Environmental variables drive differences in the beta diversity of dragonfly assemblages among urban stormwater ponds

(1)

Contents lists available atScienceDirect

Ecological Indicators

journal homepage:www.elsevier.com/locate/ecolind

Original Articles

Environmental variables drive di ﬀ erences in the beta diversity of dragon ﬂ y assemblages among urban stormwater ponds

F. Johansson

^a,⁎

, L.M. Bini

^b

, P. Coi ﬀ ard

^a

, R. Svanbäck

^a

, J. Wester

^a

, J. Heino

^c

aDepartment of Ecology and Genetics, Animal Ecology, Uppsala University, Norbyvägen 18D, 752 36 Uppsala, Sweden

bDepartamento de Ecologia, Universidade Federal de Goiás, Goiânia 74001-970, GO, Brazil

cFinnish Environment Institute, Freshwater Centre, Paavo Havaksen Tie 3, FI-90570 Oulu, Finland

A R T I C L E I N F O

Keywords:

Beta diversity

Compositional dissimilarity Environmental gradients Generalised dissimilarity modelling Geographic distance

Odonata Urban ecology

A B S T R A C T

Stormwater ponds are beneficial to urban landscapes because these man-made systems can reduce the negative effects offlooding in urban areas and restrain the distribution of pollutants. In addition, these systems are especially important to maintain the biodiversity of urban landscapes. Here, we sampled a set of 18 stormwater ponds in the city of Uppsala in Sweden to test the relationship between beta diversity of adult dragonflies and environmental factors (local and land use variables). We analysed the total beta diversity and its two components: replacement and richness difference. We recorded 31 species of Odonata, comprising 61% of the Odonata species in the province of Uppland in Sweden. By itself, this result indicates the importance of stormwater ponds in contributing to biodiversity in urban areas. The richness difference component of beta diversity was higher than the replacement component. Results from generalized dissimilarity models indicated that the richness difference component was mainly related with pond area and total vegetation cover (aquatic vegetation plus vegetation surrounding ponds). Focusing on different vegetation variables separately, models indicated that the beta diversity components were significantly correlated with percentage cover offloating algae scums, emergent aquatic macrophytes and tall shore vegetation. These results are consistent with what is known about the ecology of dragonflies, including the importance of aerial plant structures for perching, shelter from terrestrial and aquatic predators, and for providing oviposition sites. We also found that the stormwater ponds harboured a large part of the regional species pool. These systems are therefore important havens of biodiversity in urban landscapes. Our results also indicate that the management of different types of vegetation is key to maximize the potential of these systems in maintaining regional biodiversity.

1. Introduction

In order to understand the processes underlying variation in biological diversity, it is important to have a measure that portrays changes in species composition between local assemblages (Anderson et al., 2011). One such measure is beta diversity, which, if measured appro- priately, can be used to explain how changes in species composition across sites translate into gamma diversity (i.e. species diversity in a large region consisting of many sites). One of the earliest formations of this idea wasWhittaker’s (1960)proposition of alpha, beta and gamma diversity. Beta diversity can thus be understood as the element of regional diversity that accumulates from compositional diﬀerences between local species assemblages (Anderson et al., 2011). Many measures of beta diversity have been suggested to capture changes in diversity between sites (Legendre and Legendre, 2012). Among the

most informative decompositions of total beta diversity are compositional dissimilarity due to richness diﬀerences (Brich) and dissimilarity due to species replacement (Brepl). The former refers to loss or gain of species between sites, while the latter refers to replacement of species between sites (Podani and Schmera, 2011; Carvalho et al., 2012;

Legendre, 2014).

Knowledge of correlates of beta diversity components such as Brich and Brepl provides valuable information about biodiversity variation for ecologists, conservation biologists and environmental managers (Angeler, 2013; Socolar et al., 2016; Bush et al., 2016; Rocha et al., 2018). For example, if we are creating or preserving a set of habitats in an urban landscape, knowledge on beta diversity might allow us to understand how species loss and gains or species replacement between sites are aﬀected by distances in the environmental variables and in the geographical spaces. The focus on beta diversity in basic and applied

https://doi.org/10.1016/j.ecolind.2019.105529

Received 6 February 2019; Received in revised form 31 May 2019; Accepted 26 June 2019

⁎Corresponding author.

E-mail address:[email protected](F. Johansson).

Available online 04 July 2019

T

(2)

contexts is also highly beneﬁcial because variation in alpha diversity may fall short in providing adequate information about the factors structuring ecological communities (Socolar et al., 2016), and because beta diversity provides a bridge between alpha and gamma diversity (Anderson et al., 2011). In addition, the Brich and Brepl components may respond diﬀerently to ecological gradients, thereby proving important complementary information about the factors shaping biodiversity variation and its conservation implications (Heino et al., 2019.

Urban habitats suffer from a severe loss of biodiversity (McDonald et al., 2013; Oertli, 2018). Nevertheless, considerable efforts have been done to mitigate such losses, for example, by creating blue-green areas in cities (Lepczyk et al., 2017). One type of such blue-green areas are man-made ponds. Such ponds might harbour a high level of aquatic biodiversity which is affected by various environmental variables, such as built-up area, pond depth and size, as well as pond aquatic and terrestrial vegetation (Goertzen and Suhling, 2013; Briers, 2014;

Céréghino et al., 2014; Hassall, 2014; Heino et al., 2017; Hill et al., 2017; Holtmann et al., 2018; Hill et al., 2018). Stormwater ponds are a special category of man-made ponds. These ponds are used to reduce the negative effects of strong waterflow during heavy rains in urban areas. As such, they not only reduce the run-offfrom impervious surfaces, but also retain the polluted runoff, reducing the impact of pollutants on other urban areas (Heal et al., 2006; Villareal et al., 2004;

Gallagher, 2011; Blicharska and Johansson, 2016). Since urban areas have large amounts of impervious surfaces, these ponds are highly valuable for negative impact on biodiversity in urban environments, such as medium to large-sized cities. Though there is an increasing number of studies on how biodiversity in stormwater ponds is aﬀected by environmental variables (Bishop et al., 2000; Scher and Thiery, 2005; Greenway et al., 2007; Le Viol et al., 2009; Briers, 2014; Vincent and Kirkwood, 2014; Hassall and Anderson, 2015), few of these studies have focused on beta diversity in urban areas. For example, the stormwater ponds studies byScher and Thiery (2005) and Le Viol et al.

(2009)focused on motorway stormwater ponds and did not estimate beta diversity components such as Brich and Brepl. If the goal for conservation biologists is to create and protect biodiversity of aquatic environments in the urban landscape, knowledge on components of beta diversity in stormwater ponds is important, because it provides on information on how environmental factors aﬀects diﬀerent components of biodiversity (Gál et al., 2019).

The purpose of this study was to explore how environmental factors in an urban area correlate with total beta diversity and its richness difference (Brich) and replacement (Brepl) components. We did this by sampling adult damselflies and dragonflies (hereafter collectively re- ferred to as dragonflies (Odonata)) in stormwater ponds in the city of Uppsala in Central Sweden. These systems are thought to be highly variable in terms of several environmental factors and biotic pressures, e.g., area, depth, abundance of aquatic plants, predators and surrounding landscape features (Scher and Thiery, 2005, Le Viol et al., 2009, Hassall and Anderson, 2015; Holtmann et al., 2018). We focused on dragonflies, as they are known to be good ecological indicators of general biodiversity (Oertli et al., 2002; Samways, 2008). Therefore, we hypothesized that both beta diversity components of dragonfly assemblages would exhibit strong responses to environmental differentiation between sites.

2. Methods 2.1. Study system

The study area was located in the city of Uppsala, Central Sweden.

The city has 150 000 inhabitants and covers an area of 26 km². Within the city borders, there are 18 stormwater ponds that areﬁlled with water all year round. All these ponds were surveyed in this study (Fig. 1). The age of these man-made ponds ranged between 4 and 13 years and their size varied from 431 to 11092 m² (Table 1). All

ponds had some percentage of shore vegetation and all ponds except one had some aquatic vegetation. The majority of the ponds are situated at the border of the city, but we chose to call them urban ponds since the majority are situated in areas with developed urban land (Fig. 1).

2.2. Species occurrence

In each pond, dragonfly (Odonata) abundance was recorded ap- proximately every second week over a 10-week period from 29 May until 5 August 2018, corresponding tofive visits per each pond. The abundance counts were distributed over the season because the emergence of individuals differ between species (Dijkstra and Lewington, 2006). The 10 weeks cover the emergence period of all species found at this latitude (Sahlén, 1996). Abundance estimates of dragonflies at each pond were made by two people walking slowly around the ponds and counting the number of individuals of each species present. Most species were identified visually from a distance using binoculars (when necessary). However, for some species, identifications were made after capture using a dragonfly net with a diameter of 50 cm. These individuals were released immediately after identification. During counts of adults, three estimates were done for each species: total number of adults, number of mating pairs and number of egg laying individuals.

All these three estimates were signiﬁcantly correlated for the most abundant species (unpublished results), therefore, we used the total number of individuals of each species as our estimate of abundance.

Because the activities of dragonfly and damselfly species are dependent on weather conditions, no counts were done during cloudy, windy or rainy days. For the analyses, the week with the highest number of individuals observed was used for each species. Hence, for early emerging species this was one of thefirst weeks, while for late emerging species this was one of the last weeks. Species richness and composition were based on the total number of species and individuals observed at a pond.

2.3. Predictor variables

We used both local and land use variables as predictors. Eleven local pond variables were measured (Table 1). Maximum pond depth was estimated in thefield with a meter stick; pond area was estimated from Google maps using the software Qgis 2.18.18 (Qgis, 2018); age of the ponds was obtained from the municipality of Uppsala and the Swedish Transport Administration; and conductivity was measured with a por- table conductivity meter (WTW, Cond. 3210). In addition, three shoreline vegetation and three aquatic vegetation variables were estimated. The percentage of shoreline vegetation was quantified visually by estimating the percentage cover of three classes of plant height (0–10 cm, 11–100 cm, > 100 cm) in an area of covering 2 m perpendi- cular to the shoreline. The percentage was estimated in 10% classes, ranging from 0 to 100% coverage. Vegetation cover offloating algae scums,floating plants and emergent plants in the ponds was visually estimated in 10% classes, ranging from 0 to 100% coverage. Note that these can together exceed 100% because of the overlap in plant distribution. In addition, we constructed an index summarising the values of all vegetation variables into one plant abundance variable. In practice, we summarised the percentage values of the classes from different vegetation variables to obtain an index ranging from 0 (theoretical minimum) to 600 (theoretical maximum). The following land use variables were estimated from Google satellite and OpenStreetMap data using the software Qgis 2.18.18: developed land (impervious surfaces), forest and open land around ponds. These estimates were based on the proportion of area occupied by each land-use category within a radius of 250 m from the centre of the pond. We did not include presence and absence offish in our study since the statistical approach we used (see below) cannot consider binary variables. However, at-test on dragonfly richness estimated as alpha diversity, showed no difference between fish and nofish ponds (t = 0.52, P = 0.30; F. Johansson unpublished).

(3)

3. Statistical methods

We calculated dragonfly beta diversity components using the Sørensen dissimilarity coefficient on presence-absence data. All species from rare to very abundant ones were included in our analyses. To do so, we followed the approach independently devised by Podani and Schmera (2011) and Carvalho et al. (2012). In this approach, total beta diversity is decomposed into replacement and richness difference components: Btotal = Brepl + Brich. Btotal reflects both species replacement and loss-gain; Brepl shows replacement of species composi- tions alone, and Brich shows species loss-gain or richness differences alone. We produced symmetric dissimilarity matrices (18 × 18) based on total beta diversity and each of the two components using the‘beta’ function in the R package BAT (Cardoso et al., 2015).

Using each of these three matrices as responses, we modelled variation in biological dissimilarities using Generalized Dissimilarity Modelling (GDM; Ferrier et al., 2017). GDM is a matrix regression- based method for modelling variation in pairwise dissimilarities, ac- commodating the typical nonlinearities in ecological datasets.

According toFerrier et al. (2017), these nonlinearities occur for two reasons. First, compositional dissimilarity may increase non-mono- tonically with increasing environmental distance between sites. Second, compositional turnover rates may vary across environmental gradients.

Similar to other generalized linear models, the GDM model is specified based on two functions: (i) a link function defining the relationship between the response (i.e. compositional dissimilarity between sites) and the linear predictor (i.e. between-sites distances based on any predictor variable, including geographic distance between sites), and (ii) a variance function defining how the variance of the response de- pends on the predicted mean (Ferrier et al., 2017). We ran the GDM models, plotted the I-splines (that are monotone cubic spline functions) for geographic distance and each predictor variable and assessed the importance of the predictor variables on the response dissimilarities using the functions‘gdm’and‘gdm.varImp’available in the R package gdm (Manion et al., 2017). As Ferrier et al. (2017) mentioned, the nonlinear monotonic functionsfitted for environmental variables and geographic distance are derived from I‐spline basis functions. They show (1) the maximum height reached by each function which Fig. 1.Map of Uppsala city showing the location of the 18 stormwater ponds (© Lantmäteriet).

(4)

indicates the total amount of compositional turnover associated with the given environmental gradient (or geographic distance), while holding all other variables constant; and (2) the slope of each function which indicates the rate of compositional turnover, and how this rate varies along the given environmental gradient (or geographic distance).

We did not standardize the predictor variables in our focal analyses, as different authors have followed a similar approach (e.g., Fitzpatrick et al., 2013), and because this facilitates understanding variation in beta diversity along actual ecological gradients (e.g., Heino et al., 2019). Finally, we used 100 permutations to assess the significance of the full models and predictor variables. We showed the model with all predictors and three alternative models with a reduced set of variables, omitting the variable with the lowest importance at each step. We ran separate models for all predictor variables and geographic distance, as well as for separate vegetation variables and geographic distance to show the subtle effects of vegetation composition on dragonfly assem- blage composition.

4. Results

We found a total of 31 taxa of Odonata in the 18 ponds studied (Table 2). This number represents 61% of the Odonata species in the province of Uppland in Sweden. If we exclude strictly lotic species which probably do not breed in the ponds, the number of species found represents 55% of the fauna in the province. The species richness per pond ranged from 3 to 20, with an average of 10 species per pond. The most common species wereLibellula quadrimaculata(17 ponds),Sym- petrum vulgatum(17 ponds) Lestes sponsa (15 ponds) andCoenagrion puella(14 ponds). Six species were only recorded in one pond (Lestes virens,Platycnemis pennipes,Aeshna mixta,Orthetrum coerulescens,Leu- corrhinia pectoralisandLeucorrhinia rubicunda;Table 2).

There was considerable variation in species composition in pairwise Table1 Localpondandlandusevariableestimatesforeachofthe18stormwaterponds(seeFig.1forpondlocations). PondDepthAreaAgeConductivityFishShoreveg.Shoreveg.Shoreveg.AlgaeFloatingveg.Emergentveg.DevelopedlandForestOpenland ID(cm)(m2)(years)(µS/cm)(+/-)(1–10cm)(%)(11–100cm)(%)(>100cm)(%)(%)(%)(%)(%)(%)(%) 4150153913873090906002060191467 512576614659120405010502043561 11110116411953010402000037063 12605911277408080900108041158 13452146610020809010608023077 14120109645381201003030204096 151507799408090906001020221167 161502631944619070200202030169 179011,0938253711001002010205014086 181157597569180701010302016480 191252166948411010010101029071 2712518911931170603010704031186 28100975111038090907030703030070 3012010991123571102030402060181864 3112511421190213040501003046054 3215036411118314060306002042058 33120215111267010302060002098 3412553344311010010002989

Table 2

List of dragonﬂy species recorded in 18 stormwater ponds in the city of Uppsala, Central Sweden. Also shown are the number of ponds where a species was observed and the maximum abundance of each species per visit.

Taxa Number of ponds Maximum abundance

Aeshna cyanea 2 1

Aeshnasp. 2 2

Aeshna grandis 11 6

Aeshna juncea 2 3

Aeshna mixta 1 1

Aeshna serrata 4 6

Brachytron pratense 4 10

Cordulia aenea 3 3

Coenagrion hastulatum 11 18

Coenagrion lunulatum 3 7

Coenagrion puella 14 50

Coenagrion pulchellum 11 69

Calopteryx splendens 2 1

Calopteryx virgo 5 1

Enallagma cyathigerum 12 100

Erythromma najas 5 30

Ischnura elegans 9 77

Libellula depressa 6 2

Leucorrhinia pectoralis 1 4

Libellula quadrimaculata 17 10

Leucorrhinia rubicunda 1 1

Lestes sponsa 15 119

Lestes virens 1 1

Orthetrum coerulescens 1 1

Platycnemis pennipes 1 1

Sympetrum danae 3 7

Sympetrumﬂaveolum 7 18

Sympecma fusca 3 2

Somatochlora metallica 4 2

Sympetrum sanguineum 8 12

Sympetrum vulgatum 17 51

(5)

comparisons of dragonfly assemblages (Fig. 2). The average total beta diversity was 0.483. This average was mainly due to the richness differences component (0.312), rather than the replacement component (0.171), indicating that much of the variation in beta diversity is related to richness differences between the ponds. It has to be emphasised that the richness difference component is not strictly equivalent to alpha diversity, but to the part of compositional dissimilarity not related to the replacement component.

For total beta diversity, the full GDM model showed that plant index and pond area were the most inﬂuential and signiﬁcant variables (Table 3,Fig. 3). This model explained 40.32% of the deviance in total beta diversity. In addition, two alternative models excluding pond age (Model-1) or pond age and forest area (Model-2) had the same explained deviance. Open land was also excluded from a third model (Model-3). The replacement component was best explained by the plant index, but the full model explained only a minor part of deviance (10%:

Fig. 2.Heatmaps showing pairwise dissimilarities for total (Btotal), replacement (Brepl) and richness diﬀerence (Brich) components of stormwater pond dragonﬂies.

Table 3

Results of the GDM analysis on the relationship between biotic dissimilarities (beta total, replacement and richness diﬀerence) and environmental variables while controlling for geographical distance. Note that‘plants’is an index of the abundance of all plant cover classes. Shown are four models, the one with all predictors included and three additional ones with the weakest variable removed at each step. Model p-values are based on 100 permutations. The sig- niﬁcant predictors (p < 0.05) are shown by bolded font. NA = variable not assessed in the model.

Btotal Full Model Model-1 Model-2 Model-3

Model deviance 8.573 8.573 8.573 8.750

Percent explained 40.320 40.320 40.320 39.086

Model p-value 0.070 0.010 0.020 0.020

Impacts

Geographic distance 8.991 9.290 9.290 9.015

Depth 1.291 1.291 1.291 0.968

Area 24.897 24.897 24.897 30.210

Age 0.000 NA NA NA

Conductivity 0.400 0.400 0.400 0.317

Developed 3.244 3.244 3.244 5.130

Forest 0.000 0.000 NA NA

Open 3.061 3.061 3.061 NA

Plants 48.413 48.413 48.413 53.207

Brepl Full Model Model-1 Model-2 Model-3

Model deviance 23.183 23.183 23.183 23.183

Model p-value 0.130 0.120 0.020 0.040

Impacts

Depth 9.103 9.103 9.216 9.206

Area 0.000 NA NA NA

Age 0.000 0.000 NA NA

Conductivity 10.955 10.955 11.219 11.206

Developed 4.989 4.989 4.989 4.984

Forest 0.016 0.016 0.016 NA

Open 0.519 0.519 0.519 0.526

Plants 34.753 34.753 34.789 35.282

Brich Full Model Model-1 Model-2 Model-3

Model deviance 23.269 23.269 23.280 23.326

Model p-value 0.030 0.060 0.040 0.030

Impacts

Depth 13.920 13.920 13.769 13.771

Area 54.814 54.814 55.764 55.454

Age 1.360 1.508 1.392 1.238

Conductivity 0.175 0.175 NA NA

Developed 0.760 0.760 0.695 NA

Forest 0.000 NA NA NA

Open 2.340 2.340 2.230 2.188

Plants 27.884 27.884 27.758 27.272

(6)

Table 3). The three alternative models excluding the least important variables had almost the same amount of explained deviance. On the other hand, the richness diﬀerence component was most strongly associated with pond area and plant index, which explained 22.09% of the deviance in the full model (Table 3). The three alternative models omitting the variables with the lowest importance were almost as strong as the full model with all predictor variables.

The second set of GDM models including the different vegetation variables and geographic distance showed that, for total beta diversity, algae, emergent plants and geographic distance were significant predictors of beta diversity (Table 4,Fig. 4). This model explained more than half of the deviance (54.57%). The three alternative models with the reduced sets of predictor variables also explained a large amount of deviance. The replacement component was poorly explained by the predictor variables, and none were significant in the full model (Table 4). In the three alternative models, some individual predictor variables were significant, but explained a low percentage of the deviance. The richness difference component was best explained by algae and geographic distance, and the full model explained 29.3% of deviance (Table 4). In general, the alternative models explained similar amounts of deviance, and algae, short shore plants and geographic

distance were the signiﬁcant predictors.

5. Discussion

Our study is one of the few that have explored how environmental factors affect the richness difference (Brich) and replacement (Brepl) components of beta diversity in aquatic ecosystems (Radkova et al., 2014; Specziar et al., 2018). We found that both beta diversity components varied widely, but the richness difference component was much more responsive to ecological and geographical differentiation than the replacement component. Thus, our hypothesis was partially supported by the results. In other words, the environmental variables causing changes between dragonfly assemblages are mainly a result of differences in species richness between ponds rather than replacement of species between ponds. Few studies have compared Brich and Brepl in aquatic systems. In contrast to our study,Radkova et al. (2014) and Specziar et al. (2018)found that Brepl was higher than Brich. However, those two studies compared the beta diversity components across contrasting habitats, such asflowing and standing water or lakes and streams, respectively. In such systems, many species have very different niches, occurring in either lotic or lentic systems. In our study, we Fig. 3.Graphical results for total beta diversity of dragonflies from generalised dissimilarity modelling (GDM). Thefirst two plots show observed compositional dissimilarity vs predicted ecological distance and observed compositional dissimilarities vs predicted compositional dissimilarities based on the GDM model. The remaining seven plots show the responses of compositional dissimilarities to the environmental predictor variables and geographical distance. See text for further information.

(7)

compared the beta diversity components in a set of ponds, which, when compared toﬂowing versus standing waters, are likely more similar in physical and chemical conditions. Hence, we suggest that the higher Brich compared to Brepl in our study is a result of relatively more similar environmental conditions compared to these two previous studies (Radkova et al., 2014; Specziar et al., 2018).

In the full model, distance between ponds and plant index were associated with total beta diversity. An increase in geographic distance between ponds could result in a high compositional dissimilarity because individuals have larger distances to cover or because distantly located ponds diﬀer in environmental conditions (Heino et al., 2017).

Since dragonflies are good dispersers (especially considering the small spatial extent of this study), low dispersal is probably not an important mechanism to explain this result. Therefore, we suggest that it might be due to unmeasured (and spatially autocorrelated) environmental variables that might differ between ponds. The association between total beta diversity and our plant index, which was a composite of many vegetation variables, is supported by previous studies showing that vegetation affects biodiversity of dragonflies as well as other aquatic invertebrates (Dodson et al., 2005; Hassall et al., 2011; Jansen et al.

2018; LeGall et al., 2018).

Interestingly, the Brich component was related to pond area. The species-area relationship is one of the most iconic patterns in ecology (MacArthur and Wilson, 1967). However, area is not as an important predictor of species richness in freshwater as in terrestrial systems, and the relationship is usually lower or even absent for small freshwater systems (Oertli et al., 2002; Hassall et al., 2011; but seeBiggs et al., 2005). In addition, the strength of the relationship might diﬀer between taxonomic groups. For example, in a study on 60 ponds,Oertli et al.

(2002)found that pond size explained 31% of the variability in dragonfly species richness, while there was no relationship between pond size and species richness for amphibians and aquatic beetles. The absence of a strong relationship between species richness and area in freshwater systems has been suggested to be a second order effect, such as interactions among species that might vary between ponds differing in size and connectivity (Scheffer et al., 2006). However, we note that we estimated diversity as beta diversity, whereas most other studies that have tested the relationship between area and diversity have focused directly on alpha diversity (Oertli et al., 2002; Biggs et al., 2005;

Hassall et al., 2011). Nevertheless, the importance of pond area to the Brich component of dragonﬂies is an interesting result, which parallels the one obtained for alpha diversity byOertli et al. (2002)in Switzer- land.

Beta diversity showed a significant relationship with emergent vegetation. Emergent vegetation has multiple positive effects on dragonflies, and species differ in their use of such plants in a number of ways (Corbet, 1980). First, the above-water plant structure provides locations for perching adults, shelter from potential terrestrial predators, and oviposition substrate (Corbet, 1980). Second, the below water surface structure of these plants provides shelter from aquatic predators (Thompson, 1987), as well as structure for egg deposition (Corbet, 1980). We also found that beta diversity was affected by algae cover.

Some of dragonfly species usefloating algae for egg laying (Rowe, 1988; Wolf and Waltz, 1988), while others avoid ponds with a high degree of algae cover (Rowe, 1988). Our results suggest that algae cover may affect species richness since Brich, but not Brepl, was affected by floating algae. Finally, we found that shore vegetation taller than 100 cm affected the Brich component, and we suggest that a high amount of vegetation around a pond provides spots for perching, resting and shelter from predators (Corbet, 1980; Roquette and Thompson, 2007). However, if ponds are too shady as a result of vegetation > 100 cm, the diversity of dragonflies decreases (Remsburg et al., 2008).

Several of our environmental predictor variables were not associated with beta diversity, including pond depth, age, conductivity and land use. Conductivity levels were high in our ponds, but were within the range found in other similar studies (Scher and Thiery, 2005, Le Viol et al., 2009). Conductivity is usually associated with nutrient levels and/or salt concentrations. However, we did not estimate nutrients, such as phosphorous and nitrogen, or salts, such as sulphate and chlorine, and cannot therefore speculate which of these individual variables contributed the most to the high conductivity levels we found in our study ponds. Nevertheless, conductivity did not affect beta diversity of dragonflies significantly in this study. Since our study was performed in an urban landscape, it was surprising that developed land was not associated with beta diversity. For example, many studies in urban landscapes have indicated that the amount of developed land has negative impacts on biodiversity of aquatic insects, including that of dragonflies (Goertzen and Suhling, 2013; Heino et al., 2017; Thornhill et al., 2017; Holtmann et al., 2018). One reason for the absence of such a relationship in our study could be that the majority of our ponds were situated in the outer margin of the city, and hence they had a low percentage of developed land (maximum = 46.5%) around them.

Although we only sampled 18 stormwater ponds, they harboured 61% of the Odonata species found to date in the province of Uppland in Sweden. In addition, the numbers of species found represents about half of the species found in Sweden (48%). Hence, these ponds are Table 4

Results of the GDM analysis on the relationship between biotic dissimilarities and vegetation variables while controlling for geographical distance. Note that there are six diﬀerent vegetation variables. Shown are four models, the one with all predictors included and three additional ones with the weakest variable removed at each step. Model p-values are based on 100 permutations. The signiﬁcant predictors (p < 0.05) are shown by bolded font. NA = variable not assessed in the model.

Btotal Full Model Model-1 Model-2 Model-3

Model 6.525 6.631 6.690 6.960

Percent 54.572 53.840 53.428 51.545

Model p-value < 0.010 < 0.010 < 0.010 < 0.01 Impacts

Shore 1–10 cm 0.028 0.765 NA NA

Shore 10–100 cm 1.342 NA NA NA

Shore > 100 cm 2.879 5.423 5.664 4.631

Algae 17.328 17.106 16.590 24.961

Floating 3.741 3.346 3.524 NA

Emergent 13.572 13.621 22.639 44.332

Brepl Full Model Model-1 Model-2 Model-3

Model deviance 23.839 23.839 23.853 23.864

Model p-value 0.120 0.140 0.030 0.030

Impacts

Shore 1–10 cm 0.000 NA NA NA

Shore 10–100 cm 0.697 0.697 NA NA

Shore > 100 cm 15.366 15.366 37.282 36.970

Algae 0.688 0.688 0.496 NA

Floating 4.826 4.826 4.484 4.200

Emergent 31.457 31.976 32.543 33.156

Brich Full Model Model-1 Model-2 Model-3

Model deviance 21.111 21.218 21.354 21.591

Model p-value 0.010 < 0.010 < 0.010 < 0.010 Impacts

Geographic distance 2.535 2.182 2.779 NA

Shore 1–10 cm 11.263 11.346 11.746 12.359

Shore 10–100 cm 3.499 3.501 3.324 3.123

Shore > 100 cm 1.218 NA NA NA

Algae 27.103 26.222 35.052 34.451

Floating 1.922 1.575 NA NA

Emergent 2.781 2.543 5.149 6.258

(8)

important to maintain biodiversity in urban environments. This high biodiversity corresponds to that found in other studies of stormwater ponds in urban and rural areas (Scher and Thiery, 2005, Le Viol et al.

2009; Hassall and Anderson, 2015; Holtmann et al., 2018). In fact, stormwater ponds can have as high or higher biodiversity as natural ponds and other water bodies (Hassall and Anderson, 2015, Le Viol et al., 2009, Holtmann et al., 2018). Hence, our study, as well as those cited above, supports previous conclusions that man-made ponds such as stormwater ponds may favour and help maintain regional biodiversity. Thus, preserving and creating new ponds is especially important nowadays as there has been a large decrease in wetland habitats during the last decade (Gardner et al., 2015). Wetlands are also among the habitats with a great risk of biodiversity loss and have a high proportion of biodiversity relative to their global abundance (Dudgeon et al., 2007).

We encourage the creation of more stormwater ponds since they provide two important purposes: (1) they offer beneficial ecosystem services such as water retention and purification, and (2) they con- tribute to the maintenance of aquatic biodiversity. However, based on previous studies (Scher and Thiery, 2005) and on our results, it is

important that the stormwater ponds are designed in a biodiversity- friendly way, regarding bottom substratum and pond vegetation. In addition, it is important that toxic contaminants such as heavy metals and pesticide are kept at low levels in stormwater ponds. If not, these ponds might act as ecological traps, i.e. they attract adult dragonflies that choose to oviposit in these habitats, but the offspring suffers from higher mortality compared to when breeding in natural habitats. For example,Sievers et al. (2018)found that tadpoles had lower survival in mesocosms simulating stormwater ponds with higher contaminant levels. Finally, if we want to maintain regional biodiversity, we should try to maximize beta diversity. To do so, ponds with different vegetation types that would differ in species composition (i.e. high beta diversity of dragonflies) are required.

Acknowledgements

Work by LMB has been continuously supported by CNPq grants (304314/2014-5) and the Institute for Science and Technology (INCT) in Ecology, Evolution and Biodiversity Conservation (MCTIC/CNpq proc. 465610/2014-5 and FAPEG). JH acknowledges the support for Fig. 4.Graphical results for total beta diversity of dragonﬂies from generalised dissimilarity modelling (GDM). Theﬁrst two plots show observed compositional dissimilarity vs predicted ecological distance and observed compositional dissimilarities vs predicted compositional dissimilarities based on the GDM model. The remaining seven plots show the responses of compositional dissimilarities to the vegetation predictor variables and geographical distance. See text for further information.

(9)

freshwater biodiversity research from the Finnish Environment Institute. FJ and RS was supported by the Swedish Research Council.

Thanks to Jaelle Brealey for comments on a previous version of this article.

Appendix A. Supplementary data

Supplementary data to this article can be found online athttps://

doi.org/10.1016/j.ecolind.2019.105529.

References

Angeler, D.G., 2013. Revealing a conservation challenge through partitioned long-time beta diversity: increasing turnover and decreasing nestedness of boreal lake metacommunities. Divers. Distrib. 19, 772–781.

Anderson, M.J., Crist, T.O., Chase, J.M., Vellend, M., Inouye, B.D., Freestone, A.L., Sanders, N.J., Cornell, H.V., Comita, L.S., Davies, K.F., Harrison, S.P., Kraft, N.J., Stegen, J.C., Swenson, N.G., 2011. Navigating the multiple meanings ofβdiversity: a roadmap for the practicing ecologist. Ecol. Lett. 14, 19–28.

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. Freshwat. Ecosyst. 15, 693–714.

Bishop, et al., 2000. Contamination and Wildlife communities in stormwater retention ponds in Guelph and the greater Toronto area, Ontario, 1997 and 1998 Part I– Wildlife communities. Water Qual. Res. J. Can. 35, 399–435.

Blicharska, M., Johansson, F., 2016. Urban ponds for people and by people. In: Francis, R.

A., Millington, J.D.A., Chadwick, M.A. (Eds.), Urban Landscape Ecology: Science, Policy and Practice. Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN, 164–180.

Briers, R.A., 2014. Invertebrate communities and environmental conditions in a series of urban drainage ponds in Eastern Scotland: Implications for biodiversity and conservation value of SUDS. Soil Air Water 42, 193–200.

Bush, A. et al. 2016. Current uses of beta-diversity in biodiversity conservation: a response to Socolar et al. TREE. 31, 337–338.

Cardoso, P., Rigal, F., Carvalho, J.C., 2015. BAT–biodiversity assessment tools, an R package for the measurement and estimation of alpha and beta taxon, phylogenetic and functional diversity. Method. Ecol. Evol. 6, 232–236.

Carvalho, J.C., Cardoso, P., Gomes, P., 2012. Determining the relative roles of species replacement and species richness diﬀerences in generating beta-diversity patterns.

Glob. Ecol. Biogeogr. 21, 760–771.

Céréghino, R., Boix, D., Cauchie, H.M., et al., 2014. The ecological role of ponds in a changing world. Hydrobiologia 723, 1–6.

Corbet, P.S., 1980. Biology of odonata. Annu. Rev. Entomol. 25, 189–217.

Dijkstra, K.-D.B., Lewington, R., 2006. Field Guide to the Dragonﬂies of Britain and Europe. British Wildlife Publishing, Dorset, England.

Dodson, S.I., Lillie, R.A., Will-Wolf, S., 2005. Land use, water chemistry, aquatic vegetation, and zooplankton community structure of shallow lakes. Ecol. Appl. 15, 1191–1198.

Dudgeon, D., Arthington, A.H., Gessner, M.O., Kawabata, Z., Knowler, D.J., Lévêque, C., Naiman, R.J., Prieur-Richard, A.H., Soto, D., Stiassny, M.L., Sullivan, C.A., Ferrier, S., Manion, G., Elith, J., Richardson, K., 2007. Using generalized dissimilarity modelling to analyse and predict patterns of beta diversity in regional biodiversity assessment.

Divers. Distrib. 13, 252–264.

Ferrier, S., Manion, G., Elith, J., Richardson, K., 2017. Using generalized dissimilarity modelling to analyse and predict patterns of beta diversity in regional biodiversity assessment. Divers. Distrib. 13, 252–264.

Fitzpatrick, M.C., Sanders, N.J., Normand, S., Svenning, J.C., Ferrier, S., Gove, A.D., Dunn, R.R., 2013. Environmental and historical imprints on beta diversity: insights from variation in rates of species turnover along gradients. Proc. R. Soc. B 280, 20131201.

Greenway, M., Jenkins, D., Polson, C., 2007. Macrophyte zonation in stormwater wetlands: getting it right! A case study from subtropical Australia. Water Sci. Technol.

56, 223–231.

Gál, B., Szivák, I., Heino, J., Schmera, D., 2019. The eﬀect of urbanization on freshwater macroinvertebrates–knowledge gaps and future research directions. Ecol. Ind. 104, 357–364.

Gallagher, M.T., 2011. Watershed-scale analysis of pollutant distributions in stormwater management ponds. Urban Ecosystem 14, 469–484.

Gardner, R.C., Barchiesi, S., Beltrame, C., Finlayson, C.M., Galewski, T., Harrison, I., Paganini, M., Perennou, C., Pritchard, D.E., Rosenqvist, A., Walpole, M. 2015. State of the World’s Wetlands and their Services to People: A compilation of recent analyses.

Ramsar Brieﬁng Note no. 7. Gland, Switzerland: Ramsar Convention Secretariat.

Goertzen, D., Suhling, F., 2013. Promoting dragonﬂy diversity in cities: major determinants and implications for urban pond design. J. Insect Conserv. 17, 399–409.

Hassall, C., 2014. The ecology and biodiversity of urban ponds. WIREs Water 1, 187–206.

Hassall, C., Hollingshead, J., Hull, A., 2011. Environmental correlates of plant and invertebrate species richness in ponds. Biodivers. Conserv. 20, 3189–3222.

Hassall, C., Anderson, S., 2015. Stormwater ponds can contain comparable biodiversity to unmanaged wetlands in urban areas. Hydrobiologia 745, 137–149.

Heal, K.V., Hepburn, D.A., Lunn, R.J., 2006. Sediment management in sustainable urban drainage system ponds. Water Sci Technol. 53, 219–227.

Heino, J., Bini, L.M., Andersson, J., Bergsten, Johannes, Bjelke, U., Johansson, F., 2017a.

Unravelling the correlates of species richness and ecological uniqueness in a meta- community of urban pond insects. Ecol. Indicators 73, 422–431.

Heino, J., Alahuhta, J., Ala-Hulkko, T., Antikainen, H., Bini, L.M., Bonada, N., Datry, T.,

Erős, T., Hjort, J., Kotavaara, O., Melo, A.S., Soininen, J., 2017b. Integrating dispersal proxies in ecological and environmental research in the freshwater realm. Environ.

Rev. 25, 334–349.

Heino, J., Alahuhta, J., Fattorini, S., Schmera, D., 2019. Predicting beta diversity of terrestrial and aquatic beetles using ecogeographical variables: insights from the replacement and richness diﬀerence components. J. Biogeogr. 46, 304–315.

Hill, M.J., Heino, J., Thornhill, I., Ryves, D.B., Wood, P.J., 2017. Eﬀects of dispersal mode on the environmental and spatial correlates of nestedness and species turnover in pond communities. Oikos 126, 1575–1585.

Hill, M.J., et al., 2018. Community heterogeneity of aquatic macroinvertebrates in urban ponds at a multi-city scale. Landscape Ecol. 33, 389–405.

Holtmann, L., et al., 2018. Stormwater ponds promote dragonﬂy (Odonata) species richness and density in urban areas. Ecol. Eng. 118, 1–11.

Jansen, A., et al., 2018. Simple pond restoration measures increase dragonﬂy (Insecta:

Odonata) diversity. Biodivers. Conserv. 27, 2311–2328.

Gall, Le, et al., 2018. Determinants of landscape-scale factors of pond odonate assemblages. Freshw. Biol. 63, 306–317.

Le Viol, I., Mocq, J., Julliard, R., Kerbiriou, C., 2009. The contribution of motorway stormwater retention ponds to the biodiversity of aquatic macroinvertebrates. Biol.

Conserv. 142 (12), 3163–3171.

Legendre, P., 2014. Replacement and richness diﬀerence components. Glob. Ecol.

Biogeogr. 23, 1324–1334.

Legendre, P., Legendre, L., 2012. Numerical Ecology, third ed. Elsevier.

Lepczyk, et al., 2017. Biodiversity in the city: fundamental questions for understanding the ecology of urban green spaces for biodiversity conservation. Bioscience 67, 799–807.

Manion, G., Lisk, M., Ferrier, S., Nieto-Lugilde, D., Mokany, K., Fitzpatrick, M.C., 2017 gdm: Generalized Dissimilarity Modeling. R package version 1.3.4. https://CRAN.R- project.org/package=gdm.

MacArthur, R.H., Wilson, E.O., 1967. The Theory of Island Biogeography. Princeton University Press, Princeton, N.J.

McDonald, R.I., et al., 2013. In Urbanization, Biodiversity and Ecosystem Services:

Challenges and Opportunities. Chapter 3. Urbanization and Global Trends in Biodiversity and Ecosystem Services.

Oertli, B., et al., 2002. Does size matter? The relationship between pond area and biodiversity. Biol. Conserv. 104, 59–70.

Oertli, B., 2018. Freshwater biodiversity conservation: the role of artiﬁcial ponds in the 21st century. Aquat. Conserv. Mar. Freshwater Ecosyst. 28, 264–269.

Podani, J., Schmera, D., 2011. A new conceptual and methodological framework for exploring and explaining pattern in presence–absence data. Oikos 120, 1625–1638.

QGIS Development Team. 2018. QGIS Geographic Information System. Open Source Geospatial Foundation Project. http://qgis.osgeo.org.

Radkova, V., et al., 2014. The importance of species replacement and richness diﬀerences in small-scale diversity patterns of aquatic macroinvertebrates in spring fens.

Limnologica 47, 52–61.

Remsburg, et al., 2008. Shade alone reduces adult dragonﬂy (Odonata: Libellulidae) abundance. J. Insect Behav. 21, 460–468.

Rocha, M.P., Bini, L.M., Domisch, S., Tolonen, K.T., Jyrkänkallio-Mikkola, J., Soininen, J., Hjort, J., Heino, J., 2018. Local environment and space drive multiple facets of stream macroinvertebrate beta diversity. J. Biogeogr. 45, 2744–2754.

Rouquette, J.R., Thompson, D.J., 2007. Roosting site selection in the endangered dam- selﬂy, Coenagrion mercuriale, and implications for habitat design. J. Insect Conserv.

11, 187–193.

Rowe, R.J., 1988. Alternative oviposition behaviours in three New Zealand cordulii dragonﬂies: their adaptive signiﬁcance and implications for male mating tactics.

Zool. J. Linn Soc. 92, 43–66.

Samways, M.J., 2008. Dragonflies as focal organisms in contemporary conservation biology. In: Córdoba-Aguilar, A. (Ed.), Dragonflies and damselflies. Oxford University Press, Oxford, pp. 97–108.

Sahlén, G., 1996. Sveriges trollsländor. Fältbiologernas förlag. Bokskorpionen.

Stockholm.

Scheﬀer, M., et al., 2006. Small habitat size and isolation can promote species richness:

second-order eﬀects on biodiversity in shallow lakes and ponds. Oikos 112, 227–231.

Scher, O., Thiery, A., 2005. Odonata, Amphibia and environmental characteristics in motorway stormwater retention ponds (Southern France). Hydrobiologia 551, 237–251.

Sievers, M., et al., 2018. Stormwater wetlands can function as ecological traps for urban frogs. Ecol. Appl. 28, 1106–1115.

Socolar, J.B., et al., 2016. How should beta-diversity inform biodiversity conservation?

TREE 31, 67–80.

Specziar, A., et al., 2018. Environmental and spatial drivers of beta diversity components of chironomid metacommunities in contrasting freshwater systems. Hydrobiologia 819, 123–143.

Thompson, D.J., 1987. Regulation of damselﬂy populations: the eﬀects of weed density on larval mortality due to predation. Freshw. Biol. 17, 367–371.

Thornhill, et al., 2017. Local and landscape scale determinants of macroinvertebrate assemblages and their conservation value in ponds across an urban land-use gradient.

Biodivers. Conserv. 26, 1065–1086.

Villareal, E.I., et al., 2004. Inner city stormwater control using a combination of best management practices. Ecol. Eng. 22, 279–288.

Vincent, J., Kirkwood, A.E., 2014. Variability of water quality, metals and phytoplankton community structure in urban stormwater ponds along a vegetation gradient. Urban Ecosystems 17, 839–853.

Whittaker, R.H., 1960. Vegetation of the Siskiyou Mountains, Oregon and California.

Ecol. Monogr. 30, 279–338.

Wolf, L.L., Waltz, E.C., 1988. Oviposition site selection and spatial predictability of female white-faced dragonﬂies (Leucorrhinia intacta) (Odonata: Libellulidae). Ethology 78, 306–332.