Size–Abundance Relationships of Freshwater Macroinvertebrates in Two Contrasting Floodplain Channels of Rhone River

: Body size is perhaps the most fundamental property of an organism and its relationship with abundance is one of the most studied relationships in ecology. Although numerous studies have examined these relationships in local communities, few have investigated how they vary at different temporal and spatial scales. We investigated the relationship between body size and abundance of local macroinvertebrate communities in two ﬂoodplain channels of the French upper Rhone River. The two channels differ in their vegetation coverage (high vs. low vegetation) and hydrological regimes. The shapes of the size–abundance relationship were similar between channels on a yearly basis but differed when compared between months. The variation in local size–abundance relationships between months was related to variation in the functional diversity across time. Our ﬁndings suggest that local size–abundance relationships are able to quantitatively describe temporal changes in community structure, showing the importance of relating diversity with ecosystem function in a more realistic context.


Introduction
Body size is a key component of the diversity and structure of animal assemblages [1,2] as it encapsulates many ecological traits of species [3,4]. One of the most important relationships of body size is with abundance, as recognized in both terrestrial and aquatic ecology [5][6][7]. The relationship between body size and abundance, according to Damuth's Rule, scales with a scaling exponent of −0.75 [8][9][10]. Body size (or mass) is also related to metabolic rate with a scaling exponent of 0.75 (i.e., metabolic theory in ecology (MTE) [11]). Furthermore, body size of the species is negatively related to abundance at the scale of local to global communities. Such size-abundance relationships [7] have remained at the core of our understanding of ecology for more than 20 years [12][13][14][15][16].
The power function of −0.75 implies that the total energy or resource used (as a function of abundance and metabolism) by all the different-sized species is independent of their body size, known as the "energetic equivalence hypothesis (EER)" [17]. This pattern of size-abundance relationships has been confirmed using global-scale data sets compiled from the literature, such as the global size-abundance relationship (GSDR following [7]). However, these studies are observational and are subject to the normal limitations of an observational study, making it difficult to draw strong conclusions of the mechanisms that control size-abundance relationships. Indeed, other studies suggest that using smaller scales data sets (i.e., local communities' data [18,19]), such as local size-abundance relation-selected channels is very different. Molottes channel (MOLO) is located in the bypassed section of the Brégnier-Cordon hydroelectric scheme where the regulated flow delivered by the diversion dam is 150 m 3 s −1 from 1 June to 31 August, 100 m 3 s −1 from 1 September to 31 October, and 80 m 3 s −1 during winter [40]. This floodplain channel is influenced by backflow entries generated by water level elevation (i.e., return flow from the bypassed river section) in relation to the functioning of the hydroelectric power station. Ponton channel (PONT) is located downstream of the hydroelectric scheme and hydropeaking usually causes daily discharge fluctuations from Monday to Friday (between~200 and 500 m 3 s −1 ). Upstream flow connections occur during high floods for less than one day per year for both floodplain channels on average [41].

Studied Sites
The two Rhone floodplain channels, which are former braided side-arms, are presently both connected downstream with the main river, but disconnected upstream, at average discharge. In this sector of the French Upper Rhone, between Geneva and Lyon, the mean summer discharge is about ~700 m 3 s −1 ) (see Figure 1). The hydrological context of the two selected channels is very different. Molottes channel (MOLO) is located in the bypassed section of the Brégnier-Cordon hydroelectric scheme where the regulated flow delivered by the diversion dam is 150 m 3 s −1 from 1 June to 31 August, 100 m 3 s −1 from 1 September to 31 October, and 80 m 3 s −1 during winter [40]. This floodplain channel is influenced by backflow entries generated by water level elevation (i.e., return flow from the bypassed river section) in relation to the functioning of the hydroelectric power station. Ponton channel (PONT) is located downstream of the hydroelectric scheme and hydropeaking usually causes daily discharge fluctuations from Monday to Friday (between ~200 and ~500 m 3 s −1 ). Upstream flow connections occur during high floods for less than one day per year for both floodplain channels on average [41]. The two floodplain channels vary in length (1220 m for MOLO and 910 m for PONT) and geographic position (4 km between the two channels). A slightly deeper gully (depth minimum ~0.30 m), through which water passes continuously, allows the maintenance of a downstream connection for both channels with the main channel during low flow periods. In contrast, the upstream part of both channels was not modified by restoration works. Because of the loss of competence of the river flow when it enters into the floodplain channel from downstream, downstream alluvial plugs, composed of a mix of sand and silt were deposited in the two channels, allowing the establishment of aquatic plants, e.g., Elodea nuttallii and Ceratophyllum demersum, which, for the MOLO channel exclusively, became denser in the downstream part of the channel. Regarding PONT, the shallow upstream part of the channel is completely emerged over 15 m during reduced The two floodplain channels vary in length (1220 m for MOLO and 910 m for PONT) and geographic position (4 km between the two channels). A slightly deeper gully (depth minimum~0.30 m), through which water passes continuously, allows the maintenance of a downstream connection for both channels with the main channel during low flow periods. In contrast, the upstream part of both channels was not modified by restoration works. Because of the loss of competence of the river flow when it enters into the floodplain channel from downstream, downstream alluvial plugs, composed of a mix of sand and silt were deposited in the two channels, allowing the establishment of aquatic plants, e.g., Elodea nuttallii and Ceratophyllum demersum, which, for the MOLO channel exclusively, became denser in the downstream part of the channel. Regarding PONT, the shallow upstream part of the channel is completely emerged over 15 m during reduced discharge periods. During high discharge periods, low-to-moderate water depth (<50 cm) covers the previously emerged part. Aquatic vegetation development is generally low in the PONT channel and mainly restricted to the upstream and mid-channel parts. Furthermore, aquatic vegetation cover provides egg-laying sites for, e.g., gastropods, and a substrate for periphyton, which, in turn, is highly influential for other organisms. Energy flow of these habitats is also influenced by periodic desiccation of the shallowest parts (e.g., [42]) or by stable and low flow conditions [43,44]. Despite these highly diverse habitats, there have been no modeling efforts to study how the different flow conditions affect the transfer of energy of these channels.
Finally, basic information on physicochemical data (e.g., water temperature, conductivity, and dissolved oxygen) of the water column in MOLO and PONT channel for each month (e.g., April, May, June, July) were also collected (see Table 1). These abiotic data are not analyzed in this paper but they have a profound impact on the size-abundance relationships.

Sampling Method and Ecological Traits
A total of 24 samples was used, where 3 samples were collected from each of 2 given floodplain channel sites on the same date of each month (across 4 months). For each channel site, the 3 samples were collected from 3 random spots along a 30 m transect (in a straight line). Specifically, benthic invertebrate samples were taken in the middle part of the two floodplain channels over four dates in early 2017: April, May, June and July. The period chosen for sampling corresponds to a time of much higher diversity and productivity compared with winter or fall. After July, many aquatic insects with annual voltinism emerge [45] however, others undergo extreme conditions (e.g., dissolved oxygen and temperature conditions) where the abundance of the macroinvertebrates decreases, thus limiting productivity and functional diversity.
Benthic invertebrates living among aquatic macrophytes and in or on the sediment/litter upper layer were both sampled within a 0.5 × 0.5 m metal frame with a hand net (mesh size 500 µm). The collected samples of benthic invertebrates were weighed on a balance as wet weight (g) for each individual present in the sample. Benthic invertebrates were then sorted under a stereomicroscope, where they were counted (i.e., abundance) and identified to genus or species level, when possible. Diptera were identified up to the family level.
Feeding type was extracted for each taxon from the freshwater ecology database available online (http://www.freshwaterecology.info, accessed on 31 July 2021). It consists in assigning a positive score (from 0 to 10) that describes the affinity of a taxon for predefined feeding categories, i.e., filtering collectors, gathering collectors, grazers, shredders and predators (see Appendix B- Table A4).

Statistical Analysis
A linear model was fitted to the raw data of average body size relative to their average abundance (both variables in log 10 , see Figure 2) per taxon, in order to test the size-abundance relationship of freshwater macroinvertebrate among the two channels. This approach to studying the size-abundance relationships was based on calculating local size-abundance relationship (LSDR), which plots average size and average abundance of each taxon (see [7]). Least-squares regression (LSR) was fitted on average wet weight and average abundance (number of individuals): i. among the two channels (all months included), ii. among the two channels and across months, and iii. among the two channels and across functional feeding groups.

Statistical Analysis
A linear model was fitted to the raw data of average body size relative to their average abundance (both variables in log10, see Figure 2) per taxon, in order to test the size-abundance relationship of freshwater macroinvertebrate among the two channels. This approach to studying the size-abundance relationships was based on calculating local size-abundance relationship (LSDR), which plots average size and average abundance of each taxon (see [7]). Least-squares regression (LSR) was fitted on average wet weight and average abundance (number of individuals): i. among the two channels (all months included), ii. among the two channels and across months, and iii. among the two channels and across functional feeding groups. We used least-squares regression (LSR) for analyzing body size and abundance relationships. LSR is more appropriate than reduced major axis regression when measurement error in the independent variable (body mass) is less than that of the dependent variable (abundance), as is the case in our study [46,47]. We used log10transformation to normalize the data variation, and to permit proportional, linear relationships to be readily discerned [48,49]. Significant differences between scaling exponents (slopes) and intercepts (elevations) were estimated by ANCOVA (with body mass as a covariate).

Results
Both LSDRMOLO and LSDRPONT showed significant relationships between body weight and abundance of macroinvertebrate taxon ( Figure 2, Table 2). The scaling slopes were shallower than the energetic equivalence hypothesis value of −0.75 (−0.39 and −0.33, for MOLO and PONT, respectively), thus the slopes and intercepts were not significantly different between the two channels (Table A1). We used least-squares regression (LSR) for analyzing body size and abundance relationships. LSR is more appropriate than reduced major axis regression when measurement error in the independent variable (body mass) is less than that of the dependent variable (abundance), as is the case in our study [46,47]. We used log 10 -transformation to normalize the data variation, and to permit proportional, linear relationships to be readily discerned [48,49]. Significant differences between scaling exponents (slopes) and intercepts (elevations) were estimated by ANCOVA (with body mass as a covariate).

Results
Both LSDR MOLO and LSDR PONT showed significant relationships between body weight and abundance of macroinvertebrate taxon ( Figure 2, Table 2). The scaling slopes were shallower than the energetic equivalence hypothesis value of −0.75 (−0.39 and −0.33, for MOLO and PONT, respectively), thus the slopes and intercepts were not significantly different between the two channels (Table A1). Table 2. Results of the LSR analyses of log 10 abundance in relation to log 10 body weight in two floodplain channels of the Rhone River.    Finally, the scaling slopes of LSDR MOLO and LSDR PONT across functional feeding groups showed similar patterns for both channels, with the exception of the predators and shredders (Figure 4, Table 4). Specifically, gathering collectors showed steeper scaling slopes for LSDR MOLO and LSDR PONT (−0.60 and −0.45, respectively), whereas filtering collectors (−0.27 and −0.35, respectively) and grazers (−0.31 and −0.28, respectively) resulted in less steep slopes for both LSDR MOLO and LSDR PONT . The scaling slopes across functional feeding groups were not significantly different for the two channels, while the intercepts were different for grazers, filterers and gatherers (Appendix A- Table A3). Finally, predators and shredders showed different LSDR slopes, where only predators of PONT (−0.16) and shredders of MOLO (−0.49) were significant.   However, at a monthly scale, the average body size and average abundance across functional feeding groups of MOLO and PONT were consistent with the monthly patterns of LSDR MOLO

Patterns of LSDR MOLO and LSDR PONT (All Months Included)
The patterns of the LSDR across the two floodplain channels, considering the whole dataset including all the months, were both statistically significant, with similar slopes (i.e., −0.39 and −0.33 for MOLO and PONT, respectively) that were significantly shallower than predicted by Damuth [8][9][10]. Our patterns were consistent with several studies that suggested that the relationship between body size and abundance is much shallower than the value of −0.75 for aquatic macroinvertebrate communities (e.g., [26,[50][51][52]) and other animal communities (e.g., [19,[53][54][55][56]). However, the shallower slopes found were not consistent with the EER. This deviation of LSDR provides evidence against the universality of EER, showing that the amount of energy used is not the same for all species [7]. Because the EER predicts that size-abundance relationships arise from the metabolic scaling, environmental factors that alter size-dependent metabolic demands may alter size-abundance relationships [17]. Therefore, the shallower slope of LSDR in both freshwater channels implies a shallower slope of the metabolic scaling with body size.
Indeed, the metabolic scaling slope of macroinvertebrate species in aquatic ecosystems was previously found to be shallower than the expected MTE value of 0.75 (reviewed by [57]). For reasons that are still not completely understood, the metabolic scaling slope of benthic macroinvertebrates often shows less steep slopes (i.e.,~0.30 to~0.55) than the −0.75 value. Whether this variation in macroinvertebrate metabolic scaling slope is related to environmental factors, such as water temperature [58,59] and biological factors, such as predation pressure [60,61], is less known. Moreover, this variation of metabolic responses in relation to predation threats may be temperature-dependent [62,63]. These studies showed that smaller macroinvertebrates in both freshwater springs and saltwater lagoons more strongly increased their metabolism under high temperature and predation threat, when compared with the larger taxa. However, large macroinvertebrates, that do not change their metabolism in the presence of a predation threat, were present at a higher abundance. This again implies that the slope of the size-abundance relationships of benthic macroinvertebrates should scale with a less steep slope (i.e.,~−0.30 to~−0.55), and indeed this was found for both channels considered in our study (Figure 3). The size-dependent effects of abiotic and biotic factors on metabolic rate may change the metabolic scaling slope and, as a consequence, it may influence the size-abundance relationship slope. This may explain the competitive advantage of large macroinvertebrate species over smaller ones also observed by other studies [64,65].

Patterns of LSDR MOLO and LSDR PONT across Months
The size-abundance relationship varied at the monthly scale for both channels (i.e., LSDR MOLO and LSDR PONT ). The size-abundance relationship across the two sites showed different, site-specific patterns for April and May, while LSDRs for June and July followed comparable patterns. Our findings were consistent with other studies that showed a temporal [24,33] and spatial (reviewed by [6]) variation in size-abundance relationship of local communities. This pattern implies that time and space are both able to shape the local communities, resulting in different shapes of size-abundance relationships, for which differently sized species acquire energy differently.
PONT showed a slightly steeper size-abundance relationship for the months of May and June, while MOLO showed a slightly steeper size-abundance relationship only for April, compared with the other months. These patterns show an increase in the abundance of small taxa during May and June for PONT, and April for MOLO. Small aquatic macroinvertebrates may have reached their highest abundance during May and June for MOLO and during April for PONT. This is because reproduction and the appearance of juveniles (gastropods in particular) tend to peak during this period of time (i.e., from April to July) for most macroinvertebrate species [66]. However, this peak for each species or functional group varies because of differences in life cycles and resource availability. Thus, the LSDRs changed in a different way for MOLO and PONT across time, following the proposed hypothesis based on the specifics of reproduction and juvenile appearance, implying that juveniles of the species present in PONT peaked at a different moment across the four months than those in MOLO. Therefore, our findings suggest that although the size-abundance relationships of the two sites were broadly the same when looking at all months combined, this result does not hold when considered at the monthly scale.

Patterns of LSDR MOLO and LSDR PONT across Functional Diversity
The size-abundance relationship of local communities at both sites taking functional diversity into account, showed the same shape with different intercepts, with the exception of the predators and shredders. LSDR PONT showed significantly higher abundance of filtering and gathering collectors, independent of body size, than LSDR MOLO . However, LSDR MOLO showed significantly higher abundance of grazers regardless of their body size than LSDR PONT . Our findings therefore showed different energy flow patterns for the two channels. Filtering and gathering collectors in PONT and grazers in MOLO are shown here to have higher energy flow compared with the other functional groups. This suggests a difference in the energy flow across the functional groups, depending on differences in the habitat structure of the two channels [67]. To repeat, MOLO was characterized by slow and less active backflows from the by-passed section of the river, whereas PONT was subjected to more frequent oscillations associated with hydropeaking in the full river and characterized by lower vegetation density.
The various macroinvertebrate functional feeding groups translate their diversity in terms of food acquisition strategies. Specifically, in our study the filtering feeders were mostly bivalves, gathering collectors were mostly insects and some gastropods, grazers were mostly gastropods, and shredders were mostly crustaceans. The difference of the intercept of the functional groups independent of their body size implies higher energy flow for the functional feeding groups with higher intercepts [62]. Our findings showed that the energy flow of filtering and gathering collectors was higher in PONT, while the energy flow of grazers was higher in MOLO.
Our study indicates that energy transfers through floodplain feeding groups are linked to and may change as a function of the vegetation cover and the water flow conditions in the two channels. Specifically, the functional groups of filtering collectors, gathering collectors and grazers showed different patterns in energy transfers between the two channels [68]. These variations suggest that water movement (enhanced by hydropeaking) may increase the suspended matter load as food for filtering and gathering feeders in MOLO [66]. Low and more stable flow conditions may increase temperature and promote biofilm/algal growth upon macrophytes for grazing in PONT [69]. Because the energy flow in floodplain channels probably relies on hydrology in addition to their connection (or reconnection in the case of restoration) with the main channel, it would be possible to adapt water releases from dams in the Rhone River to control energy transfers in floodplain channels [70]. Restricting hydropeaking (see [71]) might suffice to allow the proliferation of grazers that would tend to increase the energy transfers and thus support more top predators for PONT than MOLO channel.
Additionally, functional feeding groups were able to also explain the different monthly LSDR slopes of the two channels. Specifically, small-sized filtering and gathering collectors showed higher abundance for MOLO in May and June, whereas small-sized grazers showed higher abundance for PONT in April. Therefore, our findings suggest that the LSDR shape of most functional feeding groups was apparently the same, but they showed a different energy flow (i.e., higher abundance regardless of their body size). Furthermore, the body size and abundance of functional feeding groups when considered at a monthly scale were able to explain the monthly pattern of LSDRs in the two channels. Our finding confirms that the changes in habitat conditions across time and space, such as vegetation coverage and water flow, are closely related to the functional diversity and LSDR shape.

Conclusions
Size-abundance relationships link individual-and population-level energetics with the structure of ecological communities based on principles that are generally applicable. Broadly evaluating the variation of size-abundance relationship in local communities (i.e., LSDR) and understanding how it arises from biological and ecological processes at population-and community-levels would represent a major advance in our present understanding of ecosystems. This study offers a potential method of using size-abundance relationships for describing community structure, and the underlying processes across time, space, habitat type and functional diversity. This methodological approach links biodiversity and ecosystem functioning processes, improving our ability to make predictions of how environmental change will shape aquatic ecosystems, and is relevant for research and management policies for the conservation of biodiversity.

Acknowledgments:
We thank all the research teams involved in the monitoring of the Rhône River ecological restoration (RhônEco program) for making data available and for discussions. We thank the Syndicat du Haut-Rhône for their technical support. Furthermore, we would like to thank Pablo Timoner for his help on classifying the species into functional feeding groups. We are grateful for the helpful suggestions and comments by three anonymous reviewers.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A
In order to quantify the difference between the two channel sites we have used ANCOVA (p value) for slope and intercept comparison between LSDR MOLO and LSDR PONT : across all months, and across each month. Specifically, the differences between slopes were assessed by comparing 95% CI. When the slopes were not significantly different, the differences between elevations were estimated by ANCOVA (with body mass as a covariate). Least-squares regression (LSR) was fitted on average body size and average abundance (number of individuals): i. among the two channels (all months included- Table A1), ii. among the two channels and across months (Table A2), and iii. among the two channels and across functional feeding group (Table A3).
a Significance of slope differences; b Significance of intercept differences; * p < 0.05; ** p < 0.005; ns-not significant; -not measurable. Table A3. p value for slope and intercept comparison of the LSDR analyses in Table 3. The differences among slopes were assessed by comparing 95% CI. When the slopes were not significantly different, the differences between elevations were estimated by ANCOVA (with body mass as a covariate).

Feeding Types
a Significance of slope differences; b Significance of intercept differences; * p < 0.05; ns-not significant; -not measurable.

Appendix B
This table contains the estimated average wet weight, abundance and feeding traits for 54 taxa collected across 2 floodplain channels in the French upper Rhone River. Below are the column names and a brief description of them: Channel = channel name (MOLO-Molottes and PONT-Ponton; Month = month of sampling; Site = sampling site in each channel (1, 2, or 3); Taxon = species name; FFG = functional feeding groups; Abundance = total abundance per taxon.