Shifts in Riparian Plant Life Forms Following Flow Regulation

: Flow regulation a ﬀ ects bordering riparian plant communities worldwide, but how di ﬀ erent plant life forms are a ﬀ ected by river regulation still needs further research. In northern Sweden, we selected 10 rivers ranging from free-ﬂowing to low, moderately, and highly regulated ones. In 94 reaches across those rivers, we evaluated the relative abundance of woody and herbaceous (i.e., graminoids and forbs) life forms, their species richness, and their relative presence. We also explored which, and to what extent, hydrological variables drove species assembly within each life form. The relative abundance and species richness of each life form decreased across river categories with increasing levels of regulation. This was particularly apparent in herbaceous life forms, and the most drastic decreases were observed in all life forms in moderately or highly regulated reaches. Additionally, when river regulation increased, the relative presence of many species from all life forms decreased. Unlike woody species, only a few new herbaceous species appeared in regulated reaches. A canonical correspondence analyses (CCA) revealed that a wide range of hydrological variables explained the occurrence of woody species, while fewer variables explained variation in the graminoid and forb life forms. We conclude that ﬂow regulation and its intensity result into clear shifts in the relative abundance of di ﬀ erent life forms, as well as in changes of within-group species richness and composition. Consequently, the modiﬁcation of certain ﬂow attributes in ﬂow regulation schemes, as well as the intensity of these modiﬁcations, may alter the ratio between herbaceous and woody species, ultimately impacting the functions and beneﬁts derived from each life form.


Introduction
It is obvious that changed flow regimes affect plant growth, performance, and distributions [1,2]. However, since studies on many species at once are relatively scarce or restricted to smaller spatial scales, many questions about community responses are still unanswered. This is unfortunate, as general frameworks (as opposed to species-specific or site-specific frameworks) are essential to develop flow management guidelines for regional or national water management plans. To overcome the difficulties of working at a species level, classification into functional groups has been advocated and used in other ecological theories [3]. Functional groups include species sharing certain traits, resulting in common  We obtained hydrological and floristic data from 94 reaches across 10 rivers from [16], as available in the Dryad repository [19]. The hydrological data consisted of 16 variables that characterize ecologically relevant facets of the annual hydrograph at a reach scale (e.g., the magnitude, timing, duration, frequency, and rates of change of the flows; Table 1 and Annex 1). As an overall measure of the degree of regulation, we calculated the percentage of the natural mean annual discharge potentially stored in reservoirs upstream from the reach. Based on the degree of regulation, we categorized reaches into free-flowing (<0%; n = 63) and with a low (>0% and <15%; n = 14), moderate (>15% and <60%; n = 8), and high (>60%; n = 9) degree of regulation. The floristic data of each reach consisted of plant species' presence-absence, as recorded along a 200 m long section by the Landscape Ecology research group (Umeå University, Sweden) (Annex 2). Species were classified into the simplest life form groups: woody, graminoid, and forb species. The few number of woody species compared to the herbaceous ones prevented us from subdividing this group into trees and shrubs. Conversely, the large number of herbaceous species led us to subdivide them into graminoids and forbs. For the sake of simplicity, we discarded other more complex groupings (e.g., [6]). Hydrological series were either recorded (the longest from 1909 and the shortest from 1988) or modeled (when no recorded data were available; from 1999 to 2011). Floristic data were collected during the nineties by the Landscape Ecology Group at Umeå University. The reader is referred to the [16] for more details on the dataset. Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.

Hydrological Variable
Free  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1 Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.     Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.    Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.    Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1 Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.  Table 1. Mean and standard deviation of the hydrological variables characterizing each flow regulation group. "n" indicates the number of sites included within each flow regulation group. Letters "a," "b," "c," and "d" identify significant differences between groups (p < 0.05), while the trend across groups is indicated with an arrow.

Data Analysis
Differences between flow regulation groups (i.e., free-flowing, low regulated, moderately regulated, and highly regulated) for each hydrological variable were tested with a Kruskal-Wallis test followed by a multiple comparison BH (Benjamini-Hochberg) test. This was appropriate because the data were not distributed normally. We used the R package "dplyr" [20] in R 3.6.1.
Three vegetation metrics were computed. First, we computed the 'species richness'(#) within the woody, graminoid, and forb functional groups for each regulation group as the number of species from each life form present in free-flowing, low-, moderately-, and highly-regulated reaches. Second, we computed the 'relative abundance' (%) of each life form for each regulation group (i.e., mix of plant life forms) as the ratio between the species richness within each life form (see above) and the total number of species (including woody species, graminoids and forbs) present at each regulation group. Third, we computed the 'relative presence'(%) of the species within each life form in each flow regulation group as the number of reaches from a regulation group where a species was present in relation to the total number of reaches of that regulation group.
Afterwards, we fitted linear regressions for each individual species with 'relative presence' as dependent and the 'percentage of regulation' as independent variables, and we extracted the regression coefficients using the lme4 package in R [21]. A positive regression coefficient indicates that the species benefits from regulation (i.e., the presence of the species is more likely when regulation intensifies), whereas a negative coefficient indicates that the species is disfavored by regulation (i.e., the presence of the species is less likely when regulation intensifies). From the pool of all species, we also noted whether each species from each life form were (i.e., appeared) or were not (i.e., disappeared) in reaches beyond 15% regulation (i.e., moderately and highly regulated). For the interpretation of the results, we clustered the individual species in the following response groups: positive response (i.e., species with regression coefficients >2), negative response (i.e., regression coefficient <−2), impassive response (i.e., regression coefficient <2 and >−2), present only when regulation is beyond moderate (>15%), present only when regulation is high (>60%), absent when regulation is beyond moderate (>15%), and absent when regulation is high (>60%). By 'regulation beyond moderate' we mean in moderately regulated reaches, in highly regulated reaches, or in both, whereas by 'high regulation,' we mean only in highly regulated reaches. Subsequently, we calculated the percentage of species in each response group for each life form group.
Finally, we ran three canonical correspondence analyses (CCA), explored to what extent the hydrological variables drive species assembly within each life form, and identified which hydrological variables play key roles. We used the presence-absence-species-by-reaches dataset and the hydrological data per reach as the input and tested the significance of the CCA models and axes using the R package "vegan" [22]. The relative importance of each hydrological variable was evaluated using the loadings on the first three axes that were extracted. Higher loadings indicate a higher importance of that variable for that axis and thus for the spread of species along that axis. We considered 0.45 as the threshold of significance [23]. For the interpretation of the results, we indicated species by their response groups to flow regulation. Statistical analyses that were performed in R used the 3.6.1 version [24].

Hydrological Variables
All hydrological variables differed significantly (p < 0.05) among flow regulation groups except for the duration of the minimum monthly flows (Table 1). Multiple comparisons tests showed that differences between regulation groups reflected decreased long-time scale (i.e., intra-annual) flow variation but increased short-time scale (i.e., daily) flow variations. That is, when comparing groups with increasing levels of flow regulation, we observed an increased magnitude of the minimum monthly flows, an increased duration of the maximum monthly flows, an increased frequency of the daily reversals and extreme low flows, and increased rates of change in the daily discharge, as well as delayed maximum monthly flows and two-year floods. Furthermore, we observed a decreased magnitude of the maximum monthly flows and two-year floods, a decreased duration of the extreme low flows and two-year floods, and a decreased frequency of monthly reversals and early minimum monthly and extreme low flows (Table 1). In summary, these changes indicated that regulation stabilizes intra-annual flows but results in highly fluctuating and faster daily flows, earlier minimum and later maximum flows, and shorter extreme high and low flow events

Riparian Plant Species and Life Forms
Across the 94 study reaches, 372 plant species consisting of 44 woody species, 107 graminoids, and 221 forbs were found. In each life form, species richness decreased across categories with increasing flow regulation (Figure 2a). When comparing free-flowing conditions to high flow regulation, species richness decreased, on average, by 20% for woody species, 40% for graminoids, and 45% for forbs (Figure 2a). While the decrease was rather gradual for woody species, there was a more sudden decrease in species richness between free-flowing and 'low regulation' for the graminoid and forb species (Figure 2a). The diversity of the forb species substantially contributed to the diversity in natural riparian zones, with the ratio of relative abundance between woody, graminoid, and forb species being 13/29/58 (Figure 2b). Due to a slight increase in relative abundance of woody species and a decrease in the relative abundance of herbaceous species (graminoids and forbs), the herbaceous:woody ratio decreased from 6.7 in the free flowing category to 4.4, 5.7, and 5.1 in the low, moderate, and high flow regulation categories, respectively (Figure 2b).
Across all functional groups, the relative presence of most species correlated negatively with the degree of flow regulation (Figures 3a and 4; Annexes 3 and 4). A relatively higher percentage of woody species had a strong negative correlation compared to graminoids and forbs, among which, many showed an impassive response. That is, negative responses were observed for 61%, 44%, and 36% of woody, graminoid, and forb species, respectively (Figure 3a). Additionally, the mean response within those groups differed, with the average regression coefficient being −8 (±6.6 standard deviation) for woody species, −6 (±6.1 SD) for graminoids, and −4 (±4.4 SD) for forbs. The relative presence of fewer species correlated positively with the degree of flow regulation, and the same fraction of species in each life form group showed a positive response to increased regulation (i.e., 11% woody species, 16% graminoids, and 13% forbs) (Figure 3a). The strength of the relationship did not differ much between life forms (on average +3.0 (±3.0 SD)). While almost half of the graminoid and forb species were impassive to regulation, only a quarter of the woody species were impassive (Figure 3a).
A high number of species was solely found in free-flowing reaches, irrespective of life form. Furthermore, 70% of all woody, 46% of all graminoid, and 53% of all forb species found in the entire study grew along reaches with a free-flowing and/or low-regulated flow regime. A lower fraction of species was sensitive to high regulation, as 9% of all woody, 20% of all graminoid, and 6% of all forb species were absent in reaches with highly regulated flow regime (Figure 3b and Annex 3). In contrast, only a few graminoid and forb species were restricted from free-flowing and/or low-regulated reaches, as 4% of all graminoid and 5% of all forb species only occurred in reaches with an above moderate flow regulation, with, respectively, 1% and 2% of the species occurring exclusively in reaches with high flow regulation regimes (Figure 3b and Annex 3). No woody species were restricted to reaches with a higher flow regimes.     The three CCA models (i.e., for woody species, graminoids, and forbs) were statistically significant ( Table 2). The first two CCA components were significant in each CCA model, whereas the third and fourth components were significant for the woody and forb species but not for the graminoid species (Table 2). The highest cumulative variance was observed in woody species, for which the selected hydrological variables explained up to 52% of the variance in the first three axes, whereas 49% and 40% variance was explained by the first three axes for graminoids and forbs, respectively ( Table 2). Woody species showing positive, negative, and impassive trends in response to increased degrees of flow regulation were well segregated in CCA space. Negative trends were found for willow Salix glauca L., mosses (Lycopodium selago (L.) Bernh. ex Schrank & Mart. and L. annotinum L.), heaths (Vaccinium microcarpum Aiton, Vaccinium oxycoccos L., Andromeda polifolia L., and Ledum palustre L.), Betula nana L., Linnaea borealis L., and Rosa majalis Herrm., whereas most willows (Salix herbacea L., Salix lapponum L. and Salix pentandra L., Salix lanata L., Salix triandra L., Salix hastata L., and Salix cinerea L.) along with other shrubs such as the black and red currants (Ribes ningrum L. and Ribes rubrum L.), raspberry (Rubus idaeus L.), and trees such as birches (Betula pendula Roth and Betula pubescens Ehrh.) showed marked positive or impassive trends with regulation (Figures 4 and 5; Annex 5). Hydrological variables that were strongly correlated to woody species arrangement in CCA space were related to the percentage of regulation, extreme low flow conditions (duration, frequency, and timing), monthly flows (magnitude of maximum and minimum), and daily flows (frequency of reversals and fall rates) ( Figure 5 and Table 2). In general, woody species with positive and impassive trends were separated along the second axis that aligned with a higher percentages of regulation and related hydrological characteristics such as shorter, more frequent, and earlier extreme low flow periods; lower maximum but higher minimum monthly flows; and more frequent and faster daily flow reversals. These latter two had highest loading on the second axis ( Figure 5 and Table 2). The opposite was true for species characterized by negative trends. The three CCA models (i.e., for woody species, graminoids, and forbs) were statistically significant ( Table 2). The first two CCA components were significant in each CCA model, whereas the third and fourth components were significant for the woody and forb species but not for the graminoid species ( Table 2). The highest cumulative variance was observed in woody species, for which the selected hydrological variables explained up to 52% of the variance in the first three axes, whereas 49% and 40% variance was explained by the first three axes for graminoids and forbs, respectively (Table 2). Woody species showing positive, negative, and impassive trends in response to increased degrees of flow regulation were well segregated in CCA space. Negative trends were found for willow Salix glauca L., mosses (Lycopodium selago (L.) Bernh. ex Schrank & Mart. and L., annotinum L.), heaths (Vaccinium microcarpum Aiton, Vaccinium oxycoccos L., Andromeda polifolia L., and Ledum palustre L.), Betula nana L., Linnaea borealis L., and Rosa majalis Herrm., whereas most willows (Salix herbacea L., Salix lapponum L. and Salix pentandra L., Salix lanata L., Salix triandra L., Salix hastata L., and Salix cinerea L.) along with other shrubs such as the black and red currants (Ribes ningrum L. and Ribes rubrum L.), raspberry (Rubus idaeus L.), and trees such as birches (Betula pendula Roth and Betula pubescens Ehrh.) showed marked positive or impassive trends with regulation (Figures 4 and 5; Annex 5). Hydrological variables that were strongly correlated to woody species arrangement in CCA space were related to the percentage of regulation, extreme low flow conditions (duration, frequency, and timing), monthly flows (magnitude of maximum and minimum), and daily flows (frequency of reversals and fall rates) ( Figure 5 and Table 2). In general, woody species with positive and impassive trends were separated along the second axis that aligned with a higher percentages of regulation and related hydrological characteristics such as shorter, more frequent, and earlier extreme low flow periods; lower maximum but higher minimum monthly flows; and more frequent and faster daily flow reversals. These latter two had highest loading on the second axis ( Figure 5 and Table 2). The opposite was true for species characterized by negative trends.   7  10  13  16  19  22  25  28  31  34  37  40  43  46  49  52  55  58  61  64  67  70  73  76  79  82  85  88  91  94  97  100  103  106  109  112  115  118  121  124  127  130  133  136  139  142  145  148  151  154  157  160  163  166  169  172  175  178  181  184  187  190  193  196  199  202  205  208  211  214  217 Table 1 and Annex 1.  Table 1 and Annex 1.
Graminoid species with different responses to flow regulation were not as well segregated in the CCA space as woody species. However, species with positive trends were located more to the left of the first axis ( Figure 5). Among graminoids that showed negative trends to increased flow regulation, we found several Carex species (Carex acuta L., Carex Lasiocarpa Ehrh., Carex rostrata and Carex canescens L., but many Carex species were impassive to regulation (e.g., Carex elongata L., Carex bigelowii Torr. ex Schwein., Carex ornithopoda Willd., Carex capitata L., Carex saxatilis L., Carex digitata L., Carex livida (Wahlenb.) Willd., and Carex loliacea L.). Species from the genus Festuca sp. (e.g., Festuca vivipara (L.) Sm., Festuca rubra L., and Festuca ovina L.) also responded positively to regulation. Other graminoids with positive trends to increased flow regulation were: Eriophorum scheuchzeri Hoppe, Calamagrostis chalybaea (Laest.) Fr. and Calamagrostis stricta (Timm.) Koeler, Juncus trifidus L., Alopecurus aequalis Sobol., Juncus trifidus L., and Agrostis mertensii (Trin.) Kuntze (Figures 4 and 5; Annex 5). From the CCA, it became clear that a considerably smaller number of hydrological variables explained this arrangement for graminoids on the first two axes compared to woody species. Positive and impassive trends of graminoids aligned with higher percentages of regulation and, subsequently, more frequent extreme low flows and delayed floods ( Figure 5). The opposite was true for species with negative trends to increased flow regulation.
Finally, forb species with positive, negative, and impassive trends in response to increased flow regulation were mixed in the CCA space ( Figure 5). For example, Sedum rosea L., Erigeron borealis L., Cerastium alpinum L., Persicaria foliosa (H.Lindb.) Kitag. Cicuta virosa L., and Arabis arenosa L. showed positive trends, whereas Euphrasia frígida Pugsley, Euphrasia stricta Kunth, Melampyrum pretense L., Persicaria lapathifolia (L.) Delarbre, and Spergula arvensis L. showed negative trends (Figure 4), but all were scattered throughout the biplot ( Figure 5 and Annex 5). A couple of hydrological variables correlated to forb species arrangement, with variation along the first axis being related to decreased intra-annual flow fluctuation (i.e., frequency of monthly reversals) and variation along the second axis being related to an increased duration of the extreme low flows ( Figure 5 and Table 2). Unlike the CCA of the other life forms, the variable 'percentage of regulation' had a minimal effect on species distribution, as indicated by much lower axes loadings for this variable in the forb CCA compared to the CCA's of the other life forms (Table 2).

Discussion and Conclusions
The high spatial and temporal heterogeneity of environmental conditions such as hydrology, geomorphology, light, and temperature determine the richness and distribution of plants along riparian zones of free-flowing rivers. Hydrological factors, however, are seen as the most important ones [25]. In boreal streams, hydrological heterogeneity results from a flow pattern with a strong seasonality and ice disturbance. During fall and winter, flows are low, while during spring, large and snowmelt-driven floods shape rich riparian plant communities. Characteristic riparian forbs, graminoids, shrubs, and trees are arranged as clear belts at different elevations, parallel along the channel because each life form is thought to have a different resistance to the hydrological forces of the spring flood [26]. Naturally recurrent ice disturbance usually restrains the development of woody plants at lower elevations, while it favors annual and biennial species (i.e., mostly forbs) or perennial hemicryptophytes that die back to ground level during winter. The natural flow conditions imposed by the spring flood can explain the typically high ratio between herbaceous and woody species in the riparian areas of boreal free-flowing rivers [27].
In our study area, river regulation is mostly due to hydropower production, and the selected hydrological variables showed that it drastically alters flow regimes at long and short times scales. At long time scales, it evens out differences in monthly flows, induces shorter, more frequent, and earlier extreme low flows, and it also delays floods. At short time scales, regulation enhances daily flow fluctuations, with hydropeaking in some extreme cases [28]. Our results reinforce other studies demonstrating that flow alterations that are associated with regulation translate into declined riparian diversity and modified riparian community composition [29][30][31]. Our study showed how flow regulation through changes in various flow attributes exerts differing effects over specific riparian life forms and breaks the balance between them that exists in riparian communities along free flowing rivers.
The substantial decrease in the herbaceous:woody species ratio, as observed in regulated reaches in our study, suggests stronger impacts of regulation on the richness of graminoids and forbs compared to woody species. Unlike the authors of [32], who described the almost complete disappearance of shrubs and trees over the herbaceous species following dam-induced changes in the winter flooding in the Yangtze River, in our study, the richness of herbaceous species (the sum of graminoids and forbs) declined more strongly with increased regulation compared to woody species (both in absolute numbers as well as a net reduction, i.e., the species that appear minus the ones that disappear). Herbs, characterized by short generation times and ruderal life strategies, seem to be especially affected by long-term flow stabilization, as such species benefit from natural formation of disturbance patches [33]. Such bare patches allow for germination, and other studies on the effect of river regulation on species diversity have frequently associated reduced species richness with difficulties in germination and establishment [32][33][34][35]. In addition to recruitment, those bare patches may change the competition among woody, graminoid, and forb species. In riparian systems, frequent occurrence of disturbances could relax competition for resources, especially light [12,36]. The observed shift towards less short-lived species in our regulated rivers is therefore consistent with life-history theories regarding adaptations to disturbance [33,37]. Empirical studies in riparian zones further confirm that annuals increase following increased flooding and/or give way to perennials with increased time since major flood disturbance [30,32,[38][39][40]. A field experiment in riparian zones in the semiarid savanna showed a comparable recruitment of grasses and forbs in treatments with and without water addition and disturbance [12]. In our study however, graminoids were more responsive to regulation than forbs.
The results of our CCA suggested that graminoid community composition seems to be particularly affected by regulation through delayed floods and more extreme low flow events, whereas forb community composition was less clearly affected by most hydrological variables, as only intra-annual flow stabilization was able to explain some of the variance. This somewhat contrasts results by the authors of [41,42] who found that herbaceous species are very responsive to changes in flow regime. The variability in the woody, graminoid, and forb community compositions that was unexplained by hydrological variables suggests that other environmental variables may co-determine community composition. Our results point that such other variables gain importance in herbaceous communities. In addition to hydrology, the variables of land-use, nutrient availability, soil moisture, air temperature, and light availability have been shown to be particularly critical for riparian community composition [43][44][45]. In addition to intra-annual flow stabilization, the alteration of many other hydrological variables was found to determine the community composition of woody species in our study. The community composition of woody species appeared to be sensitive to both long term (intra-annual) and short-term fluctuations in flows. On the one hand, the desynchronization of a species' life story stages with intra-annual flow regime patterns may strongly affect woody species, which highly depend on recruitment from seeds rather than clonal dispersal-as is common among herbaceous species [46,47]. Seed dispersal and establishment are highly dependent on flood schedules, and other studies in the same area [16] have described important impacts of regulation on woody species as being linked to, among other phenological traits, their particular dispersal and germination periods. On the other extreme, very frequent and rapid variations of daily flows related to intense hydropower production schemes may hamper woody species germination and survival [48,49]. Such short-term changes in flows cause repeated wetting and, ultimately, nearly permanent moist or inundated soils. In general, woody, graminoid, and forb species have intrinsically different water requirements and tolerances during most life stages, with graminoids and forbs being better adapted to waterlogged environments than woody plants [50,51].
In addition the described negative effects of regulation, the fact that few species from all life forms increased their presence along regulated reaches and that some graminoids and forbs only occurred at higher levels of flow regulation, demonstrated that meager positive effects may be also expected following regulation. According to the deletion/addition/replacement theory (i.e., assembly rules by the authors of [52]), species-specific differences in the ability to tolerate hydrological alterations may underlay the observed wide range of species responses. Eventually, this results in a non-random sorting of species along hydrological gradients [2,53] and across reaches with different degrees of regulation. Numerous examples from scientific literature have shown that woody species (both native and non-native) frequently take advantage of the new hydrological conditions and invade regulated reaches (e.g., [54,55]). Though we found positive responses to increased flow regulation among woody species, there were no woody species that uniquely occurred in reaches with some degree of regulation. This may indicate dispersal limitations [56] or that requirements for establishment were not met under regulated conditions. Conversely, although of a relatively small number, the fact that new graminoids and forbs appeared in regulated reaches suggested that there is a source of new species from the upstream and upland herbaceous flora that can find success under the new hydrological conditions. However, no exotic species were found among new species appearing only in regulated reaches, which, on the other hand, is not surprising at this high latitude [57].
Woody, graminoid, and forb species coexist in the riparian areas, and flow regulation may affect those life forms differently, thus leading to unbalances in community composition. Our results showed that riparian diversity in boreal rivers is driven by graminoid and forb life forms, as well as that regulation, due to a reduction of graminoid and forb diversity, results in overall poorer communities. This agrees with trends in other plant communities suffering from pressures such as heavy logging and overgrazing [9]. When population and community structure are altered, an array of ecosystem functions can be affected too, in particular when shifts in community structure includes (major) shifts in plant functional types [58,59]. For example, impoverished riparian woodlands may ultimately stimulate bank collapse through reduced rooting structures, decreased nitrogen assimilation, and changed stream hydrology, temperature, and primary production [60]. Similarly, the impoverishment of the herbaceous riparian community may affect erosion-related processes like trapping and the storing processes of suspended sediment, or provisions of habitat and refuge for biota. This impoverishment may also induce changes in phosphorous assimilation [61]. Since riparian community composition also affects macroinvertebrate and fish assemblages, changes in the ratio between herbaceous and woody species following flow regulation may have cascading effects. For instance, salmonid fishes could be particularly negatively affected by changes in herbaceous community composition [62]. Moreover, unbalances between life forms may also influence the success of the restoration of degraded systems [63]. In conclusion, the modification of certain flow attributes in flow regulation schemes, as well as the intensity of these modifications, result in shifts in the relative abundance of woody, graminoid, and forb life forms, as well as changes of the within-group species richness and community composition. Consequently, flow regulation may alter the characteristic herbaceous:woody species balance of riparian plant communities, ultimately impacting the functions and benefits derived from riparian zones.
Supplementary Materials: http://www.mdpi.com/1999-4907/11/5/518/s1. Annex 1. List of hydrological variables. Annex 2. List of species. Annex 3. Species from each life form group which appeared or disappeared compared as the intensity of flow regulation increases, compared to the free-flowing species pool. Annex 4. Code for each forb species according to the slope of the fitted line from the regression between the mean percentage of regulation and the relative presence of each species (see Figure 4). Code 1 corresponds to the lowest slope and code 221 to the highest slope. Annex 5. Species scores for the three first CCA components, the slope of the fitted line from a linear simple regression between the mean percentage of regulation and the relative presence at the study reaches of each species from each life form group, and the assigned response according to the slope (i.e., negative, positive or impassive). The list of species names is in the Annex 2.