Ecological theory suggests that multiple environmental factors shape local species assemblages by progressively filtering species from the regional species pool to local communities (Baattrup-Pedersen et al., 2015), and one of the main challenges in community ecology is understanding the processes driving the functional structure of the species assemblage (Villégér et al., 2010). Three types of ecological processes, or filters, produce non-random patterns in the community structure, acting on species traits, rather than on species themselves: the abiotic environment, biotic interactions, and dispersal (McGill et al., 2006; Violle et al., 2007; Lukács et al., 2019). Species traits are morphological, physiological, and phenological features measurable at the species or individual level (Violle et al., 2007). Abiotic environment acts as filter selecting species with suitable traits that can persist in the given habitat; therefore, increasing similarity in trait composition and reducing functional diversity (MacArthur and Levins, 1967; Kraft et al., 2008). Biotic interactions, or niche differentiation, usually prevail on a finer scale where co-existing species have small overlap in their functional niches, usually by competitive exclusion; leading to divergent traits and increased functional diversity (Mason et al., 2007). The third filter influences the community structure by selecting species according to their ability to disperse to a site (Paz et al., 2021). Interaction of these three processes is complex, and can lead to shifts in the community structure, changing the abundance and/or presence of species (Cadotte and Tucker, 2017). Disturbances could lead to changes in the interaction and relative importance of filters in the community structure, acting directly through the loss of species, or indirectly by modifying environmental conditions (Baattrup-Pedersen et al., 2003; Rambaud et al., 2009). According to the stress-dominance hypothesis, environmental filtering is important under stressful, disturbed conditions, while biotic interactions are more important in structuring communities in less disturbed environments (Weiher and Keddy, 1995; Swenson and Enquist, 2007; Lukács et al., 2019).

The establishment and structure of macrophyte communities in rivers are largely determined by physical factors (i.e., water current, light availability, bottom substrate) that combine the effects of water and sediment chemistry on community composition (Butcher, 1933; Canfield and Hoyer, 1988; Baattrup-Pedersen and Riis, 1999; Riis et al., 2001; Riis and Biggs, 2003; Daniel et al., 2006; Janauer et al., 2010; O'Hare et al., 2010; Steffen et al., 2013). Macrophyte communities are strongly influenced by river hydrology, reflecting both anthropogenic and natural disturbances, where these environmental factors have stronger effect on the trait composition than on the species composition of the community (Papastergiadou et al., 2016; Göthe et al., 2017; Baattrup-Pedersen et al., 2018; Bejarano et al., 2018). Due to the great potential that functional diversity and trait composition of macrophyte communities might have in bioindication of hydro-morphological disturbances in running water-bodies, and its potential importance in river management, there is a growing number of studies investigating the influence of these factors on the trait distribution and/or functional diversity of macrophyte communities (e.g., Baattrup-Pedersen et al., 2016; Göthe et al., 2017; Lukács et al., 2019; Manolaki et al., 2020; Paz et al., 2021; Stefanidis et al., 2021). However, there is still a lack of data on the effects that hydro-morphological disturbances have on the functional diversity and trait composition of macrophyte communities in a large, heavily modified river, such as Danube. After the millennia of human activities on its banks, Danube is far from being natural river (Olson and Krug, 2020; Janauer et al., 2021). The most extensive interferences with Danube course are dams of the hydro-power plants, causing discontinuities in river flow and disruption of longitudinal connectivity, and consequently a series of environmental changes, including decreased water flow, increased sedimentation, increased transparency, etc. (Birk et al., 2012). It was found that free-running sections of the Danube River are poorer in number and abundance of macrophyte species than the impounded reaches (Janauer et al., 2021).

In the middle section of the Danube River (the same section that was analyzed here), it was found that species diversity and the abundance of macrophyte assemblages increased in the downstream direction towards the dams (i.e., in the run of the river reservoirs), and were negatively correlated to the riparian zone width (Vukov et al., 2018). In the impoundments, discontinuity in flow imposed by anthropogenic alteration—damming, and loss of riparian shading, alleviate many of the constrains to the plant growth found in the free-running section (water velocity is lower, sedimentation is increased, transparency is higher), while the higher retention time leads to accumulation of nutrients and eutrophication that contribute to macrophyte biomass production (Smith et al., 2006; O’Hare et al., 2010; Nõges et al., 2016). It was concluded that macrophytic cover and species diversity in the main channel of the Danube River in its middle course indicates major anthropogenic impact—damming, and consequent eutrophication (Vukov et al., 2018). Since the environmental filters act on species traits, rather than on species themselves, the main idea behind this study was to analyze the response of the trait based functional diversity indices and trait structure of macrophyte communities to the set of environmental variables used in previous study (Vukov et al., 2018) in order to compare it to the response observed previously for the diversity of macrophyte communities on a species level.

Therefore, the aims of this study were: 1) to investigate the effects that hydro-morphological disturbances, both natural and anthropogenic, have on the functional diversity of macrophyte communities in the middle section of the Danube River; and 2) to identify macrophyte traits that respond to the analyzed hydro-morphological factors. Achieving these objectives will provide an insight into the prevailing ecological processes shaping macrophyte communities in the heavily modified water bodies, with possible implications regarding the river bioassessment and management.

Original data on the abundance and distribution of 49 macrophyte species (Supplementary Table S1), and on eleven environmental parameters (Table 1, Supplementary Table S2), in 947 sampling units were collected during the survey of the Danube River in Serbia done in the summer of 2014 (Vukov et al., 2018). A survey of macrophytes was conducted according to European Standard EN 14184:2014 (2014), using a five-level scale for species abundance estimation (1—very rare; 2—rare; 3—frequent; 4—abundant; 5—very abundant) in relation to the total volume and length of the sampling unit. Studied section of the Danube River is located between river km 1433 and 846. Survey was done in the river’s main channel and in the side channels and side arms with surface inflow. Studied river reach was divided into 1081 contiguous sampling units that were 1 km long in the main channel, and of varying length in the side channels and side arms. Left and right riversides were surveyed separately. Survey was done in the whole length of the sampling unit visually and/or by raking, using small boat. In this study we have used data from 947 sampling units in which macrophytes have been recorded.

Studied section of the Danube River is located between river km 1433 and 846. The upstream part of the surveyed reach (Pannonian Plain Danube) runs through a floodplain landscape (avg. slope is 0.04‰), with an average width of approximately 750 m, and a mean depth of 6 m. At km 1071, the Danube enters Đerdap gorge (Iron Gate, slope gradient of the riverbed ranges from 0.04‰ to 0.25‰, average width of the river is 750 m). From km 931 the river again runs through a floodplain landscape (avg. slope 0.04‰). Average width is 830 m, and mean depth is 8.5 m. In the Đerdap section, two hydropower plants with large dams were built: Iron Gate 1 at km 943 (60 m high, 1278 m bank to bank, operational since 1972); Iron Gate 2 at km 863 (35 m high, 412 m bank to bank, operational since 1984). These modifications resulted in creation of two contiguous runs of river reservoirs. The entire studied river reach is part of the Danube navigation corridor.

Data on 18 macrophyte traits and 65 trait states were extracted from the literature (Borhidi, 1993; Willby et al., 2000; Table 2). Traits: Rooting at nodes, High below: above-ground biomass, Evergreen leaf, and Amphibius, had binary states (0 meaning the trait is absent, and 1 meaning the trait is present). States of other traits had values: 0 for absence, 1 for occasional but not general presence, and 2 for general presence of the trait state (Willby et al., 2000; Baattrup-Pedersen et al., 2016).

The environmental variables included in the RDA model (Figure 1A) explained 36.7% of the variation in functional diversity metrics (R ² adjusted = 0.367). The first RDA axis alone explained 35.48% of variation in the response data. It is worth mentioning that the first residual eigenvalue (PC1 representing 46.13% of the variance in response data) is larger than the first canonical eigenvalue (RDA1) in the obtained RDA model, indicating that there could be other factors responsible for the variability in the response data. Variables: river_km (r = −0.793), dam_dst (r = −0.752), and ripz_w (r = −0.545) were strongly correlated with the first RDA axis, whereas the highest correlation with the second axis was detected for trnspr (r = −0.595) and sdm_txt (r = −0.653). Permutation tests of RDA results showed that the RDA model, as well as the first RDA axis, was statistically highly significant. According to the forward selection procedure, eight variables are enough to reach the R ² adjusted threshold. Their order of importance is: river_km, trnspr, dam_dst, sdm_txt. cnnctv, bnk_txt, ripz_w, prt_bnk. Highest absolute values (above 0.30) of the correlation coefficients extracted from the RDA were recorded between all functional diversity metrics and variables: river_km, dam_dst, and ripz_w (Supplementary Table S3). The strongest correlation was found between river_km and FDiv (−0.518), followed by FRed (−0.514), FEve (−0.513), FDis (−0.490), RaoQ (−0.475), FD (−0.452), and FRic (−0.378). The correlation coefficient between dam_dst and FDiv was −0.486; followed by FRed (−0.484), FEve (−0.473), FDis (−0.448), RaoQ (−0.439), FD (−0.429), and FRic (−0.410). Correlation between ripz_w and diversity metrics was: FRed (−0.376), RaoQ (−0.332), FDis (−0.328), FD (−0.325), FDiv (−0.315), FRic (−0.310), and FEve (−0.304).

Variation Partitioning revealed that the hydro-morphological features of the environment explained 33.1% of the functional diversity metrics in total; 19.8% was the part of the variability jointly explained with anthropogenic alterations, while for 13.3% hydro-morphology was solely responsible. Anthropogenic alterations of the river habitat explained 23.3% of the variability in the functional diversity metrics in total; 19.8% of the variability was explained in combination with hydro-morphology, and for 3.5%, anthropogenic alterations were solely responsible (Figure 1B). Permutation tests of all fragments indicate their high significance.

Fourth-corner tests revealed a total of 83 statistically significant associations (out of 704 possible) between macrophyte trait states and environmental variables. Thirty-one of them were positive and 52 were negative correlations (Figure 2A). Output of the fourth-corner analysis was corrected for multiple testing using the FDR (false discovery rate) procedure, leaving six out of 83 associations (Figure 2B). Trait state plinsct (pollination by insects) was positively associated with the environmental variable trnspr (water transparency), and negatively associated with river_km (distance from the river’s mouth) and dam_dst (distance from the dam). Trait state G (generalists) was negatively associated with dam_dst (distance from the dam). BF3 (high body flexibility) was negatively associated with cnnctv (level of connectivity to the river’s main channel), as well as trait state rhzm (dominant mode of reproduction is by rhizomes).

Our study assessed the effects of hydro-morphology and hydro-morphological alterations on functional diversity and trait structure of macrophyte communities in the Danube River in its middle course. The results showed that the examined predictors were responsible for substantial portion of the explained variance of functional diversity metrics, indicating the importance of hydro-morphological disturbances, both natural and anthropogenic, in shaping the macrophyte communities in running waters. This finding corresponds with the results of previous studies done in different types of running water-bodies (Lukács et al., 2019; Manolaki et al., 2020; Paz et al., 2021). Also, our results indicated that there could be other important environmental factors, since the effects of e.g., nutrient levels and other physicochemical factors, were out of scope of this study.

We have found that distance from the mouth, distance from the dam and riparian zone width, along the first RDA axes, and water transparency and riverbed granulometry, along the second RDA axes, were most important factors constraining functional diversity metrics. The values of all functional diversity indices increased in the downstream direction, towards the dam, in the river reach with narrow/none riparian zone, and with fine material on the riverbed. Similar result was found for the species diversity in the same reach of the Danube, where the abundance and species diversity of macrophyte communities showed the same pattern (Vukov et al., 2018). The Variation Partitioning analysis showed that the combined effect of morpho-hydrological factors (according to RDA, river’s directional flow as the most important) and anthropogenic alterations (according to RDA, damming as the most important) explained largest part of the variability in functional diversity metrics, larger than any of them explained separately. Individual relation of each of the functional diversity metrics revealed their similar response to the variables representing the distance from the mouth and the distance from the dam. In both cases the highest correlation was found between those factors and Functional Divergence (FDiv), Functional Redundancy (FRed), and Functional Evenness (FEve). Higher values of FDiv, found in the downstream direction, toward the dam indicate higher degree of niche differentiation, and consequent low resource competition. It is suggested that communities with high functional divergence may have increased ecosystem function as a result of more efficient resource use (Mason et al., 2005). Higher FRed indicates the similarity in trait states, meaning that some species within the community perform similar functions, improving its resilience to the loss of species and ecosystem functions (Ricotta et al., 2016). High FEve indicates the high degree of the trait abundance distribution of the community in the niche space, allowing effective utilization of the entire range of resources available to it, hence providing increased productivity and stability of the community (Mason et al., 2005). Width of the riparian zone was also found to be the important environmental factor constraining functional diversity metrics, with the strongest correlation to Functional Redundancy, as well as with two metrics of functional dissimilarity—Rao quadratic entropy (RaoQ) and functional dispersion (FDis). Manolaki et al. (2020) found that low values of functional dissimilarity related to the riparian zone width are consequence of habitat homogeneity. Therefore, its high values found here, might be the result of habitat heterogeneity, and consequent heterogeneity of niches (wide riparian zone—shaded, narrow/none riparian zone—open) found in the impoundments (Vukov et al., 2018).

Distance from the mouth is both spatial and typological factor, indicating the catchment size and the position of the studied section in the context of the whole river, it comprises the river flow parameters—discharge and flow velocity, which provide longitudinal connectivity for the river reaches (Vannote et al., 1980; Franklin et al., 2008). In contrast, dams are disruption of longitudinal connectivity of the river, and cause a series of environmental changes, including decreased water flow, increased sedimentation and increased water transparency, accumulation of nutrients and eutrophication, contributing to macrophyte biomass production (O’Hare et al., 2010; Birk et al., 2012). Increased values of all functional diversity metrics in the downstream direction and towards dam indicate shift from predominant effect of abiotic environment to the prevalence of biotic interactions as filters shaping structure of macrophyte communities (MacArthur and Levins, 1967; Stubbs and Wilson, 2004; Mason et al., 2007; Kraft et al., 2008; Lukács et al., 2019). In the impoundments, high values of functional diversity metrics confirm the findings of previous studies, that biotic filters—competition and limiting similarity (i.e., niche differentiation) play role in aquatic ecosystems under high productivity (Engelhardt and Ritchie, 2001; Fu et al., 2014). According to the stress-dominance hypothesis, environmental filtering is important under stressful conditions, while competitive interactions are more important in the benign environments (Weiher and Keddy, 1995; Swenson and Enquist, 2007). Therefore, the conditions in the impoundments found in the surveyed section of the Danube River, although the result of major anthropogenic modification of the river habitat, enabled the development of rich and functionally diverse aquatic vegetation.

Although fourth-corner tests revealed quite large number of significant associations between cumulated weighted means of macrophyte trait states (CWMs) and environmental variables, after performing FDR correction, only five of them remained. Pollination by insects was trait positively associated with higher water transparency, and negatively associated with the distance from the mouth and distance from the dam. This association indicates the assemblages of entomophilous, also mainly amphibious species that are established in the impoundments, were the water transparency is higher. In this case the association with the higher water transparency is indirect, while the conditions in the impoundments (narrow/none riparian zone, shallows along the banks), enable their growth as well as pollination by various insects preferring open, sunny habitats. Generalist strategy was negatively associated with the distance from the dam. Generalists are group of stress-tolerant widely distributed plants that are able to thrive in a wide variety of environmental conditions and to use different resources, and therefore dominate the communities (Borhidi, 1993). Having the wide variety of traits, they contribute to the functional diversity, and consequent resilience of the communities (Zografou et al., 2020). High body flexibility and dominant mode of reproduction by rhizomes were negatively associated with the level of connectivity to the river’s main channel, placing them in the side channels and side arms. Rhizomes provide clonal growth, which can theoretically lead to unlimited horizonal spreading of the population, competition avoidance, and exploitation and saving of resources (Klimešová, 2018; Manolaki et al., 2020). They also provide the lateral mobility and the capacity to survive unfavorable conditions (Baattrup-Pedersen et al., 2016). Since the surveyed side water-bodies dry up during the low water level, plants with rhizomes are fit to survive, and due to their capacity to reproduce by rhizomes, particularly when the water level is too high during the flowering periods, enable their populations to thrive. High body flexibility is essentially related to stronger water currents (Baattrup-Pedersen et al., 2016) which are not found in the side arms and side channels. Species assemblages with this trait are abundant in the side water-bodies with frequent and intense changes in the water level due to the hydropeaking in the hydropower plant, and are resilient to the vertical movement of water.

Our study confirmed that macrophyte functional diversity metrics respond well to the natural and anthropogenic hydro-morphological disturbances, and are valuable indicators of prevailing ecological processes shaping macrophyte communities in the large rivers. These metrics might prove to be useful diagnostic tools that are needed to identify cause(s) of ecological degradation and to guide the choice of relevant management measures. The potential of CWMs as indicators of environmental stressors is relatively low. According to Lukács et al. (2019) this shortcoming might be improved by applying more precise continuous traits and more specific and relevant gradients in the future.