Diatom Biodiversity in Karst Springs of Mediterranean Geographic Areas with Contrasting Characteristics: Islands vs Mainland

: Karst ecosystems are considered as priority environments for the protection of biodiversity on a global scale. This study provides a ﬁrst comparative analysis of epilithic diatom ﬂora from karst springs in two Mediterranean geographic areas (Spain and Italy) with contrasting characteristics (islands vs mainland). We investigated twenty-three springs with di ﬀ erent anthropogenic impact levels once in the winter season between 2007 and 2017 (N = 23). A total of 176 diatom taxa (56 genera) were found of which 101 (44 genera) were observed in single sites. A general good biotic integrity was revealed by structural indices (species richness, diversity and evenness). However, crenophilous species were generally present and abundant in less impacted springs. Comparing islands and mainland, signiﬁcant di ﬀ erences were found in species composition and diversity (H’) based on multivariate analyses (global R = 0.610; p = 0.001) and t -test ( t = 2.304; p = 0.031). Discharge and Cl − were the most signiﬁcant variables in determining diatom assemblages. Our results conﬁrm the role of springs as multiple ecotones and refuges for rare species and suggest that the geographic insularity may be an important factor in maintaining diatom biodiversity.


Introduction
Karst springs are freshwater environments of great ecological value and strategic water resources in the Mediterranean area since they provide valuable ecosystem services, among which habitat for high biodiversity and drinking water supply [1]. In this geographic area, due to the strong climatic seasonality and lower water availability during dry periods (summer-autumn), several large springs have been captured since ancient times and they are almost the exclusive drinking water source for many urban centres [2,3].
Karst ecosystems have been indicated as priority environments for the protection of biodiversity on a global scale [4]. Unlike other freshwater ecosystems, several springs are still relatively natural systems. This is typically due to artificial (concrete) water collection basins that occupy part of the spring-head.
Finally, in Friuli, the six springs studied cover an altitudinal range of 454-1249 m a.s.l and are perennial and typical rheocrenic systems. The springs BM and GD are included in the list of geosites of regional and national interest, respectively [34]. The site GD is classified as "Alpine Cave" (habitat code SC1) [35]. The springs RR is captured for drinking purposes.   The three springs studied in Majorca Island are located in the Tramuntana mountain range, in the northern part of the island, and cover an altitudinal range of 56-471 m a.s.l. The Tramuntana Mountains extend parallel to the north-western coast and present numerous springs with different Water 2019, 11, 2602 4 of 21 sizes and discharge [33]. The springs studied in this area are perennial and rheocrenic systems, and some of them are captured for drinking purposes.
In Sardinia, the eight springs studied are located in different areas of the island and distributed over an elevational range of 55-780 m a.s.l. They are mostly perennial and rheocrenic systems. Only two springs, SG and VA are classified as limno-rheocrenic and seasonal, respectively. The springs SG and DN are "Natural Monuments" according to Regional Law 31/1998 that defines the areas of noteworthy naturalistic value for conservation and protection in Sardinia. Five springs are located in Sites of Community Importance (SCI) for the Mediterranean biogeographic region: UL and FO (ITB 021107 "Monte Albo") SG (ITB022212 "Supramonte di Oliena, Orgosolo e Urzulei-Su Sercone"), VA (ITB021156 "Monte Gonare") and DN (ITB041111 "Monte Linas-Marganai"). Most of these springs are captured for drinking purposes.
In the Veneto Region, almost all springs (five out of six) are located in the south-eastern Alpine foothills, in a geographic area called Lessinia. One spring (PR) is located on the Monte Baldo mountain range (south-eastern Alps). They cover an elevation range of 183-675 m a.s.l. These springs are perennial systems and are virtually all captured (tapped) to some extent. Three springs (BR, BA, and TO) are classified as rheo-limnocrenic, and three springs (FC, LF and PR) as limno-rheocrenic systems. This is typically due to artificial (concrete) water collection basins that occupy part of the spring-head.
Finally, in Friuli, the six springs studied cover an altitudinal range of 454-1249 m a.s.l and are perennial and typical rheocrenic systems. The springs BM and GD are included in the list of geosites of regional and national interest, respectively [34]. The site GD is classified as "Alpine Cave" (habitat code SC1) [35]. The springs RR is captured for drinking purposes.

Sampling
Sampling was carried out once in the winter season between 2007 and 2017 (N = 23). The geographic position and altitude of the springs were recorded by a GPS. Water temperature, pH and conductivity were measured in situ using digital multi-parametric probes [24,26,32].
Anthropogenic disturbances were evaluated in an ordinal scale with disturbance scores: 1 = low; 2 = medium; 3 = high [24]. Shading conditions and current velocity were assessed using a five-score scale [36]. In most springs, discharge was measured by the volumetric method using a graduated bucket and a chronometer. Repeated measures were made close to the diatom sampling points. Only in some large springs (TR, BM and AL), discharge was obtained by estimates based on current velocity/area ratio. The discharge measured during sampling was reported using a seven-range scale (<0.5; 0.5-1; 1-5; 5-10; 10-50; >50; >100 L sec -1 ).
Diatom and water samples for physical and chemical analyses were collected simultaneously. Diatoms were collected with a toothbrush from hard natural substrata (five stones) along a longitudinal transect from the emergence point up to a max distance of 5 m following the European Standard [37]. All samples were preserved with a formaldehyde solution (4% v/v) immediately after sampling. Some typically selected sampling points in each geographic region are reported in Figure 2.
Water samples were collected at the emergence point using 1-L polypropylene and polyethylene bottles and were transported in cold and dark conditions for the laboratory analyses. Further details for the methods used are given in previous studies [24,26,32].

Data Processing and Statistical Analyses
The structure of diatom assemblages was examined by species richness (R), Shannon-Wiener diversity index (H') [52] and Pielou's evenness index (J') [53], calculated using OMNIDIA 6.0 software [54]. Differences in structural indices between islands and the mainland were evaluated by a paired t-test [55,56] and after verification of normal distribution with the Shapiro-Wilktest [57] using R 3.6.1 [58]. For this test probability (p) <0.05 was considered significant.
All diatom data were converted in relative abundances (RA) for the statistical analyses. To minimize the influence of rare species, only the species with RA ≥5% in at least two sites are included in the analyses.
Ecological preferences of the most abundant diatom species (RA ≥5%) for pH, trophic state, salinity and moisture were attributed according to Van Dam et al. [59]. The taxa with a wider ecological range were equally placed among the respective autecological levels to obtain a simplified framework.
The similarity among diatom assemblages of the four geographic regions was analysed using a non-metric multidimensional scaling ordination (nMDS) [60]. The ordination was performed on a Bray-Curtis similarity matrix of species data log (x+1) transformed [61]. The significance of the differences was tested by a one-way analysis of similarities (ANOSIM) [62]. For this analysis, p <0.05 was considered significant. The ANOSIM pairwise test was also performed for each pair of geographic regions. nMDS and ANOSIM were performed using the software PRIMER 6 [63].
Relationships between diatom species and environmental variables were explored by a Canonical Correspondence Analysis (CCA) [64] after the previous assessment of the length of the gradient (>4) by means of a Detrended Correspondence Analysis (DCA) [65] of diatom data using Canoco 4.5 [66]. The two matrices, for physical and chemical data (except for pH), and for diatom data, were log (x+1) transformed. All canonical axes were used to assess the significant variables through analyses by means of a Monte Carlo test (499 permutations).
The indicator value method (IndVal) [67] was used to identify the most characteristic species of three groups defined by the CCA. Indicator values were calculated using abundance and frequency of occurrence of each species in each group as the input data. The species with p <0.05 based on 499 permutations from the Monte Carlo test were considered to be indicator species. The IndVal was performed by means of the R package indicspecies (ver. 1.7.1).

Environmental Variables
The values of the environmental variables measured and analysed for all springs studied are presented in Table 2. The water temperature ranged from 6.7 to 16.

Species Composition and Structure of Diatom Assemblages
The complete list of species was reported in the supplementary table. A total of 176 taxa from 56 genera were found of which 5 (5 genera) were centric and 171 (51 genera) pennate. Ellerbeckia arenaria was the most abundant centric species with a maximum RA = 5.3%. The most rich-species genera were Gomphonema (20), Nitzschia (17), and Navicula (14), followed by Achnanthidium (10), Diploneis (9), and Sellaphora (6). Overall, the species common to samples collected in the islands were 22 and those common to samples collected in the mainland were 11.
The taxa found in single sites were 101 from 44 genera (57% of the total species). Among these, the most abundant were Cymbella tridentina, Eunotia arcubus, Fallacia insociabilis, F. muraloides, Gomphonema aff. cymbelliclinum, Hannaea arcus, Nitzschia frustulum, Platessa conspicua and Psammothidium grischunum. The number of taxa found in singles sites was higher in the islands (66) than the mainland (35). The number of rare taxa found in single sites was also higher in the islands (62) than the mainland (24). Distribution, the number of taxa found in single sites and the number of rare taxa (RA <5%) found in singles sites are presented in Figures 3 and 4.   Table 1.  Table 1.    Table 1.  Table 1.
The results of the structural indices are reported in Table 4

Comparison among Diatom Assemblages
The nMDS ordination showed a clear separation of the diatom assemblages in the springs of Friuli from those of all other regions ( Figure 5). The springs RR in Friuli, VA in Sardinia and TO in Veneto were also quite separated from their respective groups. Significant differences among assemblages were confirmed by the ANOSIM test (global R = 0.610; p = 0.001) and pairwise tests (Table 5).
Water 2019, 11, x FOR PEER REVIEW 13 of 22 Figure 5. Non-metric multidimensional scaling (nMDS) ordination plot for diatom assemblages from the springs studied. All codes of the sites are reported in Table 1.

Relationships with Environmental Variables
In the CCA analysis (Figure 6), the first two axes accounted for 60.5% of the total variance of diatom species and environmental data (axis 1: 44.8% and axis 2: 15.7%). Ordination data distinguished three main groups of sites and species. The first group, situated in the right side of the plot includes the sites of the Friuli, with higher discharge and more alkaline waters. This group is composed of Achnanthidium lineare, A. pyrenaicum, Denticula tenuis, Gomphonema elegantissimum, and G. tergestinum. The second group, situated at the upper left part of the plot contains the sites of the islands, characterized by higher temperature and mineralization of water (Cl − and conductivity). This group is composed of Amphora pediculus, Caloneis fontinalis, Cocconeis euglypta, C. pseudolineata, Meridion circulare, Humidophila contenta, Kolbesia gessneri, Planothidium frequentissimum, and Pseudostaurosira alvareziae. The third group, situated at the bottom left part of the plot, include the sites of the Veneto, characterized by higher shading and impacts. This group is composed of Amphora inariensis, Sellaphora nigri, and S. seminulum. Overall, the significant variables for diatom assemblages were discharge (p = 0.018) and Cl − (p = 0.002).
The IndVAL analysis identified a total of eight indicator species for the three distinct groups of sites based on the CCA analysis (Figures 7-34, Table 6).  Table 1. The t-test, performed on the structural indices, indicated significant differences only in diversity as expressed by Shannon-Wiener's index (t = 2.304; p = 0.031).

Relationships with Environmental Variables
In the CCA analysis (Figure 6), the first two axes accounted for 60.5% of the total variance of diatom species and environmental data (axis 1: 44.8% and axis 2: 15.7%). Ordination data distinguished three main groups of sites and species. The first group, situated in the right side of the plot includes the sites of the Friuli, with higher discharge and more alkaline waters. This group is composed of Achnanthidium lineare, A. pyrenaicum, Denticula tenuis, Gomphonema elegantissimum, and G. tergestinum. The second group, situated at the upper left part of the plot contains the sites of the islands, characterized by higher temperature and mineralization of water (Cl − and conductivity). This group is composed of Amphora pediculus, Caloneis fontinalis, Cocconeis euglypta, C. pseudolineata, Meridion circulare, Humidophila contenta, Kolbesia gessneri, Planothidium frequentissimum, and Pseudostaurosira alvareziae. The third group, situated at the bottom left part of the plot, include the sites of the Veneto, characterized by higher shading and impacts. This group is composed of Amphora inariensis, Sellaphora nigri, and S. seminulum. Overall, the significant variables for diatom assemblages were discharge (p = 0.018) and Cl − (p = 0.002).  The IndVAL analysis identified a total of eight indicator species for the three distinct groups of sites based on the CCA analysis ( Figure 7, Table 6).

Main Characteristics of the Diatom Assemblages
Overall, the taxa abundant and common to more than six sites were Achnanthidium lineare, A. minutissimum, A. pyrenaicum, Amphora pediculus, Caloneis fontinalis, Cocconeis euglypta, C. pseudolineata, Denticula tenuis, Humidophila contenta, Meridion circulare, Planothidium frequentissimum, P. lanceolatum, and Sellaphora nigri. Most of these taxa were found in karst springs of different geographic areas in Europe, such as France [68], Poland [69], Austria [70], Slovenia [21], and Bosnia and Herzegovina [23]. By contrast, a large number of taxa occurred in single sites and most of them were rare (RA <5%). Our observations are supported by previous studies on springs in Italy [32,36] and France [68]. The high number of taxa found in single sites suggests a general marked heterogeneity of assemblages but also a specificity of each site at a regional level, since these species often have high affinity with specific abiotic conditions [32]. This aspect also underlines the role of the springs as refuges for many different species as recently reported by Taxböck et al. [7]. Several species with a strong affinity to spring environments or crenophilous species were found: Caloneis fontinalis, Cymbella tridentina, Diploneis krammeri Lange-Bertalot & Reichardt, Eunotia arcubus, Eunotia minor (Kützing) Grunow, Geissleria gereckei Cantonati & Lange-Bertalot, Gomphonema elegantissimum, Gomphonema lateripunctatum Reichardt & Lange-Bertalot, Meridion circulare, and Odontium mesodon. Most of them were recorded in karst springs both in high and low-altitude, with different disturbance levels e.g., [18,71,72]. Further, Meridion circulare and Odontium mesodon constitute a common association of species of karst springs of different geographic areas. In our study, the crenophilous species were generally more abundant in springs with low impact: C. fontinalis at SG and BR (score =3-5), C. tridentina at AL (score = 2), E. arcubus at SV (score = 3) G. elegantissimum at BM (score =2), M. circulare at UL, VA and RR (score = 3) and O. mesodon at RR (score =3). Cymbella tridentina, found only at AL (1249 m a.s.l.) in Friuli, seems to be the only crenophilous species with a distribution restricted to high-altitude springs in good accordance with previous studies that reported this species from the uppermost sections of carbonate spring-fed streams of the Alps at elevations >1200 m a.s.l. [18,73]. Many taxa are not closely linked to the aquatic environment according to Van Dam et al. [59]. Examples of the pseudaerial/euaerial species, belonging to categories 4+5, (i.e., species occurring in temporarily dry places or outside water bodies) were Achnanthes coarctata Whilst being absent in the springs of Friuli, characterized by high discharge, these species were observed mainly in springs with low discharge in Sardinia and Veneto. In these sites, these species can find suitable parts of substrata, temporarily uncovered by the water film due to the low discharge and water abstraction. Their presence underlines the role of springs as multiple ecotones, in particular, the land-water transition.
The diatom flora of the springs studied was composed by a small group of centric species rarely observed in spring environments: Aulacoseira sp., Cyclotella meneghiniana Kützing, Ellerbeckia arenaria, Melosira varians C. Agardh, and Pleurosira laevis (Ehrenberg) Compère. These centric diatoms were found only in the Sardinian springs, generally with very low abundance, and could be favoured by low current velocity. Similar findings were reported in previous studies on thermo-mineral springs of the island [74,75].

Structure of the Diatom Assemblages
In general, the springs studied hosted rich and diversified diatom communities with a balanced distribution of the species as expressed by species richness (R), diversity (H'), and evenness (J') indices. High species richness (≥25 taxa) was found at several sites, in Sardinia and Friuli, both with higher naturalness and strongly modified by water abstraction systems. In fact, the coexistence of numerous species seems to depend on a complex combination of local abiotic factors and contrastingly disturbed microhabitats, besides the degree of naturalness of the site [8,32]. The values of species richness were comparable among spring types, i.e., R-L vs LR, except for the spring SG, which showed a higher number of species. This is in agreement with an extensive study carried out on different spring types in the southeastern Alps [76].
The t-test performed on structural indices revealed significant differences between islands and the mainland only in diversity as expressed by the Shannon index (H').

Comparison of Diatom Assemblages from Islands and Mainland
The nMDS ordination and ANOSIM test performed on the most abundant taxa did not show a real distinctiveness among diatom assemblages from islands and the mainland. However, the diatom assemblages from the sites of Friuli formed a well-separated cluster and revealed high dissimilarity with those from all other regions. Differences in this group of springs seem to be attributable to specific local factors, such as hydrological features (high discharge and current velocity) and lower mineralization level of the water. In fact, these factors are considered among the most important drivers for the growth and distribution of diatom species [17]. In the nMDS plot, the springs RR, VA and TO were also quite separated from their groups, respectively in Friuli, Sardinia, and Veneto. In Sardinia, VA was the only spring with a seasonal regime. The water flow permanence was an important driver for diatom flora in this site as reported in a previous study [32]. Further, this factor was indicated as a relevant local hydrological factor for diatom assemblages in springs [8,77]. By contrast, differences for the springs RR and TO seem less clear and are probably due to a combination of different factors.
Considering the whole data set, we found further differences in diatom assemblages from the islands and the mainland. For example, the number of taxa found in single sites and the number of rare taxa (also found in single sites) were higher in springs of the islands than those of the mainland. This suggests that the role of springs as refuges for rare species may be greater in the islands and that the condition of geographic insularity may be important in maintaining a high level of biodiversity. In fact, a limited number of sites and low abundance were indicated as characteristics of rare diatom species [78]. In addition, springs are considered in themselves water islands with specific and well-differentiated biocoenoses due to their scattered distribution [8][9][10][11][12][13][14][15][16][17][18][19] and the presence of species with restricted distributions are expected in more isolated freshwater systems because of geographic, physical, and chemical barriers [8].

Relationships Diatoms-Environmental Variables
Among the environmental variables measured and analysed in this study, discharge and Cl − , respectively associated with the sites of the islands and Friuli, explained the significant amount of variance in diatom assemblages according to CCA analysis. Our results seem to reflect well the characteristics of these sites. For example, water mineralization was reported as an important feature of the springs in Majorca Island [26,79]. In addition, the low discharge was indicated among the most important environmental variables determining the diatom assemblages in low-elevation carbonate springs [72].
The IndVal analysis revealed indicator species quite consistent with the two major gradients highlighted by the CCA analysis. The indicator taxa for the springs of Friuli (group 1) were Achnanthidium pyrenaicum, A. lineare, Gomphonema elegantissimum and G. tergestinum. The first two species of Achnanthidium genus were observed in all sites of Friuli. These species are reported as well-adapted at high discharge and current velocity in the literature e.g., [45,80]. Gomphonema tergestinum is also reported as a species present in fast-flowing running waters [45] while G. elegantissimum seems to prefer moderate current velocity [80] but in these springs may be favoured by the availability of nutrients. For the springs of the islands (group 2) Cocconeis euglypta was the only indicator species. This species seems to be typical of flowing waters with medium-high mineralization [81,82]. However, habitat and ecology of this species are not yet well known because it was not consistently identified over time e.g., [45]. The springs of Veneto (group 3) were characterized by Amphora inariensis, Sellaphora seminulum and S. nigri. Amphora inariensis was found in undisturbed and sometimes moderately impacted freshwater environments [45]. Sellaphora seminulum was reported in springs and intermittently-drying small water bodies but the ecology of this species is not still precisely known [45]. Instead, S. nigri seems to be related to conditions of greater anthropogenic impact and was indicated as a well-known indicator of nutrient enrichment [51]. This species was reported in low-elevation carbonate springs affected by anthropogenic disturbance [72]. However, NGS data suggest great genetic diversity, and the consequent likely necessity of future revisions of its taxonomy and ecology [45].

Conclusions
This study provides the first comparison of diatom biodiversity in karst springs with different anthropogenic impact levels in two Mediterranean geographic areas with contrasting characteristics (islands and mainland). In general, the biotic integrity level was good both in sites with greater naturalness and in those strongly modified by water abstraction systems. However, crenophilous species were generally present and abundant in less impacted springs. The diatom flora was very heterogeneous and significant differences were found in the diversity of assemblages between islands and the mainland. The springs in the islands also recorded the highest number of species observed in single sites, as well as of rare species in single sites. Based on the most abundant species, significant differences were observed according to our initial hypothesis. Discharge and Cl − were two important variables in the springs studied and significant factors influencing diatom assemblages. Our results confirm the role of springs as refuges for rare species and suggest that the geographic insularity may be an important factor in maintaining diatom biodiversity. However, Mediterranean karst springs are fragile ecosystems increasingly exposed to natural and anthropogenic pressures. Morphological alterations and uncontrolled water abstraction as a result of inadequate management may cause a reduction of their ecological integrity and a loss of biodiversity in the future.