An increase in stream water temperatures is being detected worldwide because of global warming. Mean increases of at least 0.5°C per decade have been observed (Daufresne et al., 2004; Durance and Ormerod, 2009; O’Reilly et al., 2015; Liu et al., 2020; Niedrist and Füreder, 2020). As global warming continues, stream temperature predictions suggest rises of 1.5°C by 2030, and, in some regions of the world, up to 4°C by 2050 (Wanders et al., 2019). Increased water temperatures in rivers can be accelerated by anthropogenic structures like weirs and dams, which slow the flow and promote warming of the upper water layers (Schneider et al., 2013; Zaidel et al., 2021). Water temperature is directly related to oxygen concentrations which are crucial for biota in these systems (Piatka et al., 2021). Greater seasonal variation in water temperature and higher temperature extremes in summer can result in severe stress on riverine organisms (Durance and Ormerod, 2009; van Vliet et al., 2013), with cold stenothermic species being most heavily affected (Fullerton et al., 2018). Consequences include the decline of cold-water adapted native species and a potential increase of opportunistic invasive species (Daufresne et al., 2004; Haidekker and Hering, 2008; Rahel and Olden, 2008). While cold-stenothermic fishes such as salmonids can quickly move to more oxygen-rich cold-water patches (reviewed in Smialek et al., 2021), these suitable thermal habitats likely become scarcer, more distant and less accessible in future climatic scenarios (Kuhn et al., 2021). For other organisms such as macroinvertebrates (MIV) during the aquatic phases of their life cycles, reaching potentially cooler water may even be harder given their limited mobility compared to fish, which may leave them more exposed to short-term extremes in water temperature. Especially in hot and dry summers, when the predicted maximum temperatures have the potential to exceed the upper tolerance levels of many MIV species (Leigh et al., 2015), pronounced changes in MIV community composition can be expected (Lento et al., 2021). MIV play a key role in regulating functional processes in streams and are an essential food source for higher trophic levels (Wallace and Webster, 1996; Huryn, 1998; Ruetz et al., 2002). Therefore, MIV community changes may have consequences for the entire food web and the resilience of riverine ecosystems.

The effects of thermal-induced seasonal changes in stream community composition are not fully clear. Still, shifts driven by species-specific thermal tolerances, especially in those having their entire life cycle in aquatic environments—such as gammarids—(Kolding and Fenchel, 1979), should be expected. Gammarids are ideal indicators to study stream temperature effects in aquatic habitats given their functional importance for these systems (Boeker and Geist, 2015) as well as their sensitivity to habitat structure and temperature, with clearly different thermal thresholds between species (Wijnhoven et al., 2003; Cottin et al., 2012; Halle et al., 2016).

The aim of this study was to analyze how spatio-temporal thermal changes may affect MIV communities, placing special emphasis on the comparison of native vs. non-native gammarids (Gammarus roeselii and Dikerogammarus villosus, respectively). We carried out sampling in combination with a cross-experiment (Figure 1) in a single restored stream system with three differentiated thermal regions [TR: cold (C), intermixed (I) and warm (W)] during spring and summer, to test the following hypotheses:

1) Differences in temperature between the three thermal regions will result in pronounced differences in MIV abundance and community composition.

In order to test our three hypotheses, we designed a dual experiment (Figure 1) to differentiate spatial vs. seasonal thermal effects on overall macroinvertebrate community composition and the targeted gammarids. The dual experiment consisted of 1) the characterization of the in-situ macroinvertebrate community during two seasons in three thermal regions (TR) of an interconnected stream system (hypotheses I and II), and 2) a cross-exposure of two gammarid species typically found in that stream system to investigate their survival in the 3 TR (hypothesis III), using standardized exposure to ambient stream water in the so-called “salmonid-egg floating boxes” (SEFLOBs, Pander and Geist, 2010).

This study examined the influence of water temperature and other important habitat variables in warm, cold and intermixed thermal regions on the composition of the MIV communities and the presence and survival of native versus non-native gammarids. Although very much influenced by season and the presence of macrophytes or gravel as substratum, temperature most strongly affected MIV species composition. Indigenous gammarids preferred colder thermal regions, whilst the invasive one, D. villosus, was dominant in warmer thermal regions. These results indicate that during rising water temperatures, community composition can change irrespective of other factors which is in line with our hypothesis I. It is well known that macroinvertebrate community composition can widely be predicted based on structural habitat (Poff, 1997; Gore et al., 2001; Cortes et al., 2011) and that a high structural habitat heterogeneity, particularly characteristic of the study system, explains high species diversity (Pander et al., 2016; Pander et al., 2018). Our findings provide evidence that temperature is also a strong driver often impairing structural habitat quality, particularly affecting cold-stenothermic species negatively. This is indicated by the highest R-value detected by PERMANOVA compared to the seasonal and habitat (GR-vs. MP-dominated habitats) factors tested in each of the thermal regions. Additionally, the comparatively low individual numbers of indigenous gammarids and high numbers of the invasive D. villosus in warm thermal region W compared to cold C (where indigenous gammarids were more dominant) clearly mirror the temperature effect. Other factors such as O₂, current speed, water depth, or substratum composition were less important in this study since they exerted similar effects throughout all thermal regions and were overall differentiated on a relatively low level.

We hypothesized (hypothesis II) that such differences in community composition will be most pronounced during the warmest season when maximum water temperatures and upper temperature tolerance thresholds of certain species are reached. Due to their evolutionary adaptation, species driving community changes at increased temperatures are often Mediterranean and Ponto-Caspian species which have already spread throughout the middle and northern regions of Europe (Hesselschwerdt and Wantzen, 2018). Some of them are potentially invasive, as evident for D. villosus in this study. D. villosus is known to have colonized all of the studied TRs since 2010, when a large-scale floodplain restoration project connected all aquatic habitats (Pander et al., 2016). In 2010, the species was detected in higher densities close to the source (main River Danube) in GR-dominated habitats with low-to-medium current speed. In contrast, currently, the species prefer MP-dominated habitats in the W thermal regions of the system. However, these habitats are also located close to the Danube River - from where this species originally invaded the system - excluding a potential distance to source effect as a driver of spatial distribution patterns (Pander et al., 2016; Pander et al., 2018). The MP-dominated habitats are characterized by less current speed resulting from the denser vegetation compared to habitats without macrophytes. The slower current results in increased fine sediment accumulation, which seem particularly attractive for D. villosus combined with high temperatures. However, since all TRs comprised habitats with and without macrophytes, this specific habitat structure does not play a major role in the composition of the gammarid community among the different TRs.

Either other gammarid species such as G. pulex or G. roeselii, which are known to also colonize MP-dominated habitats, are already above their preferred temperature range or unable to compete with D. villosus when temperatures are high and were therefore only recorded at low densities in warm TR. This explanation is not fully supported by the cross-exposure experiment that was used to test hypothesis III. Here, D. villosus as well as the tested indigenous gammarid species had highest survival rates in SEFLOBs exposed in the cold thermal region C. This indicates favorable conditions for both species there and it contrasts the situation found in the TR habitats where probably additional factors such as competition for food, microhabitat and predation play a role. Predation can be excluded in our study for both G. roeselii and D. villosus as well since they were locked in their compartments not affecting each other nor being affected from invaders outside of the compartments. We cannot exclude cannibalism, which is reported for D. villosus, however, we have no indication for this effect in our study. We neither could observe direct cannibalism during the SEFLOBs maintenance, nor did we find gnawed carcasses of D. villosus in the compartments. Alternatively, the mortality rate in cold water could be related to the fact that the specimens are less active and so their predatory behaviour is weak. Also, cold water has higher levels of oxygen saturation which may positively affect gammarid survival. Preference of D. villosus for cold water was to some extend also found in other studies, suggesting that this species can tolerate a wide range of temperatures (Wijnhoven et al., 2003) and that it can cope better with colder temperatures compared to other gammarids (e.g., Maazouzi et al., 2011). In contrast to these laboratory studies, both exposed gammarids herein seemed to persisted better in the SEFLOBs exposed in the cold thermal region, contrasting the results found in the natural habitats of the TRs. However, we have to state that overall mortality in the compartments was relatively high for both species. Findings generated in laboratory studies do not seem to be directly transferable to the natural environment where other factors such as direct and indirect competition, the availability of oxygen and shelter act together and may influence principle physiological tolerances of the competing species, potentially pushing indigenous species out of their preferred habitats. MP-dominated habitats in the thermal region C had a higher presence of indigenous gammarids as also detected in a study carried out at Lake Constance (Hesselschwerdt and Wantzen, 2018). This is in contrast with a laboratory study about the effects of direct interactions of invasive and non-invasive gammarids in combination with a fish predator. Beggel et al. (2016) demonstrated that D. villosus is more assertive when it comes to hiding in niches to avoid predation and we would therefore have expected that D. villosus would preferably colonize the macrophytic habitats since a high predation risk due to the rich fish population in the thermal regions is likely (Pander et al., 2015b). In the study herein, temperature in combination with structure may have affected competition between species, influencing indigenous species’ assertiveness in terms of greater competitive strength in cold and MP-dominated habitats.

Reaching thermal limits for macroinvertebrate communities can depend on different factors such as the initial species community, source habitats from which permanent colonization can be driven and the degree of temperature-related alteration these communities already experienced over time. Since it is known that the Danube system has been particularly affected by rising water temperatures for decades (Bonacci et al., 2008; Markovic et al., 2013), a warm-water adapted macroinvertebrate community could already be found in all TRs. This is evident from the absence of species favoring cold water conditions such as plecopterans of the genera Isoperla, which were quite regularly detected in that system until a decade ago (Pander et al., 2016; Pander et al., 2018). Such effects are potentially strongest during time periods when temperatures repeatedly exceed the thermal limits of cold-stenothermic species. This effect may be disguised in our dataset because several of those species have now disappeared from the system without the possibility of recolonization due to the general warming of the source habitats. However, our dataset still indicates a shift in the community composition in the warm TR where temperatures reached the detected maxima and resistant species of the genus Assellus and Proassellus along with D. villosus occurred in higher numbers compared to cooler time periods.

In light of the ongoing climate change with increasing mean temperatures and increasing numbers of extremes, it becomes obvious that novel MIV communities will arise, largely driven by species with higher temperature optima. In a globalized world, this can be dramatically boosted by the translocation of species from warmer regions, often happening accidentally and unnoticed (Walther et al., 2009; Brandner et al., 2015; Ricciardi et al., 2021), resulting in novel communities being permanently invaded and fast-changing (Rodriguez, 2006; Walther et al., 2009; Belnap et al., 2012). Up to now, it is not clear to which extent indigenous highly specialized species can cope with these temperature changes and ongoing invasions.

This study provides evidence that regional thermal effects on MIV community composition may exceed seasonal or meso-habitat effects, commonly known to govern such composition. Specifically, rising water temperatures have the potential to favour invasive D. villosus, with higher thermal tolerances. In light of climate change, our findings indicate that an impairment of habitat quality by increased temperatures may provide competitive advantage for invasive macroinvertebrate species. Consideration of shifts in community composition related to temperature needs to become integrated with biological response patterns related to other, better-characterized stressors such as morphological and flow degradation, chemical pollution and fine sediment. Such an approach is crucial for effective conservation and restoration of native biodiversity and for a realistic prediction of the ability to reach policy target settings for aquatic ecosystems.