Environmental and spatial patterns as drivers of littoral macroinvertebrate assemblages in patchily distributed mountain lakes: Contribution to typology design

Limnologica 62 (2017) 57–67

Contents lists available at ScienceDirect

Limnologica journal homepage: www.elsevier.com/locate/limno

Environmental and spatial patterns as drivers of littoral macroinvertebrate assemblages in patchily distributed mountain lakes: Contribution to typology design F. García-Criado ∗ , C. Martínez-Sanz, L.F. Valladares, C. Fernández-Aláez Faculty of Biology and Environmental Science, University of León, 24071 León, Spain

a r t i c l e

i n f o

Article history: Received 16 December 2015 Received in revised form 8 November 2016 Accepted 9 November 2016 Keywords: Alpine ponds Macrobenthic invertebrates Spatial autocorrelation Ecotypes Water Framework Directive

a b s t r a c t Biological assemblages are affected by both environmental and spatial processes. Spatial autocorrelation can be specially marked in discrete ecosystems patchily distributed over a large region (e.g., lakes arranged in districts). Lake typologies are exclusively based on environmental features, but we hardly know to what extent spatial patterns can hinder their implementation. We analysed the role of environmental factors and spatial autocorrelation in shaping littoral macroinvertebrate communities of 51 mountain lakes from a large Spanish region in order to test: 1) the suitability of the variables currently used to construct typologies; 2) the influence of spatial patterns on typology implementation. Biologically meaningful types of lakes were created and described by means of cluster analysis (Jaccard index) and multiple discriminant analysis. Water permanence, substrate type and vegetation were the main drivers of the assemblage composition. The cluster analysis and Mantel tests showed that spatial patterns did not generally hamper recognizing lake types. Only in the district with lakes closest to each other (Sanabria Natural Park), spatial autocorrelation was strong enough to overcome the effects of some factors (substrate type), but not others (water permanence). © 2016 Elsevier GmbH. All rights reserved.

1. Introduction Most organisms are non-randomly distributed. Communities have a spatial structure resulting from a large number of processes acting at different spatial scales. Local conditions are essential drivers of community composition, as shown for lakes by a number of studies conducted on macroinvertebrates (Free et al., 2009; Hinden et al., 2005; Johnson and Goedkoop, 2002). However, the composition of a local community is not only the result of environmental factors. Metapopulation dynamics and large-scale biogeographical and historical constrictions may play a role as well. Following the principles of metacommunities (Leibold et al., 2004), in small-scale studies, among-site differences in community structure can be attributed to environmental conditions of each particular site (including species interactions and environmental pressures), and to the importance of dispersal. With increasing geographic distance, dispersal limitation is likely to increase. This can cause spatial autocorrelation in the assemblages, that is, a tendency of neighbouring sites to harbour similar biotic assemblages

∗ Corresponding author at: Area of Ecology, Faculty of Biology and Environmental Science, University of León, Campus de Vegazana s/n. 24071, León, Spain. E-mail address: [email protected] (F. García-Criado). http://dx.doi.org/10.1016/j.limno.2016.11.002 0075-9511/© 2016 Elsevier GmbH. All rights reserved.

(Legendre and Legendre, 1998). Therefore, in broad-scale studies, we might expect closely connected sites to harbour more similar assemblages than sites further apart although the intensity of the effect depends on the characteristics (e.g., dispersal ability) of the organisms (Beisner et al., 2006; Borthagaray et al., 2015; Rádková et al., 2014; Razeng et al., 2016; Shurin et al., 2009). Statistical models of species distribution neglecting this aspect of ecology can lead to mis-estimations (Dormann, 2007). Since the 80 s, a growing concern about the spatial structure of communities has arisen in the scientific community. This concern has resulted in a number of studies aimed at disentangling the effects of spatial processes and environmental conditions (see a meta-analysis in Cottenie, 2005), some of them dealing with aquatic invertebrate communities (Briers and Biggs, 2005; Rádková et al., 2014). This issue can be particularly relevant in discrete ecosystems such as lakes, even more if sites have a patchy distribution, with lakes clustered in lake districts as is usual in mountain lakes of glacial origin (Catalan et al., 2009a). Monitoring of freshwater is usually undertaken at large spatial scales. The Water Framework Directive (WFD, Directive 2000/60/EC, 2000) established the catchment as the management unit, although monitoring programmes (e.g., typologies) are often designed for larger areas (national scale). At this scale, assemblage structure may be influenced by broad-scale spatial patterns

58

F. García-Criado et al. / Limnologica 62 (2017) 57–67

Fig. 1. Map of Castilla y León with the situation of the lakes included in the study. Lakes are named after the district (when assigned to one) or after the province they belong to.

resulting from dispersal limitation and historical effects, as well as by environmental differences. Most monitoring programmes are based on typologies aimed at reducing variability. For that purpose, they use a number of variables to account for both environmental features and geographical position (ecoregions). The need of differentiating ecoregions is widely accepted. In this respect, Kernan et al. (2009) found a strong geographical influence on the composition of several assemblages from mountain lakes at a pan-European scale and recommended dividing Europe into three regions. However, spatial patterns might arise at smaller spatial scales (Briers and Biggs, 2005; Mykrä et al., 2007), especially if patchiness of waterbodies is high. A crucial point is checking to which extent spatial patterns caused by the proximity of waterbodies (and not by differences in local, environmental conditions) may hamper the use of a typology and if this effect is noticed at the level of taxonomic resolution used in bioassessment, usually genus or family in the case of benthic invertebrates. Studies of this type must be undertaken at the catchment rather than pan-European scale (Catalan et al., 2009a). It is also necessary to clarify which environmental variables are to be selected to create ecotypes. Several attempts have been made to create typologies for management of European lakes (Kagalou and Leonardos, 2009; Kolada et al., 2005; Moss et al., 2003; Little et al., 2006). These typological schemes are often based on general limnological knowledge, usually skewed toward pelagic environment and biological groups (phytoplankton and zooplankton). It is hardly known to which extent they have a biological meaning for littoral macroinvertebrates, which might strongly depend on factors currently overlooked such as substrate type (Zenker and Baier, 2009). We need to gain knowledge on the factors shaping these assemblages at the spatial scale and taxonomical level used in management. A great deal of work has been done in mountain lakes since the approval of the WFD (for example, Kernan et al., 2009). The distribution of the studied areas, however, is uneven. There are many contributions for central Europe (e.g., Free et al., 2009; Füreder et al., 2006; Hinden et al., 2005; Oertli et al., 2008)

but not so for Spain, where research has mainly focused on the Pyrenees (e.g., De Mendoza and Catalan, 2010) and, to a lesser extent, the Central Range (Toro et al., 2006). Castilla y León is an extensive region largely coincident with one of the water management units in Spain: the Spanish part of the Duero River Basin. Its area is larger than several EU countries and comparable to many other river basins in Europe. The findings of this research, therefore, might be useful for management in other river catchments across Europe. A peculiar feature of the study area, common to many other mountain regions, is the presence of a number of patchily distributed mountain ponds and small lakes. Applying current typologies to littoral macoinvertebrate assemblages in these waterbodies poses several potential difficulties. Firstly, it is not known whether the variables used in these typological schemes are relevant for macroinvertebrates. Secondly, the patchy distribution of mountain lakes (in groups or districts) might create spatial patterns capable of blurring among-ecotype differences in the assemblages. The aim of this study was to determine which environmental factors influenced the composition of littoral macroinvertebrate assemblages in the mountain lakes of the region as a means to check whether the typological schemes commonly proposed for European lakes are appropriate for this assemblage. In particular, certain habitat variables usually overlooked, such as substrate type, might be relevant for littoral macroinvertebrates. We also aimed to check the role of spatial autocorrelation (community similarity among lakes of the same district) on the assemblage composition. We hypothesized that lakes in the same district would tend to show similar taxonomic compositions regardless of the type they belong to. 2. Material and methods 2.1. Study area Castilla y León (94,223 km2 ) is dominated by a vast, flat, central area (around 700 to 800 ma.s.l. on average) surrounded by moun-

F. García-Criado et al. / Limnologica 62 (2017) 57–67 Table 1 Number of lakes sampled (n) and mean, minimum and maximum distances between pairs of lakes within each district. Distances were measured from the center of the lake. Lake district

n

mean dist. (km)

min. dist. (km)

max. dist. (km)

Sanabria (SA) Gredos (GR) Urbión (UR) Neila (NE) Fuentes Carrionas (FC)

14 9 3 7 5

5.5 17.9 1.4 3.3 3.4

0.2 0.3 1.2 0.1 0.2

11.9 37.9 1.6 7.3 7.1

tains (up to 2,600 ma.s.l.). Most of it, nearly 79,000 km2 , is drained by the Duero River and its tributaries. Fifty-one mountain lakes and ponds were selected for this study (Fig. 1). The site selection included waterbodies with different size (from 0.3 to 12 ha), depth (from 0.3 m to 14 m maximum depth on the sampling date), water permanence (temporary and permanent), catchment lithology (calcareous and siliceous catchments), substrate granulometry in the littoral zone (from soft sediments to stony substrates), and altitude (from 1400 to 2,200 ma.s.l.). The shallowest, usually temporary, ponds supported dense beds of either submerged macrophytes, emergent macrophytes or both. Macrophytes were scarce or absent in the rest of the sites with the exception of Isoetes, which was present and often abundant in many lakes. This set of sites included both ponds and lakes according to some frequently used definitions (Biggs et al., 2005; Céréghino et al., 2008). In this paper, they are referred as lakes for simplicity. The distribution of the lakes in the region is patchy. Most of them tend to concentrate in four alpine areas which are on average 230 km apart (between 180 and 325 km, measured as Euclidean distance between centroids). In one of these areas (Neila and Urbión in the map), lakes are arranged in two distinct clusters (betweengroup distance 20 km, maximum within-group distance 7 km) and could be taken as separate groups. Thus, we have considered five areas (hereafter named as lake districts): Sanabria (SA), Gredos (GR), Fuentes Carrionas (FC), Neila (NE), and Urbión (UR). Several lakes within each district have been selected for the study (see Table 1 for additional information). Isolated lakes or those far apart from a district (Euclidean distance to the district higher than the average within-district distance) were considered separately. In this paper, the lakes are named after the district (when assigned to one) or after the province (Bu, Le, So). No attempt to select reference sites was made for two reasons: 1) only a few lakes in the area are free of environmental pressures, 2) it is not known whether the impact of pressures such as extensive grazing or fish introductions is strong enough to overcome natural patterns. Instead, the statistical analyses conducted in this study were used to identify sites with communities shaped by pressures rather than by natural factors (see the results section). Eight sites were thus removed from further analyses. 2.2. Sampling and variables measured The data used in this paper come from a single visit to each lake between 2004 and 2008. When data from several dates (years) were available, only the most recent sampling was taken. Previously, we checked that samples from the same lake and different year had similar assemblages (cluster analysis, not shown), indicating that inter-annual differences in the assemblage composition was small in relation to among-lake differences. All the samples were taken between mid June and mid July in order to minimize seasonal variations. Each lake is taken as a sample for statistical purposes. A number of variables were measured to characterize lake types: latitude, longitude, altitude, lake area, depth (maximum depth in the sampling date), water permanence (temporary vs. permanent),

59

and alkalinity (as a surrogate for catchment lithology). The characteristics of the substrate in the littoral zone (up to a depth of around 1 m) were expressed as percentage of the sampling area covered by silt, sand, gravel, and cobbles. The littoral vegetation was characterized by two attributes: percentage cover of submerged macrophytes (hydrophytes, hereafter) and percentage cover of emergent macrophytes (helophytes). Macroinvertebrates were collected by kick and sweep sampling with a D-frame net (FBA standard, mesh size 500 ␮m) following a multihabitat time-limited sampling (Collinson et al., 1995; Briers and Biggs, 2005). Five habitats were roughly considered: emergent macrophytes, submerged macrophytes other than Isoetes, soft sediments, sandy bottoms and stony bottoms (gravel or cobbles), the latter three with or without Isoetes but with no other macrophyte. A three to five minutes total sampling time for each lake (depending on the lake area) was shared proportionally among the main habitats of the littoral zone. These samples were pooled in order to obtain a single, integrated sample per lake. Specimens were preserved in 96◦ ethanol. The macroinvertebrates were sorted in the laboratory at 10 x magnification. When necessary, subsamples were taken until at least 600 specimens were collected and one sixth of the sample processed. In this case, an additional search for rare taxa was made on the rest of the sample. Taxonomic identification was done to genus level whenever possible. Exceptions were Sphaeriidae, Coenagrionidae, Aeshnidae (to family), Limnephilidae (to tribe) and Diptera (mostly to subfamily). Oligochaeta were only identified to class and were removed from the data set. 2.3. Data processing 2.3.1. Factors influencing the assemblage composition In order to simultaneously check which factors were more relevant in shaping the assemblage composition and create a set of lake types with biological meaning, a two-step statistical process was applied. Firstly, biological data (taxa present in each lake) were used to create clusters of sites from between-site dissimilarities as measured by Jaccard (average linkage method). The cluster analysis was performed with the Community Analysis Package (CAP), version 3.11. Secondly, a multiple discriminant analysis (MDA) was conducted to identify the abiotic factors driving the cluster membership (SPSS v.19). This analysis allowed us to describe the environmental characteristics of each cluster by identifying the variables differing among them. The variables set used in MDA included those commonly used to create ecotypes for monitoring purposes (longitude, latitude, altitude, lake area, maximum depth, water permanence, conductivity, alkalinity), as well as some additional variables related to habitat characteristics (percentage silt area in the littoral zone, percentage stony area in the littoral zone, coverage of submerged hydrophytes, and coverage of emergent vegetation). Non-normal variables were log-transformed or arcsin-transformed (in the case of percentages). As recommended by Legendre and Legendre (1998), no preliminary selection of variables was made in MDA. There are two main reasons for it: first, stepwise selection of explanatory variables does not guarantee that the best set of explanatory variables will be found; second, corelated variables can help in interpreting community patterns. The resulting MDA model was used for two purposes: 1) identifying which variables differed among clusters and thus describing lake types, 2) predicting to which cluster samples belonged to according to their environmental characteristics (cross-validation or “leaveone-out” option) and estimating the percentage of lakes correctly classified by the cluster analysis. The percentage of correctly classified sites was used as an aid to select a cut level in the cluster analysis. One-way ANOVA was used to test for among-cluster differences in the abiotic variables. Post-hoc, pair-wise differences were tested by applying the Scheffé test. The statistical procedures

60

F. García-Criado et al. / Limnologica 62 (2017) 57–67

so far described (clustering plus MDA) allowed us to create and characterize lake types. In an attempt to describe the assemblages associated to each lake type, taxa closely associated with each type were defined by calculating their indicator value (IndVal) as proposed by Dufrêne and Legendre (1997): IndValij =

Nindij Nindi

×

Nsitesij Nsitesj

× 100

where Nindij is the number of individual of taxon i in cluster j, Nindi is the total number of individuals in all the clusters, Nsitesij is the number of sites in cluster j where species i is present and Nsitesj is total number of sites in that cluster. 2.3.2. Spatial patterns In spatially autocorrelated assemblages, we might expect sites close to each other to support more similar assemblages. Therefore, we firstly constructed two matrices, one of geographical (physical) distances between pairs of lakes, and another of assemblage similarity between pairs of sites measured by the Jaccard index. The correlation between both matrices was assessed by Mantel tests (PASSaGE 2 software). The significance of the test was evaluated with a randomization procedure (10,000 permutations). Finally, in order to summarize the environmental characteristics of the lakes, a principal component analysis (PCA) was made using the same environmental variables as in MDA. This analysis was used to check whether spatial autocorrelation could be better explained by environmental differences (that is, sites with similar assemblages were also environmentally more similar) or by geographical distance itself (as a result of dispersal or historical processes). 3. Results Ninety-nine taxa (84 genera) were identified. Richness values per lake were between 5 and 36 (average 18.8). The lowest values correspond to severely impacted sites. Fig. 2 shows the result of the cluster analysis, including the two cut levels considered as well as information on the pressures observed. In general terms, environmental pressures (fish presence, livestock in the catchment, or dammed lakes) did not have an appreciable effect on the community composition since lakes with and without pressures clustered together. However, there were some noticeable exceptions. Eight sites were classified apart from the rest (clusters 4, 5 and 6 in Fig. 2). These clusters included the five lakes in the study area which were intensely and repeatedly stocked with salmonids (Salmo trutta and Onchorrhynchus mykiss) until late 90 s for sport fishing (NE-2, NE-3, NE-5, Bu-1 and UR-1), as well as the only one in the study area (Le-3) with carps (Cyprinus carpio). These sites were not otherwise different from those in the remaining clusters, as shown by the values of the environmental variables (Fig. 3). They could be described as the sites of the study most severely impaired by fish stocking. Sites Bu-2 and Le-8 are temporary ponds which could be considered as misclassified items. The eight sites in clusters 4 to 6 were removed from further analyses. Results referring to the three remaining clusters are explained next. Three major groups were created by the cluster analysis (clusters 1 to 3, Fig. 2). As deduced from the MDA results (Table 2) and the values of the variables within each cluster (Fig. 3), cluster 1 included permanent lakes with coarse substrate in the littoral zone (that is, with little or no silt) and located at high altitude (above 1,800 ma.s.l.); cluster 2 included permanent lakes at medium altitude and with varying substrate granulometry and vegetation cover, whereas lakes in cluster 3 were very shallow, mostly temporary and with a dense helophyte cover. The MDA model classified 79.1% of the sites in the correct group.

Table 2 Summary of results of the MDA applied to the first cut level of the cluster analysis (clusters 1 to 3 in Fig. 2). Only variables significantly differing among clusters (ANOVA) are included. Discriminant function

Variance explained

p-value

Variables correlated with each function

1

72.1%

<0.001

2

27.9%

0.001

helophyte cover, silt, water permanence, cobbles, depth, area, hydrophytes altitude, longitude

A step further in the clustering created a 5-group scheme (1.1, 1.2, 2.1, 2.2, and 3.1, Fig. 2) and a single sample segregated from group 3 (3.2). The interpretation of the MDA results was essentially the same, but the number of correctly classified sites was lower (61.9%), mostly because cluster 2.1 was meaningless (only 14.3% of the sites in this group was correctly classified). The most remarkable feature of this second clustering level was the creation of a group (named 2.2) which included almost all the permanent ponds (but not the temporary ones) located in Sanabria Natural Park and none from outside the park. This supports the assumption that lakes in this district have an assemblage with particular features. The spatial pattern suggested by the cluster analysis was confirmed by the Mantel test, which revealed a significant negative relationship between assemblage similarity (Jaccard) and geographical distance (r = −0,204, p < 0,001), although the low value of the correlation coefficient points to the existence of other factors also influencing the relationship. Mantel test carried out after removing the lakes in Sanabria (n = 37) was only marginally significant (r = −0.077, p = 0.047), showing that the spatial pattern observed was almost entirely due to the Sanabria district and not a widespread pattern across the whole study area. This spatial pattern might be a result of either geographical distance (among-lake proximity) or environmental autocorrelation (sites in Sanabria might be different from the rest in environmental features). The PCA graph (Fig. 4) revealed two major environmental gradients across the study area: altitude (axis 1) and a complex factor related to water permanence, vegetation and substrate type (axis 2). These two axes explained 93,3% of the variance. The lakes in Sanabria did not form an environmentally distinct group. They are located within a narrow range of altitude, near the lowest extreme of the gradient in the region (right end of axis 1), but some other lakes were at similar (or even lower) elevations. On the contrary, it included sites with varying characteristics of substrate type, vegetation and water permanence and thus widespread over PCA-axis 2. This spatial autocorrelation might have conditioned the preliminary typological scheme described above. Therefore, we undertook a new cluster analysis after removing the sites in Sanabria to check whether the typological scheme stood, and could thus be considered appropriate for the study area. The resulting dendrogram was quite similar to the one in Fig. 2 and the same typological scheme arose from it. Only the first discriminant function of MDA was significant (p < 0,001). The new three-group scheme was explained by differences (in order according to MDA) in silt, cobbles, helophytes, water persistence, area and hydrophytes (see Fig. 5 for a description of clusters and ANOVA results). Among-cluster differences in altitude were also significant (ANOVA F = 6.25, p = 0.006), but only between groups 1 and 3 (Scheffé test, p = 0.007). The corresponding types could be described as: 1) permanent, relatively deep lakes with coarse-substrate in the littoral zone (no silt) and no or very little vegetation (cluster 1); 2) permanent lakes with some silt in the littoral zone and intermediate depth and vegetation cover (cluster 2); 3) temporary, shallow ponds with silty substrates and high veg-

F. García-Criado et al. / Limnologica 62 (2017) 57–67

61

Fig. 2. Dendrogram based on Jaccard dissimilarity. The two cut levels considered in the study are shown (level one, groups 1 to 3; level 2, groups 1.1 to 3.2). Additional information on environmental pressures (E.P.) is provided for each lake. Lakes without evident environmental pressures are left blank in the corresponding column. D = dammed lakes; F = fish presence; L = livestock in the lake catchment.

Table 3 Taxa most closely related to each lake type, as shown by IndVal (in parenthesis, percentage values). Only taxa with IndVal > 30% for any single type are shown. type 1 (permanent, no silt)

type 2 (permanent, silt)

type 3 (temporary)

Stenophylacini (86.3) Oulimnius (55.2) Pisidium (49.9) Chironominae (37.1) Plectrocnemia (36.5) Hexatomini (36.4%)

Nebrioporus (72.4) Sigara (58.0) Sialis (50.8) Hydroporus (48.6) Athripsodes (47.7) Orthocladiinae (43.8) Aeshna/Anax (43.2) Cloeon (38.4) Erpobdella (37.5) Habrophlebia (37.5) Thraulus (37.5) Arctocorisa (37.5) Gyrinus (35.2) Notonecta (31.2)

Helochares (87.7) Lestes (86.2) Hesperocorixa (77.8) Tabanidae (65.8) Ceratopogoninae (60.5) Tanypodinae (58.0) Sympetrum (52.1) Dytiscus (51.5) Hygrotus (51.1) Agabus (47.0) Gerris (45.6)

etation cover (cluster 3). Taxa with highest IndVal for each cluster (or lake type) are shown in Table 3. 4. Discussion The data set included around half the mountain lakes or ponds catalogued by the regional administration and should give useful

information for developing and implementing a typology for mountain lakes in the region. Furthermore, the relatively large spatial scale of the study was appropriate to provide some clues on issues of general interest, such as which factors should be taken into account to develop typologies meaningful for littoral macroinvertebrates or the relative role of environmental factors and geographical distance in shaping these assemblages in mountain lakes. 4.1. Environmental factors affecting macroinvertebrate assemblages The selection of variables for the study was based on management criteria, but most of them had previously been found to be important for littoral macroinvertebrates. However, generalizations about their individual role are difficult to make because the results are not consistent across studies. Differences in the spatial scale involved, the amplitude of the environmental gradients or the type of systems studied result in different outcomes. For instance, elevation has often been considered a major factor (Füreder et al., 2006; Hinden et al., 2005; Wissinger et al., 2016), but not always (Oertli et al., 2008; Martínez-Sanz et al., 2012a). Similarly, at least macroinvertebrate richness has been found to be related to lake area (Biggs et al., 2005; Céréghino et al., 2008; Hamerlík et al., 2014; Heino, 2000), although there are also exceptions (Hinden et al., 2005; Oertli et al., 2002). Also alkalinity has been occasion-

62

F. García-Criado et al. / Limnologica 62 (2017) 57–67

Fig. 3. Box plots showing the values of the variables for the six clusters. Only variables significantly differing among clusters 1 to 3 are represented. The box on the right-top of each graph gives the ANOVA results and the differences between pairs of groups when they were significant (Scheffé test). Groups 4 to 6 are only represented for informative purposes, but they were excluded from the ANOVA. The plots represent the median (central line), 25 and 75 percentiles (box), and maximum and minimum excepting outliers and extreme values (whiskers). Outlier and extreme values are shown as points and asterisks, respectively.

ally considered a relevant factor for macroinvertebrate in mountain ponds (Boggero and Lencioni, 2006; Little et al., 2006). None of these factors seemed to be a strong driver of taxonomical composition in the study area, although the results might differ provided that wider environmental gradients were available in the region,

as shown for macroinvertebrate richness when very large lakes are included (Martínez-Sanz et al., 2012b). Instead, we found three factors to be the main responsible for differences in the assemblages: water permanence (temporary vs permanent waterbodies), substrate type (silt vs coarser substrates), and vegetation (helophyte

F. García-Criado et al. / Limnologica 62 (2017) 57–67

63

Fig. 4. PCA graph (axes I and II) showing the main environmental patterns in the study area. The same variables as in MDA were used in this analysis. The clusters in Fig. 2 are represented by symbols as follows: cluster 1.1 (solid squares), 1.2 (empty squares), 2.1 (empty circles), 2.2 (solid circles), 3 (empty triangles) and 4 to 6 (X-marks).

and hydrophyte) cover. Water permanence is a critical variable conditioning lake functioning and its relevance for macroinvertebrate assemblages is beyond discussion (Collinson et al., 1995; Della Bella et al., 2005; Hassall et al., 2011; Urban, 2004), as it is for other biological groups. Substrate granulometry or vegetation cover or type, in contrast, have only occasionally been mentioned as major drivers of macroinvertebrate assemblages in mountain lakes in spite of their potential influence on littoral macroinvertebrates (Free et al., 2009). The role of habitat type (especially substrate) deserves particular focus. Substrate could be expected to be determinant in studies at low spatial scale but not so in regional studies. A good example is the work by White and Irvine (2003) on lakes across Ireland. They found differences in the macroinvertabrate assemblages among mesohabitats of the same lake, but such variation was nested in the among-lake variability. At large spatial scales, differences in macroinvertebrate assemblages can be attributed to a great extent to regional variables, with habitat becoming a secondary or complementary factor, as found by Johnson and Goedkoop (2002) in Sweden. Even at relatively small spatial scales geographical patterns can sometimes override the influence of habitat-type variables (Stoffels et al., 2005). Given the relatively large spatial scale of our study area, geographical-driven differences in the assemblage (among districts, for instance) a similar result might have been expected. However, habitat variables were stronger drivers of assemblage structure than variables operating at regional scale (summarized here by longitude and latitude). Our data set and the statistical approach did not make it possible to disentangle the individual contribution of the variables mentioned above. Neither did we aim to analyze the response of the assemblage to individual environmental factors. Instead, we identified taxa associated to each of the three types of lakes defined: temporary systems with silty substrates (they are shallow, small, and with high plant cover), permanent lakes with only coarse substrate (and, at the same time, deeper and with no or very little vegetation) and permanent lakes with some silt in the littoral zone (and with intermediate depth and plant cover). Temporary (silty,

vegetated) systems, for instance, were mainly characterized by the presence of Helochares (almost all of them H. punctatus), Lestes and Hesperocorixa (H. castanea, H. linnaei and H. sahlberghi). All these taxa were predominantly, but not exclusively, found in temporary systems. This is consistent with a general pattern observed in alpine ponds, in which species of temporary ponds tend to be a subset of the surrounding permanent systems with particular biological traits to resist drying (Bazzanti et al., 2000; Wissinger et al., 2016). Lestes, for example, is particularly adapted to temporary waters through their capability to overwinter as eggs, which are resistant to dessication (Corbet and Brooks, 2008). Corixids (Hesperocorixa among them) are often reported from shallow systems, whether temporary or permanent but with special preference for fishless ponds (Jansson, 1986). The strong association between Helochares and temporary ponds, in contrast, finds no support in literature (Hansen, 1982; Valladares et al., 2002). Genera associated to permanent lakes seemed to be less type-specific (lower IndVal). Many of them are not even restricted to standing waters. Oulimnius (mostly O. perezi, but also O. rivularis and O. troglodytes) and Pisidium, which were found to be closely related to permanent lakes with coarse substrate and little or no vegetation, are taxa tolerating a wide range of habitat types but often found in rivers on a variety of substrate types (Tachet et al., 2002; Millán et al., 2014). In fact, none of them was exclusive of this lake type. The dytiscid Nebrioporus (mostly N. fabressei or N. carinatus), the genus most closely associated to permanent systems with silty substrate, is also a rheophilic taxon often found in mountain river pools (Millán et al., 2014).

4.2. Contribution to a lake typology Most of the proposals to classify lakes into types have appeared after the approval of the Water Framework Directive. This has conditioned the criteria for defining lake types, which must follow either “System A” or the less rigid “System B” (Annex II), and has resulted in similar schemes for different regions. Current lake typologies usually generate types on the basis of ecoregion, altitude, area, mixing type (often with depth as a surrogate), conductivity,

64

F. García-Criado et al. / Limnologica 62 (2017) 57–67

Fig. 5. Box plots showing the values of the variables for the three definitive lake types (clusters 1 to 3, after removing sites in Sanabria). Only variables significantly differing among clusters are represented. The box on the right-top of each graph gives the ANOVA results and the differences between pairs of groups when they were significant (Scheffé test). Values for clusters 4 to 6 are not represented because they are the same as in Fig. 3. The plots represent the median (central line), 25 and 75 percentiles (box), and maximum and minimum excepting outliers and extreme values (whiskers). Outlier and extreme values are shown as points and asterisks, respectively.

and alkalinity or catchment geology (Kagalou and Leonardos, 2009; Kolada et al., 2005; Little et al., 2006; Moss et al., 2003; Søndergaard et al., 2005). However, whether they are biologically meaningful has seldom been tested. The same stands for the official typology for implementing the WFD in Spain (Orden ARM/2656/2008),

specifically created for a defined set of lakes (those with area above 8 ha). One conclusion of our results is that some relevant variables are overlooked by usual typologies, in particular water permanence and substrate type. Water permanence is often missing in WFD-

F. García-Criado et al. / Limnologica 62 (2017) 57–67

based typologies because they tend to neglect small waterbodies. It is just a matter of scope whether water permanence should be considered or not. Substrate type is also consistently missing in typological classifications (but it is a possibility considered by Annex II of the WFD). However, it has been recognized as a main driver factor for benthic macroinvertebrates (Free et al., 2009). Here we stress the convenience of taking substrate type into account, as proposed for German lakes by Zenker and Baier (2009). This would make typologies more meaningful for littoral organisms, such as macroinvertebrates or macrophytes. In contrast, it is very likely that some other variables used to construct typologies may be meaningless for littoral macroinvertebrate assemblages. Stratification (or depth, as a surrogate) is a strong determining factor of lake functioning. However, it is probably of little relevance for littoral macroinvertebrates (Hinden et al., 2005; Zenker and Baier, 2009). Including depth in a typology for littoral macroinvertebrates of mountain lakes (with classes above and below 10 m, as proposed by the Spanish scheme, for instance) would create superfluous classes (not corresponding to differences in macroinvertebrate assemblage composition). So could happen with other factors, such as size, although our data set was not good enough to derive conclusions on them, as explained above. The problem for typology development is about the thresholds. Hamerlík et al. (2014) proposed a 2-ha threshold whereas Catalan et al. (2009b) reported differences in taxonomic composition driven by lake area with a threshold at around 3 ha. Both of these values seem too low for Castilla y León mountain lakes where no such threshold is evident (this study; Martínez-Sanz et al., 2012a). Zenker and Baier (2009) found differences in macroinvertebratebased measures between lakes above and below 10 ha. In any case, area is likely to have an influence on whole-lake richness, but might contribute little to configure the composition of littoral assemblages. The evidences of the importance of altitude for lake biological communities are abundant (De Mendoza and Catalan, 2010; Hinden et al., 2005), and its inclusion in lake typologies is mandatory. There is a need to define meaningful thresholds. Our results are insufficient for such a task because, in the available range (14002,200 ma.s.l.), altitude did not seem to affect the macroinvertebrate assemblages. In fact, thresholds proposed by different authors are above this range. De Mendoza and Catalan (2010) reported changes in the littoral macroinvertebrate assemblages in Pyrenean lakes from altitudes around 2,500 ma.s.l., whereas Catalan et al. (2009b) proposed a 190-days ice-cover duration (around 2,300 ma.s.l.) as a significant threshold for several communities. Similarly, decreases in macroinvertebrate abundance, richness or diversity have been reported for lakes in the Alps from 2200-2,600 ma.s.l. (Füreder et al., 2006). All these studies focused on alpine lakes located above the tree line, but mountain ponds can be found over a wider altitude gradient, probably making it necessary to consider an additional threshold at lower altitude to distinguish lakes subjected to freezing from the rest. Spanish typology considers two altitude classes: medium (1000-1,500 ma.s.l.) and high (above 1,500 ma.s.l.) mountain lakes. We found no support for these figures, but we had very few sites below 1500 m to draw conclusions. Unfortunately, there are few studies on mountain lakes at medium or low elevation and there is no information to define such a threshold or even to decide if it is biologically relevant. In summary, several recommendations can be made to advance in the design of a mountain-lake typology meaningful for littoral macroinvertebrates. Substrate type in the littoral zone should be taken into account by incorporating additional classes within permanent lakes (at least two, with and without silty areas). Of course, were monitoring programmes extended to small waterbodies, as desirable, it would be necessary to add classes for temporary waterbodies. On the other hand, there is no need to incorporate

65

mixing type or depth as typological variables since they create spurious types (without correspondence with differences in the macroinvertebrate assemblage composition). Certainly, different biological groups respond to different environmental variables (Allen et al., 1999; Hassall et al., 2011; Hrabik et al., 2005; Kernan et al., 2009). Therefore, a general, rigid typological scheme is not likely to account for all these peculiarities and, consequently, to perform equally in all the cases unless all the potential drivers of assemblage composition (for all the biological groups) are incorporated. This would create an enormous amount of lake types (demanding increased efforts in defining reference conditions and quality criteria) although many of them could be irrelevant for one or several biological groups. Perhaps, typological schemes should be more flexible and allow for different types for each biological group. Neale and Rippey (2008) go beyond and state that largescale typologies based on environmental factors, as promoted by the WFD, are unreliable and recommend using multivariate-based (and biologically specifical) classifications. 4.3. Spatial patterns It is well known that both spatial patterns and environmental conditions may simultaneously have an influence on lake communities (Briers and Biggs, 2005; Cottenie, 2005; Rádková et al., 2014). Our results, however, showed a prevalence of local environmental conditions in shaping macroinvertebrate assemblages. Spatial autocorrelation played a secondary role and was only apparent in one of the lake districts. Whether spatial patterns become apparent or not is partially a matter of scale. In large spatial-scale studies, spatial autocorrelation is more likely to occur. At a panEuropean scale, Kernan et al. (2009) found that location explained more of species variance than the proximal environment, although such tendency was not so evident for littoral macroinvertebrates. They concluded that regional scale approaches would be more robust than pan-European schemes for classification or prediction of biological assemblages. In the other extreme, we might expect dispersal limitation to be less important (and thus spatial autocorrelation less apparent) in studies conducted in small regions, as shown for different groups (Capers et al., 2010; Cottenie et al., 2003; Urban, 2004; Waterkeyn et al., 2008). It is not evident at which spatial scale among-site distance becomes a factor strong enough to hamper the implementation of monitoring programmes. Kernan et al. (2009) suggested dividing Europe into three limno-regions, but spatial autocorrelation independent of environmental conditions has also been reported in studies at smaller scales in ponds (Briers and Biggs, 2005) and rivers (Mykrä et al., 2007). In such large a region as Castilla y León, and given the peripheral and patchy distribution of mountain lakes, a strong spatial autocorrelation could have also been expected, forcing within-district assemblage similarity to be higher than within-type similarity. In general, no such pattern was observed in most of the area at the taxonomical level used in the study (mostly genus): sites from so far apart districts as Gredos, Fuentes Carrionas and Urbión grouped together in the cluster analysis. This supports the tacit assumption, not evident beforehand, that a common typology can be applied to a set of lakes patchily distributed over a large region. Water permanence, in particular, seems to be a factor capable of overriding the homogenizing effect of dispersal (Urban, 2004). In our study, it proved to be strong enough a driver to create distinct lake types (temporary vs permanent) regardless their geographical position, even in those districts with lakes very densely distributed and without significant mountain barriers (Sanabria; Neila). However, not all the factors will perform similarly. Substrate type, for example, was a factor strong enough to differentiate two permanent lake types (with and without silt) in most of the study area, but failed to do it in Sanabria. A probable explanation for this result is

66

F. García-Criado et al. / Limnologica 62 (2017) 57–67

that among-lake connectivity by dispersal was comparatively high. These results suggest a trade-off between geographical proximity and strength of the environmental drivers which, in turn, depends on the biological relevance of the variable and on the amplitude of the environmental gradient in a given area. For management purposes, it would be convenient to quantify both connectivity by dispersal (or geographical proximity) and strength of environmental drivers and establish thresholds helping to know when one or the other will prevail. Acknowledgments This study was supported by the Government of Castilla y León (UMC1/04) and the Spanish Ministry of Education and Science (CGL2006-03927). References Allen, A.P., Whittier, T.R., Larsen, D.P., Kaufmann, P.R., O’Connor, R.J., Hughes, R.M., et al., 1999. Concordance of taxonomic composition patterns across multiple lake assemblages: effects of scale, body size, and land use. Can. J. Fish. Aquat.Sci. 56, 2029–2040. Bazzanti, M., Seminara, M., Baldoni, S., Stella, A., 2000. Macroinvertebrates and environmental factors of some temporary and permanent ponds in Italy. Verhandlungen Internationalen Verein Limnologie 27, 936–941. Beisner, B.E., Peres-Neto, P.R., Lindström, E.S., Barnett, A., Longhi, M.L., 2006. The role of environmental and spatial processes in structuring lake communities from bacteria to fish. Ecology 87, 2985–2991. Biggs, J., Williams, P., Whitfield, M., Nicolet, P., Weatherby, A., 2005. 15 years of pond assessment in Britain: results and lessons learned from the work of Pond Conservation. Aquat. Conserv. Mar. Freshwater Ecosyst. 15, 693–714. Boggero, A., Lencioni, V., 2006. Macroinvertebrate assemblages of high altitude lakes, inlets and outlets in the southern Alps. Archiv für Hydrobiologie 165, 37–61. Borthagaray, A.I., Berazategui, M., Arim, M., 2015. Disentangling the effects of local and regional processes on biodiversity patterns through taxon-contingent metacommunity network analysis. Oikos 124, 1383–1390. Briers, R.A., Biggs, J., 2005. Spatial patterns in pond invertebrate communities: separating environmental and distance effects. Aquat. Conserv. Mar. Freshw. Ecosyst. 15, 549–557. Céréghino, R., Biggs, J., Oertli, B., Declerck, S., 2008. The ecology of European ponds: defining the characteristics of a neglected freshwater habitat. Hydrobiologia 597, 1–6. Capers, R.S., Selsky, R., Bugbee, G.J., 2010. The relative importance of local conditions and regional processes in structuring aquatic plant communities. Freshw. Biol. 55, 952–966. Catalan, J., Curtis, C.J., Kernan, M., 2009a. Remote European mountain lake ecosystems: regionalisation and ecological status. Freshw. Biol. 54, 2419–2432. Catalan, J., Barbieri, M.G., Bartumeus, F., Bituˇsík, P., Botev, I., Brancelj, A., et al., 2009b. Ecological thresholds in European alpine lakes. Freshw. Biol. 54, 2494–2517. Collinson, N.H., Biggs, J., Corfield, A., Hodson, M.T., Walker, M., Whitfield, M., et al., 1995. Temporary and permanent ponds: an assessment of the effects of drying out on the conservation value of aquatic macroinvertebrate communities. Biol. Conserv. 74, 125–133. Corbet, P.S., Brooks, S.J., 2008. Dragonflies. Collins, London. Cottenie, K., Michels, E., Nuytten, N., De Meester, L., 2003. Zooplankton metacommunity structure: regional vs local processes in highly interconnected ponds. Ecology 84, 991–1000. Cottenie, K., 2005. Integrating environmental and spatial processes in ecological community dynamics. Ecol. Lett. 8, 1175–1182. De Mendoza, G., Catalan, J., 2010. Lake macroinvertebrates and the altitudinal environmental gradient in the Pyrenees. Hydrobiologia 648, 51–72. Della Bella, V., Bazzanti, M., Chiarotti, F., 2005. Macroinvertebrate dive,sity and conservation of Mediterranean ponds in Italy: water permanence and mesohabitat influence. Aquat. Conserv. Mar. Freshw. Ecosyst. 15, 583–600. Directive 2000/60/EC, 2000. Water Framework Directive of the European Parliament and of the Council, of 23 October 2000, establishing a framework for Community action in the field of water policy. Off. J. Eur. Commun. L327, 1–72. Dormann, C.F., 2007. Effects of incorporating spatial autocorrelation into the analysis of species distribution data. Global Ecol. Biogeogr. 16, 129–138. Dufrêne, M., Legendre, P., 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecol. Monogr. 67, 345–366. Füreder, L., Ettinger, R., Boggero, A., Thaler, B., Thies, H., 2006. Macroinvertebrate diversity in Alpine lakes: effects of altitude and catchment properties. Hydrobiologia 562, 123–144. Free, G., Solimini, A.G., Rossaro, B., Marziali, L., Giacchini, R., Paracchini, B., et al., 2009. Modelling lake macroinvertebrate species in the shallow sublittoral: relative roles of habitat, lake morphology, aquatic chemistry and sediment composition. Hydrobiologia 633, 123–136.

Hamerlík, L., Svitok, M., Novikmec, M., Oˇcadlík, M., Bituˇsik, P., 2014. Local, among-site, and regional diversity patterns of benthic macroinvertebrates in high altitude waterbodies: do ponds differ from lakes? Hydrobiologia 723, 41–52. Hansen, M., 1982. Revisional notes on some European Helochares Muls. (Coleoptera: Hydrophilidae). Entomologica scandinavica 13, 201–211. Hassall, C., Hollinshead, J., Hull, A., 2011. Environmental correlates of plant and invertebrate species richness in ponds. Biodivers. Conserv. 20, 3189–3222. Heino, J., 2000. Lentic macroinvertebrate assemblage structure along gradients in spatial heterogeneity, habitat size and water chemistry. Hydrobiologia 418, 229–242. Hinden, H., Oertli, B., Menetrey, N., Sager, L., Lachavanne, J.-B., 2005. Alpine pond biodiversity: what are the related environmental variables? Aquat. Conserv. Mar. Freshw. Ecosyst. 15, 613–624. Hrabik, T.R., Greenfield, B.K., Lewis, D.B., Pollard, A.I., Wilson, K.A., Kratz, T.K., 2005. Landscape-scale variation in taxonomic diversity in four groups of aquatic organisms: the influence of physical, chemical, and biological properties. Ecosystems 8, 301–317. Jansson, A., 1986. The Corixidae (Heteroptera) of Europe and some adjacent regions. Acta Entomol. Fennica 47, 1–94. Johnson, R.K., Goedkoop, W., 2002. Littoral macroinvertebrate communities: spatial scale and ecological relationships. Freshw. Biol. 47, 1840–1854. Kagalou, I., Leonardos, I., 2009. Typology, classification and management issues of Greek lakes: implication of the Water Framework Directive (2000/60/CE). Environ. Monit. Assess. 150, 469–484. Kernan, M., Ventura, M., Bituˇsík, P., Brancelj, A., Clarke, G., Velle, G., et al., 2009. Regionalisation of remote European mountain lake ecosystems according to their biota: environmental versus geographical patterns. Freshw. Biol. 54, 2470–2493. Kolada, A., Soszka, H., Cydzik, D., Golub, M., 2005. Abiotic typology of Polish lakes. Limnologica 35, 145–150. Legendre, P., Legendre, L., 1998. Numerical Ecology, second ed. Elsevier Science, B.V Amsterdam. Leibold, M.A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J.M., Hoopes, M.F., et al., 2004. The metacommunity concept: a framework for multi-scale community ecology. Ecol. Lett. 7, 601–613. Little, R., Bowman, J.J., McGarrigle, M.L., Free, G., Donnelly, K., Tierney, D., et al., 2006. Implementation of the Water Framework Directive: investigations for establishing a lake tipology using littoral macroinvertebrates in the Republic of Ireland. Verhandlungen des Internationalen Verein Limnologie 29, 2192–2196. Martínez-Sanz, C., Fernández-Aláez, C., García-Criado, F., 2012a. Richness of littoral macroinvertebrate communities in mountain ponds from NW Spain: what factors does it depend on? J. Limnol. 71, 154–163. Martínez-Sanz, C., Cenzano, C.S.S., Fernández-Aláez, M., García-Criado, F., 2012b. Relative contribution of small mountain ponds to regional richness of littoral macroinvertebrates and the implications for conservation. Aquat. Conserv. 22, 155–164. Millán, A., Sánchez-Fernández, D., Abellán, P., Picazo, F., Carbonell, J.A., Lobo, J.M., ˜ peninsular. Ribera, I., 2014. Atlas de los coleópteros acuáticos de Espana Ministerio de Agricultura, Alimentación y Medio Ambiente, Madrid. Moss, B., Stephen, D., Álvarez, C., Bécares, E., van de Bund, W., et al., 2003. The determination of ecological status in shallow lakes −a tested system (ECOFRAME) for implementation of the European Water Framework Directive. Aquat. Conserv. 13, 507–549. Mykrä, H., Heino, J., Muotka, T., 2007. Scale-related patterns in the spatial and environmental components of stream macroinvertebrate assemblage variation. Global Ecol. Biogeogr. 16, 149–159. Neale, M.W., Rippey, B., 2008. A comparison of environmental and biological site classifications for the prediction of macroinvertebrate communities of lakes in Northern Ireland. Aquat. Conserv. 18, 729–741. Oertli, B., Joye, D.A., Castella, E., Juge, R., Cambin, D., Lachavanne, J.-B., 2002. Does size matter? The relationship between pond area and biodiversity. Biol. Conserv. 104, 59–70. Oertli, B., Indermuehle, N., Angélibert, S., Hinden, H., Stoll, A., 2008. Macroinvertebrate assemblages in 25 high alpine ponds of the Swiss National Park (Cirque of Macun) and relation to environmental variables. Hydrobiologia 597, 29–41. Orden ARM 2656/2008, de 10 de septiembre, por la que se aprueba la instrucción de planificación hidrológica. B.O.E. 229. Rádková, V., Bojková, J., Kˇroupalová, V., Schenková, J., Syrovátka, V., Horsák, M., 2014. The role of dispersal mode and habitat specialisation in metacommunity structuring of aquatic macroinvertebrats in isolated spring fens. Freshw. Biol. 59, 2256–2267. ˜ Razeng, E., Morán-Ordónez, A., Box, J.B., Thompson, R., Davis, J., Sunnucks, P., 2016. A potential role for overland dispersal in shaping aquatic invertebrate communities in arid regions. Freshw. Biol. 61, 745–757. Søndergaard, M., Jeppesen, E., Jensen, J.P., Amsinck, S.L., 2005. Water framework directive: ecological classification of danish lakes. J. Appl. Ecol. 42, 616–629. Shurin, J.B., Cottenie, K., Hillebrand, H., 2009. Spatial autocorrelation and dispersal limitation in freshwater organisms. Oecologia 159, 151–159. Stoffels, R.J., Clarke, K.R., Closs, G.P., 2005. Spatial scale and benthic community organisation in the littoral zones of large oligotrophic lakes: potential for cross-scale interactions. Freshw. Biol. 50, 1131–1145. Tachet, H., Richoux, P., Bournaud, M., Usseglio-Polatera, P., 2002. Invertébrés d’eau douce. Systematique, biologie, écologie. CNRS editions, Paris.

F. García-Criado et al. / Limnologica 62 (2017) 57–67 Toro, M., Granados, I., Robles, S., Montes, C., 2006. High mountain lakes of the central range (Iberian peninsula): regional limnology & environmental changes. Limnetica 25, 217–251. Urban, M.C., 2004. Disturbance heterogeneity determines freshwater metacommunity structure. Ecology 85, 2971–2978. Valladares, L.F., Garrido, J., García-Criado, F., 2002. The assemblages of aquatic Coleoptera from shallow lakes in the northern Iberian Meseta: influence of environmental variables. Eur. J. Entomol. 99, 289–298. Waterkeyn, A., Grillas, P., Vanschoenwinkel, B., Brendonck, L., 2008. Invertebrate community patterns in Mediterranean temporary wetlands along hydroperiod and salinity gradients. Freshw. Biol. 53, 1808–1822.

67

White, J., Irvine, K., 2003. The use of littoral mesohabitats and their macroinvertebrate assemblages in the ecological assessment of lakes. Aquat. Conserv. 13, 331–351. Wissinger, S.A., Oertli, B., Rosset, V., 2016. Invertebrate communities of alpine ponds. In: Batzer, D., Boix, D. (Eds.), Invertebrates in Freshwater Wetlands. An International Perspective on Their Ecology. Springer, pp. 55–103. Zenker, A., Baier, B., 2009. Relevance of abiotic criteria used in German lake typology for macroinvertebrate fauna. Hydrobiologia 636, 379–392.