hydrological end-members has a strong potential to predict biodiversity shifts under climate-change related modifications of water resources (Marchina et al. 2020a). Indeed, the climatic shifts occurring in alpine areas also modify the relative and absolute contribution from rainfall and meltwater to river networks, with consequent ecological effects (Huss et al. 2017). However, while a contrasting meltwater contribution in spring brooks and glacier springs was strongly associated to a gradient of physical harshness (see also Khamis et al. 2014; Lencioni 2018), the same meltwater/rainwater gradient did not fully explain the striking differences among icy seep communities that we found in our study. The physical and chemical conditions of spring habitats result from a complex interplay between hydrological resources contribution, landform type, and other climatic and topographic drivers. As such, the contribution from meltwater and rainfall, as estimated with isotope-based mixing models, only indirectly influences invertebrate communities. Yet, further investigations on the correlation between meltwater/rainfall contribution and the community metrics would gain more insights about the changing meltwater and rainwater influence on freshwater communities, particularly in non-glacial systems.

As expected, the investigated springs encompassed gradients of turbidity, discharge, and water temperature at different elevations. These parameters shaped our index of physical harshness. High turbidity and discharge were unique for glacier springs, whereas the cold waters (< 1.5 °C) in these habitats and in icy seeps (4 to sevenfold colder than spring brooks) can be attributed to the influence of glaciers and permafrost (Carturan 2016; Brighenti et al. 2021b). Overall, the diversity of macroinvertebrates was negatively related to the physical harshness of the habitat. This partially contrasts with other studies, where diversity peaks were related to intermediate physical harshness (e.g., Füreder 2007; Jacobsen et al. 2012). However, these studies were conducted on alpine streams, and it is likely that different processes (e.g., dispersal limitations, environmental filtering, higher resource-consumer mismatch) drive community metrics in springs (Talluto et al. 2024).

In contrast to physical conditions, that were strongly driven by the origin of springs, the chemical habitat was strongly controlled by the interaction between the spring origin and the underlying geology. Indeed, we found higher solute concentrations, with the dominance of SO₄²⁻ among anions, in springs belonging to the Ötztal unit. In this geological group, trace elements including Ni, Zn, Mn, Al, Co, and Y, were enriched in icy seeps and glacier springs. For example, values of Ni, which can be toxic for aquatic organisms (EU 2007; Custer et al. 2016), exceeded 3–11 times the reference standards for drinking water quality (20 µg L⁻¹; EU 2020), with the highest concentration (221 µg L⁻¹) at the Hochjoch/Giogo alto icy seep (HJ-RG). These differences in trace metal concentrations shaped our index of chemical harshness, that in turn had a strong influence on macroinvertebrate communities. Indeed, a gradient of decreasing diversity as a function of increasing Ni, Fe, Mn, and Co concentrations was associated to a decreasing prevalence of EPT taxa in icy seeps, paired with an increase of the relative abundance of species belonging to the genus Diamesa. Icy seep communities similar to those found in spring brook ones and composed of a variety of EPT and Diptera taxa, were those located within the Campo fault geological domain, and that at the Saldur/Saldura catchment (Ötztal unit). For the latter, sampling activities (for safety reasons) could be performed only two-hundred meters below the actual spring. Consequently, a large diversity at this site might have been related to the fact that we sampled a stream community rather than a spring community (rhithron, sensu Ward 1994). In icy seeps belonging to the Hochjoch/Giogo alto (HJ) and Lazaun (LZ) catchments, a physical harshness comparable with spring brooks suggests that the lower diversity and a large prevalence of Diamesa specimens in these icy seep communities are driven by the strong chemical harshness. Unfortunately, the plausible relation between a geology-driven chemical harshness and icy seep community assembly can only be hypothesized, given the low number of icy seeps investigated.

In general, the diversity and composition of macroinvertebrate communities resulted from a combination of physical and chemical harshness. Where both harshness types were low, as in spring brooks, diversity was high and the community was composed of abundant Ephemeroptera, Plecoptera, Trichoptera, and Diptera taxa, with Diamesa spp. underrepresented in the community (in spring brooks 1.5%). Under increasing physical and chemical harshness, the amount of EPTD taxa decreased and the abundance of Diamesa spp. increased (in icy seeps 17.6%). Under highest chemical (some icy seeps) and/or physical (glacier springs) harshness, macroinvertebrates were mainly represented by the Chironomidae Diamesa spp. (in glacier springs 78.9%) with higher affinity to glacier-fed streams (D. steinboecki, D. zernyi, D. goetghebueri, D. latitarsis; Lencioni 2018). Thus, the mutual interplay between physical (turbidity, bedload transport, water temperature) and chemical (trace element concentrations) harshness is a primary driver for alpine spring macroinvertebrates.

Different studies highlight the potential role of icy seeps as climate refugia (e.g., Brighenti et al. 2021b). While the occurrence of cold-adapted populations was demonstrated in icy seeps from different mountain areas globally for Plecoptera and Trichoptera species, the knowledge on Chironomidae was hitherto restricted to subfamilies/genera (Brighenti et al. 2021c; Reato et al. 2024). In this work, we were able to dig deeper into the species composition of this family and estimate the presence of the only species considered as unique for glacier springs: Diamesa steinboecki (Lencioni 2018; Wilkes et al. 2023). Local extinction of this species has been already reported in the Italian Alps, testifying its sensitivity to warming waters and decreasing glacier cover (Lencioni 2018; Lencioni et al. 2022). Besides glacier springs, we could find D. steinboecki in one icy seep (Hochjoch/Giogo alto, HJ). However, the recovery of only one specimen hinders drawing any conclusion about a resident population. Yet, that icy seep was the only site in the entire Hochjoch/Giogo alto (HJ) and Lazaun (LZ) catchments where we could find D. steinboecki, including the outflow of a very large glacier (Hochjochferner), located about 2 km in air distance and hydrologically disconnected to the icy seep. Also, there were no glaciers in the catchment underlain by the Hochjoch/Giogo alto icy seep (HJ-RG), and the occasional presence of D. steinboecki larvae due to active movements along river network pathways can be excluded. Since we can exclude a sampling bias (i.e., specimen stuck in the net from a previous sampling), the occurrence of that specimen in that icy seep would be (i) an occasional reproductive success from an adult female, passively transported from one of the glaciers located in the surroundings, or (ii) the presence of a resident population in that site, where larvae would be living at low densities in the spring, and adults would find suitable thermal conditions on the cold grounds typical of permafrost environments (e.g., Gobbi et al. 2021a). Under the second hypothesis, why D. steinboecki could not be found in the other icy seeps too? This might be related to the very cold waters (0.6–0.9 °C throughout the season; Brighenti, unpublished) at HJ-RG, but these were not or only slightly colder (by max 0.6 °C) than those of the other icy seeps. Instead, the particularity of the Hochjoch/Giogo alto icy seep (HJ-RG) is its sharp chemical harshness, as previously highlighted. This harshness was not only related to the concentrations of dissolved trace elements (Ni, Mn, Al, Y), but also to a peculiar benthic environment covered with particular white inorganic ‘coatings’. In other intact rock glacier streams, this type of ‘coating’ was demonstrated to be strongly enriched in heavy metals and rare earth elements (As, Al, Ni, Cu, Y, Li; Thies et al. 2013; Wanner et al. 2023). From an ecological perspective, this benthic substrate may represent an extreme environment for the development of benthic biofilms, which is a key food resource for alpine invertebrates (Niedrist and Füreder 2017). Also, the ingestion of toxic elements covering organic detritus and biofilms might be harmful for sensitive aquatic organisms (Lencioni et al. 2023). Thus, we could speculate that D. steinboecki may be resistant to heavy metals, as previously demonstrated (in glacier springs) for D. zernyi (Lencioni et al. 2023) and organic synthetic compounds for D. tonsa and D. zernyi (Lencioni 2018). Notably, these two species were also found at the Hochjoch/Giogo alto icy seep (HJ-RG) (but again, with single specimens). The pre-adaptation of D. steinboecki, as well as the other (low-density) co-dwelling taxa (including also Rhipholophus sp., Enchytraeidae, and Acarina), to high heavy metal concentrations was never investigated but can be only hypothesized based on their occurrence in glacier springs with high concentrations of the same heavy metals. Evidence of a physiological link between cold resistance and extraordinarily high toxic resistance of kryal Diamesa species has been emphasized also by Trenti et al. (2022). Thus, we speculate that a strong chemical filtering may drop down the densities of chemically resistant and cold-adapted organisms and hinder the presence of several species typical of “warmer” springs and not resistant to harsh chemistry. The necessity of D. steinboecki to live in habitats with very low invertebrate diversity, as also suggested by the outcomes of the species associations analyses, is generally attributed to its inability to cope with competition in the community (Lencioni 2018).

Icy seeps may become increasingly important refugia also for cold-adapted taxa that are not exclusive of glacier springs (Brighenti et al. 2021b). These include species strongly associated with a large glacier influence, like D. goetghebueri, D. zernyi and D. latitarsis (Lencioni 2018) and found at non-negligible densities in the investigated icy seeps communities. In addition, our data indicate that icy seeps are vital habitats for species like D. tonsa, Pseudokiefferiella parva or Eukiefferiella brevivalvar, with high relative abundances and a central arrangement in the species association network of this spring type, as well. Other cold-adapted species that are commonly found in relatively cool (< 4–6 °C) spring brooks, and that might be threatened by rising temperatures and the colonization from downstream communities (Niedrist and Füreder 2021), may find additional climatic refuge in icy seeps (e.g., Eukieferiella minor, Heleniella ornaticollis, Paratrichocladius rufiventris, Tvetenia calvecens, Tvetenia bavarica and Thienemaniella majuscola in our study). Furthermore, an increasing frequency of droughts and flow intermittency are predicted for Alpine areas due to climate change, with predicted strong ecological effects for benthic invertebrate communities (e.g., Piano et al. 2019; Chanut et al. 2023). Rock glaciers and other landforms that support icy seeps also host aquifers, and the storage capacity of these aquifers is known to increase along with permafrost ice loss (Hayashi 2020). Consequently, icy seeps may also represent climate refugia for cold-adapted species that cannot bear droughts and flow intermittency. However, this combined thermal and hydrological refugial role of icy seeps can only be hypothesized, as more research is needed to assess their actual ecological characteristics within alpine stream networks.

Environmental filtering and species interactions inside and between communities dictates the assembly processes in alpine river networks (Brown et al. 2018). While the former process generally prevails in the harsh glacier springs, the ability of different species to coexist more strongly prevails in spring brooks. With their intermediate harshness, icy seeps encompass a wide gradient of mildness/harshness (Brighenti et al. 2021c), where environmental filtering and species associations play a mutual role in species sorting. In our study, this was evident in the analysis of taxa associations. Not surprisingly, given the low density and diversity, taxa associations were rare in glacier springs. While the possible positive associations, when normalised with the total number of associations, were in the same order of magnitude in icy seeps and spring brooks, the negative ones were much higher in spring brooks than in icy seeps. This may imply that, in icy seeps, co-dwelling dominates over competition, at least more than in spring brooks. This may be related to the intermediate-disturbance conditions of the physical and chemical habitat in icy seeps. The same suboptimal conditions may also make these springs more resistant against new colonizers, and this might be an important prerequisite for their refugial character. Within this context, the biogeographical role of ‘white-coated’ icy seeps, where the environmental filtering is strongest and thus species colonization from spring brook communities is unlikely, is still to be investigated further.

The species colonisation in new habitat patches from insect taxa is strongly controlled also by the ability of non-aquatic adults to disperse and effectively reproduce. In our study, we investigated different spring types located sufficiently close to each other to allow between-habitat colonisation even from poor dispersers. The establishment of new populations by mass effect can only occur if suitable habitat conditions can be found also during the adult stages. This is also valid for the survival of cold-adapted species under warming air temperature (Lencioni et al. 2022). The cold ground conditions typically found at intact rock glaciers may provide suitable microclimatic conditions for adults of cold specialists, as previously demonstrated for fully terrestrial invertebrates (Millar et al. 2015; Gobbi et al. 2021b), and perhaps an additional filter against the colonization of downstream communities.

Alpine aquatic ecosystems are considered by popular perception as uncontaminated and pristine, with no or limited influence from human pressures. However, stressors like hydropower production, land-use changes, water abstractions, winter and summer tourism, and fish stocking represent increasing threats to these environments (Knight 2022; Schmeller et al. 2022). Habitat fragmentation and water exploitation are increasing in the European Alps and other mountain areas, and the interaction of local pressures and larger-scale drivers (such as changing climate) has synergistic and complex ecological effects (Elsen et al. 2020). These drivers affect the resilience of mountain habitats, and their capacity to support ecosystem services such as local and downstream water security, the functionality of river networks and their cultural value (Drenkhan et al. 2023). Hence, the protection of alpine environments and their unique geomorphological and ecological diversity, represents an increasing challenge for conservation (Schmeller et al. 2022). Globally, the present cover of protected areas only partially includes those streams that will offer climate refugia for cold-water specialists (Wilkes et al. 2023). In some mountain areas, the local legislation offers some status of protection to cold-specialist aquatic invertebrates threatened by climate change (e.g., Lednia tumana, listed in the US Endangered Species Act, 2019). However, none of the species investigated in this study have a specific protection status. Given the strong affinity to cold-adapted spring taxa to glaciers and rock glaciers, the protection of cold-adapted species and glacier stream specialists should be necessarily associated to the preservation of the hydrological processes occurring at these landforms. For glaciers, the deterioration of these processes (meltwater production) generally (but see Anacona et al. 2018) depends on large-scale drivers (climate change), against which local efforts should be dedicated to avoid additional pressure. In contrast, the climatic resistance of permafrost and the water storage capacity of rock glaciers and other landforms are supposed to make icy seeps less sensitive to climatic warming in the short-time (Wagner et al. 2021). However, the low discharge often found at these springs, coupled with the limited availability of the habitat extent, makes icy seeps particularly sensitive to local-scale drivers. As such, a sustainable management of their water should be a strategical priority, and particularly the prevention of water abstractions. The little knowledge on the ecology of alpine stream species hinders the development of detailed conservation strategies (Gobbi et al. 2021a). Thus, the active and precautionary protection of glaciers, rock glaciers and their aquatic environments should be paired with an increased knowledge on these ecosystems. Given the low densities at which such spring invertebrate populations dwell, and the scarcity of available habitats for cryophile biodiversity, the ecological effects of research should account for little-impact to no-impact strategies (Lencioni and Gobbi 2021). Efficient tools such as the “climate refugia conservation cycle” (Morelli et al. 2016) offer methods for nature conservation, while ensuring a good compromise with the needs from local communities. This is a particularly difficult goal in relatively densely populated and frequented mountain areas, like the European Alps. There, the conflicts between water use and environmental flow requirements are predicted to exacerbate, due to a long-term decline of water availability, flow-intermittency, and droughts frequency.