Identifying environmental drivers of benthic diatom diversity: the case of Mediterranean mountain ponds

Introduction

Ponds are one of the most striking features in high elevation landscapes. Although these aquatic ecosystems represent a small area in mountainous regions, they are a key component in safeguarding aquatic biodiversity (Biggs et al., 2005; Martínez Sanz et al., 2012; Takaoka, 2015). Several studies have examined environmental factors influencing the biodiversity of high-elevation temperate lakes and ponds (Hinden et al., 2005; Catalan et al., 2006; Füreder et al., 2006; Robinson & Oertli, 2009; Martínez Sanz et al., 2012; Ilg & Oertli, 2014), but little attention has been paid to Mediterranean mountain systems (Escrivà, Armengol & Mezquita, 2010). In the Mediterranean region, these habitats are extremely sensitive to environmental stressors due to their long history of human settlement and impacts (which has led to significant alterations in ecological status), hence they play an important role for the assessment of the impact of environmental changes at a local level (Rosset, Lehmann & Oertli, 2010). Among the aquatic communities inhabiting these systems, diatoms are widely used to detect such alterations. In particular, periphytic diatoms constitute one of the most important components of algal assemblages in aquatic habitats in terms of diversity, biomass and ecosystem metabolism. However, the understanding of how diatom communities respond to environmental changes is still limited (Teittinen et al., 2016), and accordingly, the recent literature has a focus on identifying factors that control the diversity of diatoms in freshwater ecosystems (Marazzi et al., 2017).

Mountain ponds exhibit a high microbial phylogenetic singularity, constituting reservoirs of great evolutionary potential (Casamayor, 2017). These systems are now considered as biodiversity hotspots (Zaharescu et al., 2016). There are two scientific research lines in this regard; one line states that diatoms are ubiquitous and the composition of the community is determined by environmental conditions (Finlay & Fenchel, 1999; Finlay, 2002). Teittinen & Soininen (2015) found that dispersal limitations does not appear to influence diatom diversity at narrow scales, as species simply occur where their habitat requirements are met. The second line suggests that, on a larger scale, freshwater diatoms show biogeographic patterns that are related to dispersal processes, climate, and evolutionary history (Hillebr & Blenckner, 2002; Vanormelingen, Verleyen & Vyverman, 2008). According to Passy (2010), large-scale biodiversity patterns in freshwater protists are mostly resource-driven, compared to more complex organisms that are more strongly influenced by climate. These contrasting results demonstrate that different factors may affect species diversity at different spatial scales (Bolgovics et al., 2016); and hence, that climate operates at regional spatial scales while nutrients, pH and conductivity, among others, are factors that operate at the local scale (Teittinen & Soininen, 2015). In addition, it is commonly recognised that in regions with highly connected waterbodies, lentic diatom communities exhibit high diversity values (Vyverman et al., 2007; Soininen et al., 2009). This second line has been proposed as the best hypothesis with which to explain changes in diatom richness along the elevational gradient (Wang et al., 2011).

The exploration of elevational trends in diversity is essential for determining broad-scale distribution patterns (Soininen, 2012). In this sense, it becomes critical to understand the relative contribution of different environmental drivers impacting on taxa diversity, and how they connect to different scale gradients (Zaharescu et al., 2016).

Within this framework, Soininen & Teittinen (2019) have recently summarized the main open questions concerning the spatial ecology of diatoms. This paper answers specifically their question 1 dealing with the key factors explaining diversity patterns in diatoms. Particularly, the present study addresses the following questions: (i) Which is the main abiotic predictor of diatom diversity in mountain ponds? (ii) Are there dispersal limitations for diatom taxa in these ecosystems? (iii) Do rare taxa tend to occur in more isolated systems?

Materials and Methods

During spring 2017 a biological survey was conducted (Junta de Andalucía approval no. SGYB/AF) in 45 mountain ponds, both temporary and permanent (median elevation: 1203 m a.s.l., range: 225–2520 m a.s.l.) located in Andalusia (south of Spain, Fig. 1). We consider temporary ponds those that permanently dried every year, whereas permanent ponds are those that do not dry for at least a decade (Chase, 2003). Our recent results (Rodríguez Alcalá et al., 2019) show that changes in diatom composition are mostly driven by conductivity and water depth; therefore, turbidity and conductivity were the unique environmental variables that were measured in situ since they are considered as excellent proxies of ecological status in continental lakes (Stenger-Kovács et al., 2018).

Periphytic diatoms were collected from each system and samples were processed and analyzed in the laboratory by following the method explained in Blanco & Becares (2006). Diatoms were cleaned and mounted on permanent microscope slides according to the European protocol EN14407. Due to the scarcity of diatoms in certain samples, a homogeneous count of ca. 100 individuals (diatom valves) were identified on each slide at the lowest possible taxonomic level following Hofmann, Werum & Lange-Bertalot (2011) and references therein. Counting 100 individuals has been found to be representative in terms of statistical reliability in ecological studies (Bate & Newall, 1998; Fatela & Taborda, 2002; Razumovskii, 2004). Uniqueness of diatom communities(U) of each pond i was calculated as: ${\displaystyle U_i=\sum _j{A}_j\left(1-O_j\right)}$

where A_j and O_j are the abundance (relative) and the occurrence (rescaled to the [0–1] range), respectively, of the jth taxon. U values vary between 0 (all of the species of the sample are also found in all of the other ponds) and 100 (none of the taxa that were found occur in any other sample). Diatom diversity was calculated using Shannon–Wiener index: ${\displaystyle H=-\frac{\sum n_i}{n}ln\left(\frac{n_i}{n}\right)}$

where n_i/n is the proportion of the ith taxon in the sample. Spatial clustering was tested with Nearest Neighbours analysis, whereas spatial autocorrelation was tested using Moran’s statistic. Finally, the degree of isolation (I) of each pond i was calculated with a Gaussian kernel density estimator (Hammer, Harper & Ryan, 2001): ${\displaystyle I_i=1-\left(\pi r^{-2}\sum _i{10}^{{\left(-d_i\right)}^{2}2r^{-2}}\right)}$

where d_i is the Euclidean distance between pairs of locations, and r is the scale radius of the kernel, here set to 0.1.

To analyse the response of the diatom diversity and the uniqueness of diatoms with respect to abiotic factors, a Generalized Linear Model (GLM) ANCOVA model was calculated using hydroperiod and substrate type as categorical factors, and the other variables (limnological and geographical data) as continuous predictors. The resulting variables were selected based on a ‘best subsets’ algorithm (Neter et al., 1996). To compare the performance of the different models that were generated, we used Akaike’s Information Criterion (AIC).

Finally, distance matrices were computed using appropriate metrics (Sorensen’s index for floristic data, geographic distance for geographic data and Euclidean distance for elevational and limnological data) and distance decay of similarity (DDS) was evaluated by means of Partial Mantel tests (Legendre & Legendre, 1998) which consist on Pearson’s correlation tests between pairs of matrices based on random permutations.

Computations were performed by using Statistica (Statistica 13.0, StatSoft Inc., Tulsa, OK, USA) and PAST 3.19 (Hammer, Harper & Ryan, 2001) softwares. Raw data are available at DOI 10.6084/m9.figshare.11356262.

Results

Most of the studied ponds (71% of the systems) were temporary, most of them (58%) presented limestone substrates, 22% presented Trias substrata and 20% presented siliceous substrata (Table 1). Pond waters were characterised by a wide range of mineralisation levels (0.03–10.40 mScm⁻¹) and by moderate turbidity values (median: 11.32 NTU, range: 1.26–1000.00 NTU).

A total of 238 diatom species (including subspecific taxa) were found, belonging to 64 different genera. Diatom assemblages were diverse (median Hα: 1.90, range: 0.16–2.84), and 25 species reached abundances>1% over the whole dataset. Although there was a statistically significant clustering in sampling locations (nearest neighbour analysis, R = 0.72, p < 0.001), no spatial autocorrelation was detected for diatom diversity along the distance gradient (Moran’s statistic, p > 0.05). Maximal richness was observed in Charco del Cornillejo, where 31 species were identified in a 100 valve count (Table 1). Dominant taxa included alkaliphilous adnate species such as Achnanthidium pyrenaicum or Amphora pediculus, and in general oligo-mesotrophilous taxa (Achnanthidium minutissimum, Nitzschia perminuta, Fragilaria pararumpens) were relatively abundant and widespread in the analyzed systems. Based on the low conductivity values that were observed, most of the ponds studied could be characterized as fresh to subsaline ecosystems with respect to the diatom communities they contain.

Average occurrence was 6.4%, with 47% of taxa only appearing in a single pond. On the contrary, the ubiquitous Achnanthidium minutissimum was present in 56% of the samples. Three unknown diatom species were found in these ponds, recently described by us as new taxa, e.g., Blanco et al. (2019). Maximal uniqueness was observed in Charca de los Llanos de Líbar, where 97% of the individuals that were identified belong to taxa not found elsewhere, with the occurrence of A. jackii having an abundance of 84%.

GLM models

The obtained results showed the lowest AIC values for GLM models with a Gaussian error distribution and an identity link function. Three variables were selected as the best predictors for diatom diversity: elevation (p = 0.02), uniqueness (p = 0.03), and hydroperiod (p = 0.09), with no significant interactions. Figure 2 describes the relationships between these variables: diversity increases monotonically with elevation, and with uniqueness but only in temporary systems, where diversity values were higher on average. Isolation had no apparent relationship with diversity (Fig. 3).

Floristic uniqueness depended significantly on water conductivity (p = 0.01), isolation (p = 0.03) and diversity (p = 0.01). An increase in floristic uniqueness with increasing isolation (Fig. 4) and conductivity (data available at DOI 10.6084/m9.figshare.11356262) was only evident for temporal ponds. In general, variable relationships are monotonic in temporal systems, whereas unimodal patterns tend to occur in permanent systems (Fig. 2).

Discussion

Diatom diversity predictors in Andalusian mountain ponds

The results obtained in this study show that the most diverse diatom communities appear in high elevation ponds. The relationship between diatom diversity metrics and elevation has been previously analysed in the literature, but only a few publications report a clear and significant trend (Teittinen et al., 2016); only Wang et al. (2011) observed a monotonic decrease of diatom diversity with elevation in stony streams. The lack of general trends in this pattern suggests that diatom diversity could be more affected by local environmental factors rather than by climatic variables that are associated with elevation (Hillebrand & Blenckner, 2002). The presence of a dominant, constraining environmental variable in the dataset may also obscure this relationship (Teittinen et al., 2016). However, in our case, neither conductivity nor turbidity were observed to be significantly affecting diatom diversity. This contrasts with the literature consensus (Soininen & Heino, 2007; Teittinen & Soininen, 2015; Jyrkänkallio-Mikkola, Heino & Soininen, 2016) reporting on how different environmental variables affect this metric. At a regional scale, the effect of other variables on diatom diversity, such as pH or nutrients, has been found to be very weak or nonlinear (Wang et al., 2017). This is probably because historical factors explain significantly more regarding the observed diversity patterns than do the contemporary environmental conditions (Vyverman et al., 2007; Vilmi, Karjalainen & Heino, 2017). Therefore, although in high elevation areas the prediction is related to an increase of diatom richness as a consequence of the global change (Rosset, Lehmann & Oertli, 2010), human activities tend to yield water quality impairment that could be masking this pattern.

Table 2:

Partial Mantel tests results comparing the floristic distance matrix (Sorensen index) with the other distance matrices calculated.

Compared matrix	When controlling for similarities given in:	R	p value
Elevational	Environmental	0.03	<0.01
Elevational	Geographical	0.07	0.80
Environmental	Elevational	0.08	<0.01
Environmental	Geographical	0.10	0.31
Geographical	Elevational	0.07	<0.01
Geographical	Environmental	0.07	<0.05

DOI: 10.7717/peerj.8825/table-2

Conclusions

Although they are often excluded in conservation studies, microorganisms (e.g., periphytic diatoms) arise as essential components of biodiversity in mountain regions (Casamayor, 2017). We found that diatom diversity in mountain ponds essentially responds to geographic factors that mask any plausible contribution of environmental or elevational gradients. The link that is observed between isolation, uniqueness and diversity indicates that isolated ponds are clear targets if floristic singularities are a conservation goal (Vilmi, Karjalainen & Heino, 2017). In conclusion, periphytic diatom communities in the studied mountain ponds are likely to be driven by regional factors, and therefore are mostly shaped through dispersal limitation. Further studies with a focus on exploring the drivers of diatom diversity and uniqueness at smaller temporal and spatial scales are required.