Theories explaining community assembly assume that biotic and abiotic filters sort species into communities based on the values of their traits and are thus based on between-species trait variability (BTV). Nevertheless, these filters act on individuals rather than on species. Consequently, the selection is also influenced by intraspecific trait variability (ITV) and its drivers. These drivers may be abiotic (e.g., water availability) or biotic (e.g., competition). Although closely related species should have similar traits, many of them coexist.

We investigated the relative magnitudes of BTV and ITV in coexisting closely related species and how their individual traits differ under different drivers of ITV.

We manipulated conditions in a greenhouse pot experiment with four common Carex species, where individuals of each species originated from four source localities. Individuals were grown in factorial combinations of two moisture levels, with and without a competitor (grass species Holcus lanatus, a frequent competitor). We analysed the variability of six morphological traits on individuals in the greenhouse and three morphological traits in the source localities.

Species identity was the main determinant of differences in most traits. Competition exerted a greater effect than water availability. For leaf dry matter content (LDMC) and vegetative height, competition’s effect even exceeded the variability among species. On the contrary, for specific leaf area (SLA) and clonal spread, the interspecific differences exceeded ITV induced by experimental treatments. SLA measured in the greenhouse closely correlated with values measured in field populations, while LDMC did not. The variability caused by source locality of ramets in the greenhouse was small, although sometimes significant.

Closely related species differ in their traits, but for some traits, ITV can exceed BTV. We can expect that ITV can modify the processes of community assembly, particularly among co-existing closely related species.

Species and localities

For our greenhouse experiment, we chose four common Carex species (Table S1 in Appendix) able to co-occur although they differ especially in preference of soil moisture (Ellenberg-type indicator value for moisture is 4 for C. caryophyllea, 5 for C. pilulifera, 6 for C. pallescens and 8 for C. panicea; Chytrý et al., 2018). These species are all perennial herbs, hemicryptophytes, commonly distributed in the South Bohemia region, where they were recorded in all 12 × 11 km quadrats of grid mapping and in the majority of their 6 × 5.5 km sub-quadrats (Pladias – database of Czech flora and vegetation, www.pladias.cz, access June 13, 2024; Wild et al., 2019). These species are able to co-occur, although they also often occur separately, especially in the case of C. caryophyllea which prefers drier habitats. Although these Carex species frequently co-occur in relatively small plots (2x2 m, personal observation), they seldom achieve high cover, and the main competitor is commonly some dominant grass. The average cover of our Carex species in phytocenological relevés from Pladias – the database of Czech flora and vegetation (only relevés where the species is present considered) is relatively low (3.6% C. caryophyllea, 7.9% C. panicea, 2.4% C. pallescens, 2.5% C. pilulifera) with a maximum recorded cover of 38% (i.e., degree 3 of the Braun-Blanquet scale) for all species but C. panicea, which can become dominant (but only rarely and in special types of peatland meadows) (www.pladias.cz, access June 13, 2024; Chytrý et al., 2021). We used populations from four source localities (size 1-3 ha) of the South Bohemia region for each species. One locality contained all four species in mixtures, demonstrating they can coexist; together we used individuals from seven localities (Table S1 in Appendix). All localities were semi-natural meadows with regular mowing management, mostly once per year with the exception of one locality, which was not managed. From each population, we took 20 individuals in the form of small ramets (young vegetative rosettes, average fresh weight = 0.58 ± 0.43 g (mean ± SD), average height = 5.05 ± 2.82 cm) at the beginning of April 2020 for the greenhouse experiment.

Design of greenhouse experiment

Each individual was taken from the field to the greenhouse and left in water to develop roots for one week. After that, we planted each individual in its own pot (upper diameter 16 cm, lower diameter 10 cm, height 15 cm, volume 2 l) with peat substrate and sand mixture in the ratio of 1:2. From each population, 20 ramets were grown placing them randomly in five blocks in factorial combinations of two moisture levels (water treatment), and with and without competition of Holcus lanatus (competition treatment) (Figure 1a). Thus, we had four experimental treatment type combinations: CH – competition and high water treatment, CL – competition and low water treatment, NH – no-competition and high water treatment, NL – no-competition and low water treatment. In all our source localities (and similar localities in the area) sedges grew in the matrix of grasses, which dominate the meadows and thus are their main competitors. Holcus lanatus (Ellenberg-type indicator value for moisture is 6; Chytrý et al., 2018) is their typical representative. It is a clonal hemicryptophyte, typically with a height similar to Carex species from our experiment. It is one of the most frequently co-occurring species with Carex in our source localities; it is also suitable for experiments, as its germination rate is quite high and it establishes quickly (Tammaru et al., 2021). We used seeds from a commercial supplier (Planta Naturalis). For the competition treatment, ca. 45 seeds of Holcus lanatus were sown in each pot together with the Carex ramet at the time of ramet planting. After two weeks, Holcus lanatus was thinned to 15 individuals per pot. One week after thinning, when Holcus lanatus was established, the water treatment was started. While the high water treatment consisted of keeping the saucer below the pot (volume of 380 ml) permanently filled with water, in the low water treatment meant saucers were only filled with water once the substrate was dried out. Because we could not measure the soil water content directly in every pot, we checked the effect of our water availability treatment on Carex biomass aboveground (18% decrease in low water treatment, F_1,312=7.43, p=0.007), belowground (24% decrease in low water treatment, F_1,309=19.76, p<0.001) and in total (22% decrease in low water treatment, F_1,307=14.95, p<0.001) to prove that the low water treatment was sufficient to impose water stress. Because we used nutrient poor substrate (peat with sand), with fertility insufficient for H. lanatus, we used 100 ml of fertilizer (15 ml of Vitality Komplex Agro on 1 l of water) on each pot two times. All plants were harvested from July 29 to August 1, 2020. They were washed free of soil, separated into aboveground (leaves, stems, and flowers) and belowground (roots and rhizomes) parts, and processed for measuring functional traits as described below.

Measured functional traits and other variables

In the greenhouse experiment, we measured six functional traits characterizing different functions of Carex plants (Figure 1b). Vegetative Height was measured as the highest leaf apparently forming a rosette without stretching. To calculate SLA and LDMC, we took the fresh and dry weights of two leaves per individual and measured their area in ImageJ 1.x (Schneider et al., 2012). The number of ramets created by the mother plant in each pot (Clonal Spread further on) determined the ability of clonal multiplication. Belowground and aboveground biomass (with total biomass the sum of above- and belowground biomass) from each pot was oven-dried at 60°C for 48 hours and subsequently weighed. From these values, we calculated Root/Shoot Ratio (i.e., dry biomass of only roots divided by dry aboveground biomass) and Aboveground/Belowground Biomass (i.e., dry aboveground biomass divided by dry belowground biomass of both roots and rhizomes) to differentiate belowground biomass as only roots, or including rhizomes, and to compare them with the aboveground biomass.

To compare functional traits measured on plants grown under specific greenhouse conditions and experimentally influenced by water and competition treatments with traits measured in the source localities, we also measured three functional traits (Vegetative Height, SLA and LDMC) of 10 individuals per population sampled from the field (i.e., field study) at the same time the greenhouse experiment was harvested, selected in the field according to Pérez-Harguindeguy et al. (2016).

Data analysis

Inter- and Intraspecific trait variability. We used Linear Mixed-Effects Models (LMM) with F tests of type III ANOVA using R version 4.4.0, package “lmerTest” (Kuznetsova et al., 2020) to explain measured values of individual traits in the greenhouse experiment based on species identity (Species), source locality (Locality), experimental treatment (Competition and Water) and all their interactions. These factors were taken as fixed with experimental block (Block) as random effect because Block had significant effect on three out of six tested traits (Table S2 in Appendix). For this analysis, after a check of the data normality, we used Log-transformed data for SLA, Vegetative Height, Root/Shoot Ratio and Aboveground/Belowground Biomass and Square Root-transformed data for Clonal Spread. The same test was used for aboveground, belowground and total biomass of Carex (all values Log-transformed) to check the effectivity of the water treatment. To examine BTV and ITV for plants from the field study, we used values measured in the field of Vegetative Height (Log-transformed), SLA, and LDMC in linear models (LM) with an F test of ANOVA, where we tested the effect of Species, Locality, and their interaction. For graphical representation (Figure 2), we calculated critical values for F (for individual factors both for greenhouse experiment and field study) reflecting appropriate degrees of freedom of each factor and α = 0.05.

            Because we wanted to compare the explained variability of the greenhouse experiment with that of the field study represented by Sum of Squares, we reanalysed the LMM for the greenhouse experiment without the random factor Block and using a LM with F test of ANOVA, as in the case of the field study. The effect of Block, which was not included in this new analysis, was thus included in residuals of the model. Explained variability was composed by the Sum of Squares of Species, Locality, Experiment (i.e., Competition, Water and their interaction; only for greenhouse experiment), Interactions (i.e., all interactions with Species and Locality; because the number of tested interactions was high, we summed up the Sum of Squares of all interactions with Species and Locality together for clarity) and Residuals (i.e., variability explained by other factors than were tested in the model, including the effect of Block for the greenhouse experiment). While the variability among Species expressed BTV, ITV included variability of Locality, Experiment, and Interactions. The variability of individuals from the same population (the same species from the same locality) under the same treatment combination is reflected by the Residuals (including the unexplained ITV, as well as sampling and measurement errors).

Although one of our aims was to compare BTW and ITV, we should be aware that this comparison is context dependent and also depends on the design of the experimental setup. In particular, the explained variability depends on the degrees of freedom (generally, the explained variability increases with the degrees of freedom used). Nevertheless, the explained variability (expressed using Sum of Squares) is additive, which is a great advantage. Also, because the same design and data analysis model was used for all the traits, it is ideal for comparison of variability explained by individual factors in various traits. For direct comparison of effects of individual factors (i.e., species identity, individual experimental factors, source locality), the simple F-value might be more useful as a measure of the effect size. It is the ratio of variability explained in given experimental settings by an individual factor to its expected value under the null hypothesis that it does not have any effect. The disadvantage of F-value is that it is not additive. 

Correlations of traits between greenhouse experiment and field study. To assess how well the functional trait differences among and within species in nature were reflected in the values of their traits in the greenhouse experiment, we calculated the average values of functional traits from the field and greenhouse for each population and four experimental treatment type combinations (CH, CL, NH, NL). Subsequently, we compared the values from the field with values from the greenhouse using Major axis regression (MA) in R-package “lmodel2” (Legendre, 2018). We also calculated between group (i.e., between-species) and within group (i.e., within-species between localities) Pearson correlations between trait values from the field and greenhouse using function “statsBy” from the R-package “psych” (Revelle, 2024), where the grouping variable was Carex species. Nevertheless, especially in the case of between group correlations, the statistical test was not very relevant because it was calculated only from four values (i.e., four Carex species) which cannot give significant values for such a dataset. Thus, it was rather used to identify trends without testing for statistical significance.

            For all figures, we used R-package ‘ggplot2’ (Wickham, 2016) and represented all data as non-transformed for easier visual interpretation.