On the influence of water conductivity, pH and climate on bryophyte assemblages in Catalan semi-natural springs

Bryophytes are some of the most sensitive biological indicators of environmental change. Springs have a significant presence of bryophytes and so are ideal habitats for studying their relationship with the environment. We tested whether bryophyte assemblages can be explained with macro-, meso- and micro-ecological variables (i.e. seasonal climate, altitude, water pH and conductivity) sampling bryophytes from 198 semi-natural springs distributed along montane regions in the north-eastern Iberian Peninsula. We tested the influence of environmental variables on bryophyte assemblages in springs using sparse Partial Least Squares. Our results show that variability in bryophyte assemblages is explained by seasonal climate (temperature and precipitation from winter, spring, summer and autumn and temperature and precipitation seasonality), altitude and water conductivity. The results obtained by the present study will be useful for predicting bryophyte diversity in springs using simple and easy to obtain variables such as climate, water pH and conductivity.


Introduction
Bryophytes are very sensitive to environmental conditions (Kapfer et al., 2012), mainly because they do not have a well-developed cuticle or epidermis, and hence most of them can absorb water and nutrients through their entire surface. This is why bryophytes are a very interesting group to study the effects of environmental variability and pollution on community composition. In fact, bryophytes have previously been used as bioindicators of air pollution (Slack, 1990;Suren & Ormerod, 1998;Suren & Duncan, 1999) and water quality (Arts, 1990;Vanderpoorten & Klein 1999;Ceschin et al., 2012).
One of the habitats that can be dominated by bryophytes is the headwaters of rivers and springs. In these environments, climate, altitude, water pH and conductivity have been suggested to determine which species can become established (Tomaselli et al., 2011;Ceschin et al., 2012;Tessler et al., 2014;Vieira et al., 2014). However, other micro-and mesoscale environmental variables such as substrate or habitat heterogeneity (Suren, 1996;Strohbach et al., 2009;Vieira et al., 2014), aspect and slope (Pentecost & Zhang, 2006) or disturbance regime of floods and droughts may also be factors explaining variation in bryophyte assemblages in aquatic and semi-aquatic habitats (Suren & Ormerod, 1998;Suren & Duncan, 1999). In similar habitats, such as peatlands, some studies have suggested that bryophyte assemblages depend on gradients of temperature and precipitation (Gignac, 1994), water pH and conductivity (Gignac, 1992) or shade and permafrost (Belland & Vitt, 1995). Other studies carried out in temperate ecosystems of England (Callaghan & Ashton, 2008a), coastal zones of Canada (Belland, 2005) and semi-arid zones of Australia (Eldridge & Tozer, 1997) indicate that the different bryophyte assemblages depended on precipitation, soil pH and geology, showing a clear differentiation between communities located over calcareous and non-calcareous lithologies (Callaghan & Ashton, 2008b;Spitale et al., 2009;Virtanen et al., 2009). Additionally, bryophyte communities have also been suggested to vary along altitudinal gradients (Wolf, 1993;Andrew & Rodgerson, 2003;Bruun et al., 2006;Grytnes et al., 2006;Ah-Peng et al., 2007;Grau et al., 2007;Spitale, 2016), which can be a proxy for differences in topology and radiation amongst others.
Notwithstanding all this research on bryophyte ecology, the study of differences in bryophyte assemblages along gradients of altitude, seasonal temperature and precipitation, water pH and conductivity combined, is still very limited in the region of Catalonia and especially in semi-natural or man-made springs . In our study region (Catalonia, NE Iberian Peninsula), the ecology of bryophytes has been poorly studied (but see Peñuelas, 1983;Peñuelas & Sabater, 1987), especially concerning the interaction between macro-, meso-and microecological variables. Hence, the aim of this study was to test to what extent macro-, meso-and microecological variables such as climate (seasonal temperatures and precipitation and their seasonality), altitude and water pH and conductivity can explain variation in bryophyte species assemblage in springs within a relatively large climate and water chemistry gradient. To achieve our aim, we surveyed springs and their bryophyte communities in montane regions of north-eastern Iberian Peninsula (Catalonia, NE Iberian Peninsula). Based on results provided by previous studies, we hypothesised that water pH and conductivity would play a main role determining bryophyte communities of semi-natural springs, due to the direct contact between the bryophytes and the water, and species differences in tolerance ranges for these properties. In addition, because in the Mediterranean climate, summer is the warmest and driest season of the year, we hypothesised that climate from the summer season (dry and warm season) would also significantly explain variability in species composition of spring bryophyte communities. Altitude was also hypothesised to be an important factor for determining spring bryophyte assemblage as it can reflect differences in topography, radiation and climate.

Study area and experimental design
We surveyed 198 water springs following a 95 km south-to-north and a 50 km east-to-west gradient along the north-eastern Iberian Peninsula territory (Figure 1), comprising 4 different mountain ranges: the Central Littoral mountain range, Montseny-Guilleries, Lluçanès and Garrotxa. The springs were unevenly distributed across the different mountain ranges: 101 springs occurred in the Montseny-Guilleries, 55 in the Central Littoral mountain range, 32 in Lluçanès and 10 in Garrotxa. Climate also differs amongst these mountain ranges. The Central Littoral mountain range has a typical maritime Mediterranean climate with mild summers and winters, while the climate is Mediterranean humid and sub-humid in Montseny-Guilleries and Lluçanès and Mediterranean pre-Pyreneeal in Garrotxa (Martín-Vide, 1992), being more continental. Lithologically, the springs of the Central Littoral mountain range are located over different types of granite, except for the eastern side which is dominated by metamorphic rocks Fernández-Martínez et al., 2016). In Montseny, springs are located over granite, metamorphic and calcareous rocks, while in Guilleries, most of them are over granite. Lluçanès is completely calcareous and Garrotxa is divided into calcareous and volcanic rocks. This rich lithology produces a large variability in the characteristics of water from springs, affecting dissolved ions and thus conductivity and pH .
Most of the springs we surveyed are small humanmade constructions to collect water from underground and release it through a spout ( Figure S1). Some of the springs drained water mines (collecting water from aquifers), while others were built in naturally occurring springs. However, in all of the surveyed springs, flowing water did not receive any sanitary treatment. The morphology of our springs did not vary greatly (for a typical example, see Figure S1). Almost all of them contained the same microhabitats such as a spout, wet rock walls, soil ground/walls and a sink where the water makes a little pool. Surveys of the springs were carried out during spring and autumn seasons of the years 2013 and 2015, and although many other springs were surveyed, only those that had water flowing at the time of the survey were sampled and included in this study. We also excluded from the analysis any springs whose water flow was controlled by a tap. Therefore, the springs we sampled contained bryophytes that were continuously receiving water for most of the year, apart from springs in which drought had temporarily interrupted water flow. Water pH and conductivity of the flowing water was analysed in the field with pH-meter (ORION, Thermo-Scientific, Waltham, MA, USA), previously calibrated with standard solutions at pH 7 and 10, while electrical conductivity was measured using a conductivity-meter (WTW, Xylem Inc., Weilheim, Germany) calibrated at 25°C and with a standard solution of 1314 µS cm -1 .
To determine the bryophyte communities (as the presence or absence data), we took a sample of the different bryophyte species present in the spring that were either in direct contact with the water flow, receiving water drops or under water (see yellow line in Figure S1). We did not standardise our bryophyte measurements by the total area sampled because (1) sampled area was lower than 0.9 m 2 for most of the springs, for which differences in scale were minimal, (2) species-area relationship was missing for a subset of our dataset ( Figure S2) and (3) surveying a predetermined area would have missed some species present in the springs. Species number per spring was generally very low, most of them having between 1 and 4 species ( Figure S3). We used identification keys in Smith (1978Smith ( , 1990 and Casas et al. (2001Casas et al. ( , 2004 and follow the nomenclature established by Hill et al. (2006). Some of the samples could not be identified to species level (3%, 18 of 538 samples). To study the variation in bryophyte assemblages, we used those bryophyte species that appeared in more than 10 springs (13 species in total), to ensure enough replicates per species.
Using the geographical coordinates of the springs, obtained in the field with a GPS, we extracted average seasonal (winter, spring, summer and autumn) temperatures and precipitation, as well as mean annual temperature (MAT) and precipitation (MAP), from the Digital Climatic Atlas of Catalonia (Pons, 1996;Ninyerola et al., 2000; available at http://www.opengis.uab.cat/acdc/index.htm) with a resolution of 180 m. Seasonal climate was calculated as the average temperature and total precipitation values of December to February for winter, March to May for spring, June to August for summer and September to November for autumn. Using seasonal climate, we also calculated temperature and precipitation seasonality as the coefficient of variation of seasonal temperatures and precipitation. Altitude was obtained using a GPS in the field.

Statistical analyses
To test whether bryophyte assemblages depend on environmental conditions, we performed a sparse Partial Least Squares (sPLS) analysis (Lê Cao et al., 2008), fitted by the regression mode, where water pH and conductivity, seasonal temperatures and precipitation, temperature and precipitation seasonality and altitude were the environmental matrix of predictors and the binary variables of bryophyte species presence were the response matrix to be predicted. Because the response matrix (species presence) was a binary variable, we fitted the model after applying a log-ratio transformation. PLS family methods work by finding the underlying relationships between two matrices (X and Y, predictors and responses) by extracting latent variables to model the covariance structures of the two spaces. PLS models attempt to estimate axes or variates in the X matrix that explains the maximum variance in Y matrix. These methods are particularly useful when the X matrix has more variables than observations, and when high multicollinearity is present among X values. Additionally, sparse PLS methods include a process of selection of the predictor variables to facilitate biological interpretation of the results. The sPLS model was fitted using the "spls" function in mixOmics R package (Le Cao et al., 2017). Tuning of the model suggested that extracting more than two variates did not improve the prediction of the response matrix (total Q-value saturates at the second variate extracted). To visualise the results of the sPLS analysis, we plotted a Clustered Image Map (CIM), using the "cim" function in mixOmics, which shows the correlations between the predictor and the response variables while clustering both based on their similarity on the multivariate space.
To look for significant differences of the variables selected by the sPLS analysis among species, we performed analyses of variance (ANOVA) in which we related the environmental variables selected in the sPLS as a function of the species. This analysis allows us to estimate ranges of tolerance for the species and to test whether different species share their ecological niche or not. Differences amongst species were tested using the post-hoc Tukey HSD test. These analyses were based on the 82 springs in which at least one of the most frequent species was found. All ANOVAs accomplished the required statistical assumptions of normality, homoscedasticity and independence of the residuals. All the statistical analysis was performed using R statistical software (R Core Team, 2015).

Water chemistry and climate of the studied springs
We found a relatively large gradient in water chemistry and climatological features of the surveyed springs. MAT ranged from 7.3 to 15.7°C, MAP from 555 to 1180 mm year -1 , pH from 5.7 to 8.8, conductivity from 19 to 2160 µS cm -1 and altitude from 100 to 1570 m a.s.l. The springs of the Littoral mountain range are located at lower altitudes, being on average 328 m above sea level (a.s.l.), and, therefore, experienced the highest temperatures (13.9°C on average) ( Table 1). Springs from Lluçanès, Montseny-Guilleries and Garrotxa are found at intermediate altitudes, between 550 and 850 m a.s.l., and recorded temperatures between 11 and 12.5°C, although some of Montseny springs experienced relatively high temperatures (up to 15.7°C). Annual precipitation in La Garrotxa is higher than in any other of the studied regions (1049 mm), followed by Montseny-Guilleries, whereas Lluçanès and the Littoral mountain range are drier (Table 1).
In general, pH of the springs surveyed was slightly alkaline. However, pH in Garrotxa's springs is higher than those from the other regions studied (Table 1). Conductivity, though, is much more variable than pH. Lluçanès' springs are, on average, those with highest conductivity (1059 µS cm -1 ), whereas springs from the Littoral mountain range and Garrotxa are found in the mid-range (around 700 µS cm -1 ), and those from Montseny-Guilleries have the lowest conductivity values of all (286 µS cm -1 ) ( Table 1).

Relationships between species presence and environmental variables
The sPLS model included conductivity, altitude, spring, summer and autumn temperature, winter, spring, summer and autumn precipitation and temperature and precipitation seasonality as important predictors of bryophyte species assemblages in springs ( Figure 2). The first variate explained 51.1% of the variability in the predictors and 12.9% of the variability of the bryophyte assemblages. As shown by Figure  S4, no single species or spring was driving the results of our sPLS model. This variate was positively related to seasonal precipitation, altitude and temperature seasonality and negatively related to conductivity, seasonal temperatures and precipitation seasonality ( Figure S5). The second variate explained 28.1% of the variability in the predictors and 11.0% of the variability of the bryophyte assemblages. This variate was positively related mainly to summer precipitation and altitude and negatively related mainly to seasonal temperatures and precipitation seasonality.
Species were clustered into two big clusters (1 and 2) with two smaller clusters each (a and b, Figure 2). The first cluster of species (1a), formed by Eucladium verticillatum, Pohlia melanodon, Didymodon tophaceus and Pellia endiviifolia, was found to be related to warm and especially dry climate and high water conductivity. Their presence was also negatively related to altitude of the springs and temperature seasonality and, less strongly, positively related to precipitation seasonality. Palustriella commutata and Cratoneuron filicinum, which formed cluster 1b, were found to be associated with dry and cold climate with low temperature and precipitation seasonality, but with rainy summers and located at high altitude. Palustriella commutata was also positively related to high water conductivity, while no relationship was found with C. filicinum.
Kindbergia praelonga, Oxyrrhynchium speciosum and Conocephalum conicum formed cluster 2a and were more likely to be found in springs that have a warm and relatively dry climate in spring, and especially in summer, and high water conductivity (except C. conicum). The species of this cluster were also found in springs of low altitude and low seasonal Table 1 Average values for altitude, climate (mean ± standard error of the mean [SE]), water pH and conductivity and bryophyte species richness of the springs and average values per region. Lluçanès is a calcareous region, Montseny and the Central Littoral mountain range have granitic and metamorphic lithology and Garrotxa springs are located over volcanic and calcareous rocks. Acronyms are MAT for mean annual temperature and MAP for mean annual precipitation. Units are degrees Celsius for MAT, mm year -1 for MAP, µS cm -1 for conductivity and metres for altitude. variability of temperature and precipitation (low continentality). The last cluster of species (2b) was formed by Plagiomnium undulatum, Brachythecium rivulare, Bryum pseudotriquetrum and Platyhypnidium riparioides. These species were the group of mosses related to the springs with the coldest and wettest climate, highest temperature seasonality but lowest precipitation seasonality and highest altitude and having also low water conductivity. Plagiomnium undulatum, however, was not so strongly related to cold climate. Of all studied species, P. endiviifolia and C. conicum were the species with the lower correlations ( Figure 2). Table 2 shows the average values per species for the environmental variables included in the sPLS. Significant differences appear amongst species for all of the variables, which further corroborates results from the sPLS. This indicates that ecological niche separation occurs within the environmental variables considered.

Discussion
Our results show that macroclimate, altitude and water conductivity can explain variability in bryophyte assemblages in semi-natural and man-made springs. The distribution of some of the most abundant bryophyte species in our springs is clearly separated into different ecological niches, defined by the environmental variables we considered. In particular, our findings suggest that models predicting species distributions could be improved by including micro-and mesoscale environmental variables such as water conductivity and altitude in addition to macroscale variables such as climate. However, other environmental variables considered in previous studies (Longton et al., 1983;Cattaneo & Fortin, 2000;Tessler et al., 2014;Corbera et al., 2015), such as main ions (Ca 2+ , Na + , Mg 2+ , K + , SO 4 2-, Cl -, PO 3 -4 ), pollutants such as nitrate (Vanderpoorten & Klein, 1999) or type of microhabitat and its heterogeneity (Suren, 1996) could help to differentiate the ecological niche of bryophyte species even further. Despite that, the present study clearly shows several general trends in bryophyte assemblages in springs.

Determinants of bryophyte assemblages
The role of precipitation on bryophyte assemblages is evident in previous studies (Eldridge & Tozer, 1997;Gignac, 2001;Callaghan & Ashton, 2008a) although some other papers highlight that the number of rainy days per year is sometimes more important than the total amount of precipitation (Ratcliffe, 1968;Callaghan & Ashton, 2008b) because the time that the bryophytes are dry is more determinant than the total amount of water they receive (Pentecost & Zhang, 2006;Proctor et al., 2007). However, this fact may not be so relevant in our study, because most of the sampled springs provide water almost constantly throughout the year. We found that B. rivulare, P. riparioides and B. pseudotriquetrum, which are species associated with streams and believed not to tolerate desiccation, are those with the highest affinity for high precipitation (Figure 2, Table 3). However, other species with supposedly equally high affinity for humidity like P. endiviifolia, P. commutata, C. filicinum and E. verticillatum (Table 3), do not show such a strong positive association with precipitation, but even show the opposite (Figure 2). Kapfer et al. (2012) suggested that the length of time water resides in the aquifers of zones with lower precipitation is longer than in zones with higher precipitation. A longer time in contact with the rock causes the water to dissolve more salts and increase its conductivity. This could explain the presence of known calcicolous species, associated with high conductivity water, in springs with low precipitation  ( Figure 2). Despite the fact that bryophytes are able to tolerate wide ranges of temperature, we found different average seasonal temperatures and temperature seasonality (which relates to continentality) per species. This effect of temperature on bryophyte assemblages may also be mediated through its effect on the water availability of the spring. Almost all of our springs had forested areas nearby and which typically consume large quantities of subsoil water, especially under high temperatures (Fernández-Martínez et al., 2014), thus reducing water runoff. Also, continentality increases temperature extremes throughout the year, which can be an important factor favouring certain species versus others, depending on their temperature tolerance ranges.
Some studies have shown a great polarisation in the distribution of bryophytes in relation to calcareous and non-calcareous lithology (Callaghan & Ashton, 2008a;Virtanen et al., 2009) because of big differences in the cellular exchange capacity of Ca 2+ (Bates, 1982). In our study, species with affinity for calcareous lithology, like D. tophaceus and E. verticillatum (Table  3), have been found in springs with high water conductivity (on average, 972 and 886 µS cm -1 , respectively), while calcifugous species like B. rivulare (Longton et al., 1983) have been found in springs with low water conductivity (292 cm -1 ). However, in contrast to other studies carried out with bryophyte communities from peat lands and in the headwaters of rivers (Gignac, 1992;Tessler et al., 2014), we did not find any significant relationship between water pH and bryophyte assemblages (Figure 2). This fact may be the result of having analysed only the 13 most frequent bryophyte species that were found in the springs. As these species are more frequent, they are also presumed to be more likely to be generalists and to tolerate wider ranges of water pH. Further work, based on more intensive sampling and focused on less frequent species, will help elucidate how water pH, conductivity and other environmental variables (e.g. climate, topography and ionic composition of water) affect bryophyte assemblages in Mediterranean springs. However, our results indicate that water conductivity could be a very suitable variable to include in species distribution models to improve their predictions on humid or sub-humid Table 2 Species average (± standard error of the mean) values for environmental variables that were selected in the sPLS for explaining variability in species distribution. Letters (a, b, c, d, e, f) indicate groups according to the post-hoc Tukey HSD test for multiple comparisons. All ANOVA tables were statistically significant at the <0.001 level. Units were metres for altitude, mm year -1 for precipitation variables and Celsius degrees for temperature variables and (µS cm -1 ) for conductivity. Precipitation seasonality (PS) and temperature seasonality (TS) were unit less. Abbreviations: T indicates temperature, P indicates precipitation, and seasons were indicated in lowercase letters as winter (w), spring (sp), summer (sm) and autumn (a). bryophytes, given the fact that it is relatively easy to obtain (it could be inferred from lithology) and significantly separates bryophyte species.

Conclusions
Our results clearly demonstrate that bryophyte assemblages in semi-natural springs can be explained using widely available variables such as mean seasonal temperatures and precipitation, temperature and precipitation seasonality (intra-annual variation), altitude and water conductivity of the springs. Our results also point out the need to include micro-and mesoscale variables, in addition to macroscale variables to help improve the prediction of species distribution models.