Introduction

Understanding how planktonic communities function and how their performance relates to diversity is of increasing interest, as aquatic habitats are experiencing increasing environmental stress of anthropogenic origin. Lakes are particularly sensitive to environmental change, the long-term processes of global warming (Woolway et al., 2022) and salinization (Cunillera-Montcusí et al., 2022; Hintz et al., 2022) pose serious challenges for the communities inhabiting these systems. The situation in shallow lakes is even more alarming, with an anticipated rise in evaporation rates (Woolway et al., 2020) and numerous effects on community structure due to the varying responses of different organism groups to temperature increase (Meerhoff et al., 2012). For phytoplankton, major primary producer of the aquatic food web, climate change mediated impacts on the local environment cause complex effects on its diversity and functioning that necessitates more efficient and more elaborate methodologies (Salmaso & Tolotti, 2021).

In order to achieve a deeper understanding of these effects, the range of analytical methods has been considerably expanded. Recent advancements in environmental DNA (eDNA) techniques led to an unprecedented development in our ability to explore community composition in any type of aquatic ecosystem (Takahashi et al., 2023). The progress in molecular tools is also highly relevant in the case of phytoplankton (Johnson & Martiny, 2015), whose detailed assessment using traditional microscopic analysis is challenging due to the huge number of taxa and great phenotypic plasticity. Moreover, molecular approaches are also more sensitive for detecting rare species, which can disproportionately influence community dynamics (Wilhelm et al., 2014). However, even though there is a globally increasing amount of phylogenetic data on various protist communities in aquatic environments (Burki et al., 2021; Lopes dos Santos et al., 2022), the ecological information contained in these data sets about community functioning has not yet been fully exploited.

The variation of phytoplankton diversity and community structure along environmental gradients has been extensively studied in both marine (Marinov et al., 2010; Righetti et al., 2019) and freshwater environments (Reynolds, 2006). Over the years, it has also become a widely accepted notion that assessing diversity on purely taxonomic terms is less effective in revealing linkages between community composition, ecosystem processes, and changes in the abiotic environment. In this regard, functional trait-based approaches in phytoplankton ecology (Litchman & Klausmeier, 2008) offer a robust framework to promote more meaningful data interpretation and conclusions (Edwards et al., 2013; Pálffy et al., 2013). The diversity of phytoplankton functional traits can be particularly important in maintaining food web and ecosystem stability (Vallina et al., 2017; Ceulemans et al., 2019). The concept of functional diversity has also been adapted to eDNA-based studies in some cases, however, most of these instances rely on Reynolds’ functional classification (Lv et al., 2023; Hanžek et al., 2024). Ramond et al. (2019) studied trait-based diversity of protists in coastal ecosystems using eDNA data, but to our knowledge, similar studies have not yet been conducted on freshwater phytoplankton.

Substantial effort has also been devoted to elucidating links between diversity and ecosystem functioning (Tilman et al., 2014). Several studies on biodiversity-ecosystem functioning relationships rely on the assessment of community-scale resource use efficiency (RUE), commonly expressed as community biomass corrected for the availability of limiting resources (Hodapp et al., 2019). As is the case with many other organism groups, phytoplankton diversity and RUE were found to be interrelated, but the nature of this relationship varies widely among studies, probably due to differences in local factors, community composition, spatiotemporal scale, and analytical approach. While taxonomic richness was found to have a positive influence (Ptacnik et al., 2008; Yang et al., 2022), higher evenness can lead to a decrease in RUE (Filstrup et al., 2014; Xu et al., 2023), suggesting that species composition and species dominance can have contrasting impacts in this respect. These seemingly contradicting observations might be resolved by employing functional trait-based concepts. The dominance of species with specific traits related to nutrient utilization elicits a selection effect, which results in high RUE at low species evenness (Otero et al., 2020), whereas a higher number of functional groups can enhance functioning in the form of increased carbon uptake through complementary resource use (Behl et al., 2011). These findings imply that combining eDNA data with trait-based information could provide deeper insights into the links between changes in phytoplankton diversity and functioning in response to environmental variability.

Soda pans, the shallow saline waters of Central Europe dominated by sodium (Na +) and carbonate ions (CO32– and HCO3–), represent ideal sites to investigate the impact of environmental stress. These water bodies cover a wide range of stress gradients, with salinity, turbidity, alkalinity and nutrient concentrations (Boros et al., 2017) varying both spatially (Horváth et al., 2014) and temporally, following the actual weather conditions (Márton et al., 2023). Salinity and inorganic turbidity in soda pans were found to be major drivers of zooplankton taxonomic richness and functioning (Horváth et al. 2014). We can similarly expect that abiotic stress has a considerable impact on the functional composition of the phytoplankton in these environments, but relevant studies have hitherto mostly dealt with benthic diatom composition (Stenger-Kovács et al., 2014) and pico-sized algae (Somogyi et al., 2017), while the diversity-functioning link has yet remained unexplored. Salinity and turbidity might affect algal diversity in different ways, presumably leading to divergent consequences for community functioning. While higher salinity was found to increase taxonomic diversity and evenness as a result of compensatory growth (Flöder et al., 2010), it can also decrease functional group diversity (Afonina & Tashlykova, 2024). Turbidity can enhance the strength of bottom-up control through increased light limitation (Allende et al., 2009) and, at the same time, dampen top-down control by hindering grazing efficiency (Levine et al., 2005). How these two dominant factors influence various aspects of phytoplankton diversity at the level of functional traits, and how these influences affect RUE through ecological mechanisms such as niche complementarity or functional redundancy is still poorly understood.

In the present work, we intend to analyze an eDNA data set acquired from a network of soda pans to assess the impact of the abiotic environment on the functional diversity of the phytoplankton. The data were previously analyzed in terms of the taxonomic diversity of prokaryotic and eukaryotic microorganism groups (Márton et al., 2023). Here we focus on how the trait-based functional composition of photosynthetic microorganisms changes along the stress gradients of salinity and turbidity (1), what pattern of functional diversity and evenness these changes elicit in the phytoplankton (2) and how this trait-based diversity relates to RUE (3) in order to gain new insights into phytoplankton-related ecosystem functioning in inland saline waters. As the effect mechanisms of salinity and turbidity differ, we expect that changes in these two factors affect different aspects of the niche space occupied by the phytoplankton, indirectly resulting in divergent impacts on RUE.

Materials and methods

Study area, sampling and data collection

The sampling was conducted in April 2017 (3–6) and April 2018 (2–4) in 26 soda pans of the Neusiedlersee-Seewinkel National Park, situated in eastern Austria near the Hungarian border. Salinity in the ponds varies between sub- (0.5–3 g/L) and hyposaline (3–20 g/L) values (Boros et al., 2014). The amount of precipitation differed between the two years: the first quarter of 2017 was drier than the same period in 2018. Measurements of abiotic variables as well as the collection of plankton samples (n = 44) and the procedure of eDNA metabarcoding (including DNA extraction, amplification, and sequencing) were performed as described in detail in Márton et al. (2023). For some of the samples, either the 16S or the 18S rRNA metabarcoding data were of insufficient quality, therefore those samples had to be excluded from the analysis. This step was necessary, since the present study focused on total phytoplankton RUE and diversity patterns, which required both 16S (cyanobacterial) and 18S (algal) data for each sample included in the analysis. The abiotic variables considered in the analysis included total suspended solids (TSS, a proxy for turbidity), total nitrogen (TN) and total phosphorus (TP) concentration, conductivity (a proxy for salinity) and pH (Table 1). Phytoplankton RUE was determined by dividing chlorophyll a concentration (a proxy for phytoplankton biomass) by TN values. While TP is often used for this purpose, we decided to refrain from using it, since a large portion of TP in these lakes is bound to inorganic suspended particles and generally nitrogen is the limiting nutrient in these systems (Boros et al., 2021). As TSS also contained phytoplankton biomass, it was corrected by assuming a 1:100 ratio between chlorophyll a concentration and phytoplankton dry weight, and subtracting this value from TSS to estimate algal-free TSS (V.-Balogh et al., 2009). Any mention in the text about the TSS values of the results refers to corrected, algal-free TSS.

Table 1 Minimum and maximum values of abiotic variables and chlorophyll a, measured during the sampling periods in the 26 investigated soda pans (n = 44)
Full size table

Community and trait matrices

All analyses were based on phylogenetic data described in Márton et al. (2023) and accessible in the NCBI SRA repository within the BioProject accession PRJNA748202. Prokaryotic and microeukaryotic community composition was determined by amplifying the V4 region of the 16S rRNA gene and the V7 region of the 18S rRNA gene, respectively. The V7 region was selected as it offers a suitable balance between fragment length (260–360 bp) and taxonomic resolution. Polymerase chain reactions and amplicon sequencing (Illumina MiSeq platform) were performed by LGC Genomics (Berlin, Germany). Bioinformatic analysis was performed with mothur v1.43.0 (Schloss et al., 2009), using MiSeq SOP (Kozich et al., 2013). Primers were removed from the sequences, singletons were excluded from the dataset. Reads assigned to non-target specific lineages (e.g., Archaea, Chloroplast, Mitochondria, unknown) were removed from the dataset. Operational taxonomic units (OTUs) were picked at 99% similarity threshold levels. To account for variation in sequencing depth, both 16S and 18S rRNA datasets were rarefied to the sample with the lowest number of reads (8620 reads/sample for 16S and 2432 reads/sample for 18S), allowing for standardized comparisons of relative abundances across samples. The ARB-SILVA SSU reference database was used for the alignment of the sequence reads. Taxonomic assignment of 16S rRNA OTUs was performed with TaxAss workflow (Rohwer et al., 2018), 18S rRNA OTUs taxonomic assignment took place using the PR2 reference sequence database. OTU (operational taxonomic unit) matrices were filtered for cyanobacteria from 16S rRNA, and for algae from 18S rRNA sequence data.

Based on expert knowledge and taxonomic classification (John et al., 2021; Guiry & Guiry, 2024), each cyanobacterial and algal OTU was classified in two different ways in order to create an OTU-trait matrix. First, each OTU was sorted into functional groups (Table S1) to assess the relationship between functional composition and environmental drivers using canonical correspondence analysis (CCA, detailed under ‘Statistical analyses’). Thus, prokaryotic OTUs were classified as non-nitrogen-fixing cyanobacteria or nitrogen-fixing cyanobacteria. Eukaryotes were sorted into the following functional groups: non-motile unicellular chlorophytes, non-motile multicellular chlorophytes, planktonic diatoms, autotrophic flagellates, flagellates with phagotrophic potential, other ochrophytes (not diatoms or flagellates), picoeukaryotes and attached (benthic or periphytic) forms. Next, each cyanobacterial and algal OTU was assigned to specific categories within each of the following functional traits: growth form (unicell, filamentous or colony-forming), photosynthetic pigment composition (containing chlorophyll a, chlorophyll a and b, chlorophyll a and c, chlorophyll a and phycobilipigments or chlorophyll a, c and phycobilipigments), flagellated form (presence/absence), mixotrophic potential (presence/absence), pico size (presence/absence) and habitat preference (planktonic or attached (benthic or periphytic)). The selected traits are linked to a wide range of ecological functions. Among the size-related traits, growth form affects zooplankton grazing efficiency (Beardall et al., 2009), and pico-sized algae and cyanobacteria are important components of the microbial food web (Callieri & Stockner, 2002). Variance in pigment composition promotes niche partitioning and coexistence (Burson et al., 2019; Spaak & De Laender, 2021), whereas mixotrophy can confer competitive advantage and alter nutrient fluxes under light and nutrient limitation (Jones, 2000). The inclusion of habitat preference was also necessary, as shallow lakes provide ideal conditions for a meroplanktonic lifestyle, significantly contributing to water column primary production (Schallenberg & Burns, 2004). This trait assignment resulted in an OTU-trait matrix with each row containing the trait categories for a specific OTU (matrices for both 16S and 18S-based OTUs are included as separate supplementary csv files). This OTU-trait matrix paired with the community matrix containing the relative OTU abundances was used to determine functional diversity (detailed under ‘Statistical analyses’). The original 16S and 18S OTU tables, relative OTU abundances, cyanobacterial and algal trait matrices along with the environmental data is publicly available on Figshare (https://doi.org/10.6084/m9.figshare.29128589.v1) and also as supplementary csv files.

Statistical analyses

All analyses were conducted using R version 4.4.0 statistical and programming environment (R Core Team, 2024), the R script containing the full analysis is available in a public GitHub repository (https://github.com/k-palffy/Seewinkel_saline_lakes_analysis). To quantify functional diversity, we used four trait-based indices, representing different attributes of the functional trait space: the number of OTUs with unique trait combinations (UTC; Erős et al., 2009), functional richness (FRic; Villéger et al., 2008), functional dispersion (FDis; Laliberté & Legendre, 2010), and functional evenness (FEve; Villéger et al., 2008). UTC is the functional equivalent of species richness, representing the number of functionally singular species. FRic, FDis and FEve are distance-based metrics, describing specific aspects of the multidimensional trait space determined by the trait-based distances between the species of a community. FRic is the equivalent of the convex hull volume of the space occupied by the species, representing the multidimensional version of trait range. FDis is the average spread of the species in the trait space, calculated as the mean distance of individual species to the centroid of all species in the community. FEve expresses both the evenness of species relative abundances and that of their distribution in trait space, determined from the minimum spanning tree connecting all the species. Interspecific distances were determined on the basis of OTU-specific trait values of the OTU-trait matrix using the Gower index (Gower, 1971; modified by Podani, 1999). All trait-based indices (UTC, FRic, FDis, and FEve) were calculated using the R package ‘FD’ (Laliberté et al., 2014). While FRic is solely a presence-based index, we determined both abundance-weighted and unweighted, presence-based versions of FDis and FEve. The unweighted FDis and FEve was used to describe pure functional dispersion and evenness, undistorted by species dominance. OTU relative abundances for cyanobacteria and algae are based on separate datasets (i.e., 16S and 18S rRNA data, respectively), which were derived from separate bioinformatic processes (Márton et al. 2024) and thus could not be directly merged. Therefore, abundance-weighted FDis and FEve were calculated separately for algae and cyanobacteria, while the unweighted indices were calculated from the combined presence-absence matrix of the two data sets. In order to acquire all three aspects of distance-based diversity for algae, FRic was calculated for both the algal and the combined presence-absence matrix. In addition to trait-based metrics, we also determined Pielou’s evenness (J) (Pielou, 1966) for algal unique trait combinations. Indices specifically for cyanobacteria were eventually not used in the analyses, since some samples (n = 11) contained a single cyanobacterial UTC, for which FRic, FDis and FEve cannot be determined.

The CCA, used to assess the relationship between environmental variables and functional group composition, was performed using the R package ‘vegan’ (Oksanen et al., 2022). Significant environmental variables were identified using backward and forward stepwise model selection using permutation tests (n = 999). Due to the origin of the eDNA data sets, cyanobacterial and algal abundances would have required separate CCAs. However, since cyanobacteria were composed of only two functional groups (nitrogen-fixers and non-nitrogen-fixers), resulting in a single canonical axis, the relationship between cyanobacterial groups and abiotic variables was analyzed using correlation analysis and a scatterplot matrix (Fig. S1.). Environmental variables spanning multiple orders of magnitude (TP, TN, and algal-free TSS) have been log-transformed prior to analysis.

To analyze the association between phytoplankton functional diversity metrics, resource use efficiency and the abiotic environment, we applied three different approaches. First, a canonical correlation analysis (CCorA) was performed using the ‘CCA’ package (González et al., 2008). CCorA is a multidimensional technique to explore relationships between two sets of quantitative variables, which also proved to be an effective method in ecological research (Anderson & Willis, 2003). In our analysis, water depth, algal-free TSS, TN, conductivity and pH were treated as one set of (environmental) variables, while UTC, FRic, FDis, FEve, J and RUE composed the other set of (community-related) variables. CCorA is based on the following computations:

$$U^{1} = a_{1}^{1} X^{1} + a_{2}^{1} X^{2} + \cdots + a_{p}^{1} X^{p}$$
$$V^{1} = b_{1}^{1} Y^{1} + b_{2}^{1} Y^{2} + \cdots + b_{q}^{1} Y^{q}$$

where X and Y are the environmental and community-related variables, respectively, p and q are the number of variables within each set, and U1 and V1 are the first canonical variates (linear combinations of the variables). Parameter vectors a1 and b1 contain values (the canonical weights) that maximize the correlation between the canonical variates:

$$\rho_{1} = cor\left( {U^{1} ,V^{1} } \right)$$

where ρ1 is the first canonical correlation. Further canonical variates and correlations are computed iteratively in a stepwise manner. Successive iterations yield decreasing correlations, we therefore used the first two canonical variates for visualization and analysis. Prior to analysis, each variable was checked for normality using density plots. Variables exhibiting non-normality were Box-Cox transformed, and all variables were subsequently standardized (mean = 0, SD = 1) to ensure comparability and avoid scale effects.

The results of the CCorA were visualized in a canonical loadings plot, which displays the correlations (loadings) of each variable with the first two canonical axes (U1, U2). In the canonical loadings plot, variables are projected inside a circle of radius 1 (the correlation circle, indicating maximum correlation). Those occurring within the ring between the correlation circle and a radius of 0.5 are considered most important in the total amount of correlation between the environmental and community-related variables. The relative positions of the variable vectors reflect their associations in the canonical space: variables pointing in similar directions are positively correlated with each other and with the corresponding canonical axis, while variables pointing in opposite directions are negatively correlated. To relate the variables to the observed samples, we also included the CCorA sample scores projected onto the canonical axes. Sample scores were rescaled by dividing each axis by its maximum absolute value in order to position all samples within the unit circle of the plot. This allows for direct visual comparison of the directionality between samples and variables: samples near the direction of a variable's loading vector have relatively higher values for that variable.

Next, we conducted a pairwise assessment of the variables with Pearson correlation analysis and scatterplot matrices. In the final step, we used the results of the CCorA and the Pearson correlation analysis to select the environmental and community-related variables with the strongest correlation, and performed regression analysis to describe the relationships. The relationships were analyzed using either linear or generalized additive models. In order to deal with heterogeneity, models were also tested with a variance structure, in which case linear relationships were assessed with generalized least squares models, and non-linear relationships were tested with generalized additive mixed models. Best model fits were identified using Akaike’s information criterion (AIC) (Sakamoto et al., 1986). Model fitting was performed using the packages ‘stats’ (R Core Team, 2024), ‘nlme’ (Pinheiro et al., 2023) and ‘mgcv’ (Wood et al., 2016).

The diversity-RUE relationship was also assessed using non-linear quantile regression (Koenker & Park, 1996). This method was chosen as initial analyses showed that RUE varied in a unimodal manner and with high variability along the diversity values. A Gaussian distribution was applied to describe the relationship between the upper limit of RUE and functional diversity metrics, using package ‘quantreg’ (Koenker, 2023), with the following equation:

$$Y = d \times \exp \left[ { - b \times \left( {X - e} \right)^{2} } \right]$$

where Y is the response variable (RUE), X is the independent variable (a functional diversity index), d is the maximum value for Y, e is the value of X at which Y reaches its maximum and b determines the slope at the inflection points of the unimodal curve. Only the diversity indices explaining the most variation in CCorA were selected and tested as independent variables.

Results

The occurrence and relative abundances of functional groups showed similar variability across sites in both sampling periods (Fig. 1). Considering inter-annual differences, cyanobacteria reached generally higher relative abundances in the dryer spring of 2017, with nitrogen-fixers less abundant than non-nitrogen-fixing taxa. The functional groups of eukaryotic algae showed high variance in both years, generally, planktonic diatoms and multicellular chlorophytes occurred with the lowest relative abundances compared to unicellular chlorophytes and most other groups. We found no clear inter-annual differences in the relative abundances of each group except for planktonic diatoms, which were absent from the 2018 samples.

Fig. 1.
Fig. 1.The alternative text for this image may have been generated using AI.
Full size image

Violin plots showing the relative abundance of OTU reads in functional groups of prokaryotic (a) and eukaryotic (b) photoautotrophs in the spring of 2017 (n = 22) and 2018 (n = 22) in the investigated soda pans. The horizontal axes were set to a logarithmic scale to better visualize differences of several orders of magnitudes. Larger black dots represent group means. “Cyanobacteria” refers to non-nitrogen-fixing cyanobacteria.

As for the relationship between functional groups and environmental variables, non-nitrogen-fixing cyanobacterial OTU abundance showed significant positive correlation with conductivity (0.509; P < 0.001) and pH (0.408; P < 0.01), while the abundance of nitrogen-fixing cyanobacteria positively correlated with conductivity (0.307; P < 0.05) and total nitrogen (0.377; P < 0.05) (Fig. S1). Conductivity, which significantly correlated with both groups, showed a correlation of 0.54 (P < 0.001) with total cyanobacterial abundance (Fig. 2a). The CCA analysis revealed that conductivity and algal-free TSS showed a significant correlation with algal functional group composition (Fig. 2b), the first two canonical axes explaining 18.5% of total variance. Among the functional groups, planktonic diatoms and autotrophic flagellates occurred in samples with higher conductivity and lower algal-free TSS, multicellular chlorophytes and picoeukaryotes were more characteristic in environments with high algal-free TSS and moderate conductivity. Mixotrophs tended to occur at low conductivity and moderate algal-free TSS, while attached (benthic and periphytic) forms and non-flagellated, non-diatom ochrophytes were more common at low algal-free TSS and moderate conductivity.

Fig. 2.
Fig. 2.The alternative text for this image may have been generated using AI.
Full size image

a Scatterplot of conductivity versus total cyanobacterial OTU abundance. Axis scales are logarithmic. b Symmetric biplot of canonical correspondence analysis illustrating the relationship between functional groups of algae and environmental variables. Numbers in parentheses in the axis titles represent the amount of total variance explained by the respective canonical axis. Significant and insignificant environmental variables are indicated with dark and light blue color, respectively. cond: conductivity; log_af_TSS: algal-free total suspended solids (log-transformed); log_TN: total nitrogen (log-transformed); log_TP: total phosphorus (log-transformed); Z: water depth; Multi_chloro: non-motile multicellular chlorophytes; Non-mot_Uni_Chloro: non-motile unicellular chlorophytes.

According to the canonical correlation analysis, conductivity, algal-free TSS, and TN showed the strongest associations with the first two canonical axes, whereas among the community-related variables, RUE, FDispr and FEvepr explained the most variation in the canonical relationships (Fig. 3). Correlation between the first canonical variates is 0.84 (P = 0.0017, Table S2), as also illustrated by the scatterplot of the canonical variates for environmental and community-related variables (Fig. S2). While RUE and FEvepr was positively associated with algal-free TSS, FDispr was more related to conductivity and TN. Samples of 2017 showed generally higher values of RUE and FDispr than the samples of 2018.

Fig. 3.
Fig. 3.The alternative text for this image may have been generated using AI.
Full size image

Correlation circle plot of canonical correlation analysis (CCorA) on the group of environmental variables, and phytoplankton diversity metrics and resource use efficiency (RUE) as the group of community-related variables. Dashed circles represent 0.5 and 1.0 correlation circles, respectively. Sample scores (filled points) were rescaled to fit their spread to that of the variable loadings. Cond: conductivity; afTSS: algal-free total suspended solids; TN: total nitrogen; Z: water depth; UTC: number of unique trait combination; FRic: functional richness; FEve: functional evenness; FDis; functional dispersion; J: Pielou’s evenness of algae; subscript ‘pr’: presence-based index; subscript ‘alga’: algal abundance-weighted index (except algal presence-based FRicalga).

The scatterplot matrices and Pearson coefficients indicated similar relationships (Fig S3a–c). RUE and FEvepr had a significant positive correlation with algal-free TSS (0.588, P < 0.001, and 0.348, P < 0.05, respectively), while FDispr positively correlated with TN (0.387, P < 0.01) and conductivity (0.346, P < 0.05) (Fig. S3a). Among the other diversity metrics, only FRicalga showed negative correlations with conductivity (− 0.384, P < 0.05) and pH (− 0.311, P < 0.05). With regard to the regression models tested, generalized least squares models showed the lowest AIC scores (all model formulae and AIC tests are included in the R script available in the associated GitHub repository). Based on the model coefficients, algal-free TSS had a significant positive effect (P < 0.0001), and conductivity had a weak negative effect on RUE (P < 0.05, Table 2). These effects were also apparent in Fig. 4a, where higher conductivity tended to decrease RUE along the algal-free TSS gradient. FEvepr was found to be significantly positively affected by algal-free TSS (P < 0.01, Table 2; Fig. 4b). As for FDispr, TN and conductivity had a positive effect, with an additional significant negative interaction effect between them (P < 0.05, Table 2; Fig. 4c).

Table 2 Summary statistics of generalized least squares models describing the relationships between community-related metrics as dependent variables and environmental factors as independent variables (af-TSS alga-free suspended solids, TN total nitrogen, RUE resource use efficiency, FDispr presence-based functional dispersion, FEvepr presence-based functional evenness, varPower power of variance covariate structure, varFixed fixed weights structure, log base 10 logarithm, SE standard error, SD standard deviation, RMSE root mean squared error)
Full size table
Fig. 4.
Fig. 4.The alternative text for this image may have been generated using AI.
Full size image

Scatterplots of community-related variables as a function of significant environmental variables, inferred from generalized least squares (GLS) models (Table 2). Lines represent GLS model fits. a Resource use efficiency (RUE) versus algal-free total suspended solids, color-graded based on conductivity values. b Presence-based functional evenness (FEvepr) versus algal-free total suspended solids. c Presence-based functional dispersion (FDispr) versus total nitrogen (TN), color-graded based on conductivity values. On figures ‘a’ and ‘c’, GLS regression lines were determined for the median of the conductivity values.

When analyzing FDispr and RUE, we found that the 80% quantile of the unimodal relationship between these two variables was significant (based on non-linear quantile regression; Table 3). Higher values and variability in RUE occurred at intermediate FDispr, declining toward both lower and higher values (Fig. 5). As already shown in the CCorA plot (Fig. 3), communities in 2018 tended to show lower FDispr than in 2017. The same quantile regression fitted on the other influential diversity metric, FEvepr showed no significant relationship with RUE (non-significant estimate for inflection point slope, Table S3). Based on Fig S4, this was caused by the skewed distribution of FEvepr, having most values in the lower half of its range while showing high RUE variance toward higher FEvepr.

Table 3 Summary statistics of non-linear quantile regression coefficients at the 80% quantile, with functional dispersion (FDispr) as the independent variable and resource use efficiency (RUE) as the dependent variable (b slope at the inflection points, d maximum value for RUE, e FDispr value at which RUE reaches its maximum)
Full size table
Fig. 5.
Fig. 5.The alternative text for this image may have been generated using AI.
Full size image

Resource use efficiency as a function of presence-based functional dispersion (FDispr). The area in green, bordered with a dashed line represents the range below the 80% non-linear regression quantile, using a Gaussian distribution.

Discussion

Our study showed that environmental factors can influence different aspects of trait-based functional diversity in distinct ways, and that these influences differ from their effects on resource use efficiency in these saline lakes. While mean functional distances within the communities (FDis) changed along salinity and TN, the distribution of species within the trait space (FEve) was more affected by turbidity. RUE also showed an increasing trend toward higher turbidity but exhibited a unimodal (hump-shaped) relationship with functional diversity. This suggests that linking changes in community structure to community performance and understanding the observed patterns in RUE requires careful consideration. The complexity of these relationships stems not only from the fact that RUE might be differently linked to various aspects of diversity, but also from differing sensitivities of these aspects to various environmental gradients.

Functional composition of the communities was mostly affected by salinity, followed by turbidity (Fig. 2; Fig. S1). Increasing salinity positively influenced cyanobacteria, planktonic diatoms and autotrophic flagellates, whereas mixotroph abundances were reduced. The relative abundance of planktonic diatoms was particularly low, but their occurrence at higher salinity is in agreement with earlier observations in Central European soda pans, identifying conductivity as a major factor affecting total (benthic and planktonic) diatom diversity (Stenger-Kovács et al., 2023). Available data also shows that motile (raphid), stress-tolerant benthic species form the dominant diatom group in these turbid systems (Stenger-Kovács et al., 2014), thus, higher sensitivity to environmental stress and lower silicate uptake efficiency of planktonic diatoms might both contribute to their low abundance. Turbidity favored chlorophytes, picoeukaryotes in particular, which is a well-documented phenomenon in shallow lakes (Somogyi et al., 2017). In terms of traits, it is important to note that turbidity seemed to have an impact on pigment dominance, and also on motility, as flagellates became rare at high algal-free TSS. The most likely impact of inorganic suspended solids on these communities is associated with light attenuation, favoring planktonic chlorophytes under turbid, and benthic or periphytic algae (diatoms and chlorophytes) under clear-water conditions. In general, the distribution of functional groups implies that the main abiotic stressors had different effects on the dominance of functional traits, presumably both leading to a shrinking niche space.

Beyond the level of functional groups, species-specific trait diversity showed more nuanced patterns along the environmental gradients. Turbidity, salinity, and to some extent TN, proved to be crucial factors in this respect, which is similar to what other studies found for other organism groups in these environments (Horváth et al., 2014; Márton et al., 2023). Interestingly, abundance-weighted functional diversity measures of the eukaryotic phytoplankton did not show any significant relationship with abiotic variables. On the other hand, presence-based metrics of the total community (algae and cyanobacteria) responded more strongly to environmental variation. This highlights that eDNA-based assessments of phytoplankton should incorporate both 16S and 18S rRNA data to yield meaningful insights into functional diversity. Presence-based dispersion and evenness showed the highest sensitivity to environmental change, but these two aspects of the niche space were influenced in different ways.

Functional dispersion showed an increasing trend toward higher TN and conductivity (Fig. 4c; Table 2), which indicates that higher trophic status and moderate salinity promoted the occupation of a wider trait space, enabling a variety of functional strategies. The higher end of the conductivity range in the studied lakes represents intermediate salinity, similar to brackish waters. The fact that conductivity had a positive relationship with FDis and a negative relationship with RUE implies that intermediate salinity might affect composition and general physiological responses of the phytoplankton differently. In contrast with extremely saline lakes dominated by few halotolerant species (Padisák & Naselli-Flores, 2021), moderate salinity might provide an environment that permits coexistence of diverse trait strategies, species with various salinity optima and tolerances (Muylaert et al., 2009), thereby enhancing functional dispersion. There can be substantial interspecific differences in the minimum, maximum and optimum values of salinity tolerance, but there is also a certain amount of overlap in the tolerance range of various species (Orizar & Lewandowska, 2025). This overlap implies that environments with intermediate salinity range might be suitable for more species than those of extremely low or extremely high salinity. Such a pattern was observed in the Baltic Sea (Telesh et al., 2011), although ionic composition in marine waters differs from that of soda pans. The marginally negative interaction between TN and conductivity, however, suggests a possible ecological trade-off, limiting trait diversification. This trade-off might be explained by compositional shifts: salinity affected both algal functional group composition and cyanobacterial abundance, whereas TN showed a weak correlation only with nitrogen-fixing cyanobacteria and had no significant effect on the functional composition of algae. The positive relationship between TN and nitrogen-fixers is common, as diazotrophy can significantly contribute to the nitrogen pool (Horváth et al., 2013) and reflects a potentially important biological feedback mechanism. The proliferation of nitrogen-fixing cyanobacteria is also a sign of nitrogen limitation typical of these environments (Boros et al., 2021). Even though nutrient limitation causes competitive exclusion and leads to dominance shifts over time, the observed increase in presence-based diversity implies coexistence and a broader trait pool.

Presence-based functional evenness showed a clearer response, exhibiting an increasing trend toward higher turbidity (Fig. 4b; Table 2). Higher trait-based evenness means a more even coverage of functional space occupied by the species, independent of the amount of functional dispersion in the community. In the studied pond network, low and high turbidity levels promoted the dominance of distinct functional groups, thus, the observed increase in FEve comes from the difference in evenness within those groups, not in the community in general. Light limitation from high amounts of suspended solids can reduce phytoplankton diversity (Varga et al., 2024), which was also evident in our study through the dominance of non-motile chlorophytes. In contrast, size distribution appeared to be less affected by increasing turbidity, with both pico-sized and larger uni- or multicellular taxa remaining similarly dominant under more turbid conditions. As zooplankton grazing can lead to shifts in prey size distribution (e.g., Ger et al., 2019; Kunzmann et al., 2019), we can assume that high turbidity might have impaired top-down control, as found in other studies (e.g., Robinson et al., 2010). Levine et al. (2005) demonstrated that while daphnids were inherently less efficient in grazing small-celled algae and large filamentous cyanobacteria, increased turbidity led to an overall reduction in clearance of phytoplankton, regardless of size. Therefore, a plausible explanation for the observed increase in functional evenness might be that turbid conditions indirectly induced an increase in functional complementarity and a reduction in size-related functional overlap within an otherwise diminishing niche space.

Algal-free total suspended solids, as a measure of turbidity, had a similarly strong positive impact on phytoplankton resource use efficiency (Figs. 3, 4a; Table 2). Just as in the case of FEve, reduced zooplankton grazing is a plausible explanation, as inorganic turbidity can limit ingestion rates (Hart, 1988; Lukić et al., 2020) and act as a strong selective force for grazers (Robinson et al., 2010; York et al., 2011), facilitating the accumulation of phytoplankton biomass. These findings support our hypothesis that high turbidity as a cause and high RUE by phytoplankton with a wide size range as a consequence can imply impaired grazing. Another cause could be increased cellular chlorophyll a content induced by light limitation under turbid conditions. However, determining whether this was a community-wide physiological response or merely a consequence of chlorophyte dominance would require a more detailed study.

In contrast with turbidity, salinity had a marginally negative effect on RUE. A similarly weak, negative salinity-RUE relationship was found along coastal salinity gradients (Olli et al., 2023), where the effect of salinity was mediated by diversity. This finding supports our assumption that environmental impacts on RUE cannot be interpreted without simultaneously taking diversity into account. It is thus highly probable that the decrease in RUE at higher FDis is a direct diversity effect, and salinity is only an indirect driver of RUE. In experimental microalgal communities, a positive correlation between diversity and ecosystem functioning was found to diminish with increasing salinity stress (Steudel et al., 2012), which also supports the ostensibly contrasting conductivity effects on FDis and RUE in our study. A possible trade-off of increasing functional dispersion is that it can hinder community functioning above a threshold, when vastly dissimilar species might have an overall negative effect. Such a phenomenon was found in terrestrial ecosystems (Le Bagousse-Pinguet et al., 2021), but whether this is the direct cause of low phytoplankton RUE at high FDis requires further research. Most trait-based phytoplankton studies rely on easily identifiable morphological, structural or behavioral traits. However, species-specific physiological or metabolic response curves along various combinations of environmental gradients are highly varied and so far studied only for a small number of selected species (e.g., Bestion et al., 2018; Lewington‐Pearce et al., 2019). Hence, even if phytoplankton shows high FDis based on traits that are commonly used in functional diversity research (e.g., size, pigment dominance, motility), species-specific growth curves in that community can be similarly diverse, potentially resulting in suboptimal resource use. Experimental evidence shows that functionally diverse phytoplankton can exhibit underyielding due to the dominance of fast-growing but low-productive species (Schmidtke et al., 2010). This trade-off between growth rate and total biovolume is a possible explanation for the unimodal pattern observed in our study. Low RUE was also associated with low turbidity and a concurrently low functional evenness. Under these conditions, limited resource use is highly probable, because low turbidity facilitates more efficient zooplankton grazing, and low FEve indicates an unbalanced use of niche space. The unimodal variation in RUE with changing FDis might thus occur from a complex interplay of salinity, turbidity and diversity effects, which cannot be explained by classical biodiversity-ecosystem functioning theory.

Our finding that various functional diversity metrics showed different patterns along the main abiotic factors carries a substantive message. Despite the general notion that community composition has a strong influence over ecosystem functioning (D’Alelio et al., 2016; Amorim & Moura, 2021), field observations necessitate sophisticated consideration of local conditions, because the putative direct link between diversity and functioning is suppressed by a multitude of impacts arising from environmental heterogeneity. This is in close agreement with the notion that the drivers of environmental change need to be thoroughly examined in biodiversity-ecosystem functioning research to better predict the response of ecosystems to shifts in diversity (Laender et al., 2016). Moreover, experimental evidence for positive biodiversity effects is based on initial biodiversity, which obviously differs from the realized diversity of natural environments (Hagan et al., 2021). Another crucial aspect is seasonality and phytoplankton dynamics, which can further complicate interpretation. Spring phytoplankton blooms are relatively common in soda pans across the broader region (Boros et al., 2016; Szabó et al., 2020), but whether our sampling dates coincided with such events cannot be confirmed due to the lack of seasonal data for the investigated lakes. A better understanding of the link between diversity and functioning requires extending the spatiotemporal scope and placing more emphasis on the role of the regional species pool. This can be of particular importance in phytoplankton communities with rapid changes in composition and with a direct effect of species pool size on seasonal dynamics (Pálffy & Smeti, 2024). Furthermore, the effect of top-down control is often neglected in phytoplankton field studies, future studies should therefore also investigate the role of zooplankton in mediating the diversity-functioning relationship in saline environments.

In general, eDNA metabarcoding studies on phototrophic microorganisms still have certain limitations compared to traditional microscopic analysis. Although metagenomic approaches can provide greater detail in composition, especially in communities characterized by a large fraction of pico-sized taxa, they are still prone to yielding lower taxonomic resolution. Comparative studies imply that while metabarcoding is more efficient in detecting rare, less well-known genera, the identification of some common taxa is presently not always feasible on the species level (Akcaalan et al., 2023). This limits trait-based analyses, which might explain the lack of significant progress in this regard. Despite these shortcomings, based on our results we can conclude that eDNA can be a useful tool to detect spatiotemporal patterns of functional diversity in phytoplankton, especially in highly turbid environments, where microscopic identification becomes complicated. The ongoing debate on how diversity contributes to ecosystem functioning might benefit from the trait-based interpretation of metabarcoding data, and this still unexplored potential could direct metagenomics research on phytoplankton from questions dealing with “what is where” toward shedding more light on how environmental factors drive community composition and performance.