Discussion

Phylogeny explains most of trait variations

Linking patterns to processes in community ecology is often challenging, especially in field-based studies. Several authors have suggested to use trait variations to infer community assembly mechanisms (McGill, 2008; Taudiere & Violle, 2016; Violle et al., 2007, 2012). To tease apart the multiple factors that drive community assembly in aquatic systems, we partitioned the variance of several traits among three distinct, yet interacting, scales, i.e. space, time, and phylogeny. Importantly, studies simultaneously taking different scales into account remain rare (but see e.g. Henn et al., 2018; R. Liu et al., 2018; Luo et al., 2016), and to our knowledge, none of them have so far considered an interannual effect over several years.
In the present study, variance partitioning among tested scales is uniform among traits, indicating that the processes determining these different trait values are similar. Community assembly mechanisms operating through time (year) and space (regional scale) are responsible for a low percentage of total trait variance, in line with the observations formerly reported by Fu et al. (2013). These authors also partitioned trait variance, but across spatial scales only, and they found that large spatial scale (i.e. regional scale including different lakes) does not account much for the root/shoot ratio, and leaf number variations in the macrophyte species Potamogeton maackianus . In our work, trait distributions are fairly conserved among sites and suggests a filter operating on the overall distribution of trait values within the region. Then, we showed that intra-site variations (inter and intraspecific scales) were overwhelmingly dominant in comparison with inter-site variations, for all traits. In other words, the trait means were variable among species in each site, and among individuals in each species. Thus, a mechanism participated in spreading trait values around the optimum values within sites. This observed pattern can arise from several non-exclusive hypotheses. First, adaptation to habitat heterogeneity within sites can lead to different optimal trait values depending on the micro-site considered. Second, biotic interactions can disperse trait values to limit the similarity between individuals, thus reducing competition. As highlighted by Messier et al. (2010), both niche and neutral theory can play a role in community assembly, since a filter select individuals based on their trait values rather than their species identity, finally constraining trait values at regional scale.
Interestingly, phylogeny accounted for the most of all trait variations, with in particular intraspecific trait variations being higher (56.29% on average) than interspecific variations in most traits. However, the relative importance of interspecific versus intraspecific trait variability was trait-depend which is consistent with previous studies showing that the relative amount of interspecific versusintraspecific trait variability differ among studies and traits (see Albert, Thuiller, Yoccoz, Douzet, et al., 2010; Fajardo & McIntire, 2010; Roche et al., 2004). We found LDMC and specific internode mass to vary more between species than within species. This indicates that traits related to resources conservation are highly conserved among species, allowing different species to specialize and co-occur along habitat gradients (different conditions among ponds). However, growth and resources acquisition traits (height, SLA, internode length, maximum root length) varied more within species than between species, in line with Siefert et al. (2015) who found that “whole-plant traits” (i.e. plant height, and architecture) were more variable within population than between species in a global-scale meta-analysis. This high amount of intraspecific trait variability allows individuals to better resist to filters, and the environmental heterogeneity at very fine scale (within-pond habitat variations) can lead to high variation in phenotypic trait values. Another non-exclusive hypothesis is that, in case of macrophytes, architectural traits can display multiple trait values since individuals are much less constrained by gravitational force than terrestrial species (Fu et al., 2013).
We confirmed that partitioning trait variance is a relevant approach to detect at which scale operate the most decisive processes in plant community assembly without scale dependency issues (McGill, 2008). Moreover, we underlined the strong importance of ITV in shaping community assembly (Jiang et al., 2016; Violle et al., 2012), and we then argue that not only species mean and community-aggregated trait values (Schamp et al., 2007; Stubbs & Bastow Wilson, 2004) but also trait distributions have to be analyzed when studying community dynamics and structure (de la Riva et al., 2016; Henn et al., 2018). Including ITV in community ecology can improve the detection of processes underlying species coexistence at local scale (Doudová & Douda, 2020; Fu et al., 2014; Violle et al., 2009).

Intraspecific trait variation in response to both biotic and abiotic variables

Trait based approaches are widely used in the literature to investigate individual responses to abiotic and biotic filters separately (Violle et al., 2007), although individuals usually cope with both types of filters within plant communities. As a consequence, our vision of ITV in a context of multiple constraints remains partial. In our study, we propose to fill this literature gap by simultaneously testing trait responses to different abiotic and biotic variables.
Contrary to our expectations, in pond living individuals, all traits did not respond to both abiotic and biotic variables. Height, SLA, and internode length on the one hand, and specific internode mass on the other hand, were respectively influenced by abiotic or biotic variables only, although these traits are known to usually respond to these two types of variables (Bittebiere & Mony, 2015; Loughnan & Gilbert, 2017; Rusch et al., 2011; Santamaría, 2002). However, all other traits variations (i.e. LDMC, maximum root length, and specific root mass) were due to both biotic and abiotic variables, as predicted. These different patterns of trait responses indicate that, in a context of multiple constraints, individual plant does not respond as a whole, but instead, each trait has a singular role. This finding challenges the validity of conclusions that assume responses at a whole-plant level (De Kroon et al., 2005), and suggests that survival within multi-constrained habitat, relies on low phenotypic integration (Murren, 2002) allowing the expression of multivariate phenotypes. Overall, our study thus underlines the relevance of multi-traits approaches integrating both biotic and abiotic variables, as also stated by Chalmandrier et al. (2022), to efficiently capture plant response strategies, and ultimately identify mechanisms of species coexistence at local scale. Besides, our observations suggest that patterns of trait variations measured in situ should be carefully interpreted as they may not reflect responses to one type of filter only, either biotic or abiotic (Duarte et al., 1986; Kõrs et al., s. d.; Su et al., 2019).
In the ponds of the Iles Kerguelen, each trait category was involved in the response to specific habitat variables. First, traits related to nutrient acquisition and light acquisition as SLA did respond to water nutrients. More precisely, SLA decreased when water nutrients increased, suggesting that macrophyte species may be plastic to either maximize the capture of light when higher nutrient concentrations stimulate phytoplankton growth reducing light penetration in water (Rabalais, 2002), or maximize the capture of nutrients (Lavorel & Garnier, 2002). Root length, also involved in nutrient acquisition, was positively impacted by species abundances. Competition for nutrient acquisition would favor root traits divergence and force plants to optimize resource uptake (Ferguson et al., 2016; Manschadi et al., 2006). This result indicates that root trait variations can be useful in detecting responses to biotic competition, although they are poorly considered in field studies due to the difficulty of sampling (Colom & Baucom, 2020). Secondly, traits related to resources conservation, as LDMC, specific internode, and root masses were negatively impacted by competition. Indeed, resource storage decreased within connection internodes and roots in response to species abundance and/or phytoplankton biomass, perhaps reflecting an escape strategy towards less competitive micro-sites for nutrients. To a lesser extent, LDMC also responded to sediment nutrients, which is consistent with the literature (Fu, 2020; Garnier et al., 2001). Previous studies suggested indeed that an increase in nutrient availability may induce leaf trait variability towards a low conservation of nutrients, as a decrease in LDMC for instance (Poorter & Garnier, 2007). Lastly, internodes elongated with water depth, probably as part of harsh conditions avoidance strategy (Bittebiere & Mony, 2015; Callaghan et al., 1992; Louâpre et al., 2012).
As a result, while many previous studies focused on the effects of one or few variables on species, we argue to consider several biotic and abiotic variables in future studies to better understand the effects of environmental changes on plant communities (Cavieres et al., 2014; Crain et al., 2004; Klanderud et al., 2015; Mitchell et al., 2017; Morris et al., 2020; Wood et al., 2012).

Effects of trait values and their relationships on macrophyte performance

The SEM approach allowed us to highlight that all traits, except LDMC, had either direct or indirect effects on individual performance. This observation is especially important as most published studies neglect the relationships between traits when dealing with their influence on individual performance (Engbersen et al., 2022; Pywell et al., 2003), in particular in aquatic ecosystems (Fu et al., 2014; Ma et al., 2022). Our study thus supports the current view in functional ecology underlying the relevance of multi-traits approaches taking trait relationships into account (Albert, Thuiller, Yoccoz, Douzet, et al., 2010; Bittebiere et al., 2019; Blonder et al., 2014; Gustafsson & Norkko, 2019).
As traits are interrelated, adjustments in trait values must be coordinated to produce a properly functioning organism, also meaning that patterns of trait correlations can greatly change across environments (Murren, 2002). To maintain optimal fitness, it is expected that plant individual should present high correlations among functionally related traits (high integration within functional unit), and lower correlations among functionally different traits (among functional units), conferring flexibility yet coordination to individuals (Nicotra et al., 1997). This is consistent with our results where the highest correlations were found among traits related to resource storage (specific internode mass, specific root mass), or traits related to resource acquisition/conservation strategies (SLA, LDMC).
Comparing the respective effects of traits on individual performance, we observe that clonal traits act more intensely and more directly on performance than expected, suggesting their crucial importance for individual survival. These traits allow to forage for suitable micro-habitats (Suzuki & Stuefer, 1999) and store resources (Dong et al., 2010), which is essential to maintain within constrained habitats. Pooling direct and indirect effects, growth and storage traits (i.e. height, specific internode mass, and specific root mass) have stronger positive effects than architectural traits (root and internode length), which is consistent with the results presented in Wildová et al. (2007). Individual growth and storage traits are directly related to individual size, subsequently affecting individual biomass, whereas architectural traits are more related to colonization ability and allocations between organs, thus affecting more indirectly the individual performance (Wildová et al., 2007). Besides, traits related to resource storage have here stronger effects (amplified by the correlation between specific internode mass and specific root mass) than traits related to resource acquisition (SLA), probably implying a priority in investing in resource conservation within constrained habitats like ponds of the sub-Antarctic region.
Finally, abiotic variables can have cascading effects on individual performance through their influence on trait variations. Indeed, changes in abiotic values from season to season and from year to year, within constrained habitats like polar ponds, would greatly impact plant trait values. For example, an increase in water temperature reduces root length, indirectly and negatively affecting individual performance. Thus, although rising water temperatures can stimulate macrophyte performance within the optimum thermal range (Dar et al., 2014), it may present a negative impact within sub-Antarctic ponds, suggesting that this optimal thermal range can be overpassed (Frenot et al., 2006). Similarly, variations of water depth affect plant performance, through its effects on internode length. As water temperatures keep warming up and precipitations keep decreasing (Walther et al., 2002), one can predict that the consequences for the macrophyte community structure and its productivity will be severe.